Merge pull request #1337 from DanielGavin/parser-fix

Add Matrix_Type as literal type on "core:odin"

.github/ISSUE_TEMPLATE/bug_report.md

@@ -11,8 +11,8 @@ assignees: ''
 
 Please provide any relevant information about your setup. This is important in case the issue is not reproducible except for under certain conditions.
 
-* Operating System:
-* Please paste `odin version` output:
+* Operating System & Odin Version:
+* Please paste `odin report` output:
 
 ## Expected Behavior
 

.github/workflows/ci.yml

@@ -6,31 +6,42 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v1
-      - name: Download LLVM
-        run: sudo apt-get install llvm-11 clang-11 llvm
+      - name: Download LLVM, botan
+        run: sudo apt-get install llvm-11 clang-11 llvm libbotan-2-dev botan
       - name: build odin
         run: make release
       - name: Odin version
         run: ./odin version
         timeout-minutes: 1
+      - name: Odin report
+        run: ./odin report
+        timeout-minutes: 1
       - name: Odin check
         run: ./odin check examples/demo/demo.odin -vet
         timeout-minutes: 10
       - name: Odin run
         run: ./odin run examples/demo/demo.odin
         timeout-minutes: 10
+      - name: Odin run -debug
+        run: ./odin run examples/demo/demo.odin -debug
+        timeout-minutes: 10
       - name: Core library tests
         run: |
           cd tests/core
           make
         timeout-minutes: 10
+      - name: Vendor library tests
+        run: |
+          cd tests/vendor
+          make
+        timeout-minutes: 10
   build_macOS:
     runs-on: macos-latest
     steps:
       - uses: actions/checkout@v1
-      - name: Download LLVM and setup PATH
+      - name: Download LLVM, botan and setup PATH
         run: |
-          brew install llvm@11
+          brew install llvm@11 botan
           echo "/usr/local/opt/llvm@11/bin" >> $GITHUB_PATH
           TMP_PATH=$(xcrun --show-sdk-path)/user/include
           echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
@@ -39,12 +50,28 @@ jobs:
       - name: Odin version
         run: ./odin version
         timeout-minutes: 1
+      - name: Odin report
+        run: ./odin report
+        timeout-minutes: 1
       - name: Odin check
         run: ./odin check examples/demo/demo.odin -vet
         timeout-minutes: 10
       - name: Odin run
         run: ./odin run examples/demo/demo.odin
         timeout-minutes: 10
+      - name: Odin run -debug
+        run: ./odin run examples/demo/demo.odin -debug
+        timeout-minutes: 10
+      - name: Core library tests
+        run: |
+          cd tests/core
+          make
+        timeout-minutes: 10
+      - name: Vendor library tests
+        run: |
+          cd tests/vendor
+          make
+        timeout-minutes: 10
   build_windows:
     runs-on: windows-latest
     steps:
@@ -57,6 +84,9 @@ jobs:
       - name: Odin version
         run: ./odin version
         timeout-minutes: 1
+      - name: Odin report
+        run: ./odin report
+        timeout-minutes: 1
       - name: Odin check
         shell: cmd
         run: |
@@ -69,6 +99,12 @@ jobs:
           call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
           odin run examples/demo/demo.odin
         timeout-minutes: 10
+      - name: Odin run -debug
+        shell: cmd
+        run: |
+          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin run examples/demo/demo.odin -debug
+        timeout-minutes: 10
       - name: Core library tests
         shell: cmd
         run: |
@@ -76,6 +112,13 @@ jobs:
           cd tests\core
           call build.bat
         timeout-minutes: 10
+      - name: Vendor library tests
+        shell: cmd
+        run: |
+          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          cd tests\vendor
+          call build.bat
+        timeout-minutes: 10
       - name: core:math/big tests
         shell: cmd
         run: |

.github/workflows/nightly.yml

@@ -93,7 +93,7 @@ jobs:
       - uses: actions/checkout@v1
       - uses: actions/setup-python@v2
         with:
-          python-version: '3.x'
+          python-version: '3.8.x'
 
       - name: Install B2 CLI
         shell: bash
@@ -127,16 +127,23 @@ jobs:
           BUCKET: ${{ secrets.B2_BUCKET }}
           DAYS_TO_KEEP: ${{ secrets.B2_DAYS_TO_KEEP }}
         run: |
+          echo Authorizing B2 account
           b2 authorize-account "$APPID" "$APPKEY"
-
+          
+          echo Uploading artifcates to B2
           chmod +x ./ci/upload_create_nightly.sh
           ./ci/upload_create_nightly.sh "$BUCKET" windows-amd64 windows_artifacts/
           ./ci/upload_create_nightly.sh "$BUCKET" ubuntu-amd64 ubuntu_artifacts/
           ./ci/upload_create_nightly.sh "$BUCKET" macos-amd64 macos_artifacts/
 
+          echo Deleting old artifacts in B2
           python3 ci/delete_old_binaries.py "$BUCKET" "$DAYS_TO_KEEP"
 
+          echo Creating nightly.json
           python3 ci/create_nightly_json.py "$BUCKET" > nightly.json
+          
+          echo Uploading nightly.json
           b2 upload-file "$BUCKET" nightly.json nightly.json
 
+          echo Clear B2 account info
           b2 clear-account

Makefile

@@ -23,6 +23,8 @@ ifeq ($(OS), Linux)
     LLVM_CONFIG=llvm-config-11
     ifneq ($(shell which llvm-config-11 2>/dev/null),)
         LLVM_CONFIG=llvm-config-11
+    else ifneq ($(shell which llvm-config-11-64 2>/dev/null),)
+        LLVM_CONFIG=llvm-config-11-64
     else
         ifneq ($(shell llvm-config --version | grep '^11\.'),)
             LLVM_CONFIG=llvm-config
@@ -40,6 +42,9 @@ all: debug demo
 demo:
 	./odin run examples/demo/demo.odin
 
+report:
+	./odin report
+
 debug:
 	$(CC) src/main.cpp src/libtommath.cpp $(DISABLED_WARNINGS) $(CFLAGS) -g $(LDFLAGS) -o odin
 

README.md

@@ -78,7 +78,7 @@ Get live support and talk with other odiners on the Odin Discord.
 
 ### References
 
-#### [Language Specification](https://odin-lang.org/ref/spec)
+#### [Language Specification](https://odin-lang.org/docs/spec/)
 
 The official Odin Language specification.
 

bin/nasm/windows/LICENSE

@@ -0,0 +1,29 @@
+NASM is now licensed under the 2-clause BSD license, also known as the
+simplified BSD license.
+
+    Copyright 1996-2010 the NASM Authors - All rights reserved.
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following
+    conditions are met:
+
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+      copyright notice, this list of conditions and the following
+      disclaimer in the documentation and/or other materials provided
+      with the distribution.
+      
+      THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
+      CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
+      INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+      MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+      DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+      CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+      SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+      NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+      LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+      HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+      CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+      OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
+      EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

build.bat

@@ -79,4 +79,4 @@ if %release_mode% EQU 0 odin run examples/demo
 
 del *.obj > NUL 2> NUL
 
-:end_of_build
+:end_of_build

core/c/libc/complex.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/ctype.odin

@@ -2,6 +2,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/errno.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -38,6 +40,20 @@ when ODIN_OS == "windows" {
 	ERANGE :: 34
 }
 
+when ODIN_OS == "darwin" {
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		@(link_name="__error")
+		_get_errno :: proc() -> ^int ---
+	}
+
+	// Unknown
+	EDOM   :: 33
+	EILSEQ :: 92
+	ERANGE :: 34
+}
+
 // Odin has no way to make an identifier "errno" behave as a function call to
 // read the value, or to produce an lvalue such that you can assign a different
 // error value to errno. To work around this, just expose it as a function like

core/c/libc/math.odin

@@ -6,6 +6,8 @@ import "core:intrinsics"
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/setjmp.odin

@@ -4,10 +4,11 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
-
 when ODIN_OS == "windows" {
 	@(default_calling_convention="c")
 	foreign libc {

core/c/libc/signal.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -32,7 +34,20 @@ when ODIN_OS == "windows" {
 	SIGTERM :: 15
 }
 
-when ODIN_OS == "linux" || ODIN_OS == "freebsd" || ODIN_OS == "darwin" {
+when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
+	SIG_ERR  :: rawptr(~uintptr(0))
+	SIG_DFL  :: rawptr(uintptr(0))
+	SIG_IGN  :: rawptr(uintptr(1)) 
+
+	SIGABRT  :: 6
+	SIGFPE   :: 8
+	SIGILL   :: 4
+	SIGINT   :: 2
+	SIGSEGV  :: 11
+	SIGTERM  :: 15
+}
+
+when ODIN_OS == "darwin" {
 	SIG_ERR  :: rawptr(~uintptr(0))
 	SIG_DFL  :: rawptr(uintptr(0))
 	SIG_IGN  :: rawptr(uintptr(1)) 

core/c/libc/stdio.odin

@@ -2,6 +2,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -67,7 +69,7 @@ when ODIN_OS == "linux" {
 	SEEK_CUR      :: 1
 	SEEK_END      :: 2
 
-	TMP_MAX       :: 10000
+	TMP_MAX       :: 308915776
 
 	foreign libc {
 		stderr: ^FILE
@@ -76,6 +78,36 @@ when ODIN_OS == "linux" {
 	}
 }
 
+when ODIN_OS == "darwin" {
+	fpos_t :: distinct i64
+	
+	_IOFBF        :: 0
+	_IOLBF        :: 1
+	_IONBF        :: 2
+
+	BUFSIZ        :: 1024
+
+	EOF           :: int(-1)
+
+	FOPEN_MAX     :: 20
+
+	FILENAME_MAX  :: 1024
+
+	L_tmpnam      :: 1024
+
+	SEEK_SET      :: 0
+	SEEK_CUR      :: 1
+	SEEK_END      :: 2
+
+	TMP_MAX       :: 308915776
+
+	foreign libc {
+		@(link_name="__stderrp") stderr: ^FILE
+		@(link_name="__stdinp")  stdin:  ^FILE
+		@(link_name="__stdoutp") stdout: ^FILE
+	}
+}
+
 @(default_calling_convention="c")
 foreign libc {
 	// 7.21.4 Operations on files

core/c/libc/stdlib.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -33,7 +35,23 @@ when ODIN_OS == "linux" {
 	}
 
 	MB_CUR_MAX :: #force_inline proc() -> size_t {
-		return __ctype_get_mb_cur_max()
+		return size_t(__ctype_get_mb_cur_max())
+	}
+}
+
+
+when ODIN_OS == "darwin" {
+	RAND_MAX :: 0x7fffffff
+
+	// GLIBC and MUSL only
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		___mb_cur_max :: proc() -> int ---
+	}
+
+	MB_CUR_MAX :: #force_inline proc() -> size_t {
+		return size_t(___mb_cur_max())
 	}
 }
 

core/c/libc/string.odin

@@ -6,6 +6,8 @@ import "core:runtime"
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/threads.odin

@@ -136,3 +136,8 @@ when ODIN_OS == "linux" {
 		tss_set       :: proc(key: tss_t, val: rawptr) -> int ---
 	}
 }
+
+
+when ODIN_OS == "darwin" {
+	// TODO: find out what this is meant to be!
+}

core/c/libc/time.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -43,7 +45,7 @@ when ODIN_OS == "windows" {
 	}
 }
 
-when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
+when ODIN_OS == "linux" || ODIN_OS == "freebsd" || ODIN_OS == "darwin" {
 	@(default_calling_convention="c")
 	foreign libc {
 		// 7.27.2 Time manipulation functions
@@ -75,7 +77,7 @@ when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
 
 	tm :: struct {
 		tm_sec, tm_min, tm_hour, tm_mday, tm_mon, tm_year, tm_wday, tm_yday, tm_isdst: int,
-		_: long,
-		_: rawptr,
+		tm_gmtoff: long,
+		tm_zone: rawptr,
 	}
 }

core/c/libc/uchar.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/wchar.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

core/c/libc/wctype.odin

@@ -4,6 +4,8 @@ package libc
 
 when ODIN_OS == "windows" {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == "darwin" {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -14,10 +16,15 @@ when ODIN_OS == "windows" {
 }
 
 when ODIN_OS == "linux" {
-	wctrans_t :: distinct rawptr
+	wctrans_t :: distinct intptr_t
 	wctype_t  :: distinct ulong
 }
 
+when ODIN_OS == "darwin" {
+	wctrans_t :: distinct int
+	wctype_t  :: distinct u32
+}
+
 @(default_calling_convention="c")
 foreign libc {
 	// 7.30.2.1 Wide character classification functions

core/compress/common.odin

@@ -1,5 +1,3 @@
-package compress
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
 	Made available under Odin's BSD-3 license.
@@ -7,10 +5,11 @@ package compress
 	List of contributors:
 		Jeroen van Rijn: Initial implementation, optimization.
 */
+package compress
 
 import "core:io"
-import "core:image"
 import "core:bytes"
+import "core:runtime"
 
 /*
 	These settings bound how much compression algorithms will allocate for their output buffer.
@@ -51,11 +50,8 @@ Error :: union {
 	ZLIB_Error,
 	GZIP_Error,
 	ZIP_Error,
-	/*
-		This is here because png.load will return a this type of error union,
-		as it may involve an I/O error, a Deflate error, etc.
-	*/
-	image.Error,
+
+	runtime.Allocator_Error,
 }
 
 General_Error :: enum {
@@ -69,7 +65,6 @@ General_Error :: enum {
 	Incompatible_Options,
 	Unimplemented,
 
-
 	/*
 		Memory errors
 	*/

core/crypto/README.md

@@ -0,0 +1,89 @@
+# crypto
+A crypto library for the Odin language
+
+## Supported
+This library offers various algorithms implemented in Odin.
+Please see the chart below for the options.  
+
+## Hashing algorithms
+| Algorithm                                                                                                    |                  |
+|:-------------------------------------------------------------------------------------------------------------|:-----------------|
+| [BLAKE](https://web.archive.org/web/20190915215948/https://131002.net/blake)                                 | ✔️ |
+| [BLAKE2B](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | ✔️ |
+| [BLAKE2S](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | ✔️ |
+| [GOST](https://datatracker.ietf.org/doc/html/rfc5831)                                                        | ✔️ |
+| [Grøstl](http://www.groestl.info/Groestl.zip)                                                                | ✔️ |
+| [HAVAL](https://web.archive.org/web/20150111210116/http://labs.calyptix.com/haval.php)                       | ✔️ |
+| [JH](https://www3.ntu.edu.sg/home/wuhj/research/jh/index.html)                                               | ✔️ |
+| [Keccak](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                           | ✔️ |
+| [MD2](https://datatracker.ietf.org/doc/html/rfc1319)                                                         | ✔️ |
+| [MD4](https://datatracker.ietf.org/doc/html/rfc1320)                                                         | ✔️ |
+| [MD5](https://datatracker.ietf.org/doc/html/rfc1321)                                                         | ✔️ |
+| [RIPEMD](https://homes.esat.kuleuven.be/~bosselae/ripemd160.html)                                            | ✔️ |
+| [SHA-1](https://datatracker.ietf.org/doc/html/rfc3174)                                                       | ✔️ |
+| [SHA-2](https://csrc.nist.gov/csrc/media/publications/fips/180/2/archive/2002-08-01/documents/fips180-2.pdf) | ✔️ |
+| [SHA-3](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | ✔️ |
+| [SHAKE](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | ✔️ |
+| [SM3](https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02)                                           | ✔️ |
+| [Streebog](https://datatracker.ietf.org/doc/html/rfc6986)                                                    | ✔️ |
+| [Tiger](https://www.cs.technion.ac.il/~biham/Reports/Tiger/)                                                 | ✔️ |
+| [Tiger2](https://www.cs.technion.ac.il/~biham/Reports/Tiger/)                                                | ✔️ |
+| [Whirlpool](https://web.archive.org/web/20171129084214/http://www.larc.usp.br/~pbarreto/WhirlpoolPage.html)  | ✔️ |
+
+#### High level API
+Each hash algorithm contains a procedure group named `hash`, or if the algorithm provides more than one digest size `hash_<size>`\*.  
+Included in these groups are four procedures.
+* `hash_string` - Hash a given string and return the computed hash. Just calls `hash_bytes` internally
+* `hash_bytes` - Hash a given byte slice and return the computed hash
+* `hash_stream` - Takes a stream from io.Stream and returns the computed hash from it
+* `hash_file` - Takes a file handle and returns the computed hash from it. A second optional boolean parameter controls if the file is streamed (this is the default) or read at once (set to true)
+
+\* On some algorithms there is another part to the name, since they might offer control about additional parameters.  
+For instance, `HAVAL` offers different sizes as well as three different round amounts.  
+Computing a 256-bit hash with 3 rounds is therefore achieved by calling `haval.hash_256_3(...)`.
+
+#### Low level API
+The above mentioned procedures internally call three procedures: `init`, `update` and `final`.
+You may also directly call them, if you wish.
+
+#### Example
+```odin
+package crypto_example
+
+// Import the desired package
+import "core:crypto/md4"
+
+main :: proc() {
+    input := "foo"
+
+    // Compute the hash, using the high level API
+    computed_hash := md4.hash(input)
+
+    // Compute the hash, using the low level API
+    ctx: md4.Md4_Context
+    computed_hash_low: [16]byte
+    md4.init(&ctx)
+    md4.update(&ctx, transmute([]byte)input)
+    md4.final(&ctx, computed_hash_low[:])
+}
+```
+For example uses of all available algorithms, please see the tests within `tests/core/crypto`.
+
+#### Thread safety
+The crypto package is not thread-safe at the moment. This may change in the future.
+
+### Disclaimer
+The algorithms were ported out of curiosity and due to interest in the field.
+We have not had any of the code verified by a third party or tested/fuzzed by any automatic means.
+Whereever we were able to find official test vectors, those were used to verify the implementation.
+We do not recommend using them in a production environment, without any additional testing and/or verification.
+
+### ToDo
+* Ciphers (Symmetric, Asymmetric)
+* MACs (Message Authentication Code)
+* CSPRNGs (Cryptographically Secure PseudoRandom Number Generator)
+* KDFs (Key Derivation Function)
+* KEAs (Key Exchange Algorithm)
+
+### License
+This library is made available under the BSD-3 license.

core/crypto/_fiat/README.md

@@ -0,0 +1,35 @@
+# fiat
+
+This package contains low level arithmetic required to implement certain
+cryptographic primitives, ported from the [fiat-crypto project][1]
+along with some higher-level helpers.
+
+## Notes
+
+fiat-crypto gives the choice of 3 licenses for derived works.  The 1-Clause
+BSD license is chosen as it is compatible with Odin's existing licensing.
+
+The routines are intended to be timing-safe, as long as the underlying
+integer arithmetic is constant time.  This is true on most systems commonly
+used today, with the notable exception of WASM.
+
+While fiat-crypto provides both output targeting both 32-bit and 64-bit
+architectures, only the 64-bit versions were used, as 32-bit architectures
+are becoming increasingly uncommon and irrelevant.
+
+With the current Odin syntax, the Go output is trivially ported in most
+cases and was used as the basis of the port.
+
+In the future, it would be better to auto-generate Odin either directly
+by adding an appropriate code-gen backend written in Coq, or perhaps by
+parsing the JSON output.
+
+As this is a port rather than autogenerated output, none of fiat-crypto's
+formal verification guarantees apply, unless it is possible to prove binary
+equivalence.
+
+For the most part, alterations to the base fiat-crypto generated code was
+kept to a minimum, to aid auditability.  This results in a somewhat
+ideosyncratic style, and in some cases minor performance penalties.
+
+[1]: https://github.com/mit-plv/fiat-crypto

core/crypto/_fiat/fiat.odin

@@ -0,0 +1,24 @@
+package fiat
+
+// This package provides various helpers and types common to all of the
+// fiat-crypto derived backends.
+
+// This code only works on a two's complement system.
+#assert((-1 & 3) == 3)
+
+u1 :: distinct u8
+i1 :: distinct i8
+
+cmovznz_u64 :: #force_inline proc "contextless" (arg1: u1, arg2, arg3: u64) -> (out1: u64) {
+	x1 := (u64(arg1) * 0xffffffffffffffff)
+	x2 := ((x1 & arg3) | ((~x1) & arg2))
+	out1 = x2
+	return
+}
+
+cmovznz_u32 :: #force_inline proc "contextless" (arg1: u1, arg2, arg3: u32) -> (out1: u32) {
+	x1 := (u32(arg1) * 0xffffffff)
+	x2 := ((x1 & arg3) | ((~x1) & arg2))
+	out1 = x2
+	return
+}

core/crypto/_fiat/field_curve25519/field.odin

@@ -0,0 +1,138 @@
+package field_curve25519
+
+import "core:crypto"
+import "core:mem"
+
+fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
+	return transmute(^Loose_Field_Element)(arg1)
+}
+
+fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_Field_Element) -> ^Tight_Field_Element {
+	return transmute(^Tight_Field_Element)(arg1)
+}
+
+fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	// Ignore the unused bit by copying the input and masking the bit off
+	// prior to deserialization.
+	tmp1: [32]byte = ---
+	copy_slice(tmp1[:], arg1[:])
+	tmp1[31] &= 127
+
+	_fe_from_bytes(out1, &tmp1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+}
+
+fe_equal :: proc "contextless" (arg1, arg2: ^Tight_Field_Element) -> int {
+	tmp2: [32]byte = ---
+
+	fe_to_bytes(&tmp2, arg2)
+	ret := fe_equal_bytes(arg1, &tmp2)
+
+	mem.zero_explicit(&tmp2, size_of(tmp2))
+
+	return ret
+}
+
+fe_equal_bytes :: proc "contextless" (arg1: ^Tight_Field_Element, arg2: ^[32]byte) -> int {
+	tmp1: [32]byte = ---
+
+	fe_to_bytes(&tmp1, arg1)
+
+	ret := crypto.compare_constant_time(tmp1[:], arg2[:])
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+
+	return ret
+}
+
+fe_carry_pow2k :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element, arg2: uint) {
+	// Special case: `arg1^(2 * 0) = 1`, though this should never happen.
+	if arg2 == 0 {
+		fe_one(out1)
+		return
+	}
+
+	fe_carry_square(out1, arg1)
+	for _ in 1..<arg2 {
+		fe_carry_square(out1, fe_relax_cast(out1))
+	}
+}
+
+fe_carry_opp :: #force_inline proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	fe_opp(fe_relax_cast(out1), arg1)
+	fe_carry(out1, fe_relax_cast(out1))
+}
+
+fe_carry_invsqrt :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) -> int {
+	// Inverse square root taken from Monocypher.
+
+	tmp1, tmp2, tmp3: Tight_Field_Element = ---, ---, ---
+
+	// t0 = x^((p-5)/8)
+	// Can be achieved with a simple double & add ladder,
+	// but it would be slower.
+	fe_carry_pow2k(&tmp1, arg1, 1)
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 2)
+	fe_carry_mul(&tmp2, arg1, fe_relax_cast(&tmp2))
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp1), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp1, fe_relax_cast(&tmp1), 1)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 5)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 10)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp3, fe_relax_cast(&tmp2), 20)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp3), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp2), 10)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 50)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp3, fe_relax_cast(&tmp2), 100)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp3), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp2), 50)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp1, fe_relax_cast(&tmp1), 2)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp1), arg1)
+
+	// quartic = x^((p-1)/4)
+	quartic := &tmp2
+	fe_carry_square(quartic, fe_relax_cast(&tmp1))
+	fe_carry_mul(quartic, fe_relax_cast(quartic), arg1)
+
+	// Serialize quartic once to save on repeated serialization/sanitization.
+	quartic_buf: [32]byte = ---
+	fe_to_bytes(&quartic_buf, quartic)
+	check := &tmp3
+
+	fe_one(check)
+	p1 := fe_equal_bytes(check, &quartic_buf)
+	fe_carry_opp(check, check)
+	m1 := fe_equal_bytes(check, &quartic_buf)
+	fe_carry_opp(check, &SQRT_M1)
+	ms := fe_equal_bytes(check, &quartic_buf)
+
+	// if quartic == -1 or sqrt(-1)
+	// then  isr = x^((p-1)/4) * sqrt(-1)
+	// else  isr = x^((p-1)/4)
+	fe_carry_mul(out1, fe_relax_cast(&tmp1), fe_relax_cast(&SQRT_M1))
+	fe_cond_assign(out1, &tmp1, (m1|ms) ~ 1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+	mem.zero_explicit(&tmp2, size_of(tmp2))
+	mem.zero_explicit(&tmp3, size_of(tmp3))
+	mem.zero_explicit(&quartic_buf, size_of(quartic_buf))
+
+	return p1 | m1
+}
+
+fe_carry_inv :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	tmp1: Tight_Field_Element
+
+	fe_carry_square(&tmp1, arg1)
+	_ = fe_carry_invsqrt(&tmp1, fe_relax_cast(&tmp1))
+	fe_carry_square(&tmp1, fe_relax_cast(&tmp1))
+	fe_carry_mul(out1, fe_relax_cast(&tmp1), arg1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+}

core/crypto/_fiat/field_curve25519/field51.odin

@@ -0,0 +1,616 @@
+// The BSD 1-Clause License (BSD-1-Clause)
+//
+// Copyright (c) 2015-2020 the fiat-crypto authors (see the AUTHORS file)
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     1. Redistributions of source code must retain the above copyright
+//        notice, this list of conditions and the following disclaimer.
+//
+// THIS SOFTWARE IS PROVIDED BY the fiat-crypto authors "AS IS"
+// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL Berkeley Software Design,
+// Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+package field_curve25519
+
+// The file provides arithmetic on the field Z/(2^255-19) using
+// unsaturated 64-bit integer arithmetic.  It is derived primarily
+// from the machine generated Golang output from the fiat-crypto project.
+//
+// While the base implementation is provably correct, this implementation
+// makes no such claims as the port and optimizations were done by hand.
+// At some point, it may be worth adding support to fiat-crypto for
+// generating Odin output.
+//
+// TODO:
+//  * When fiat-crypto supports it, using a saturated 64-bit limbs
+//    instead of 51-bit limbs will be faster, though the gains are
+//    minimal unless adcx/adox/mulx are used.
+
+import fiat "core:crypto/_fiat"
+import "core:math/bits"
+
+Loose_Field_Element :: distinct [5]u64
+Tight_Field_Element :: distinct [5]u64
+
+SQRT_M1 := Tight_Field_Element{
+	1718705420411056,
+	234908883556509,
+	2233514472574048,
+	2117202627021982,
+	765476049583133,
+}
+
+_addcarryx_u51 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0x7ffffffffffff)
+	x3 := fiat.u1((x1 >> 51))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u51 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 51))
+	x3 := (u64(x1) & 0x7ffffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+fe_carry_mul :: proc (out1: ^Tight_Field_Element, arg1, arg2: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(arg1[4], (arg2[4] * 0x13))
+	x4, x3 := bits.mul_u64(arg1[4], (arg2[3] * 0x13))
+	x6, x5 := bits.mul_u64(arg1[4], (arg2[2] * 0x13))
+	x8, x7 := bits.mul_u64(arg1[4], (arg2[1] * 0x13))
+	x10, x9 := bits.mul_u64(arg1[3], (arg2[4] * 0x13))
+	x12, x11 := bits.mul_u64(arg1[3], (arg2[3] * 0x13))
+	x14, x13 := bits.mul_u64(arg1[3], (arg2[2] * 0x13))
+	x16, x15 := bits.mul_u64(arg1[2], (arg2[4] * 0x13))
+	x18, x17 := bits.mul_u64(arg1[2], (arg2[3] * 0x13))
+	x20, x19 := bits.mul_u64(arg1[1], (arg2[4] * 0x13))
+	x22, x21 := bits.mul_u64(arg1[4], arg2[0])
+	x24, x23 := bits.mul_u64(arg1[3], arg2[1])
+	x26, x25 := bits.mul_u64(arg1[3], arg2[0])
+	x28, x27 := bits.mul_u64(arg1[2], arg2[2])
+	x30, x29 := bits.mul_u64(arg1[2], arg2[1])
+	x32, x31 := bits.mul_u64(arg1[2], arg2[0])
+	x34, x33 := bits.mul_u64(arg1[1], arg2[3])
+	x36, x35 := bits.mul_u64(arg1[1], arg2[2])
+	x38, x37 := bits.mul_u64(arg1[1], arg2[1])
+	x40, x39 := bits.mul_u64(arg1[1], arg2[0])
+	x42, x41 := bits.mul_u64(arg1[0], arg2[4])
+	x44, x43 := bits.mul_u64(arg1[0], arg2[3])
+	x46, x45 := bits.mul_u64(arg1[0], arg2[2])
+	x48, x47 := bits.mul_u64(arg1[0], arg2[1])
+	x50, x49 := bits.mul_u64(arg1[0], arg2[0])
+	x51, x52 := bits.add_u64(x13, x7, u64(0x0))
+	x53, _ := bits.add_u64(x14, x8, u64(fiat.u1(x52)))
+	x55, x56 := bits.add_u64(x17, x51, u64(0x0))
+	x57, _ := bits.add_u64(x18, x53, u64(fiat.u1(x56)))
+	x59, x60 := bits.add_u64(x19, x55, u64(0x0))
+	x61, _ := bits.add_u64(x20, x57, u64(fiat.u1(x60)))
+	x63, x64 := bits.add_u64(x49, x59, u64(0x0))
+	x65, _ := bits.add_u64(x50, x61, u64(fiat.u1(x64)))
+	x67 := ((x63 >> 51) | ((x65 << 13) & 0xffffffffffffffff))
+	x68 := (x63 & 0x7ffffffffffff)
+	x69, x70 := bits.add_u64(x23, x21, u64(0x0))
+	x71, _ := bits.add_u64(x24, x22, u64(fiat.u1(x70)))
+	x73, x74 := bits.add_u64(x27, x69, u64(0x0))
+	x75, _ := bits.add_u64(x28, x71, u64(fiat.u1(x74)))
+	x77, x78 := bits.add_u64(x33, x73, u64(0x0))
+	x79, _ := bits.add_u64(x34, x75, u64(fiat.u1(x78)))
+	x81, x82 := bits.add_u64(x41, x77, u64(0x0))
+	x83, _ := bits.add_u64(x42, x79, u64(fiat.u1(x82)))
+	x85, x86 := bits.add_u64(x25, x1, u64(0x0))
+	x87, _ := bits.add_u64(x26, x2, u64(fiat.u1(x86)))
+	x89, x90 := bits.add_u64(x29, x85, u64(0x0))
+	x91, _ := bits.add_u64(x30, x87, u64(fiat.u1(x90)))
+	x93, x94 := bits.add_u64(x35, x89, u64(0x0))
+	x95, _ := bits.add_u64(x36, x91, u64(fiat.u1(x94)))
+	x97, x98 := bits.add_u64(x43, x93, u64(0x0))
+	x99, _ := bits.add_u64(x44, x95, u64(fiat.u1(x98)))
+	x101, x102 := bits.add_u64(x9, x3, u64(0x0))
+	x103, _ := bits.add_u64(x10, x4, u64(fiat.u1(x102)))
+	x105, x106 := bits.add_u64(x31, x101, u64(0x0))
+	x107, _ := bits.add_u64(x32, x103, u64(fiat.u1(x106)))
+	x109, x110 := bits.add_u64(x37, x105, u64(0x0))
+	x111, _ := bits.add_u64(x38, x107, u64(fiat.u1(x110)))
+	x113, x114 := bits.add_u64(x45, x109, u64(0x0))
+	x115, _ := bits.add_u64(x46, x111, u64(fiat.u1(x114)))
+	x117, x118 := bits.add_u64(x11, x5, u64(0x0))
+	x119, _ := bits.add_u64(x12, x6, u64(fiat.u1(x118)))
+	x121, x122 := bits.add_u64(x15, x117, u64(0x0))
+	x123, _ := bits.add_u64(x16, x119, u64(fiat.u1(x122)))
+	x125, x126 := bits.add_u64(x39, x121, u64(0x0))
+	x127, _ := bits.add_u64(x40, x123, u64(fiat.u1(x126)))
+	x129, x130 := bits.add_u64(x47, x125, u64(0x0))
+	x131, _ := bits.add_u64(x48, x127, u64(fiat.u1(x130)))
+	x133, x134 := bits.add_u64(x67, x129, u64(0x0))
+	x135 := (u64(fiat.u1(x134)) + x131)
+	x136 := ((x133 >> 51) | ((x135 << 13) & 0xffffffffffffffff))
+	x137 := (x133 & 0x7ffffffffffff)
+	x138, x139 := bits.add_u64(x136, x113, u64(0x0))
+	x140 := (u64(fiat.u1(x139)) + x115)
+	x141 := ((x138 >> 51) | ((x140 << 13) & 0xffffffffffffffff))
+	x142 := (x138 & 0x7ffffffffffff)
+	x143, x144 := bits.add_u64(x141, x97, u64(0x0))
+	x145 := (u64(fiat.u1(x144)) + x99)
+	x146 := ((x143 >> 51) | ((x145 << 13) & 0xffffffffffffffff))
+	x147 := (x143 & 0x7ffffffffffff)
+	x148, x149 := bits.add_u64(x146, x81, u64(0x0))
+	x150 := (u64(fiat.u1(x149)) + x83)
+	x151 := ((x148 >> 51) | ((x150 << 13) & 0xffffffffffffffff))
+	x152 := (x148 & 0x7ffffffffffff)
+	x153 := (x151 * 0x13)
+	x154 := (x68 + x153)
+	x155 := (x154 >> 51)
+	x156 := (x154 & 0x7ffffffffffff)
+	x157 := (x155 + x137)
+	x158 := fiat.u1((x157 >> 51))
+	x159 := (x157 & 0x7ffffffffffff)
+	x160 := (u64(x158) + x142)
+	out1[0] = x156
+	out1[1] = x159
+	out1[2] = x160
+	out1[3] = x147
+	out1[4] = x152
+}
+
+fe_carry_square :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := (arg1[4] * 0x13)
+	x2 := (x1 * 0x2)
+	x3 := (arg1[4] * 0x2)
+	x4 := (arg1[3] * 0x13)
+	x5 := (x4 * 0x2)
+	x6 := (arg1[3] * 0x2)
+	x7 := (arg1[2] * 0x2)
+	x8 := (arg1[1] * 0x2)
+	x10, x9 := bits.mul_u64(arg1[4], x1)
+	x12, x11 := bits.mul_u64(arg1[3], x2)
+	x14, x13 := bits.mul_u64(arg1[3], x4)
+	x16, x15 := bits.mul_u64(arg1[2], x2)
+	x18, x17 := bits.mul_u64(arg1[2], x5)
+	x20, x19 := bits.mul_u64(arg1[2], arg1[2])
+	x22, x21 := bits.mul_u64(arg1[1], x2)
+	x24, x23 := bits.mul_u64(arg1[1], x6)
+	x26, x25 := bits.mul_u64(arg1[1], x7)
+	x28, x27 := bits.mul_u64(arg1[1], arg1[1])
+	x30, x29 := bits.mul_u64(arg1[0], x3)
+	x32, x31 := bits.mul_u64(arg1[0], x6)
+	x34, x33 := bits.mul_u64(arg1[0], x7)
+	x36, x35 := bits.mul_u64(arg1[0], x8)
+	x38, x37 := bits.mul_u64(arg1[0], arg1[0])
+	x39, x40 := bits.add_u64(x21, x17, u64(0x0))
+	x41, _ := bits.add_u64(x22, x18, u64(fiat.u1(x40)))
+	x43, x44 := bits.add_u64(x37, x39, u64(0x0))
+	x45, _ := bits.add_u64(x38, x41, u64(fiat.u1(x44)))
+	x47 := ((x43 >> 51) | ((x45 << 13) & 0xffffffffffffffff))
+	x48 := (x43 & 0x7ffffffffffff)
+	x49, x50 := bits.add_u64(x23, x19, u64(0x0))
+	x51, _ := bits.add_u64(x24, x20, u64(fiat.u1(x50)))
+	x53, x54 := bits.add_u64(x29, x49, u64(0x0))
+	x55, _ := bits.add_u64(x30, x51, u64(fiat.u1(x54)))
+	x57, x58 := bits.add_u64(x25, x9, u64(0x0))
+	x59, _ := bits.add_u64(x26, x10, u64(fiat.u1(x58)))
+	x61, x62 := bits.add_u64(x31, x57, u64(0x0))
+	x63, _ := bits.add_u64(x32, x59, u64(fiat.u1(x62)))
+	x65, x66 := bits.add_u64(x27, x11, u64(0x0))
+	x67, _ := bits.add_u64(x28, x12, u64(fiat.u1(x66)))
+	x69, x70 := bits.add_u64(x33, x65, u64(0x0))
+	x71, _ := bits.add_u64(x34, x67, u64(fiat.u1(x70)))
+	x73, x74 := bits.add_u64(x15, x13, u64(0x0))
+	x75, _ := bits.add_u64(x16, x14, u64(fiat.u1(x74)))
+	x77, x78 := bits.add_u64(x35, x73, u64(0x0))
+	x79, _ := bits.add_u64(x36, x75, u64(fiat.u1(x78)))
+	x81, x82 := bits.add_u64(x47, x77, u64(0x0))
+	x83 := (u64(fiat.u1(x82)) + x79)
+	x84 := ((x81 >> 51) | ((x83 << 13) & 0xffffffffffffffff))
+	x85 := (x81 & 0x7ffffffffffff)
+	x86, x87 := bits.add_u64(x84, x69, u64(0x0))
+	x88 := (u64(fiat.u1(x87)) + x71)
+	x89 := ((x86 >> 51) | ((x88 << 13) & 0xffffffffffffffff))
+	x90 := (x86 & 0x7ffffffffffff)
+	x91, x92 := bits.add_u64(x89, x61, u64(0x0))
+	x93 := (u64(fiat.u1(x92)) + x63)
+	x94 := ((x91 >> 51) | ((x93 << 13) & 0xffffffffffffffff))
+	x95 := (x91 & 0x7ffffffffffff)
+	x96, x97 := bits.add_u64(x94, x53, u64(0x0))
+	x98 := (u64(fiat.u1(x97)) + x55)
+	x99 := ((x96 >> 51) | ((x98 << 13) & 0xffffffffffffffff))
+	x100 := (x96 & 0x7ffffffffffff)
+	x101 := (x99 * 0x13)
+	x102 := (x48 + x101)
+	x103 := (x102 >> 51)
+	x104 := (x102 & 0x7ffffffffffff)
+	x105 := (x103 + x85)
+	x106 := fiat.u1((x105 >> 51))
+	x107 := (x105 & 0x7ffffffffffff)
+	x108 := (u64(x106) + x90)
+	out1[0] = x104
+	out1[1] = x107
+	out1[2] = x108
+	out1[3] = x95
+	out1[4] = x100
+}
+
+fe_carry :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := arg1[0]
+	x2 := ((x1 >> 51) + arg1[1])
+	x3 := ((x2 >> 51) + arg1[2])
+	x4 := ((x3 >> 51) + arg1[3])
+	x5 := ((x4 >> 51) + arg1[4])
+	x6 := ((x1 & 0x7ffffffffffff) + ((x5 >> 51) * 0x13))
+	x7 := (u64(fiat.u1((x6 >> 51))) + (x2 & 0x7ffffffffffff))
+	x8 := (x6 & 0x7ffffffffffff)
+	x9 := (x7 & 0x7ffffffffffff)
+	x10 := (u64(fiat.u1((x7 >> 51))) + (x3 & 0x7ffffffffffff))
+	x11 := (x4 & 0x7ffffffffffff)
+	x12 := (x5 & 0x7ffffffffffff)
+	out1[0] = x8
+	out1[1] = x9
+	out1[2] = x10
+	out1[3] = x11
+	out1[4] = x12
+}
+
+fe_add :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := (arg1[0] + arg2[0])
+	x2 := (arg1[1] + arg2[1])
+	x3 := (arg1[2] + arg2[2])
+	x4 := (arg1[3] + arg2[3])
+	x5 := (arg1[4] + arg2[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_sub :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := ((0xfffffffffffda + arg1[0]) - arg2[0])
+	x2 := ((0xffffffffffffe + arg1[1]) - arg2[1])
+	x3 := ((0xffffffffffffe + arg1[2]) - arg2[2])
+	x4 := ((0xffffffffffffe + arg1[3]) - arg2[3])
+	x5 := ((0xffffffffffffe + arg1[4]) - arg2[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_opp :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := (0xfffffffffffda - arg1[0])
+	x2 := (0xffffffffffffe - arg1[1])
+	x3 := (0xffffffffffffe - arg1[2])
+	x4 := (0xffffffffffffe - arg1[3])
+	x5 := (0xffffffffffffe - arg1[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_cond_assign :: proc "contextless" (out1, arg1: ^Tight_Field_Element, arg2: int) {
+	x1 := fiat.cmovznz_u64(fiat.u1(arg2), out1[0], arg1[0])
+	x2 := fiat.cmovznz_u64(fiat.u1(arg2), out1[1], arg1[1])
+	x3 := fiat.cmovznz_u64(fiat.u1(arg2), out1[2], arg1[2])
+	x4 := fiat.cmovznz_u64(fiat.u1(arg2), out1[3], arg1[3])
+	x5 := fiat.cmovznz_u64(fiat.u1(arg2), out1[4], arg1[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_to_bytes :: proc "contextless" (out1: ^[32]byte, arg1: ^Tight_Field_Element) {
+	x1, x2 := _subborrowx_u51(0x0, arg1[0], 0x7ffffffffffed)
+	x3, x4 := _subborrowx_u51(x2, arg1[1], 0x7ffffffffffff)
+	x5, x6 := _subborrowx_u51(x4, arg1[2], 0x7ffffffffffff)
+	x7, x8 := _subborrowx_u51(x6, arg1[3], 0x7ffffffffffff)
+	x9, x10 := _subborrowx_u51(x8, arg1[4], 0x7ffffffffffff)
+	x11 := fiat.cmovznz_u64(x10, u64(0x0), 0xffffffffffffffff)
+	x12, x13 := _addcarryx_u51(0x0, x1, (x11 & 0x7ffffffffffed))
+	x14, x15 := _addcarryx_u51(x13, x3, (x11 & 0x7ffffffffffff))
+	x16, x17 := _addcarryx_u51(x15, x5, (x11 & 0x7ffffffffffff))
+	x18, x19 := _addcarryx_u51(x17, x7, (x11 & 0x7ffffffffffff))
+	x20, _ := _addcarryx_u51(x19, x9, (x11 & 0x7ffffffffffff))
+	x22 := (x20 << 4)
+	x23 := (x18 * u64(0x2))
+	x24 := (x16 << 6)
+	x25 := (x14 << 3)
+	x26 := (u8(x12) & 0xff)
+	x27 := (x12 >> 8)
+	x28 := (u8(x27) & 0xff)
+	x29 := (x27 >> 8)
+	x30 := (u8(x29) & 0xff)
+	x31 := (x29 >> 8)
+	x32 := (u8(x31) & 0xff)
+	x33 := (x31 >> 8)
+	x34 := (u8(x33) & 0xff)
+	x35 := (x33 >> 8)
+	x36 := (u8(x35) & 0xff)
+	x37 := u8((x35 >> 8))
+	x38 := (x25 + u64(x37))
+	x39 := (u8(x38) & 0xff)
+	x40 := (x38 >> 8)
+	x41 := (u8(x40) & 0xff)
+	x42 := (x40 >> 8)
+	x43 := (u8(x42) & 0xff)
+	x44 := (x42 >> 8)
+	x45 := (u8(x44) & 0xff)
+	x46 := (x44 >> 8)
+	x47 := (u8(x46) & 0xff)
+	x48 := (x46 >> 8)
+	x49 := (u8(x48) & 0xff)
+	x50 := u8((x48 >> 8))
+	x51 := (x24 + u64(x50))
+	x52 := (u8(x51) & 0xff)
+	x53 := (x51 >> 8)
+	x54 := (u8(x53) & 0xff)
+	x55 := (x53 >> 8)
+	x56 := (u8(x55) & 0xff)
+	x57 := (x55 >> 8)
+	x58 := (u8(x57) & 0xff)
+	x59 := (x57 >> 8)
+	x60 := (u8(x59) & 0xff)
+	x61 := (x59 >> 8)
+	x62 := (u8(x61) & 0xff)
+	x63 := (x61 >> 8)
+	x64 := (u8(x63) & 0xff)
+	x65 := fiat.u1((x63 >> 8))
+	x66 := (x23 + u64(x65))
+	x67 := (u8(x66) & 0xff)
+	x68 := (x66 >> 8)
+	x69 := (u8(x68) & 0xff)
+	x70 := (x68 >> 8)
+	x71 := (u8(x70) & 0xff)
+	x72 := (x70 >> 8)
+	x73 := (u8(x72) & 0xff)
+	x74 := (x72 >> 8)
+	x75 := (u8(x74) & 0xff)
+	x76 := (x74 >> 8)
+	x77 := (u8(x76) & 0xff)
+	x78 := u8((x76 >> 8))
+	x79 := (x22 + u64(x78))
+	x80 := (u8(x79) & 0xff)
+	x81 := (x79 >> 8)
+	x82 := (u8(x81) & 0xff)
+	x83 := (x81 >> 8)
+	x84 := (u8(x83) & 0xff)
+	x85 := (x83 >> 8)
+	x86 := (u8(x85) & 0xff)
+	x87 := (x85 >> 8)
+	x88 := (u8(x87) & 0xff)
+	x89 := (x87 >> 8)
+	x90 := (u8(x89) & 0xff)
+	x91 := u8((x89 >> 8))
+	out1[0] = x26
+	out1[1] = x28
+	out1[2] = x30
+	out1[3] = x32
+	out1[4] = x34
+	out1[5] = x36
+	out1[6] = x39
+	out1[7] = x41
+	out1[8] = x43
+	out1[9] = x45
+	out1[10] = x47
+	out1[11] = x49
+	out1[12] = x52
+	out1[13] = x54
+	out1[14] = x56
+	out1[15] = x58
+	out1[16] = x60
+	out1[17] = x62
+	out1[18] = x64
+	out1[19] = x67
+	out1[20] = x69
+	out1[21] = x71
+	out1[22] = x73
+	out1[23] = x75
+	out1[24] = x77
+	out1[25] = x80
+	out1[26] = x82
+	out1[27] = x84
+	out1[28] = x86
+	out1[29] = x88
+	out1[30] = x90
+	out1[31] = x91
+}
+
+_fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	x1 := (u64(arg1[31]) << 44)
+	x2 := (u64(arg1[30]) << 36)
+	x3 := (u64(arg1[29]) << 28)
+	x4 := (u64(arg1[28]) << 20)
+	x5 := (u64(arg1[27]) << 12)
+	x6 := (u64(arg1[26]) << 4)
+	x7 := (u64(arg1[25]) << 47)
+	x8 := (u64(arg1[24]) << 39)
+	x9 := (u64(arg1[23]) << 31)
+	x10 := (u64(arg1[22]) << 23)
+	x11 := (u64(arg1[21]) << 15)
+	x12 := (u64(arg1[20]) << 7)
+	x13 := (u64(arg1[19]) << 50)
+	x14 := (u64(arg1[18]) << 42)
+	x15 := (u64(arg1[17]) << 34)
+	x16 := (u64(arg1[16]) << 26)
+	x17 := (u64(arg1[15]) << 18)
+	x18 := (u64(arg1[14]) << 10)
+	x19 := (u64(arg1[13]) << 2)
+	x20 := (u64(arg1[12]) << 45)
+	x21 := (u64(arg1[11]) << 37)
+	x22 := (u64(arg1[10]) << 29)
+	x23 := (u64(arg1[9]) << 21)
+	x24 := (u64(arg1[8]) << 13)
+	x25 := (u64(arg1[7]) << 5)
+	x26 := (u64(arg1[6]) << 48)
+	x27 := (u64(arg1[5]) << 40)
+	x28 := (u64(arg1[4]) << 32)
+	x29 := (u64(arg1[3]) << 24)
+	x30 := (u64(arg1[2]) << 16)
+	x31 := (u64(arg1[1]) << 8)
+	x32 := arg1[0]
+	x33 := (x31 + u64(x32))
+	x34 := (x30 + x33)
+	x35 := (x29 + x34)
+	x36 := (x28 + x35)
+	x37 := (x27 + x36)
+	x38 := (x26 + x37)
+	x39 := (x38 & 0x7ffffffffffff)
+	x40 := u8((x38 >> 51))
+	x41 := (x25 + u64(x40))
+	x42 := (x24 + x41)
+	x43 := (x23 + x42)
+	x44 := (x22 + x43)
+	x45 := (x21 + x44)
+	x46 := (x20 + x45)
+	x47 := (x46 & 0x7ffffffffffff)
+	x48 := u8((x46 >> 51))
+	x49 := (x19 + u64(x48))
+	x50 := (x18 + x49)
+	x51 := (x17 + x50)
+	x52 := (x16 + x51)
+	x53 := (x15 + x52)
+	x54 := (x14 + x53)
+	x55 := (x13 + x54)
+	x56 := (x55 & 0x7ffffffffffff)
+	x57 := u8((x55 >> 51))
+	x58 := (x12 + u64(x57))
+	x59 := (x11 + x58)
+	x60 := (x10 + x59)
+	x61 := (x9 + x60)
+	x62 := (x8 + x61)
+	x63 := (x7 + x62)
+	x64 := (x63 & 0x7ffffffffffff)
+	x65 := u8((x63 >> 51))
+	x66 := (x6 + u64(x65))
+	x67 := (x5 + x66)
+	x68 := (x4 + x67)
+	x69 := (x3 + x68)
+	x70 := (x2 + x69)
+	x71 := (x1 + x70)
+	out1[0] = x39
+	out1[1] = x47
+	out1[2] = x56
+	out1[3] = x64
+	out1[4] = x71
+}
+
+fe_relax :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	x4 := arg1[3]
+	x5 := arg1[4]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_carry_scmul_121666 :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(0x1db42, arg1[4])
+	x4, x3 := bits.mul_u64(0x1db42, arg1[3])
+	x6, x5 := bits.mul_u64(0x1db42, arg1[2])
+	x8, x7 := bits.mul_u64(0x1db42, arg1[1])
+	x10, x9 := bits.mul_u64(0x1db42, arg1[0])
+	x11 := ((x9 >> 51) | ((x10 << 13) & 0xffffffffffffffff))
+	x12 := (x9 & 0x7ffffffffffff)
+	x13, x14 := bits.add_u64(x11, x7, u64(0x0))
+	x15 := (u64(fiat.u1(x14)) + x8)
+	x16 := ((x13 >> 51) | ((x15 << 13) & 0xffffffffffffffff))
+	x17 := (x13 & 0x7ffffffffffff)
+	x18, x19 := bits.add_u64(x16, x5, u64(0x0))
+	x20 := (u64(fiat.u1(x19)) + x6)
+	x21 := ((x18 >> 51) | ((x20 << 13) & 0xffffffffffffffff))
+	x22 := (x18 & 0x7ffffffffffff)
+	x23, x24 := bits.add_u64(x21, x3, u64(0x0))
+	x25 := (u64(fiat.u1(x24)) + x4)
+	x26 := ((x23 >> 51) | ((x25 << 13) & 0xffffffffffffffff))
+	x27 := (x23 & 0x7ffffffffffff)
+	x28, x29 := bits.add_u64(x26, x1, u64(0x0))
+	x30 := (u64(fiat.u1(x29)) + x2)
+	x31 := ((x28 >> 51) | ((x30 << 13) & 0xffffffffffffffff))
+	x32 := (x28 & 0x7ffffffffffff)
+	x33 := (x31 * 0x13)
+	x34 := (x12 + x33)
+	x35 := fiat.u1((x34 >> 51))
+	x36 := (x34 & 0x7ffffffffffff)
+	x37 := (u64(x35) + x17)
+	x38 := fiat.u1((x37 >> 51))
+	x39 := (x37 & 0x7ffffffffffff)
+	x40 := (u64(x38) + x22)
+	out1[0] = x36
+	out1[1] = x39
+	out1[2] = x40
+	out1[3] = x27
+	out1[4] = x32
+}
+
+// The following routines were added by hand, and do not come from fiat-crypto.
+
+fe_zero :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 0
+	out1[1] = 0
+	out1[2] = 0
+	out1[3] = 0
+	out1[4] = 0
+}
+
+fe_one :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 1
+	out1[1] = 0
+	out1[2] = 0
+	out1[3] = 0
+	out1[4] = 0
+}
+
+fe_set :: proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	x4 := arg1[3]
+	x5 := arg1[4]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_cond_swap :: proc "contextless" (out1, out2: ^Tight_Field_Element, arg1: int) {
+	mask := -u64(arg1)
+	x := (out1[0] ~ out2[0]) & mask
+	x1, y1 := out1[0] ~ x, out2[0] ~ x
+	x = (out1[1] ~ out2[1]) & mask
+	x2, y2 := out1[1] ~ x, out2[1] ~ x
+	x = (out1[2] ~ out2[2]) & mask
+	x3, y3 := out1[2] ~ x, out2[2] ~ x
+	x = (out1[3] ~ out2[3]) & mask
+	x4, y4 := out1[3] ~ x, out2[3] ~ x
+	x = (out1[4] ~ out2[4]) & mask
+	x5, y5 := out1[4] ~ x, out2[4] ~ x
+	out1[0], out2[0] = x1, y1
+	out1[1], out2[1] = x2, y2
+	out1[2], out2[2] = x3, y3
+	out1[3], out2[3] = x4, y4
+	out1[4], out2[4] = x5, y5
+}

core/crypto/_fiat/field_poly1305/field.odin

@@ -0,0 +1,66 @@
+package field_poly1305
+
+import "core:crypto/util"
+import "core:mem"
+
+fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
+	return transmute(^Loose_Field_Element)(arg1)
+}
+
+fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_Field_Element) -> ^Tight_Field_Element {
+	return transmute(^Tight_Field_Element)(arg1)
+}
+
+fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, arg2: byte, sanitize: bool = true) {
+	// fiat-crypto's deserialization routine effectively processes a
+	// single byte at a time, and wants 256-bits of input for a value
+	// that will be 128-bits or 129-bits.
+	//
+	// This is somewhat cumbersome to use, so at a minimum a wrapper
+	// makes implementing the actual MAC block processing considerably
+	// neater.
+
+	assert(len(arg1) == 16)
+
+	when ODIN_ARCH == "386" || ODIN_ARCH == "amd64" {
+		// While it may be unwise to do deserialization here on our
+		// own when fiat-crypto provides equivalent functionality,
+		// doing it this way provides a little under 3x performance
+		// improvement when optimization is enabled.
+		src_p := transmute(^[2]u64)(&arg1[0])
+		lo := src_p[0]
+		hi := src_p[1]
+
+		// This is inspired by poly1305-donna, though adjustments were
+		// made since a Tight_Field_Element's limbs are 44-bits, 43-bits,
+		// and 43-bits wide.
+		//
+		// Note: This could be transplated into fe_from_u64s, but that
+		// code is called once per MAC, and is non-criticial path.
+		hibit := u64(arg2) << 41 // arg2 << 128
+		out1[0] = lo & 0xfffffffffff
+		out1[1] = ((lo >> 44) | (hi << 20)) & 0x7ffffffffff
+		out1[2] = ((hi >> 23) & 0x7ffffffffff) | hibit
+	} else {
+		tmp: [32]byte
+		copy_slice(tmp[0:16], arg1[:])
+		tmp[16] = arg2
+
+		_fe_from_bytes(out1, &tmp)
+		if sanitize {
+			// This is used to deserialize `s` which is confidential.
+			mem.zero_explicit(&tmp, size_of(tmp))
+		}
+	}
+}
+
+fe_from_u64s :: proc "contextless" (out1: ^Tight_Field_Element, lo, hi: u64) {
+	tmp: [32]byte
+	util.PUT_U64_LE(tmp[0:8], lo)
+	util.PUT_U64_LE(tmp[8:16], hi)
+
+	_fe_from_bytes(out1, &tmp)
+
+	// This routine is only used to deserialize `r` which is confidential.
+	mem.zero_explicit(&tmp, size_of(tmp))
+}

core/crypto/_fiat/field_poly1305/field4344.odin

@@ -0,0 +1,356 @@
+// The BSD 1-Clause License (BSD-1-Clause)
+//
+// Copyright (c) 2015-2020 the fiat-crypto authors (see the AUTHORS file)
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     1. Redistributions of source code must retain the above copyright
+//        notice, this list of conditions and the following disclaimer.
+//
+// THIS SOFTWARE IS PROVIDED BY the fiat-crypto authors "AS IS"
+// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL Berkeley Software Design,
+// Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+package field_poly1305
+
+// This file provides arithmetic on the field Z/(2^130 - 5) using
+// unsaturated 64-bit integer arithmetic.  It is derived primarily
+// from the machine generate Golang output from the fiat-crypto project.
+//
+// While the base implementation is provably correct, this implementation
+// makes no such claims as the port and optimizations were done by hand.
+// At some point, it may be worth adding support to fiat-crypto for
+// generating Odin output.
+
+import fiat "core:crypto/_fiat"
+import "core:math/bits"
+
+Loose_Field_Element :: distinct [3]u64
+Tight_Field_Element :: distinct [3]u64
+
+_addcarryx_u44 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0xfffffffffff)
+	x3 := fiat.u1((x1 >> 44))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u44 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 44))
+	x3 := (u64(x1) & 0xfffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+_addcarryx_u43 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0x7ffffffffff)
+	x3 := fiat.u1((x1 >> 43))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u43 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 43))
+	x3 := (u64(x1) & 0x7ffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+fe_carry_mul :: proc (out1: ^Tight_Field_Element, arg1, arg2: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(arg1[2], (arg2[2] * 0x5))
+	x4, x3 := bits.mul_u64(arg1[2], (arg2[1] * 0xa))
+	x6, x5 := bits.mul_u64(arg1[1], (arg2[2] * 0xa))
+	x8, x7 := bits.mul_u64(arg1[2], arg2[0])
+	x10, x9 := bits.mul_u64(arg1[1], (arg2[1] * 0x2))
+	x12, x11 := bits.mul_u64(arg1[1], arg2[0])
+	x14, x13 := bits.mul_u64(arg1[0], arg2[2])
+	x16, x15 := bits.mul_u64(arg1[0], arg2[1])
+	x18, x17 := bits.mul_u64(arg1[0], arg2[0])
+	x19, x20 := bits.add_u64(x5, x3, u64(0x0))
+	x21, _ := bits.add_u64(x6, x4, u64(fiat.u1(x20)))
+	x23, x24 := bits.add_u64(x17, x19, u64(0x0))
+	x25, _ := bits.add_u64(x18, x21, u64(fiat.u1(x24)))
+	x27 := ((x23 >> 44) | ((x25 << 20) & 0xffffffffffffffff))
+	x28 := (x23 & 0xfffffffffff)
+	x29, x30 := bits.add_u64(x9, x7, u64(0x0))
+	x31, _ := bits.add_u64(x10, x8, u64(fiat.u1(x30)))
+	x33, x34 := bits.add_u64(x13, x29, u64(0x0))
+	x35, _ := bits.add_u64(x14, x31, u64(fiat.u1(x34)))
+	x37, x38 := bits.add_u64(x11, x1, u64(0x0))
+	x39, _ := bits.add_u64(x12, x2, u64(fiat.u1(x38)))
+	x41, x42 := bits.add_u64(x15, x37, u64(0x0))
+	x43, _ := bits.add_u64(x16, x39, u64(fiat.u1(x42)))
+	x45, x46 := bits.add_u64(x27, x41, u64(0x0))
+	x47 := (u64(fiat.u1(x46)) + x43)
+	x48 := ((x45 >> 43) | ((x47 << 21) & 0xffffffffffffffff))
+	x49 := (x45 & 0x7ffffffffff)
+	x50, x51 := bits.add_u64(x48, x33, u64(0x0))
+	x52 := (u64(fiat.u1(x51)) + x35)
+	x53 := ((x50 >> 43) | ((x52 << 21) & 0xffffffffffffffff))
+	x54 := (x50 & 0x7ffffffffff)
+	x55 := (x53 * 0x5)
+	x56 := (x28 + x55)
+	x57 := (x56 >> 44)
+	x58 := (x56 & 0xfffffffffff)
+	x59 := (x57 + x49)
+	x60 := fiat.u1((x59 >> 43))
+	x61 := (x59 & 0x7ffffffffff)
+	x62 := (u64(x60) + x54)
+	out1[0] = x58
+	out1[1] = x61
+	out1[2] = x62
+}
+
+fe_carry_square :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := (arg1[2] * 0x5)
+	x2 := (x1 * 0x2)
+	x3 := (arg1[2] * 0x2)
+	x4 := (arg1[1] * 0x2)
+	x6, x5 := bits.mul_u64(arg1[2], x1)
+	x8, x7 := bits.mul_u64(arg1[1], (x2 * 0x2))
+	x10, x9 := bits.mul_u64(arg1[1], (arg1[1] * 0x2))
+	x12, x11 := bits.mul_u64(arg1[0], x3)
+	x14, x13 := bits.mul_u64(arg1[0], x4)
+	x16, x15 := bits.mul_u64(arg1[0], arg1[0])
+	x17, x18 := bits.add_u64(x15, x7, u64(0x0))
+	x19, _ := bits.add_u64(x16, x8, u64(fiat.u1(x18)))
+	x21 := ((x17 >> 44) | ((x19 << 20) & 0xffffffffffffffff))
+	x22 := (x17 & 0xfffffffffff)
+	x23, x24 := bits.add_u64(x11, x9, u64(0x0))
+	x25, _ := bits.add_u64(x12, x10, u64(fiat.u1(x24)))
+	x27, x28 := bits.add_u64(x13, x5, u64(0x0))
+	x29, _ := bits.add_u64(x14, x6, u64(fiat.u1(x28)))
+	x31, x32 := bits.add_u64(x21, x27, u64(0x0))
+	x33 := (u64(fiat.u1(x32)) + x29)
+	x34 := ((x31 >> 43) | ((x33 << 21) & 0xffffffffffffffff))
+	x35 := (x31 & 0x7ffffffffff)
+	x36, x37 := bits.add_u64(x34, x23, u64(0x0))
+	x38 := (u64(fiat.u1(x37)) + x25)
+	x39 := ((x36 >> 43) | ((x38 << 21) & 0xffffffffffffffff))
+	x40 := (x36 & 0x7ffffffffff)
+	x41 := (x39 * 0x5)
+	x42 := (x22 + x41)
+	x43 := (x42 >> 44)
+	x44 := (x42 & 0xfffffffffff)
+	x45 := (x43 + x35)
+	x46 := fiat.u1((x45 >> 43))
+	x47 := (x45 & 0x7ffffffffff)
+	x48 := (u64(x46) + x40)
+	out1[0] = x44
+	out1[1] = x47
+	out1[2] = x48
+}
+
+fe_carry :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := arg1[0]
+	x2 := ((x1 >> 44) + arg1[1])
+	x3 := ((x2 >> 43) + arg1[2])
+	x4 := ((x1 & 0xfffffffffff) + ((x3 >> 43) * 0x5))
+	x5 := (u64(fiat.u1((x4 >> 44))) + (x2 & 0x7ffffffffff))
+	x6 := (x4 & 0xfffffffffff)
+	x7 := (x5 & 0x7ffffffffff)
+	x8 := (u64(fiat.u1((x5 >> 43))) + (x3 & 0x7ffffffffff))
+	out1[0] = x6
+	out1[1] = x7
+	out1[2] = x8
+}
+
+fe_add :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := (arg1[0] + arg2[0])
+	x2 := (arg1[1] + arg2[1])
+	x3 := (arg1[2] + arg2[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_sub :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := ((0x1ffffffffff6 + arg1[0]) - arg2[0])
+	x2 := ((0xffffffffffe + arg1[1]) - arg2[1])
+	x3 := ((0xffffffffffe + arg1[2]) - arg2[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_opp :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := (0x1ffffffffff6 - arg1[0])
+	x2 := (0xffffffffffe - arg1[1])
+	x3 := (0xffffffffffe - arg1[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_cond_assign :: proc "contextless" (out1, arg1: ^Tight_Field_Element, arg2: bool) {
+	x1 := fiat.cmovznz_u64(fiat.u1(arg2), out1[0], arg1[0])
+	x2 := fiat.cmovznz_u64(fiat.u1(arg2), out1[1], arg1[1])
+	x3 := fiat.cmovznz_u64(fiat.u1(arg2), out1[2], arg1[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_to_bytes :: proc "contextless" (out1: ^[32]byte, arg1: ^Tight_Field_Element) {
+	x1, x2 := _subborrowx_u44(0x0, arg1[0], 0xffffffffffb)
+	x3, x4 := _subborrowx_u43(x2, arg1[1], 0x7ffffffffff)
+	x5, x6 := _subborrowx_u43(x4, arg1[2], 0x7ffffffffff)
+	x7 := fiat.cmovznz_u64(x6, u64(0x0), 0xffffffffffffffff)
+	x8, x9 := _addcarryx_u44(0x0, x1, (x7 & 0xffffffffffb))
+	x10, x11 := _addcarryx_u43(x9, x3, (x7 & 0x7ffffffffff))
+	x12, _ := _addcarryx_u43(x11, x5, (x7 & 0x7ffffffffff))
+	x14 := (x12 << 7)
+	x15 := (x10 << 4)
+	x16 := (u8(x8) & 0xff)
+	x17 := (x8 >> 8)
+	x18 := (u8(x17) & 0xff)
+	x19 := (x17 >> 8)
+	x20 := (u8(x19) & 0xff)
+	x21 := (x19 >> 8)
+	x22 := (u8(x21) & 0xff)
+	x23 := (x21 >> 8)
+	x24 := (u8(x23) & 0xff)
+	x25 := u8((x23 >> 8))
+	x26 := (x15 + u64(x25))
+	x27 := (u8(x26) & 0xff)
+	x28 := (x26 >> 8)
+	x29 := (u8(x28) & 0xff)
+	x30 := (x28 >> 8)
+	x31 := (u8(x30) & 0xff)
+	x32 := (x30 >> 8)
+	x33 := (u8(x32) & 0xff)
+	x34 := (x32 >> 8)
+	x35 := (u8(x34) & 0xff)
+	x36 := u8((x34 >> 8))
+	x37 := (x14 + u64(x36))
+	x38 := (u8(x37) & 0xff)
+	x39 := (x37 >> 8)
+	x40 := (u8(x39) & 0xff)
+	x41 := (x39 >> 8)
+	x42 := (u8(x41) & 0xff)
+	x43 := (x41 >> 8)
+	x44 := (u8(x43) & 0xff)
+	x45 := (x43 >> 8)
+	x46 := (u8(x45) & 0xff)
+	x47 := (x45 >> 8)
+	x48 := (u8(x47) & 0xff)
+	x49 := u8((x47 >> 8))
+	out1[0] = x16
+	out1[1] = x18
+	out1[2] = x20
+	out1[3] = x22
+	out1[4] = x24
+	out1[5] = x27
+	out1[6] = x29
+	out1[7] = x31
+	out1[8] = x33
+	out1[9] = x35
+	out1[10] = x38
+	out1[11] = x40
+	out1[12] = x42
+	out1[13] = x44
+	out1[14] = x46
+	out1[15] = x48
+	out1[16] = x49
+}
+
+_fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	x1 := (u64(arg1[16]) << 41)
+	x2 := (u64(arg1[15]) << 33)
+	x3 := (u64(arg1[14]) << 25)
+	x4 := (u64(arg1[13]) << 17)
+	x5 := (u64(arg1[12]) << 9)
+	x6 := (u64(arg1[11]) * u64(0x2))
+	x7 := (u64(arg1[10]) << 36)
+	x8 := (u64(arg1[9]) << 28)
+	x9 := (u64(arg1[8]) << 20)
+	x10 := (u64(arg1[7]) << 12)
+	x11 := (u64(arg1[6]) << 4)
+	x12 := (u64(arg1[5]) << 40)
+	x13 := (u64(arg1[4]) << 32)
+	x14 := (u64(arg1[3]) << 24)
+	x15 := (u64(arg1[2]) << 16)
+	x16 := (u64(arg1[1]) << 8)
+	x17 := arg1[0]
+	x18 := (x16 + u64(x17))
+	x19 := (x15 + x18)
+	x20 := (x14 + x19)
+	x21 := (x13 + x20)
+	x22 := (x12 + x21)
+	x23 := (x22 & 0xfffffffffff)
+	x24 := u8((x22 >> 44))
+	x25 := (x11 + u64(x24))
+	x26 := (x10 + x25)
+	x27 := (x9 + x26)
+	x28 := (x8 + x27)
+	x29 := (x7 + x28)
+	x30 := (x29 & 0x7ffffffffff)
+	x31 := fiat.u1((x29 >> 43))
+	x32 := (x6 + u64(x31))
+	x33 := (x5 + x32)
+	x34 := (x4 + x33)
+	x35 := (x3 + x34)
+	x36 := (x2 + x35)
+	x37 := (x1 + x36)
+	out1[0] = x23
+	out1[1] = x30
+	out1[2] = x37
+}
+
+fe_relax :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+// The following routines were added by hand, and do not come from fiat-crypto.
+
+fe_zero :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 0
+	out1[1] = 0
+	out1[2] = 0
+}
+
+fe_set :: #force_inline proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_cond_swap :: proc "contextless" (out1, out2: ^Tight_Field_Element, arg1: bool) {
+	mask := -u64(arg1)
+	x := (out1[0] ~ out2[0]) & mask
+	x1, y1 := out1[0] ~ x, out2[0] ~ x
+	x = (out1[1] ~ out2[1]) & mask
+	x2, y2 := out1[1] ~ x, out2[1] ~ x
+	x = (out1[2] ~ out2[2]) & mask
+	x3, y3 := out1[2] ~ x, out2[2] ~ x
+	out1[0], out2[0] = x1, y1
+	out1[1], out2[1] = x2, y2
+	out1[2], out2[2] = x3, y3
+}

core/crypto/_sha3/_sha3.odin

@@ -0,0 +1,169 @@
+package _sha3
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Keccak hashing algorithm, standardized as SHA3 in <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf>
+    To use the original Keccak padding, set the is_keccak bool to true, otherwise it will use SHA3 padding.
+*/
+
+import "../util"
+
+ROUNDS :: 24
+
+Sha3_Context :: struct {
+    st: struct #raw_union {
+        b: [200]u8,
+        q: [25]u64,
+    },
+    pt:        int,
+    rsiz:      int,
+    mdlen:     int,
+    is_keccak: bool,
+}
+
+keccakf :: proc "contextless" (st: ^[25]u64) {
+    keccakf_rndc := [?]u64 {
+        0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
+        0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
+        0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
+        0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
+        0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
+        0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
+        0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
+        0x8000000000008080, 0x0000000080000001, 0x8000000080008008,
+    }
+
+    keccakf_rotc := [?]i32 {
+        1,  3,  6,  10, 15, 21, 28, 36, 45, 55, 2,  14,
+        27, 41, 56, 8,  25, 43, 62, 18, 39, 61, 20, 44,
+    }
+
+    keccakf_piln := [?]i32 {
+        10, 7,  11, 17, 18, 3, 5,  16, 8,  21, 24, 4,
+        15, 23, 19, 13, 12, 2, 20, 14, 22, 9,  6,  1,
+    }
+
+    i, j, r: i32 = ---, ---, ---
+    t: u64       = ---
+    bc: [5]u64   = ---
+
+    when ODIN_ENDIAN != "little" {
+        v: uintptr = ---
+        for i = 0; i < 25; i += 1 {
+            v := uintptr(&st[i])
+            st[i] = u64((^u8)(v + 0)^ << 0)  | u64((^u8)(v + 1)^ << 8)  |
+                    u64((^u8)(v + 2)^ << 16) | u64((^u8)(v + 3)^ << 24) |
+                    u64((^u8)(v + 4)^ << 32) | u64((^u8)(v + 5)^ << 40) |
+                    u64((^u8)(v + 6)^ << 48) | u64((^u8)(v + 7)^ << 56)
+        }
+    }
+
+    for r = 0; r < ROUNDS; r += 1 {
+        // theta
+        for i = 0; i < 5; i += 1 {
+            bc[i] = st[i] ~ st[i + 5] ~ st[i + 10] ~ st[i + 15] ~ st[i + 20]
+        }
+
+        for i = 0; i < 5; i += 1 {
+            t = bc[(i + 4) % 5] ~ util.ROTL64(bc[(i + 1) % 5], 1)
+            for j = 0; j < 25; j += 5 {
+                st[j + i] ~= t
+            }
+        }
+
+        // rho pi
+        t = st[1]
+        for i = 0; i < 24; i += 1 {
+            j = keccakf_piln[i]
+            bc[0] = st[j]
+            st[j] = util.ROTL64(t, u64(keccakf_rotc[i]))
+            t = bc[0]
+        }
+
+        // chi
+        for j = 0; j < 25; j += 5 {
+            for i = 0; i < 5; i += 1 {
+                bc[i] = st[j + i]
+            }
+            for i = 0; i < 5; i += 1 {
+                st[j + i] ~= ~bc[(i + 1) % 5] & bc[(i + 2) % 5]
+            }
+        }
+
+        st[0] ~= keccakf_rndc[r]
+    }
+
+    when ODIN_ENDIAN != "little" {
+        for i = 0; i < 25; i += 1 {
+            v = uintptr(&st[i])
+            t = st[i]
+            (^u8)(v + 0)^ = (t >> 0)  & 0xff
+            (^u8)(v + 1)^ = (t >> 8)  & 0xff
+            (^u8)(v + 2)^ = (t >> 16) & 0xff
+            (^u8)(v + 3)^ = (t >> 24) & 0xff
+            (^u8)(v + 4)^ = (t >> 32) & 0xff
+            (^u8)(v + 5)^ = (t >> 40) & 0xff
+            (^u8)(v + 6)^ = (t >> 48) & 0xff
+            (^u8)(v + 7)^ = (t >> 56) & 0xff
+        }
+    }
+}
+
+init :: proc "contextless" (c: ^Sha3_Context) {
+    for i := 0; i < 25; i += 1 {
+        c.st.q[i] = 0
+    }
+    c.rsiz = 200 - 2 * c.mdlen
+}
+
+update :: proc "contextless" (c: ^Sha3_Context, data: []byte) {
+    j := c.pt
+    for i := 0; i < len(data); i += 1 {
+        c.st.b[j] ~= data[i]
+        j += 1
+        if j >= c.rsiz {
+            keccakf(&c.st.q)
+            j = 0
+        }
+    }
+    c.pt = j
+}
+
+final :: proc "contextless" (c: ^Sha3_Context, hash: []byte) {
+    if c.is_keccak {
+        c.st.b[c.pt] ~= 0x01
+    } else {
+        c.st.b[c.pt] ~= 0x06
+    }
+    
+    c.st.b[c.rsiz - 1] ~= 0x80
+    keccakf(&c.st.q)
+    for i := 0; i < c.mdlen; i += 1 {
+        hash[i] = c.st.b[i]
+    }
+}
+
+shake_xof :: proc "contextless" (c: ^Sha3_Context) {
+    c.st.b[c.pt]       ~= 0x1F
+    c.st.b[c.rsiz - 1] ~= 0x80
+    keccakf(&c.st.q)
+    c.pt = 0
+}
+
+shake_out :: proc "contextless" (c: ^Sha3_Context, hash: []byte) {
+    j := c.pt
+    for i := 0; i < len(hash); i += 1 {
+        if j >= c.rsiz {
+            keccakf(&c.st.q)
+            j = 0
+        }
+        hash[i] = c.st.b[j]
+        j += 1
+    }
+    c.pt = j
+}

core/crypto/_tiger/_tiger.odin

@@ -0,0 +1,410 @@
+package _tiger
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Tiger hashing algorithm, as defined in <https://www.cs.technion.ac.il/~biham/Reports/Tiger/>
+*/
+
+import "../util"
+
+T1 := [?]u64 {
+	0x02aab17cf7e90c5e, 0xac424b03e243a8ec, 0x72cd5be30dd5fcd3, 0x6d019b93f6f97f3a,
+	0xcd9978ffd21f9193, 0x7573a1c9708029e2, 0xb164326b922a83c3, 0x46883eee04915870,
+	0xeaace3057103ece6, 0xc54169b808a3535c, 0x4ce754918ddec47c, 0x0aa2f4dfdc0df40c,
+	0x10b76f18a74dbefa, 0xc6ccb6235ad1ab6a, 0x13726121572fe2ff, 0x1a488c6f199d921e,
+	0x4bc9f9f4da0007ca, 0x26f5e6f6e85241c7, 0x859079dbea5947b6, 0x4f1885c5c99e8c92,
+	0xd78e761ea96f864b, 0x8e36428c52b5c17d, 0x69cf6827373063c1, 0xb607c93d9bb4c56e,
+	0x7d820e760e76b5ea, 0x645c9cc6f07fdc42, 0xbf38a078243342e0, 0x5f6b343c9d2e7d04,
+	0xf2c28aeb600b0ec6, 0x6c0ed85f7254bcac, 0x71592281a4db4fe5, 0x1967fa69ce0fed9f,
+	0xfd5293f8b96545db, 0xc879e9d7f2a7600b, 0x860248920193194e, 0xa4f9533b2d9cc0b3,
+	0x9053836c15957613, 0xdb6dcf8afc357bf1, 0x18beea7a7a370f57, 0x037117ca50b99066,
+	0x6ab30a9774424a35, 0xf4e92f02e325249b, 0x7739db07061ccae1, 0xd8f3b49ceca42a05,
+	0xbd56be3f51382f73, 0x45faed5843b0bb28, 0x1c813d5c11bf1f83, 0x8af0e4b6d75fa169,
+	0x33ee18a487ad9999, 0x3c26e8eab1c94410, 0xb510102bc0a822f9, 0x141eef310ce6123b,
+	0xfc65b90059ddb154, 0xe0158640c5e0e607, 0x884e079826c3a3cf, 0x930d0d9523c535fd,
+	0x35638d754e9a2b00, 0x4085fccf40469dd5, 0xc4b17ad28be23a4c, 0xcab2f0fc6a3e6a2e,
+	0x2860971a6b943fcd, 0x3dde6ee212e30446, 0x6222f32ae01765ae, 0x5d550bb5478308fe,
+	0xa9efa98da0eda22a, 0xc351a71686c40da7, 0x1105586d9c867c84, 0xdcffee85fda22853,
+	0xccfbd0262c5eef76, 0xbaf294cb8990d201, 0xe69464f52afad975, 0x94b013afdf133e14,
+	0x06a7d1a32823c958, 0x6f95fe5130f61119, 0xd92ab34e462c06c0, 0xed7bde33887c71d2,
+	0x79746d6e6518393e, 0x5ba419385d713329, 0x7c1ba6b948a97564, 0x31987c197bfdac67,
+	0xde6c23c44b053d02, 0x581c49fed002d64d, 0xdd474d6338261571, 0xaa4546c3e473d062,
+	0x928fce349455f860, 0x48161bbacaab94d9, 0x63912430770e6f68, 0x6ec8a5e602c6641c,
+	0x87282515337ddd2b, 0x2cda6b42034b701b, 0xb03d37c181cb096d, 0xe108438266c71c6f,
+	0x2b3180c7eb51b255, 0xdf92b82f96c08bbc, 0x5c68c8c0a632f3ba, 0x5504cc861c3d0556,
+	0xabbfa4e55fb26b8f, 0x41848b0ab3baceb4, 0xb334a273aa445d32, 0xbca696f0a85ad881,
+	0x24f6ec65b528d56c, 0x0ce1512e90f4524a, 0x4e9dd79d5506d35a, 0x258905fac6ce9779,
+	0x2019295b3e109b33, 0xf8a9478b73a054cc, 0x2924f2f934417eb0, 0x3993357d536d1bc4,
+	0x38a81ac21db6ff8b, 0x47c4fbf17d6016bf, 0x1e0faadd7667e3f5, 0x7abcff62938beb96,
+	0xa78dad948fc179c9, 0x8f1f98b72911e50d, 0x61e48eae27121a91, 0x4d62f7ad31859808,
+	0xeceba345ef5ceaeb, 0xf5ceb25ebc9684ce, 0xf633e20cb7f76221, 0xa32cdf06ab8293e4,
+	0x985a202ca5ee2ca4, 0xcf0b8447cc8a8fb1, 0x9f765244979859a3, 0xa8d516b1a1240017,
+	0x0bd7ba3ebb5dc726, 0xe54bca55b86adb39, 0x1d7a3afd6c478063, 0x519ec608e7669edd,
+	0x0e5715a2d149aa23, 0x177d4571848ff194, 0xeeb55f3241014c22, 0x0f5e5ca13a6e2ec2,
+	0x8029927b75f5c361, 0xad139fabc3d6e436, 0x0d5df1a94ccf402f, 0x3e8bd948bea5dfc8,
+	0xa5a0d357bd3ff77e, 0xa2d12e251f74f645, 0x66fd9e525e81a082, 0x2e0c90ce7f687a49,
+	0xc2e8bcbeba973bc5, 0x000001bce509745f, 0x423777bbe6dab3d6, 0xd1661c7eaef06eb5,
+	0xa1781f354daacfd8, 0x2d11284a2b16affc, 0xf1fc4f67fa891d1f, 0x73ecc25dcb920ada,
+	0xae610c22c2a12651, 0x96e0a810d356b78a, 0x5a9a381f2fe7870f, 0xd5ad62ede94e5530,
+	0xd225e5e8368d1427, 0x65977b70c7af4631, 0x99f889b2de39d74f, 0x233f30bf54e1d143,
+	0x9a9675d3d9a63c97, 0x5470554ff334f9a8, 0x166acb744a4f5688, 0x70c74caab2e4aead,
+	0xf0d091646f294d12, 0x57b82a89684031d1, 0xefd95a5a61be0b6b, 0x2fbd12e969f2f29a,
+	0x9bd37013feff9fe8, 0x3f9b0404d6085a06, 0x4940c1f3166cfe15, 0x09542c4dcdf3defb,
+	0xb4c5218385cd5ce3, 0xc935b7dc4462a641, 0x3417f8a68ed3b63f, 0xb80959295b215b40,
+	0xf99cdaef3b8c8572, 0x018c0614f8fcb95d, 0x1b14accd1a3acdf3, 0x84d471f200bb732d,
+	0xc1a3110e95e8da16, 0x430a7220bf1a82b8, 0xb77e090d39df210e, 0x5ef4bd9f3cd05e9d,
+	0x9d4ff6da7e57a444, 0xda1d60e183d4a5f8, 0xb287c38417998e47, 0xfe3edc121bb31886,
+	0xc7fe3ccc980ccbef, 0xe46fb590189bfd03, 0x3732fd469a4c57dc, 0x7ef700a07cf1ad65,
+	0x59c64468a31d8859, 0x762fb0b4d45b61f6, 0x155baed099047718, 0x68755e4c3d50baa6,
+	0xe9214e7f22d8b4df, 0x2addbf532eac95f4, 0x32ae3909b4bd0109, 0x834df537b08e3450,
+	0xfa209da84220728d, 0x9e691d9b9efe23f7, 0x0446d288c4ae8d7f, 0x7b4cc524e169785b,
+	0x21d87f0135ca1385, 0xcebb400f137b8aa5, 0x272e2b66580796be, 0x3612264125c2b0de,
+	0x057702bdad1efbb2, 0xd4babb8eacf84be9, 0x91583139641bc67b, 0x8bdc2de08036e024,
+	0x603c8156f49f68ed, 0xf7d236f7dbef5111, 0x9727c4598ad21e80, 0xa08a0896670a5fd7,
+	0xcb4a8f4309eba9cb, 0x81af564b0f7036a1, 0xc0b99aa778199abd, 0x959f1ec83fc8e952,
+	0x8c505077794a81b9, 0x3acaaf8f056338f0, 0x07b43f50627a6778, 0x4a44ab49f5eccc77,
+	0x3bc3d6e4b679ee98, 0x9cc0d4d1cf14108c, 0x4406c00b206bc8a0, 0x82a18854c8d72d89,
+	0x67e366b35c3c432c, 0xb923dd61102b37f2, 0x56ab2779d884271d, 0xbe83e1b0ff1525af,
+	0xfb7c65d4217e49a9, 0x6bdbe0e76d48e7d4, 0x08df828745d9179e, 0x22ea6a9add53bd34,
+	0xe36e141c5622200a, 0x7f805d1b8cb750ee, 0xafe5c7a59f58e837, 0xe27f996a4fb1c23c,
+	0xd3867dfb0775f0d0, 0xd0e673de6e88891a, 0x123aeb9eafb86c25, 0x30f1d5d5c145b895,
+	0xbb434a2dee7269e7, 0x78cb67ecf931fa38, 0xf33b0372323bbf9c, 0x52d66336fb279c74,
+	0x505f33ac0afb4eaa, 0xe8a5cd99a2cce187, 0x534974801e2d30bb, 0x8d2d5711d5876d90,
+	0x1f1a412891bc038e, 0xd6e2e71d82e56648, 0x74036c3a497732b7, 0x89b67ed96361f5ab,
+	0xffed95d8f1ea02a2, 0xe72b3bd61464d43d, 0xa6300f170bdc4820, 0xebc18760ed78a77a,
+}
+
+T2 := [?]u64 {
+	0xe6a6be5a05a12138, 0xb5a122a5b4f87c98, 0x563c6089140b6990, 0x4c46cb2e391f5dd5,
+	0xd932addbc9b79434, 0x08ea70e42015aff5, 0xd765a6673e478cf1, 0xc4fb757eab278d99,
+	0xdf11c6862d6e0692, 0xddeb84f10d7f3b16, 0x6f2ef604a665ea04, 0x4a8e0f0ff0e0dfb3,
+	0xa5edeef83dbcba51, 0xfc4f0a2a0ea4371e, 0xe83e1da85cb38429, 0xdc8ff882ba1b1ce2,
+	0xcd45505e8353e80d, 0x18d19a00d4db0717, 0x34a0cfeda5f38101, 0x0be77e518887caf2,
+	0x1e341438b3c45136, 0xe05797f49089ccf9, 0xffd23f9df2591d14, 0x543dda228595c5cd,
+	0x661f81fd99052a33, 0x8736e641db0f7b76, 0x15227725418e5307, 0xe25f7f46162eb2fa,
+	0x48a8b2126c13d9fe, 0xafdc541792e76eea, 0x03d912bfc6d1898f, 0x31b1aafa1b83f51b,
+	0xf1ac2796e42ab7d9, 0x40a3a7d7fcd2ebac, 0x1056136d0afbbcc5, 0x7889e1dd9a6d0c85,
+	0xd33525782a7974aa, 0xa7e25d09078ac09b, 0xbd4138b3eac6edd0, 0x920abfbe71eb9e70,
+	0xa2a5d0f54fc2625c, 0xc054e36b0b1290a3, 0xf6dd59ff62fe932b, 0x3537354511a8ac7d,
+	0xca845e9172fadcd4, 0x84f82b60329d20dc, 0x79c62ce1cd672f18, 0x8b09a2add124642c,
+	0xd0c1e96a19d9e726, 0x5a786a9b4ba9500c, 0x0e020336634c43f3, 0xc17b474aeb66d822,
+	0x6a731ae3ec9baac2, 0x8226667ae0840258, 0x67d4567691caeca5, 0x1d94155c4875adb5,
+	0x6d00fd985b813fdf, 0x51286efcb774cd06, 0x5e8834471fa744af, 0xf72ca0aee761ae2e,
+	0xbe40e4cdaee8e09a, 0xe9970bbb5118f665, 0x726e4beb33df1964, 0x703b000729199762,
+	0x4631d816f5ef30a7, 0xb880b5b51504a6be, 0x641793c37ed84b6c, 0x7b21ed77f6e97d96,
+	0x776306312ef96b73, 0xae528948e86ff3f4, 0x53dbd7f286a3f8f8, 0x16cadce74cfc1063,
+	0x005c19bdfa52c6dd, 0x68868f5d64d46ad3, 0x3a9d512ccf1e186a, 0x367e62c2385660ae,
+	0xe359e7ea77dcb1d7, 0x526c0773749abe6e, 0x735ae5f9d09f734b, 0x493fc7cc8a558ba8,
+	0xb0b9c1533041ab45, 0x321958ba470a59bd, 0x852db00b5f46c393, 0x91209b2bd336b0e5,
+	0x6e604f7d659ef19f, 0xb99a8ae2782ccb24, 0xccf52ab6c814c4c7, 0x4727d9afbe11727b,
+	0x7e950d0c0121b34d, 0x756f435670ad471f, 0xf5add442615a6849, 0x4e87e09980b9957a,
+	0x2acfa1df50aee355, 0xd898263afd2fd556, 0xc8f4924dd80c8fd6, 0xcf99ca3d754a173a,
+	0xfe477bacaf91bf3c, 0xed5371f6d690c12d, 0x831a5c285e687094, 0xc5d3c90a3708a0a4,
+	0x0f7f903717d06580, 0x19f9bb13b8fdf27f, 0xb1bd6f1b4d502843, 0x1c761ba38fff4012,
+	0x0d1530c4e2e21f3b, 0x8943ce69a7372c8a, 0xe5184e11feb5ce66, 0x618bdb80bd736621,
+	0x7d29bad68b574d0b, 0x81bb613e25e6fe5b, 0x071c9c10bc07913f, 0xc7beeb7909ac2d97,
+	0xc3e58d353bc5d757, 0xeb017892f38f61e8, 0xd4effb9c9b1cc21a, 0x99727d26f494f7ab,
+	0xa3e063a2956b3e03, 0x9d4a8b9a4aa09c30, 0x3f6ab7d500090fb4, 0x9cc0f2a057268ac0,
+	0x3dee9d2dedbf42d1, 0x330f49c87960a972, 0xc6b2720287421b41, 0x0ac59ec07c00369c,
+	0xef4eac49cb353425, 0xf450244eef0129d8, 0x8acc46e5caf4deb6, 0x2ffeab63989263f7,
+	0x8f7cb9fe5d7a4578, 0x5bd8f7644e634635, 0x427a7315bf2dc900, 0x17d0c4aa2125261c,
+	0x3992486c93518e50, 0xb4cbfee0a2d7d4c3, 0x7c75d6202c5ddd8d, 0xdbc295d8e35b6c61,
+	0x60b369d302032b19, 0xce42685fdce44132, 0x06f3ddb9ddf65610, 0x8ea4d21db5e148f0,
+	0x20b0fce62fcd496f, 0x2c1b912358b0ee31, 0xb28317b818f5a308, 0xa89c1e189ca6d2cf,
+	0x0c6b18576aaadbc8, 0xb65deaa91299fae3, 0xfb2b794b7f1027e7, 0x04e4317f443b5beb,
+	0x4b852d325939d0a6, 0xd5ae6beefb207ffc, 0x309682b281c7d374, 0xbae309a194c3b475,
+	0x8cc3f97b13b49f05, 0x98a9422ff8293967, 0x244b16b01076ff7c, 0xf8bf571c663d67ee,
+	0x1f0d6758eee30da1, 0xc9b611d97adeb9b7, 0xb7afd5887b6c57a2, 0x6290ae846b984fe1,
+	0x94df4cdeacc1a5fd, 0x058a5bd1c5483aff, 0x63166cc142ba3c37, 0x8db8526eb2f76f40,
+	0xe10880036f0d6d4e, 0x9e0523c9971d311d, 0x45ec2824cc7cd691, 0x575b8359e62382c9,
+	0xfa9e400dc4889995, 0xd1823ecb45721568, 0xdafd983b8206082f, 0xaa7d29082386a8cb,
+	0x269fcd4403b87588, 0x1b91f5f728bdd1e0, 0xe4669f39040201f6, 0x7a1d7c218cf04ade,
+	0x65623c29d79ce5ce, 0x2368449096c00bb1, 0xab9bf1879da503ba, 0xbc23ecb1a458058e,
+	0x9a58df01bb401ecc, 0xa070e868a85f143d, 0x4ff188307df2239e, 0x14d565b41a641183,
+	0xee13337452701602, 0x950e3dcf3f285e09, 0x59930254b9c80953, 0x3bf299408930da6d,
+	0xa955943f53691387, 0xa15edecaa9cb8784, 0x29142127352be9a0, 0x76f0371fff4e7afb,
+	0x0239f450274f2228, 0xbb073af01d5e868b, 0xbfc80571c10e96c1, 0xd267088568222e23,
+	0x9671a3d48e80b5b0, 0x55b5d38ae193bb81, 0x693ae2d0a18b04b8, 0x5c48b4ecadd5335f,
+	0xfd743b194916a1ca, 0x2577018134be98c4, 0xe77987e83c54a4ad, 0x28e11014da33e1b9,
+	0x270cc59e226aa213, 0x71495f756d1a5f60, 0x9be853fb60afef77, 0xadc786a7f7443dbf,
+	0x0904456173b29a82, 0x58bc7a66c232bd5e, 0xf306558c673ac8b2, 0x41f639c6b6c9772a,
+	0x216defe99fda35da, 0x11640cc71c7be615, 0x93c43694565c5527, 0xea038e6246777839,
+	0xf9abf3ce5a3e2469, 0x741e768d0fd312d2, 0x0144b883ced652c6, 0xc20b5a5ba33f8552,
+	0x1ae69633c3435a9d, 0x97a28ca4088cfdec, 0x8824a43c1e96f420, 0x37612fa66eeea746,
+	0x6b4cb165f9cf0e5a, 0x43aa1c06a0abfb4a, 0x7f4dc26ff162796b, 0x6cbacc8e54ed9b0f,
+	0xa6b7ffefd2bb253e, 0x2e25bc95b0a29d4f, 0x86d6a58bdef1388c, 0xded74ac576b6f054,
+	0x8030bdbc2b45805d, 0x3c81af70e94d9289, 0x3eff6dda9e3100db, 0xb38dc39fdfcc8847,
+	0x123885528d17b87e, 0xf2da0ed240b1b642, 0x44cefadcd54bf9a9, 0x1312200e433c7ee6,
+	0x9ffcc84f3a78c748, 0xf0cd1f72248576bb, 0xec6974053638cfe4, 0x2ba7b67c0cec4e4c,
+	0xac2f4df3e5ce32ed, 0xcb33d14326ea4c11, 0xa4e9044cc77e58bc, 0x5f513293d934fcef,
+	0x5dc9645506e55444, 0x50de418f317de40a, 0x388cb31a69dde259, 0x2db4a83455820a86,
+	0x9010a91e84711ae9, 0x4df7f0b7b1498371, 0xd62a2eabc0977179, 0x22fac097aa8d5c0e,
+}
+
+T3 := [?]u64 {
+	0xf49fcc2ff1daf39b, 0x487fd5c66ff29281, 0xe8a30667fcdca83f, 0x2c9b4be3d2fcce63,
+	0xda3ff74b93fbbbc2, 0x2fa165d2fe70ba66, 0xa103e279970e93d4, 0xbecdec77b0e45e71,
+	0xcfb41e723985e497, 0xb70aaa025ef75017, 0xd42309f03840b8e0, 0x8efc1ad035898579,
+	0x96c6920be2b2abc5, 0x66af4163375a9172, 0x2174abdcca7127fb, 0xb33ccea64a72ff41,
+	0xf04a4933083066a5, 0x8d970acdd7289af5, 0x8f96e8e031c8c25e, 0xf3fec02276875d47,
+	0xec7bf310056190dd, 0xf5adb0aebb0f1491, 0x9b50f8850fd58892, 0x4975488358b74de8,
+	0xa3354ff691531c61, 0x0702bbe481d2c6ee, 0x89fb24057deded98, 0xac3075138596e902,
+	0x1d2d3580172772ed, 0xeb738fc28e6bc30d, 0x5854ef8f63044326, 0x9e5c52325add3bbe,
+	0x90aa53cf325c4623, 0xc1d24d51349dd067, 0x2051cfeea69ea624, 0x13220f0a862e7e4f,
+	0xce39399404e04864, 0xd9c42ca47086fcb7, 0x685ad2238a03e7cc, 0x066484b2ab2ff1db,
+	0xfe9d5d70efbf79ec, 0x5b13b9dd9c481854, 0x15f0d475ed1509ad, 0x0bebcd060ec79851,
+	0xd58c6791183ab7f8, 0xd1187c5052f3eee4, 0xc95d1192e54e82ff, 0x86eea14cb9ac6ca2,
+	0x3485beb153677d5d, 0xdd191d781f8c492a, 0xf60866baa784ebf9, 0x518f643ba2d08c74,
+	0x8852e956e1087c22, 0xa768cb8dc410ae8d, 0x38047726bfec8e1a, 0xa67738b4cd3b45aa,
+	0xad16691cec0dde19, 0xc6d4319380462e07, 0xc5a5876d0ba61938, 0x16b9fa1fa58fd840,
+	0x188ab1173ca74f18, 0xabda2f98c99c021f, 0x3e0580ab134ae816, 0x5f3b05b773645abb,
+	0x2501a2be5575f2f6, 0x1b2f74004e7e8ba9, 0x1cd7580371e8d953, 0x7f6ed89562764e30,
+	0xb15926ff596f003d, 0x9f65293da8c5d6b9, 0x6ecef04dd690f84c, 0x4782275fff33af88,
+	0xe41433083f820801, 0xfd0dfe409a1af9b5, 0x4325a3342cdb396b, 0x8ae77e62b301b252,
+	0xc36f9e9f6655615a, 0x85455a2d92d32c09, 0xf2c7dea949477485, 0x63cfb4c133a39eba,
+	0x83b040cc6ebc5462, 0x3b9454c8fdb326b0, 0x56f56a9e87ffd78c, 0x2dc2940d99f42bc6,
+	0x98f7df096b096e2d, 0x19a6e01e3ad852bf, 0x42a99ccbdbd4b40b, 0xa59998af45e9c559,
+	0x366295e807d93186, 0x6b48181bfaa1f773, 0x1fec57e2157a0a1d, 0x4667446af6201ad5,
+	0xe615ebcacfb0f075, 0xb8f31f4f68290778, 0x22713ed6ce22d11e, 0x3057c1a72ec3c93b,
+	0xcb46acc37c3f1f2f, 0xdbb893fd02aaf50e, 0x331fd92e600b9fcf, 0xa498f96148ea3ad6,
+	0xa8d8426e8b6a83ea, 0xa089b274b7735cdc, 0x87f6b3731e524a11, 0x118808e5cbc96749,
+	0x9906e4c7b19bd394, 0xafed7f7e9b24a20c, 0x6509eadeeb3644a7, 0x6c1ef1d3e8ef0ede,
+	0xb9c97d43e9798fb4, 0xa2f2d784740c28a3, 0x7b8496476197566f, 0x7a5be3e6b65f069d,
+	0xf96330ed78be6f10, 0xeee60de77a076a15, 0x2b4bee4aa08b9bd0, 0x6a56a63ec7b8894e,
+	0x02121359ba34fef4, 0x4cbf99f8283703fc, 0x398071350caf30c8, 0xd0a77a89f017687a,
+	0xf1c1a9eb9e423569, 0x8c7976282dee8199, 0x5d1737a5dd1f7abd, 0x4f53433c09a9fa80,
+	0xfa8b0c53df7ca1d9, 0x3fd9dcbc886ccb77, 0xc040917ca91b4720, 0x7dd00142f9d1dcdf,
+	0x8476fc1d4f387b58, 0x23f8e7c5f3316503, 0x032a2244e7e37339, 0x5c87a5d750f5a74b,
+	0x082b4cc43698992e, 0xdf917becb858f63c, 0x3270b8fc5bf86dda, 0x10ae72bb29b5dd76,
+	0x576ac94e7700362b, 0x1ad112dac61efb8f, 0x691bc30ec5faa427, 0xff246311cc327143,
+	0x3142368e30e53206, 0x71380e31e02ca396, 0x958d5c960aad76f1, 0xf8d6f430c16da536,
+	0xc8ffd13f1be7e1d2, 0x7578ae66004ddbe1, 0x05833f01067be646, 0xbb34b5ad3bfe586d,
+	0x095f34c9a12b97f0, 0x247ab64525d60ca8, 0xdcdbc6f3017477d1, 0x4a2e14d4decad24d,
+	0xbdb5e6d9be0a1eeb, 0x2a7e70f7794301ab, 0xdef42d8a270540fd, 0x01078ec0a34c22c1,
+	0xe5de511af4c16387, 0x7ebb3a52bd9a330a, 0x77697857aa7d6435, 0x004e831603ae4c32,
+	0xe7a21020ad78e312, 0x9d41a70c6ab420f2, 0x28e06c18ea1141e6, 0xd2b28cbd984f6b28,
+	0x26b75f6c446e9d83, 0xba47568c4d418d7f, 0xd80badbfe6183d8e, 0x0e206d7f5f166044,
+	0xe258a43911cbca3e, 0x723a1746b21dc0bc, 0xc7caa854f5d7cdd3, 0x7cac32883d261d9c,
+	0x7690c26423ba942c, 0x17e55524478042b8, 0xe0be477656a2389f, 0x4d289b5e67ab2da0,
+	0x44862b9c8fbbfd31, 0xb47cc8049d141365, 0x822c1b362b91c793, 0x4eb14655fb13dfd8,
+	0x1ecbba0714e2a97b, 0x6143459d5cde5f14, 0x53a8fbf1d5f0ac89, 0x97ea04d81c5e5b00,
+	0x622181a8d4fdb3f3, 0xe9bcd341572a1208, 0x1411258643cce58a, 0x9144c5fea4c6e0a4,
+	0x0d33d06565cf620f, 0x54a48d489f219ca1, 0xc43e5eac6d63c821, 0xa9728b3a72770daf,
+	0xd7934e7b20df87ef, 0xe35503b61a3e86e5, 0xcae321fbc819d504, 0x129a50b3ac60bfa6,
+	0xcd5e68ea7e9fb6c3, 0xb01c90199483b1c7, 0x3de93cd5c295376c, 0xaed52edf2ab9ad13,
+	0x2e60f512c0a07884, 0xbc3d86a3e36210c9, 0x35269d9b163951ce, 0x0c7d6e2ad0cdb5fa,
+	0x59e86297d87f5733, 0x298ef221898db0e7, 0x55000029d1a5aa7e, 0x8bc08ae1b5061b45,
+	0xc2c31c2b6c92703a, 0x94cc596baf25ef42, 0x0a1d73db22540456, 0x04b6a0f9d9c4179a,
+	0xeffdafa2ae3d3c60, 0xf7c8075bb49496c4, 0x9cc5c7141d1cd4e3, 0x78bd1638218e5534,
+	0xb2f11568f850246a, 0xedfabcfa9502bc29, 0x796ce5f2da23051b, 0xaae128b0dc93537c,
+	0x3a493da0ee4b29ae, 0xb5df6b2c416895d7, 0xfcabbd25122d7f37, 0x70810b58105dc4b1,
+	0xe10fdd37f7882a90, 0x524dcab5518a3f5c, 0x3c9e85878451255b, 0x4029828119bd34e2,
+	0x74a05b6f5d3ceccb, 0xb610021542e13eca, 0x0ff979d12f59e2ac, 0x6037da27e4f9cc50,
+	0x5e92975a0df1847d, 0xd66de190d3e623fe, 0x5032d6b87b568048, 0x9a36b7ce8235216e,
+	0x80272a7a24f64b4a, 0x93efed8b8c6916f7, 0x37ddbff44cce1555, 0x4b95db5d4b99bd25,
+	0x92d3fda169812fc0, 0xfb1a4a9a90660bb6, 0x730c196946a4b9b2, 0x81e289aa7f49da68,
+	0x64669a0f83b1a05f, 0x27b3ff7d9644f48b, 0xcc6b615c8db675b3, 0x674f20b9bcebbe95,
+	0x6f31238275655982, 0x5ae488713e45cf05, 0xbf619f9954c21157, 0xeabac46040a8eae9,
+	0x454c6fe9f2c0c1cd, 0x419cf6496412691c, 0xd3dc3bef265b0f70, 0x6d0e60f5c3578a9e,
+}
+
+T4 := [?]u64 {
+	0x5b0e608526323c55, 0x1a46c1a9fa1b59f5, 0xa9e245a17c4c8ffa, 0x65ca5159db2955d7,
+	0x05db0a76ce35afc2, 0x81eac77ea9113d45, 0x528ef88ab6ac0a0d, 0xa09ea253597be3ff,
+	0x430ddfb3ac48cd56, 0xc4b3a67af45ce46f, 0x4ececfd8fbe2d05e, 0x3ef56f10b39935f0,
+	0x0b22d6829cd619c6, 0x17fd460a74df2069, 0x6cf8cc8e8510ed40, 0xd6c824bf3a6ecaa7,
+	0x61243d581a817049, 0x048bacb6bbc163a2, 0xd9a38ac27d44cc32, 0x7fddff5baaf410ab,
+	0xad6d495aa804824b, 0xe1a6a74f2d8c9f94, 0xd4f7851235dee8e3, 0xfd4b7f886540d893,
+	0x247c20042aa4bfda, 0x096ea1c517d1327c, 0xd56966b4361a6685, 0x277da5c31221057d,
+	0x94d59893a43acff7, 0x64f0c51ccdc02281, 0x3d33bcc4ff6189db, 0xe005cb184ce66af1,
+	0xff5ccd1d1db99bea, 0xb0b854a7fe42980f, 0x7bd46a6a718d4b9f, 0xd10fa8cc22a5fd8c,
+	0xd31484952be4bd31, 0xc7fa975fcb243847, 0x4886ed1e5846c407, 0x28cddb791eb70b04,
+	0xc2b00be2f573417f, 0x5c9590452180f877, 0x7a6bddfff370eb00, 0xce509e38d6d9d6a4,
+	0xebeb0f00647fa702, 0x1dcc06cf76606f06, 0xe4d9f28ba286ff0a, 0xd85a305dc918c262,
+	0x475b1d8732225f54, 0x2d4fb51668ccb5fe, 0xa679b9d9d72bba20, 0x53841c0d912d43a5,
+	0x3b7eaa48bf12a4e8, 0x781e0e47f22f1ddf, 0xeff20ce60ab50973, 0x20d261d19dffb742,
+	0x16a12b03062a2e39, 0x1960eb2239650495, 0x251c16fed50eb8b8, 0x9ac0c330f826016e,
+	0xed152665953e7671, 0x02d63194a6369570, 0x5074f08394b1c987, 0x70ba598c90b25ce1,
+	0x794a15810b9742f6, 0x0d5925e9fcaf8c6c, 0x3067716cd868744e, 0x910ab077e8d7731b,
+	0x6a61bbdb5ac42f61, 0x93513efbf0851567, 0xf494724b9e83e9d5, 0xe887e1985c09648d,
+	0x34b1d3c675370cfd, 0xdc35e433bc0d255d, 0xd0aab84234131be0, 0x08042a50b48b7eaf,
+	0x9997c4ee44a3ab35, 0x829a7b49201799d0, 0x263b8307b7c54441, 0x752f95f4fd6a6ca6,
+	0x927217402c08c6e5, 0x2a8ab754a795d9ee, 0xa442f7552f72943d, 0x2c31334e19781208,
+	0x4fa98d7ceaee6291, 0x55c3862f665db309, 0xbd0610175d53b1f3, 0x46fe6cb840413f27,
+	0x3fe03792df0cfa59, 0xcfe700372eb85e8f, 0xa7be29e7adbce118, 0xe544ee5cde8431dd,
+	0x8a781b1b41f1873e, 0xa5c94c78a0d2f0e7, 0x39412e2877b60728, 0xa1265ef3afc9a62c,
+	0xbcc2770c6a2506c5, 0x3ab66dd5dce1ce12, 0xe65499d04a675b37, 0x7d8f523481bfd216,
+	0x0f6f64fcec15f389, 0x74efbe618b5b13c8, 0xacdc82b714273e1d, 0xdd40bfe003199d17,
+	0x37e99257e7e061f8, 0xfa52626904775aaa, 0x8bbbf63a463d56f9, 0xf0013f1543a26e64,
+	0xa8307e9f879ec898, 0xcc4c27a4150177cc, 0x1b432f2cca1d3348, 0xde1d1f8f9f6fa013,
+	0x606602a047a7ddd6, 0xd237ab64cc1cb2c7, 0x9b938e7225fcd1d3, 0xec4e03708e0ff476,
+	0xfeb2fbda3d03c12d, 0xae0bced2ee43889a, 0x22cb8923ebfb4f43, 0x69360d013cf7396d,
+	0x855e3602d2d4e022, 0x073805bad01f784c, 0x33e17a133852f546, 0xdf4874058ac7b638,
+	0xba92b29c678aa14a, 0x0ce89fc76cfaadcd, 0x5f9d4e0908339e34, 0xf1afe9291f5923b9,
+	0x6e3480f60f4a265f, 0xeebf3a2ab29b841c, 0xe21938a88f91b4ad, 0x57dfeff845c6d3c3,
+	0x2f006b0bf62caaf2, 0x62f479ef6f75ee78, 0x11a55ad41c8916a9, 0xf229d29084fed453,
+	0x42f1c27b16b000e6, 0x2b1f76749823c074, 0x4b76eca3c2745360, 0x8c98f463b91691bd,
+	0x14bcc93cf1ade66a, 0x8885213e6d458397, 0x8e177df0274d4711, 0xb49b73b5503f2951,
+	0x10168168c3f96b6b, 0x0e3d963b63cab0ae, 0x8dfc4b5655a1db14, 0xf789f1356e14de5c,
+	0x683e68af4e51dac1, 0xc9a84f9d8d4b0fd9, 0x3691e03f52a0f9d1, 0x5ed86e46e1878e80,
+	0x3c711a0e99d07150, 0x5a0865b20c4e9310, 0x56fbfc1fe4f0682e, 0xea8d5de3105edf9b,
+	0x71abfdb12379187a, 0x2eb99de1bee77b9c, 0x21ecc0ea33cf4523, 0x59a4d7521805c7a1,
+	0x3896f5eb56ae7c72, 0xaa638f3db18f75dc, 0x9f39358dabe9808e, 0xb7defa91c00b72ac,
+	0x6b5541fd62492d92, 0x6dc6dee8f92e4d5b, 0x353f57abc4beea7e, 0x735769d6da5690ce,
+	0x0a234aa642391484, 0xf6f9508028f80d9d, 0xb8e319a27ab3f215, 0x31ad9c1151341a4d,
+	0x773c22a57bef5805, 0x45c7561a07968633, 0xf913da9e249dbe36, 0xda652d9b78a64c68,
+	0x4c27a97f3bc334ef, 0x76621220e66b17f4, 0x967743899acd7d0b, 0xf3ee5bcae0ed6782,
+	0x409f753600c879fc, 0x06d09a39b5926db6, 0x6f83aeb0317ac588, 0x01e6ca4a86381f21,
+	0x66ff3462d19f3025, 0x72207c24ddfd3bfb, 0x4af6b6d3e2ece2eb, 0x9c994dbec7ea08de,
+	0x49ace597b09a8bc4, 0xb38c4766cf0797ba, 0x131b9373c57c2a75, 0xb1822cce61931e58,
+	0x9d7555b909ba1c0c, 0x127fafdd937d11d2, 0x29da3badc66d92e4, 0xa2c1d57154c2ecbc,
+	0x58c5134d82f6fe24, 0x1c3ae3515b62274f, 0xe907c82e01cb8126, 0xf8ed091913e37fcb,
+	0x3249d8f9c80046c9, 0x80cf9bede388fb63, 0x1881539a116cf19e, 0x5103f3f76bd52457,
+	0x15b7e6f5ae47f7a8, 0xdbd7c6ded47e9ccf, 0x44e55c410228bb1a, 0xb647d4255edb4e99,
+	0x5d11882bb8aafc30, 0xf5098bbb29d3212a, 0x8fb5ea14e90296b3, 0x677b942157dd025a,
+	0xfb58e7c0a390acb5, 0x89d3674c83bd4a01, 0x9e2da4df4bf3b93b, 0xfcc41e328cab4829,
+	0x03f38c96ba582c52, 0xcad1bdbd7fd85db2, 0xbbb442c16082ae83, 0xb95fe86ba5da9ab0,
+	0xb22e04673771a93f, 0x845358c9493152d8, 0xbe2a488697b4541e, 0x95a2dc2dd38e6966,
+	0xc02c11ac923c852b, 0x2388b1990df2a87b, 0x7c8008fa1b4f37be, 0x1f70d0c84d54e503,
+	0x5490adec7ece57d4, 0x002b3c27d9063a3a, 0x7eaea3848030a2bf, 0xc602326ded2003c0,
+	0x83a7287d69a94086, 0xc57a5fcb30f57a8a, 0xb56844e479ebe779, 0xa373b40f05dcbce9,
+	0xd71a786e88570ee2, 0x879cbacdbde8f6a0, 0x976ad1bcc164a32f, 0xab21e25e9666d78b,
+	0x901063aae5e5c33c, 0x9818b34448698d90, 0xe36487ae3e1e8abb, 0xafbdf931893bdcb4,
+	0x6345a0dc5fbbd519, 0x8628fe269b9465ca, 0x1e5d01603f9c51ec, 0x4de44006a15049b7,
+	0xbf6c70e5f776cbb1, 0x411218f2ef552bed, 0xcb0c0708705a36a3, 0xe74d14754f986044,
+	0xcd56d9430ea8280e, 0xc12591d7535f5065, 0xc83223f1720aef96, 0xc3a0396f7363a51f,
+}
+
+Tiger_Context :: struct {
+    a:      u64,
+	b:      u64,
+	c:      u64,
+	x:      [64]byte,
+	nx:     int,
+	length: u64,
+	ver:    int,
+}
+
+round :: #force_inline proc "contextless" (a, b, c, x, mul: u64) -> (u64, u64, u64) {
+	a, b, c := a, b, c
+	c ~= x
+	a -= T1[c & 0xff] ~ T2[(c >> 16) & 0xff] ~ T3[(c >> 32) & 0xff] ~ T4[(c >> 48) & 0xff]
+	b += T4[(c >> 8) & 0xff] ~ T3[(c >> 24) & 0xff] ~ T2[(c >> 40) & 0xff] ~ T1[(c >> 56) & 0xff]
+	b *= mul
+	return a, b, c
+}
+
+pass :: #force_inline proc "contextless" (a, b, c: u64, d: []u64, mul: u64) -> (x, y, z: u64) {
+	x, y, z = round(a, b, c, d[0], mul)
+	y, z, x = round(y, z, x, d[1], mul)
+	z, x, y = round(z, x, y, d[2], mul)
+	x, y, z = round(x, y, z, d[3], mul)
+	y, z, x = round(y, z, x, d[4], mul)
+	z, x, y = round(z, x, y, d[5], mul)
+	x, y, z = round(x, y, z, d[6], mul)
+	y, z, x = round(y, z, x, d[7], mul)
+	return
+}
+
+key_schedule :: #force_inline proc "contextless" (x: []u64) {
+	x[0] -= x[7] ~ 0xa5a5a5a5a5a5a5a5
+	x[1] ~= x[0]
+	x[2] += x[1]
+	x[3] -= x[2] ~ ((~x[1]) << 19)
+	x[4] ~= x[3]
+	x[5] += x[4]
+	x[6] -= x[5] ~ ((~x[4]) >> 23)
+	x[7] ~= x[6]
+	x[0] += x[7]
+	x[1] -= x[0] ~ ((~x[7]) << 19)
+	x[2] ~= x[1]
+	x[3] += x[2]
+	x[4] -= x[3] ~ ((~x[2]) >> 23)
+	x[5] ~= x[4]
+	x[6] += x[5]
+	x[7] -= x[6] ~ 0x0123456789abcdef
+}
+
+compress :: #force_inline proc "contextless" (ctx: ^Tiger_Context, data: []byte) {
+	a := ctx.a
+	b := ctx.b
+	c := ctx.c
+	x := util.cast_slice([]u64, data)
+	ctx.a, ctx.b, ctx.c = pass(ctx.a, ctx.b, ctx.c, x, 5)
+	key_schedule(x)
+	ctx.c, ctx.a, ctx.b = pass(ctx.c, ctx.a, ctx.b, x, 7)
+	key_schedule(x)
+	ctx.b, ctx.c, ctx.a = pass(ctx.b, ctx.c, ctx.a, x, 9)
+	ctx.a ~= a
+	ctx.b -= b
+	ctx.c += c
+}
+
+init :: proc "contextless" (ctx: ^Tiger_Context) {
+	ctx.a = 0x0123456789abcdef
+	ctx.b = 0xfedcba9876543210
+	ctx.c = 0xf096a5b4c3b2e187
+}
+
+update :: proc(ctx: ^Tiger_Context, input: []byte) {
+	p := make([]byte, len(input))
+	copy(p, input)
+
+	length     := len(p)
+	ctx.length += u64(length)
+	if ctx.nx > 0 {
+		n := len(p)
+		if n > 64 - ctx.nx {
+			n = 64 - ctx.nx
+		}
+		copy(ctx.x[ctx.nx:ctx.nx + n], p[:n])
+		ctx.nx += n
+		if ctx.nx == 64 {
+			compress(ctx, ctx.x[:64 - 1])
+			ctx.nx = 0
+		}
+		p = p[n:]
+	}
+	for len(p) >= 64 {
+		compress(ctx, p[:64])
+		p = p[64:]
+	}
+	if len(p) > 0 {
+		ctx.nx = copy(ctx.x[:], p)
+	}
+}
+
+final :: proc(ctx: ^Tiger_Context, hash: []byte) {
+	length := ctx.length
+	tmp: [64]byte
+	if ctx.ver == 1 {
+		tmp[0] = 0x01
+	} else {
+		tmp[0] = 0x80
+	}
+
+	size := length & 0x3f
+	if size < 56 {
+		update(ctx, tmp[:56 - size])
+	} else {
+		update(ctx, tmp[:64 + 56 - size])
+	}
+
+	length <<= 3
+	for i := uint(0); i < 8; i += 1 {
+		tmp[i] = byte(length >> (8 * i))
+	}
+	update(ctx, tmp[:8])
+
+	for i := uint(0); i < 8; i += 1 {
+		tmp[i]      = byte(ctx.a >> (8 * i))
+		tmp[i + 8]  = byte(ctx.b >> (8 * i))
+		tmp[i + 16] = byte(ctx.c >> (8 * i))
+	}
+	copy(hash[:], tmp[:len(hash)])
+}

core/crypto/blake/blake.odin

@@ -0,0 +1,637 @@
+package blake
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the BLAKE hashing algorithm, as defined in <https://web.archive.org/web/20190915215948/https://131002.net/blake>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc "contextless" (data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc "contextless" (data: []byte) -> [28]byte {
+    hash: [28]byte
+    ctx: Blake256_Context
+    ctx.is224 = true
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+    hash: [28]byte
+    ctx: Blake256_Context
+    ctx.is224 = true
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc "contextless" (data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc "contextless" (data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Blake256_Context
+    ctx.is224 = false
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Blake256_Context
+    ctx.is224 = false
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc "contextless" (data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc "contextless" (data: []byte) -> [48]byte {
+    hash: [48]byte
+    ctx: Blake512_Context
+    ctx.is384 = true
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+    hash: [48]byte
+    ctx: Blake512_Context
+    ctx.is384 = true
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc "contextless" (data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc "contextless" (data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: Blake512_Context
+    ctx.is384 = false
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: Blake512_Context
+    ctx.is384 = false
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc "contextless" (ctx: ^$T) {
+    when T == Blake256_Context {
+        if ctx.is224 {
+            ctx.h[0] = 0xc1059ed8
+            ctx.h[1] = 0x367cd507
+            ctx.h[2] = 0x3070dd17
+            ctx.h[3] = 0xf70e5939
+            ctx.h[4] = 0xffc00b31
+            ctx.h[5] = 0x68581511
+            ctx.h[6] = 0x64f98fa7
+            ctx.h[7] = 0xbefa4fa4
+        } else {
+            ctx.h[0] = 0x6a09e667
+            ctx.h[1] = 0xbb67ae85
+            ctx.h[2] = 0x3c6ef372
+            ctx.h[3] = 0xa54ff53a
+            ctx.h[4] = 0x510e527f
+            ctx.h[5] = 0x9b05688c
+            ctx.h[6] = 0x1f83d9ab
+            ctx.h[7] = 0x5be0cd19
+        }
+    } else when T == Blake512_Context {
+        if ctx.is384 {
+            ctx.h[0] = 0xcbbb9d5dc1059ed8
+            ctx.h[1] = 0x629a292a367cd507
+            ctx.h[2] = 0x9159015a3070dd17
+            ctx.h[3] = 0x152fecd8f70e5939
+            ctx.h[4] = 0x67332667ffc00b31
+            ctx.h[5] = 0x8eb44a8768581511
+            ctx.h[6] = 0xdb0c2e0d64f98fa7
+            ctx.h[7] = 0x47b5481dbefa4fa4
+        } else {
+            ctx.h[0] = 0x6a09e667f3bcc908
+            ctx.h[1] = 0xbb67ae8584caa73b
+            ctx.h[2] = 0x3c6ef372fe94f82b
+            ctx.h[3] = 0xa54ff53a5f1d36f1
+            ctx.h[4] = 0x510e527fade682d1
+            ctx.h[5] = 0x9b05688c2b3e6c1f
+            ctx.h[6] = 0x1f83d9abfb41bd6b
+            ctx.h[7] = 0x5be0cd19137e2179
+        }
+    }
+}
+
+update :: proc "contextless" (ctx: ^$T, data: []byte) {
+    data := data
+    when T == Blake256_Context {
+        if ctx.nx > 0 {
+            n := copy(ctx.x[ctx.nx:], data)
+            ctx.nx += n
+            if ctx.nx == BLOCKSIZE_256 {
+                block256(ctx, ctx.x[:])
+                ctx.nx = 0
+            }
+            data = data[n:]
+        }
+        if len(data) >= BLOCKSIZE_256 {
+            n := len(data) &~ (BLOCKSIZE_256 - 1)
+            block256(ctx, data[:n])
+            data = data[n:]
+        }
+        if len(data) > 0 {
+            ctx.nx = copy(ctx.x[:], data)
+        }
+    } else when T == Blake512_Context {
+        if ctx.nx > 0 {
+            n := copy(ctx.x[ctx.nx:], data)
+            ctx.nx += n
+            if ctx.nx == BLOCKSIZE_512 {
+                block512(ctx, ctx.x[:])
+                ctx.nx = 0
+            }
+            data = data[n:]
+        }
+        if len(data) >= BLOCKSIZE_512 {
+            n := len(data) &~ (BLOCKSIZE_512 - 1)
+            block512(ctx, data[:n])
+            data = data[n:]
+        }
+        if len(data) > 0 {
+            ctx.nx = copy(ctx.x[:], data)
+        }
+    }
+}
+
+final :: proc "contextless" (ctx: ^$T, hash: []byte) {
+    when T == Blake256_Context {
+        tmp: [65]byte
+    } else when T == Blake512_Context {
+        tmp: [129]byte
+    }
+    nx     := u64(ctx.nx)
+    tmp[0]  = 0x80
+    length := (ctx.t + nx) << 3
+
+    when T == Blake256_Context {
+        if nx == 55 {
+            if ctx.is224 {
+                write_additional(ctx, {0x80})
+            } else {
+                write_additional(ctx, {0x81})
+            }
+        } else {
+            if nx < 55 {
+                if nx == 0 {
+                    ctx.nullt = true
+                }
+                write_additional(ctx, tmp[0 : 55 - nx])
+            } else { 
+                write_additional(ctx, tmp[0 : 64 - nx])
+                write_additional(ctx, tmp[1:56])
+                ctx.nullt = true
+            }
+            if ctx.is224 {
+                write_additional(ctx, {0x00})
+            } else {
+                write_additional(ctx, {0x01})
+            }
+        }
+
+        for i : uint = 0; i < 8; i += 1 {
+            tmp[i] = byte(length >> (56 - 8 * i))
+        }
+        write_additional(ctx, tmp[0:8])
+
+        h := ctx.h[:]
+        if ctx.is224 {
+            h = h[0:7]
+        }
+        for s, i in h {
+            hash[i * 4]     = byte(s >> 24)
+            hash[i * 4 + 1] = byte(s >> 16)
+            hash[i * 4 + 2] = byte(s >> 8)
+            hash[i * 4 + 3] = byte(s)
+        }
+    } else when T == Blake512_Context {
+        if nx == 111 {
+            if ctx.is384 {
+                write_additional(ctx, {0x80})
+            } else {
+                write_additional(ctx, {0x81})
+            }
+        } else {
+            if nx < 111 {
+                if nx == 0 {
+                    ctx.nullt = true
+                }
+                write_additional(ctx, tmp[0 : 111 - nx])
+            } else { 
+                write_additional(ctx, tmp[0 : 128 - nx])
+                write_additional(ctx, tmp[1:112])
+                ctx.nullt = true
+            }
+            if ctx.is384 {
+                write_additional(ctx, {0x00})
+            } else {
+                write_additional(ctx, {0x01})
+            }
+        }
+
+        for i : uint = 0; i < 16; i += 1 {
+            tmp[i] = byte(length >> (120 - 8 * i))
+        }
+        write_additional(ctx, tmp[0:16])
+
+        h := ctx.h[:]
+        if ctx.is384 {
+            h = h[0:6]
+        }
+        for s, i in h {
+            hash[i * 8]     = byte(s >> 56)
+            hash[i * 8 + 1] = byte(s >> 48)
+            hash[i * 8 + 2] = byte(s >> 40)
+            hash[i * 8 + 3] = byte(s >> 32)
+            hash[i * 8 + 4] = byte(s >> 24)
+            hash[i * 8 + 5] = byte(s >> 16)
+            hash[i * 8 + 6] = byte(s >> 8)
+            hash[i * 8 + 7] = byte(s)
+        }
+    }
+}
+
+SIZE_224 :: 28
+SIZE_256 :: 32
+SIZE_384 :: 48
+SIZE_512 :: 64
+BLOCKSIZE_256 :: 64
+BLOCKSIZE_512 :: 128
+
+Blake256_Context :: struct {
+    h:     [8]u32,
+    s:     [4]u32,
+    t:     u64,
+    x:     [64]byte,
+    nx:    int,
+    is224: bool,
+    nullt: bool,
+}
+
+Blake512_Context :: struct {
+    h:     [8]u64,
+    s:     [4]u64,
+    t:     u64,
+    x:     [128]byte,
+    nx:    int,
+    is384: bool,
+    nullt: bool,
+}
+
+SIGMA := [?]int {
+    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+    14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3,
+    11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4,
+    7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8,
+    9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13,
+    2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9,
+    12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11,
+    13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10,
+    6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5,
+    10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0,
+}
+
+U256 := [16]u32 {
+    0x243f6a88, 0x85a308d3, 0x13198a2e, 0x03707344,
+    0xa4093822, 0x299f31d0, 0x082efa98, 0xec4e6c89,
+    0x452821e6, 0x38d01377, 0xbe5466cf, 0x34e90c6c,
+    0xc0ac29b7, 0xc97c50dd, 0x3f84d5b5, 0xb5470917,
+}
+
+U512 := [16]u64 {
+    0x243f6a8885a308d3, 0x13198a2e03707344, 0xa4093822299f31d0, 0x082efa98ec4e6c89,
+    0x452821e638d01377, 0xbe5466cf34e90c6c, 0xc0ac29b7c97c50dd, 0x3f84d5b5b5470917,
+    0x9216d5d98979fb1b, 0xd1310ba698dfb5ac, 0x2ffd72dbd01adfb7, 0xb8e1afed6a267e96,
+    0xba7c9045f12c7f99, 0x24a19947b3916cf7, 0x0801f2e2858efc16, 0x636920d871574e69,
+}
+
+G256 :: #force_inline proc "contextless" (a, b, c, d: u32, m: [16]u32, i, j: int) -> (u32, u32, u32, u32) {
+    a, b, c, d := a, b, c, d
+    a += m[SIGMA[(i % 10) * 16 + (2 * j)]] ~ U256[SIGMA[(i % 10) * 16 + (2 * j + 1)]]
+    a += b
+    d ~= a
+    d = d << (32 - 16) | d >> 16
+    c += d
+    b ~= c
+    b = b << (32 - 12) | b >> 12
+    a += m[SIGMA[(i % 10) * 16 + (2 * j + 1)]] ~ U256[SIGMA[(i % 10) * 16 + (2 * j)]]
+    a += b
+    d ~= a
+    d = d << (32 - 8) | d >> 8
+    c += d
+    b ~= c
+    b = b << (32 - 7) | b >> 7
+    return a, b, c, d
+}
+
+G512 :: #force_inline proc "contextless" (a, b, c, d: u64, m: [16]u64, i, j: int) -> (u64, u64, u64, u64) {
+    a, b, c, d := a, b, c, d
+    a += m[SIGMA[(i % 10) * 16 + (2 * j)]] ~ U512[SIGMA[(i % 10) * 16 + (2 * j + 1)]]
+    a += b
+    d ~= a
+    d = d << (64 - 32) | d >> 32
+    c += d
+    b ~= c
+    b = b << (64 - 25) | b >> 25
+    a += m[SIGMA[(i % 10) * 16 + (2 * j + 1)]] ~ U512[SIGMA[(i % 10) * 16 + (2 * j)]]
+    a += b
+    d ~= a
+    d = d << (64 - 16) | d >> 16
+    c += d
+    b ~= c
+    b = b << (64 - 11) | b >> 11
+    return a, b, c, d
+}
+
+block256 :: proc "contextless" (ctx: ^Blake256_Context, p: []byte) #no_bounds_check {
+    i, j: int     = ---, ---
+    v, m: [16]u32 = ---, ---
+    p := p
+    for len(p) >= BLOCKSIZE_256 {
+        v[0]  = ctx.h[0]
+        v[1]  = ctx.h[1]
+        v[2]  = ctx.h[2]
+        v[3]  = ctx.h[3]
+        v[4]  = ctx.h[4]
+        v[5]  = ctx.h[5]
+        v[6]  = ctx.h[6]
+        v[7]  = ctx.h[7]
+        v[8]  = ctx.s[0] ~ U256[0]
+        v[9]  = ctx.s[1] ~ U256[1]
+        v[10] = ctx.s[2] ~ U256[2]
+        v[11] = ctx.s[3] ~ U256[3]
+        v[12] = U256[4]
+        v[13] = U256[5]
+        v[14] = U256[6]
+        v[15] = U256[7]
+
+        ctx.t += 512
+        if !ctx.nullt {
+            v[12] ~= u32(ctx.t)
+            v[13] ~= u32(ctx.t)
+            v[14] ~= u32(ctx.t >> 32)
+            v[15] ~= u32(ctx.t >> 32)
+        }
+
+        for i, j = 0, 0; i < 16; i, j = i+1, j+4 {
+            m[i] = u32(p[j]) << 24 | u32(p[j + 1]) << 16 | u32(p[j + 2]) << 8 | u32(p[j + 3])
+        }
+
+        for i = 0; i < 14; i += 1 {
+            v[0], v[4], v[8],  v[12] = G256(v[0], v[4], v[8],  v[12], m, i, 0)
+            v[1], v[5], v[9],  v[13] = G256(v[1], v[5], v[9],  v[13], m, i, 1)
+            v[2], v[6], v[10], v[14] = G256(v[2], v[6], v[10], v[14], m, i, 2)
+            v[3], v[7], v[11], v[15] = G256(v[3], v[7], v[11], v[15], m, i, 3)
+            v[0], v[5], v[10], v[15] = G256(v[0], v[5], v[10], v[15], m, i, 4)
+            v[1], v[6], v[11], v[12] = G256(v[1], v[6], v[11], v[12], m, i, 5)
+            v[2], v[7], v[8],  v[13] = G256(v[2], v[7], v[8],  v[13], m, i, 6)
+            v[3], v[4], v[9],  v[14] = G256(v[3], v[4], v[9],  v[14], m, i, 7)
+        }
+
+        for i = 0; i < 8; i += 1 {
+            ctx.h[i] ~= ctx.s[i % 4] ~ v[i] ~ v[i + 8]
+        }
+        p = p[BLOCKSIZE_256:]
+    }
+}
+
+block512 :: proc "contextless" (ctx: ^Blake512_Context, p: []byte) #no_bounds_check {
+    i, j: int     = ---, ---
+    v, m: [16]u64 = ---, ---
+    p := p
+    for len(p) >= BLOCKSIZE_512 {
+        v[0]  = ctx.h[0]
+        v[1]  = ctx.h[1]
+        v[2]  = ctx.h[2]
+        v[3]  = ctx.h[3]
+        v[4]  = ctx.h[4]
+        v[5]  = ctx.h[5]
+        v[6]  = ctx.h[6]
+        v[7]  = ctx.h[7]
+        v[8]  = ctx.s[0] ~ U512[0]
+        v[9]  = ctx.s[1] ~ U512[1]
+        v[10] = ctx.s[2] ~ U512[2]
+        v[11] = ctx.s[3] ~ U512[3]
+        v[12] = U512[4]
+        v[13] = U512[5]
+        v[14] = U512[6]
+        v[15] = U512[7]
+
+        ctx.t += 1024
+        if !ctx.nullt {
+            v[12] ~= ctx.t
+            v[13] ~= ctx.t
+            v[14] ~= 0
+            v[15] ~= 0
+        }
+
+        for i, j = 0, 0; i < 16; i, j = i + 1, j + 8 {
+            m[i] = u64(p[j]) << 56     | u64(p[j + 1]) << 48 | u64(p[j + 2]) << 40 | u64(p[j + 3]) << 32 | 
+                   u64(p[j + 4]) << 24 | u64(p[j + 5]) << 16 | u64(p[j + 6]) << 8  | u64(p[j + 7])
+        }
+        for i = 0; i < 16; i += 1 {
+            v[0], v[4], v[8],  v[12] = G512(v[0], v[4], v[8],  v[12], m, i, 0)
+            v[1], v[5], v[9],  v[13] = G512(v[1], v[5], v[9],  v[13], m, i, 1)
+            v[2], v[6], v[10], v[14] = G512(v[2], v[6], v[10], v[14], m, i, 2)
+            v[3], v[7], v[11], v[15] = G512(v[3], v[7], v[11], v[15], m, i, 3)
+            v[0], v[5], v[10], v[15] = G512(v[0], v[5], v[10], v[15], m, i, 4)
+            v[1], v[6], v[11], v[12] = G512(v[1], v[6], v[11], v[12], m, i, 5)
+            v[2], v[7], v[8],  v[13] = G512(v[2], v[7], v[8],  v[13], m, i, 6)
+            v[3], v[4], v[9],  v[14] = G512(v[3], v[4], v[9],  v[14], m, i, 7)
+        }
+
+        for i = 0; i < 8; i += 1 {
+            ctx.h[i] ~= ctx.s[i % 4] ~ v[i] ~ v[i + 8]
+        }
+        p = p[BLOCKSIZE_512:]
+    }
+}
+
+write_additional :: proc "contextless" (ctx: ^$T, data: []byte) {
+	ctx.t -= u64(len(data)) << 3
+    update(ctx, data)
+}

core/crypto/blake2b/blake2b.odin

@@ -0,0 +1,101 @@
+package blake2b
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the BLAKE2B hashing algorithm.
+    BLAKE2B and BLAKE2B share the implementation in the _blake2 package.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_blake2"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [64]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: _blake2.Blake2b_Context
+    cfg: _blake2.Blake2_Config
+    cfg.size = _blake2.BLAKE2B_SIZE
+    ctx.cfg  = cfg
+    _blake2.init(&ctx)
+    _blake2.update(&ctx, data)
+    _blake2.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: _blake2.Blake2b_Context
+    cfg: _blake2.Blake2_Config
+    cfg.size = _blake2.BLAKE2B_SIZE
+    ctx.cfg  = cfg
+    _blake2.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _blake2.update(&ctx, buf[:read])
+        } 
+    }
+    _blake2.final(&ctx, hash[:])
+    return hash, true 
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+Blake2b_Context :: _blake2.Blake2b_Context
+
+init :: proc(ctx: ^_blake2.Blake2b_Context) {
+    _blake2.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_blake2.Blake2b_Context, data: []byte) {
+    _blake2.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_blake2.Blake2b_Context, hash: []byte) {
+    _blake2.final(ctx, hash)
+}

core/crypto/blake2s/blake2s.odin

@@ -0,0 +1,101 @@
+package blake2s
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the BLAKE2S hashing algorithm.
+    BLAKE2B and BLAKE2B share the implementation in the _blake2 package.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_blake2"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [32]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: _blake2.Blake2s_Context
+    cfg: _blake2.Blake2_Config
+    cfg.size = _blake2.BLAKE2S_SIZE
+    ctx.cfg  = cfg
+    _blake2.init(&ctx)
+    _blake2.update(&ctx, data)
+    _blake2.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: _blake2.Blake2s_Context
+    cfg: _blake2.Blake2_Config
+    cfg.size = _blake2.BLAKE2S_SIZE
+    ctx.cfg  = cfg
+    _blake2.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _blake2.update(&ctx, buf[:read])
+        } 
+    }
+    _blake2.final(&ctx, hash[:])
+    return hash, true 
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+Blake2s_Context :: _blake2.Blake2b_Context
+
+init :: proc(ctx: ^_blake2.Blake2s_Context) {
+    _blake2.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_blake2.Blake2s_Context, data: []byte) {
+    _blake2.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_blake2.Blake2s_Context, hash: []byte) {
+    _blake2.final(ctx, hash)
+}

core/crypto/chacha20/chacha20.odin

@@ -0,0 +1,581 @@
+package chacha20
+
+import "core:crypto/util"
+import "core:math/bits"
+import "core:mem"
+
+KEY_SIZE :: 32
+NONCE_SIZE :: 12
+XNONCE_SIZE :: 24
+
+_MAX_CTR_IETF :: 0xffffffff
+
+_BLOCK_SIZE :: 64
+_STATE_SIZE_U32 :: 16
+_ROUNDS :: 20
+
+_SIGMA_0 : u32 : 0x61707865
+_SIGMA_1 : u32 : 0x3320646e
+_SIGMA_2 : u32 : 0x79622d32
+_SIGMA_3 : u32 : 0x6b206574
+
+Context :: struct {
+	_s: [_STATE_SIZE_U32]u32,
+
+	_buffer: [_BLOCK_SIZE]byte,
+	_off: int,
+
+	_is_ietf_flavor: bool,
+	_is_initialized: bool,
+}
+
+init :: proc (ctx: ^Context, key, nonce: []byte) {
+	if len(key) != KEY_SIZE {
+		panic("crypto/chacha20: invalid ChaCha20 key size")
+	}
+	if n_len := len(nonce); n_len != NONCE_SIZE && n_len != XNONCE_SIZE {
+		panic("crypto/chacha20: invalid (X)ChaCha20 nonce size")
+	}
+
+	k, n := key, nonce
+
+	// Derive the XChaCha20 subkey and sub-nonce via HChaCha20.
+	is_xchacha := len(nonce) == XNONCE_SIZE
+	if is_xchacha {
+		sub_key := ctx._buffer[:KEY_SIZE]
+		_hchacha20(sub_key, k, n)
+		k = sub_key
+		n = n[16:24]
+	}
+
+	ctx._s[0] = _SIGMA_0
+	ctx._s[1] = _SIGMA_1
+	ctx._s[2] = _SIGMA_2
+	ctx._s[3] = _SIGMA_3
+	ctx._s[4] = util.U32_LE(k[0:4])
+	ctx._s[5] = util.U32_LE(k[4:8])
+	ctx._s[6] = util.U32_LE(k[8:12])
+	ctx._s[7] = util.U32_LE(k[12:16])
+	ctx._s[8] = util.U32_LE(k[16:20])
+	ctx._s[9] = util.U32_LE(k[20:24])
+	ctx._s[10] = util.U32_LE(k[24:28])
+	ctx._s[11] = util.U32_LE(k[28:32])
+	ctx._s[12] = 0
+	if !is_xchacha {
+		ctx._s[13] = util.U32_LE(n[0:4])
+		ctx._s[14] = util.U32_LE(n[4:8])
+		ctx._s[15] = util.U32_LE(n[8:12])
+	} else {
+		ctx._s[13] = 0
+		ctx._s[14] = util.U32_LE(n[0:4])
+		ctx._s[15] = util.U32_LE(n[4:8])
+
+		// The sub-key is stored in the keystream buffer.  While
+		// this will be overwritten in most circumstances, explicitly
+		// clear it out early.
+		mem.zero_explicit(&ctx._buffer, KEY_SIZE)
+	}
+
+	ctx._off = _BLOCK_SIZE
+	ctx._is_ietf_flavor = !is_xchacha
+	ctx._is_initialized = true
+}
+
+seek :: proc (ctx: ^Context, block_nr: u64) {
+	assert(ctx._is_initialized)
+
+	if ctx._is_ietf_flavor {
+		if block_nr > _MAX_CTR_IETF {
+			panic("crypto/chacha20: attempted to seek past maximum counter")
+		}
+	} else {
+		ctx._s[13] = u32(block_nr >> 32)
+	}
+	ctx._s[12] = u32(block_nr)
+	ctx._off = _BLOCK_SIZE
+}
+
+xor_bytes :: proc (ctx: ^Context, dst, src: []byte) {
+	assert(ctx._is_initialized)
+
+	// TODO: Enforcing that dst and src alias exactly or not at all
+	// is a good idea, though odd aliasing should be extremely uncommon.
+
+	src, dst := src, dst
+	if dst_len := len(dst); dst_len < len(src) {
+		src = src[:dst_len]
+	}
+
+	for remaining := len(src); remaining > 0; {
+		// Process multiple blocks at once
+		if ctx._off == _BLOCK_SIZE {
+			if nr_blocks := remaining / _BLOCK_SIZE; nr_blocks > 0 {
+				direct_bytes := nr_blocks * _BLOCK_SIZE
+				_do_blocks(ctx, dst, src, nr_blocks)
+				remaining -= direct_bytes
+				if remaining == 0 {
+					return
+				}
+				dst = dst[direct_bytes:]
+				src = src[direct_bytes:]
+			}
+
+			// If there is a partial block, generate and buffer 1 block
+			// worth of keystream.
+			_do_blocks(ctx, ctx._buffer[:], nil, 1)
+			ctx._off = 0
+		}
+
+		// Process partial blocks from the buffered keystream.
+		to_xor := min(_BLOCK_SIZE - ctx._off, remaining)
+		buffered_keystream := ctx._buffer[ctx._off:]
+		for i := 0; i < to_xor; i = i + 1 {
+			dst[i] = buffered_keystream[i] ~ src[i]
+		}
+		ctx._off += to_xor
+		dst = dst[to_xor:]
+		src = src[to_xor:]
+		remaining -= to_xor
+	}
+}
+
+keystream_bytes :: proc (ctx: ^Context, dst: []byte) {
+	assert(ctx._is_initialized)
+
+	dst := dst
+	for remaining := len(dst); remaining > 0; {
+		// Process multiple blocks at once
+		if ctx._off == _BLOCK_SIZE {
+			if nr_blocks := remaining / _BLOCK_SIZE; nr_blocks > 0 {
+				direct_bytes := nr_blocks * _BLOCK_SIZE
+				_do_blocks(ctx, dst, nil, nr_blocks)
+				remaining -= direct_bytes
+				if remaining == 0 {
+					return
+				}
+				dst = dst[direct_bytes:]
+			}
+
+			// If there is a partial block, generate and buffer 1 block
+			// worth of keystream.
+			_do_blocks(ctx, ctx._buffer[:], nil, 1)
+			ctx._off = 0
+		}
+
+		// Process partial blocks from the buffered keystream.
+		to_copy := min(_BLOCK_SIZE - ctx._off, remaining)
+		buffered_keystream := ctx._buffer[ctx._off:]
+		copy(dst[:to_copy], buffered_keystream[:to_copy])
+		ctx._off += to_copy
+		dst = dst[to_copy:]
+		remaining -= to_copy
+	}
+}
+
+reset :: proc (ctx: ^Context) {
+	mem.zero_explicit(&ctx._s, size_of(ctx._s))
+	mem.zero_explicit(&ctx._buffer, size_of(ctx._buffer))
+
+	ctx._is_initialized = false
+}
+
+_do_blocks :: proc (ctx: ^Context, dst, src: []byte, nr_blocks: int) {
+	// Enforce the maximum consumed keystream per nonce.
+	//
+	// While all modern "standard" definitions of ChaCha20 use
+	// the IETF 32-bit counter, for XChaCha20 most common
+	// implementations allow for a 64-bit counter.
+	//
+	// Honestly, the answer here is "use a MRAE primitive", but
+	// go with common practice in the case of XChaCha20.
+	if ctx._is_ietf_flavor {
+		if u64(ctx._s[12]) + u64(nr_blocks) > 0xffffffff {
+			panic("crypto/chacha20: maximum ChaCha20 keystream per nonce reached")
+		}
+	} else {
+		ctr := (u64(ctx._s[13]) << 32) | u64(ctx._s[12])
+		if _, carry := bits.add_u64(ctr, u64(nr_blocks), 0); carry != 0 {
+			panic("crypto/chacha20: maximum XChaCha20 keystream per nonce reached")
+		}
+	}
+
+	dst, src := dst, src
+	x := &ctx._s
+	for n := 0; n < nr_blocks; n = n + 1 {
+		x0, x1, x2, x3 := _SIGMA_0, _SIGMA_1, _SIGMA_2, _SIGMA_3
+		x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15 := x[4], x[5], x[6], x[7], x[8], x[9], x[10], x[11], x[12], x[13], x[14], x[15]
+
+		for i := _ROUNDS; i > 0; i = i - 2 {
+			// Even when forcing inlining manually inlining all of
+			// these is decently faster.
+
+			// quarterround(x, 0, 4, 8, 12)
+			x0 += x4
+			x12 ~= x0
+			x12 = util.ROTL32(x12, 16)
+			x8 += x12
+			x4 ~= x8
+			x4 = util.ROTL32(x4, 12)
+			x0 += x4
+			x12 ~= x0
+			x12 = util.ROTL32(x12, 8)
+			x8 += x12
+			x4 ~= x8
+			x4 = util.ROTL32(x4, 7)
+
+			// quarterround(x, 1, 5, 9, 13)
+			x1 += x5
+			x13 ~= x1
+			x13 = util.ROTL32(x13, 16)
+			x9 += x13
+			x5 ~= x9
+			x5 = util.ROTL32(x5, 12)
+			x1 += x5
+			x13 ~= x1
+			x13 = util.ROTL32(x13, 8)
+			x9 += x13
+			x5 ~= x9
+			x5 = util.ROTL32(x5, 7)
+
+			// quarterround(x, 2, 6, 10, 14)
+			x2 += x6
+			x14 ~= x2
+			x14 = util.ROTL32(x14, 16)
+			x10 += x14
+			x6 ~= x10
+			x6 = util.ROTL32(x6, 12)
+			x2 += x6
+			x14 ~= x2
+			x14 = util.ROTL32(x14, 8)
+			x10 += x14
+			x6 ~= x10
+			x6 = util.ROTL32(x6, 7)
+
+			// quarterround(x, 3, 7, 11, 15)
+			x3 += x7
+			x15 ~= x3
+			x15 = util.ROTL32(x15, 16)
+			x11 += x15
+			x7 ~= x11
+			x7 = util.ROTL32(x7, 12)
+			x3 += x7
+			x15 ~= x3
+			x15 = util.ROTL32(x15, 8)
+			x11 += x15
+			x7 ~= x11
+			x7 = util.ROTL32(x7, 7)
+
+			// quarterround(x, 0, 5, 10, 15)
+			x0 += x5
+			x15 ~= x0
+			x15 = util.ROTL32(x15, 16)
+			x10 += x15
+			x5 ~= x10
+			x5 = util.ROTL32(x5, 12)
+			x0 += x5
+			x15 ~= x0
+			x15 = util.ROTL32(x15, 8)
+			x10 += x15
+			x5 ~= x10
+			x5 = util.ROTL32(x5, 7)
+
+			// quarterround(x, 1, 6, 11, 12)
+			x1 += x6
+			x12 ~= x1
+			x12 = util.ROTL32(x12, 16)
+			x11 += x12
+			x6 ~= x11
+			x6 = util.ROTL32(x6, 12)
+			x1 += x6
+			x12 ~= x1
+			x12 = util.ROTL32(x12, 8)
+			x11 += x12
+			x6 ~= x11
+			x6 = util.ROTL32(x6, 7)
+
+			// quarterround(x, 2, 7, 8, 13)
+			x2 += x7
+			x13 ~= x2
+			x13 = util.ROTL32(x13, 16)
+			x8 += x13
+			x7 ~= x8
+			x7 = util.ROTL32(x7, 12)
+			x2 += x7
+			x13 ~= x2
+			x13 = util.ROTL32(x13, 8)
+			x8 += x13
+			x7 ~= x8
+			x7 = util.ROTL32(x7, 7)
+
+			// quarterround(x, 3, 4, 9, 14)
+			x3 += x4
+			x14 ~= x3
+			x14 = util.ROTL32(x14, 16)
+			x9 += x14
+			x4 ~= x9
+			x4 = util.ROTL32(x4, 12)
+			x3 += x4
+			x14 ~= x3
+			x14 = util.ROTL32(x14, 8)
+			x9 += x14
+			x4 ~= x9
+			x4 = util.ROTL32(x4, 7)
+		}
+
+		x0 += _SIGMA_0
+		x1 += _SIGMA_1
+		x2 += _SIGMA_2
+		x3 += _SIGMA_3
+		x4 += x[4]
+		x5 += x[5]
+		x6 += x[6]
+		x7 += x[7]
+		x8 += x[8]
+		x9 += x[9]
+		x10 += x[10]
+		x11 += x[11]
+		x12 += x[12]
+		x13 += x[13]
+		x14 += x[14]
+		x15 += x[15]
+
+		// While the "correct" answer to getting more performance out of
+		// this is "use vector operations", support for that is currently
+		// a work in progress/to be designed.
+		//
+		// Until dedicated assembly can be written leverage the fact that
+		// the callers of this routine ensure that src/dst are valid.
+
+		when ODIN_ARCH == "386" || ODIN_ARCH == "amd64" {
+			// util.PUT_U32_LE/util.U32_LE are not required on little-endian
+			// systems that also happen to not be strict about aligned
+			// memory access.
+
+			dst_p := transmute(^[16]u32)(&dst[0])
+			if src != nil {
+				src_p := transmute(^[16]u32)(&src[0])
+				dst_p[0] = src_p[0] ~ x0
+				dst_p[1] = src_p[1] ~ x1
+				dst_p[2] = src_p[2] ~ x2
+				dst_p[3] = src_p[3] ~ x3
+				dst_p[4] = src_p[4] ~ x4
+				dst_p[5] = src_p[5] ~ x5
+				dst_p[6] = src_p[6] ~ x6
+				dst_p[7] = src_p[7] ~ x7
+				dst_p[8] = src_p[8] ~ x8
+				dst_p[9] = src_p[9] ~ x9
+				dst_p[10] = src_p[10] ~ x10
+				dst_p[11] = src_p[11] ~ x11
+				dst_p[12] = src_p[12] ~ x12
+				dst_p[13] = src_p[13] ~ x13
+				dst_p[14] = src_p[14] ~ x14
+				dst_p[15] = src_p[15] ~ x15
+				src = src[_BLOCK_SIZE:]
+			} else {
+				dst_p[0] = x0
+				dst_p[1] = x1
+				dst_p[2] = x2
+				dst_p[3] = x3
+				dst_p[4] = x4
+				dst_p[5] = x5
+				dst_p[6] = x6
+				dst_p[7] = x7
+				dst_p[8] = x8
+				dst_p[9] = x9
+				dst_p[10] = x10
+				dst_p[11] = x11
+				dst_p[12] = x12
+				dst_p[13] = x13
+				dst_p[14] = x14
+				dst_p[15] = x15
+			}
+			dst = dst[_BLOCK_SIZE:]
+		} else {
+			#no_bounds_check {
+				if src != nil {
+					util.PUT_U32_LE(dst[0:4], util.U32_LE(src[0:4]) ~ x0)
+					util.PUT_U32_LE(dst[4:8], util.U32_LE(src[4:8]) ~ x1)
+					util.PUT_U32_LE(dst[8:12], util.U32_LE(src[8:12]) ~ x2)
+					util.PUT_U32_LE(dst[12:16], util.U32_LE(src[12:16]) ~ x3)
+					util.PUT_U32_LE(dst[16:20], util.U32_LE(src[16:20]) ~ x4)
+					util.PUT_U32_LE(dst[20:24], util.U32_LE(src[20:24]) ~ x5)
+					util.PUT_U32_LE(dst[24:28], util.U32_LE(src[24:28]) ~ x6)
+					util.PUT_U32_LE(dst[28:32], util.U32_LE(src[28:32]) ~ x7)
+					util.PUT_U32_LE(dst[32:36], util.U32_LE(src[32:36]) ~ x8)
+					util.PUT_U32_LE(dst[36:40], util.U32_LE(src[36:40]) ~ x9)
+					util.PUT_U32_LE(dst[40:44], util.U32_LE(src[40:44]) ~ x10)
+					util.PUT_U32_LE(dst[44:48], util.U32_LE(src[44:48]) ~ x11)
+					util.PUT_U32_LE(dst[48:52], util.U32_LE(src[48:52]) ~ x12)
+					util.PUT_U32_LE(dst[52:56], util.U32_LE(src[52:56]) ~ x13)
+					util.PUT_U32_LE(dst[56:60], util.U32_LE(src[56:60]) ~ x14)
+					util.PUT_U32_LE(dst[60:64], util.U32_LE(src[60:64]) ~ x15)
+					src = src[_BLOCK_SIZE:]
+				} else {
+					util.PUT_U32_LE(dst[0:4], x0)
+					util.PUT_U32_LE(dst[4:8], x1)
+					util.PUT_U32_LE(dst[8:12], x2)
+					util.PUT_U32_LE(dst[12:16], x3)
+					util.PUT_U32_LE(dst[16:20], x4)
+					util.PUT_U32_LE(dst[20:24], x5)
+					util.PUT_U32_LE(dst[24:28], x6)
+					util.PUT_U32_LE(dst[28:32], x7)
+					util.PUT_U32_LE(dst[32:36], x8)
+					util.PUT_U32_LE(dst[36:40], x9)
+					util.PUT_U32_LE(dst[40:44], x10)
+					util.PUT_U32_LE(dst[44:48], x11)
+					util.PUT_U32_LE(dst[48:52], x12)
+					util.PUT_U32_LE(dst[52:56], x13)
+					util.PUT_U32_LE(dst[56:60], x14)
+					util.PUT_U32_LE(dst[60:64], x15)
+				}
+				dst = dst[_BLOCK_SIZE:]
+			}
+		}
+
+		// Increment the counter.  Overflow checking is done upon
+		// entry into the routine, so a 64-bit increment safely
+		// covers both cases.
+		new_ctr := ((u64(ctx._s[13]) << 32) | u64(ctx._s[12])) + 1
+		x[12] = u32(new_ctr)
+		x[13] = u32(new_ctr >> 32)
+	}
+}
+
+_hchacha20 :: proc (dst, key, nonce: []byte) {
+	x0, x1, x2, x3 := _SIGMA_0, _SIGMA_1, _SIGMA_2, _SIGMA_3
+	x4 := util.U32_LE(key[0:4])
+	x5 := util.U32_LE(key[4:8])
+	x6 := util.U32_LE(key[8:12])
+	x7 := util.U32_LE(key[12:16])
+	x8 := util.U32_LE(key[16:20])
+	x9 := util.U32_LE(key[20:24])
+	x10 := util.U32_LE(key[24:28])
+	x11 := util.U32_LE(key[28:32])
+	x12 := util.U32_LE(nonce[0:4])
+	x13 := util.U32_LE(nonce[4:8])
+	x14 := util.U32_LE(nonce[8:12])
+	x15 := util.U32_LE(nonce[12:16])
+
+	for i := _ROUNDS; i > 0; i = i - 2 {
+		// quarterround(x, 0, 4, 8, 12)
+		x0 += x4
+		x12 ~= x0
+		x12 = util.ROTL32(x12, 16)
+		x8 += x12
+		x4 ~= x8
+		x4 = util.ROTL32(x4, 12)
+		x0 += x4
+		x12 ~= x0
+		x12 = util.ROTL32(x12, 8)
+		x8 += x12
+		x4 ~= x8
+		x4 = util.ROTL32(x4, 7)
+
+		// quarterround(x, 1, 5, 9, 13)
+		x1 += x5
+		x13 ~= x1
+		x13 = util.ROTL32(x13, 16)
+		x9 += x13
+		x5 ~= x9
+		x5 = util.ROTL32(x5, 12)
+		x1 += x5
+		x13 ~= x1
+		x13 = util.ROTL32(x13, 8)
+		x9 += x13
+		x5 ~= x9
+		x5 = util.ROTL32(x5, 7)
+
+		// quarterround(x, 2, 6, 10, 14)
+		x2 += x6
+		x14 ~= x2
+		x14 = util.ROTL32(x14, 16)
+		x10 += x14
+		x6 ~= x10
+		x6 = util.ROTL32(x6, 12)
+		x2 += x6
+		x14 ~= x2
+		x14 = util.ROTL32(x14, 8)
+		x10 += x14
+		x6 ~= x10
+		x6 = util.ROTL32(x6, 7)
+
+		// quarterround(x, 3, 7, 11, 15)
+		x3 += x7
+		x15 ~= x3
+		x15 = util.ROTL32(x15, 16)
+		x11 += x15
+		x7 ~= x11
+		x7 = util.ROTL32(x7, 12)
+		x3 += x7
+		x15 ~= x3
+		x15 = util.ROTL32(x15, 8)
+		x11 += x15
+		x7 ~= x11
+		x7 = util.ROTL32(x7, 7)
+
+		// quarterround(x, 0, 5, 10, 15)
+		x0 += x5
+		x15 ~= x0
+		x15 = util.ROTL32(x15, 16)
+		x10 += x15
+		x5 ~= x10
+		x5 = util.ROTL32(x5, 12)
+		x0 += x5
+		x15 ~= x0
+		x15 = util.ROTL32(x15, 8)
+		x10 += x15
+		x5 ~= x10
+		x5 = util.ROTL32(x5, 7)
+
+		// quarterround(x, 1, 6, 11, 12)
+		x1 += x6
+		x12 ~= x1
+		x12 = util.ROTL32(x12, 16)
+		x11 += x12
+		x6 ~= x11
+		x6 = util.ROTL32(x6, 12)
+		x1 += x6
+		x12 ~= x1
+		x12 = util.ROTL32(x12, 8)
+		x11 += x12
+		x6 ~= x11
+		x6 = util.ROTL32(x6, 7)
+
+		// quarterround(x, 2, 7, 8, 13)
+		x2 += x7
+		x13 ~= x2
+		x13 = util.ROTL32(x13, 16)
+		x8 += x13
+		x7 ~= x8
+		x7 = util.ROTL32(x7, 12)
+		x2 += x7
+		x13 ~= x2
+		x13 = util.ROTL32(x13, 8)
+		x8 += x13
+		x7 ~= x8
+		x7 = util.ROTL32(x7, 7)
+
+		// quarterround(x, 3, 4, 9, 14)
+		x3 += x4
+		x14 ~= x3
+		x14 = util.ROTL32(x14, 16)
+		x9 += x14
+		x4 ~= x9
+		x4 = util.ROTL32(x4, 12)
+		x3 += x4
+		x14 ~= x3
+		x14 = util.ROTL32(x14, 8)
+		x9 += x14
+		x4 ~= x9
+		x4 = util.ROTL32(x4, 7)
+	}
+
+	util.PUT_U32_LE(dst[0:4], x0)
+	util.PUT_U32_LE(dst[4:8], x1)
+	util.PUT_U32_LE(dst[8:12], x2)
+	util.PUT_U32_LE(dst[12:16], x3)
+	util.PUT_U32_LE(dst[16:20], x12)
+	util.PUT_U32_LE(dst[20:24], x13)
+	util.PUT_U32_LE(dst[24:28], x14)
+	util.PUT_U32_LE(dst[28:32], x15)
+}

core/crypto/chacha20poly1305/chacha20poly1305.odin

@@ -0,0 +1,146 @@
+package chacha20poly1305
+
+import "core:crypto"
+import "core:crypto/chacha20"
+import "core:crypto/poly1305"
+import "core:crypto/util"
+import "core:mem"
+
+KEY_SIZE :: chacha20.KEY_SIZE
+NONCE_SIZE :: chacha20.NONCE_SIZE
+TAG_SIZE :: poly1305.TAG_SIZE
+
+_P_MAX :: 64 * 0xffffffff // 64 * (2^32-1)
+
+_validate_common_slice_sizes :: proc (tag, key, nonce, aad, text: []byte) {
+	if len(tag) != TAG_SIZE {
+		panic("crypto/chacha20poly1305: invalid destination tag size")
+	}
+	if len(key) != KEY_SIZE {
+		panic("crypto/chacha20poly1305: invalid key size")
+	}
+	if len(nonce) != NONCE_SIZE {
+		panic("crypto/chacha20poly1305: invalid nonce size")
+	}
+
+	#assert(size_of(int) == 8 || size_of(int) <= 4)
+	when size_of(int) == 8 {
+		// A_MAX = 2^64 - 1 due to the length field limit.
+		// P_MAX = 64 * (2^32 - 1) due to the IETF ChaCha20 counter limit.
+		//
+		// A_MAX is limited by size_of(int), so there is no need to
+		// enforce it. P_MAX only needs to be checked on 64-bit targets,
+		// for reasons that should be obvious.
+		if text_len := len(text); text_len > _P_MAX {
+			panic("crypto/chacha20poly1305: oversized src data")
+		}
+	}
+}
+
+_PAD: [16]byte
+_update_mac_pad16 :: #force_inline proc (ctx: ^poly1305.Context, x_len: int) {
+	if pad_len := 16 - (x_len & (16-1)); pad_len != 16 {
+		poly1305.update(ctx, _PAD[:pad_len])
+	}
+}
+
+encrypt :: proc (ciphertext, tag, key, nonce, aad, plaintext: []byte) {
+	_validate_common_slice_sizes(tag, key, nonce, aad, plaintext)
+	if len(ciphertext) != len(plaintext) {
+		panic("crypto/chacha20poly1305: invalid destination ciphertext size")
+	}
+
+	stream_ctx: chacha20.Context = ---
+	chacha20.init(&stream_ctx, key, nonce)
+
+	// otk = poly1305_key_gen(key, nonce)
+	otk: [poly1305.KEY_SIZE]byte = ---
+	chacha20.keystream_bytes(&stream_ctx, otk[:])
+	mac_ctx: poly1305.Context = ---
+	poly1305.init(&mac_ctx, otk[:])
+	mem.zero_explicit(&otk, size_of(otk))
+
+	aad_len, ciphertext_len := len(aad), len(ciphertext)
+
+	// There is nothing preventing aad and ciphertext from overlapping
+	// so auth the AAD before encrypting (slightly different from the
+	// RFC, since the RFC encrypts into a new buffer).
+	//
+	// mac_data = aad | pad16(aad)
+	poly1305.update(&mac_ctx, aad)
+	_update_mac_pad16(&mac_ctx, aad_len)
+
+	// ciphertext = chacha20_encrypt(key, 1, nonce, plaintext)
+	chacha20.seek(&stream_ctx, 1)
+	chacha20.xor_bytes(&stream_ctx, ciphertext, plaintext)
+	chacha20.reset(&stream_ctx) // Don't need the stream context anymore.
+
+	// mac_data |= ciphertext | pad16(ciphertext)
+	poly1305.update(&mac_ctx, ciphertext)
+	_update_mac_pad16(&mac_ctx, ciphertext_len)
+
+	// mac_data |= num_to_8_le_bytes(aad.length)
+	// mac_data |= num_to_8_le_bytes(ciphertext.length)
+	l_buf := otk[0:16] // Reuse the scratch buffer.
+	util.PUT_U64_LE(l_buf[0:8], u64(aad_len))
+	util.PUT_U64_LE(l_buf[8:16], u64(ciphertext_len))
+	poly1305.update(&mac_ctx, l_buf)
+
+	// tag = poly1305_mac(mac_data, otk)
+	poly1305.final(&mac_ctx, tag) // Implicitly sanitizes context.
+}
+
+decrypt :: proc (plaintext, tag, key, nonce, aad, ciphertext: []byte) -> bool {
+	_validate_common_slice_sizes(tag, key, nonce, aad, ciphertext)
+	if len(ciphertext) != len(plaintext) {
+		panic("crypto/chacha20poly1305: invalid destination plaintext size")
+	}
+
+	// Note: Unlike encrypt, this can fail early, so use defer for
+	// sanitization rather than assuming control flow reaches certain
+	// points where needed.
+
+	stream_ctx: chacha20.Context = ---
+	chacha20.init(&stream_ctx, key, nonce)
+
+	// otk = poly1305_key_gen(key, nonce)
+	otk: [poly1305.KEY_SIZE]byte = ---
+	chacha20.keystream_bytes(&stream_ctx, otk[:])
+	defer chacha20.reset(&stream_ctx)
+
+	mac_ctx: poly1305.Context = ---
+	poly1305.init(&mac_ctx, otk[:])
+	defer mem.zero_explicit(&otk, size_of(otk))
+
+	aad_len, ciphertext_len := len(aad), len(ciphertext)
+
+	// mac_data = aad | pad16(aad)
+	// mac_data |= ciphertext | pad16(ciphertext)
+	// mac_data |= num_to_8_le_bytes(aad.length)
+	// mac_data |= num_to_8_le_bytes(ciphertext.length)
+	poly1305.update(&mac_ctx, aad)
+	_update_mac_pad16(&mac_ctx, aad_len)
+	poly1305.update(&mac_ctx, ciphertext)
+	_update_mac_pad16(&mac_ctx, ciphertext_len)
+	l_buf := otk[0:16] // Reuse the scratch buffer.
+	util.PUT_U64_LE(l_buf[0:8], u64(aad_len))
+	util.PUT_U64_LE(l_buf[8:16], u64(ciphertext_len))
+	poly1305.update(&mac_ctx, l_buf)
+
+	// tag = poly1305_mac(mac_data, otk)
+	derived_tag := otk[0:poly1305.TAG_SIZE] // Reuse the scratch buffer again.
+	poly1305.final(&mac_ctx, derived_tag) // Implicitly sanitizes context.
+
+	// Validate the tag in constant time.
+	if crypto.compare_constant_time(tag, derived_tag) != 1 {
+		// Zero out the plaintext, as a defense in depth measure.
+		mem.zero_explicit(raw_data(plaintext), ciphertext_len)
+		return false
+	}
+
+	// plaintext = chacha20_decrypt(key, 1, nonce, ciphertext)
+	chacha20.seek(&stream_ctx, 1)
+	chacha20.xor_bytes(&stream_ctx, plaintext, ciphertext)
+
+	return true
+}

core/crypto/crypto.odin

@@ -0,0 +1,52 @@
+package crypto
+
+import "core:mem"
+
+// compare_constant_time returns 1 iff a and b are equal, 0 otherwise.
+//
+// The execution time of this routine is constant regardless of the contents
+// of the slices being compared, as long as the length of the slices is equal.
+// If the length of the two slices is different, it will early-return 0.
+compare_constant_time :: proc "contextless" (a, b: []byte) -> int {
+	// If the length of the slices is different, early return.
+	//
+	// This leaks the fact that the slices have a different length,
+	// but the routine is primarily intended for comparing things
+	// like MACS and password digests.
+	n := len(a)
+	if n != len(b) {
+		return 0
+	}
+
+	return compare_byte_ptrs_constant_time(raw_data(a), raw_data(b), n)
+}
+
+// compare_byte_ptrs_constant_time returns 1 iff the bytes pointed to by
+// a and b are equal, 0 otherwise.
+//
+// The execution time of this routine is constant regardless of the
+// contents of the memory being compared.
+compare_byte_ptrs_constant_time :: proc "contextless" (a, b: ^byte, n: int) -> int {
+	x := mem.slice_ptr(a, n)
+	y := mem.slice_ptr(b, n)
+
+	v: byte
+	for i in 0..<n {
+		v |= x[i] ~ y[i]
+	}
+
+	// After the loop, v == 0 iff a == b.  The subtraction will underflow
+	// iff v == 0, setting the sign-bit, which gets returned.
+	return int((u32(v)-1) >> 31)
+}
+
+// rand_bytes fills the dst buffer with cryptographic entropy taken from
+// the system entropy source.  This routine will block if the system entropy
+// source is not ready yet.  All system entropy source failures are treated
+// as catastrophic, resulting in a panic.
+rand_bytes :: proc (dst: []byte) {
+	// zero-fill the buffer first
+	mem.zero_explicit(raw_data(dst), len(dst))
+
+	_rand_bytes(dst)
+}

core/crypto/gost/gost.odin

@@ -0,0 +1,360 @@
+package gost
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the GOST hashing algorithm, as defined in RFC 5831 <https://datatracker.ietf.org/doc/html/rfc5831>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [32]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Gost_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Gost_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc "contextless" (ctx: ^Gost_Context) {
+    sbox: [8][16]u32 = {
+        { 10, 4,  5,  6,  8,  1,  3,  7,  13, 12, 14, 0,  9,  2,  11, 15 },
+        { 5,  15, 4,  0,  2,  13, 11, 9,  1,  7,  6,  3,  12, 14, 10, 8  },
+        { 7,  15, 12, 14, 9,  4,  1,  0,  3,  11, 5,  2,  6,  10, 8,  13 },
+        { 4,  10, 7,  12, 0,  15, 2,  8,  14, 1,  6,  5,  13, 11, 9,  3  },
+        { 7,  6,  4,  11, 9,  12, 2,  10, 1,  8,  0,  14, 15, 13, 3,  5  },
+        { 7,  6,  2,  4,  13, 9,  15, 0,  10, 1,  5,  11, 8,  14, 12, 3  },
+        { 13, 14, 4,  1,  7,  0,  5,  10, 3,  12, 8,  15, 6,  2,  9,  11 },
+        { 1,  3,  10, 9,  5,  11, 4,  15, 8,  6,  7,  14, 13, 0,  2,  12 },
+    }
+
+    i := 0
+    for a := 0; a < 16; a += 1 {
+        ax := sbox[1][a] << 15
+        bx := sbox[3][a] << 23
+        cx := sbox[5][a]
+        cx = (cx >> 1) | (cx << 31)
+        dx := sbox[7][a] << 7
+        for b := 0; b < 16; b, i = b + 1, i + 1 {
+            SBOX_1[i] = ax | (sbox[0][b] << 11)
+            SBOX_2[i] = bx | (sbox[2][b] << 19)
+            SBOX_3[i] = cx | (sbox[4][b] << 27)
+            SBOX_4[i] = dx | (sbox[6][b] << 3)
+        }
+    }
+}
+
+update :: proc(ctx: ^Gost_Context, data: []byte) {
+    length := byte(len(data))
+    j: byte
+
+    i := ctx.partial_bytes
+    for i < 32 && j < length {
+        ctx.partial[i] = data[j]
+        i, j = i + 1, j + 1
+    }
+
+    if i < 32 {
+        ctx.partial_bytes = i
+        return
+    }
+    bytes(ctx, ctx.partial[:], 256)
+
+    for (j + 32) < length {
+        bytes(ctx, data[j:], 256)
+        j += 32
+    }
+
+    i = 0
+    for j < length {
+        ctx.partial[i] = data[j]
+        i, j = i + 1, j + 1
+    }
+    ctx.partial_bytes = i
+}
+
+final :: proc(ctx: ^Gost_Context, hash: []byte) {
+    if ctx.partial_bytes > 0 {
+        mem.set(&ctx.partial[ctx.partial_bytes], 0, 32 - int(ctx.partial_bytes))
+        bytes(ctx, ctx.partial[:], u32(ctx.partial_bytes) << 3)
+    }
+  
+    compress(ctx.hash[:], ctx.len[:])
+    compress(ctx.hash[:], ctx.sum[:])
+
+    for i, j := 0, 0; i < 8; i, j = i + 1, j + 4 {
+        hash[j]     = byte(ctx.hash[i])
+        hash[j + 1] = byte(ctx.hash[i] >> 8)
+        hash[j + 2] = byte(ctx.hash[i] >> 16)
+        hash[j + 3] = byte(ctx.hash[i] >> 24)
+    }
+}
+
+/*
+    GOST implementation
+*/
+
+Gost_Context :: struct {
+    sum:           [8]u32,
+    hash:          [8]u32,
+    len:           [8]u32,
+    partial:       [32]byte,
+    partial_bytes: byte,
+}
+
+SBOX_1: [256]u32
+SBOX_2: [256]u32
+SBOX_3: [256]u32
+SBOX_4: [256]u32
+
+ENCRYPT_ROUND :: #force_inline proc "contextless" (l, r, t, k1, k2: u32) -> (u32, u32, u32) {
+    l, r, t := l, r, t
+    t  = (k1) + r
+    l ~= SBOX_1[t & 0xff] ~ SBOX_2[(t >> 8) & 0xff] ~ SBOX_3[(t >> 16) & 0xff] ~ SBOX_4[t >> 24]
+    t  = (k2) + l
+    r ~= SBOX_1[t & 0xff] ~ SBOX_2[(t >> 8) & 0xff] ~ SBOX_3[(t >> 16) & 0xff] ~ SBOX_4[t >> 24]
+    return l, r, t
+}
+
+ENCRYPT :: #force_inline proc "contextless" (a, b, c: u32, key: []u32) -> (l, r, t: u32) {
+    l, r, t = ENCRYPT_ROUND(a, b, c, key[0], key[1])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[2], key[3])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[4], key[5])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[6], key[7])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[0], key[1])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[2], key[3])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[4], key[5])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[6], key[7])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[0], key[1])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[2], key[3])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[4], key[5])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[6], key[7])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[7], key[6])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[5], key[4])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[3], key[2])
+    l, r, t = ENCRYPT_ROUND(l, r, t, key[1], key[0])
+    t = r
+    r = l
+    l = t
+    return
+}
+
+bytes :: proc(ctx: ^Gost_Context, buf: []byte, bits: u32) {
+    a, c: u32
+    m: [8]u32
+
+    for i, j := 0, 0; i < 8; i += 1 {
+        a = u32(buf[j]) | u32(buf[j + 1]) << 8 | u32(buf[j + 2]) << 16 | u32(buf[j + 3]) << 24
+        j += 4
+        m[i] = a
+        c = a + c + ctx.sum[i]
+        ctx.sum[i] = c
+        c = c < a ? 1 : 0
+    }
+
+    compress(ctx.hash[:], m[:])
+    ctx.len[0] += bits
+    if ctx.len[0] < bits {
+        ctx.len[1] += 1
+    }
+}
+
+compress :: proc(h, m: []u32) {
+    key, u, v, w, s: [8]u32
+
+    copy(u[:], h)
+    copy(v[:], m)
+
+    for i := 0; i < 8; i += 2 {
+        w[0] = u[0] ~ v[0]
+        w[1] = u[1] ~ v[1]
+        w[2] = u[2] ~ v[2]
+        w[3] = u[3] ~ v[3]
+        w[4] = u[4] ~ v[4]
+        w[5] = u[5] ~ v[5]
+        w[6] = u[6] ~ v[6]
+        w[7] = u[7] ~ v[7]
+
+        key[0] = (w[0] & 0x000000ff)       | (w[2] & 0x000000ff) <<  8 | (w[4] & 0x000000ff) << 16 | (w[6] & 0x000000ff) << 24
+        key[1] = (w[0] & 0x0000ff00) >>  8 | (w[2] & 0x0000ff00)       | (w[4] & 0x0000ff00) <<  8 | (w[6] & 0x0000ff00) << 16
+        key[2] = (w[0] & 0x00ff0000) >> 16 | (w[2] & 0x00ff0000) >>  8 | (w[4] & 0x00ff0000)       | (w[6] & 0x00ff0000) <<  8
+        key[3] = (w[0] & 0xff000000) >> 24 | (w[2] & 0xff000000) >> 16 | (w[4] & 0xff000000) >>  8 | (w[6] & 0xff000000)
+        key[4] = (w[1] & 0x000000ff)       | (w[3] & 0x000000ff) <<  8 | (w[5] & 0x000000ff) << 16 | (w[7] & 0x000000ff) << 24
+        key[5] = (w[1] & 0x0000ff00) >>  8 | (w[3] & 0x0000ff00)       | (w[5] & 0x0000ff00) <<  8 | (w[7] & 0x0000ff00) << 16
+        key[6] = (w[1] & 0x00ff0000) >> 16 | (w[3] & 0x00ff0000) >>  8 | (w[5] & 0x00ff0000)       | (w[7] & 0x00ff0000) <<  8
+        key[7] = (w[1] & 0xff000000) >> 24 | (w[3] & 0xff000000) >> 16 | (w[5] & 0xff000000) >>  8 | (w[7] & 0xff000000)
+
+        r := h[i]
+        l := h[i + 1]
+        t: u32
+        l, r, t = ENCRYPT(l, r, 0, key[:])
+
+        s[i] = r
+        s[i + 1] = l
+
+        if i == 6 {
+            break
+        }
+
+        l    = u[0] ~ u[2]
+        r    = u[1] ~ u[3]
+        u[0] = u[2]
+        u[1] = u[3]
+        u[2] = u[4]
+        u[3] = u[5]
+        u[4] = u[6]
+        u[5] = u[7]
+        u[6] = l
+        u[7] = r
+
+        if i == 2 {
+            u[0] ~= 0xff00ff00
+            u[1] ~= 0xff00ff00
+            u[2] ~= 0x00ff00ff
+            u[3] ~= 0x00ff00ff
+            u[4] ~= 0x00ffff00
+            u[5] ~= 0xff0000ff
+            u[6] ~= 0x000000ff
+            u[7] ~= 0xff00ffff
+        }
+
+        l    = v[0]
+        r    = v[2]
+        v[0] = v[4]
+        v[2] = v[6]
+        v[4] = l ~ r
+        v[6] = v[0] ~ r
+        l    = v[1]
+        r    = v[3]
+        v[1] = v[5]
+        v[3] = v[7]
+        v[5] = l ~ r
+        v[7] = v[1] ~ r
+    }
+
+    u[0] = m[0] ~ s[6]
+    u[1] = m[1] ~ s[7]
+    u[2] = m[2] ~ (s[0] << 16) ~ (s[0] >> 16) ~ (s[0] & 0xffff) ~ 
+        (s[1] & 0xffff) ~ (s[1] >> 16) ~ (s[2] << 16) ~ s[6] ~ (s[6] << 16) ~
+        (s[7] & 0xffff0000) ~ (s[7] >> 16)
+    u[3] = m[3] ~ (s[0] & 0xffff) ~ (s[0] << 16) ~ (s[1] & 0xffff) ~
+        (s[1] << 16) ~ (s[1] >> 16) ~ (s[2] << 16) ~ (s[2] >> 16) ~
+        (s[3] << 16) ~ s[6] ~ (s[6] << 16) ~ (s[6] >> 16) ~ (s[7] & 0xffff) ~
+        (s[7] << 16) ~ (s[7] >> 16)
+    u[4] = m[4] ~
+        (s[0] & 0xffff0000) ~ (s[0] << 16) ~ (s[0] >> 16) ~
+        (s[1] & 0xffff0000) ~ (s[1] >> 16) ~ (s[2] << 16) ~ (s[2] >> 16) ~
+        (s[3] << 16) ~ (s[3] >> 16) ~ (s[4] << 16) ~ (s[6] << 16) ~
+        (s[6] >> 16) ~(s[7] & 0xffff) ~ (s[7] << 16) ~ (s[7] >> 16)
+    u[5] = m[5] ~ (s[0] << 16) ~ (s[0] >> 16) ~ (s[0] & 0xffff0000) ~
+        (s[1] & 0xffff) ~ s[2] ~ (s[2] >> 16) ~ (s[3] << 16) ~ (s[3] >> 16) ~
+        (s[4] << 16) ~ (s[4] >> 16) ~ (s[5] << 16) ~  (s[6] << 16) ~
+        (s[6] >> 16) ~ (s[7] & 0xffff0000) ~ (s[7] << 16) ~ (s[7] >> 16)
+    u[6] = m[6] ~ s[0] ~ (s[1] >> 16) ~ (s[2] << 16) ~ s[3] ~ (s[3] >> 16) ~
+        (s[4] << 16) ~ (s[4] >> 16) ~ (s[5] << 16) ~ (s[5] >> 16) ~ s[6] ~
+        (s[6] << 16) ~ (s[6] >> 16) ~ (s[7] << 16)
+    u[7] = m[7] ~ (s[0] & 0xffff0000) ~ (s[0] << 16) ~ (s[1] & 0xffff) ~
+        (s[1] << 16) ~ (s[2] >> 16) ~ (s[3] << 16) ~ s[4] ~ (s[4] >> 16) ~
+        (s[5] << 16) ~ (s[5] >> 16) ~ (s[6] >> 16) ~ (s[7] & 0xffff) ~
+        (s[7] << 16) ~ (s[7] >> 16)
+
+    v[0] = h[0] ~ (u[1] << 16) ~ (u[0] >> 16)
+    v[1] = h[1] ~ (u[2] << 16) ~ (u[1] >> 16)
+    v[2] = h[2] ~ (u[3] << 16) ~ (u[2] >> 16)
+    v[3] = h[3] ~ (u[4] << 16) ~ (u[3] >> 16)
+    v[4] = h[4] ~ (u[5] << 16) ~ (u[4] >> 16)
+    v[5] = h[5] ~ (u[6] << 16) ~ (u[5] >> 16)
+    v[6] = h[6] ~ (u[7] << 16) ~ (u[6] >> 16)
+    v[7] = h[7] ~ (u[0] & 0xffff0000) ~ (u[0] << 16) ~ (u[7] >> 16) ~ (u[1] & 0xffff0000) ~ (u[1] << 16) ~ (u[6] << 16) ~ (u[7] & 0xffff0000)
+
+    h[0] = (v[0] & 0xffff0000) ~ (v[0] << 16) ~ (v[0] >> 16) ~ (v[1] >> 16) ~
+        (v[1] & 0xffff0000) ~ (v[2] << 16) ~ (v[3] >> 16) ~ (v[4] << 16) ~
+        (v[5] >> 16) ~ v[5] ~ (v[6] >> 16) ~ (v[7] << 16) ~ (v[7] >> 16) ~
+        (v[7] & 0xffff)
+    h[1] = (v[0] << 16) ~ (v[0] >> 16) ~ (v[0] & 0xffff0000) ~ (v[1] & 0xffff) ~
+        v[2] ~ (v[2] >> 16) ~ (v[3] << 16) ~ (v[4] >> 16) ~ (v[5] << 16) ~
+        (v[6] << 16) ~ v[6] ~ (v[7] & 0xffff0000) ~ (v[7] >> 16)
+    h[2] = (v[0] & 0xffff) ~ (v[0] << 16) ~ (v[1] << 16) ~ (v[1] >> 16) ~
+        (v[1] & 0xffff0000) ~ (v[2] << 16) ~ (v[3] >> 16) ~ v[3] ~ (v[4] << 16) ~
+        (v[5] >> 16) ~ v[6] ~ (v[6] >> 16) ~ (v[7] & 0xffff) ~ (v[7] << 16) ~
+        (v[7] >> 16)
+    h[3] = (v[0] << 16) ~ (v[0] >> 16) ~ (v[0] & 0xffff0000) ~
+        (v[1] & 0xffff0000) ~ (v[1] >> 16) ~ (v[2] << 16) ~ (v[2] >> 16) ~ v[2] ~
+        (v[3] << 16) ~ (v[4] >> 16) ~ v[4] ~ (v[5] << 16) ~ (v[6] << 16) ~
+        (v[7] & 0xffff) ~ (v[7] >> 16)
+    h[4] = (v[0] >> 16) ~ (v[1] << 16) ~ v[1] ~ (v[2] >> 16) ~ v[2] ~
+        (v[3] << 16) ~ (v[3] >> 16) ~ v[3] ~ (v[4] << 16) ~ (v[5] >> 16) ~
+        v[5] ~ (v[6] << 16) ~ (v[6] >> 16) ~ (v[7] << 16)
+    h[5] = (v[0] << 16) ~ (v[0] & 0xffff0000) ~ (v[1] << 16) ~ (v[1] >> 16) ~
+        (v[1] & 0xffff0000) ~ (v[2] << 16) ~ v[2] ~ (v[3] >> 16) ~ v[3] ~
+        (v[4] << 16) ~ (v[4] >> 16) ~ v[4] ~ (v[5] << 16) ~ (v[6] << 16) ~
+        (v[6] >> 16) ~ v[6] ~ (v[7] << 16) ~ (v[7] >> 16) ~ (v[7] & 0xffff0000)
+    h[6] = v[0] ~ v[2] ~ (v[2] >> 16) ~ v[3] ~ (v[3] << 16) ~ v[4] ~
+        (v[4] >> 16) ~ (v[5] << 16) ~ (v[5] >> 16) ~ v[5] ~ (v[6] << 16) ~
+        (v[6] >> 16) ~ v[6] ~ (v[7] << 16) ~ v[7]
+    h[7] = v[0] ~ (v[0] >> 16) ~ (v[1] << 16) ~ (v[1] >> 16) ~ (v[2] << 16) ~
+        (v[3] >> 16) ~ v[3] ~ (v[4] << 16) ~ v[4] ~ (v[5] >> 16) ~ v[5] ~
+        (v[6] << 16) ~ (v[6] >> 16) ~ (v[7] << 16) ~ v[7]
+}

core/crypto/groestl/groestl.odin

@@ -0,0 +1,564 @@
+package groestl
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the GROESTL hashing algorithm, as defined in <http://www.groestl.info/Groestl.zip>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc(data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc(data: []byte) -> [28]byte {
+    hash: [28]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 224
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+    hash: [28]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 224
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 256
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 256
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc(data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc(data: []byte) -> [48]byte {
+    hash: [48]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 384
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+    hash: [48]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 384
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 512
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: Groestl_Context
+    ctx.hashbitlen = 512
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Groestl_Context) {
+    assert(ctx.hashbitlen == 224 || ctx.hashbitlen == 256 || ctx.hashbitlen == 384 || ctx.hashbitlen == 512, "hashbitlen must be set to 224, 256, 384 or 512")
+    if ctx.hashbitlen <= 256 {
+        ctx.rounds    = 10
+        ctx.columns   = 8
+        ctx.statesize = 64
+    } else {
+        ctx.rounds    = 14
+        ctx.columns   = 16
+        ctx.statesize = 128
+    }
+    for i := 8 - size_of(i32); i < 8; i += 1 {
+        ctx.chaining[i][ctx.columns - 1] = byte(ctx.hashbitlen >> (8 * (7 - uint(i))))
+    }
+}
+
+update :: proc(ctx: ^Groestl_Context, data: []byte) {
+    databitlen := len(data) * 8
+    msglen     := databitlen / 8
+    rem        := databitlen % 8
+
+    i: int
+    assert(ctx.bits_in_last_byte == 0)
+
+    if ctx.buf_ptr != 0 {
+        for i = 0; ctx.buf_ptr < ctx.statesize && i < msglen; i, ctx.buf_ptr =  i + 1, ctx.buf_ptr + 1 {
+            ctx.buffer[ctx.buf_ptr] = data[i]
+        }
+
+        if ctx.buf_ptr < ctx.statesize {
+            if rem != 0 {
+                ctx.bits_in_last_byte    = rem
+                ctx.buffer[ctx.buf_ptr]  = data[i]
+                ctx.buf_ptr             += 1
+            }
+            return
+        }
+
+        ctx.buf_ptr = 0
+        transform(ctx, ctx.buffer[:], u32(ctx.statesize))
+    }
+
+    transform(ctx, data[i:], u32(msglen - i))
+    i += ((msglen - i) / ctx.statesize) * ctx.statesize
+    for i < msglen {
+        ctx.buffer[ctx.buf_ptr] = data[i]
+        i, ctx.buf_ptr          = i + 1, ctx.buf_ptr + 1
+    }
+    
+    if rem != 0 {
+        ctx.bits_in_last_byte    = rem
+        ctx.buffer[ctx.buf_ptr]  = data[i]
+        ctx.buf_ptr             += 1
+    }
+}
+
+final :: proc(ctx: ^Groestl_Context, hash: []byte) {
+    hashbytelen := ctx.hashbitlen / 8
+
+    if ctx.bits_in_last_byte != 0 {
+        ctx.buffer[ctx.buf_ptr - 1] &= ((1 << uint(ctx.bits_in_last_byte)) - 1) << (8 - uint(ctx.bits_in_last_byte))
+        ctx.buffer[ctx.buf_ptr - 1] ~= 0x1 << (7 - uint(ctx.bits_in_last_byte))
+    } else {
+        ctx.buffer[ctx.buf_ptr]  = 0x80
+        ctx.buf_ptr             += 1
+    }
+
+    if ctx.buf_ptr > ctx.statesize - 8 {
+        for ctx.buf_ptr < ctx.statesize {
+            ctx.buffer[ctx.buf_ptr]  = 0
+            ctx.buf_ptr             += 1
+        }
+        transform(ctx, ctx.buffer[:], u32(ctx.statesize))
+        ctx.buf_ptr = 0
+    }
+
+    for ctx.buf_ptr < ctx.statesize - 8 {
+        ctx.buffer[ctx.buf_ptr]  = 0
+        ctx.buf_ptr             += 1
+    }
+
+    ctx.block_counter += 1
+    ctx.buf_ptr        = ctx.statesize
+
+    for ctx.buf_ptr > ctx.statesize - 8 {
+        ctx.buf_ptr              -= 1
+        ctx.buffer[ctx.buf_ptr]   = byte(ctx.block_counter)
+        ctx.block_counter       >>= 8
+    }
+
+    transform(ctx, ctx.buffer[:], u32(ctx.statesize))
+    output_transformation(ctx)
+
+    for i, j := ctx.statesize - hashbytelen , 0; i < ctx.statesize; i, j = i + 1, j + 1 {
+        hash[j] = ctx.chaining[i % 8][i / 8]
+    }
+}
+
+/*
+    GROESTL implementation
+*/
+
+SBOX := [256]byte {
+    0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5,
+    0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
+    0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0,
+    0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
+    0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc,
+    0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
+    0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a,
+    0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
+    0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0,
+    0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
+    0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b,
+    0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
+    0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85,
+    0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
+    0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5,
+    0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
+    0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17,
+    0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
+    0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88,
+    0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
+    0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c,
+    0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
+    0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9,
+    0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
+    0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6,
+    0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
+    0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e,
+    0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
+    0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94,
+    0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
+    0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68,
+    0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
+}
+
+SHIFT := [2][2][8]int {
+    {{0, 1, 2, 3, 4, 5, 6, 7},  {1, 3, 5, 7,  0, 2, 4, 6}},
+    {{0, 1, 2, 3, 4, 5, 6, 11}, {1, 3, 5, 11, 0, 2, 4, 6}},
+}
+
+Groestl_Context :: struct {
+    chaining:          [8][16]byte,
+    block_counter:     u64,
+    hashbitlen:        int,
+    buffer:            [128]byte,
+    buf_ptr:           int,
+    bits_in_last_byte: int,
+    columns:           int,
+    rounds:            int,
+    statesize:         int,
+}
+
+Groestl_Variant :: enum {
+    P512  = 0, 
+    Q512  = 1, 
+    P1024 = 2, 
+    Q1024 = 3,
+}
+
+MUL2 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return (b >> 7) != 0 ? (b << 1) ~ 0x1b : (b << 1)
+}
+
+MUL3 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return MUL2(b) ~ b
+}
+
+MUL4 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return MUL2(MUL2(b))
+}
+
+MUL5 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return MUL4(b) ~ b
+}
+
+MUL6 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return MUL4(b) ~ MUL2(b)
+}
+
+MUL7 :: #force_inline proc "contextless"(b: byte) -> byte {
+    return MUL4(b) ~ MUL2(b) ~ b
+}
+
+sub_bytes :: #force_inline proc (x: [][16]byte, columns: int) {
+    for i := 0; i < 8; i += 1 {
+        for j := 0; j < columns; j += 1 {
+            x[i][j] = SBOX[x[i][j]]
+        }
+    }
+}
+
+shift_bytes :: #force_inline proc (x: [][16]byte, columns: int, v: Groestl_Variant) {
+    temp: [16]byte
+    R := &SHIFT[int(v) / 2][int(v) & 1]
+
+    for i := 0; i < 8; i += 1 {
+        for j := 0; j < columns; j += 1 {
+            temp[j] = x[i][(j + R[i]) % columns]
+        }
+        for j := 0; j < columns; j += 1 {
+            x[i][j] = temp[j]
+        }
+    }
+}
+
+mix_bytes :: #force_inline proc (x: [][16]byte, columns: int) {
+    temp: [8]byte
+
+    for i := 0; i < columns; i += 1 {
+        for j := 0; j < 8; j += 1 {
+            temp[j] =  MUL2(x[(j + 0) % 8][i]) ~
+                       MUL2(x[(j + 1) % 8][i]) ~
+                       MUL3(x[(j + 2) % 8][i]) ~
+                       MUL4(x[(j + 3) % 8][i]) ~
+                       MUL5(x[(j + 4) % 8][i]) ~
+                       MUL3(x[(j + 5) % 8][i]) ~
+                       MUL5(x[(j + 6) % 8][i]) ~
+                       MUL7(x[(j + 7) % 8][i])
+        }
+        for j := 0; j < 8; j += 1 {
+            x[j][i] = temp[j]
+        }
+    }
+}
+
+p :: #force_inline proc (ctx: ^Groestl_Context, x: [][16]byte) {
+    v := ctx.columns == 8 ? Groestl_Variant.P512 : Groestl_Variant.P1024
+    for i := 0; i < ctx.rounds; i += 1 {
+        add_roundconstant(x, ctx.columns, byte(i), v)
+        sub_bytes(x, ctx.columns)
+        shift_bytes(x, ctx.columns, v)
+        mix_bytes(x, ctx.columns)
+    }
+}
+
+q :: #force_inline proc (ctx: ^Groestl_Context, x: [][16]byte) {
+    v := ctx.columns == 8 ? Groestl_Variant.Q512 : Groestl_Variant.Q1024
+    for i := 0; i < ctx.rounds; i += 1 {
+        add_roundconstant(x, ctx.columns, byte(i), v)
+        sub_bytes(x, ctx.columns)
+        shift_bytes(x, ctx.columns, v)
+        mix_bytes(x, ctx.columns)
+    }
+}
+
+transform :: proc(ctx: ^Groestl_Context, input: []byte, msglen: u32) {
+    tmp1, tmp2: [8][16]byte
+    input, msglen := input, msglen
+
+    for msglen >= u32(ctx.statesize) {
+        for i := 0; i < 8; i += 1 {
+            for j := 0; j < ctx.columns; j += 1 {
+                tmp1[i][j] = ctx.chaining[i][j] ~ input[j * 8 + i]
+                tmp2[i][j] = input[j * 8 + i]
+            }
+        }
+
+        p(ctx, tmp1[:])
+        q(ctx, tmp2[:])
+
+        for i := 0; i < 8; i += 1 {
+            for j := 0; j < ctx.columns; j += 1 {
+                ctx.chaining[i][j] ~= tmp1[i][j] ~ tmp2[i][j]
+            }
+        }
+
+        ctx.block_counter += 1
+        msglen            -= u32(ctx.statesize)
+        input              = input[ctx.statesize:]
+    }
+}
+
+output_transformation :: proc(ctx: ^Groestl_Context) {
+    temp: [8][16]byte
+
+    for i := 0; i < 8; i += 1 {
+        for j := 0; j < ctx.columns; j += 1 {
+            temp[i][j] = ctx.chaining[i][j]
+        }
+    }
+
+    p(ctx, temp[:])
+
+    for i := 0; i < 8; i += 1 {
+        for j := 0; j < ctx.columns; j += 1 {
+            ctx.chaining[i][j] ~= temp[i][j]
+        }
+    }
+}
+
+add_roundconstant :: proc(x: [][16]byte, columns: int, round: byte, v: Groestl_Variant) {
+    switch (i32(v) & 1) {
+        case 0: 
+            for i := 0; i < columns; i += 1 {
+                x[0][i] ~= byte(i << 4) ~ round
+            }
+        case 1:
+            for i := 0; i < columns; i += 1 {
+                for j := 0; j < 7; j += 1 {
+                    x[j][i] ~= 0xff
+                }
+            }
+            for i := 0; i < columns; i += 1 {
+                x[7][i] ~= byte(i << 4) ~ 0xff ~ round
+            }
+    }
+}

core/crypto/jh/jh.odin

@@ -0,0 +1,495 @@
+package jh
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the JH hashing algorithm, as defined in <https://www3.ntu.edu.sg/home/wuhj/research/jh/index.html>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc(data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc(data: []byte) -> [28]byte {
+    hash: [28]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 224
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+    hash: [28]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 224
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 256
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 256
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc(data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc(data: []byte) -> [48]byte {
+    hash: [48]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 384
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+    hash: [48]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 384
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 512
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: Jh_Context
+    ctx.hashbitlen = 512
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Jh_Context) {
+    assert(ctx.hashbitlen == 224 || ctx.hashbitlen == 256 || ctx.hashbitlen == 384 || ctx.hashbitlen == 512, "hashbitlen must be set to 224, 256, 384 or 512")
+    ctx.H[1] = byte(ctx.hashbitlen)      & 0xff
+    ctx.H[0] = byte(ctx.hashbitlen >> 8) & 0xff
+    F8(ctx)
+}
+
+update :: proc(ctx: ^Jh_Context, data: []byte) {
+    databitlen     := u64(len(data)) * 8
+    ctx.databitlen += databitlen
+    i              := u64(0)
+
+    if (ctx.buffer_size > 0) && ((ctx.buffer_size + databitlen) < 512) {
+        if (databitlen & 7) == 0 {
+            copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3)])
+        } else {
+            copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3) + 1])
+        } 
+        ctx.buffer_size += databitlen
+        databitlen = 0
+    }
+
+    if (ctx.buffer_size > 0 ) && ((ctx.buffer_size + databitlen) >= 512) {
+        copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3)])
+        i      = 64 - (ctx.buffer_size >> 3)
+        databitlen = databitlen - (512 - ctx.buffer_size)
+        F8(ctx)
+        ctx.buffer_size = 0
+    }
+
+    for databitlen >= 512 {
+        copy(ctx.buffer[:], data[i:i + 64])
+        F8(ctx)
+        i += 64
+        databitlen -= 512
+    }
+
+    if databitlen > 0 {
+        if (databitlen & 7) == 0 {
+            copy(ctx.buffer[:], data[i:i + ((databitlen & 0x1ff) >> 3)])
+        } else {
+            copy(ctx.buffer[:], data[i:i + ((databitlen & 0x1ff) >> 3) + 1])
+        }
+        ctx.buffer_size = databitlen
+    }
+}
+
+final :: proc(ctx: ^Jh_Context, hash: []byte) {
+    if ctx.databitlen & 0x1ff == 0 {
+        for i := 0; i < 64; i += 1 {
+            ctx.buffer[i] = 0
+        }
+        ctx.buffer[0]  = 0x80
+        ctx.buffer[63] = byte(ctx.databitlen)       & 0xff
+        ctx.buffer[62] = byte(ctx.databitlen >> 8)  & 0xff
+        ctx.buffer[61] = byte(ctx.databitlen >> 16) & 0xff
+        ctx.buffer[60] = byte(ctx.databitlen >> 24) & 0xff
+        ctx.buffer[59] = byte(ctx.databitlen >> 32) & 0xff
+        ctx.buffer[58] = byte(ctx.databitlen >> 40) & 0xff
+        ctx.buffer[57] = byte(ctx.databitlen >> 48) & 0xff
+        ctx.buffer[56] = byte(ctx.databitlen >> 56) & 0xff
+        F8(ctx)
+    } else {
+        if ctx.buffer_size & 7 == 0 {
+            for i := (ctx.databitlen & 0x1ff) >> 3; i < 64; i += 1 {
+                ctx.buffer[i] = 0
+            }
+        } else {
+            for i := ((ctx.databitlen & 0x1ff) >> 3) + 1; i < 64; i += 1 {
+                ctx.buffer[i] = 0
+            }
+        }
+        ctx.buffer[(ctx.databitlen & 0x1ff) >> 3] |= 1 << (7 - (ctx.databitlen & 7))
+        F8(ctx)
+        for i := 0; i < 64; i += 1 {
+            ctx.buffer[i] = 0
+        }
+        ctx.buffer[63] = byte(ctx.databitlen)       & 0xff
+        ctx.buffer[62] = byte(ctx.databitlen >> 8)  & 0xff
+        ctx.buffer[61] = byte(ctx.databitlen >> 16) & 0xff
+        ctx.buffer[60] = byte(ctx.databitlen >> 24) & 0xff
+        ctx.buffer[59] = byte(ctx.databitlen >> 32) & 0xff
+        ctx.buffer[58] = byte(ctx.databitlen >> 40) & 0xff
+        ctx.buffer[57] = byte(ctx.databitlen >> 48) & 0xff
+        ctx.buffer[56] = byte(ctx.databitlen >> 56) & 0xff
+        F8(ctx)
+    }
+    switch ctx.hashbitlen {
+        case 224: copy(hash[:], ctx.H[100:128])
+        case 256: copy(hash[:], ctx.H[96:128])
+        case 384: copy(hash[:], ctx.H[80:128])
+        case 512: copy(hash[:], ctx.H[64:128])
+    }
+}
+
+/*
+    JH implementation
+*/
+
+ROUNDCONSTANT_ZERO := [64]byte {
+    0x6, 0xa, 0x0, 0x9, 0xe, 0x6, 0x6, 0x7,
+    0xf, 0x3, 0xb, 0xc, 0xc, 0x9, 0x0, 0x8,
+    0xb, 0x2, 0xf, 0xb, 0x1, 0x3, 0x6, 0x6,
+    0xe, 0xa, 0x9, 0x5, 0x7, 0xd, 0x3, 0xe,
+    0x3, 0xa, 0xd, 0xe, 0xc, 0x1, 0x7, 0x5,
+    0x1, 0x2, 0x7, 0x7, 0x5, 0x0, 0x9, 0x9,
+    0xd, 0xa, 0x2, 0xf, 0x5, 0x9, 0x0, 0xb,
+    0x0, 0x6, 0x6, 0x7, 0x3, 0x2, 0x2, 0xa,
+}
+
+SBOX := [2][16]byte {
+    {9, 0,  4, 11, 13, 12, 3, 15, 1,  10, 2, 6, 7,  5,  8,  14},
+    {3, 12, 6, 13, 5,  7,  1, 9,  15, 2,  0, 4, 11, 10, 14, 8},
+}
+
+Jh_Context :: struct {
+    hashbitlen:    int,
+    databitlen:    u64,
+    buffer_size:   u64,
+    H:             [128]byte,
+    A:             [256]byte,
+    roundconstant: [64]byte,
+    buffer:        [64]byte,
+}
+
+E8_finaldegroup :: proc(ctx: ^Jh_Context) {
+    t0,t1,t2,t3: byte
+    tem: [256]byte
+    for i := 0; i < 128; i += 1 {
+        tem[i]       = ctx.A[i << 1]
+        tem[i + 128] = ctx.A[(i << 1) + 1]
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.H[i] = 0
+    }
+    for i := 0; i < 256; i += 1 {
+        t0 = (tem[i] >> 3) & 1
+        t1 = (tem[i] >> 2) & 1
+        t2 = (tem[i] >> 1) & 1
+        t3 = (tem[i] >> 0) & 1
+
+        ctx.H[uint(i) >> 3]         |= t0 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 256) >> 3] |= t1 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 512) >> 3] |= t2 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 768) >> 3] |= t3 << (7 - (uint(i) & 7))
+    }
+}
+
+update_roundconstant :: proc(ctx: ^Jh_Context) {
+    tem: [64]byte
+    t: byte
+    for i := 0; i < 64; i += 1 {
+        tem[i] = SBOX[0][ctx.roundconstant[i]]
+    }
+    for i := 0; i < 64; i += 2 {
+        tem[i + 1] ~= ((tem[i]   << 1)   ~ (tem[i]   >> 3)   ~ ((tem[i]   >> 2) & 2))   & 0xf
+        tem[i]     ~= ((tem[i + 1] << 1) ~ (tem[i + 1] >> 3) ~ ((tem[i + 1] >> 2) & 2)) & 0xf
+    }
+    for i := 0; i < 64; i += 4 {
+        t          = tem[i + 2]
+        tem[i + 2] = tem[i + 3]
+        tem[i + 3] = t
+    }
+    for i := 0; i < 32; i += 1 {
+        ctx.roundconstant[i]      = tem[i << 1]
+        ctx.roundconstant[i + 32] = tem[(i << 1) + 1]
+    }
+    for i := 32; i < 64; i += 2 {
+        t                        = ctx.roundconstant[i]
+        ctx.roundconstant[i]     = ctx.roundconstant[i + 1]
+        ctx.roundconstant[i + 1] = t
+    }
+}
+
+R8 :: proc(ctx: ^Jh_Context) {
+    t: byte
+    tem, roundconstant_expanded: [256]byte
+    for i := u32(0); i < 256; i += 1 {
+        roundconstant_expanded[i] = (ctx.roundconstant[i >> 2] >> (3 - (i & 3)) ) & 1
+    }
+    for i := 0; i < 256; i += 1 {
+        tem[i] = SBOX[roundconstant_expanded[i]][ctx.A[i]]
+    }
+    for i := 0; i < 256; i += 2 {
+        tem[i+1] ~= ((tem[i]   << 1)   ~ (tem[i]   >> 3)   ~ ((tem[i]   >> 2) & 2))   & 0xf
+        tem[i]   ~= ((tem[i + 1] << 1) ~ (tem[i + 1] >> 3) ~ ((tem[i + 1] >> 2) & 2)) & 0xf
+    }
+    for i := 0; i < 256; i += 4 {
+        t        = tem[i + 2]
+        tem[i+2] = tem[i + 3]
+        tem[i+3] = t
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.A[i]       = tem[i << 1]
+        ctx.A[i + 128] = tem[(i << 1) + 1]
+    }
+    for i := 128; i < 256; i += 2 {
+        t            = ctx.A[i]
+        ctx.A[i]     = ctx.A[i + 1]
+        ctx.A[i + 1] = t
+    }
+}
+
+E8_initialgroup :: proc(ctx: ^Jh_Context) {
+    t0, t1, t2, t3: byte
+    tem:            [256]byte
+    for i := u32(0); i < 256; i += 1 {
+        t0     = (ctx.H[i >> 3]   >> (7 - (i & 7)))       & 1
+        t1     = (ctx.H[(i + 256) >> 3] >> (7 - (i & 7))) & 1
+        t2     = (ctx.H[(i + 512) >> 3] >> (7 - (i & 7))) & 1
+        t3     = (ctx.H[(i + 768) >> 3] >> (7 - (i & 7))) & 1
+        tem[i] = (t0 << 3) | (t1 << 2) | (t2 << 1) | (t3 << 0)
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.A[i << 1]       = tem[i]
+        ctx.A[(i << 1) + 1] = tem[i + 128]
+    }
+}
+
+E8 :: proc(ctx: ^Jh_Context) {
+    for i := 0; i < 64; i += 1 {
+        ctx.roundconstant[i] = ROUNDCONSTANT_ZERO[i]
+    }
+    E8_initialgroup(ctx)
+    for i := 0; i < 42; i += 1 {
+        R8(ctx)
+        update_roundconstant(ctx)
+    }
+    E8_finaldegroup(ctx)
+}
+
+F8 :: proc(ctx: ^Jh_Context) {
+    for i := 0; i < 64; i += 1 {
+        ctx.H[i] ~= ctx.buffer[i]
+    }
+    E8(ctx)
+    for i := 0; i < 64; i += 1 {
+        ctx.H[i + 64] ~= ctx.buffer[i]
+    }
+}

core/crypto/keccak/keccak.odin

@@ -0,0 +1,281 @@
+package keccak
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the Keccak hashing algorithm.
+    This is done because the padding in the SHA3 standard was changed by the NIST, resulting in a different output.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_sha3"
+
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc(data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc(data: []byte) -> [28]byte {
+    hash: [28]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 28
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+    hash: [28]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 28
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc(data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc(data: []byte) -> [48]byte {
+    hash: [48]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 48
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+    hash: [48]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 48
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 64
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 64
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+Sha3_Context :: _sha3.Sha3_Context
+
+init :: proc(ctx: ^_sha3.Sha3_Context) {
+    ctx.is_keccak = true
+    _sha3.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_sha3.Sha3_Context, data: []byte) {
+    _sha3.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_sha3.Sha3_Context, hash: []byte) {
+    _sha3.final(ctx, hash)
+}

core/crypto/md2/md2.odin

@@ -0,0 +1,160 @@
+package md2
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the MD2 hashing algorithm, as defined in RFC 1319 <https://datatracker.ietf.org/doc/html/rfc1319>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [16]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [16]byte {
+	hash: [16]byte
+	ctx: Md2_Context
+    // init(&ctx) No-op
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([16]byte, bool) {
+	hash: [16]byte
+	ctx: Md2_Context
+	// init(&ctx) No-op
+	buf := make([]byte, 512)
+	defer delete(buf)
+	read := 1
+	for read > 0 {
+	    read, _ = s->impl_read(buf)
+	    if read > 0 {
+			update(&ctx, buf[:read])
+	    } 
+	}
+	final(&ctx, hash[:])
+	return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+	if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+@(warning="Init is a no-op for MD2")
+init :: proc(ctx: ^Md2_Context) {
+	// No action needed here
+}
+
+update :: proc(ctx: ^Md2_Context, data: []byte) {
+	for i := 0; i < len(data); i += 1 {
+		ctx.data[ctx.datalen] = data[i]
+		ctx.datalen += 1
+		if (ctx.datalen == 16) {
+			transform(ctx, ctx.data[:])
+			ctx.datalen = 0
+		}
+	}
+}
+
+final :: proc(ctx: ^Md2_Context, hash: []byte) {
+	to_pad := byte(16 - ctx.datalen)
+    for ctx.datalen < 16 {
+        ctx.data[ctx.datalen] = to_pad
+		ctx.datalen += 1
+    }
+	transform(ctx, ctx.data[:])
+	transform(ctx, ctx.checksum[:])
+    for i := 0; i < 16; i += 1 {
+        hash[i] = ctx.state[i]
+    }
+}
+
+/*
+    MD2 implementation
+*/
+
+Md2_Context :: struct {
+    data:     [16]byte,
+    state:    [16 * 3]byte,
+    checksum: [16]byte,
+    datalen:  int,
+}
+
+PI_TABLE := [?]byte {
+	41,  46,  67,  201, 162, 216, 124, 1,   61,  54,  84,  161, 236, 240, 6,
+	19,  98,  167, 5,   243, 192, 199, 115, 140, 152, 147, 43,  217, 188, 76,
+	130, 202, 30,  155, 87,  60,  253, 212, 224, 22,  103, 66,  111, 24,  138, 
+	23,  229, 18,  190, 78,  196, 214, 218, 158, 222, 73,  160, 251, 245, 142,
+	187, 47,  238, 122, 169, 104, 121, 145, 21,  178, 7,   63,  148, 194, 16,
+	137, 11,  34,  95,  33,  128, 127, 93,  154, 90,  144, 50,  39,  53,  62, 
+	204, 231, 191, 247, 151, 3,   255, 25,  48,  179, 72,  165, 181, 209, 215,
+	94,  146, 42,  172, 86,  170, 198, 79,  184, 56,  210, 150, 164, 125, 182,
+	118, 252, 107, 226, 156, 116, 4,   241, 69,  157, 112, 89,  100, 113, 135,
+	32,  134, 91,  207, 101, 230, 45,  168, 2,   27,  96,  37,  173, 174, 176,
+	185, 246, 28,  70,  97,  105, 52,  64,  126, 15,  85,  71,  163, 35,  221,
+	81,  175, 58,  195, 92,  249, 206, 186, 197, 234, 38,  44,  83,  13,  110,
+	133, 40,  132, 9,   211, 223, 205, 244, 65,  129, 77,  82,  106, 220, 55,
+	200, 108, 193, 171, 250, 36,  225, 123, 8,   12,  189, 177, 74,  120, 136,
+	149, 139, 227, 99,  232, 109, 233, 203, 213, 254, 59,  0,   29,  57,  242,
+	239, 183, 14,  102, 88,  208, 228, 166, 119, 114, 248, 235, 117, 75,  10,
+	49,  68,  80,  180, 143, 237, 31,  26,  219, 153, 141, 51,  159, 17,  131,
+	20,
+}
+
+transform :: proc(ctx: ^Md2_Context, data: []byte) {
+    j,k,t: byte
+	for j = 0; j < 16; j += 1 {
+		ctx.state[j + 16] = data[j]
+		ctx.state[j + 16 * 2] = (ctx.state[j + 16] ~ ctx.state[j])
+	}
+	t = 0
+	for j = 0; j < 16 + 2; j += 1 {
+		for k = 0; k < 16 * 3; k += 1 {
+			ctx.state[k] ~= PI_TABLE[t]
+			t = ctx.state[k]
+		}
+		t = (t + j) & 0xff
+	}
+	t = ctx.checksum[16 - 1]
+	for j = 0; j < 16; j += 1 {
+		ctx.checksum[j] ~= PI_TABLE[data[j] ~ t]
+		t = ctx.checksum[j]
+	}
+}

core/crypto/md4/md4.odin

@@ -0,0 +1,241 @@
+package md4
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+        Jeroen van Rijn: Context design to be able to change from Odin implementation to bindings.
+
+    Implementation of the MD4 hashing algorithm, as defined in RFC 1320 <https://datatracker.ietf.org/doc/html/rfc1320>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [16]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: Md4_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: Md4_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Md4_Context) {
+    ctx.state[0] = 0x67452301
+    ctx.state[1] = 0xefcdab89
+    ctx.state[2] = 0x98badcfe
+    ctx.state[3] = 0x10325476
+}
+
+update :: proc(ctx: ^Md4_Context, data: []byte) {
+    for i := 0; i < len(data); i += 1 {
+        ctx.data[ctx.datalen] = data[i]
+        ctx.datalen += 1
+        if(ctx.datalen == BLOCK_SIZE) {
+            transform(ctx, ctx.data[:])
+            ctx.bitlen += 512
+            ctx.datalen = 0
+        }
+    }
+}
+
+final :: proc(ctx: ^Md4_Context, hash: []byte) {
+    i := ctx.datalen
+    if ctx.datalen < 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < 56 {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+    } else if ctx.datalen >= 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < BLOCK_SIZE {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+        transform(ctx, ctx.data[:])
+        mem.set(&ctx.data, 0, 56)
+    }
+
+    ctx.bitlen  += u64(ctx.datalen * 8)
+    ctx.data[56] = byte(ctx.bitlen)
+    ctx.data[57] = byte(ctx.bitlen >> 8)
+    ctx.data[58] = byte(ctx.bitlen >> 16)
+    ctx.data[59] = byte(ctx.bitlen >> 24)
+    ctx.data[60] = byte(ctx.bitlen >> 32)
+    ctx.data[61] = byte(ctx.bitlen >> 40)
+    ctx.data[62] = byte(ctx.bitlen >> 48)
+    ctx.data[63] = byte(ctx.bitlen >> 56)
+    transform(ctx, ctx.data[:])
+
+    for i = 0; i < 4; i += 1 {
+        hash[i]      = byte(ctx.state[0] >> (i * 8)) & 0x000000ff
+        hash[i + 4]  = byte(ctx.state[1] >> (i * 8)) & 0x000000ff
+        hash[i + 8]  = byte(ctx.state[2] >> (i * 8)) & 0x000000ff
+        hash[i + 12] = byte(ctx.state[3] >> (i * 8)) & 0x000000ff
+    }
+}
+
+/*
+    MD4 implementation
+*/
+
+BLOCK_SIZE  :: 64
+
+Md4_Context :: struct {
+    data:    [64]byte,
+    state:   [4]u32,
+    bitlen:  u64,
+    datalen: u32,
+}
+
+/*
+    @note(zh): F, G and H, as mentioned in the RFC, have been inlined into FF, GG 
+    and HH respectively, instead of declaring them separately.
+*/
+
+FF :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + ((b & c) | (~b & d)) + x, s)
+}
+
+GG :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + ((b & c) | (b & d) | (c & d)) + x + 0x5a827999, s)
+}
+
+HH :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + (b ~ c ~ d) + x + 0x6ed9eba1, s)
+}
+
+transform :: proc(ctx: ^Md4_Context, data: []byte) {
+    a, b, c, d, i, j: u32
+    m: [16]u32
+
+    for i, j = 0, 0; i < 16; i += 1 {
+        m[i] = u32(data[j]) | (u32(data[j + 1]) << 8) | (u32(data[j + 2]) << 16) | (u32(data[j + 3]) << 24)
+        j += 4
+    }
+
+    a = ctx.state[0]
+    b = ctx.state[1]
+    c = ctx.state[2]
+    d = ctx.state[3]
+
+    a = FF(a, b, c, d, m[0],  3)
+    d = FF(d, a, b, c, m[1],  7)
+    c = FF(c, d, a, b, m[2],  11)
+    b = FF(b, c, d, a, m[3],  19)
+    a = FF(a, b, c, d, m[4],  3)
+    d = FF(d, a, b, c, m[5],  7)
+    c = FF(c, d, a, b, m[6],  11)
+    b = FF(b, c, d, a, m[7],  19)
+    a = FF(a, b, c, d, m[8],  3)
+    d = FF(d, a, b, c, m[9],  7)
+    c = FF(c, d, a, b, m[10], 11)
+    b = FF(b, c, d, a, m[11], 19)
+    a = FF(a, b, c, d, m[12], 3)
+    d = FF(d, a, b, c, m[13], 7)
+    c = FF(c, d, a, b, m[14], 11)
+    b = FF(b, c, d, a, m[15], 19)
+
+    a = GG(a, b, c, d, m[0],  3)
+    d = GG(d, a, b, c, m[4],  5)
+    c = GG(c, d, a, b, m[8],  9)
+    b = GG(b, c, d, a, m[12], 13)
+    a = GG(a, b, c, d, m[1],  3)
+    d = GG(d, a, b, c, m[5],  5)
+    c = GG(c, d, a, b, m[9],  9)
+    b = GG(b, c, d, a, m[13], 13)
+    a = GG(a, b, c, d, m[2],  3)
+    d = GG(d, a, b, c, m[6],  5)
+    c = GG(c, d, a, b, m[10], 9)
+    b = GG(b, c, d, a, m[14], 13)
+    a = GG(a, b, c, d, m[3],  3)
+    d = GG(d, a, b, c, m[7],  5)
+    c = GG(c, d, a, b, m[11], 9)
+    b = GG(b, c, d, a, m[15], 13)
+
+    a = HH(a, b, c, d, m[0],  3)
+    d = HH(d, a, b, c, m[8],  9)
+    c = HH(c, d, a, b, m[4],  11)
+    b = HH(b, c, d, a, m[12], 15)
+    a = HH(a, b, c, d, m[2],  3)
+    d = HH(d, a, b, c, m[10], 9)
+    c = HH(c, d, a, b, m[6],  11)
+    b = HH(b, c, d, a, m[14], 15)
+    a = HH(a, b, c, d, m[1],  3)
+    d = HH(d, a, b, c, m[9],  9)
+    c = HH(c, d, a, b, m[5],  11)
+    b = HH(b, c, d, a, m[13], 15)
+    a = HH(a, b, c, d, m[3],  3)
+    d = HH(d, a, b, c, m[11], 9)
+    c = HH(c, d, a, b, m[7],  11)
+    b = HH(b, c, d, a, m[15], 15)
+
+    ctx.state[0] += a
+    ctx.state[1] += b
+    ctx.state[2] += c
+    ctx.state[3] += d
+}

core/crypto/md5/md5.odin

@@ -0,0 +1,263 @@
+package md5
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the MD5 hashing algorithm, as defined in RFC 1321 <https://datatracker.ietf.org/doc/html/rfc1321>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [16]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: Md5_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: Md5_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Md5_Context) {
+    ctx.state[0] = 0x67452301
+    ctx.state[1] = 0xefcdab89
+    ctx.state[2] = 0x98badcfe
+    ctx.state[3] = 0x10325476
+}
+
+update :: proc(ctx: ^Md5_Context, data: []byte) {
+    for i := 0; i < len(data); i += 1 {
+        ctx.data[ctx.datalen] = data[i]
+        ctx.datalen += 1
+        if(ctx.datalen == BLOCK_SIZE) {
+            transform(ctx, ctx.data[:])
+            ctx.bitlen += 512
+            ctx.datalen = 0
+        }
+    }
+}
+
+final :: proc(ctx: ^Md5_Context, hash: []byte){
+    i : u32
+    i = ctx.datalen
+
+    if ctx.datalen < 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < 56 {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+    } else if ctx.datalen >= 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < BLOCK_SIZE {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+        transform(ctx, ctx.data[:])
+        mem.set(&ctx.data, 0, 56)
+    }
+
+    ctx.bitlen  += u64(ctx.datalen * 8)
+    ctx.data[56] = byte(ctx.bitlen)
+    ctx.data[57] = byte(ctx.bitlen >> 8)
+    ctx.data[58] = byte(ctx.bitlen >> 16)
+    ctx.data[59] = byte(ctx.bitlen >> 24)
+    ctx.data[60] = byte(ctx.bitlen >> 32)
+    ctx.data[61] = byte(ctx.bitlen >> 40)
+    ctx.data[62] = byte(ctx.bitlen >> 48)
+    ctx.data[63] = byte(ctx.bitlen >> 56)
+    transform(ctx, ctx.data[:])
+
+    for i = 0; i < 4; i += 1 {
+        hash[i]      = byte(ctx.state[0] >> (i * 8)) & 0x000000ff
+        hash[i + 4]  = byte(ctx.state[1] >> (i * 8)) & 0x000000ff
+        hash[i + 8]  = byte(ctx.state[2] >> (i * 8)) & 0x000000ff
+        hash[i + 12] = byte(ctx.state[3] >> (i * 8)) & 0x000000ff
+    }
+}
+
+/*
+    MD4 implementation
+*/
+
+BLOCK_SIZE  :: 64
+
+Md5_Context :: struct {
+    data:    [BLOCK_SIZE]byte,
+    state:   [4]u32,
+    bitlen:  u64,
+    datalen: u32,
+}
+
+/*
+    @note(zh): F, G, H and I, as mentioned in the RFC, have been inlined into FF, GG, HH 
+    and II respectively, instead of declaring them separately.
+*/
+
+FF :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + ((b & c) | (~b & d)) + m + t, s)
+}
+
+GG :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + ((b & d) | (c & ~d)) + m + t, s)
+}
+
+HH :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + (b ~ c ~ d) + m + t, s)
+}
+
+II :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + (c ~ (b | ~d)) + m + t, s)
+}
+
+transform :: proc(ctx: ^Md5_Context, data: []byte) {
+    i, j: u32
+    m: [16]u32
+
+    for i, j = 0, 0; i < 16; i+=1 {
+        m[i] = u32(data[j]) + u32(data[j + 1]) << 8 + u32(data[j + 2]) << 16 + u32(data[j + 3]) << 24
+        j += 4
+    }
+
+    a := ctx.state[0]
+    b := ctx.state[1]
+    c := ctx.state[2]
+    d := ctx.state[3]
+
+    a = FF(a, b, c, d, m[0],   7, 0xd76aa478)
+    d = FF(d, a, b, c, m[1],  12, 0xe8c7b756)
+    c = FF(c, d, a, b, m[2],  17, 0x242070db)
+    b = FF(b, c, d, a, m[3],  22, 0xc1bdceee)
+    a = FF(a, b, c, d, m[4],   7, 0xf57c0faf)
+    d = FF(d, a, b, c, m[5],  12, 0x4787c62a)
+    c = FF(c, d, a, b, m[6],  17, 0xa8304613)
+    b = FF(b, c, d, a, m[7],  22, 0xfd469501)
+    a = FF(a, b, c, d, m[8],   7, 0x698098d8)
+    d = FF(d, a, b, c, m[9],  12, 0x8b44f7af)
+    c = FF(c, d, a, b, m[10], 17, 0xffff5bb1)
+    b = FF(b, c, d, a, m[11], 22, 0x895cd7be)
+    a = FF(a, b, c, d, m[12],  7, 0x6b901122)
+    d = FF(d, a, b, c, m[13], 12, 0xfd987193)
+    c = FF(c, d, a, b, m[14], 17, 0xa679438e)
+    b = FF(b, c, d, a, m[15], 22, 0x49b40821)
+
+    a = GG(a, b, c, d, m[1],   5, 0xf61e2562)
+    d = GG(d, a, b, c, m[6],   9, 0xc040b340)
+    c = GG(c, d, a, b, m[11], 14, 0x265e5a51)
+    b = GG(b, c, d, a, m[0],  20, 0xe9b6c7aa)
+    a = GG(a, b, c, d, m[5],   5, 0xd62f105d)
+    d = GG(d, a, b, c, m[10],  9, 0x02441453)
+    c = GG(c, d, a, b, m[15], 14, 0xd8a1e681)
+    b = GG(b, c, d, a, m[4],  20, 0xe7d3fbc8)
+    a = GG(a, b, c, d, m[9],   5, 0x21e1cde6)
+    d = GG(d, a, b, c, m[14],  9, 0xc33707d6)
+    c = GG(c, d, a, b, m[3],  14, 0xf4d50d87)
+    b = GG(b, c, d, a, m[8],  20, 0x455a14ed)
+    a = GG(a, b, c, d, m[13],  5, 0xa9e3e905)
+    d = GG(d, a, b, c, m[2],   9, 0xfcefa3f8)
+    c = GG(c, d, a, b, m[7],  14, 0x676f02d9)
+    b = GG(b, c, d, a, m[12], 20, 0x8d2a4c8a)
+
+    a = HH(a, b, c, d, m[5],   4, 0xfffa3942)
+    d = HH(d, a, b, c, m[8],  11, 0x8771f681)
+    c = HH(c, d, a, b, m[11], 16, 0x6d9d6122)
+    b = HH(b, c, d, a, m[14], 23, 0xfde5380c)
+    a = HH(a, b, c, d, m[1],   4, 0xa4beea44)
+    d = HH(d, a, b, c, m[4],  11, 0x4bdecfa9)
+    c = HH(c, d, a, b, m[7],  16, 0xf6bb4b60)
+    b = HH(b, c, d, a, m[10], 23, 0xbebfbc70)
+    a = HH(a, b, c, d, m[13],  4, 0x289b7ec6)
+    d = HH(d, a, b, c, m[0],  11, 0xeaa127fa)
+    c = HH(c, d, a, b, m[3],  16, 0xd4ef3085)
+    b = HH(b, c, d, a, m[6],  23, 0x04881d05)
+    a = HH(a, b, c, d, m[9],   4, 0xd9d4d039)
+    d = HH(d, a, b, c, m[12], 11, 0xe6db99e5)
+    c = HH(c, d, a, b, m[15], 16, 0x1fa27cf8)
+    b = HH(b, c, d, a, m[2],  23, 0xc4ac5665)
+
+    a = II(a, b, c, d, m[0],   6, 0xf4292244)
+    d = II(d, a, b, c, m[7],  10, 0x432aff97)
+    c = II(c, d, a, b, m[14], 15, 0xab9423a7)
+    b = II(b, c, d, a, m[5],  21, 0xfc93a039)
+    a = II(a, b, c, d, m[12],  6, 0x655b59c3)
+    d = II(d, a, b, c, m[3],  10, 0x8f0ccc92)
+    c = II(c, d, a, b, m[10], 15, 0xffeff47d)
+    b = II(b, c, d, a, m[1],  21, 0x85845dd1)
+    a = II(a, b, c, d, m[8],   6, 0x6fa87e4f)
+    d = II(d, a, b, c, m[15], 10, 0xfe2ce6e0)
+    c = II(c, d, a, b, m[6],  15, 0xa3014314)
+    b = II(b, c, d, a, m[13], 21, 0x4e0811a1)
+    a = II(a, b, c, d, m[4],   6, 0xf7537e82)
+    d = II(d, a, b, c, m[11], 10, 0xbd3af235)
+    c = II(c, d, a, b, m[2],  15, 0x2ad7d2bb)
+    b = II(b, c, d, a, m[9],  21, 0xeb86d391)
+
+    ctx.state[0] += a
+    ctx.state[1] += b
+    ctx.state[2] += c
+    ctx.state[3] += d
+}

core/crypto/poly1305/poly1305.odin

@@ -0,0 +1,163 @@
+package poly1305
+
+import "core:crypto"
+import "core:crypto/util"
+import field "core:crypto/_fiat/field_poly1305"
+import "core:mem"
+
+KEY_SIZE :: 32
+TAG_SIZE :: 16
+
+_BLOCK_SIZE :: 16
+
+sum :: proc (dst, msg, key: []byte) {
+	ctx: Context = ---
+
+	init(&ctx, key)
+	update(&ctx, msg)
+	final(&ctx, dst)
+}
+
+verify :: proc (tag, msg, key: []byte) -> bool {
+	ctx: Context = ---
+	derived_tag: [16]byte = ---
+
+	if len(tag) != TAG_SIZE {
+		panic("crypto/poly1305: invalid tag size")
+	}
+
+	init(&ctx, key)
+	update(&ctx, msg)
+	final(&ctx, derived_tag[:])
+
+	return crypto.compare_constant_time(derived_tag[:], tag) == 1
+}
+
+Context :: struct {
+	_r: field.Tight_Field_Element,
+	_a: field.Tight_Field_Element,
+	_s: field.Tight_Field_Element,
+
+	_buffer: [_BLOCK_SIZE]byte,
+	_leftover: int,
+
+	_is_initialized: bool,
+}
+
+init :: proc (ctx: ^Context, key: []byte) {
+	if len(key) != KEY_SIZE {
+		panic("crypto/poly1305: invalid key size")
+	}
+
+	// r = le_bytes_to_num(key[0..15])
+	// r = clamp(r) (r &= 0xffffffc0ffffffc0ffffffc0fffffff)
+	tmp_lo := util.U64_LE(key[0:8]) & 0x0ffffffc0fffffff
+	tmp_hi := util.U64_LE(key[8:16]) & 0xffffffc0ffffffc
+	field.fe_from_u64s(&ctx._r, tmp_lo, tmp_hi)
+
+	// s = le_bytes_to_num(key[16..31])
+	field.fe_from_bytes(&ctx._s, key[16:32], 0)
+
+	// a = 0
+	field.fe_zero(&ctx._a)
+
+	// No leftover in buffer
+	ctx._leftover = 0
+
+	ctx._is_initialized = true
+}
+
+update :: proc (ctx: ^Context, data: []byte) {
+	assert(ctx._is_initialized)
+
+	msg := data
+	msg_len := len(data)
+
+	// Handle leftover
+	if ctx._leftover > 0 {
+		want := min(_BLOCK_SIZE - ctx._leftover, msg_len)
+		copy_slice(ctx._buffer[ctx._leftover:], msg[:want])
+		msg_len = msg_len - want
+		msg = msg[want:]
+		ctx._leftover = ctx._leftover + want
+		if ctx._leftover < _BLOCK_SIZE {
+			return
+		}
+		_blocks(ctx, ctx._buffer[:])
+		ctx._leftover = 0
+	}
+
+	// Process full blocks
+	if msg_len >= _BLOCK_SIZE {
+		want := msg_len & (~int(_BLOCK_SIZE - 1))
+		_blocks(ctx, msg[:want])
+		msg = msg[want:]
+		msg_len = msg_len - want
+	}
+
+	// Store leftover
+	if msg_len > 0 {
+		// TODO: While -donna does it this way, I'm fairly sure that
+		// `ctx._leftover == 0` is an invariant at this point.
+		copy(ctx._buffer[ctx._leftover:], msg)
+		ctx._leftover = ctx._leftover + msg_len
+	}
+}
+
+final :: proc (ctx: ^Context, dst: []byte) {
+	assert(ctx._is_initialized)
+
+	if len(dst) != TAG_SIZE {
+		panic("poly1305: invalid destination tag size")
+	}
+
+	// Process remaining block
+	if ctx._leftover > 0 {
+		ctx._buffer[ctx._leftover] = 1
+		for i := ctx._leftover + 1; i < _BLOCK_SIZE; i = i + 1 {
+			ctx._buffer[i] = 0
+		}
+		_blocks(ctx, ctx._buffer[:], true)
+	}
+
+	// a += s
+	field.fe_add(field.fe_relax_cast(&ctx._a), &ctx._a, &ctx._s) // _a unreduced
+	field.fe_carry(&ctx._a, field.fe_relax_cast(&ctx._a)) // _a reduced
+
+	// return num_to_16_le_bytes(a)
+	tmp: [32]byte = ---
+	field.fe_to_bytes(&tmp, &ctx._a)
+	copy_slice(dst, tmp[0:16])
+
+	reset(ctx)
+}
+
+reset :: proc (ctx: ^Context) {
+	mem.zero_explicit(&ctx._r, size_of(ctx._r))
+	mem.zero_explicit(&ctx._a, size_of(ctx._a))
+	mem.zero_explicit(&ctx._s, size_of(ctx._s))
+	mem.zero_explicit(&ctx._buffer, size_of(ctx._buffer))
+
+	ctx._is_initialized = false
+}
+
+_blocks :: proc (ctx: ^Context, msg: []byte, final := false) {
+	n: field.Tight_Field_Element = ---
+	final_byte := byte(!final)
+
+	data := msg
+	data_len := len(data)
+	for data_len >= _BLOCK_SIZE {
+		// n = le_bytes_to_num(msg[((i-1)*16)..*i*16] | [0x01])
+		field.fe_from_bytes(&n, data[:_BLOCK_SIZE], final_byte, false)
+
+		// a += n
+		field.fe_add(field.fe_relax_cast(&ctx._a), &ctx._a, &n) // _a unreduced
+
+		// a = (r * a) % p
+		field.fe_carry_mul(&ctx._a, field.fe_relax_cast(&ctx._a), field.fe_relax_cast(&ctx._r)) // _a reduced
+
+		data = data[_BLOCK_SIZE:]
+		data_len = data_len - _BLOCK_SIZE
+	}
+}

core/crypto/rand_generic.odin

@@ -0,0 +1,7 @@
+package crypto
+
+when ODIN_OS != "linux" {
+	_rand_bytes :: proc (dst: []byte) {
+		unimplemented("crypto: rand_bytes not supported on this OS")
+	}
+}

core/crypto/rand_linux.odin

@@ -0,0 +1,37 @@
+package crypto
+
+import "core:fmt"
+import "core:os"
+import "core:sys/unix"
+
+_MAX_PER_CALL_BYTES :: 33554431 // 2^25 - 1
+
+_rand_bytes :: proc (dst: []byte) {
+	dst := dst
+	l := len(dst)
+
+	for l > 0 {
+		to_read := min(l, _MAX_PER_CALL_BYTES)
+		ret := unix.sys_getrandom(raw_data(dst), to_read, 0)
+		if ret < 0 {
+			switch os.Errno(-ret) {
+			case os.EINTR:
+				// Call interupted by a signal handler, just retry the
+				// request.
+				continue
+			case os.ENOSYS:
+				// The kernel is apparently prehistoric (< 3.17 circa 2014)
+				// and does not support getrandom.
+				panic("crypto: getrandom not available in kernel")
+			case:
+				// All other failures are things that should NEVER happen
+				// unless the kernel interface changes (ie: the Linux
+				// developers break userland).
+				panic(fmt.tprintf("crypto: getrandom failed: %d", ret))
+			}
+		}
+
+		l -= ret
+		dst = dst[ret:]
+	}
+}

core/crypto/ripemd/ripemd.odin

@@ -0,0 +1,834 @@
+package ripemd
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation for the RIPEMD hashing algorithm as defined in <https://homes.esat.kuleuven.be/~bosselae/ripemd160.html>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string_128 will hash the given input and return the
+// computed hash
+hash_string_128 :: proc(data: string) -> [16]byte {
+    return hash_bytes_128(transmute([]byte)(data))
+}
+
+// hash_bytes_128 will hash the given input and return the
+// computed hash
+hash_bytes_128 :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: Ripemd128_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_128 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_128 :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: Ripemd128_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_128 will read the file provided by the given handle
+// and compute a hash
+hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream_128(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_128(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash_128 :: proc {
+    hash_stream_128,
+    hash_file_128,
+    hash_bytes_128,
+    hash_string_128,
+}
+
+// hash_string_160 will hash the given input and return the
+// computed hash
+hash_string_160 :: proc(data: string) -> [20]byte {
+    return hash_bytes_160(transmute([]byte)(data))
+}
+
+// hash_bytes_160 will hash the given input and return the
+// computed hash
+hash_bytes_160 :: proc(data: []byte) -> [20]byte {
+    hash: [20]byte
+    ctx: Ripemd160_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_160 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_160 :: proc(s: io.Stream) -> ([20]byte, bool) {
+    hash: [20]byte
+    ctx: Ripemd160_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_160 will read the file provided by the given handle
+// and compute a hash
+hash_file_160 :: proc(hd: os.Handle, load_at_once := false) -> ([20]byte, bool) {
+    if !load_at_once {
+        return hash_stream_160(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_160(buf[:]), ok
+        }
+    }
+    return [20]byte{}, false
+}
+
+hash_160 :: proc {
+    hash_stream_160,
+    hash_file_160,
+    hash_bytes_160,
+    hash_string_160,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Ripemd256_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Ripemd256_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_320 will hash the given input and return the
+// computed hash
+hash_string_320 :: proc(data: string) -> [40]byte {
+    return hash_bytes_320(transmute([]byte)(data))
+}
+
+// hash_bytes_320 will hash the given input and return the
+// computed hash
+hash_bytes_320 :: proc(data: []byte) -> [40]byte {
+    hash: [40]byte
+    ctx: Ripemd320_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_320 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_320 :: proc(s: io.Stream) -> ([40]byte, bool) {
+    hash: [40]byte
+    ctx: Ripemd320_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_320 will read the file provided by the given handle
+// and compute a hash
+hash_file_320 :: proc(hd: os.Handle, load_at_once := false) -> ([40]byte, bool) {
+    if !load_at_once {
+        return hash_stream_320(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_320(buf[:]), ok
+        }
+    }
+    return [40]byte{}, false
+}
+
+hash_320 :: proc {
+    hash_stream_320,
+    hash_file_320,
+    hash_bytes_320,
+    hash_string_320,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^$T) {
+    when T == Ripemd128_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3] = S0, S1, S2, S3
+    } else when T == Ripemd160_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4] = S0, S1, S2, S3, S4
+    } else when T == Ripemd256_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3] = S0, S1, S2, S3
+        ctx.s[4], ctx.s[5], ctx.s[6], ctx.s[7] = S5, S6, S7, S8
+    } else when T == Ripemd320_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4] = S0, S1, S2, S3, S4
+        ctx.s[5], ctx.s[6], ctx.s[7], ctx.s[8], ctx.s[9] = S5, S6, S7, S8, S9
+    }
+}
+
+update :: proc(ctx: ^$T, data: []byte) {
+    ctx.tc += u64(len(data))
+    data := data
+    if ctx.nx > 0 {
+        n := len(data)
+
+        when T == Ripemd128_Context {
+            if n > RIPEMD_128_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_128_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd160_Context {
+            if n > RIPEMD_160_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_160_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd256_Context{
+            if n > RIPEMD_256_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_256_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd320_Context{
+            if n > RIPEMD_320_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_320_BLOCK_SIZE - ctx.nx
+            }
+        }
+
+        for i := 0; i < n; i += 1 {
+            ctx.x[ctx.nx + i] = data[i]
+        }
+
+        ctx.nx += n
+        when T == Ripemd128_Context {
+            if ctx.nx == RIPEMD_128_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd160_Context {
+            if ctx.nx == RIPEMD_160_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd256_Context{
+            if ctx.nx == RIPEMD_256_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd320_Context{
+            if ctx.nx == RIPEMD_320_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        }
+        data = data[n:]
+    }
+    n := block(ctx, data)
+    data = data[n:]
+    if len(data) > 0 {
+        ctx.nx = copy(ctx.x[:], data)
+    }
+}
+
+final :: proc(ctx: ^$T, hash: []byte) {
+    d := ctx
+    tc := d.tc
+    tmp: [64]byte
+    tmp[0] = 0x80
+
+    if tc % 64 < 56 {
+        update(d, tmp[0:56 - tc % 64])
+    } else {
+        update(d, tmp[0:64 + 56 - tc % 64])
+    }
+
+    tc <<= 3
+    for i : u32 = 0; i < 8; i += 1 {
+        tmp[i] = byte(tc >> (8 * i))
+    }
+
+    update(d, tmp[0:8])
+
+    when T == Ripemd128_Context {
+        size :: RIPEMD_128_SIZE
+    } else when T == Ripemd160_Context {
+        size :: RIPEMD_160_SIZE
+    } else when T == Ripemd256_Context{
+        size :: RIPEMD_256_SIZE
+    } else when T == Ripemd320_Context{
+        size :: RIPEMD_320_SIZE
+    }
+
+    digest: [size]byte
+    for s, i in d.s {
+        digest[i * 4]     = byte(s)
+        digest[i * 4 + 1] = byte(s >> 8)
+        digest[i * 4 + 2] = byte(s >> 16)
+        digest[i * 4 + 3] = byte(s >> 24)
+    }
+    copy(hash[:], digest[:])
+}
+
+
+/*
+    RIPEMD implementation
+*/
+
+Ripemd128_Context :: struct {
+	s:  [4]u32,
+	x:  [RIPEMD_128_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd160_Context :: struct {
+	s:  [5]u32,
+	x:  [RIPEMD_160_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd256_Context :: struct {
+	s:  [8]u32,
+	x:  [RIPEMD_256_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd320_Context :: struct {
+	s:  [10]u32,
+	x:  [RIPEMD_320_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+RIPEMD_128_SIZE       :: 16
+RIPEMD_128_BLOCK_SIZE :: 64
+RIPEMD_160_SIZE       :: 20
+RIPEMD_160_BLOCK_SIZE :: 64
+RIPEMD_256_SIZE       :: 32
+RIPEMD_256_BLOCK_SIZE :: 64
+RIPEMD_320_SIZE       :: 40
+RIPEMD_320_BLOCK_SIZE :: 64
+
+S0 :: 0x67452301
+S1 :: 0xefcdab89
+S2 :: 0x98badcfe
+S3 :: 0x10325476
+S4 :: 0xc3d2e1f0
+S5 :: 0x76543210
+S6 :: 0xfedcba98
+S7 :: 0x89abcdef
+S8 :: 0x01234567
+S9 :: 0x3c2d1e0f
+
+RIPEMD_128_N0 := [64]uint {
+	0, 1,  2,  3,  4,  5,  6,  7, 8,  9, 10, 11, 12, 13, 14, 15,
+	7, 4,  13, 1,  10, 6,  15, 3, 12, 0, 9,  5,  2,  14, 11, 8,
+	3, 10, 14, 4,  9,  15, 8,  1, 2,  7, 0,  6,  13, 11, 5,  12,
+	1, 9,  11, 10, 0,  8,  12, 4, 13, 3, 7,  15, 14, 5,  6,  2,
+}
+
+RIPEMD_128_R0 := [64]uint {
+	11, 14, 15, 12, 5,  8,  7,  9,  11, 13, 14, 15, 6,  7,  9,  8,
+	7,  6,  8,  13, 11, 9,  7,  15, 7,  12, 15, 9,  11, 7,  13, 12,
+	11, 13, 6,  7,  14, 9,  13, 15, 14, 8,  13, 6,  5,  12, 7,  5,
+	11, 12, 14, 15, 14, 15, 9,  8,  9,  14, 5,  6,  8,  6,  5,  12,
+}
+
+RIPEMD_128_N1 := [64]uint {
+	5, 14, 7, 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12,
+	6, 11, 3, 7, 0, 13, 5, 10, 14, 15, 8, 12, 4, 9, 1, 2,
+	15, 5, 1, 3, 7, 14, 6, 9, 11, 8, 12, 2, 10, 0, 4, 13,
+	8, 6, 4, 1, 3, 11, 15, 0, 5, 12, 2, 13, 9, 7, 10, 14,
+}
+
+RIPEMD_128_R1 := [64]uint {
+	8, 9, 9, 11, 13, 15, 15, 5, 7, 7, 8, 11, 14, 14, 12, 6,
+	9, 13, 15, 7, 12, 8, 9, 11, 7, 7, 12, 7, 6, 15, 13, 11,
+	9, 7, 15, 11, 8, 6, 6, 14, 12, 13, 5, 14, 13, 13, 7, 5,
+	15, 5, 8, 11, 14, 14, 6, 14, 6, 9, 12, 9, 12, 5, 15, 8,
+}
+
+RIPEMD_160_N0 := [80]uint {
+	0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	7, 4, 13, 1, 10, 6, 15, 3, 12, 0, 9, 5, 2, 14, 11, 8,
+	3, 10, 14, 4, 9, 15, 8, 1, 2, 7, 0, 6, 13, 11, 5, 12,
+	1, 9, 11, 10, 0, 8, 12, 4, 13, 3, 7, 15, 14, 5, 6, 2,
+	4, 0, 5, 9, 7, 12, 2, 10, 14, 1, 3, 8, 11, 6, 15, 13,
+}
+
+RIPEMD_160_R0 := [80]uint {
+	11, 14, 15, 12, 5, 8, 7, 9, 11, 13, 14, 15, 6, 7, 9, 8,
+	7, 6, 8, 13, 11, 9, 7, 15, 7, 12, 15, 9, 11, 7, 13, 12,
+	11, 13, 6, 7, 14, 9, 13, 15, 14, 8, 13, 6, 5, 12, 7, 5,
+	11, 12, 14, 15, 14, 15, 9, 8, 9, 14, 5, 6, 8, 6, 5, 12,
+	9, 15, 5, 11, 6, 8, 13, 12, 5, 12, 13, 14, 11, 8, 5, 6,
+}
+
+RIPEMD_160_N1 := [80]uint {
+	5, 14, 7, 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12,
+	6, 11, 3, 7, 0, 13, 5, 10, 14, 15, 8, 12, 4, 9, 1, 2,
+	15, 5, 1, 3, 7, 14, 6, 9, 11, 8, 12, 2, 10, 0, 4, 13,
+	8, 6, 4, 1, 3, 11, 15, 0, 5, 12, 2, 13, 9, 7, 10, 14,
+	12, 15, 10, 4, 1, 5, 8, 7, 6, 2, 13, 14, 0, 3, 9, 11,
+}
+
+RIPEMD_160_R1 := [80]uint {
+	8, 9, 9, 11, 13, 15, 15, 5, 7, 7, 8, 11, 14, 14, 12, 6,
+	9, 13, 15, 7, 12, 8, 9, 11, 7, 7, 12, 7, 6, 15, 13, 11,
+	9, 7, 15, 11, 8, 6, 6, 14, 12, 13, 5, 14, 13, 13, 7, 5,
+	15, 5, 8, 11, 14, 14, 6, 14, 6, 9, 12, 9, 12, 5, 15, 8,
+	8, 5, 12, 9, 12, 5, 14, 6, 8, 13, 6, 5, 15, 13, 11, 11,
+}
+
+block :: #force_inline proc (ctx: ^$T, p: []byte) -> int {
+    when T == Ripemd128_Context {
+    	return ripemd_128_block(ctx, p)
+    }
+    else when T == Ripemd160_Context {
+    	return ripemd_160_block(ctx, p)
+    }
+    else when T == Ripemd256_Context {
+    	return ripemd_256_block(ctx, p)
+    }
+    else when T == Ripemd320_Context {
+    	return ripemd_320_block(ctx, p)
+    }
+}
+
+ripemd_128_block :: proc(ctx: ^$T, p: []byte) -> int {
+	n := 0
+	x: [16]u32 = ---
+	alpha: u32 = ---
+	p := p
+	for len(p) >= RIPEMD_128_BLOCK_SIZE {
+		a, b, c, d := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3]
+		aa, bb, cc, dd := a, b, c, d
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_128_N0[i]]
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb & dd | cc &~ dd) + x[RIPEMD_128_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd= dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 32 {
+			alpha = a + (d ~ (b & (c~d))) + x[RIPEMD_128_N0[i]] + 0x5a827999
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb | ~cc)) + x[RIPEMD_128_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 48 {
+			alpha = a + (d ~ (b | ~c)) + x[RIPEMD_128_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb & (cc~dd))) + x[RIPEMD_128_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 64 {
+			alpha = a + (c ~ (d & (b~c))) + x[RIPEMD_128_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_128_N1[i]]
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		c = ctx.s[1] + c + dd
+		ctx.s[1] = ctx.s[2] + d + aa
+		ctx.s[2] = ctx.s[3] + a + bb
+		ctx.s[3] = ctx.s[0] + b + cc
+		ctx.s[0] = c
+		p = p[RIPEMD_128_BLOCK_SIZE:]
+		n += RIPEMD_128_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_160_block :: proc(ctx: ^$T, p: []byte) -> int {
+    n := 0
+	x: [16]u32 = ---
+	alpha, beta: u32 = ---, ---
+	p := p
+	for len(p) >= RIPEMD_160_BLOCK_SIZE {
+		a, b, c, d, e := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4]
+		aa, bb, cc, dd, ee := a, b, c, d, e
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_160_N0[i]]
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ (cc | ~dd)) + x[RIPEMD_160_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 32 {
+			alpha = a + (b&c | ~b&d) + x[RIPEMD_160_N0[i]] + 0x5a827999
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&dd | cc&~dd) + x[RIPEMD_160_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 48 {
+			alpha = a + (b | ~c ~ d) + x[RIPEMD_160_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb | ~cc ~ dd) + x[RIPEMD_160_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 64 {
+			alpha = a + (b&d | c&~d) + x[RIPEMD_160_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&cc | ~bb&dd) + x[RIPEMD_160_N1[i]] + 0x7a6d76e9
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 80 {
+			alpha = a + (b ~ (c | ~d)) + x[RIPEMD_160_N0[i]] + 0xa953fd4e
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_160_N1[i]]
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		dd += c + ctx.s[1]
+		ctx.s[1] = ctx.s[2] + d + ee
+		ctx.s[2] = ctx.s[3] + e + aa
+		ctx.s[3] = ctx.s[4] + a + bb
+		ctx.s[4] = ctx.s[0] + b + cc
+		ctx.s[0] = dd
+		p = p[RIPEMD_160_BLOCK_SIZE:]
+		n += RIPEMD_160_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_256_block :: proc(ctx: ^$T, p: []byte) -> int {
+	n := 0
+	x: [16]u32 = ---
+	alpha: u32 = ---
+	p := p
+	for len(p) >= RIPEMD_256_BLOCK_SIZE {
+		a, b, c, d := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3]
+		aa, bb, cc, dd := ctx.s[4], ctx.s[5], ctx.s[6], ctx.s[7]
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_128_N0[i]]
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb & dd | cc &~ dd) + x[RIPEMD_128_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd= dd, alpha, bb, cc
+			i += 1
+		}
+		t := a
+		a = aa
+		aa = t
+		for i < 32 {
+			alpha = a + (d ~ (b & (c~d))) + x[RIPEMD_128_N0[i]] + 0x5a827999
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb | ~cc)) + x[RIPEMD_128_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = b
+		b = bb
+		bb = t
+		for i < 48 {
+			alpha = a + (d ~ (b | ~c)) + x[RIPEMD_128_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb & (cc~dd))) + x[RIPEMD_128_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = c
+		c = cc
+		cc = t
+		for i < 64 {
+			alpha = a + (c ~ (d & (b~c))) + x[RIPEMD_128_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_128_N1[i]]
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = d
+		d = dd
+		dd = t
+		ctx.s[0] += a
+		ctx.s[1] += b
+		ctx.s[2] += c
+		ctx.s[3] += d
+		ctx.s[4] += aa
+		ctx.s[5] += bb
+		ctx.s[6] += cc
+		ctx.s[7] += dd
+		p = p[RIPEMD_256_BLOCK_SIZE:]
+		n += RIPEMD_256_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_320_block :: proc(ctx: ^$T, p: []byte) -> int {
+    n := 0
+	x: [16]u32 = ---
+	alpha, beta: u32 = ---, ---
+	p := p
+	for len(p) >= RIPEMD_320_BLOCK_SIZE {
+		a, b, c, d, e := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4]
+		aa, bb, cc, dd, ee := ctx.s[5], ctx.s[6], ctx.s[7], ctx.s[8], ctx.s[9]
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_160_N0[i]]
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ (cc | ~dd)) + x[RIPEMD_160_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t := b
+		b = bb
+		bb = t
+		for i < 32 {
+			alpha = a + (b&c | ~b&d) + x[RIPEMD_160_N0[i]] + 0x5a827999
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&dd | cc&~dd) + x[RIPEMD_160_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = d
+		d = dd
+		dd = t
+		for i < 48 {
+			alpha = a + (b | ~c ~ d) + x[RIPEMD_160_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb | ~cc ~ dd) + x[RIPEMD_160_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = a
+		a = aa
+		aa = t
+		for i < 64 {
+			alpha = a + (b&d | c&~d) + x[RIPEMD_160_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&cc | ~bb&dd) + x[RIPEMD_160_N1[i]] + 0x7a6d76e9
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = c
+		c = cc
+		cc = t
+		for i < 80 {
+			alpha = a + (b ~ (c | ~d)) + x[RIPEMD_160_N0[i]] + 0xa953fd4e
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_160_N1[i]]
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = e
+		e = ee
+		ee = t
+		ctx.s[0] += a
+		ctx.s[1] += b
+		ctx.s[2] += c
+		ctx.s[3] += d
+		ctx.s[4] += e
+		ctx.s[5] += aa
+		ctx.s[6] += bb
+		ctx.s[7] += cc
+		ctx.s[8] += dd
+		ctx.s[9] += ee
+		p = p[RIPEMD_320_BLOCK_SIZE:]
+		n += RIPEMD_320_BLOCK_SIZE
+	}
+	return n
+}

core/crypto/sha1/sha1.odin

@@ -0,0 +1,223 @@
+package sha1
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the SHA1 hashing algorithm, as defined in RFC 3174 <https://datatracker.ietf.org/doc/html/rfc3174>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [20]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [20]byte {
+    hash: [20]byte
+    ctx: Sha1_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([20]byte, bool) {
+    hash: [20]byte
+    ctx: Sha1_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([20]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [20]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Sha1_Context) {
+	ctx.state[0] = 0x67452301
+	ctx.state[1] = 0xefcdab89
+	ctx.state[2] = 0x98badcfe
+	ctx.state[3] = 0x10325476
+	ctx.state[4] = 0xc3d2e1f0
+	ctx.k[0]     = 0x5a827999
+	ctx.k[1]     = 0x6ed9eba1
+	ctx.k[2]     = 0x8f1bbcdc
+	ctx.k[3]     = 0xca62c1d6
+}
+
+update :: proc(ctx: ^Sha1_Context, data: []byte) {
+	for i := 0; i < len(data); i += 1 {
+		ctx.data[ctx.datalen] = data[i]
+		ctx.datalen += 1
+		if (ctx.datalen == BLOCK_SIZE) {
+			transform(ctx, ctx.data[:])
+			ctx.bitlen += 512
+			ctx.datalen = 0
+		}
+	}
+}
+
+final :: proc(ctx: ^Sha1_Context, hash: []byte) {
+	i := ctx.datalen
+
+	if ctx.datalen < 56 {
+		ctx.data[i] = 0x80
+        i += 1
+        for i < 56 {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+	}
+	else {
+		ctx.data[i] = 0x80
+        i += 1
+        for i < BLOCK_SIZE {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+		transform(ctx, ctx.data[:])
+		mem.set(&ctx.data, 0, 56)
+	}
+
+	ctx.bitlen  += u64(ctx.datalen * 8)
+	ctx.data[63] = u8(ctx.bitlen)
+	ctx.data[62] = u8(ctx.bitlen >> 8)
+	ctx.data[61] = u8(ctx.bitlen >> 16)
+	ctx.data[60] = u8(ctx.bitlen >> 24)
+	ctx.data[59] = u8(ctx.bitlen >> 32)
+	ctx.data[58] = u8(ctx.bitlen >> 40)
+	ctx.data[57] = u8(ctx.bitlen >> 48)
+	ctx.data[56] = u8(ctx.bitlen >> 56)
+	transform(ctx, ctx.data[:])
+
+	for j: u32 = 0; j < 4; j += 1 {
+		hash[j]      = u8(ctx.state[0] >> (24 - j * 8)) & 0x000000ff
+		hash[j + 4]  = u8(ctx.state[1] >> (24 - j * 8)) & 0x000000ff
+		hash[j + 8]  = u8(ctx.state[2] >> (24 - j * 8)) & 0x000000ff
+		hash[j + 12] = u8(ctx.state[3] >> (24 - j * 8)) & 0x000000ff
+		hash[j + 16] = u8(ctx.state[4] >> (24 - j * 8)) & 0x000000ff
+	}
+}
+
+/*
+    SHA1 implementation
+*/
+
+BLOCK_SIZE  :: 64
+
+Sha1_Context :: struct {
+    data:    [BLOCK_SIZE]byte,
+    datalen: u32,
+    bitlen:  u64,
+    state:   [5]u32,
+    k:       [4]u32,
+}
+
+transform :: proc(ctx: ^Sha1_Context, data: []byte) {
+    a, b, c, d, e, i, j, t: u32
+    m: [80]u32
+
+	for i, j = 0, 0; i < 16; i += 1 {
+        m[i] = u32(data[j]) << 24 + u32(data[j + 1]) << 16 + u32(data[j + 2]) << 8 + u32(data[j + 3])
+        j += 4
+    }
+	for i < 80 {
+		m[i] = (m[i - 3] ~ m[i - 8] ~ m[i - 14] ~ m[i - 16])
+		m[i] = (m[i] << 1) | (m[i] >> 31)
+        i += 1
+	}
+
+	a = ctx.state[0]
+	b = ctx.state[1]
+	c = ctx.state[2]
+	d = ctx.state[3]
+	e = ctx.state[4]
+
+	for i = 0; i < 20; i += 1 {
+		t = util.ROTL32(a, 5) + ((b & c) ~ (~b & d)) + e + ctx.k[0] + m[i]
+		e = d
+		d = c
+		c = util.ROTL32(b, 30)
+		b = a
+		a = t
+	}
+	for i < 40 {
+		t = util.ROTL32(a, 5) + (b ~ c ~ d) + e + ctx.k[1] + m[i]
+		e = d
+		d = c
+		c = util.ROTL32(b, 30)
+		b = a
+		a = t
+        i += 1
+	}
+	for i < 60 {
+		t = util.ROTL32(a, 5) + ((b & c) ~ (b & d) ~ (c & d)) + e + ctx.k[2] + m[i]
+		e = d
+		d = c
+		c = util.ROTL32(b, 30)
+		b = a
+		a = t
+        i += 1
+	}
+	for i < 80 {
+		t = util.ROTL32(a, 5) + (b ~ c ~ d) + e + ctx.k[3] + m[i]
+		e = d
+		d = c
+		c = util.ROTL32(b, 30)
+		b = a
+		a = t
+        i += 1
+	}
+
+	ctx.state[0] += a
+	ctx.state[1] += b
+	ctx.state[2] += c
+	ctx.state[3] += d
+	ctx.state[4] += e
+}

core/crypto/sha2/sha2.odin

@@ -0,0 +1,593 @@
+package sha2
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the SHA2 hashing algorithm, as defined in <https://csrc.nist.gov/csrc/media/publications/fips/180/2/archive/2002-08-01/documents/fips180-2.pdf>
+    and in RFC 3874 <https://datatracker.ietf.org/doc/html/rfc3874>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc(data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc(data: []byte) -> [28]byte {
+    hash: [28]byte
+	ctx: Sha256_Context
+    ctx.is224 = true
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+	hash: [28]byte
+    ctx: Sha512_Context
+    ctx.is384 = false
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+	ctx: Sha256_Context
+    ctx.is224 = false
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+	hash: [32]byte
+    ctx: Sha512_Context
+    ctx.is384 = false
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc(data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc(data: []byte) -> [48]byte {
+    hash: [48]byte
+	ctx: Sha512_Context
+    ctx.is384 = true
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+	hash: [48]byte
+    ctx: Sha512_Context
+    ctx.is384 = true
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+	ctx: Sha512_Context
+    ctx.is384 = false
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: Sha512_Context
+    ctx.is384 = false
+	init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^$T) {
+    when T == Sha256_Context {
+        if ctx.is224 {
+            ctx.h[0] = 0xc1059ed8
+            ctx.h[1] = 0x367cd507
+            ctx.h[2] = 0x3070dd17
+            ctx.h[3] = 0xf70e5939
+            ctx.h[4] = 0xffc00b31
+            ctx.h[5] = 0x68581511
+            ctx.h[6] = 0x64f98fa7
+            ctx.h[7] = 0xbefa4fa4
+        } else {
+            ctx.h[0] = 0x6a09e667
+            ctx.h[1] = 0xbb67ae85
+            ctx.h[2] = 0x3c6ef372
+            ctx.h[3] = 0xa54ff53a
+            ctx.h[4] = 0x510e527f
+            ctx.h[5] = 0x9b05688c
+            ctx.h[6] = 0x1f83d9ab
+            ctx.h[7] = 0x5be0cd19
+        }
+    } else when T == Sha512_Context {
+        if ctx.is384 {
+            ctx.h[0] = 0xcbbb9d5dc1059ed8
+            ctx.h[1] = 0x629a292a367cd507
+            ctx.h[2] = 0x9159015a3070dd17
+            ctx.h[3] = 0x152fecd8f70e5939
+            ctx.h[4] = 0x67332667ffc00b31
+            ctx.h[5] = 0x8eb44a8768581511
+            ctx.h[6] = 0xdb0c2e0d64f98fa7
+            ctx.h[7] = 0x47b5481dbefa4fa4
+        } else {
+            ctx.h[0] = 0x6a09e667f3bcc908
+            ctx.h[1] = 0xbb67ae8584caa73b
+            ctx.h[2] = 0x3c6ef372fe94f82b
+            ctx.h[3] = 0xa54ff53a5f1d36f1
+            ctx.h[4] = 0x510e527fade682d1
+            ctx.h[5] = 0x9b05688c2b3e6c1f
+            ctx.h[6] = 0x1f83d9abfb41bd6b
+            ctx.h[7] = 0x5be0cd19137e2179
+        }
+    }
+}
+
+update :: proc(ctx: ^$T, data: []byte) {
+    length := uint(len(data))
+    block_nb: uint
+    new_len, rem_len, tmp_len: uint
+    shifted_message := make([]byte, length)
+
+    when T == Sha256_Context {
+        CURR_BLOCK_SIZE :: SHA256_BLOCK_SIZE
+    } else when T == Sha512_Context {
+        CURR_BLOCK_SIZE :: SHA512_BLOCK_SIZE
+    }
+
+    tmp_len = CURR_BLOCK_SIZE - ctx.length
+    rem_len = length < tmp_len ? length : tmp_len
+    copy(ctx.block[ctx.length:], data[:rem_len])
+
+    if ctx.length + length < CURR_BLOCK_SIZE {
+        ctx.length += length
+        return
+    }
+
+    new_len = length - rem_len
+    block_nb = new_len / CURR_BLOCK_SIZE
+    shifted_message = data[rem_len:]
+
+    sha2_transf(ctx, ctx.block[:], 1)
+    sha2_transf(ctx, shifted_message, block_nb)
+
+    rem_len = new_len % CURR_BLOCK_SIZE
+    when T == Sha256_Context      {copy(ctx.block[:], shifted_message[block_nb << 6:rem_len])}
+    else when T == Sha512_Context {copy(ctx.block[:], shifted_message[block_nb << 7:rem_len])}
+
+    ctx.length = rem_len
+    when T == Sha256_Context      {ctx.tot_len += (block_nb + 1) << 6}
+    else when T == Sha512_Context {ctx.tot_len += (block_nb + 1) << 7}
+}
+
+final :: proc(ctx: ^$T, hash: []byte) {
+    block_nb, pm_len, len_b: u32
+    i: i32
+
+    when T == Sha256_Context      {CURR_BLOCK_SIZE :: SHA256_BLOCK_SIZE}
+    else when T == Sha512_Context {CURR_BLOCK_SIZE :: SHA512_BLOCK_SIZE}
+
+    when T == Sha256_Context      {block_nb = 1 + ((CURR_BLOCK_SIZE - 9)  < (ctx.length % CURR_BLOCK_SIZE) ? 1 : 0)}
+    else when T == Sha512_Context {block_nb = 1 + ((CURR_BLOCK_SIZE - 17) < (ctx.length % CURR_BLOCK_SIZE) ? 1 : 0)}
+
+    len_b = u32(ctx.tot_len + ctx.length) << 3
+    when T == Sha256_Context      {pm_len = block_nb << 6}
+    else when T == Sha512_Context {pm_len = block_nb << 7}
+
+    mem.set(rawptr(&(ctx.block[ctx.length:])[0]), 0, int(uint(pm_len) - ctx.length))
+    ctx.block[ctx.length] = 0x80
+
+    util.PUT_U32_BE(ctx.block[pm_len - 4:], len_b)
+
+    sha2_transf(ctx, ctx.block[:], uint(block_nb))
+
+    when T == Sha256_Context {
+        if ctx.is224 {
+            for i = 0; i < 7; i += 1 {util.PUT_U32_BE(hash[i << 2:], ctx.h[i])}
+        } else {
+            for i = 0; i < 8; i += 1 {util.PUT_U32_BE(hash[i << 2:], ctx.h[i])}
+        } 
+    } else when T == Sha512_Context {
+        if ctx.is384 {
+            for i = 0; i < 6; i += 1 {util.PUT_U64_BE(hash[i << 3:], ctx.h[i])}
+        } else {
+            for i = 0; i < 8; i += 1 {util.PUT_U64_BE(hash[i << 3:], ctx.h[i])}
+        } 
+    } 
+}
+
+/*
+    SHA2 implementation
+*/
+
+SHA256_BLOCK_SIZE :: 64
+SHA512_BLOCK_SIZE :: 128
+
+Sha256_Context :: struct {
+    tot_len: uint,
+    length:  uint,
+    block:   [128]byte,
+    h:       [8]u32,
+    is224:   bool,
+}
+
+Sha512_Context :: struct {
+    tot_len: uint,
+    length:  uint,
+    block:   [256]byte,
+    h:       [8]u64,
+    is384:   bool,
+}
+
+sha256_k := [64]u32 {
+    0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
+    0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
+    0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
+    0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
+    0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
+    0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
+    0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
+    0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
+    0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
+    0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
+    0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
+    0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
+    0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
+    0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
+    0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
+    0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2,
+}
+
+sha512_k := [80]u64 {
+    0x428a2f98d728ae22, 0x7137449123ef65cd,
+    0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc,
+    0x3956c25bf348b538, 0x59f111f1b605d019,
+    0x923f82a4af194f9b, 0xab1c5ed5da6d8118,
+    0xd807aa98a3030242, 0x12835b0145706fbe,
+    0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2,
+    0x72be5d74f27b896f, 0x80deb1fe3b1696b1,
+    0x9bdc06a725c71235, 0xc19bf174cf692694,
+    0xe49b69c19ef14ad2, 0xefbe4786384f25e3,
+    0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65,
+    0x2de92c6f592b0275, 0x4a7484aa6ea6e483,
+    0x5cb0a9dcbd41fbd4, 0x76f988da831153b5,
+    0x983e5152ee66dfab, 0xa831c66d2db43210,
+    0xb00327c898fb213f, 0xbf597fc7beef0ee4,
+    0xc6e00bf33da88fc2, 0xd5a79147930aa725,
+    0x06ca6351e003826f, 0x142929670a0e6e70,
+    0x27b70a8546d22ffc, 0x2e1b21385c26c926,
+    0x4d2c6dfc5ac42aed, 0x53380d139d95b3df,
+    0x650a73548baf63de, 0x766a0abb3c77b2a8,
+    0x81c2c92e47edaee6, 0x92722c851482353b,
+    0xa2bfe8a14cf10364, 0xa81a664bbc423001,
+    0xc24b8b70d0f89791, 0xc76c51a30654be30,
+    0xd192e819d6ef5218, 0xd69906245565a910,
+    0xf40e35855771202a, 0x106aa07032bbd1b8,
+    0x19a4c116b8d2d0c8, 0x1e376c085141ab53,
+    0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8,
+    0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb,
+    0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3,
+    0x748f82ee5defb2fc, 0x78a5636f43172f60,
+    0x84c87814a1f0ab72, 0x8cc702081a6439ec,
+    0x90befffa23631e28, 0xa4506cebde82bde9,
+    0xbef9a3f7b2c67915, 0xc67178f2e372532b,
+    0xca273eceea26619c, 0xd186b8c721c0c207,
+    0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178,
+    0x06f067aa72176fba, 0x0a637dc5a2c898a6,
+    0x113f9804bef90dae, 0x1b710b35131c471b,
+    0x28db77f523047d84, 0x32caab7b40c72493,
+    0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c,
+    0x4cc5d4becb3e42b6, 0x597f299cfc657e2a,
+    0x5fcb6fab3ad6faec, 0x6c44198c4a475817,
+}
+
+SHA256_CH :: #force_inline proc "contextless"(x, y, z: u32) -> u32 {
+    return (x & y) ~ (~x & z)
+}
+
+SHA256_MAJ :: #force_inline proc "contextless"(x, y, z: u32) -> u32 {
+    return (x & y) ~ (x & z) ~ (y & z)
+}
+
+SHA512_CH :: #force_inline proc "contextless"(x, y, z: u64) -> u64 {
+    return (x & y) ~ (~x & z)
+}
+
+SHA512_MAJ :: #force_inline proc "contextless"(x, y, z: u64) -> u64 {
+    return (x & y) ~ (x & z) ~ (y & z)
+}
+
+SHA256_F1 :: #force_inline proc "contextless"(x: u32) -> u32 {
+    return util.ROTR32(x, 2) ~ util.ROTR32(x, 13) ~ util.ROTR32(x, 22)
+}
+
+SHA256_F2 :: #force_inline proc "contextless"(x: u32) -> u32 {
+    return util.ROTR32(x, 6) ~ util.ROTR32(x, 11) ~ util.ROTR32(x, 25)
+}
+
+SHA256_F3 :: #force_inline proc "contextless"(x: u32) -> u32 {
+    return util.ROTR32(x, 7) ~ util.ROTR32(x, 18) ~ (x >> 3)
+}
+
+SHA256_F4 :: #force_inline proc "contextless"(x: u32) -> u32 {
+    return util.ROTR32(x, 17) ~ util.ROTR32(x, 19) ~ (x >> 10)
+}
+
+SHA512_F1 :: #force_inline proc "contextless"(x: u64) -> u64 {
+    return util.ROTR64(x, 28) ~ util.ROTR64(x, 34) ~ util.ROTR64(x, 39)
+}
+
+SHA512_F2 :: #force_inline proc "contextless"(x: u64) -> u64 {
+    return util.ROTR64(x, 14) ~ util.ROTR64(x, 18) ~ util.ROTR64(x, 41)
+}
+
+SHA512_F3 :: #force_inline proc "contextless"(x: u64) -> u64 {
+    return util.ROTR64(x, 1) ~ util.ROTR64(x, 8) ~ (x >> 7)
+}
+
+SHA512_F4 :: #force_inline proc "contextless"(x: u64) -> u64 {
+    return util.ROTR64(x, 19) ~ util.ROTR64(x, 61) ~ (x >> 6)
+}
+
+PACK32 :: #force_inline proc "contextless"(b: []byte, x: ^u32) {
+	x^ = u32(b[3]) | u32(b[2]) << 8 | u32(b[1]) << 16 | u32(b[0]) << 24
+}
+
+PACK64 :: #force_inline proc "contextless"(b: []byte, x: ^u64) {
+	x^ = u64(b[7]) | u64(b[6]) << 8 | u64(b[5]) << 16 | u64(b[4]) << 24 | u64(b[3]) << 32 | u64(b[2]) << 40 | u64(b[1]) << 48 | u64(b[0]) << 56
+}
+
+sha2_transf :: proc(ctx: ^$T, data: []byte, block_nb: uint) {
+	when T == Sha256_Context {
+		w: [64]u32
+		wv: [8]u32
+		t1, t2: u32
+	} else when T == Sha512_Context {
+		w: [80]u64
+		wv: [8]u64
+		t1, t2: u64
+	}
+
+	sub_block := make([]byte, len(data))
+	i, j: i32
+
+	for i = 0; i < i32(block_nb); i += 1 {
+		when T == Sha256_Context {
+			sub_block = data[i << 6:]
+		} else when T == Sha512_Context {
+			sub_block = data[i << 7:]
+		}
+
+		for j = 0; j < 16; j += 1 {
+			when T == Sha256_Context {
+				PACK32(sub_block[j << 2:], &w[j])
+			} else when T == Sha512_Context {
+				PACK64(sub_block[j << 3:], &w[j])
+			}
+		}
+
+		when T == Sha256_Context {
+			for j = 16; j < 64; j += 1 {
+				w[j] = SHA256_F4(w[j - 2]) + w[j - 7] + SHA256_F3(w[j - 15]) + w[j - 16]
+			}
+		} else when T == Sha512_Context {
+			for j = 16; j < 80; j += 1 {
+				w[j] = SHA512_F4(w[j - 2]) + w[j - 7] + SHA512_F3(w[j - 15]) + w[j - 16]
+			}
+		}
+
+		for j = 0; j < 8; j += 1 {
+			wv[j] = ctx.h[j]
+		}
+
+		when T == Sha256_Context {
+			for j = 0; j < 64; j += 1 {
+				t1 = wv[7] + SHA256_F2(wv[4]) + SHA256_CH(wv[4], wv[5], wv[6]) + sha256_k[j] + w[j]
+				t2 = SHA256_F1(wv[0]) + SHA256_MAJ(wv[0], wv[1], wv[2])
+				wv[7] = wv[6]
+				wv[6] = wv[5]
+				wv[5] = wv[4]
+				wv[4] = wv[3] + t1
+				wv[3] = wv[2]
+				wv[2] = wv[1]
+				wv[1] = wv[0]
+				wv[0] = t1 + t2
+			}
+		} else when T == Sha512_Context {
+			for j = 0; j < 80; j += 1 {
+				t1 = wv[7] + SHA512_F2(wv[4]) + SHA512_CH(wv[4], wv[5], wv[6]) + sha512_k[j] + w[j]
+				t2 = SHA512_F1(wv[0]) + SHA512_MAJ(wv[0], wv[1], wv[2])
+				wv[7] = wv[6]
+				wv[6] = wv[5]
+				wv[5] = wv[4]
+				wv[4] = wv[3] + t1
+				wv[3] = wv[2]
+				wv[2] = wv[1]
+				wv[1] = wv[0]
+				wv[0] = t1 + t2
+			}
+		}
+
+		for j = 0; j < 8; j += 1 {
+			ctx.h[j] += wv[j]
+		}
+	}
+}

core/crypto/sha3/sha3.odin

@@ -0,0 +1,271 @@
+package sha3
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the SHA3 hashing algorithm. The SHAKE functionality can be found in package shake.
+    If you wish to compute a Keccak hash, you can use the keccak package, it will use the original padding.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_sha3"
+
+/*
+    High level API
+*/
+
+// hash_string_224 will hash the given input and return the
+// computed hash
+hash_string_224 :: proc(data: string) -> [28]byte {
+    return hash_bytes_224(transmute([]byte)(data))
+}
+
+// hash_bytes_224 will hash the given input and return the
+// computed hash
+hash_bytes_224 :: proc(data: []byte) -> [28]byte {
+    hash: [28]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 28
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_224 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_224 :: proc(s: io.Stream) -> ([28]byte, bool) {
+    hash: [28]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 28
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([28]byte, bool) {
+    if !load_at_once {
+        return hash_stream_224(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_224(buf[:]), ok
+        }
+    }
+    return [28]byte{}, false
+}
+
+hash_224 :: proc {
+    hash_stream_224,
+    hash_file_224,
+    hash_bytes_224,
+    hash_string_224,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_384 will hash the given input and return the
+// computed hash
+hash_string_384 :: proc(data: string) -> [48]byte {
+    return hash_bytes_384(transmute([]byte)(data))
+}
+
+// hash_bytes_384 will hash the given input and return the
+// computed hash
+hash_bytes_384 :: proc(data: []byte) -> [48]byte {
+    hash: [48]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 48
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_384 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_384 :: proc(s: io.Stream) -> ([48]byte, bool) {
+    hash: [48]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 48
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([48]byte, bool) {
+    if !load_at_once {
+        return hash_stream_384(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_384(buf[:]), ok
+        }
+    }
+    return [48]byte{}, false
+}
+
+hash_384 :: proc {
+    hash_stream_384,
+    hash_file_384,
+    hash_bytes_384,
+    hash_string_384,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 64
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 64
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+Sha3_Context :: _sha3.Sha3_Context
+
+init :: proc(ctx: ^_sha3.Sha3_Context) {
+    _sha3.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_sha3.Sha3_Context, data: []byte) {
+    _sha3.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_sha3.Sha3_Context, hash: []byte) {
+    _sha3.final(ctx, hash)
+}

core/crypto/shake/shake.odin

@@ -0,0 +1,160 @@
+package shake
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the SHAKE hashing algorithm.
+    The SHA3 functionality can be found in package sha3.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_sha3"
+
+/*
+    High level API
+*/
+
+// hash_string_128 will hash the given input and return the
+// computed hash
+hash_string_128 :: proc(data: string) -> [16]byte {
+    return hash_bytes_128(transmute([]byte)(data))
+}
+
+// hash_bytes_128 will hash the given input and return the
+// computed hash
+hash_bytes_128 :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 16
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.shake_xof(&ctx)
+    _sha3.shake_out(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_128 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_128 :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 16
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.shake_xof(&ctx)
+    _sha3.shake_out(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_128 will read the file provided by the given handle
+// and compute a hash
+hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream_128(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_128(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash_128 :: proc {
+    hash_stream_128,
+    hash_file_128,
+    hash_bytes_128,
+    hash_string_128,
+}
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.shake_xof(&ctx)
+    _sha3.shake_out(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: _sha3.Sha3_Context
+    ctx.mdlen = 32
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.shake_xof(&ctx)
+    _sha3.shake_out(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+/*
+    Low level API
+*/
+
+Sha3_Context :: _sha3.Sha3_Context
+
+init :: proc(ctx: ^_sha3.Sha3_Context) {
+    _sha3.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_sha3.Sha3_Context, data: []byte) {
+    _sha3.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_sha3.Sha3_Context, hash: []byte) {
+    _sha3.shake_xof(ctx)
+    _sha3.shake_out(ctx, hash[:])
+}

core/crypto/sm3/sm3.odin

@@ -0,0 +1,227 @@
+package sm3
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the SM3 hashing algorithm, as defined in <https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../util"
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [32]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Sm3_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Sm3_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true 
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Sm3_Context) {
+    ctx.state[0] = IV[0]
+    ctx.state[1] = IV[1]
+    ctx.state[2] = IV[2]
+    ctx.state[3] = IV[3]
+    ctx.state[4] = IV[4]
+    ctx.state[5] = IV[5]
+    ctx.state[6] = IV[6]
+    ctx.state[7] = IV[7]
+}
+
+update :: proc(ctx: ^Sm3_Context, data: []byte) {
+    data := data
+    ctx.length += u64(len(data))
+
+    if ctx.bitlength > 0 {
+        n := copy(ctx.x[ctx.bitlength:], data[:])
+        ctx.bitlength += u64(n)
+        if ctx.bitlength == 64 {
+            block(ctx, ctx.x[:])
+            ctx.bitlength = 0
+        }
+        data = data[n:]
+    }
+    if len(data) >= 64 {
+        n := len(data) &~ (64 - 1)
+        block(ctx, data[:n])
+        data = data[n:]
+    }
+    if len(data) > 0 {
+        ctx.bitlength = u64(copy(ctx.x[:], data[:]))
+    }
+}
+
+final :: proc(ctx: ^Sm3_Context, hash: []byte) {
+    length := ctx.length
+
+    pad: [64]byte
+    pad[0] = 0x80
+    if length % 64 < 56 {
+        update(ctx, pad[0: 56 - length % 64])
+    } else {
+        update(ctx, pad[0: 64 + 56 - length % 64])
+    }
+
+    length <<= 3
+    util.PUT_U64_BE(pad[:], length)
+    update(ctx, pad[0: 8])
+    assert(ctx.bitlength == 0)
+
+    util.PUT_U32_BE(hash[0:],  ctx.state[0])
+    util.PUT_U32_BE(hash[4:],  ctx.state[1])
+    util.PUT_U32_BE(hash[8:],  ctx.state[2])
+    util.PUT_U32_BE(hash[12:], ctx.state[3])
+    util.PUT_U32_BE(hash[16:], ctx.state[4])
+    util.PUT_U32_BE(hash[20:], ctx.state[5])
+    util.PUT_U32_BE(hash[24:], ctx.state[6])
+    util.PUT_U32_BE(hash[28:], ctx.state[7])
+}
+
+/*
+    SM3 implementation
+*/
+
+Sm3_Context :: struct {
+    state:     [8]u32,
+    x:         [64]byte,
+    bitlength: u64,
+    length:    u64,
+}
+
+BLOCK_SIZE_IN_BYTES :: 64
+BLOCK_SIZE_IN_32    :: 16
+
+IV := [8]u32 {
+    0x7380166f, 0x4914b2b9, 0x172442d7, 0xda8a0600,
+    0xa96f30bc, 0x163138aa, 0xe38dee4d, 0xb0fb0e4e,
+}
+
+block :: proc "contextless" (ctx: ^Sm3_Context, buf: []byte) {
+    buf := buf
+
+    w:  [68]u32
+    wp: [64]u32
+
+    state0, state1, state2, state3 := ctx.state[0], ctx.state[1], ctx.state[2], ctx.state[3]
+    state4, state5, state6, state7 := ctx.state[4], ctx.state[5], ctx.state[6], ctx.state[7]
+
+    for len(buf) >= 64 {
+        for i := 0; i < 16; i += 1 {
+            j := i * 4
+            w[i] = u32(buf[j]) << 24 | u32(buf[j + 1]) << 16 | u32(buf[j + 2]) << 8 | u32(buf[j + 3])
+        }
+        for i := 16; i < 68; i += 1 {
+            p1v := w[i - 16] ~ w[i - 9] ~ util.ROTL32(w[i - 3], 15)
+            // @note(zh): inlined P1
+            w[i] = p1v ~ util.ROTL32(p1v, 15) ~ util.ROTL32(p1v, 23) ~ util.ROTL32(w[i - 13], 7) ~ w[i - 6]
+        }
+        for i := 0; i < 64; i += 1 {
+            wp[i] = w[i] ~ w[i + 4]
+        }
+
+        a, b, c, d := state0, state1, state2, state3
+        e, f, g, h := state4, state5, state6, state7
+
+        for i := 0; i < 16; i += 1 {
+            v1  := util.ROTL32(u32(a), 12)
+            ss1 := util.ROTL32(v1 + u32(e) + util.ROTL32(0x79cc4519, i), 7)
+            ss2 := ss1 ~ v1
+
+            // @note(zh): inlined FF1
+            tt1 := u32(a ~ b ~ c) + u32(d) + ss2 + wp[i]
+            // @note(zh): inlined GG1
+            tt2 := u32(e ~ f ~ g) + u32(h) + ss1 + w[i]
+
+            a, b, c, d = tt1, a, util.ROTL32(u32(b), 9), c
+            // @note(zh): inlined P0
+            e, f, g, h = (tt2 ~ util.ROTL32(tt2, 9) ~ util.ROTL32(tt2, 17)), e, util.ROTL32(u32(f), 19), g
+        }
+
+        for i := 16; i < 64; i += 1 {
+            v   := util.ROTL32(u32(a), 12)
+            ss1 := util.ROTL32(v + u32(e) + util.ROTL32(0x7a879d8a, i % 32), 7)
+            ss2 := ss1 ~ v
+
+            // @note(zh): inlined FF2
+            tt1 := u32(((a & b) | (a & c) | (b & c)) + d) + ss2 + wp[i]
+            // @note(zh): inlined GG2
+            tt2 := u32(((e & f) | ((~e) & g)) + h) + ss1 + w[i]
+
+            a, b, c, d = tt1, a, util.ROTL32(u32(b), 9), c
+            // @note(zh): inlined P0
+            e, f, g, h = (tt2 ~ util.ROTL32(tt2, 9) ~ util.ROTL32(tt2, 17)), e, util.ROTL32(u32(f), 19), g
+        }
+
+        state0 ~= a
+        state1 ~= b
+        state2 ~= c
+        state3 ~= d
+        state4 ~= e
+        state5 ~= f
+        state6 ~= g
+        state7 ~= h
+
+        buf = buf[64:]
+    }
+
+    ctx.state[0], ctx.state[1], ctx.state[2], ctx.state[3] = state0, state1, state2, state3
+    ctx.state[4], ctx.state[5], ctx.state[6], ctx.state[7] = state4, state5, state6, state7
+}

core/crypto/streebog/streebog.odin

@@ -0,0 +1,473 @@
+package streebog
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Streebog hashing algorithm, standardized as GOST R 34.11-2012 in RFC 6986 <https://datatracker.ietf.org/doc/html/rfc6986>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string_256 will hash the given input and return the
+// computed hash
+hash_string_256 :: proc(data: string) -> [32]byte {
+    return hash_bytes_256(transmute([]byte)(data))
+}
+
+// hash_bytes_256 will hash the given input and return the
+// computed hash
+hash_bytes_256 :: proc(data: []byte) -> [32]byte {
+    hash: [32]byte
+    ctx: Streebog_Context
+    ctx.is256 = true
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_256 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_256 :: proc(s: io.Stream) -> ([32]byte, bool) {
+    hash: [32]byte
+    ctx: Streebog_Context
+    ctx.is256 = true
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([32]byte, bool) {
+    if !load_at_once {
+        return hash_stream_256(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_256(buf[:]), ok
+        }
+    }
+    return [32]byte{}, false
+}
+
+hash_256 :: proc {
+    hash_stream_256,
+    hash_file_256,
+    hash_bytes_256,
+    hash_string_256,
+}
+
+// hash_string_512 will hash the given input and return the
+// computed hash
+hash_string_512 :: proc(data: string) -> [64]byte {
+    return hash_bytes_512(transmute([]byte)(data))
+}
+
+// hash_bytes_512 will hash the given input and return the
+// computed hash
+hash_bytes_512 :: proc(data: []byte) -> [64]byte {
+    hash: [64]byte
+    ctx: Streebog_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_512 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_512 :: proc(s: io.Stream) -> ([64]byte, bool) {
+    hash: [64]byte
+    ctx: Streebog_Context
+    init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            update(&ctx, buf[:read])
+        } 
+    }
+    final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+    if !load_at_once {
+        return hash_stream_512(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_512(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash_512 :: proc {
+    hash_stream_512,
+    hash_file_512,
+    hash_bytes_512,
+    hash_string_512,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^Streebog_Context) {
+	if ctx.is256 {
+		ctx.hash_size = 256
+		for _, i in ctx.h {
+			ctx.h[i] = 0x01
+		}
+	} else {
+		ctx.hash_size = 512
+	}
+	ctx.v_512[1] = 0x02
+}
+
+update :: proc(ctx: ^Streebog_Context, data: []byte) {
+	length := u64(len(data))
+	chk_size: u64
+	data := data
+	for (length > 63) && (ctx.buf_size == 0) {
+		stage2(ctx, data)
+		data = data[64:]
+		length -= 64
+	}
+
+	for length != 0 {
+		chk_size = 64 - ctx.buf_size
+		if chk_size > length {
+			chk_size = length
+		}
+		copy(ctx.buffer[ctx.buf_size:], data[:chk_size])
+		ctx.buf_size += chk_size
+		length -= chk_size
+		data = data[chk_size:]
+		if ctx.buf_size == 64 {
+			stage2(ctx, ctx.buffer[:])
+			ctx.buf_size = 0
+		}
+	}
+}
+
+final :: proc(ctx: ^Streebog_Context, hash: []byte) {
+	t: [64]byte
+	t[1] = byte((ctx.buf_size * 8) >> 8) & 0xff
+	t[0] = byte((ctx.buf_size) * 8) & 0xff
+
+	padding(ctx)
+
+	G(ctx.h[:], ctx.n[:], ctx.buffer[:])
+
+	add_mod_512(ctx.n[:], t[:], ctx.n[:])
+	add_mod_512(ctx.sigma[:], ctx.buffer[:], ctx.sigma[:])
+
+	G(ctx.h[:], ctx.v_0[:], ctx.n[:])
+	G(ctx.h[:], ctx.v_0[:], ctx.sigma[:])
+
+	if ctx.is256 {
+		copy(hash[:], ctx.h[32:])
+	} else {
+		copy(hash[:], ctx.h[:])
+	}
+}
+
+/*
+    Streebog implementation
+*/
+
+PI := [256]byte {
+	252, 238, 221, 17,  207, 110, 49,  22,  251, 196, 250, 218, 35,  197, 4,   77,
+	233, 119, 240, 219, 147, 46,  153, 186, 23,  54,  241, 187, 20,  205, 95,  193,
+	249, 24,  101, 90,  226, 92,  239, 33,  129, 28,  60,  66,  139, 1,   142, 79,
+	5,   132, 2,   174, 227, 106, 143, 160, 6,   11,  237, 152, 127, 212, 211, 31,
+	235, 52,  44,  81,  234, 200, 72,  171, 242, 42,  104, 162, 253, 58,  206, 204,
+	181, 112, 14,  86,  8,   12,  118, 18,  191, 114, 19,  71,  156, 183, 93,  135,
+	21,  161, 150, 41,  16,  123, 154, 199, 243, 145, 120, 111, 157, 158, 178, 177,
+	50,  117, 25,  61,  255, 53,  138, 126, 109, 84,  198, 128, 195, 189, 13,  87,
+	223, 245, 36,  169, 62,  168, 67,  201, 215, 121, 214, 246, 124, 34,  185, 3,
+	224, 15,  236, 222, 122, 148, 176, 188, 220, 232, 40,  80,  78,  51,  10,  74,
+	167, 151, 96,  115, 30,  0,   98,  68,  26,  184, 56,  130, 100, 159, 38,  65,
+	173, 69,  70,  146, 39,  94,  85,  47,  140, 163, 165, 125, 105, 213, 149, 59,
+	7,   88,  179, 64,  134, 172, 29,  247, 48,  55,  107, 228,	136, 217, 231, 137,
+	225, 27,  131, 73,  76,  63,  248, 254, 141, 83,  170, 144, 202, 216, 133, 97,
+	32,  113, 103, 164, 45,  43,  9,   91,  203, 155, 37,  208, 190, 229, 108, 82,
+	89,  166, 116, 210,	230, 244, 180, 192, 209, 102, 175, 194, 57,  75,  99,  182,
+}
+
+TAU := [64]byte {
+	0,  8, 16, 24, 32, 40, 48, 56,
+	1,  9, 17, 25, 33, 41, 49, 57,
+	2, 10, 18, 26, 34, 42, 50, 58,
+	3, 11, 19, 27, 35, 43, 51, 59,
+	4, 12, 20, 28, 36, 44, 52, 60,
+	5, 13, 21, 29, 37, 45, 53, 61,
+	6, 14, 22, 30, 38, 46, 54, 62,
+	7, 15, 23, 31, 39, 47, 55, 63,
+}
+
+STREEBOG_A := [64]u64 {
+	0x8e20faa72ba0b470, 0x47107ddd9b505a38, 0xad08b0e0c3282d1c, 0xd8045870ef14980e,
+	0x6c022c38f90a4c07, 0x3601161cf205268d, 0x1b8e0b0e798c13c8, 0x83478b07b2468764,
+	0xa011d380818e8f40, 0x5086e740ce47c920, 0x2843fd2067adea10, 0x14aff010bdd87508,
+	0x0ad97808d06cb404, 0x05e23c0468365a02, 0x8c711e02341b2d01, 0x46b60f011a83988e,
+	0x90dab52a387ae76f, 0x486dd4151c3dfdb9, 0x24b86a840e90f0d2, 0x125c354207487869,
+	0x092e94218d243cba, 0x8a174a9ec8121e5d, 0x4585254f64090fa0, 0xaccc9ca9328a8950,
+	0x9d4df05d5f661451, 0xc0a878a0a1330aa6, 0x60543c50de970553, 0x302a1e286fc58ca7,
+	0x18150f14b9ec46dd, 0x0c84890ad27623e0, 0x0642ca05693b9f70, 0x0321658cba93c138,
+	0x86275df09ce8aaa8, 0x439da0784e745554, 0xafc0503c273aa42a, 0xd960281e9d1d5215,
+	0xe230140fc0802984, 0x71180a8960409a42, 0xb60c05ca30204d21, 0x5b068c651810a89e,
+	0x456c34887a3805b9, 0xac361a443d1c8cd2, 0x561b0d22900e4669, 0x2b838811480723ba,
+	0x9bcf4486248d9f5d, 0xc3e9224312c8c1a0, 0xeffa11af0964ee50, 0xf97d86d98a327728,
+	0xe4fa2054a80b329c, 0x727d102a548b194e, 0x39b008152acb8227, 0x9258048415eb419d,
+	0x492c024284fbaec0, 0xaa16012142f35760, 0x550b8e9e21f7a530, 0xa48b474f9ef5dc18,
+	0x70a6a56e2440598e, 0x3853dc371220a247, 0x1ca76e95091051ad, 0x0edd37c48a08a6d8,
+	0x07e095624504536c, 0x8d70c431ac02a736, 0xc83862965601dd1b, 0x641c314b2b8ee083,
+}
+
+STREEBOG_C := [12][64]byte { 
+	{
+		0x07, 0x45, 0xa6, 0xf2, 0x59, 0x65, 0x80, 0xdd,
+		0x23, 0x4d, 0x74, 0xcc, 0x36, 0x74, 0x76, 0x05,
+		0x15, 0xd3, 0x60, 0xa4, 0x08, 0x2a, 0x42, 0xa2,
+		0x01, 0x69, 0x67, 0x92, 0x91, 0xe0, 0x7c, 0x4b,
+		0xfc, 0xc4, 0x85, 0x75, 0x8d, 0xb8, 0x4e, 0x71,
+		0x16, 0xd0, 0x45, 0x2e, 0x43, 0x76, 0x6a, 0x2f,
+		0x1f, 0x7c, 0x65, 0xc0, 0x81, 0x2f, 0xcb, 0xeb,
+		0xe9, 0xda, 0xca, 0x1e, 0xda, 0x5b, 0x08, 0xb1,
+	},
+	{
+		0xb7, 0x9b, 0xb1, 0x21, 0x70, 0x04, 0x79, 0xe6,
+		0x56, 0xcd, 0xcb, 0xd7, 0x1b, 0xa2, 0xdd, 0x55,
+		0xca, 0xa7, 0x0a, 0xdb, 0xc2, 0x61, 0xb5, 0x5c,
+		0x58, 0x99, 0xd6, 0x12, 0x6b, 0x17, 0xb5, 0x9a,
+		0x31, 0x01, 0xb5, 0x16, 0x0f, 0x5e, 0xd5, 0x61,
+		0x98, 0x2b, 0x23, 0x0a, 0x72, 0xea, 0xfe, 0xf3,
+		0xd7, 0xb5, 0x70, 0x0f, 0x46, 0x9d, 0xe3, 0x4f,
+		0x1a, 0x2f, 0x9d, 0xa9, 0x8a, 0xb5, 0xa3, 0x6f,
+	},
+	{
+		0xb2, 0x0a, 0xba, 0x0a, 0xf5, 0x96, 0x1e, 0x99,
+		0x31, 0xdb, 0x7a, 0x86, 0x43, 0xf4, 0xb6, 0xc2,
+		0x09, 0xdb, 0x62, 0x60, 0x37, 0x3a, 0xc9, 0xc1,
+		0xb1, 0x9e, 0x35, 0x90, 0xe4, 0x0f, 0xe2, 0xd3,
+		0x7b, 0x7b, 0x29, 0xb1, 0x14, 0x75, 0xea, 0xf2,
+		0x8b, 0x1f, 0x9c, 0x52, 0x5f, 0x5e, 0xf1, 0x06,
+		0x35, 0x84, 0x3d, 0x6a, 0x28, 0xfc, 0x39, 0x0a,
+		0xc7, 0x2f, 0xce, 0x2b, 0xac, 0xdc, 0x74, 0xf5,
+	},
+	{
+		0x2e, 0xd1, 0xe3, 0x84, 0xbc, 0xbe, 0x0c, 0x22,
+		0xf1, 0x37, 0xe8, 0x93, 0xa1, 0xea, 0x53, 0x34,
+		0xbe, 0x03, 0x52, 0x93, 0x33, 0x13, 0xb7, 0xd8,
+		0x75, 0xd6, 0x03, 0xed, 0x82, 0x2c, 0xd7, 0xa9,
+		0x3f, 0x35, 0x5e, 0x68, 0xad, 0x1c, 0x72, 0x9d,
+		0x7d, 0x3c, 0x5c, 0x33, 0x7e, 0x85, 0x8e, 0x48,
+		0xdd, 0xe4, 0x71, 0x5d, 0xa0, 0xe1, 0x48, 0xf9,
+		0xd2, 0x66, 0x15, 0xe8, 0xb3, 0xdf, 0x1f, 0xef,
+	},
+	{
+		0x57, 0xfe, 0x6c, 0x7c, 0xfd, 0x58, 0x17, 0x60,
+		0xf5, 0x63, 0xea, 0xa9, 0x7e, 0xa2, 0x56, 0x7a,
+		0x16, 0x1a, 0x27, 0x23, 0xb7, 0x00, 0xff, 0xdf,
+		0xa3, 0xf5, 0x3a, 0x25, 0x47, 0x17, 0xcd, 0xbf,
+		0xbd, 0xff, 0x0f, 0x80, 0xd7, 0x35, 0x9e, 0x35,
+		0x4a, 0x10, 0x86, 0x16, 0x1f, 0x1c, 0x15, 0x7f,
+		0x63, 0x23, 0xa9, 0x6c, 0x0c, 0x41, 0x3f, 0x9a,
+		0x99, 0x47, 0x47, 0xad, 0xac, 0x6b, 0xea, 0x4b,
+	},
+	{
+		0x6e, 0x7d, 0x64, 0x46, 0x7a, 0x40, 0x68, 0xfa,
+		0x35, 0x4f, 0x90, 0x36, 0x72, 0xc5, 0x71, 0xbf,
+		0xb6, 0xc6, 0xbe, 0xc2, 0x66, 0x1f, 0xf2, 0x0a,
+		0xb4, 0xb7, 0x9a, 0x1c, 0xb7, 0xa6, 0xfa, 0xcf,
+		0xc6, 0x8e, 0xf0, 0x9a, 0xb4, 0x9a, 0x7f, 0x18,
+		0x6c, 0xa4, 0x42, 0x51, 0xf9, 0xc4, 0x66, 0x2d,
+		0xc0, 0x39, 0x30, 0x7a, 0x3b, 0xc3, 0xa4, 0x6f,
+		0xd9, 0xd3, 0x3a, 0x1d, 0xae, 0xae, 0x4f, 0xae,
+	},
+	{
+		0x93, 0xd4, 0x14, 0x3a, 0x4d, 0x56, 0x86, 0x88,
+		0xf3, 0x4a, 0x3c, 0xa2, 0x4c, 0x45, 0x17, 0x35,
+		0x04, 0x05, 0x4a, 0x28, 0x83, 0x69, 0x47, 0x06,
+		0x37, 0x2c, 0x82, 0x2d, 0xc5, 0xab, 0x92, 0x09,
+		0xc9, 0x93, 0x7a, 0x19, 0x33, 0x3e, 0x47, 0xd3,
+		0xc9, 0x87, 0xbf, 0xe6, 0xc7, 0xc6, 0x9e, 0x39,
+		0x54, 0x09, 0x24, 0xbf, 0xfe, 0x86, 0xac, 0x51,
+		0xec, 0xc5, 0xaa, 0xee, 0x16, 0x0e, 0xc7, 0xf4,
+	},
+	{
+		0x1e, 0xe7, 0x02, 0xbf, 0xd4, 0x0d, 0x7f, 0xa4,
+		0xd9, 0xa8, 0x51, 0x59, 0x35, 0xc2, 0xac, 0x36,
+		0x2f, 0xc4, 0xa5, 0xd1, 0x2b, 0x8d, 0xd1, 0x69,
+		0x90, 0x06, 0x9b, 0x92, 0xcb, 0x2b, 0x89, 0xf4,
+		0x9a, 0xc4, 0xdb, 0x4d, 0x3b, 0x44, 0xb4, 0x89,
+		0x1e, 0xde, 0x36, 0x9c, 0x71, 0xf8, 0xb7, 0x4e,
+		0x41, 0x41, 0x6e, 0x0c, 0x02, 0xaa, 0xe7, 0x03,
+		0xa7, 0xc9, 0x93, 0x4d, 0x42, 0x5b, 0x1f, 0x9b,
+	},
+	{
+		0xdb, 0x5a, 0x23, 0x83, 0x51, 0x44, 0x61, 0x72,
+		0x60, 0x2a, 0x1f, 0xcb, 0x92, 0xdc, 0x38, 0x0e,
+		0x54, 0x9c, 0x07, 0xa6, 0x9a, 0x8a, 0x2b, 0x7b,
+		0xb1, 0xce, 0xb2, 0xdb, 0x0b, 0x44, 0x0a, 0x80,
+		0x84, 0x09, 0x0d, 0xe0, 0xb7, 0x55, 0xd9, 0x3c,
+		0x24, 0x42, 0x89, 0x25, 0x1b, 0x3a, 0x7d, 0x3a,
+		0xde, 0x5f, 0x16, 0xec, 0xd8, 0x9a, 0x4c, 0x94,
+		0x9b, 0x22, 0x31, 0x16, 0x54, 0x5a, 0x8f, 0x37,
+	},
+	{
+		0xed, 0x9c, 0x45, 0x98, 0xfb, 0xc7, 0xb4, 0x74,
+		0xc3, 0xb6, 0x3b, 0x15, 0xd1, 0xfa, 0x98, 0x36,
+		0xf4, 0x52, 0x76, 0x3b, 0x30, 0x6c, 0x1e, 0x7a,
+		0x4b, 0x33, 0x69, 0xaf, 0x02, 0x67, 0xe7, 0x9f,
+		0x03, 0x61, 0x33, 0x1b, 0x8a, 0xe1, 0xff, 0x1f,
+		0xdb, 0x78, 0x8a, 0xff, 0x1c, 0xe7, 0x41, 0x89,
+		0xf3, 0xf3, 0xe4, 0xb2, 0x48, 0xe5, 0x2a, 0x38,
+		0x52, 0x6f, 0x05, 0x80, 0xa6, 0xde, 0xbe, 0xab,
+	},
+	{
+		0x1b, 0x2d, 0xf3, 0x81, 0xcd, 0xa4, 0xca, 0x6b,
+		0x5d, 0xd8, 0x6f, 0xc0, 0x4a, 0x59, 0xa2, 0xde,
+		0x98, 0x6e, 0x47, 0x7d, 0x1d, 0xcd, 0xba, 0xef,
+		0xca, 0xb9, 0x48, 0xea, 0xef, 0x71, 0x1d, 0x8a,
+		0x79, 0x66, 0x84, 0x14, 0x21, 0x80, 0x01, 0x20,
+		0x61, 0x07, 0xab, 0xeb, 0xbb, 0x6b, 0xfa, 0xd8,
+		0x94, 0xfe, 0x5a, 0x63, 0xcd, 0xc6, 0x02, 0x30,
+		0xfb, 0x89, 0xc8, 0xef, 0xd0, 0x9e, 0xcd, 0x7b,
+	},
+	{
+		0x20, 0xd7, 0x1b, 0xf1, 0x4a, 0x92, 0xbc, 0x48,
+		0x99, 0x1b, 0xb2, 0xd9, 0xd5, 0x17, 0xf4, 0xfa,
+		0x52, 0x28, 0xe1, 0x88, 0xaa, 0xa4, 0x1d, 0xe7,
+		0x86, 0xcc, 0x91, 0x18, 0x9d, 0xef, 0x80, 0x5d,
+		0x9b, 0x9f, 0x21, 0x30, 0xd4, 0x12, 0x20, 0xf8,
+		0x77, 0x1d, 0xdf, 0xbc, 0x32, 0x3c, 0xa4, 0xcd,
+		0x7a, 0xb1, 0x49, 0x04, 0xb0, 0x80, 0x13, 0xd2,
+		0xba, 0x31, 0x16, 0xf1, 0x67, 0xe7, 0x8e, 0x37,
+	},
+}
+
+Streebog_Context :: struct {
+	buffer:    [64]byte,
+	h:         [64]byte,
+	n:         [64]byte,
+	sigma:     [64]byte,
+	v_0:       [64]byte,
+	v_512:     [64]byte,
+	buf_size:  u64,
+	hash_size: int,
+	is256:	   bool,
+}
+
+add_mod_512 :: proc(first_vector, second_vector, result_vector: []byte) {
+	t: i32 = 0
+	for i: i32 = 0; i < 64; i += 1 {
+		t = i32(first_vector[i]) + i32(second_vector[i]) + (t >> 8)
+		result_vector[i] = byte(t & 0xff)
+	}
+}
+
+X :: #force_inline proc(a, k, out: []byte) {
+	for i := 0; i < 64; i += 1 {
+		out[i] = a[i] ~ k[i]
+	}
+}
+
+S :: #force_inline proc(state: []byte) {
+	t: [64]byte
+	for i: i32 = 63; i >= 0; i -= 1 {
+		t[i] = PI[state[i]]
+	}
+	copy(state, t[:])
+}
+
+P :: #force_inline proc(state: []byte) {
+	t: [64]byte
+	for i: i32 = 63; i >= 0; i -= 1 {
+		t[i] = state[TAU[i]]
+	}
+	copy(state, t[:])
+}
+
+L :: #force_inline proc(state: []byte) {
+	ins := util.cast_slice([]u64, state)
+	out: [8]u64
+	for i: i32 = 7; i >= 0; i -= 1 {
+		for j: i32 = 63; j >= 0; j -= 1 {
+			if (ins[i] >> u32(j)) & 1 != 0 {
+				out[i] ~= STREEBOG_A[63 - j]
+			}	
+		}
+	}
+	copy(state, util.cast_slice([]byte, out[:]))
+}
+
+E :: #force_inline proc(K, m, state: []byte) {
+	X(m, K, state)
+	for i: i32 = 0; i < 12; i += 1 {
+		S(state)
+		P(state)
+		L(state)
+		get_key(K, i)
+		X(state, K, state)
+	}
+}
+
+get_key :: #force_inline proc(K: []byte, i: i32) {
+	X(K, STREEBOG_C[i][:], K)
+	S(K)
+	P(K)
+	L(K)
+}
+
+G :: #force_inline proc(h, N, m: []byte) {
+	t, K: [64]byte
+	X(N, h, K[:])
+	S(K[:])
+	P(K[:])
+	L(K[:])
+	E(K[:], m, t[:])
+	X(t[:], h, t[:])
+	X(t[:], m, h)
+}
+
+stage2 :: proc(ctx: ^Streebog_Context, m: []byte) {
+	G(ctx.h[:], ctx.n[:], m)
+	add_mod_512(ctx.n[:], ctx.v_512[:], ctx.n[:])
+	add_mod_512(ctx.sigma[:], m, ctx.sigma[:])
+}
+
+padding :: proc(ctx: ^Streebog_Context) {
+	if ctx.buf_size < 64 {
+		t: [64]byte
+		copy(t[:], ctx.buffer[:int(ctx.buf_size)])
+		t[ctx.buf_size] = 0x01
+		copy(ctx.buffer[:], t[:])
+	}
+}

core/crypto/tiger/tiger.odin

@@ -0,0 +1,213 @@
+package tiger
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the Tiger1 variant of the Tiger hashing algorithm as defined in <https://www.cs.technion.ac.il/~biham/Reports/Tiger/>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_tiger"
+
+/*
+    High level API
+*/
+
+// hash_string_128 will hash the given input and return the
+// computed hash
+hash_string_128 :: proc(data: string) -> [16]byte {
+    return hash_bytes_128(transmute([]byte)(data))
+}
+
+// hash_bytes_128 will hash the given input and return the
+// computed hash
+hash_bytes_128 :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_128 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_128 :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_128 will read the file provided by the given handle
+// and compute a hash
+hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream_128(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_128(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash_128 :: proc {
+    hash_stream_128,
+    hash_file_128,
+    hash_bytes_128,
+    hash_string_128,
+}
+
+// hash_string_160 will hash the given input and return the
+// computed hash
+hash_string_160 :: proc(data: string) -> [20]byte {
+    return hash_bytes_160(transmute([]byte)(data))
+}
+
+// hash_bytes_160 will hash the given input and return the
+// computed hash
+hash_bytes_160 :: proc(data: []byte) -> [20]byte {
+    hash: [20]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_160 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_160 :: proc(s: io.Stream) -> ([20]byte, bool) {
+    hash: [20]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_160 will read the file provided by the given handle
+// and compute a hash
+hash_file_160 :: proc(hd: os.Handle, load_at_once := false) -> ([20]byte, bool) {
+    if !load_at_once {
+        return hash_stream_160(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_160(buf[:]), ok
+        }
+    }
+    return [20]byte{}, false
+}
+
+hash_160 :: proc {
+    hash_stream_160,
+    hash_file_160,
+    hash_bytes_160,
+    hash_string_160,
+}
+
+// hash_string_192 will hash the given input and return the
+// computed hash
+hash_string_192 :: proc(data: string) -> [24]byte {
+    return hash_bytes_192(transmute([]byte)(data))
+}
+
+// hash_bytes_192 will hash the given input and return the
+// computed hash
+hash_bytes_192 :: proc(data: []byte) -> [24]byte {
+    hash: [24]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_192 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_192 :: proc(s: io.Stream) -> ([24]byte, bool) {
+    hash: [24]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 1
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_192 will read the file provided by the given handle
+// and compute a hash
+hash_file_192 :: proc(hd: os.Handle, load_at_once := false) -> ([24]byte, bool) {
+    if !load_at_once {
+        return hash_stream_192(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_192(buf[:]), ok
+        }
+    }
+    return [24]byte{}, false
+}
+
+hash_192 :: proc {
+    hash_stream_192,
+    hash_file_192,
+    hash_bytes_192,
+    hash_string_192,
+}
+
+/*
+    Low level API
+*/
+
+Tiger_Context :: _tiger.Tiger_Context
+
+init :: proc(ctx: ^_tiger.Tiger_Context) {
+    ctx.ver = 1
+    _tiger.init(ctx)
+}
+
+update :: proc(ctx: ^_tiger.Tiger_Context, data: []byte) {
+    _tiger.update(ctx, data)
+}
+
+final :: proc(ctx: ^_tiger.Tiger_Context, hash: []byte) {
+    _tiger.final(ctx, hash)
+}

core/crypto/tiger2/tiger2.odin

@@ -0,0 +1,213 @@
+package tiger2
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the Tiger2 variant of the Tiger hashing algorithm as defined in <https://www.cs.technion.ac.il/~biham/Reports/Tiger/>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_tiger"
+
+/*
+    High level API
+*/
+
+// hash_string_128 will hash the given input and return the
+// computed hash
+hash_string_128 :: proc(data: string) -> [16]byte {
+    return hash_bytes_128(transmute([]byte)(data))
+}
+
+// hash_bytes_128 will hash the given input and return the
+// computed hash
+hash_bytes_128 :: proc(data: []byte) -> [16]byte {
+    hash: [16]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_128 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_128 :: proc(s: io.Stream) -> ([16]byte, bool) {
+    hash: [16]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_128 will read the file provided by the given handle
+// and compute a hash
+hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([16]byte, bool) {
+    if !load_at_once {
+        return hash_stream_128(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_128(buf[:]), ok
+        }
+    }
+    return [16]byte{}, false
+}
+
+hash_128 :: proc {
+    hash_stream_128,
+    hash_file_128,
+    hash_bytes_128,
+    hash_string_128,
+}
+
+// hash_string_160 will hash the given input and return the
+// computed hash
+hash_string_160 :: proc(data: string) -> [20]byte {
+    return hash_bytes_160(transmute([]byte)(data))
+}
+
+// hash_bytes_160 will hash the given input and return the
+// computed hash
+hash_bytes_160 :: proc(data: []byte) -> [20]byte {
+    hash: [20]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_160 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_160 :: proc(s: io.Stream) -> ([20]byte, bool) {
+    hash: [20]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_160 will read the file provided by the given handle
+// and compute a hash
+hash_file_160 :: proc(hd: os.Handle, load_at_once := false) -> ([20]byte, bool) {
+    if !load_at_once {
+        return hash_stream_160(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_160(buf[:]), ok
+        }
+    }
+    return [20]byte{}, false
+}
+
+hash_160 :: proc {
+    hash_stream_160,
+    hash_file_160,
+    hash_bytes_160,
+    hash_string_160,
+}
+
+// hash_string_192 will hash the given input and return the
+// computed hash
+hash_string_192 :: proc(data: string) -> [24]byte {
+    return hash_bytes_192(transmute([]byte)(data))
+}
+
+// hash_bytes_192 will hash the given input and return the
+// computed hash
+hash_bytes_192 :: proc(data: []byte) -> [24]byte {
+    hash: [24]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    _tiger.update(&ctx, data)
+    _tiger.final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream_192 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_192 :: proc(s: io.Stream) -> ([24]byte, bool) {
+    hash: [24]byte
+    ctx: _tiger.Tiger_Context
+    ctx.ver = 2
+    _tiger.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = s->impl_read(buf)
+        if read > 0 {
+            _tiger.update(&ctx, buf[:read])
+        } 
+    }
+    _tiger.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_192 will read the file provided by the given handle
+// and compute a hash
+hash_file_192 :: proc(hd: os.Handle, load_at_once := false) -> ([24]byte, bool) {
+    if !load_at_once {
+        return hash_stream_192(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_192(buf[:]), ok
+        }
+    }
+    return [24]byte{}, false
+}
+
+hash_192 :: proc {
+    hash_stream_192,
+    hash_file_192,
+    hash_bytes_192,
+    hash_string_192,
+}
+
+/*
+    Low level API
+*/
+
+Tiger_Context :: _tiger.Tiger_Context
+
+init :: proc(ctx: ^_tiger.Tiger_Context) {
+    ctx.ver = 2
+    _tiger.init(ctx)
+}
+
+update :: proc(ctx: ^_tiger.Tiger_Context, data: []byte) {
+    _tiger.update(ctx, data)
+}
+
+final :: proc(ctx: ^_tiger.Tiger_Context, hash: []byte) {
+    _tiger.final(ctx, hash)
+}

core/crypto/util/util.odin

@@ -0,0 +1,144 @@
+package util
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Various utility procedures
+*/
+
+import "core:mem"
+
+// @note(bp): this can replace the other two
+cast_slice :: #force_inline proc "contextless" ($D: typeid/[]$DE, src: $S/[]$SE) -> D {
+    src := src
+    dst := (^mem.Raw_Slice)(&src)
+
+    when size_of(DE) < size_of(SE) {
+        when size_of(DE) % size_of(SE) == 0 {
+            dst.len /= size_of(SE) / size_of(DE)
+        } else {
+            dst.len *= size_of(SE)
+            dst.len /= size_of(DE)
+        }
+    } else when size_of(DE) > size_of(SE) {
+        when size_of(DE) % size_of(SE) == 0 {
+            dst.len *= size_of(DE) / size_of(SE)
+        } else {
+            dst.len *= size_of(SE)
+            dst.len /= size_of(DE)
+        }
+    } else when size_of(DE) != size_of(SE) {
+        #assert(size_of(DE) % size_of(SE) == 0, "Different size detected")
+        dst.len *= size_of(SE)
+        dst.len /= size_of(DE)
+    }
+
+    return (^D)(dst)^
+}
+
+bytes_to_slice :: #force_inline proc "contextless" ($T: typeid/[]$E, bytes: []byte) -> T {
+    s := transmute(mem.Raw_Slice)bytes
+    s.len /= size_of(E)
+    return transmute(T)s
+}
+
+slice_to_bytes :: #force_inline proc "contextless" (slice: $E/[]$T) -> []byte {
+    s := transmute(mem.Raw_Slice)slice
+    s.len *= size_of(T)
+    return transmute([]byte)s
+}
+
+ROTL16 :: #force_inline proc "contextless" (a, b: u16) -> u16 {
+    return ((a << b) | (a >> (16 - b)))
+}
+
+ROTR16 :: #force_inline proc "contextless" (a, b: u16) -> u16 {
+    return ((a >> b) | (a << (16 - b)))
+}
+
+ROTL32 :: #force_inline proc "contextless"(a: u32, b: int) -> u32 {
+    s := uint(b) & 31
+    return (a << s) | (a >> (32 - s))
+}
+
+ROTR32 :: #force_inline proc "contextless" (a: u32, b: int) -> u32 {
+    s := uint(b) & 31
+    return (a >> s) | (a << (32 - s))
+}
+
+ROTL64 :: #force_inline proc "contextless" (a, b: u64) -> u64 {
+    return ((a << b) | (a >> (64 - b)))
+}
+
+ROTR64 :: #force_inline proc "contextless" (a, b: u64) -> u64 {
+    return ((a >> b) | (a << (64 - b)))
+}
+
+ROTL128 :: #force_inline proc "contextless" (a, b, c, d: ^u32, n: uint) {
+    a, b, c, d := a, b, c, d
+    t := a^ >> (32 - n)
+    a^ = ((a^ << n) | (b^ >> (32 - n)))
+    b^ = ((b^ << n) | (c^ >> (32 - n)))
+    c^ = ((c^ << n) | (d^ >> (32 - n)))
+    d^ = ((d^ << n) | t)
+}
+
+U32_LE :: #force_inline proc "contextless" (b: []byte) -> u32 {
+    return u32(b[0]) | u32(b[1]) << 8 | u32(b[2]) << 16 | u32(b[3]) << 24
+}
+
+U64_LE :: #force_inline proc "contextless" (b: []byte) -> u64 {
+    return u64(b[0])       | u64(b[1]) << 8  | u64(b[2]) << 16 | u64(b[3]) << 24 |
+           u64(b[4]) << 32 | u64(b[5]) << 40 | u64(b[6]) << 48 | u64(b[7]) << 56
+}
+
+U64_BE :: #force_inline proc "contextless" (b: []byte) -> u64 {
+    return u64(b[7])       | u64(b[6]) << 8  | u64(b[5]) << 16 | u64(b[4]) << 24 |
+           u64(b[3]) << 32 | u64(b[2]) << 40 | u64(b[1]) << 48 | u64(b[0]) << 56
+}
+
+PUT_U64_LE :: #force_inline proc "contextless" (b: []byte, v: u64) {
+    b[0] = byte(v)
+    b[1] = byte(v >> 8)
+    b[2] = byte(v >> 16)
+    b[3] = byte(v >> 24)
+    b[4] = byte(v >> 32)
+    b[5] = byte(v >> 40)
+    b[6] = byte(v >> 48)
+    b[7] = byte(v >> 56)
+}
+
+PUT_U32_LE :: #force_inline proc "contextless" (b: []byte, v: u32) {
+    b[0] = byte(v)
+    b[1] = byte(v >> 8)
+    b[2] = byte(v >> 16)
+    b[3] = byte(v >> 24)
+}
+
+PUT_U32_BE :: #force_inline proc "contextless" (b: []byte, v: u32) {
+    b[0] = byte(v >> 24)
+    b[1] = byte(v >> 16)
+    b[2] = byte(v >> 8)
+    b[3] = byte(v)
+}
+
+PUT_U64_BE :: #force_inline proc "contextless" (b: []byte, v: u64) {
+    b[0] = byte(v >> 56)
+    b[1] = byte(v >> 48)
+    b[2] = byte(v >> 40)
+    b[3] = byte(v >> 32)
+    b[4] = byte(v >> 24)
+    b[5] = byte(v >> 16)
+    b[6] = byte(v >> 8)
+    b[7] = byte(v)
+}
+
+XOR_BUF :: #force_inline proc "contextless" (input, output: []byte) {
+    for i := 0; i < len(input); i += 1 {
+        output[i] ~= input[i]
+    }
+}

core/crypto/whirlpool/whirlpool.odin

@@ -0,0 +1,784 @@
+package whirlpool
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Whirlpool hashing algorithm, as defined in <https://web.archive.org/web/20171129084214/http://www.larc.usp.br/~pbarreto/WhirlpoolPage.html>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+// hash_string will hash the given input and return the
+// computed hash
+hash_string :: proc(data: string) -> [64]byte {
+    return hash_bytes(transmute([]byte)(data))
+}
+
+// hash_bytes will hash the given input and return the
+// computed hash
+hash_bytes :: proc(data: []byte) -> [64]byte {
+	hash: [64]byte
+	ctx: Whirlpool_Context
+    // init(&ctx) No-op
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_stream will read the stream in chunks and compute a
+// hash from its contents
+hash_stream :: proc(s: io.Stream) -> ([64]byte, bool) {
+	hash: [64]byte
+	ctx: Whirlpool_Context
+	// init(&ctx) No-op
+	buf := make([]byte, 512)
+	defer delete(buf)
+	read := 1
+	for read > 0 {
+	    read, _ = s->impl_read(buf)
+	    if read > 0 {
+			update(&ctx, buf[:read])
+	    } 
+	}
+	final(&ctx, hash[:])
+	return hash, true
+}
+
+// hash_file will read the file provided by the given handle
+// and compute a hash
+hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([64]byte, bool) {
+	if !load_at_once {
+        return hash_stream(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes(buf[:]), ok
+        }
+    }
+    return [64]byte{}, false
+}
+
+hash :: proc {
+    hash_stream,
+    hash_file,
+    hash_bytes,
+    hash_string,
+}
+
+/*
+    Low level API
+*/
+
+@(warning="Init is a no-op for Whirlpool")
+init :: proc(ctx: ^Whirlpool_Context) {
+	// No action needed here
+}
+
+update :: proc(ctx: ^Whirlpool_Context, source: []byte) {
+    source_pos: int
+    nn := len(source)
+    source_bits := u64(nn * 8)
+    source_gap := u32((8 - (int(source_bits & 7))) & 7)
+    buffer_rem := uint(ctx.buffer_bits & 7)
+    b: u32
+
+	for i, carry, value := 31, u32(0), u32(source_bits); i >= 0 && (carry != 0 || value != 0); i -= 1 {
+		carry += u32(ctx.bitlength[i]) + (u32(value & 0xff))
+		ctx.bitlength[i] = byte(carry)
+		carry >>= 8
+		value >>= 8
+	}
+
+	for source_bits > 8 {
+		b = u32(u32((source[source_pos] << source_gap) & 0xff) | u32((source[source_pos+1] & 0xff) >> (8 - source_gap)))
+
+		ctx.buffer[ctx.buffer_pos] |= u8(b >> buffer_rem)
+		ctx.buffer_pos += 1
+		ctx.buffer_bits += int(8 - buffer_rem)
+
+		if ctx.buffer_bits == 512 {
+			transform(ctx)
+			ctx.buffer_bits = 0
+			ctx.buffer_pos = 0
+		}
+		ctx.buffer[ctx.buffer_pos] = byte(b << (8 - buffer_rem))
+		ctx.buffer_bits += int(buffer_rem)
+		source_bits -= 8
+		source_pos += 1
+	}
+
+	if source_bits > 0 {
+		b = u32((source[source_pos] << source_gap) & 0xff)
+		ctx.buffer[ctx.buffer_pos] |= byte(b) >> buffer_rem
+	} else {b = 0}
+
+	if u64(buffer_rem) + source_bits < 8 {
+		ctx.buffer_bits += int(source_bits)
+	} else {
+		ctx.buffer_pos += 1
+		ctx.buffer_bits += 8 - int(buffer_rem)
+		source_bits -= u64(8 - buffer_rem)
+
+		if ctx.buffer_bits == 512 {
+			transform(ctx)
+			ctx.buffer_bits = 0
+			ctx.buffer_pos = 0
+		}
+		ctx.buffer[ctx.buffer_pos] = byte(b << (8 - buffer_rem))
+		ctx.buffer_bits += int(source_bits)
+	}
+}
+
+final :: proc(ctx: ^Whirlpool_Context, hash: []byte) {
+	n := ctx
+	n.buffer[n.buffer_pos] |= 0x80 >> (uint(n.buffer_bits) & 7)
+	n.buffer_pos += 1
+
+	if n.buffer_pos > 64 - 32 {
+		if n.buffer_pos < 64 {
+			for i := 0; i < 64 - n.buffer_pos; i += 1 {
+				n.buffer[n.buffer_pos + i] = 0
+			}
+		}
+		transform(ctx)
+		n.buffer_pos = 0
+	}
+
+	if n.buffer_pos < 64 - 32 {
+		for i := 0; i < (64 - 32) - n.buffer_pos; i += 1 {
+			n.buffer[n.buffer_pos + i] = 0
+		}
+	}
+	n.buffer_pos = 64 - 32
+
+	for i := 0; i < 32; i += 1 {
+		n.buffer[n.buffer_pos + i] = n.bitlength[i]
+	}
+	transform(ctx)
+
+	for i := 0; i < 8; i += 1 {
+		hash[i * 8]     = byte(n.hash[i] >> 56)
+		hash[i * 8 + 1] = byte(n.hash[i] >> 48)
+		hash[i * 8 + 2] = byte(n.hash[i] >> 40)
+		hash[i * 8 + 3] = byte(n.hash[i] >> 32)
+		hash[i * 8 + 4] = byte(n.hash[i] >> 24)
+		hash[i * 8 + 5] = byte(n.hash[i] >> 16)
+		hash[i * 8 + 6] = byte(n.hash[i] >> 8)
+		hash[i * 8 + 7] = byte(n.hash[i])
+	}
+}
+
+/*
+	Whirlpool implementation
+*/
+
+ROUNDS :: 10
+
+Whirlpool_Context :: struct {
+    bitlength:   [32]byte,
+    buffer:      [64]byte,
+    buffer_bits: int,
+    buffer_pos:  int,
+    hash:        [8]u64,
+}
+
+C0 := [256]u64 {
+	0x18186018c07830d8, 0x23238c2305af4626, 0xc6c63fc67ef991b8, 0xe8e887e8136fcdfb,
+	0x878726874ca113cb, 0xb8b8dab8a9626d11, 0x0101040108050209, 0x4f4f214f426e9e0d,
+	0x3636d836adee6c9b, 0xa6a6a2a6590451ff, 0xd2d26fd2debdb90c, 0xf5f5f3f5fb06f70e,
+	0x7979f979ef80f296, 0x6f6fa16f5fcede30, 0x91917e91fcef3f6d, 0x52525552aa07a4f8,
+	0x60609d6027fdc047, 0xbcbccabc89766535, 0x9b9b569baccd2b37, 0x8e8e028e048c018a,
+	0xa3a3b6a371155bd2, 0x0c0c300c603c186c, 0x7b7bf17bff8af684, 0x3535d435b5e16a80,
+	0x1d1d741de8693af5, 0xe0e0a7e05347ddb3, 0xd7d77bd7f6acb321, 0xc2c22fc25eed999c,
+	0x2e2eb82e6d965c43, 0x4b4b314b627a9629, 0xfefedffea321e15d, 0x575741578216aed5,
+	0x15155415a8412abd, 0x7777c1779fb6eee8, 0x3737dc37a5eb6e92, 0xe5e5b3e57b56d79e,
+	0x9f9f469f8cd92313, 0xf0f0e7f0d317fd23, 0x4a4a354a6a7f9420, 0xdada4fda9e95a944,
+	0x58587d58fa25b0a2, 0xc9c903c906ca8fcf, 0x2929a429558d527c, 0x0a0a280a5022145a,
+	0xb1b1feb1e14f7f50, 0xa0a0baa0691a5dc9, 0x6b6bb16b7fdad614, 0x85852e855cab17d9,
+	0xbdbdcebd8173673c, 0x5d5d695dd234ba8f, 0x1010401080502090, 0xf4f4f7f4f303f507,
+	0xcbcb0bcb16c08bdd, 0x3e3ef83eedc67cd3, 0x0505140528110a2d, 0x676781671fe6ce78,
+	0xe4e4b7e47353d597, 0x27279c2725bb4e02, 0x4141194132588273, 0x8b8b168b2c9d0ba7,
+	0xa7a7a6a7510153f6, 0x7d7de97dcf94fab2, 0x95956e95dcfb3749, 0xd8d847d88e9fad56,
+	0xfbfbcbfb8b30eb70, 0xeeee9fee2371c1cd, 0x7c7ced7cc791f8bb, 0x6666856617e3cc71,
+	0xdddd53dda68ea77b, 0x17175c17b84b2eaf, 0x4747014702468e45, 0x9e9e429e84dc211a,
+	0xcaca0fca1ec589d4, 0x2d2db42d75995a58, 0xbfbfc6bf9179632e, 0x07071c07381b0e3f,
+	0xadad8ead012347ac, 0x5a5a755aea2fb4b0, 0x838336836cb51bef, 0x3333cc3385ff66b6,
+	0x636391633ff2c65c, 0x02020802100a0412, 0xaaaa92aa39384993, 0x7171d971afa8e2de,
+	0xc8c807c80ecf8dc6, 0x19196419c87d32d1, 0x494939497270923b, 0xd9d943d9869aaf5f,
+	0xf2f2eff2c31df931, 0xe3e3abe34b48dba8, 0x5b5b715be22ab6b9, 0x88881a8834920dbc,
+	0x9a9a529aa4c8293e, 0x262698262dbe4c0b, 0x3232c8328dfa64bf, 0xb0b0fab0e94a7d59,
+	0xe9e983e91b6acff2, 0x0f0f3c0f78331e77, 0xd5d573d5e6a6b733, 0x80803a8074ba1df4,
+	0xbebec2be997c6127, 0xcdcd13cd26de87eb, 0x3434d034bde46889, 0x48483d487a759032,
+	0xffffdbffab24e354, 0x7a7af57af78ff48d, 0x90907a90f4ea3d64, 0x5f5f615fc23ebe9d,
+	0x202080201da0403d, 0x6868bd6867d5d00f, 0x1a1a681ad07234ca, 0xaeae82ae192c41b7,
+	0xb4b4eab4c95e757d, 0x54544d549a19a8ce, 0x93937693ece53b7f, 0x222288220daa442f,
+	0x64648d6407e9c863, 0xf1f1e3f1db12ff2a, 0x7373d173bfa2e6cc, 0x12124812905a2482,
+	0x40401d403a5d807a, 0x0808200840281048, 0xc3c32bc356e89b95, 0xecec97ec337bc5df,
+	0xdbdb4bdb9690ab4d, 0xa1a1bea1611f5fc0, 0x8d8d0e8d1c830791, 0x3d3df43df5c97ac8,
+	0x97976697ccf1335b, 0x0000000000000000, 0xcfcf1bcf36d483f9, 0x2b2bac2b4587566e,
+	0x7676c57697b3ece1, 0x8282328264b019e6, 0xd6d67fd6fea9b128, 0x1b1b6c1bd87736c3,
+	0xb5b5eeb5c15b7774, 0xafaf86af112943be, 0x6a6ab56a77dfd41d, 0x50505d50ba0da0ea,
+	0x45450945124c8a57, 0xf3f3ebf3cb18fb38, 0x3030c0309df060ad, 0xefef9bef2b74c3c4,
+	0x3f3ffc3fe5c37eda, 0x55554955921caac7, 0xa2a2b2a2791059db, 0xeaea8fea0365c9e9,
+	0x656589650fecca6a, 0xbabad2bab9686903, 0x2f2fbc2f65935e4a, 0xc0c027c04ee79d8e,
+	0xdede5fdebe81a160, 0x1c1c701ce06c38fc, 0xfdfdd3fdbb2ee746, 0x4d4d294d52649a1f,
+	0x92927292e4e03976, 0x7575c9758fbceafa, 0x06061806301e0c36, 0x8a8a128a249809ae,
+	0xb2b2f2b2f940794b, 0xe6e6bfe66359d185, 0x0e0e380e70361c7e, 0x1f1f7c1ff8633ee7,
+	0x6262956237f7c455, 0xd4d477d4eea3b53a, 0xa8a89aa829324d81, 0x96966296c4f43152,
+	0xf9f9c3f99b3aef62, 0xc5c533c566f697a3, 0x2525942535b14a10, 0x59597959f220b2ab,
+	0x84842a8454ae15d0, 0x7272d572b7a7e4c5, 0x3939e439d5dd72ec, 0x4c4c2d4c5a619816,
+	0x5e5e655eca3bbc94, 0x7878fd78e785f09f, 0x3838e038ddd870e5, 0x8c8c0a8c14860598,
+	0xd1d163d1c6b2bf17, 0xa5a5aea5410b57e4, 0xe2e2afe2434dd9a1, 0x616199612ff8c24e,
+	0xb3b3f6b3f1457b42, 0x2121842115a54234, 0x9c9c4a9c94d62508, 0x1e1e781ef0663cee,
+	0x4343114322528661, 0xc7c73bc776fc93b1, 0xfcfcd7fcb32be54f, 0x0404100420140824,
+	0x51515951b208a2e3, 0x99995e99bcc72f25, 0x6d6da96d4fc4da22, 0x0d0d340d68391a65,
+	0xfafacffa8335e979, 0xdfdf5bdfb684a369, 0x7e7ee57ed79bfca9, 0x242490243db44819,
+	0x3b3bec3bc5d776fe, 0xabab96ab313d4b9a, 0xcece1fce3ed181f0, 0x1111441188552299,
+	0x8f8f068f0c890383, 0x4e4e254e4a6b9c04, 0xb7b7e6b7d1517366, 0xebeb8beb0b60cbe0,
+	0x3c3cf03cfdcc78c1, 0x81813e817cbf1ffd, 0x94946a94d4fe3540, 0xf7f7fbf7eb0cf31c,
+	0xb9b9deb9a1676f18, 0x13134c13985f268b, 0x2c2cb02c7d9c5851, 0xd3d36bd3d6b8bb05,
+	0xe7e7bbe76b5cd38c, 0x6e6ea56e57cbdc39, 0xc4c437c46ef395aa, 0x03030c03180f061b,
+	0x565645568a13acdc, 0x44440d441a49885e, 0x7f7fe17fdf9efea0, 0xa9a99ea921374f88,
+	0x2a2aa82a4d825467, 0xbbbbd6bbb16d6b0a, 0xc1c123c146e29f87, 0x53535153a202a6f1,
+	0xdcdc57dcae8ba572, 0x0b0b2c0b58271653, 0x9d9d4e9d9cd32701, 0x6c6cad6c47c1d82b,
+	0x3131c43195f562a4, 0x7474cd7487b9e8f3, 0xf6f6fff6e309f115, 0x464605460a438c4c,
+	0xacac8aac092645a5, 0x89891e893c970fb5, 0x14145014a04428b4, 0xe1e1a3e15b42dfba,
+	0x16165816b04e2ca6, 0x3a3ae83acdd274f7, 0x6969b9696fd0d206, 0x09092409482d1241,
+	0x7070dd70a7ade0d7, 0xb6b6e2b6d954716f, 0xd0d067d0ceb7bd1e, 0xeded93ed3b7ec7d6,
+	0xcccc17cc2edb85e2, 0x424215422a578468, 0x98985a98b4c22d2c, 0xa4a4aaa4490e55ed,
+	0x2828a0285d885075, 0x5c5c6d5cda31b886, 0xf8f8c7f8933fed6b, 0x8686228644a411c2,
+}
+
+C1 := [256]u64 {
+	0xd818186018c07830, 0x2623238c2305af46, 0xb8c6c63fc67ef991, 0xfbe8e887e8136fcd,
+	0xcb878726874ca113, 0x11b8b8dab8a9626d, 0x0901010401080502, 0x0d4f4f214f426e9e,
+	0x9b3636d836adee6c, 0xffa6a6a2a6590451, 0x0cd2d26fd2debdb9, 0x0ef5f5f3f5fb06f7,
+	0x967979f979ef80f2, 0x306f6fa16f5fcede, 0x6d91917e91fcef3f, 0xf852525552aa07a4,
+	0x4760609d6027fdc0, 0x35bcbccabc897665, 0x379b9b569baccd2b, 0x8a8e8e028e048c01,
+	0xd2a3a3b6a371155b, 0x6c0c0c300c603c18, 0x847b7bf17bff8af6, 0x803535d435b5e16a,
+	0xf51d1d741de8693a, 0xb3e0e0a7e05347dd, 0x21d7d77bd7f6acb3, 0x9cc2c22fc25eed99,
+	0x432e2eb82e6d965c, 0x294b4b314b627a96, 0x5dfefedffea321e1, 0xd5575741578216ae,
+	0xbd15155415a8412a, 0xe87777c1779fb6ee, 0x923737dc37a5eb6e, 0x9ee5e5b3e57b56d7,
+	0x139f9f469f8cd923, 0x23f0f0e7f0d317fd, 0x204a4a354a6a7f94, 0x44dada4fda9e95a9,
+	0xa258587d58fa25b0, 0xcfc9c903c906ca8f, 0x7c2929a429558d52, 0x5a0a0a280a502214,
+	0x50b1b1feb1e14f7f, 0xc9a0a0baa0691a5d, 0x146b6bb16b7fdad6, 0xd985852e855cab17,
+	0x3cbdbdcebd817367, 0x8f5d5d695dd234ba, 0x9010104010805020, 0x07f4f4f7f4f303f5,
+	0xddcbcb0bcb16c08b, 0xd33e3ef83eedc67c, 0x2d0505140528110a, 0x78676781671fe6ce,
+	0x97e4e4b7e47353d5, 0x0227279c2725bb4e, 0x7341411941325882, 0xa78b8b168b2c9d0b,
+	0xf6a7a7a6a7510153, 0xb27d7de97dcf94fa, 0x4995956e95dcfb37, 0x56d8d847d88e9fad,
+	0x70fbfbcbfb8b30eb, 0xcdeeee9fee2371c1, 0xbb7c7ced7cc791f8, 0x716666856617e3cc,
+	0x7bdddd53dda68ea7, 0xaf17175c17b84b2e, 0x454747014702468e, 0x1a9e9e429e84dc21,
+	0xd4caca0fca1ec589, 0x582d2db42d75995a, 0x2ebfbfc6bf917963, 0x3f07071c07381b0e,
+	0xacadad8ead012347, 0xb05a5a755aea2fb4, 0xef838336836cb51b, 0xb63333cc3385ff66,
+	0x5c636391633ff2c6, 0x1202020802100a04, 0x93aaaa92aa393849, 0xde7171d971afa8e2,
+	0xc6c8c807c80ecf8d, 0xd119196419c87d32, 0x3b49493949727092, 0x5fd9d943d9869aaf,
+	0x31f2f2eff2c31df9, 0xa8e3e3abe34b48db, 0xb95b5b715be22ab6, 0xbc88881a8834920d,
+	0x3e9a9a529aa4c829, 0x0b262698262dbe4c, 0xbf3232c8328dfa64, 0x59b0b0fab0e94a7d,
+	0xf2e9e983e91b6acf, 0x770f0f3c0f78331e, 0x33d5d573d5e6a6b7, 0xf480803a8074ba1d,
+	0x27bebec2be997c61, 0xebcdcd13cd26de87, 0x893434d034bde468, 0x3248483d487a7590,
+	0x54ffffdbffab24e3, 0x8d7a7af57af78ff4, 0x6490907a90f4ea3d, 0x9d5f5f615fc23ebe,
+	0x3d202080201da040, 0x0f6868bd6867d5d0, 0xca1a1a681ad07234, 0xb7aeae82ae192c41,
+	0x7db4b4eab4c95e75, 0xce54544d549a19a8, 0x7f93937693ece53b, 0x2f222288220daa44,
+	0x6364648d6407e9c8, 0x2af1f1e3f1db12ff, 0xcc7373d173bfa2e6, 0x8212124812905a24,
+	0x7a40401d403a5d80, 0x4808082008402810, 0x95c3c32bc356e89b, 0xdfecec97ec337bc5,
+	0x4ddbdb4bdb9690ab, 0xc0a1a1bea1611f5f, 0x918d8d0e8d1c8307, 0xc83d3df43df5c97a,
+	0x5b97976697ccf133, 0x0000000000000000, 0xf9cfcf1bcf36d483, 0x6e2b2bac2b458756,
+	0xe17676c57697b3ec, 0xe68282328264b019, 0x28d6d67fd6fea9b1, 0xc31b1b6c1bd87736,
+	0x74b5b5eeb5c15b77, 0xbeafaf86af112943, 0x1d6a6ab56a77dfd4, 0xea50505d50ba0da0,
+	0x5745450945124c8a, 0x38f3f3ebf3cb18fb, 0xad3030c0309df060, 0xc4efef9bef2b74c3,
+	0xda3f3ffc3fe5c37e, 0xc755554955921caa, 0xdba2a2b2a2791059, 0xe9eaea8fea0365c9,
+	0x6a656589650fecca, 0x03babad2bab96869, 0x4a2f2fbc2f65935e, 0x8ec0c027c04ee79d,
+	0x60dede5fdebe81a1, 0xfc1c1c701ce06c38, 0x46fdfdd3fdbb2ee7, 0x1f4d4d294d52649a,
+	0x7692927292e4e039, 0xfa7575c9758fbcea, 0x3606061806301e0c, 0xae8a8a128a249809,
+	0x4bb2b2f2b2f94079, 0x85e6e6bfe66359d1, 0x7e0e0e380e70361c, 0xe71f1f7c1ff8633e,
+	0x556262956237f7c4, 0x3ad4d477d4eea3b5, 0x81a8a89aa829324d, 0x5296966296c4f431,
+	0x62f9f9c3f99b3aef, 0xa3c5c533c566f697, 0x102525942535b14a, 0xab59597959f220b2,
+	0xd084842a8454ae15, 0xc57272d572b7a7e4, 0xec3939e439d5dd72, 0x164c4c2d4c5a6198,
+	0x945e5e655eca3bbc, 0x9f7878fd78e785f0, 0xe53838e038ddd870, 0x988c8c0a8c148605,
+	0x17d1d163d1c6b2bf, 0xe4a5a5aea5410b57, 0xa1e2e2afe2434dd9, 0x4e616199612ff8c2,
+	0x42b3b3f6b3f1457b, 0x342121842115a542, 0x089c9c4a9c94d625, 0xee1e1e781ef0663c,
+	0x6143431143225286, 0xb1c7c73bc776fc93, 0x4ffcfcd7fcb32be5, 0x2404041004201408,
+	0xe351515951b208a2, 0x2599995e99bcc72f, 0x226d6da96d4fc4da, 0x650d0d340d68391a,
+	0x79fafacffa8335e9, 0x69dfdf5bdfb684a3, 0xa97e7ee57ed79bfc, 0x19242490243db448,
+	0xfe3b3bec3bc5d776, 0x9aabab96ab313d4b, 0xf0cece1fce3ed181, 0x9911114411885522,
+	0x838f8f068f0c8903, 0x044e4e254e4a6b9c, 0x66b7b7e6b7d15173, 0xe0ebeb8beb0b60cb,
+	0xc13c3cf03cfdcc78, 0xfd81813e817cbf1f, 0x4094946a94d4fe35, 0x1cf7f7fbf7eb0cf3,
+	0x18b9b9deb9a1676f, 0x8b13134c13985f26, 0x512c2cb02c7d9c58, 0x05d3d36bd3d6b8bb,
+	0x8ce7e7bbe76b5cd3, 0x396e6ea56e57cbdc, 0xaac4c437c46ef395, 0x1b03030c03180f06,
+	0xdc565645568a13ac, 0x5e44440d441a4988, 0xa07f7fe17fdf9efe, 0x88a9a99ea921374f,
+	0x672a2aa82a4d8254, 0x0abbbbd6bbb16d6b, 0x87c1c123c146e29f, 0xf153535153a202a6,
+	0x72dcdc57dcae8ba5, 0x530b0b2c0b582716, 0x019d9d4e9d9cd327, 0x2b6c6cad6c47c1d8,
+	0xa43131c43195f562, 0xf37474cd7487b9e8, 0x15f6f6fff6e309f1, 0x4c464605460a438c,
+	0xa5acac8aac092645, 0xb589891e893c970f, 0xb414145014a04428, 0xbae1e1a3e15b42df,
+	0xa616165816b04e2c, 0xf73a3ae83acdd274, 0x066969b9696fd0d2, 0x4109092409482d12,
+	0xd77070dd70a7ade0, 0x6fb6b6e2b6d95471, 0x1ed0d067d0ceb7bd, 0xd6eded93ed3b7ec7,
+	0xe2cccc17cc2edb85, 0x68424215422a5784, 0x2c98985a98b4c22d, 0xeda4a4aaa4490e55,
+	0x752828a0285d8850, 0x865c5c6d5cda31b8, 0x6bf8f8c7f8933fed, 0xc28686228644a411,
+}
+
+C2 := [256]u64 {
+	0x30d818186018c078, 0x462623238c2305af, 0x91b8c6c63fc67ef9, 0xcdfbe8e887e8136f,
+	0x13cb878726874ca1, 0x6d11b8b8dab8a962, 0x0209010104010805, 0x9e0d4f4f214f426e,
+	0x6c9b3636d836adee, 0x51ffa6a6a2a65904, 0xb90cd2d26fd2debd, 0xf70ef5f5f3f5fb06,
+	0xf2967979f979ef80, 0xde306f6fa16f5fce, 0x3f6d91917e91fcef, 0xa4f852525552aa07,
+	0xc04760609d6027fd, 0x6535bcbccabc8976, 0x2b379b9b569baccd, 0x018a8e8e028e048c,
+	0x5bd2a3a3b6a37115, 0x186c0c0c300c603c, 0xf6847b7bf17bff8a, 0x6a803535d435b5e1,
+	0x3af51d1d741de869, 0xddb3e0e0a7e05347, 0xb321d7d77bd7f6ac, 0x999cc2c22fc25eed,
+	0x5c432e2eb82e6d96, 0x96294b4b314b627a, 0xe15dfefedffea321, 0xaed5575741578216,
+	0x2abd15155415a841, 0xeee87777c1779fb6, 0x6e923737dc37a5eb, 0xd79ee5e5b3e57b56,
+	0x23139f9f469f8cd9, 0xfd23f0f0e7f0d317, 0x94204a4a354a6a7f, 0xa944dada4fda9e95,
+	0xb0a258587d58fa25, 0x8fcfc9c903c906ca, 0x527c2929a429558d, 0x145a0a0a280a5022,
+	0x7f50b1b1feb1e14f, 0x5dc9a0a0baa0691a, 0xd6146b6bb16b7fda, 0x17d985852e855cab,
+	0x673cbdbdcebd8173, 0xba8f5d5d695dd234, 0x2090101040108050, 0xf507f4f4f7f4f303,
+	0x8bddcbcb0bcb16c0, 0x7cd33e3ef83eedc6, 0x0a2d050514052811, 0xce78676781671fe6,
+	0xd597e4e4b7e47353, 0x4e0227279c2725bb, 0x8273414119413258, 0x0ba78b8b168b2c9d,
+	0x53f6a7a7a6a75101, 0xfab27d7de97dcf94, 0x374995956e95dcfb, 0xad56d8d847d88e9f,
+	0xeb70fbfbcbfb8b30, 0xc1cdeeee9fee2371, 0xf8bb7c7ced7cc791, 0xcc716666856617e3,
+	0xa77bdddd53dda68e, 0x2eaf17175c17b84b, 0x8e45474701470246, 0x211a9e9e429e84dc,
+	0x89d4caca0fca1ec5, 0x5a582d2db42d7599, 0x632ebfbfc6bf9179, 0x0e3f07071c07381b,
+	0x47acadad8ead0123, 0xb4b05a5a755aea2f, 0x1bef838336836cb5, 0x66b63333cc3385ff,
+	0xc65c636391633ff2, 0x041202020802100a, 0x4993aaaa92aa3938, 0xe2de7171d971afa8,
+	0x8dc6c8c807c80ecf, 0x32d119196419c87d, 0x923b494939497270, 0xaf5fd9d943d9869a,
+	0xf931f2f2eff2c31d, 0xdba8e3e3abe34b48, 0xb6b95b5b715be22a, 0x0dbc88881a883492,
+	0x293e9a9a529aa4c8, 0x4c0b262698262dbe, 0x64bf3232c8328dfa, 0x7d59b0b0fab0e94a,
+	0xcff2e9e983e91b6a, 0x1e770f0f3c0f7833, 0xb733d5d573d5e6a6, 0x1df480803a8074ba,
+	0x6127bebec2be997c, 0x87ebcdcd13cd26de, 0x68893434d034bde4, 0x903248483d487a75,
+	0xe354ffffdbffab24, 0xf48d7a7af57af78f, 0x3d6490907a90f4ea, 0xbe9d5f5f615fc23e,
+	0x403d202080201da0, 0xd00f6868bd6867d5, 0x34ca1a1a681ad072, 0x41b7aeae82ae192c,
+	0x757db4b4eab4c95e, 0xa8ce54544d549a19, 0x3b7f93937693ece5, 0x442f222288220daa,
+	0xc86364648d6407e9, 0xff2af1f1e3f1db12, 0xe6cc7373d173bfa2, 0x248212124812905a,
+	0x807a40401d403a5d, 0x1048080820084028, 0x9b95c3c32bc356e8, 0xc5dfecec97ec337b,
+	0xab4ddbdb4bdb9690, 0x5fc0a1a1bea1611f, 0x07918d8d0e8d1c83, 0x7ac83d3df43df5c9,
+	0x335b97976697ccf1, 0x0000000000000000, 0x83f9cfcf1bcf36d4, 0x566e2b2bac2b4587,
+	0xece17676c57697b3, 0x19e68282328264b0, 0xb128d6d67fd6fea9, 0x36c31b1b6c1bd877,
+	0x7774b5b5eeb5c15b, 0x43beafaf86af1129, 0xd41d6a6ab56a77df, 0xa0ea50505d50ba0d,
+	0x8a5745450945124c, 0xfb38f3f3ebf3cb18, 0x60ad3030c0309df0, 0xc3c4efef9bef2b74,
+	0x7eda3f3ffc3fe5c3, 0xaac755554955921c, 0x59dba2a2b2a27910, 0xc9e9eaea8fea0365,
+	0xca6a656589650fec, 0x6903babad2bab968, 0x5e4a2f2fbc2f6593, 0x9d8ec0c027c04ee7,
+	0xa160dede5fdebe81, 0x38fc1c1c701ce06c, 0xe746fdfdd3fdbb2e, 0x9a1f4d4d294d5264,
+	0x397692927292e4e0, 0xeafa7575c9758fbc, 0x0c3606061806301e, 0x09ae8a8a128a2498,
+	0x794bb2b2f2b2f940, 0xd185e6e6bfe66359, 0x1c7e0e0e380e7036, 0x3ee71f1f7c1ff863,
+	0xc4556262956237f7, 0xb53ad4d477d4eea3, 0x4d81a8a89aa82932, 0x315296966296c4f4,
+	0xef62f9f9c3f99b3a, 0x97a3c5c533c566f6, 0x4a102525942535b1, 0xb2ab59597959f220,
+	0x15d084842a8454ae, 0xe4c57272d572b7a7, 0x72ec3939e439d5dd, 0x98164c4c2d4c5a61,
+	0xbc945e5e655eca3b, 0xf09f7878fd78e785, 0x70e53838e038ddd8, 0x05988c8c0a8c1486,
+	0xbf17d1d163d1c6b2, 0x57e4a5a5aea5410b, 0xd9a1e2e2afe2434d, 0xc24e616199612ff8,
+	0x7b42b3b3f6b3f145, 0x42342121842115a5, 0x25089c9c4a9c94d6, 0x3cee1e1e781ef066,
+	0x8661434311432252, 0x93b1c7c73bc776fc, 0xe54ffcfcd7fcb32b, 0x0824040410042014,
+	0xa2e351515951b208, 0x2f2599995e99bcc7, 0xda226d6da96d4fc4, 0x1a650d0d340d6839,
+	0xe979fafacffa8335, 0xa369dfdf5bdfb684, 0xfca97e7ee57ed79b, 0x4819242490243db4,
+	0x76fe3b3bec3bc5d7, 0x4b9aabab96ab313d, 0x81f0cece1fce3ed1, 0x2299111144118855,
+	0x03838f8f068f0c89, 0x9c044e4e254e4a6b, 0x7366b7b7e6b7d151, 0xcbe0ebeb8beb0b60,
+	0x78c13c3cf03cfdcc, 0x1ffd81813e817cbf, 0x354094946a94d4fe, 0xf31cf7f7fbf7eb0c,
+	0x6f18b9b9deb9a167, 0x268b13134c13985f, 0x58512c2cb02c7d9c, 0xbb05d3d36bd3d6b8,
+	0xd38ce7e7bbe76b5c, 0xdc396e6ea56e57cb, 0x95aac4c437c46ef3, 0x061b03030c03180f,
+	0xacdc565645568a13, 0x885e44440d441a49, 0xfea07f7fe17fdf9e, 0x4f88a9a99ea92137,
+	0x54672a2aa82a4d82, 0x6b0abbbbd6bbb16d, 0x9f87c1c123c146e2, 0xa6f153535153a202,
+	0xa572dcdc57dcae8b, 0x16530b0b2c0b5827, 0x27019d9d4e9d9cd3, 0xd82b6c6cad6c47c1,
+	0x62a43131c43195f5, 0xe8f37474cd7487b9, 0xf115f6f6fff6e309, 0x8c4c464605460a43,
+	0x45a5acac8aac0926, 0x0fb589891e893c97, 0x28b414145014a044, 0xdfbae1e1a3e15b42,
+	0x2ca616165816b04e, 0x74f73a3ae83acdd2, 0xd2066969b9696fd0, 0x124109092409482d,
+	0xe0d77070dd70a7ad, 0x716fb6b6e2b6d954, 0xbd1ed0d067d0ceb7, 0xc7d6eded93ed3b7e,
+	0x85e2cccc17cc2edb, 0x8468424215422a57, 0x2d2c98985a98b4c2, 0x55eda4a4aaa4490e,
+	0x50752828a0285d88, 0xb8865c5c6d5cda31, 0xed6bf8f8c7f8933f, 0x11c28686228644a4,
+}
+
+C3 := [256]u64 {
+	0x7830d818186018c0, 0xaf462623238c2305, 0xf991b8c6c63fc67e, 0x6fcdfbe8e887e813,
+	0xa113cb878726874c, 0x626d11b8b8dab8a9, 0x0502090101040108, 0x6e9e0d4f4f214f42,
+	0xee6c9b3636d836ad, 0x0451ffa6a6a2a659, 0xbdb90cd2d26fd2de, 0x06f70ef5f5f3f5fb,
+	0x80f2967979f979ef, 0xcede306f6fa16f5f, 0xef3f6d91917e91fc, 0x07a4f852525552aa,
+	0xfdc04760609d6027, 0x766535bcbccabc89, 0xcd2b379b9b569bac, 0x8c018a8e8e028e04,
+	0x155bd2a3a3b6a371, 0x3c186c0c0c300c60, 0x8af6847b7bf17bff, 0xe16a803535d435b5,
+	0x693af51d1d741de8, 0x47ddb3e0e0a7e053, 0xacb321d7d77bd7f6, 0xed999cc2c22fc25e,
+	0x965c432e2eb82e6d, 0x7a96294b4b314b62, 0x21e15dfefedffea3, 0x16aed55757415782,
+	0x412abd15155415a8, 0xb6eee87777c1779f, 0xeb6e923737dc37a5, 0x56d79ee5e5b3e57b,
+	0xd923139f9f469f8c, 0x17fd23f0f0e7f0d3, 0x7f94204a4a354a6a, 0x95a944dada4fda9e,
+	0x25b0a258587d58fa, 0xca8fcfc9c903c906, 0x8d527c2929a42955, 0x22145a0a0a280a50,
+	0x4f7f50b1b1feb1e1, 0x1a5dc9a0a0baa069, 0xdad6146b6bb16b7f, 0xab17d985852e855c,
+	0x73673cbdbdcebd81, 0x34ba8f5d5d695dd2, 0x5020901010401080, 0x03f507f4f4f7f4f3,
+	0xc08bddcbcb0bcb16, 0xc67cd33e3ef83eed, 0x110a2d0505140528, 0xe6ce78676781671f,
+	0x53d597e4e4b7e473, 0xbb4e0227279c2725, 0x5882734141194132, 0x9d0ba78b8b168b2c,
+	0x0153f6a7a7a6a751, 0x94fab27d7de97dcf, 0xfb374995956e95dc, 0x9fad56d8d847d88e,
+	0x30eb70fbfbcbfb8b, 0x71c1cdeeee9fee23, 0x91f8bb7c7ced7cc7, 0xe3cc716666856617,
+	0x8ea77bdddd53dda6, 0x4b2eaf17175c17b8, 0x468e454747014702, 0xdc211a9e9e429e84,
+	0xc589d4caca0fca1e, 0x995a582d2db42d75, 0x79632ebfbfc6bf91, 0x1b0e3f07071c0738,
+	0x2347acadad8ead01, 0x2fb4b05a5a755aea, 0xb51bef838336836c, 0xff66b63333cc3385,
+	0xf2c65c636391633f, 0x0a04120202080210, 0x384993aaaa92aa39, 0xa8e2de7171d971af,
+	0xcf8dc6c8c807c80e, 0x7d32d119196419c8, 0x70923b4949394972, 0x9aaf5fd9d943d986,
+	0x1df931f2f2eff2c3, 0x48dba8e3e3abe34b, 0x2ab6b95b5b715be2, 0x920dbc88881a8834,
+	0xc8293e9a9a529aa4, 0xbe4c0b262698262d, 0xfa64bf3232c8328d, 0x4a7d59b0b0fab0e9,
+	0x6acff2e9e983e91b, 0x331e770f0f3c0f78, 0xa6b733d5d573d5e6, 0xba1df480803a8074,
+	0x7c6127bebec2be99, 0xde87ebcdcd13cd26, 0xe468893434d034bd, 0x75903248483d487a,
+	0x24e354ffffdbffab, 0x8ff48d7a7af57af7, 0xea3d6490907a90f4, 0x3ebe9d5f5f615fc2,
+	0xa0403d202080201d, 0xd5d00f6868bd6867, 0x7234ca1a1a681ad0, 0x2c41b7aeae82ae19,
+	0x5e757db4b4eab4c9, 0x19a8ce54544d549a, 0xe53b7f93937693ec, 0xaa442f222288220d,
+	0xe9c86364648d6407, 0x12ff2af1f1e3f1db, 0xa2e6cc7373d173bf, 0x5a24821212481290,
+	0x5d807a40401d403a, 0x2810480808200840, 0xe89b95c3c32bc356, 0x7bc5dfecec97ec33,
+	0x90ab4ddbdb4bdb96, 0x1f5fc0a1a1bea161, 0x8307918d8d0e8d1c, 0xc97ac83d3df43df5,
+	0xf1335b97976697cc, 0x0000000000000000, 0xd483f9cfcf1bcf36, 0x87566e2b2bac2b45,
+	0xb3ece17676c57697, 0xb019e68282328264, 0xa9b128d6d67fd6fe, 0x7736c31b1b6c1bd8,
+	0x5b7774b5b5eeb5c1, 0x2943beafaf86af11, 0xdfd41d6a6ab56a77, 0x0da0ea50505d50ba,
+	0x4c8a574545094512, 0x18fb38f3f3ebf3cb, 0xf060ad3030c0309d, 0x74c3c4efef9bef2b,
+	0xc37eda3f3ffc3fe5, 0x1caac75555495592, 0x1059dba2a2b2a279, 0x65c9e9eaea8fea03,
+	0xecca6a656589650f, 0x686903babad2bab9, 0x935e4a2f2fbc2f65, 0xe79d8ec0c027c04e,
+	0x81a160dede5fdebe, 0x6c38fc1c1c701ce0, 0x2ee746fdfdd3fdbb, 0x649a1f4d4d294d52,
+	0xe0397692927292e4, 0xbceafa7575c9758f, 0x1e0c360606180630, 0x9809ae8a8a128a24,
+	0x40794bb2b2f2b2f9, 0x59d185e6e6bfe663, 0x361c7e0e0e380e70, 0x633ee71f1f7c1ff8,
+	0xf7c4556262956237, 0xa3b53ad4d477d4ee, 0x324d81a8a89aa829, 0xf4315296966296c4,
+	0x3aef62f9f9c3f99b, 0xf697a3c5c533c566, 0xb14a102525942535, 0x20b2ab59597959f2,
+	0xae15d084842a8454, 0xa7e4c57272d572b7, 0xdd72ec3939e439d5, 0x6198164c4c2d4c5a,
+	0x3bbc945e5e655eca, 0x85f09f7878fd78e7, 0xd870e53838e038dd, 0x8605988c8c0a8c14,
+	0xb2bf17d1d163d1c6, 0x0b57e4a5a5aea541, 0x4dd9a1e2e2afe243, 0xf8c24e616199612f,
+	0x457b42b3b3f6b3f1, 0xa542342121842115, 0xd625089c9c4a9c94, 0x663cee1e1e781ef0,
+	0x5286614343114322, 0xfc93b1c7c73bc776, 0x2be54ffcfcd7fcb3, 0x1408240404100420,
+	0x08a2e351515951b2, 0xc72f2599995e99bc, 0xc4da226d6da96d4f, 0x391a650d0d340d68,
+	0x35e979fafacffa83, 0x84a369dfdf5bdfb6, 0x9bfca97e7ee57ed7, 0xb44819242490243d,
+	0xd776fe3b3bec3bc5, 0x3d4b9aabab96ab31, 0xd181f0cece1fce3e, 0x5522991111441188,
+	0x8903838f8f068f0c, 0x6b9c044e4e254e4a, 0x517366b7b7e6b7d1, 0x60cbe0ebeb8beb0b,
+	0xcc78c13c3cf03cfd, 0xbf1ffd81813e817c, 0xfe354094946a94d4, 0x0cf31cf7f7fbf7eb,
+	0x676f18b9b9deb9a1, 0x5f268b13134c1398, 0x9c58512c2cb02c7d, 0xb8bb05d3d36bd3d6,
+	0x5cd38ce7e7bbe76b, 0xcbdc396e6ea56e57, 0xf395aac4c437c46e, 0x0f061b03030c0318,
+	0x13acdc565645568a, 0x49885e44440d441a, 0x9efea07f7fe17fdf, 0x374f88a9a99ea921,
+	0x8254672a2aa82a4d, 0x6d6b0abbbbd6bbb1, 0xe29f87c1c123c146, 0x02a6f153535153a2,
+	0x8ba572dcdc57dcae, 0x2716530b0b2c0b58, 0xd327019d9d4e9d9c, 0xc1d82b6c6cad6c47,
+	0xf562a43131c43195, 0xb9e8f37474cd7487, 0x09f115f6f6fff6e3, 0x438c4c464605460a,
+	0x2645a5acac8aac09, 0x970fb589891e893c, 0x4428b414145014a0, 0x42dfbae1e1a3e15b,
+	0x4e2ca616165816b0, 0xd274f73a3ae83acd, 0xd0d2066969b9696f, 0x2d12410909240948,
+	0xade0d77070dd70a7, 0x54716fb6b6e2b6d9, 0xb7bd1ed0d067d0ce, 0x7ec7d6eded93ed3b,
+	0xdb85e2cccc17cc2e, 0x578468424215422a, 0xc22d2c98985a98b4, 0x0e55eda4a4aaa449,
+	0x8850752828a0285d, 0x31b8865c5c6d5cda, 0x3fed6bf8f8c7f893, 0xa411c28686228644,
+}
+
+C4 := [256]u64 {
+	0xc07830d818186018, 0x05af462623238c23, 0x7ef991b8c6c63fc6, 0x136fcdfbe8e887e8,
+	0x4ca113cb87872687, 0xa9626d11b8b8dab8, 0x0805020901010401, 0x426e9e0d4f4f214f,
+	0xadee6c9b3636d836, 0x590451ffa6a6a2a6, 0xdebdb90cd2d26fd2, 0xfb06f70ef5f5f3f5,
+	0xef80f2967979f979, 0x5fcede306f6fa16f, 0xfcef3f6d91917e91, 0xaa07a4f852525552,
+	0x27fdc04760609d60, 0x89766535bcbccabc, 0xaccd2b379b9b569b, 0x048c018a8e8e028e,
+	0x71155bd2a3a3b6a3, 0x603c186c0c0c300c, 0xff8af6847b7bf17b, 0xb5e16a803535d435,
+	0xe8693af51d1d741d, 0x5347ddb3e0e0a7e0, 0xf6acb321d7d77bd7, 0x5eed999cc2c22fc2,
+	0x6d965c432e2eb82e, 0x627a96294b4b314b, 0xa321e15dfefedffe, 0x8216aed557574157,
+	0xa8412abd15155415, 0x9fb6eee87777c177, 0xa5eb6e923737dc37, 0x7b56d79ee5e5b3e5,
+	0x8cd923139f9f469f, 0xd317fd23f0f0e7f0, 0x6a7f94204a4a354a, 0x9e95a944dada4fda,
+	0xfa25b0a258587d58, 0x06ca8fcfc9c903c9, 0x558d527c2929a429, 0x5022145a0a0a280a,
+	0xe14f7f50b1b1feb1, 0x691a5dc9a0a0baa0, 0x7fdad6146b6bb16b, 0x5cab17d985852e85,
+	0x8173673cbdbdcebd, 0xd234ba8f5d5d695d, 0x8050209010104010, 0xf303f507f4f4f7f4,
+	0x16c08bddcbcb0bcb, 0xedc67cd33e3ef83e, 0x28110a2d05051405, 0x1fe6ce7867678167,
+	0x7353d597e4e4b7e4, 0x25bb4e0227279c27, 0x3258827341411941, 0x2c9d0ba78b8b168b,
+	0x510153f6a7a7a6a7, 0xcf94fab27d7de97d, 0xdcfb374995956e95, 0x8e9fad56d8d847d8,
+	0x8b30eb70fbfbcbfb, 0x2371c1cdeeee9fee, 0xc791f8bb7c7ced7c, 0x17e3cc7166668566,
+	0xa68ea77bdddd53dd, 0xb84b2eaf17175c17, 0x02468e4547470147, 0x84dc211a9e9e429e,
+	0x1ec589d4caca0fca, 0x75995a582d2db42d, 0x9179632ebfbfc6bf, 0x381b0e3f07071c07,
+	0x012347acadad8ead, 0xea2fb4b05a5a755a, 0x6cb51bef83833683, 0x85ff66b63333cc33,
+	0x3ff2c65c63639163, 0x100a041202020802, 0x39384993aaaa92aa, 0xafa8e2de7171d971,
+	0x0ecf8dc6c8c807c8, 0xc87d32d119196419, 0x7270923b49493949, 0x869aaf5fd9d943d9,
+	0xc31df931f2f2eff2, 0x4b48dba8e3e3abe3, 0xe22ab6b95b5b715b, 0x34920dbc88881a88,
+	0xa4c8293e9a9a529a, 0x2dbe4c0b26269826, 0x8dfa64bf3232c832, 0xe94a7d59b0b0fab0,
+	0x1b6acff2e9e983e9, 0x78331e770f0f3c0f, 0xe6a6b733d5d573d5, 0x74ba1df480803a80,
+	0x997c6127bebec2be, 0x26de87ebcdcd13cd, 0xbde468893434d034, 0x7a75903248483d48,
+	0xab24e354ffffdbff, 0xf78ff48d7a7af57a, 0xf4ea3d6490907a90, 0xc23ebe9d5f5f615f,
+	0x1da0403d20208020, 0x67d5d00f6868bd68, 0xd07234ca1a1a681a, 0x192c41b7aeae82ae,
+	0xc95e757db4b4eab4, 0x9a19a8ce54544d54, 0xece53b7f93937693, 0x0daa442f22228822,
+	0x07e9c86364648d64, 0xdb12ff2af1f1e3f1, 0xbfa2e6cc7373d173, 0x905a248212124812,
+	0x3a5d807a40401d40, 0x4028104808082008, 0x56e89b95c3c32bc3, 0x337bc5dfecec97ec,
+	0x9690ab4ddbdb4bdb, 0x611f5fc0a1a1bea1, 0x1c8307918d8d0e8d, 0xf5c97ac83d3df43d,
+	0xccf1335b97976697, 0x0000000000000000, 0x36d483f9cfcf1bcf, 0x4587566e2b2bac2b,
+	0x97b3ece17676c576, 0x64b019e682823282, 0xfea9b128d6d67fd6, 0xd87736c31b1b6c1b,
+	0xc15b7774b5b5eeb5, 0x112943beafaf86af, 0x77dfd41d6a6ab56a, 0xba0da0ea50505d50,
+	0x124c8a5745450945, 0xcb18fb38f3f3ebf3, 0x9df060ad3030c030, 0x2b74c3c4efef9bef,
+	0xe5c37eda3f3ffc3f, 0x921caac755554955, 0x791059dba2a2b2a2, 0x0365c9e9eaea8fea,
+	0x0fecca6a65658965, 0xb9686903babad2ba, 0x65935e4a2f2fbc2f, 0x4ee79d8ec0c027c0,
+	0xbe81a160dede5fde, 0xe06c38fc1c1c701c, 0xbb2ee746fdfdd3fd, 0x52649a1f4d4d294d,
+	0xe4e0397692927292, 0x8fbceafa7575c975, 0x301e0c3606061806, 0x249809ae8a8a128a,
+	0xf940794bb2b2f2b2, 0x6359d185e6e6bfe6, 0x70361c7e0e0e380e, 0xf8633ee71f1f7c1f,
+	0x37f7c45562629562, 0xeea3b53ad4d477d4, 0x29324d81a8a89aa8, 0xc4f4315296966296,
+	0x9b3aef62f9f9c3f9, 0x66f697a3c5c533c5, 0x35b14a1025259425, 0xf220b2ab59597959,
+	0x54ae15d084842a84, 0xb7a7e4c57272d572, 0xd5dd72ec3939e439, 0x5a6198164c4c2d4c,
+	0xca3bbc945e5e655e, 0xe785f09f7878fd78, 0xddd870e53838e038, 0x148605988c8c0a8c,
+	0xc6b2bf17d1d163d1, 0x410b57e4a5a5aea5, 0x434dd9a1e2e2afe2, 0x2ff8c24e61619961,
+	0xf1457b42b3b3f6b3, 0x15a5423421218421, 0x94d625089c9c4a9c, 0xf0663cee1e1e781e,
+	0x2252866143431143, 0x76fc93b1c7c73bc7, 0xb32be54ffcfcd7fc, 0x2014082404041004,
+	0xb208a2e351515951, 0xbcc72f2599995e99, 0x4fc4da226d6da96d, 0x68391a650d0d340d,
+	0x8335e979fafacffa, 0xb684a369dfdf5bdf, 0xd79bfca97e7ee57e, 0x3db4481924249024,
+	0xc5d776fe3b3bec3b, 0x313d4b9aabab96ab, 0x3ed181f0cece1fce, 0x8855229911114411,
+	0x0c8903838f8f068f, 0x4a6b9c044e4e254e, 0xd1517366b7b7e6b7, 0x0b60cbe0ebeb8beb,
+	0xfdcc78c13c3cf03c, 0x7cbf1ffd81813e81, 0xd4fe354094946a94, 0xeb0cf31cf7f7fbf7,
+	0xa1676f18b9b9deb9, 0x985f268b13134c13, 0x7d9c58512c2cb02c, 0xd6b8bb05d3d36bd3,
+	0x6b5cd38ce7e7bbe7, 0x57cbdc396e6ea56e, 0x6ef395aac4c437c4, 0x180f061b03030c03,
+	0x8a13acdc56564556, 0x1a49885e44440d44, 0xdf9efea07f7fe17f, 0x21374f88a9a99ea9,
+	0x4d8254672a2aa82a, 0xb16d6b0abbbbd6bb, 0x46e29f87c1c123c1, 0xa202a6f153535153,
+	0xae8ba572dcdc57dc, 0x582716530b0b2c0b, 0x9cd327019d9d4e9d, 0x47c1d82b6c6cad6c,
+	0x95f562a43131c431, 0x87b9e8f37474cd74, 0xe309f115f6f6fff6, 0x0a438c4c46460546,
+	0x092645a5acac8aac, 0x3c970fb589891e89, 0xa04428b414145014, 0x5b42dfbae1e1a3e1,
+	0xb04e2ca616165816, 0xcdd274f73a3ae83a, 0x6fd0d2066969b969, 0x482d124109092409,
+	0xa7ade0d77070dd70, 0xd954716fb6b6e2b6, 0xceb7bd1ed0d067d0, 0x3b7ec7d6eded93ed,
+	0x2edb85e2cccc17cc, 0x2a57846842421542, 0xb4c22d2c98985a98, 0x490e55eda4a4aaa4,
+	0x5d8850752828a028, 0xda31b8865c5c6d5c, 0x933fed6bf8f8c7f8, 0x44a411c286862286,
+}
+
+C5 := [256]u64 {
+	0x18c07830d8181860, 0x2305af462623238c, 0xc67ef991b8c6c63f, 0xe8136fcdfbe8e887,
+	0x874ca113cb878726, 0xb8a9626d11b8b8da, 0x0108050209010104, 0x4f426e9e0d4f4f21,
+	0x36adee6c9b3636d8, 0xa6590451ffa6a6a2, 0xd2debdb90cd2d26f, 0xf5fb06f70ef5f5f3,
+	0x79ef80f2967979f9, 0x6f5fcede306f6fa1, 0x91fcef3f6d91917e, 0x52aa07a4f8525255,
+	0x6027fdc04760609d, 0xbc89766535bcbcca, 0x9baccd2b379b9b56, 0x8e048c018a8e8e02,
+	0xa371155bd2a3a3b6, 0x0c603c186c0c0c30, 0x7bff8af6847b7bf1, 0x35b5e16a803535d4,
+	0x1de8693af51d1d74, 0xe05347ddb3e0e0a7, 0xd7f6acb321d7d77b, 0xc25eed999cc2c22f,
+	0x2e6d965c432e2eb8, 0x4b627a96294b4b31, 0xfea321e15dfefedf, 0x578216aed5575741,
+	0x15a8412abd151554, 0x779fb6eee87777c1, 0x37a5eb6e923737dc, 0xe57b56d79ee5e5b3,
+	0x9f8cd923139f9f46, 0xf0d317fd23f0f0e7, 0x4a6a7f94204a4a35, 0xda9e95a944dada4f,
+	0x58fa25b0a258587d, 0xc906ca8fcfc9c903, 0x29558d527c2929a4, 0x0a5022145a0a0a28,
+	0xb1e14f7f50b1b1fe, 0xa0691a5dc9a0a0ba, 0x6b7fdad6146b6bb1, 0x855cab17d985852e,
+	0xbd8173673cbdbdce, 0x5dd234ba8f5d5d69, 0x1080502090101040, 0xf4f303f507f4f4f7,
+	0xcb16c08bddcbcb0b, 0x3eedc67cd33e3ef8, 0x0528110a2d050514, 0x671fe6ce78676781,
+	0xe47353d597e4e4b7, 0x2725bb4e0227279c, 0x4132588273414119, 0x8b2c9d0ba78b8b16,
+	0xa7510153f6a7a7a6, 0x7dcf94fab27d7de9, 0x95dcfb374995956e, 0xd88e9fad56d8d847,
+	0xfb8b30eb70fbfbcb, 0xee2371c1cdeeee9f, 0x7cc791f8bb7c7ced, 0x6617e3cc71666685,
+	0xdda68ea77bdddd53, 0x17b84b2eaf17175c, 0x4702468e45474701, 0x9e84dc211a9e9e42,
+	0xca1ec589d4caca0f, 0x2d75995a582d2db4, 0xbf9179632ebfbfc6, 0x07381b0e3f07071c,
+	0xad012347acadad8e, 0x5aea2fb4b05a5a75, 0x836cb51bef838336, 0x3385ff66b63333cc,
+	0x633ff2c65c636391, 0x02100a0412020208, 0xaa39384993aaaa92, 0x71afa8e2de7171d9,
+	0xc80ecf8dc6c8c807, 0x19c87d32d1191964, 0x497270923b494939, 0xd9869aaf5fd9d943,
+	0xf2c31df931f2f2ef, 0xe34b48dba8e3e3ab, 0x5be22ab6b95b5b71, 0x8834920dbc88881a,
+	0x9aa4c8293e9a9a52, 0x262dbe4c0b262698, 0x328dfa64bf3232c8, 0xb0e94a7d59b0b0fa,
+	0xe91b6acff2e9e983, 0x0f78331e770f0f3c, 0xd5e6a6b733d5d573, 0x8074ba1df480803a,
+	0xbe997c6127bebec2, 0xcd26de87ebcdcd13, 0x34bde468893434d0, 0x487a75903248483d,
+	0xffab24e354ffffdb, 0x7af78ff48d7a7af5, 0x90f4ea3d6490907a, 0x5fc23ebe9d5f5f61,
+	0x201da0403d202080, 0x6867d5d00f6868bd, 0x1ad07234ca1a1a68, 0xae192c41b7aeae82,
+	0xb4c95e757db4b4ea, 0x549a19a8ce54544d, 0x93ece53b7f939376, 0x220daa442f222288,
+	0x6407e9c86364648d, 0xf1db12ff2af1f1e3, 0x73bfa2e6cc7373d1, 0x12905a2482121248,
+	0x403a5d807a40401d, 0x0840281048080820, 0xc356e89b95c3c32b, 0xec337bc5dfecec97,
+	0xdb9690ab4ddbdb4b, 0xa1611f5fc0a1a1be, 0x8d1c8307918d8d0e, 0x3df5c97ac83d3df4,
+	0x97ccf1335b979766, 0x0000000000000000, 0xcf36d483f9cfcf1b, 0x2b4587566e2b2bac,
+	0x7697b3ece17676c5, 0x8264b019e6828232, 0xd6fea9b128d6d67f, 0x1bd87736c31b1b6c,
+	0xb5c15b7774b5b5ee, 0xaf112943beafaf86, 0x6a77dfd41d6a6ab5, 0x50ba0da0ea50505d,
+	0x45124c8a57454509, 0xf3cb18fb38f3f3eb, 0x309df060ad3030c0, 0xef2b74c3c4efef9b,
+	0x3fe5c37eda3f3ffc, 0x55921caac7555549, 0xa2791059dba2a2b2, 0xea0365c9e9eaea8f,
+	0x650fecca6a656589, 0xbab9686903babad2, 0x2f65935e4a2f2fbc, 0xc04ee79d8ec0c027,
+	0xdebe81a160dede5f, 0x1ce06c38fc1c1c70, 0xfdbb2ee746fdfdd3, 0x4d52649a1f4d4d29,
+	0x92e4e03976929272, 0x758fbceafa7575c9, 0x06301e0c36060618, 0x8a249809ae8a8a12,
+	0xb2f940794bb2b2f2, 0xe66359d185e6e6bf, 0x0e70361c7e0e0e38, 0x1ff8633ee71f1f7c,
+	0x6237f7c455626295, 0xd4eea3b53ad4d477, 0xa829324d81a8a89a, 0x96c4f43152969662,
+	0xf99b3aef62f9f9c3, 0xc566f697a3c5c533, 0x2535b14a10252594, 0x59f220b2ab595979,
+	0x8454ae15d084842a, 0x72b7a7e4c57272d5, 0x39d5dd72ec3939e4, 0x4c5a6198164c4c2d,
+	0x5eca3bbc945e5e65, 0x78e785f09f7878fd, 0x38ddd870e53838e0, 0x8c148605988c8c0a,
+	0xd1c6b2bf17d1d163, 0xa5410b57e4a5a5ae, 0xe2434dd9a1e2e2af, 0x612ff8c24e616199,
+	0xb3f1457b42b3b3f6, 0x2115a54234212184, 0x9c94d625089c9c4a, 0x1ef0663cee1e1e78,
+	0x4322528661434311, 0xc776fc93b1c7c73b, 0xfcb32be54ffcfcd7, 0x0420140824040410,
+	0x51b208a2e3515159, 0x99bcc72f2599995e, 0x6d4fc4da226d6da9, 0x0d68391a650d0d34,
+	0xfa8335e979fafacf, 0xdfb684a369dfdf5b, 0x7ed79bfca97e7ee5, 0x243db44819242490,
+	0x3bc5d776fe3b3bec, 0xab313d4b9aabab96, 0xce3ed181f0cece1f, 0x1188552299111144,
+	0x8f0c8903838f8f06, 0x4e4a6b9c044e4e25, 0xb7d1517366b7b7e6, 0xeb0b60cbe0ebeb8b,
+	0x3cfdcc78c13c3cf0, 0x817cbf1ffd81813e, 0x94d4fe354094946a, 0xf7eb0cf31cf7f7fb,
+	0xb9a1676f18b9b9de, 0x13985f268b13134c, 0x2c7d9c58512c2cb0, 0xd3d6b8bb05d3d36b,
+	0xe76b5cd38ce7e7bb, 0x6e57cbdc396e6ea5, 0xc46ef395aac4c437, 0x03180f061b03030c,
+	0x568a13acdc565645, 0x441a49885e44440d, 0x7fdf9efea07f7fe1, 0xa921374f88a9a99e,
+	0x2a4d8254672a2aa8, 0xbbb16d6b0abbbbd6, 0xc146e29f87c1c123, 0x53a202a6f1535351,
+	0xdcae8ba572dcdc57, 0x0b582716530b0b2c, 0x9d9cd327019d9d4e, 0x6c47c1d82b6c6cad,
+	0x3195f562a43131c4, 0x7487b9e8f37474cd, 0xf6e309f115f6f6ff, 0x460a438c4c464605,
+	0xac092645a5acac8a, 0x893c970fb589891e, 0x14a04428b4141450, 0xe15b42dfbae1e1a3,
+	0x16b04e2ca6161658, 0x3acdd274f73a3ae8, 0x696fd0d2066969b9, 0x09482d1241090924,
+	0x70a7ade0d77070dd, 0xb6d954716fb6b6e2, 0xd0ceb7bd1ed0d067, 0xed3b7ec7d6eded93,
+	0xcc2edb85e2cccc17, 0x422a578468424215, 0x98b4c22d2c98985a, 0xa4490e55eda4a4aa,
+	0x285d8850752828a0, 0x5cda31b8865c5c6d, 0xf8933fed6bf8f8c7, 0x8644a411c2868622,
+}
+
+C6 := [256]u64 {
+	0x6018c07830d81818, 0x8c2305af46262323, 0x3fc67ef991b8c6c6, 0x87e8136fcdfbe8e8,
+	0x26874ca113cb8787, 0xdab8a9626d11b8b8, 0x0401080502090101, 0x214f426e9e0d4f4f,
+	0xd836adee6c9b3636, 0xa2a6590451ffa6a6, 0x6fd2debdb90cd2d2, 0xf3f5fb06f70ef5f5,
+	0xf979ef80f2967979, 0xa16f5fcede306f6f, 0x7e91fcef3f6d9191, 0x5552aa07a4f85252,
+	0x9d6027fdc0476060, 0xcabc89766535bcbc, 0x569baccd2b379b9b, 0x028e048c018a8e8e,
+	0xb6a371155bd2a3a3, 0x300c603c186c0c0c, 0xf17bff8af6847b7b, 0xd435b5e16a803535,
+	0x741de8693af51d1d, 0xa7e05347ddb3e0e0, 0x7bd7f6acb321d7d7, 0x2fc25eed999cc2c2,
+	0xb82e6d965c432e2e, 0x314b627a96294b4b, 0xdffea321e15dfefe, 0x41578216aed55757,
+	0x5415a8412abd1515, 0xc1779fb6eee87777, 0xdc37a5eb6e923737, 0xb3e57b56d79ee5e5,
+	0x469f8cd923139f9f, 0xe7f0d317fd23f0f0, 0x354a6a7f94204a4a, 0x4fda9e95a944dada,
+	0x7d58fa25b0a25858, 0x03c906ca8fcfc9c9, 0xa429558d527c2929, 0x280a5022145a0a0a,
+	0xfeb1e14f7f50b1b1, 0xbaa0691a5dc9a0a0, 0xb16b7fdad6146b6b, 0x2e855cab17d98585,
+	0xcebd8173673cbdbd, 0x695dd234ba8f5d5d, 0x4010805020901010, 0xf7f4f303f507f4f4,
+	0x0bcb16c08bddcbcb, 0xf83eedc67cd33e3e, 0x140528110a2d0505, 0x81671fe6ce786767,
+	0xb7e47353d597e4e4, 0x9c2725bb4e022727, 0x1941325882734141, 0x168b2c9d0ba78b8b,
+	0xa6a7510153f6a7a7, 0xe97dcf94fab27d7d, 0x6e95dcfb37499595, 0x47d88e9fad56d8d8,
+	0xcbfb8b30eb70fbfb, 0x9fee2371c1cdeeee, 0xed7cc791f8bb7c7c, 0x856617e3cc716666,
+	0x53dda68ea77bdddd, 0x5c17b84b2eaf1717, 0x014702468e454747, 0x429e84dc211a9e9e,
+	0x0fca1ec589d4caca, 0xb42d75995a582d2d, 0xc6bf9179632ebfbf, 0x1c07381b0e3f0707,
+	0x8ead012347acadad, 0x755aea2fb4b05a5a, 0x36836cb51bef8383, 0xcc3385ff66b63333,
+	0x91633ff2c65c6363, 0x0802100a04120202, 0x92aa39384993aaaa, 0xd971afa8e2de7171,
+	0x07c80ecf8dc6c8c8, 0x6419c87d32d11919, 0x39497270923b4949, 0x43d9869aaf5fd9d9,
+	0xeff2c31df931f2f2, 0xabe34b48dba8e3e3, 0x715be22ab6b95b5b, 0x1a8834920dbc8888,
+	0x529aa4c8293e9a9a, 0x98262dbe4c0b2626, 0xc8328dfa64bf3232, 0xfab0e94a7d59b0b0,
+	0x83e91b6acff2e9e9, 0x3c0f78331e770f0f, 0x73d5e6a6b733d5d5, 0x3a8074ba1df48080,
+	0xc2be997c6127bebe, 0x13cd26de87ebcdcd, 0xd034bde468893434, 0x3d487a7590324848,
+	0xdbffab24e354ffff, 0xf57af78ff48d7a7a, 0x7a90f4ea3d649090, 0x615fc23ebe9d5f5f,
+	0x80201da0403d2020, 0xbd6867d5d00f6868, 0x681ad07234ca1a1a, 0x82ae192c41b7aeae,
+	0xeab4c95e757db4b4, 0x4d549a19a8ce5454, 0x7693ece53b7f9393, 0x88220daa442f2222,
+	0x8d6407e9c8636464, 0xe3f1db12ff2af1f1, 0xd173bfa2e6cc7373, 0x4812905a24821212,
+	0x1d403a5d807a4040, 0x2008402810480808, 0x2bc356e89b95c3c3, 0x97ec337bc5dfecec,
+	0x4bdb9690ab4ddbdb, 0xbea1611f5fc0a1a1, 0x0e8d1c8307918d8d, 0xf43df5c97ac83d3d,
+	0x6697ccf1335b9797, 0x0000000000000000, 0x1bcf36d483f9cfcf, 0xac2b4587566e2b2b,
+	0xc57697b3ece17676, 0x328264b019e68282, 0x7fd6fea9b128d6d6, 0x6c1bd87736c31b1b,
+	0xeeb5c15b7774b5b5, 0x86af112943beafaf, 0xb56a77dfd41d6a6a, 0x5d50ba0da0ea5050,
+	0x0945124c8a574545, 0xebf3cb18fb38f3f3, 0xc0309df060ad3030, 0x9bef2b74c3c4efef,
+	0xfc3fe5c37eda3f3f, 0x4955921caac75555, 0xb2a2791059dba2a2, 0x8fea0365c9e9eaea,
+	0x89650fecca6a6565, 0xd2bab9686903baba, 0xbc2f65935e4a2f2f, 0x27c04ee79d8ec0c0,
+	0x5fdebe81a160dede, 0x701ce06c38fc1c1c, 0xd3fdbb2ee746fdfd, 0x294d52649a1f4d4d,
+	0x7292e4e039769292, 0xc9758fbceafa7575, 0x1806301e0c360606, 0x128a249809ae8a8a,
+	0xf2b2f940794bb2b2, 0xbfe66359d185e6e6, 0x380e70361c7e0e0e, 0x7c1ff8633ee71f1f,
+	0x956237f7c4556262, 0x77d4eea3b53ad4d4, 0x9aa829324d81a8a8, 0x6296c4f431529696,
+	0xc3f99b3aef62f9f9, 0x33c566f697a3c5c5, 0x942535b14a102525, 0x7959f220b2ab5959,
+	0x2a8454ae15d08484, 0xd572b7a7e4c57272, 0xe439d5dd72ec3939, 0x2d4c5a6198164c4c,
+	0x655eca3bbc945e5e, 0xfd78e785f09f7878, 0xe038ddd870e53838, 0x0a8c148605988c8c,
+	0x63d1c6b2bf17d1d1, 0xaea5410b57e4a5a5, 0xafe2434dd9a1e2e2, 0x99612ff8c24e6161,
+	0xf6b3f1457b42b3b3, 0x842115a542342121, 0x4a9c94d625089c9c, 0x781ef0663cee1e1e,
+	0x1143225286614343, 0x3bc776fc93b1c7c7, 0xd7fcb32be54ffcfc, 0x1004201408240404,
+	0x5951b208a2e35151, 0x5e99bcc72f259999, 0xa96d4fc4da226d6d, 0x340d68391a650d0d,
+	0xcffa8335e979fafa, 0x5bdfb684a369dfdf, 0xe57ed79bfca97e7e, 0x90243db448192424,
+	0xec3bc5d776fe3b3b, 0x96ab313d4b9aabab, 0x1fce3ed181f0cece, 0x4411885522991111,
+	0x068f0c8903838f8f, 0x254e4a6b9c044e4e, 0xe6b7d1517366b7b7, 0x8beb0b60cbe0ebeb,
+	0xf03cfdcc78c13c3c, 0x3e817cbf1ffd8181, 0x6a94d4fe35409494, 0xfbf7eb0cf31cf7f7,
+	0xdeb9a1676f18b9b9, 0x4c13985f268b1313, 0xb02c7d9c58512c2c, 0x6bd3d6b8bb05d3d3,
+	0xbbe76b5cd38ce7e7, 0xa56e57cbdc396e6e, 0x37c46ef395aac4c4, 0x0c03180f061b0303,
+	0x45568a13acdc5656, 0x0d441a49885e4444, 0xe17fdf9efea07f7f, 0x9ea921374f88a9a9,
+	0xa82a4d8254672a2a, 0xd6bbb16d6b0abbbb, 0x23c146e29f87c1c1, 0x5153a202a6f15353,
+	0x57dcae8ba572dcdc, 0x2c0b582716530b0b, 0x4e9d9cd327019d9d, 0xad6c47c1d82b6c6c,
+	0xc43195f562a43131, 0xcd7487b9e8f37474, 0xfff6e309f115f6f6, 0x05460a438c4c4646,
+	0x8aac092645a5acac, 0x1e893c970fb58989, 0x5014a04428b41414, 0xa3e15b42dfbae1e1,
+	0x5816b04e2ca61616, 0xe83acdd274f73a3a, 0xb9696fd0d2066969, 0x2409482d12410909,
+	0xdd70a7ade0d77070, 0xe2b6d954716fb6b6, 0x67d0ceb7bd1ed0d0, 0x93ed3b7ec7d6eded,
+	0x17cc2edb85e2cccc, 0x15422a5784684242, 0x5a98b4c22d2c9898, 0xaaa4490e55eda4a4,
+	0xa0285d8850752828, 0x6d5cda31b8865c5c, 0xc7f8933fed6bf8f8, 0x228644a411c28686,
+}
+
+C7 := [256]u64 {
+	0x186018c07830d818, 0x238c2305af462623, 0xc63fc67ef991b8c6, 0xe887e8136fcdfbe8,
+	0x8726874ca113cb87, 0xb8dab8a9626d11b8, 0x0104010805020901, 0x4f214f426e9e0d4f,
+	0x36d836adee6c9b36, 0xa6a2a6590451ffa6, 0xd26fd2debdb90cd2, 0xf5f3f5fb06f70ef5,
+	0x79f979ef80f29679, 0x6fa16f5fcede306f, 0x917e91fcef3f6d91, 0x525552aa07a4f852,
+	0x609d6027fdc04760, 0xbccabc89766535bc, 0x9b569baccd2b379b, 0x8e028e048c018a8e,
+	0xa3b6a371155bd2a3, 0x0c300c603c186c0c, 0x7bf17bff8af6847b, 0x35d435b5e16a8035,
+	0x1d741de8693af51d, 0xe0a7e05347ddb3e0, 0xd77bd7f6acb321d7, 0xc22fc25eed999cc2,
+	0x2eb82e6d965c432e, 0x4b314b627a96294b, 0xfedffea321e15dfe, 0x5741578216aed557,
+	0x155415a8412abd15, 0x77c1779fb6eee877, 0x37dc37a5eb6e9237, 0xe5b3e57b56d79ee5,
+	0x9f469f8cd923139f, 0xf0e7f0d317fd23f0, 0x4a354a6a7f94204a, 0xda4fda9e95a944da,
+	0x587d58fa25b0a258, 0xc903c906ca8fcfc9, 0x29a429558d527c29, 0x0a280a5022145a0a,
+	0xb1feb1e14f7f50b1, 0xa0baa0691a5dc9a0, 0x6bb16b7fdad6146b, 0x852e855cab17d985,
+	0xbdcebd8173673cbd, 0x5d695dd234ba8f5d, 0x1040108050209010, 0xf4f7f4f303f507f4,
+	0xcb0bcb16c08bddcb, 0x3ef83eedc67cd33e, 0x05140528110a2d05, 0x6781671fe6ce7867,
+	0xe4b7e47353d597e4, 0x279c2725bb4e0227, 0x4119413258827341, 0x8b168b2c9d0ba78b,
+	0xa7a6a7510153f6a7, 0x7de97dcf94fab27d, 0x956e95dcfb374995, 0xd847d88e9fad56d8,
+	0xfbcbfb8b30eb70fb, 0xee9fee2371c1cdee, 0x7ced7cc791f8bb7c, 0x66856617e3cc7166,
+	0xdd53dda68ea77bdd, 0x175c17b84b2eaf17, 0x47014702468e4547, 0x9e429e84dc211a9e,
+	0xca0fca1ec589d4ca, 0x2db42d75995a582d, 0xbfc6bf9179632ebf, 0x071c07381b0e3f07,
+	0xad8ead012347acad, 0x5a755aea2fb4b05a, 0x8336836cb51bef83, 0x33cc3385ff66b633,
+	0x6391633ff2c65c63, 0x020802100a041202, 0xaa92aa39384993aa, 0x71d971afa8e2de71,
+	0xc807c80ecf8dc6c8, 0x196419c87d32d119, 0x4939497270923b49, 0xd943d9869aaf5fd9,
+	0xf2eff2c31df931f2, 0xe3abe34b48dba8e3, 0x5b715be22ab6b95b, 0x881a8834920dbc88,
+	0x9a529aa4c8293e9a, 0x2698262dbe4c0b26, 0x32c8328dfa64bf32, 0xb0fab0e94a7d59b0,
+	0xe983e91b6acff2e9, 0x0f3c0f78331e770f, 0xd573d5e6a6b733d5, 0x803a8074ba1df480,
+	0xbec2be997c6127be, 0xcd13cd26de87ebcd, 0x34d034bde4688934, 0x483d487a75903248,
+	0xffdbffab24e354ff, 0x7af57af78ff48d7a, 0x907a90f4ea3d6490, 0x5f615fc23ebe9d5f,
+	0x2080201da0403d20, 0x68bd6867d5d00f68, 0x1a681ad07234ca1a, 0xae82ae192c41b7ae,
+	0xb4eab4c95e757db4, 0x544d549a19a8ce54, 0x937693ece53b7f93, 0x2288220daa442f22,
+	0x648d6407e9c86364, 0xf1e3f1db12ff2af1, 0x73d173bfa2e6cc73, 0x124812905a248212,
+	0x401d403a5d807a40, 0x0820084028104808, 0xc32bc356e89b95c3, 0xec97ec337bc5dfec,
+	0xdb4bdb9690ab4ddb, 0xa1bea1611f5fc0a1, 0x8d0e8d1c8307918d, 0x3df43df5c97ac83d,
+	0x976697ccf1335b97, 0x0000000000000000, 0xcf1bcf36d483f9cf, 0x2bac2b4587566e2b,
+	0x76c57697b3ece176, 0x82328264b019e682, 0xd67fd6fea9b128d6, 0x1b6c1bd87736c31b,
+	0xb5eeb5c15b7774b5, 0xaf86af112943beaf, 0x6ab56a77dfd41d6a, 0x505d50ba0da0ea50,
+	0x450945124c8a5745, 0xf3ebf3cb18fb38f3, 0x30c0309df060ad30, 0xef9bef2b74c3c4ef,
+	0x3ffc3fe5c37eda3f, 0x554955921caac755, 0xa2b2a2791059dba2, 0xea8fea0365c9e9ea,
+	0x6589650fecca6a65, 0xbad2bab9686903ba, 0x2fbc2f65935e4a2f, 0xc027c04ee79d8ec0,
+	0xde5fdebe81a160de, 0x1c701ce06c38fc1c, 0xfdd3fdbb2ee746fd, 0x4d294d52649a1f4d,
+	0x927292e4e0397692, 0x75c9758fbceafa75, 0x061806301e0c3606, 0x8a128a249809ae8a,
+	0xb2f2b2f940794bb2, 0xe6bfe66359d185e6, 0x0e380e70361c7e0e, 0x1f7c1ff8633ee71f,
+	0x62956237f7c45562, 0xd477d4eea3b53ad4, 0xa89aa829324d81a8, 0x966296c4f4315296,
+	0xf9c3f99b3aef62f9, 0xc533c566f697a3c5, 0x25942535b14a1025, 0x597959f220b2ab59,
+	0x842a8454ae15d084, 0x72d572b7a7e4c572, 0x39e439d5dd72ec39, 0x4c2d4c5a6198164c,
+	0x5e655eca3bbc945e, 0x78fd78e785f09f78, 0x38e038ddd870e538, 0x8c0a8c148605988c,
+	0xd163d1c6b2bf17d1, 0xa5aea5410b57e4a5, 0xe2afe2434dd9a1e2, 0x6199612ff8c24e61,
+	0xb3f6b3f1457b42b3, 0x21842115a5423421, 0x9c4a9c94d625089c, 0x1e781ef0663cee1e,
+	0x4311432252866143, 0xc73bc776fc93b1c7, 0xfcd7fcb32be54ffc, 0x0410042014082404,
+	0x515951b208a2e351, 0x995e99bcc72f2599, 0x6da96d4fc4da226d, 0x0d340d68391a650d,
+	0xfacffa8335e979fa, 0xdf5bdfb684a369df, 0x7ee57ed79bfca97e, 0x2490243db4481924,
+	0x3bec3bc5d776fe3b, 0xab96ab313d4b9aab, 0xce1fce3ed181f0ce, 0x1144118855229911,
+	0x8f068f0c8903838f, 0x4e254e4a6b9c044e, 0xb7e6b7d1517366b7, 0xeb8beb0b60cbe0eb,
+	0x3cf03cfdcc78c13c, 0x813e817cbf1ffd81, 0x946a94d4fe354094, 0xf7fbf7eb0cf31cf7,
+	0xb9deb9a1676f18b9, 0x134c13985f268b13, 0x2cb02c7d9c58512c, 0xd36bd3d6b8bb05d3,
+	0xe7bbe76b5cd38ce7, 0x6ea56e57cbdc396e, 0xc437c46ef395aac4, 0x030c03180f061b03,
+	0x5645568a13acdc56, 0x440d441a49885e44, 0x7fe17fdf9efea07f, 0xa99ea921374f88a9,
+	0x2aa82a4d8254672a, 0xbbd6bbb16d6b0abb, 0xc123c146e29f87c1, 0x535153a202a6f153,
+	0xdc57dcae8ba572dc, 0x0b2c0b582716530b, 0x9d4e9d9cd327019d, 0x6cad6c47c1d82b6c,
+	0x31c43195f562a431, 0x74cd7487b9e8f374, 0xf6fff6e309f115f6, 0x4605460a438c4c46,
+	0xac8aac092645a5ac, 0x891e893c970fb589, 0x145014a04428b414, 0xe1a3e15b42dfbae1,
+	0x165816b04e2ca616, 0x3ae83acdd274f73a, 0x69b9696fd0d20669, 0x092409482d124109,
+	0x70dd70a7ade0d770, 0xb6e2b6d954716fb6, 0xd067d0ceb7bd1ed0, 0xed93ed3b7ec7d6ed,
+	0xcc17cc2edb85e2cc, 0x4215422a57846842, 0x985a98b4c22d2c98, 0xa4aaa4490e55eda4,
+	0x28a0285d88507528, 0x5c6d5cda31b8865c, 0xf8c7f8933fed6bf8, 0x86228644a411c286,
+}
+
+RC := [ROUNDS + 1]u64 {
+	0x0000000000000000,
+	0x1823c6e887b8014f,
+	0x36a6d2f5796f9152,
+	0x60bc9b8ea30c7b35,
+	0x1de0d7c22e4bfe57,
+	0x157737e59ff04ada,
+	0x58c9290ab1a06b85,
+	0xbd5d10f4cb3e0567,
+	0xe427418ba77d95d8,
+	0xfbee7c66dd17479e,
+	0xca2dbf07ad5a8333,
+}
+
+transform :: proc (ctx: ^Whirlpool_Context) {
+	K, block, state, L: [8]u64
+
+	for i := 0; i < 8; i += 1 {block[i] = util.U64_BE(ctx.buffer[8 * i:])}
+
+	for i := 0; i < 8; i += 1 {
+		K[i] = ctx.hash[i]
+		state[i] = block[i] ~ K[i]
+	}
+
+	for r := 1; r <= ROUNDS; r += 1 {
+		for i := 0; i < 8; i += 1 {
+			L[i] = C0[byte(K[i % 8] >> 56)] ~
+				C1[byte(K[(i + 7) % 8] >> 48)] ~
+				C2[byte(K[(i + 6) % 8] >> 40)] ~
+				C3[byte(K[(i + 5) % 8] >> 32)] ~
+				C4[byte(K[(i + 4) % 8] >> 24)] ~
+				C5[byte(K[(i + 3) % 8] >> 16)] ~
+				C6[byte(K[(i + 2) % 8] >> 8)] ~
+				C7[byte(K[(i + 1) % 8])]
+		}
+		L[0] ~= RC[r]
+
+		for i := 0; i < 8; i += 1 {K[i] = L[i]}
+
+		for i := 0; i < 8; i += 1 {
+			L[i] = C0[byte(state[i % 8] >> 56)] ~
+				C1[byte(state[(i + 7) % 8] >> 48)] ~
+				C2[byte(state[(i + 6) % 8] >> 40)] ~
+				C3[byte(state[(i + 5) % 8] >> 32)] ~
+				C4[byte(state[(i + 4) % 8] >> 24)] ~
+				C5[byte(state[(i + 3) % 8] >> 16)] ~
+				C6[byte(state[(i + 2) % 8] >> 8)] ~
+				C7[byte(state[(i + 1) % 8])] ~
+				K[i % 8]
+		}
+		for i := 0; i < 8; i += 1 {state[i] = L[i]}
+	}
+	for i := 0; i < 8; i += 1 {ctx.hash[i] ~= state[i] ~ block[i]}
+}

core/crypto/x25519/x25519.odin

@@ -0,0 +1,126 @@
+package x25519
+
+import field "core:crypto/_fiat/field_curve25519"
+import "core:mem"
+
+SCALAR_SIZE :: 32
+POINT_SIZE :: 32
+
+_BASE_POINT: [32]byte = {9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+_scalar_bit :: #force_inline proc "contextless" (s: ^[32]byte, i: int) -> u8 {
+	if i < 0 {
+		return 0
+	}
+	return (s[i>>3] >> uint(i&7)) & 1
+}
+
+_scalarmult :: proc (out, scalar, point: ^[32]byte) {
+	// Montgomery pseduo-multiplication taken from Monocypher.
+
+	// computes the scalar product
+	x1: field.Tight_Field_Element = ---
+	field.fe_from_bytes(&x1, point)
+
+	// computes the actual scalar product (the result is in x2 and z2)
+	x2, x3, z2, z3: field.Tight_Field_Element =  ---, ---, ---, ---
+	t0, t1: field.Loose_Field_Element = ---, ---
+
+	// Montgomery ladder
+	// In projective coordinates, to avoid divisions: x = X / Z
+	// We don't care about the y coordinate, it's only 1 bit of information
+	field.fe_one(&x2) // "zero" point
+	field.fe_zero(&z2)
+	field.fe_set(&x3, &x1) // "one" point
+	field.fe_one(&z3)
+
+	swap: int
+	for pos := 255-1; pos >= 0; pos = pos - 1 	{
+		// constant time conditional swap before ladder step
+		b := int(_scalar_bit(scalar, pos))
+		swap ~= b // xor trick avoids swapping at the end of the loop
+		field.fe_cond_swap(&x2, &x3, swap)
+		field.fe_cond_swap(&z2, &z3, swap)
+		swap = b // anticipates one last swap after the loop
+
+		// Montgomery ladder step: replaces (P2, P3) by (P2*2, P2+P3)
+		// with differential addition
+		//
+		// Note: This deliberately omits reductions after add/sub operations
+		// if the result is only ever used as the input to a mul/square since
+		// the implementations of those can deal with non-reduced inputs.
+		//
+		// fe_tighten_cast is only used to store a fully reduced
+		// output in a Loose_Field_Element, or to provide such a
+		// Loose_Field_Element as a Tight_Field_Element argument.
+		field.fe_sub(&t0, &x3, &z3)
+		field.fe_sub(&t1, &x2, &z2)
+		field.fe_add(field.fe_relax_cast(&x2), &x2, &z2) // x2 - unreduced
+		field.fe_add(field.fe_relax_cast(&z2), &x3, &z3) // z2 - unreduced
+		field.fe_carry_mul(&z3, &t0, field.fe_relax_cast(&x2))
+		field.fe_carry_mul(&z2, field.fe_relax_cast(&z2), &t1) // z2 - reduced
+		field.fe_carry_square(field.fe_tighten_cast(&t0), &t1) // t0 - reduced
+		field.fe_carry_square(field.fe_tighten_cast(&t1), field.fe_relax_cast(&x2)) // t1 - reduced
+		field.fe_add(field.fe_relax_cast(&x3), &z3, &z2) // x3 - unreduced
+		field.fe_sub(field.fe_relax_cast(&z2), &z3, &z2) // z2 - unreduced
+		field.fe_carry_mul(&x2, &t1, &t0) // x2 - reduced
+		field.fe_sub(&t1, field.fe_tighten_cast(&t1), field.fe_tighten_cast(&t0)) // safe - t1/t0 is reduced
+		field.fe_carry_square(&z2, field.fe_relax_cast(&z2)) // z2 - reduced
+		field.fe_carry_scmul_121666(&z3, &t1)
+		field.fe_carry_square(&x3, field.fe_relax_cast(&x3)) // x3 - reduced
+		field.fe_add(&t0, field.fe_tighten_cast(&t0), &z3) // safe - t0 is reduced
+		field.fe_carry_mul(&z3, field.fe_relax_cast(&x1), field.fe_relax_cast(&z2))
+		field.fe_carry_mul(&z2, &t1, &t0)
+	}
+	// last swap is necessary to compensate for the xor trick
+	// Note: after this swap, P3 == P2 + P1.
+	field.fe_cond_swap(&x2, &x3, swap)
+	field.fe_cond_swap(&z2, &z3, swap)
+
+	// normalises the coordinates: x == X / Z
+	field.fe_carry_inv(&z2, field.fe_relax_cast(&z2))
+	field.fe_carry_mul(&x2, field.fe_relax_cast(&x2), field.fe_relax_cast(&z2))
+	field.fe_to_bytes(out, &x2)
+
+	mem.zero_explicit(&x1, size_of(x1))
+	mem.zero_explicit(&x2, size_of(x2))
+	mem.zero_explicit(&x3, size_of(x3))
+	mem.zero_explicit(&z2, size_of(z2))
+	mem.zero_explicit(&z3, size_of(z3))
+	mem.zero_explicit(&t0, size_of(t0))
+	mem.zero_explicit(&t1, size_of(t1))
+}
+
+scalarmult :: proc (dst, scalar, point: []byte) {
+	if len(scalar) != SCALAR_SIZE {
+		panic("crypto/x25519: invalid scalar size")
+	}
+	if len(point) != POINT_SIZE {
+		panic("crypto/x25519: invalid point size")
+	}
+	if len(dst) != POINT_SIZE {
+		panic("crypto/x25519: invalid destination point size")
+	}
+
+	// "clamp" the scalar
+	e: [32]byte = ---
+	copy_slice(e[:], scalar)
+	e[0] &= 248
+	e[31] &= 127
+	e[31] |= 64
+
+	p: [32]byte = ---
+	copy_slice(p[:], point)
+
+	d: [32]byte = ---
+	_scalarmult(&d, &e, &p)
+	copy_slice(dst, d[:])
+
+	mem.zero_explicit(&e, size_of(e))
+	mem.zero_explicit(&d, size_of(d))
+}
+
+scalarmult_basepoint :: proc (dst, scalar: []byte) {
+	// TODO/perf: Switch to using a precomputed table.
+	scalarmult(dst, scalar, _BASE_POINT[:])
+}

core/encoding/json/marshal.odin

@@ -18,7 +18,7 @@ Marshal_Error :: union {
 
 marshal :: proc(v: any, allocator := context.allocator) -> (data: []byte, err: Marshal_Error) {
 	b := strings.make_builder(allocator)
-	defer if err != nil || data == nil {
+	defer if err != .None {
 		strings.destroy_builder(&b)
 	}
 
@@ -27,7 +27,7 @@ marshal :: proc(v: any, allocator := context.allocator) -> (data: []byte, err: M
 	if len(b.buf) != 0 {
 		data = b.buf[:]
 	}
-	return
+	return data, .None
 }
 
 marshal_to_builder :: proc(b: ^strings.Builder, v: any) -> Marshal_Error {
@@ -35,23 +35,6 @@ marshal_to_builder :: proc(b: ^strings.Builder, v: any) -> Marshal_Error {
 }
 
 marshal_to_writer :: proc(w: io.Writer, v: any) -> (err: Marshal_Error) {
-	write_f64 :: proc(w: io.Writer, val: f64, size: int) -> io.Error {
-		buf: [386]byte
-
-		str := strconv.append_float(buf[1:], val, 'f', 2*size, 8*size)
-		s := buf[:len(str)+1]
-		if s[1] == '+' || s[1] == '-' {
-			s = s[1:]
-		} else {
-			s[0] = '+'
-		}
-		if s[0] == '+' {
-			s = s[1:]
-		}
-
-		_ = io.write_string(w, string(s)) or_return
-		return nil
-	}	
 	if v == nil {
 		io.write_string(w, "null") or_return
 		return
@@ -177,6 +160,9 @@ marshal_to_writer :: proc(w: io.Writer, v: any) -> (err: Marshal_Error) {
 
 	case runtime.Type_Info_Relative_Slice:
 		return .Unsupported_Type
+		
+	case runtime.Type_Info_Matrix:
+		return .Unsupported_Type
 
 	case runtime.Type_Info_Array:
 		io.write_byte(w, '[') or_return

core/encoding/json/parser.odin

@@ -106,6 +106,7 @@ parse_comma :: proc(p: ^Parser) -> (do_break: bool) {
 }
 
 parse_value :: proc(p: ^Parser) -> (value: Value, err: Error) {
+	err = .None
 	token := p.curr_token
 	#partial switch token.kind {
 	case .Null:
@@ -175,6 +176,7 @@ parse_value :: proc(p: ^Parser) -> (value: Value, err: Error) {
 }
 
 parse_array :: proc(p: ^Parser) -> (value: Value, err: Error) {
+	err = .None
 	expect_token(p, .Open_Bracket) or_return
 
 	array: Array
@@ -266,15 +268,14 @@ parse_object_body :: proc(p: ^Parser, end_token: Token_Kind) -> (obj: Object, er
 			break
 		}
 	}	
-	return
+	return obj, .None
 }
 
 parse_object :: proc(p: ^Parser) -> (value: Value, err: Error) {
 	expect_token(p, .Open_Brace) or_return
 	obj := parse_object_body(p, .Close_Brace) or_return
 	expect_token(p, .Close_Brace) or_return
-	value = obj
-	return
+	return obj, .None
 }
 
 

core/encoding/json/types.odin

@@ -34,6 +34,7 @@ Specification :: enum {
 	JSON,
 	JSON5, // https://json5.org/
 	MJSON, // https://bitsquid.blogspot.com/2009/10/simplified-json-notation.html
+	Bitsquid = MJSON,
 }
 
 

core/encoding/json/unmarshal.odin

@@ -222,6 +222,7 @@ unmarsal_value :: proc(p: ^Parser, v: any) -> (err: Unmarshal_Error) {
 		advance_token(p)
 		return
 	case .False, .True:
+		advance_token(p)
 		if assign_bool(v, token.kind == .True) {
 			return
 		}

core/fmt/fmt.odin

@@ -2,7 +2,6 @@ package fmt
 
 import "core:math/bits"
 import "core:mem"
-import "core:os"
 import "core:io"
 import "core:reflect"
 import "core:runtime"
@@ -62,38 +61,6 @@ register_user_formatter :: proc(id: typeid, formatter: User_Formatter) -> Regist
 }
 
 
-fprint :: proc(fd: os.Handle, args: ..any, sep := " ") -> int {
-	w := io.to_writer(os.stream_from_handle(fd))
-	return wprint(w=w, args=args, sep=sep)
-}
-
-fprintln :: proc(fd: os.Handle, args: ..any, sep := " ") -> int {
-	w := io.to_writer(os.stream_from_handle(fd))
-	return wprintln(w=w, args=args, sep=sep)
-}
-fprintf :: proc(fd: os.Handle, fmt: string, args: ..any) -> int {
-	w := io.to_writer(os.stream_from_handle(fd))
-	return wprintf(w, fmt, ..args)
-}
-fprint_type :: proc(fd: os.Handle, info: ^runtime.Type_Info) -> (n: int, err: io.Error) {
-	w := io.to_writer(os.stream_from_handle(fd))
-	return wprint_type(w, info)
-}
-fprint_typeid :: proc(fd: os.Handle, id: typeid) -> (n: int, err: io.Error) {
-	w := io.to_writer(os.stream_from_handle(fd))
-	return wprint_typeid(w, id)
-}
-
-// print* procedures return the number of bytes written
-print   :: proc(args: ..any, sep := " ") -> int { return fprint(fd=os.stdout, args=args, sep=sep) }
-println :: proc(args: ..any, sep := " ") -> int { return fprintln(fd=os.stdout, args=args, sep=sep) }
-printf  :: proc(fmt: string, args: ..any) -> int { return fprintf(os.stdout, fmt, ..args) }
-
-eprint   :: proc(args: ..any, sep := " ") -> int { return fprint(fd=os.stderr, args=args, sep=sep) }
-eprintln :: proc(args: ..any, sep := " ") -> int { return fprintln(fd=os.stderr, args=args, sep=sep) }
-eprintf  :: proc(fmt: string, args: ..any) -> int { return fprintf(os.stderr, fmt, ..args) }
-
-
 // aprint* procedures return a string that was allocated with the current context
 // They must be freed accordingly
 aprint :: proc(args: ..any, sep := " ") -> string {
@@ -1953,6 +1920,47 @@ fmt_value :: proc(fi: ^Info, v: any, verb: rune) {
 			}
 		}
 
+	case runtime.Type_Info_Matrix:
+		io.write_string(fi.writer, "matrix[")
+		defer io.write_byte(fi.writer, ']')
+		
+		fi.indent += 1
+		
+		if fi.hash { 
+			// Printed as it is written
+			io.write_byte(fi.writer, '\n')
+			for row in 0..<info.row_count {
+				fmt_write_indent(fi)
+				for col in 0..<info.column_count {
+					if col > 0 { io.write_string(fi.writer, ", ") }
+					
+					offset := (row + col*info.elem_stride)*info.elem_size
+					
+					data := uintptr(v.data) + uintptr(offset)
+					fmt_arg(fi, any{rawptr(data), info.elem.id}, verb)
+				}
+				io.write_string(fi.writer, ",\n")
+			}
+		} else {
+			// Printed in Row-Major layout to match text layout
+			for row in 0..<info.row_count {
+				if row > 0 { io.write_string(fi.writer, "; ") }
+				for col in 0..<info.column_count {
+					if col > 0 { io.write_string(fi.writer, ", ") }
+					
+					offset := (row + col*info.elem_stride)*info.elem_size
+					
+					data := uintptr(v.data) + uintptr(offset)
+					fmt_arg(fi, any{rawptr(data), info.elem.id}, verb)
+				}
+			}
+		}
+		
+		fi.indent -= 1
+		
+		if fi.hash { 
+			fmt_write_indent(fi)
+		}
 	}
 }
 
@@ -2068,9 +2076,11 @@ fmt_arg :: proc(fi: ^Info, arg: any, verb: rune) {
 	case f32be:      fmt_float(fi, f64(a), 32, verb)
 	case f64be:      fmt_float(fi, f64(a), 64, verb)
 
+	case complex32:  fmt_complex(fi, complex128(a), 32, verb)
 	case complex64:  fmt_complex(fi, complex128(a), 64, verb)
 	case complex128: fmt_complex(fi, a, 128, verb)
 
+	case quaternion64:  fmt_quaternion(fi, quaternion256(a),  64, verb)
 	case quaternion128: fmt_quaternion(fi, quaternion256(a), 128, verb)
 	case quaternion256: fmt_quaternion(fi, a, 256, verb)
 

core/fmt/fmt_js.odin

@@ -0,0 +1,44 @@
+//+build js
+package fmt
+
+import "core:io"
+
+foreign import "odin_env"
+
+@(private="file")
+foreign odin_env {
+	write :: proc "c" (fd: u32, p: []byte) ---
+}
+
+@(private="file")
+write_vtable := &io.Stream_VTable{
+	impl_write = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
+		fd := u32(uintptr(s.stream_data))
+		write(fd, p)
+		return len(p), nil
+	},	
+}
+
+@(private="file")
+stdout := io.Writer{
+	stream = {
+		stream_vtable = write_vtable,
+		stream_data = rawptr(uintptr(1)),
+	},
+}
+@(private="file")
+stderr := io.Writer{
+	stream = {
+		stream_vtable = write_vtable,
+		stream_data = rawptr(uintptr(2)),
+	},
+}
+
+// print* procedures return the number of bytes written
+print   :: proc(args: ..any, sep := " ") -> int { return wprint(w=stdout, args=args, sep=sep) }
+println :: proc(args: ..any, sep := " ") -> int { return wprintln(w=stdout, args=args, sep=sep) }
+printf  :: proc(fmt: string, args: ..any) -> int { return wprintf(stdout, fmt, ..args) }
+
+eprint   :: proc(args: ..any, sep := " ") -> int { return wprint(w=stderr, args=args, sep=sep) }
+eprintln :: proc(args: ..any, sep := " ") -> int { return wprintln(w=stderr, args=args, sep=sep) }
+eprintf  :: proc(fmt: string, args: ..any) -> int { return wprintf(stderr, fmt, ..args) }

core/fmt/fmt_os.odin

@@ -0,0 +1,37 @@
+//+build !freestanding !js
+package fmt
+
+import "core:runtime"
+import "core:os"
+import "core:io"
+
+fprint :: proc(fd: os.Handle, args: ..any, sep := " ") -> int {
+	w := io.to_writer(os.stream_from_handle(fd))
+	return wprint(w=w, args=args, sep=sep)
+}
+
+fprintln :: proc(fd: os.Handle, args: ..any, sep := " ") -> int {
+	w := io.to_writer(os.stream_from_handle(fd))
+	return wprintln(w=w, args=args, sep=sep)
+}
+fprintf :: proc(fd: os.Handle, fmt: string, args: ..any) -> int {
+	w := io.to_writer(os.stream_from_handle(fd))
+	return wprintf(w, fmt, ..args)
+}
+fprint_type :: proc(fd: os.Handle, info: ^runtime.Type_Info) -> (n: int, err: io.Error) {
+	w := io.to_writer(os.stream_from_handle(fd))
+	return wprint_type(w, info)
+}
+fprint_typeid :: proc(fd: os.Handle, id: typeid) -> (n: int, err: io.Error) {
+	w := io.to_writer(os.stream_from_handle(fd))
+	return wprint_typeid(w, id)
+}
+
+// print* procedures return the number of bytes written
+print   :: proc(args: ..any, sep := " ") -> int { return fprint(fd=os.stdout, args=args, sep=sep) }
+println :: proc(args: ..any, sep := " ") -> int { return fprintln(fd=os.stdout, args=args, sep=sep) }
+printf  :: proc(fmt: string, args: ..any) -> int { return fprintf(os.stdout, fmt, ..args) }
+
+eprint   :: proc(args: ..any, sep := " ") -> int { return fprint(fd=os.stderr, args=args, sep=sep) }
+eprintln :: proc(args: ..any, sep := " ") -> int { return fprintln(fd=os.stderr, args=args, sep=sep) }
+eprintf  :: proc(fmt: string, args: ..any) -> int { return fprintf(os.stderr, fmt, ..args) }

core/hash/hash.odin

@@ -47,8 +47,8 @@ adler32 :: proc(data: []byte, seed := u32(1)) -> u32 #no_bounds_check {
 }
 
 @(optimization_mode="speed")
-djb2 :: proc(data: []byte) -> u32 {
-	hash: u32 = 5381
+djb2 :: proc(data: []byte, seed := u32(5381)) -> u32 {
+	hash: u32 = seed
 	for b in data {
 		hash = (hash << 5) + hash + u32(b) // hash * 33 + u32(b)
 	}
@@ -56,8 +56,8 @@ djb2 :: proc(data: []byte) -> u32 {
 }
 
 @(optimization_mode="speed")
-fnv32 :: proc(data: []byte) -> u32 {
-	h: u32 = 0x811c9dc5
+fnv32 :: proc(data: []byte, seed := u32(0x811c9dc5)) -> u32 {
+	h: u32 = seed
 	for b in data {
 		h = (h * 0x01000193) ~ u32(b)
 	}
@@ -65,8 +65,8 @@ fnv32 :: proc(data: []byte) -> u32 {
 }
 
 @(optimization_mode="speed")
-fnv64 :: proc(data: []byte) -> u64 {
-	h: u64 = 0xcbf29ce484222325
+fnv64 :: proc(data: []byte, seed := u64(0xcbf29ce484222325)) -> u64 {
+	h: u64 = seed
 	for b in data {
 		h = (h * 0x100000001b3) ~ u64(b)
 	}
@@ -74,8 +74,8 @@ fnv64 :: proc(data: []byte) -> u64 {
 }
 
 @(optimization_mode="speed")
-fnv32a :: proc(data: []byte) -> u32 {
-	h: u32 = 0x811c9dc5
+fnv32a :: proc(data: []byte, seed := u32(0x811c9dc5)) -> u32 {
+	h: u32 = seed
 	for b in data {
 		h = (h ~ u32(b)) * 0x01000193
 	}
@@ -83,8 +83,8 @@ fnv32a :: proc(data: []byte) -> u32 {
 }
 
 @(optimization_mode="speed")
-fnv64a :: proc(data: []byte) -> u64 {
-	h: u64 = 0xcbf29ce484222325
+fnv64a :: proc(data: []byte, seed := u64(0xcbf29ce484222325)) -> u64 {
+	h: u64 = seed
 	for b in data {
 		h = (h ~ u64(b)) * 0x100000001b3
 	}
@@ -92,8 +92,8 @@ fnv64a :: proc(data: []byte) -> u64 {
 }
 
 @(optimization_mode="speed")
-jenkins :: proc(data: []byte) -> u32 {
-	hash: u32 = 0
+jenkins :: proc(data: []byte, seed := u32(0)) -> u32 {
+	hash: u32 = seed
 	for b in data {
 		hash += u32(b)
 		hash += hash << 10
@@ -106,11 +106,11 @@ jenkins :: proc(data: []byte) -> u32 {
 }
 
 @(optimization_mode="speed")
-murmur32 :: proc(data: []byte) -> u32 {
+murmur32 :: proc(data: []byte, seed := u32(0)) -> u32 {
 	c1_32: u32 : 0xcc9e2d51
 	c2_32: u32 : 0x1b873593
 
-	h1: u32 = 0
+	h1: u32 = seed
 	nblocks := len(data)/4
 	p := raw_data(data)
 	p1 := mem.ptr_offset(p, 4*nblocks)
@@ -156,14 +156,12 @@ murmur32 :: proc(data: []byte) -> u32 {
 }
 
 @(optimization_mode="speed")
-murmur64 :: proc(data: []byte) -> u64 {
-	SEED :: 0x9747b28c
-
+murmur64 :: proc(data: []byte, seed := u64(0x9747b28c)) -> u64 {
 	when size_of(int) == 8 {
 		m :: 0xc6a4a7935bd1e995
 		r :: 47
 
-		h: u64 = SEED ~ (u64(len(data)) * m)
+		h: u64 = seed ~ (u64(len(data)) * m)
 		data64 := mem.slice_ptr(cast(^u64)raw_data(data), len(data)/size_of(u64))
 
 		for _, i in data64 {
@@ -198,8 +196,8 @@ murmur64 :: proc(data: []byte) -> u64 {
 		m :: 0x5bd1e995
 		r :: 24
 
-		h1 := u32(SEED) ~ u32(len(data))
-		h2 := u32(SEED) >> 32
+		h1 := u32(seed) ~ u32(len(data))
+		h2 := u32(seed) >> 32
 		data32 := mem.slice_ptr(cast(^u32)raw_data(data), len(data)/size_of(u32))
 		len := len(data)
 		i := 0
@@ -262,8 +260,8 @@ murmur64 :: proc(data: []byte) -> u64 {
 }
 
 @(optimization_mode="speed")
-sdbm :: proc(data: []byte) -> u32 {
-	hash: u32 = 0
+sdbm :: proc(data: []byte, seed := u32(0)) -> u32 {
+	hash: u32 = seed
 	for b in data {
 		hash = u32(b) + (hash<<6) + (hash<<16) - hash
 	}

core/image/common.odin

@@ -1,16 +1,17 @@
-package image
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
-	Made available under Odin's BSD-2 license.
+	Made available under Odin's BSD-3 license.
 
 	List of contributors:
 		Jeroen van Rijn: Initial implementation, optimization.
 		Ginger Bill:     Cosmetic changes.
 */
+package image
 
 import "core:bytes"
 import "core:mem"
+import "core:compress"
+import "core:runtime"
 
 Image :: struct {
 	width:         int,
@@ -25,8 +26,11 @@ Image :: struct {
 	*/
 	background:    Maybe([3]u16),
 
-	metadata_ptr:  rawptr,
-	metadata_type: typeid,
+	metadata:      Image_Metadata,
+}
+
+Image_Metadata :: union {
+	^PNG_Info,
 }
 
 /*
@@ -112,31 +116,140 @@ Option :: enum {
 }
 Options :: distinct bit_set[Option]
 
-Error :: enum {
+Error :: union {
+	General_Image_Error,
+	PNG_Error,
+
+	compress.Error,
+	compress.General_Error,
+	compress.Deflate_Error,
+	compress.ZLIB_Error,
+	runtime.Allocator_Error,
+}
+
+General_Image_Error :: enum {
+	None = 0,
+	Invalid_Image_Dimensions,
+	Image_Dimensions_Too_Large,
+	Image_Does_Not_Adhere_to_Spec,
+}
+
+PNG_Error :: enum {
 	Invalid_PNG_Signature,
 	IHDR_Not_First_Chunk,
 	IHDR_Corrupt,
 	IDAT_Missing,
 	IDAT_Must_Be_Contiguous,
 	IDAT_Corrupt,
-	PNG_Does_Not_Adhere_to_Spec,
+	IDAT_Size_Too_Large,
 	PLTE_Encountered_Unexpectedly,
 	PLTE_Invalid_Length,
 	TRNS_Encountered_Unexpectedly,
 	BKGD_Invalid_Length,
-	Invalid_Image_Dimensions,
 	Unknown_Color_Type,
 	Invalid_Color_Bit_Depth_Combo,
 	Unknown_Filter_Method,
 	Unknown_Interlace_Method,
 	Requested_Channel_Not_Present,
 	Post_Processing_Error,
+	Invalid_Chunk_Length,
+}
+
+/*
+	PNG-specific structs
+*/
+PNG_Info :: struct {
+	header: PNG_IHDR,
+	chunks: [dynamic]PNG_Chunk,
+}
+
+PNG_Chunk_Header :: struct #packed {
+	length: u32be,
+	type:   PNG_Chunk_Type,
+}
+
+PNG_Chunk :: struct #packed {
+	header: PNG_Chunk_Header,
+	data:   []byte,
+	crc:    u32be,
+}
+
+PNG_Chunk_Type :: enum u32be {
+	// IHDR must come first in a file
+	IHDR = 'I' << 24 | 'H' << 16 | 'D' << 8 | 'R',
+	// PLTE must precede the first IDAT chunk
+	PLTE = 'P' << 24 | 'L' << 16 | 'T' << 8 | 'E',
+	bKGD = 'b' << 24 | 'K' << 16 | 'G' << 8 | 'D',
+	tRNS = 't' << 24 | 'R' << 16 | 'N' << 8 | 'S',
+	IDAT = 'I' << 24 | 'D' << 16 | 'A' << 8 | 'T',
+
+	iTXt = 'i' << 24 | 'T' << 16 | 'X' << 8 | 't',
+	tEXt = 't' << 24 | 'E' << 16 | 'X' << 8 | 't',
+	zTXt = 'z' << 24 | 'T' << 16 | 'X' << 8 | 't',
+
+	iCCP = 'i' << 24 | 'C' << 16 | 'C' << 8 | 'P',
+	pHYs = 'p' << 24 | 'H' << 16 | 'Y' << 8 | 's',
+	gAMA = 'g' << 24 | 'A' << 16 | 'M' << 8 | 'A',
+	tIME = 't' << 24 | 'I' << 16 | 'M' << 8 | 'E',
+
+	sPLT = 's' << 24 | 'P' << 16 | 'L' << 8 | 'T',
+	sRGB = 's' << 24 | 'R' << 16 | 'G' << 8 | 'B',
+	hIST = 'h' << 24 | 'I' << 16 | 'S' << 8 | 'T',
+	cHRM = 'c' << 24 | 'H' << 16 | 'R' << 8 | 'M',
+	sBIT = 's' << 24 | 'B' << 16 | 'I' << 8 | 'T',
+
+	/*
+		eXIf tags are not part of the core spec, but have been ratified
+		in v1.5.0 of the PNG Ext register.
+
+		We will provide unprocessed chunks to the caller if `.return_metadata` is set.
+		Applications are free to implement an Exif decoder.
+	*/
+	eXIf = 'e' << 24 | 'X' << 16 | 'I' << 8 | 'f',
+
+	// PNG files must end with IEND
+	IEND = 'I' << 24 | 'E' << 16 | 'N' << 8 | 'D',
+
+	/*
+		XCode sometimes produces "PNG" files that don't adhere to the PNG spec.
+		We recognize them only in order to avoid doing further work on them.
+
+		Some tools like PNG Defry may be able to repair them, but we're not
+		going to reward Apple for producing proprietary broken files purporting
+		to be PNGs by supporting them.
+
+	*/
+	iDOT = 'i' << 24 | 'D' << 16 | 'O' << 8 | 'T',
+	CbGI = 'C' << 24 | 'b' << 16 | 'H' << 8 | 'I',
+}
+
+PNG_IHDR :: struct #packed {
+	width:              u32be,
+	height:             u32be,
+	bit_depth:          u8,
+	color_type:         PNG_Color_Type,
+	compression_method: u8,
+	filter_method:      u8,
+	interlace_method:   PNG_Interlace_Method,
+}
+PNG_IHDR_SIZE :: size_of(PNG_IHDR)
+#assert (PNG_IHDR_SIZE == 13)
+
+PNG_Color_Value :: enum u8 {
+	Paletted = 0, // 1 << 0 = 1
+	Color    = 1, // 1 << 1 = 2
+	Alpha    = 2, // 1 << 2 = 4
+}
+PNG_Color_Type :: distinct bit_set[PNG_Color_Value; u8]
+
+PNG_Interlace_Method :: enum u8 {
+	None  = 0,
+	Adam7 = 1,
 }
 
 /*
 	Functions to help with image buffer calculations
 */
-
 compute_buffer_size :: proc(width, height, channels, depth: int, extra_row_bytes := int(0)) -> (size: int) {
 	size = ((((channels * width * depth) + 7) >> 3) + extra_row_bytes) * height
 	return
@@ -145,7 +258,6 @@ compute_buffer_size :: proc(width, height, channels, depth: int, extra_row_bytes
 /*
 	For when you have an RGB(A) image, but want a particular channel.
 */
-
 Channel :: enum u8 {
 	R = 1,
 	G = 2,
@@ -207,8 +319,7 @@ return_single_channel :: proc(img: ^Image, channel: Channel) -> (res: ^Image, ok
 	res.depth         = img.depth
 	res.pixels        = t
 	res.background    = img.background
-	res.metadata_ptr  = img.metadata_ptr
-	res.metadata_type = img.metadata_type
+	res.metadata      = img.metadata
 
 	return res, true
 }

core/image/png/example.odin

@@ -1,9 +1,6 @@
-//+ignore
-package png
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
-	Made available under Odin's BSD-2 license.
+	Made available under Odin's BSD-3 license.
 
 	List of contributors:
 		Jeroen van Rijn: Initial implementation.
@@ -11,8 +8,9 @@ package png
 
 	An example of how to use `load`.
 */
+//+ignore
+package png
 
-import "core:compress"
 import "core:image"
 // import "core:image/png"
 import "core:bytes"
@@ -41,8 +39,8 @@ main :: proc() {
 demo :: proc() {
 	file: string
 
-	options := image.Options{} // {.return_metadata};
-	err:       compress.Error
+	options := image.Options{.return_metadata}
+	err:       image.Error
 	img:      ^image.Image
 
 	file = "../../../misc/logo-slim.png"
@@ -53,32 +51,33 @@ demo :: proc() {
 	if err != nil {
 		fmt.printf("Trying to read PNG file %v returned %v\n", file, err)
 	} else {
-		v: ^Info
-
 		fmt.printf("Image: %vx%vx%v, %v-bit.\n", img.width, img.height, img.channels, img.depth)
-		if img.metadata_ptr != nil && img.metadata_type == Info {
-			v = (^Info)(img.metadata_ptr)
 
+		if v, ok := img.metadata.(^image.PNG_Info); ok {
 			// Handle ancillary chunks as you wish.
 			// We provide helper functions for a few types.
 			for c in v.chunks {
 				#partial switch c.header.type {
 				case .tIME:
-					t, _ := core_time(c)
-					fmt.printf("[tIME]: %v\n", t)
+					if t, t_ok := core_time(c); t_ok {
+						fmt.printf("[tIME]: %v\n", t)
+					}
 				case .gAMA:
-					fmt.printf("[gAMA]: %v\n", gamma(c))
+					if gama, gama_ok := gamma(c); gama_ok {
+						fmt.printf("[gAMA]: %v\n", gama)
+					}
 				case .pHYs:
-					phys := phys(c)
-					if phys.unit == .Meter {
-						xm    := f32(img.width)  / f32(phys.ppu_x)
-						ym    := f32(img.height) / f32(phys.ppu_y)
-						dpi_x, dpi_y := phys_to_dpi(phys)
-						fmt.printf("[pHYs] Image resolution is %v x %v pixels per meter.\n", phys.ppu_x, phys.ppu_y)
-						fmt.printf("[pHYs] Image resolution is %v x %v DPI.\n", dpi_x, dpi_y)
-						fmt.printf("[pHYs] Image dimensions are %v x %v meters.\n", xm, ym)
-					} else {
-						fmt.printf("[pHYs] x: %v, y: %v pixels per unknown unit.\n", phys.ppu_x, phys.ppu_y)
+					if phys, phys_ok := phys(c); phys_ok {
+						if phys.unit == .Meter {
+							xm    := f32(img.width)  / f32(phys.ppu_x)
+							ym    := f32(img.height) / f32(phys.ppu_y)
+							dpi_x, dpi_y := phys_to_dpi(phys)
+							fmt.printf("[pHYs] Image resolution is %v x %v pixels per meter.\n", phys.ppu_x, phys.ppu_y)
+							fmt.printf("[pHYs] Image resolution is %v x %v DPI.\n", dpi_x, dpi_y)
+							fmt.printf("[pHYs] Image dimensions are %v x %v meters.\n", xm, ym)
+						} else {
+							fmt.printf("[pHYs] x: %v, y: %v pixels per unknown unit.\n", phys.ppu_x, phys.ppu_y)
+						}
 					}
 				case .iTXt, .zTXt, .tEXt:
 					res, ok_text := text(c)
@@ -93,8 +92,7 @@ demo :: proc() {
 				case .bKGD:
 					fmt.printf("[bKGD] %v\n", img.background)
 				case .eXIf:
-					res, ok_exif := exif(c)
-					if ok_exif {
+					if res, ok_exif := exif(c); ok_exif {
 						/*
 							Other than checking the signature and byte order, we don't handle Exif data.
 							If you wish to interpret it, pass it to an Exif parser.
@@ -102,20 +100,17 @@ demo :: proc() {
 						fmt.printf("[eXIf] %v\n", res)
 					}
 				case .PLTE:
-					plte, plte_ok := plte(c)
-					if plte_ok {
+					if plte, plte_ok := plte(c); plte_ok {
 						fmt.printf("[PLTE] %v\n", plte)
 					} else {
 						fmt.printf("[PLTE] Error\n")
 					}
 				case .hIST:
-					res, ok_hist := hist(c)
-					if ok_hist {
+					if res, ok_hist := hist(c); ok_hist {
 						fmt.printf("[hIST] %v\n", res)
 					}
 				case .cHRM:
-					res, ok_chrm := chrm(c)
-					if ok_chrm {
+					if res, ok_chrm := chrm(c); ok_chrm {
 						fmt.printf("[cHRM] %v\n", res)
 					}
 				case .sPLT:
@@ -147,6 +142,8 @@ demo :: proc() {
 		}
 	}
 
+	fmt.printf("Done parsing metadata.\n")
+
 	if err == nil && .do_not_decompress_image not_in options && .info not_in options {
 		if ok := write_image_as_ppm("out.ppm", img); ok {
 			fmt.println("Saved decoded image.")

core/image/png/helpers.odin

@@ -1,5 +1,3 @@
-package png
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
 	Made available under Odin's BSD-2 license.
@@ -10,6 +8,7 @@ package png
 
 	These are a few useful utility functions to work with PNG images.
 */
+package png
 
 import "core:image"
 import "core:compress/zlib"
@@ -34,15 +33,14 @@ destroy :: proc(img: ^Image) {
 	}
 
 	bytes.buffer_destroy(&img.pixels)
-	// Clean up Info.
-	free(img.metadata_ptr)
 
-	/*
-		We don't need to do anything for the individual chunks.
-		They're allocated on the temp allocator, as is info.chunks
-
-		See read_chunk.
-	*/
+	if v, ok := img.metadata.(^image.PNG_Info); ok {
+		for chunk in &v.chunks {
+			delete(chunk.data)
+		}
+		delete(v.chunks)
+		free(v)
+	}
 	free(img)
 }
 
@@ -50,46 +48,50 @@ destroy :: proc(img: ^Image) {
 	Chunk helpers
 */
 
-gamma :: proc(c: Chunk) -> f32 {
-	assert(c.header.type == .gAMA)
-	res := (^gAMA)(raw_data(c.data))^
-	when true {
-		// Returns the wrong result on old backend
-		// Fixed for -llvm-api
-		return f32(res.gamma_100k) / 100_000.0
-	} else {
-		return f32(u32(res.gamma_100k)) / 100_000.0
+gamma :: proc(c: image.PNG_Chunk) -> (res: f32, ok: bool) {
+	if c.header.type != .gAMA || len(c.data) != size_of(gAMA) {
+		return {}, false
 	}
+	gama := (^gAMA)(raw_data(c.data))^
+	return f32(gama.gamma_100k) / 100_000.0, true
 }
 
 INCHES_PER_METER :: 1000.0 / 25.4
 
-phys :: proc(c: Chunk) -> pHYs {
-	assert(c.header.type == .pHYs)
-	res := (^pHYs)(raw_data(c.data))^
-	return res
+phys :: proc(c: image.PNG_Chunk) -> (res: pHYs, ok: bool) {
+	if c.header.type != .pHYs || len(c.data) != size_of(pHYs) {
+		return {}, false
+	}
+
+	return (^pHYs)(raw_data(c.data))^, true 
 }
 
 phys_to_dpi :: proc(p: pHYs) -> (x_dpi, y_dpi: f32) {
 	return f32(p.ppu_x) / INCHES_PER_METER, f32(p.ppu_y) / INCHES_PER_METER
 }
 
-time :: proc(c: Chunk) -> tIME {
-	assert(c.header.type == .tIME)
-	res := (^tIME)(raw_data(c.data))^
-	return res
+time :: proc(c: image.PNG_Chunk) -> (res: tIME, ok: bool) {
+	if c.header.type != .tIME || len(c.data) != size_of(tIME) {
+		return {}, false
+	}
+
+	return (^tIME)(raw_data(c.data))^, true
 }
 
-core_time :: proc(c: Chunk) -> (t: coretime.Time, ok: bool) {
-	png_time := time(c)
-	using png_time
-	return coretime.datetime_to_time(
-		int(year), int(month), int(day),
-		int(hour), int(minute), int(second),
-	)
+core_time :: proc(c: image.PNG_Chunk) -> (t: coretime.Time, ok: bool) {
+	if png_time, png_ok := time(c); png_ok {
+		using png_time
+		return coretime.datetime_to_time(
+			int(year), int(month), int(day),
+			int(hour), int(minute), int(second),
+		)
+	} else {
+		return {}, false
+	}
 }
 
-text :: proc(c: Chunk) -> (res: Text, ok: bool) {
+text :: proc(c: image.PNG_Chunk) -> (res: Text, ok: bool) {
+	assert(len(c.data) == int(c.header.length))
 	#partial switch c.header.type {
 	case .tEXt:
 		ok = true
@@ -191,7 +193,7 @@ text_destroy :: proc(text: Text) {
 	delete(text.text)
 }
 
-iccp :: proc(c: Chunk) -> (res: iCCP, ok: bool) {
+iccp :: proc(c: image.PNG_Chunk) -> (res: iCCP, ok: bool) {
 	ok = true
 
 	fields := bytes.split_n(s=c.data, sep=[]u8{0}, n=3, allocator=context.temp_allocator)
@@ -227,10 +229,8 @@ iccp_destroy :: proc(i: iCCP) {
 
 }
 
-srgb :: proc(c: Chunk) -> (res: sRGB, ok: bool) {
-	ok = true
-
-	if c.header.type != .sRGB || len(c.data) != 1 {
+srgb :: proc(c: image.PNG_Chunk) -> (res: sRGB, ok: bool) {
+	if c.header.type != .sRGB || len(c.data) != size_of(sRGB_Rendering_Intent) {
 		return {}, false
 	}
 
@@ -238,10 +238,10 @@ srgb :: proc(c: Chunk) -> (res: sRGB, ok: bool) {
 	if res.intent > max(sRGB_Rendering_Intent) {
 		ok = false; return
 	}
-	return
+	return res, true
 }
 
-plte :: proc(c: Chunk) -> (res: PLTE, ok: bool) {
+plte :: proc(c: image.PNG_Chunk) -> (res: PLTE, ok: bool) {
 	if c.header.type != .PLTE {
 		return {}, false
 	}
@@ -255,7 +255,7 @@ plte :: proc(c: Chunk) -> (res: PLTE, ok: bool) {
 	return
 }
 
-splt :: proc(c: Chunk) -> (res: sPLT, ok: bool) {
+splt :: proc(c: image.PNG_Chunk) -> (res: sPLT, ok: bool) {
 	if c.header.type != .sPLT {
 		return {}, false
 	}
@@ -306,7 +306,7 @@ splt_destroy :: proc(s: sPLT) {
 	delete(s.name)
 }
 
-sbit :: proc(c: Chunk) -> (res: [4]u8, ok: bool) {
+sbit :: proc(c: image.PNG_Chunk) -> (res: [4]u8, ok: bool) {
 	/*
 		Returns [4]u8 with the significant bits in each channel.
 		A channel will contain zero if not applicable to the PNG color type.
@@ -324,7 +324,7 @@ sbit :: proc(c: Chunk) -> (res: [4]u8, ok: bool) {
 
 }
 
-hist :: proc(c: Chunk) -> (res: hIST, ok: bool) {
+hist :: proc(c: image.PNG_Chunk) -> (res: hIST, ok: bool) {
 	if c.header.type != .hIST {
 		return {}, false
 	}
@@ -346,7 +346,7 @@ hist :: proc(c: Chunk) -> (res: hIST, ok: bool) {
 	return
 }
 
-chrm :: proc(c: Chunk) -> (res: cHRM, ok: bool) {
+chrm :: proc(c: image.PNG_Chunk) -> (res: cHRM, ok: bool) {
 	ok = true
 	if c.header.length != size_of(cHRM_Raw) {
 		return {}, false
@@ -364,7 +364,7 @@ chrm :: proc(c: Chunk) -> (res: cHRM, ok: bool) {
 	return
 }
 
-exif :: proc(c: Chunk) -> (res: Exif, ok: bool) {
+exif :: proc(c: image.PNG_Chunk) -> (res: Exif, ok: bool) {
 
 	ok = true
 

core/image/png/png.odin

@@ -1,13 +1,12 @@
-package png
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
-	Made available under Odin's BSD-2 license.
+	Made available under Odin's BSD-3 license.
 
 	List of contributors:
 		Jeroen van Rijn: Initial implementation.
 		Ginger Bill:     Cosmetic changes.
 */
+package png
 
 import "core:compress"
 import "core:compress/zlib"
@@ -21,11 +20,29 @@ import "core:io"
 import "core:mem"
 import "core:intrinsics"
 
-Error     :: compress.Error
-E_General :: compress.General_Error
-E_PNG     :: image.Error
-E_Deflate :: compress.Deflate_Error
+/*
+	67_108_864 pixels max by default.
+	Maximum allowed dimensions are capped at 65535 * 65535.
+*/
+MAX_DIMENSIONS    :: min(#config(PNG_MAX_DIMENSIONS, 8192 * 8192), 65535 * 65535)
 
+/*
+	Limit chunk sizes.
+		By default: IDAT = 8k x 8k x 16-bits + 8k filter bytes.
+*/
+_MAX_IDAT_DEFAULT :: ( 8192 /* Width */ *  8192 /* Height */ * 2 /* 16-bit */) +  8192 /* Filter bytes */
+_MAX_IDAT         :: (65535 /* Width */ * 65535 /* Height */ * 2 /* 16-bit */) + 65535 /* Filter bytes */
+
+MAX_IDAT_SIZE     :: min(#config(PNG_MAX_IDAT_SIZE, _MAX_IDAT_DEFAULT), _MAX_IDAT)
+
+/*
+	For chunks other than IDAT with a variable size like `zTXT` and `eXIf`,
+	limit their size to 16 MiB each by default. Max of 256 MiB each.
+*/
+MAX_CHUNK_SIZE    :: min(#config(PNG_MAX_CHUNK_SIZE, 16_777_216), 268_435_456)
+
+
+Error     :: image.Error
 Image     :: image.Image
 Options   :: image.Options
 
@@ -34,95 +51,6 @@ Signature :: enum u64be {
 	PNG = 0x89 << 56 | 'P' << 48 | 'N' << 40 | 'G' << 32 | '\r' << 24 | '\n' << 16 | 0x1a << 8 | '\n',
 }
 
-Info :: struct {
-	header: IHDR,
-	chunks: [dynamic]Chunk,
-}
-
-Chunk_Header :: struct #packed {
-	length: u32be,
-	type:   Chunk_Type,
-}
-
-Chunk :: struct #packed {
-	header: Chunk_Header,
-	data:   []byte,
-	crc:    u32be,
-}
-
-Chunk_Type :: enum u32be {
-	// IHDR must come first in a file
-	IHDR = 'I' << 24 | 'H' << 16 | 'D' << 8 | 'R',
-	// PLTE must precede the first IDAT chunk
-	PLTE = 'P' << 24 | 'L' << 16 | 'T' << 8 | 'E',
-	bKGD = 'b' << 24 | 'K' << 16 | 'G' << 8 | 'D',
-	tRNS = 't' << 24 | 'R' << 16 | 'N' << 8 | 'S',
-	IDAT = 'I' << 24 | 'D' << 16 | 'A' << 8 | 'T',
-
-	iTXt = 'i' << 24 | 'T' << 16 | 'X' << 8 | 't',
-	tEXt = 't' << 24 | 'E' << 16 | 'X' << 8 | 't',
-	zTXt = 'z' << 24 | 'T' << 16 | 'X' << 8 | 't',
-
-	iCCP = 'i' << 24 | 'C' << 16 | 'C' << 8 | 'P',
-	pHYs = 'p' << 24 | 'H' << 16 | 'Y' << 8 | 's',
-	gAMA = 'g' << 24 | 'A' << 16 | 'M' << 8 | 'A',
-	tIME = 't' << 24 | 'I' << 16 | 'M' << 8 | 'E',
-
-	sPLT = 's' << 24 | 'P' << 16 | 'L' << 8 | 'T',
-	sRGB = 's' << 24 | 'R' << 16 | 'G' << 8 | 'B',
-	hIST = 'h' << 24 | 'I' << 16 | 'S' << 8 | 'T',
-	cHRM = 'c' << 24 | 'H' << 16 | 'R' << 8 | 'M',
-	sBIT = 's' << 24 | 'B' << 16 | 'I' << 8 | 'T',
-
-	/*
-		eXIf tags are not part of the core spec, but have been ratified
-		in v1.5.0 of the PNG Ext register.
-
-		We will provide unprocessed chunks to the caller if `.return_metadata` is set.
-		Applications are free to implement an Exif decoder.
-	*/
-	eXIf = 'e' << 24 | 'X' << 16 | 'I' << 8 | 'f',
-
-	// PNG files must end with IEND
-	IEND = 'I' << 24 | 'E' << 16 | 'N' << 8 | 'D',
-
-	/*
-		XCode sometimes produces "PNG" files that don't adhere to the PNG spec.
-		We recognize them only in order to avoid doing further work on them.
-
-		Some tools like PNG Defry may be able to repair them, but we're not
-		going to reward Apple for producing proprietary broken files purporting
-		to be PNGs by supporting them.
-
-	*/
-	iDOT = 'i' << 24 | 'D' << 16 | 'O' << 8 | 'T',
-	CbGI = 'C' << 24 | 'b' << 16 | 'H' << 8 | 'I',
-}
-
-IHDR :: struct #packed {
-	width: u32be,
-	height: u32be,
-	bit_depth: u8,
-	color_type: Color_Type,
-	compression_method: u8,
-	filter_method: u8,
-	interlace_method: Interlace_Method,
-}
-IHDR_SIZE :: size_of(IHDR)
-#assert (IHDR_SIZE == 13)
-
-Color_Value :: enum u8 {
-	Paletted = 0, // 1 << 0 = 1
-	Color    = 1, // 1 << 1 = 2
-	Alpha    = 2, // 1 << 2 = 4
-}
-Color_Type :: distinct bit_set[Color_Value; u8]
-
-Interlace_Method :: enum u8 {
-	None  = 0,
-	Adam7 = 1,
-}
-
 Row_Filter :: enum u8 {
 	None    = 0,
 	Sub     = 1,
@@ -135,22 +63,22 @@ PLTE_Entry    :: [3]u8
 
 PLTE :: struct #packed {
 	entries: [256]PLTE_Entry,
-	used: u16,
+	used:    u16,
 }
 
 hIST :: struct #packed {
 	entries: [256]u16,
-	used: u16,
+	used:    u16,
 }
 
 sPLT :: struct #packed {
-	name: string,
-	depth: u8,
+	name:    string,
+	depth:   u8,
 	entries: union {
 		[][4]u8,
 		[][4]u16,
 	},
-	used: u16,
+	used:    u16,
 }
 
 // Other chunks
@@ -223,14 +151,14 @@ Exif :: struct {
 }
 
 iCCP :: struct {
-	name: string,
+	name:    string,
 	profile: []u8,
 }
 
 sRGB_Rendering_Intent :: enum u8 {
-	Perceptual = 0,
+	Perceptual            = 0,
 	Relative_colorimetric = 1,
-	Saturation = 2,
+	Saturation            = 2,
 	Absolute_colorimetric = 3,
 }
 
@@ -245,16 +173,30 @@ ADAM7_Y_SPACING := []int{ 8,8,8,4,4,2,2 }
 
 // Implementation starts here
 
-read_chunk :: proc(ctx: ^$C) -> (chunk: Chunk, err: Error) {
-	ch, e := compress.read_data(ctx, Chunk_Header)
+read_chunk :: proc(ctx: ^$C) -> (chunk: image.PNG_Chunk, err: Error) {
+	ch, e := compress.read_data(ctx, image.PNG_Chunk_Header)
 	if e != .None {
-		return {}, E_General.Stream_Too_Short
+		return {}, compress.General_Error.Stream_Too_Short
 	}
 	chunk.header = ch
 
+	/*
+		Sanity check chunk size
+	*/
+	#partial switch ch.type {
+	case .IDAT:
+		if ch.length > MAX_IDAT_SIZE {
+			return {}, image.PNG_Error.IDAT_Size_Too_Large
+		}
+	case:
+		if ch.length > MAX_CHUNK_SIZE {
+			return {}, image.PNG_Error.Invalid_Chunk_Length
+		}
+	}
+
 	chunk.data, e = compress.read_slice(ctx, int(ch.length))
 	if e != .None {
-		return {}, E_General.Stream_Too_Short
+		return {}, compress.General_Error.Stream_Too_Short
 	}
 
 	// Compute CRC over chunk type + data
@@ -264,39 +206,68 @@ read_chunk :: proc(ctx: ^$C) -> (chunk: Chunk, err: Error) {
 
 	crc, e3 := compress.read_data(ctx, u32be)
 	if e3 != .None {
-		return {}, E_General.Stream_Too_Short
+		return {}, compress.General_Error.Stream_Too_Short
 	}
 	chunk.crc = crc
 
 	if chunk.crc != u32be(computed_crc) {
-		return {}, E_General.Checksum_Failed
+		return {}, compress.General_Error.Checksum_Failed
 	}
 	return chunk, nil
 }
 
-read_header :: proc(ctx: ^$C) -> (IHDR, Error) {
+copy_chunk :: proc(src: image.PNG_Chunk, allocator := context.allocator) -> (dest: image.PNG_Chunk, err: Error) {
+	if int(src.header.length) != len(src.data) {
+		return {}, .Invalid_Chunk_Length
+	}
+
+	dest.header = src.header
+	dest.crc    = src.crc
+	dest.data   = make([]u8, dest.header.length, allocator) or_return
+
+	copy(dest.data[:], src.data[:])
+	return
+}
+
+append_chunk :: proc(list: ^[dynamic]image.PNG_Chunk, src: image.PNG_Chunk, allocator := context.allocator) -> (err: Error) {
+	if int(src.header.length) != len(src.data) {
+		return .Invalid_Chunk_Length
+	}
+
+	c := copy_chunk(src, allocator) or_return
+	length := len(list)
+	append(list, c)
+	if len(list) != length + 1 {
+		// Resize during append failed.
+		return mem.Allocator_Error.Out_Of_Memory
+	}
+
+	return
+}
+
+read_header :: proc(ctx: ^$C) -> (image.PNG_IHDR, Error) {
 	c, e := read_chunk(ctx)
 	if e != nil {
 		return {}, e
 	}
 
-	header := (^IHDR)(raw_data(c.data))^
+	header := (^image.PNG_IHDR)(raw_data(c.data))^
 	// Validate IHDR
 	using header
-	if width == 0 || height == 0 {
-		return {}, E_PNG.Invalid_Image_Dimensions
+	if width == 0 || height == 0 || u128(width) * u128(height) > MAX_DIMENSIONS {
+		return {}, .Invalid_Image_Dimensions
 	}
 
 	if compression_method != 0 {
-		return {}, E_General.Unknown_Compression_Method
+		return {}, compress.General_Error.Unknown_Compression_Method
 	}
 
 	if filter_method != 0 {
-		return {}, E_PNG.Unknown_Filter_Method
+		return {}, .Unknown_Filter_Method
 	}
 
 	if interlace_method != .None && interlace_method != .Adam7 {
-		return {}, E_PNG.Unknown_Interlace_Method
+		return {}, .Unknown_Interlace_Method
 
 	}
 
@@ -314,7 +285,7 @@ read_header :: proc(ctx: ^$C) -> (IHDR, Error) {
 			}
 		}
 		if !allowed {
-			return {}, E_PNG.Invalid_Color_Bit_Depth_Combo
+			return {}, .Invalid_Color_Bit_Depth_Combo
 		}
 	case 2, 4, 6:
 		/*
@@ -322,7 +293,7 @@ read_header :: proc(ctx: ^$C) -> (IHDR, Error) {
 			Allowed bit depths: 8 and 16
 		*/
 		if bit_depth != 8 && bit_depth != 16 {
-			return {}, E_PNG.Invalid_Color_Bit_Depth_Combo
+			return {}, .Invalid_Color_Bit_Depth_Combo
 		}
 	case 3:
 		/*
@@ -337,17 +308,17 @@ read_header :: proc(ctx: ^$C) -> (IHDR, Error) {
 			}
 		}
 		if !allowed {
-			return {}, E_PNG.Invalid_Color_Bit_Depth_Combo
+			return {}, .Invalid_Color_Bit_Depth_Combo
 		}
 
 	case:
-		return {}, E_PNG.Unknown_Color_Type
+		return {}, .Unknown_Color_Type
 	}
 
 	return header, nil
 }
 
-chunk_type_to_name :: proc(type: ^Chunk_Type) -> string {
+chunk_type_to_name :: proc(type: ^image.PNG_Chunk_Type) -> string {
 	t := transmute(^u8)type
 	return strings.string_from_ptr(t, 4)
 }
@@ -377,7 +348,7 @@ load_from_file :: proc(filename: string, options := Options{}, allocator := cont
 		return load_from_slice(data, options)
 	} else {
 		img = new(Image)
-		return img, E_General.File_Not_Found
+		return img, compress.General_Error.File_Not_Found
 	}
 }
 
@@ -391,7 +362,7 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 	}
 
 	if .alpha_drop_if_present in options && .alpha_add_if_missing in options {
-		return {}, E_General.Incompatible_Options
+		return {}, compress.General_Error.Incompatible_Options
 	}
 
 	if .do_not_expand_channels in options {
@@ -402,27 +373,25 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 		img = new(Image)
 	}
 
-	info := new(Info)
-	img.metadata_ptr  = info
-	img.metadata_type = typeid_of(Info)
+	info := new(image.PNG_Info)
+	img.metadata = info
 
 	signature, io_error := compress.read_data(ctx, Signature)
 	if io_error != .None || signature != .PNG {
-		return img, E_PNG.Invalid_PNG_Signature
+		return img, .Invalid_PNG_Signature
 	}
 
 	idat: []u8
 	idat_b: bytes.Buffer
-	idat_length := u32be(0)
 	defer bytes.buffer_destroy(&idat_b)
 
-	c:		Chunk
-	ch:     Chunk_Header
+	idat_length := u64(0)
+
+	c:		image.PNG_Chunk
+	ch:     image.PNG_Chunk_Header
 	e:      io.Error
 
-	header:	IHDR
-
-	info.chunks.allocator = context.temp_allocator
+	header:	image.PNG_IHDR
 
 	// State to ensure correct chunk ordering.
 	seen_ihdr := false; first := true
@@ -433,7 +402,7 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 	seen_iend := false
 
 	_plte := PLTE{}
-	trns := Chunk{}
+	trns := image.PNG_Chunk{}
 
 	final_image_channels := 0
 
@@ -443,16 +412,16 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 		// Peek at next chunk's length and type.
 		// TODO: Some streams may not provide seek/read_at
 
-		ch, e = compress.peek_data(ctx, Chunk_Header)
+		ch, e = compress.peek_data(ctx, image.PNG_Chunk_Header)
 		if e != .None {
-			return img, E_General.Stream_Too_Short
+			return img, compress.General_Error.Stream_Too_Short
 		}
 		// name := chunk_type_to_name(&ch.type); // Only used for debug prints during development.
 
 		#partial switch ch.type {
 		case .IHDR:
 			if seen_ihdr || !first {
-				return {}, E_PNG.IHDR_Not_First_Chunk
+				return {}, .IHDR_Not_First_Chunk
 			}
 			seen_ihdr = true
 
@@ -481,14 +450,14 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 			}
 
 			if img.channels == 0 || img.depth == 0 {
-				return {}, E_PNG.IHDR_Corrupt
+				return {}, .IHDR_Corrupt
 			}
 
 			img.width  = int(header.width)
 			img.height = int(header.height)
 
 			using header
-			h := IHDR{
+			h := image.PNG_IHDR{
 				width              = width,
 				height             = height,
 				bit_depth          = bit_depth,
@@ -498,28 +467,30 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 				interlace_method   = interlace_method,
 			}
 			info.header = h
+
 		case .PLTE:
 			seen_plte = true
 			// PLTE must appear before IDAT and can't appear for color types 0, 4.
 			ct := transmute(u8)info.header.color_type
 			if seen_idat || ct == 0 || ct == 4 {
-				return img, E_PNG.PLTE_Encountered_Unexpectedly
+				return img, .PLTE_Encountered_Unexpectedly
 			}
 
 			c = read_chunk(ctx) or_return
 
 			if c.header.length % 3 != 0 || c.header.length > 768 {
-				return img, E_PNG.PLTE_Invalid_Length
+				return img, .PLTE_Invalid_Length
 			}
 			plte_ok: bool
 			_plte, plte_ok = plte(c)
 			if !plte_ok {
-				return img, E_PNG.PLTE_Invalid_Length
+				return img, .PLTE_Invalid_Length
 			}
 
 			if .return_metadata in options {
-				append(&info.chunks, c)
+				append_chunk(&info.chunks, c) or_return
 			}
+
 		case .IDAT:
 			// If we only want image metadata and don't want the pixel data, we can early out.
 			if .return_metadata not_in options && .do_not_decompress_image in options {
@@ -528,11 +499,11 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 			}
 			// There must be at least 1 IDAT, contiguous if more.
 			if seen_idat {
-				return img, E_PNG.IDAT_Must_Be_Contiguous
+				return img, .IDAT_Must_Be_Contiguous
 			}
 
 			if idat_length > 0 {
-				return img, E_PNG.IDAT_Must_Be_Contiguous
+				return img, .IDAT_Must_Be_Contiguous
 			}
 
 			next := ch.type
@@ -540,22 +511,29 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 				c = read_chunk(ctx) or_return
 
 				bytes.buffer_write(&idat_b, c.data)
-				idat_length += c.header.length
+				idat_length += u64(c.header.length)
 
-				ch, e = compress.peek_data(ctx, Chunk_Header)
+				if idat_length > MAX_IDAT_SIZE {
+					return {}, image.PNG_Error.IDAT_Size_Too_Large
+				}
+
+				ch, e = compress.peek_data(ctx, image.PNG_Chunk_Header)
 				if e != .None {
-					return img, E_General.Stream_Too_Short
+					return img, compress.General_Error.Stream_Too_Short
 				}
 				next = ch.type
 			}
+
 			idat = bytes.buffer_to_bytes(&idat_b)
 			if int(idat_length) != len(idat) {
-				return {}, E_PNG.IDAT_Corrupt
+				return {}, .IDAT_Corrupt
 			}
 			seen_idat = true
+
 		case .IEND:
 			c = read_chunk(ctx) or_return
 			seen_iend = true
+
 		case .bKGD:
 
 			// TODO: Make sure that 16-bit bKGD + tRNS chunks return u16 instead of u16be
@@ -563,14 +541,14 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 			c = read_chunk(ctx) or_return
 			seen_bkgd = true
 			if .return_metadata in options {
-				append(&info.chunks, c)
+				append_chunk(&info.chunks, c) or_return
 			}
 
 			ct := transmute(u8)info.header.color_type
 			switch ct {
 				case 3: // Indexed color
 					if c.header.length != 1 {
-						return {}, E_PNG.BKGD_Invalid_Length
+						return {}, .BKGD_Invalid_Length
 					}
 					col := _plte.entries[c.data[0]]
 					img.background = [3]u16{
@@ -580,26 +558,27 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 					}
 				case 0, 4: // Grayscale, with and without Alpha
 					if c.header.length != 2 {
-						return {}, E_PNG.BKGD_Invalid_Length
+						return {}, .BKGD_Invalid_Length
 					}
 					col := u16(mem.slice_data_cast([]u16be, c.data[:])[0])
 					img.background = [3]u16{col, col, col}
 				case 2, 6: // Color, with and without Alpha
 					if c.header.length != 6 {
-						return {}, E_PNG.BKGD_Invalid_Length
+						return {}, .BKGD_Invalid_Length
 					}
 					col := mem.slice_data_cast([]u16be, c.data[:])
 					img.background = [3]u16{u16(col[0]), u16(col[1]), u16(col[2])}
 			}
+
 		case .tRNS:
 			c = read_chunk(ctx) or_return
 
 			if .Alpha in info.header.color_type {
-				return img, E_PNG.TRNS_Encountered_Unexpectedly
+				return img, .TRNS_Encountered_Unexpectedly
 			}
 
 			if .return_metadata in options {
-				append(&info.chunks, c)
+				append_chunk(&info.chunks, c) or_return
 			}
 
 			/*
@@ -622,20 +601,20 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 				}
 			}
 			trns = c
+
 		case .iDOT, .CbGI:
 			/*
 				iPhone PNG bastardization that doesn't adhere to spec with broken IDAT chunk.
 				We're not going to add support for it. If you have the misfortunte of coming
-				across one of these files, use a utility to defry it.s
+				across one of these files, use a utility to defry it.
 			*/
-			return img, E_PNG.PNG_Does_Not_Adhere_to_Spec
+			return img, .Image_Does_Not_Adhere_to_Spec
+
 		case:
 			// Unhandled type
 			c = read_chunk(ctx) or_return
-
 			if .return_metadata in options {
-				// NOTE: Chunk cata is currently allocated on the temp allocator.
-				append(&info.chunks, c)
+				append_chunk(&info.chunks, c) or_return
 			}
 
 			first = false
@@ -648,7 +627,7 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 	}
 
 	if !seen_idat {
-		return img, E_PNG.IDAT_Missing
+		return img, .IDAT_Missing
 	}
 
 	/*
@@ -685,7 +664,7 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 
 	buf_len := len(buf.buf)
 	if expected_size != buf_len {
-		return {}, E_PNG.IDAT_Corrupt
+		return {}, .IDAT_Corrupt
 	}
 
 	/*
@@ -752,7 +731,9 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 		// We need to create a new image buffer
 		dest_raw_size := compute_buffer_size(int(header.width), int(header.height), out_image_channels, 8)
 		t := bytes.Buffer{}
-		resize(&t.buf, dest_raw_size)
+		if !resize(&t.buf, dest_raw_size) {
+			return {}, mem.Allocator_Error.Out_Of_Memory
+		}
 
 		i := 0; j := 0
 
@@ -831,7 +812,9 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 		// We need to create a new image buffer
 		dest_raw_size := compute_buffer_size(int(header.width), int(header.height), out_image_channels, 16)
 		t := bytes.Buffer{}
-		resize(&t.buf, dest_raw_size)
+		if !resize(&t.buf, dest_raw_size) {
+			return {}, mem.Allocator_Error.Out_Of_Memory
+		}
 
 		p16 := mem.slice_data_cast([]u16, temp.buf[:])
 		o16 := mem.slice_data_cast([]u16, t.buf[:])
@@ -1028,7 +1011,9 @@ load_from_context :: proc(ctx: ^$C, options := Options{}, allocator := context.a
 		// We need to create a new image buffer
 		dest_raw_size := compute_buffer_size(int(header.width), int(header.height), out_image_channels, 8)
 		t := bytes.Buffer{}
-		resize(&t.buf, dest_raw_size)
+		if !resize(&t.buf, dest_raw_size) {
+			return {}, mem.Allocator_Error.Out_Of_Memory
+		}
 
 		p := mem.slice_data_cast([]u8, temp.buf[:])
 		o := mem.slice_data_cast([]u8, t.buf[:])
@@ -1524,7 +1509,7 @@ defilter_16 :: proc(params: ^Filter_Params) -> (ok: bool) {
 	return
 }
 
-defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^IHDR, options: Options) -> (err: compress.Error) {
+defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^image.PNG_IHDR, options: Options) -> (err: Error) {
 	input    := bytes.buffer_to_bytes(filter_bytes)
 	width    := int(header.width)
 	height   := int(header.height)
@@ -1535,7 +1520,9 @@ defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^IHDR, option
 	bytes_per_channel := depth == 16 ? 2 : 1
 
 	num_bytes := compute_buffer_size(width, height, channels, depth == 16 ? 16 : 8)
-	resize(&img.pixels.buf, num_bytes)
+	if !resize(&img.pixels.buf, num_bytes) {
+		return mem.Allocator_Error.Out_Of_Memory
+	}
 
 	filter_ok: bool
 
@@ -1560,7 +1547,7 @@ defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^IHDR, option
 		}
 		if !filter_ok {
 			// Caller will destroy buffer for us.
-			return E_PNG.Unknown_Filter_Method
+			return .Unknown_Filter_Method
 		}
 	} else {
 		/*
@@ -1575,7 +1562,9 @@ defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^IHDR, option
 			if x > 0 && y > 0 {
 				temp: bytes.Buffer
 				temp_len := compute_buffer_size(x, y, channels, depth == 16 ? 16 : 8)
-				resize(&temp.buf, temp_len)
+				if !resize(&temp.buf, temp_len) {
+					return mem.Allocator_Error.Out_Of_Memory
+				}
 
 				params := Filter_Params{
 					src      = input,
@@ -1598,7 +1587,7 @@ defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^IHDR, option
 
 				if !filter_ok {
 					// Caller will destroy buffer for us.
-					return E_PNG.Unknown_Filter_Method
+					return .Unknown_Filter_Method
 				}
 
 				t := temp.buf[:]

core/intrinsics/intrinsics.odin

@@ -39,6 +39,7 @@ sqrt :: proc(x: $T) -> T where type_is_float(T) ---
 mem_copy                 :: proc(dst, src: rawptr, len: int) ---
 mem_copy_non_overlapping :: proc(dst, src: rawptr, len: int) ---
 mem_zero                 :: proc(ptr: rawptr, len: int) ---
+mem_zero_volatile        :: proc(ptr: rawptr, len: int) ---
 
 
 fixed_point_mul     :: proc(lhs, rhs: $T, #const scale: uint) -> T where type_is_integer(T) ---

core/io/io.odin

@@ -443,7 +443,7 @@ read_at_least :: proc(r: Reader, buf: []byte, min: int) -> (n: int, err: Error)
 	for n < min && err == nil {
 		nn: int
 		nn, err = read(r, buf[n:])
-		n += n
+		n += nn
 	}
 
 	if n >= min {

core/io/util.odin

@@ -116,14 +116,14 @@ write_encoded_rune :: proc(w: Writer, r: rune, write_quote := true, n_written: ^
 		write_byte(w, '\'', &n) or_return
 	}
 	switch r {
-	case '\a': write_string(w, `\a"`, &n) or_return
-	case '\b': write_string(w, `\b"`, &n) or_return
-	case '\e': write_string(w, `\e"`, &n) or_return
-	case '\f': write_string(w, `\f"`, &n) or_return
-	case '\n': write_string(w, `\n"`, &n) or_return
-	case '\r': write_string(w, `\r"`, &n) or_return
-	case '\t': write_string(w, `\t"`, &n) or_return
-	case '\v': write_string(w, `\v"`, &n) or_return
+	case '\a': write_string(w, `\a`, &n) or_return
+	case '\b': write_string(w, `\b`, &n) or_return
+	case '\e': write_string(w, `\e`, &n) or_return
+	case '\f': write_string(w, `\f`, &n) or_return
+	case '\n': write_string(w, `\n`, &n) or_return
+	case '\r': write_string(w, `\r`, &n) or_return
+	case '\t': write_string(w, `\t`, &n) or_return
+	case '\v': write_string(w, `\v`, &n) or_return
 	case:
 		if r < 32 {
 			write_string(w, `\x`, &n) or_return
@@ -132,8 +132,8 @@ write_encoded_rune :: proc(w: Writer, r: rune, write_quote := true, n_written: ^
 			s := strconv.append_bits(buf[:], u64(r), 16, true, 64, strconv.digits, nil)
 			switch len(s) {
 			case 0: write_string(w, "00", &n) or_return
-			case 1: write_byte(w, '0', &n)             or_return
-			case 2: write_string(w, s, &n)    or_return
+			case 1: write_byte(w, '0',    &n) or_return
+			case 2: write_string(w, s,    &n) or_return
 			}
 		} else {
 			write_rune(w, r, &n) or_return

core/log/multi_logger.odin

@@ -13,6 +13,8 @@ create_multi_logger :: proc(logs: ..Logger) -> Logger {
 }
 
 destroy_multi_logger :: proc(log : ^Logger) {
+	data := (^Multi_Logger_Data)(log.data)
+	delete(data.loggers)
 	free(log.data)
 	log^ = nil_logger()
 }

core/math/big/api.odin

@@ -32,6 +32,9 @@ add :: proc {
 		int_add_digit :: proc(dest, a: ^Int, digit: DIGIT, allocator := context.allocator) -> (err: Error)
 	*/
 	int_add_digit,
+	rat_add_rat,
+	rat_add_int,
+	int_add_rat,
 }
 
 /*
@@ -46,6 +49,9 @@ sub :: proc {
 		int_sub_digit :: proc(dest, a: ^Int, digit: DIGIT) -> (err: Error)
 	*/
 	int_sub_digit,
+	rat_sub_rat,
+	rat_sub_int,
+	int_sub_rat,
 }
 
 /*
@@ -67,6 +73,10 @@ is_zero :: proc {
 		int_is_zero :: proc(a: ^Int) -> bool
 	*/
 	int_is_zero,
+	/*
+		rat_is_zero :: proc(a: ^Rat) -> bool
+	*/
+	rat_is_zero,
 }
 
 is_positive :: proc {
@@ -74,6 +84,7 @@ is_positive :: proc {
 		int_is_positive :: proc(a: ^Int) -> bool
 	*/
 	int_is_positive,
+	rat_is_positive,
 }
 is_pos :: is_positive
 
@@ -82,6 +93,7 @@ is_negative :: proc {
 		int_is_negative :: proc(a: ^Int) -> bool
 	*/
 	int_is_negative,
+	rat_is_negative,
 }
 is_neg :: is_negative
 
@@ -90,6 +102,7 @@ is_even :: proc {
 		int_is_even :: proc(a: ^Int) -> bool
 	*/
 	int_is_even,
+	rat_is_even,
 }
 
 is_odd :: proc {
@@ -97,6 +110,7 @@ is_odd :: proc {
 		int_is_odd :: proc(a: ^Int) -> bool
 	*/
 	int_is_odd,
+	rat_is_odd,
 }
 
 is_power_of_two :: proc {

core/math/big/helpers.odin

@@ -44,7 +44,20 @@ int_set_from_integer :: proc(dest: ^Int, src: $T, minimize := false, allocator :
 	return #force_inline internal_int_set_from_integer(dest, src, minimize)
 }
 
-set :: proc { int_set_from_integer, int_copy, int_atoi, }
+set :: proc { 
+	int_set_from_integer, 
+	int_copy, 
+	int_atoi, 
+
+	rat_set_f64, 
+	rat_set_f32, 
+	rat_set_f16, 
+	rat_set_u64, 
+	rat_set_i64,
+	rat_set_int, 
+	rat_set_digit, 
+	rat_set_rat, 
+}
 
 /*
 	Copy one `Int` to another.
@@ -66,7 +79,10 @@ int_copy :: proc(dest, src: ^Int, minimize := false, allocator := context.alloca
 
 	return #force_inline internal_int_copy(dest, src, minimize)
 }
-copy :: proc { int_copy, }
+copy :: proc { 
+	int_copy, 
+	rat_copy,
+}
 
 /*
 	In normal code, you can also write `a, b = b, a`.
@@ -77,7 +93,7 @@ int_swap :: proc(a, b: ^Int) {
 	assert_if_nil(a, b)
 	#force_inline internal_swap(a, b)
 }
-swap :: proc { int_swap, }
+swap :: proc { int_swap, rat_swap }
 
 /*
 	Set `dest` to |`src`|.
@@ -98,7 +114,7 @@ int_abs :: proc(dest, src: ^Int, allocator := context.allocator) -> (err: Error)
 platform_abs :: proc(n: $T) -> T where intrinsics.type_is_integer(T) {
 	return n if n >= 0 else -n
 }
-abs :: proc{ int_abs, platform_abs, }
+abs :: proc{ int_abs, platform_abs, rat_abs }
 
 /*
 	Set `dest` to `-src`.
@@ -115,7 +131,7 @@ int_neg :: proc(dest, src: ^Int, allocator := context.allocator) -> (err: Error)
 
 	return #force_inline internal_int_neg(dest, src)
 }
-neg :: proc { int_neg, }
+neg :: proc { int_neg, rat_neg }
 
 /*
 	Helpers to extract values from the `Int`.
@@ -788,7 +804,10 @@ destroy_constants :: proc() {
 }
 
 
-assert_if_nil :: proc{assert_if_nil_int}
+assert_if_nil :: proc{
+	assert_if_nil_int,
+	assert_if_nil_rat,
+}
 
 assert_if_nil_int :: #force_inline proc(integers: ..^Int, loc := #caller_location) {
 	for i in integers {
@@ -796,3 +815,8 @@ assert_if_nil_int :: #force_inline proc(integers: ..^Int, loc := #caller_locatio
 	}
 }
 
+assert_if_nil_rat :: #force_inline proc(rationals: ..^Rat, loc := #caller_location) {
+	for r in rationals {
+		assert(r != nil, "(nil)", loc)
+	}
+}

core/math/big/internal.odin

@@ -1042,7 +1042,10 @@ internal_is_initialized :: proc { internal_int_is_initialized, }
 internal_int_is_zero :: #force_inline proc(a: ^Int) -> (zero: bool) {
 	return a.used == 0
 }
-internal_is_zero :: proc { internal_int_is_zero }
+internal_is_zero :: proc { 
+	internal_rat_is_zero,
+	internal_int_is_zero,
+}
 
 /*
 	This procedure will return `true` if the `Int` is positive, `false` if not.
@@ -1865,7 +1868,10 @@ internal_int_destroy :: proc(integers: ..^Int) {
 		a = &Int{}
 	}
 }
-internal_destroy :: proc{ internal_int_destroy, }
+internal_destroy :: proc{ 
+	internal_int_destroy, 
+	internal_rat_destroy, 
+}
 
 /*
 	Helpers to set an `Int` to a specific value.
@@ -1950,13 +1956,14 @@ internal_copy :: proc { internal_int_copy, }
 	This helper swaps completely.
 */
 internal_int_swap :: #force_inline proc(a, b: ^Int) {
-	a := a; b := b
-
 	a.used,  b.used  = b.used,  a.used
 	a.sign,  b.sign  = b.sign,  a.sign
 	a.digit, b.digit = b.digit, a.digit
 }
-internal_swap :: proc { internal_int_swap, }
+internal_swap :: proc { 
+	internal_int_swap, 
+	internal_rat_swap,
+}
 
 /*
 	Set `dest` to |`src`|.

core/math/big/public.odin

@@ -152,9 +152,18 @@ int_mul :: proc(dest, src, multiplier: ^Int, allocator := context.allocator) ->
 	return #force_inline internal_int_mul(dest, src, multiplier)
 }
 
-mul :: proc { int_mul, int_mul_digit, }
+mul :: proc { 
+	int_mul, 
+	int_mul_digit, 
+	rat_mul_rat,
+	rat_mul_int,
+	int_mul_rat,
+}
+
+int_sqr :: proc(dest, src: ^Int) -> (err: Error) { return mul(dest, src, src) }
+rat_sqr :: proc(dest, src: ^Rat) -> (err: Error) { return mul(dest, src, src) }
+sqr :: proc { int_sqr, rat_sqr }
 
-sqr :: proc(dest, src: ^Int) -> (err: Error) { return mul(dest, src, src) }
 
 /*
 	divmod.
@@ -200,7 +209,13 @@ int_div_digit :: proc(quotient, numerator: ^Int, denominator: DIGIT, allocator :
 	_ = #force_inline internal_divmod(quotient, numerator, denominator) or_return
 	return
 }
-div :: proc { int_div, int_div_digit, }
+div :: proc { 
+	int_div, 
+	int_div_digit, 
+	rat_div_rat,
+	rat_div_int,
+	int_div_rat,
+}
 
 /*
 	remainder = numerator % denominator.

core/math/big/rat.odin

@@ -0,0 +1,540 @@
+package math_big
+
+import "core:builtin"
+import "core:intrinsics"
+import "core:math"
+
+Rat :: struct {
+	a, b: Int,
+}
+
+rat_set_f64 :: proc(dst: ^Rat, f: f64, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	context.allocator = allocator
+	
+	EXP_MASK :: 1<<11 - 1
+	
+	bits := transmute(u64)f
+	mantissa := bits & (1<<52 - 1)
+	exp := int((bits>>52) & EXP_MASK)
+	
+	int_set_from_integer(&dst.b, 1) or_return
+	
+	switch exp {
+	case EXP_MASK:
+		dst.a.flags += {.Inf}
+		return
+	case 0:
+		exp -= 1022
+	case:
+		mantissa |= 1<<52
+		exp -= 1023
+	}
+	
+	shift := 52 - exp
+	
+	for mantissa&1 == 0 && shift > 0 {
+		mantissa >>= 1
+		shift -= 1
+	}
+	
+	int_set_from_integer(&dst.a, mantissa) or_return
+	dst.a.sign = .Negative if f < 0 else .Zero_or_Positive
+	
+	if shift > 0 {
+		internal_int_shl_digit(&dst.b, shift) or_return
+	} else {
+		internal_int_shl_digit(&dst.a, -shift) or_return
+	}
+	
+	return internal_rat_norm(dst)
+}
+
+rat_set_f32 :: proc(dst: ^Rat, f: f32, allocator := context.allocator) -> (err: Error) {
+	return rat_set_f64(dst, f64(f), allocator)
+}
+rat_set_f16 :: proc(dst: ^Rat, f: f16, allocator := context.allocator) -> (err: Error) {
+	return rat_set_f64(dst, f64(f), allocator)
+}
+
+
+rat_set_frac :: proc{rat_set_frac_digit, rat_set_frac_int}
+
+rat_set_frac_digit :: proc(dst: ^Rat, a, b: DIGIT, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	if b == 0 {
+		return .Division_by_Zero
+	}
+	context.allocator = allocator
+	internal_set(&dst.a, a) or_return
+	internal_set(&dst.b, b) or_return
+	return internal_rat_norm(dst)
+}
+
+rat_set_frac_int :: proc(dst: ^Rat, a, b: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	assert_if_nil(a, b)
+	if internal_is_zero(b) {
+		return .Division_by_Zero
+	}
+	context.allocator = allocator
+	internal_set(&dst.a, a) or_return
+	internal_set(&dst.b, b) or_return
+	return internal_rat_norm(dst)
+}
+
+rat_set_int :: proc(dst: ^Rat, a: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	assert_if_nil(a)
+	context.allocator = allocator
+	internal_set(&dst.a, a) or_return
+	internal_set(&dst.b, 1) or_return
+	return
+}
+
+rat_set_digit :: proc(dst: ^Rat, a: DIGIT, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	context.allocator = allocator
+	internal_set(&dst.a, a) or_return
+	internal_set(&dst.b, 1) or_return
+	return
+}
+
+rat_set_rat :: proc(dst, x: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x)
+	context.allocator = allocator
+	internal_set(&dst.a, &x.a) or_return
+	internal_set(&dst.b, &x.b) or_return
+	return
+}
+
+rat_set_u64 :: proc(dst: ^Rat, x: u64, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	context.allocator = allocator
+	internal_set(&dst.a, x) or_return
+	internal_set(&dst.a, 1) or_return
+	return
+}
+rat_set_i64 :: proc(dst: ^Rat, x: i64, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst)
+	context.allocator = allocator
+	internal_set(&dst.a, x) or_return
+	internal_set(&dst.a, 1) or_return
+	return
+}
+
+rat_copy :: proc(dst, src: ^Rat, minimize := false, allocator := context.allocator) -> (err: Error) {
+	if (dst == src) { return nil }
+	
+	assert_if_nil(dst, src)
+	context.allocator = allocator
+	int_copy(&dst.a, &src.a, minimize, allocator) or_return
+	int_copy(&dst.b, &src.b, minimize, allocator) or_return
+	internal_rat_norm(dst) or_return
+	return nil
+}
+
+internal_rat_destroy :: proc(rationals: ..^Rat) {
+	rationals := rationals
+
+	for z in &rationals {
+		internal_int_destroy(&z.a, &z.b)
+	}
+}
+
+internal_rat_norm :: proc(z: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(z)
+	context.allocator = allocator
+	switch {
+	case internal_is_zero(&z.a):
+		z.a.sign = .Zero_or_Positive
+		fallthrough
+	case internal_is_zero(&z.b):
+		int_set_from_integer(&z.b, 1) or_return
+	case:
+		sign := z.a.sign
+		z.a.sign = .Zero_or_Positive
+		z.b.sign = .Zero_or_Positive
+		
+		f := &Int{}
+		internal_int_gcd(f, &z.a, &z.b) or_return
+		if !internal_int_equals_digit(f, 1) {
+			f.sign = .Zero_or_Positive
+			internal_int_div(&z.a, &z.a, f) or_return
+			internal_int_div(&z.b, &z.b, f) or_return
+		}
+		z.a.sign = sign	
+	}
+	return
+}
+
+rat_swap :: proc(a, b: ^Rat) {
+	assert_if_nil(a, b)
+	#force_inline internal_swap(a, b)
+}
+
+internal_rat_swap :: #force_inline proc(a, b: ^Rat) {
+	internal_int_swap(&a.a, &b.a)
+	internal_int_swap(&a.b, &b.b)
+}
+
+rat_sign :: proc(z: ^Rat) -> Sign {
+	if z == nil {
+		return .Zero_or_Positive
+	}
+	return z.a.sign
+}
+
+rat_is_int :: proc(z: ^Rat) -> bool {
+	assert_if_nil(z)
+	return internal_is_zero(&z.a) || internal_int_equals_digit(&z.b, 1)
+}
+
+rat_is_zero :: proc(z: ^Rat) -> bool {
+	return internal_rat_is_zero(z)
+}
+internal_rat_is_zero :: #force_inline proc(z: ^Rat) -> bool {
+	assert_if_nil(z)
+	return internal_is_zero(&z.a)
+}
+
+internal_int_mul_denom :: proc(dst, x, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x, y)
+	context.allocator = allocator
+	switch {
+	case internal_is_zero(x) && internal_is_zero(y):
+		return internal_set(dst, 1) 
+	case internal_is_zero(x):
+		return internal_set(dst, y)
+	case internal_is_zero(y):
+		return internal_set(dst, x)
+	}
+	return int_mul(dst, x, y)
+}
+
+internal_int_scale_denom :: proc(dst, x, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x, y)
+	if internal_is_zero(y) {
+		return internal_set(dst, x)
+	}
+	int_mul(dst, x, y) or_return
+	dst.sign = x.sign
+	return
+}
+
+
+rat_add_rat :: proc(dst, x, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x, y)
+	context.allocator = allocator
+	
+	a1, a2: Int
+	defer internal_destroy(&a1, &a2)
+	
+	internal_int_scale_denom(&a1, &x.a, &y.b)  or_return
+	internal_int_scale_denom(&a2, &y.a, &x.b)  or_return
+	int_add(&dst.a, &a1, &a2)                  or_return
+	internal_int_mul_denom(&dst.b, &x.b, &y.b) or_return
+	return internal_rat_norm(dst)
+}
+
+rat_sub_rat :: proc(dst, x, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x, y)
+	context.allocator = allocator
+	
+	a1, a2 := &Int{}, &Int{}
+	defer internal_destroy(a1, a2)
+	
+	internal_int_scale_denom(a1, &x.a, &y.b)   or_return
+	internal_int_scale_denom(a2, &y.a, &x.b)   or_return
+	int_sub(&dst.a, a1, a2)                    or_return
+	internal_int_mul_denom(&dst.b, &x.b, &y.b) or_return
+	return internal_rat_norm(dst)
+}
+
+rat_mul_rat :: proc(dst, x, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x, y)
+	context.allocator = allocator
+	
+	if x == y {
+		internal_sqr(&dst.a, &x.a)         or_return
+		if internal_is_zero(&x.b) {
+			internal_set(&dst.b, 1)    or_return
+		} else {
+			internal_sqr(&dst.a, &x.b) or_return
+		}
+		return
+	}
+	
+	int_sub(&dst.a, &x.a, &y.a)                or_return
+	internal_int_mul_denom(&dst.b, &x.b, &y.b) or_return
+	return internal_rat_norm(dst)
+}
+
+rat_div_rat :: proc(dst, x, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	if internal_rat_is_zero(y) {
+		return .Division_by_Zero
+	}
+	context.allocator = allocator
+	
+	a, b := &Int{}, &Int{}
+	defer internal_destroy(a, b)
+	
+	internal_int_scale_denom(a, &x.a, &y.b) or_return
+	internal_int_scale_denom(b, &y.a, &x.b) or_return
+	internal_set(&dst.a, a) or_return
+	internal_set(&dst.b, b) or_return
+	internal_int_abs(&dst.a, &dst.a) 
+	internal_int_abs(&dst.b, &dst.b) 
+	dst.a.sign = .Negative if a.sign != b.sign else .Zero_or_Positive
+	return internal_rat_norm(dst)
+}
+
+
+rat_abs :: proc(dst, x: ^Rat, allocator := context.allocator) -> (err: Error) {
+	rat_set_rat(dst, x, allocator)          or_return
+	internal_abs(&dst.a, &dst.a, allocator) or_return
+	return
+}
+rat_neg :: proc(dst, x: ^Rat, allocator := context.allocator) -> (err: Error) {
+	rat_set_rat(dst, x, allocator)          or_return
+	internal_neg(&dst.a, &dst.a, allocator) or_return
+	return
+}
+
+
+rat_is_positive :: proc(z: ^Rat, allocator := context.allocator) -> (ok: bool, err: Error) {
+	assert_if_nil(z)
+	a := int_is_positive(&z.a, allocator) or_return
+	b := int_is_positive(&z.b, allocator) or_return
+	return !(a ~ b), nil
+}
+rat_is_negative :: proc(z: ^Rat, allocator := context.allocator) -> (ok: bool, err: Error) {
+	assert_if_nil(z)
+	a := int_is_positive(&z.a, allocator) or_return
+	b := int_is_positive(&z.b, allocator) or_return
+	return (a ~ b), nil
+}
+
+rat_is_even :: proc(z: ^Rat, allocator := context.allocator) -> (ok: bool, err: Error) {
+	assert_if_nil(z)
+	if rat_is_int(z) {
+		return int_is_even(&z.a, allocator)
+	}
+	return false, nil
+}
+rat_is_odd :: proc(z: ^Rat, allocator := context.allocator) -> (ok: bool, err: Error) {
+	assert_if_nil(z)
+	if rat_is_int(z) {
+		return int_is_odd(&z.a, allocator)
+	}
+	return false, nil
+}
+
+rat_to_f16 :: proc(z: ^Rat, allocator := context.allocator) -> (f: f16, exact: bool, err: Error) {
+	assert_if_nil(z)
+	return internal_rat_to_float(f16, z, allocator)
+}
+rat_to_f32 :: proc(z: ^Rat, allocator := context.allocator) -> (f: f32, exact: bool, err: Error) {
+	assert_if_nil(z)
+	return internal_rat_to_float(f32, z, allocator)
+}
+rat_to_f64 :: proc(z: ^Rat, allocator := context.allocator) -> (f: f64, exact: bool, err: Error) {
+	assert_if_nil(z)
+	return internal_rat_to_float(f64, z, allocator)
+}
+
+internal_rat_to_float :: proc($T: typeid, z: ^Rat, allocator := context.allocator) -> (f: T, exact: bool, err: Error)  where intrinsics.type_is_float(T) {
+	FSIZE :: 8*size_of(T)
+	when FSIZE == 16 {
+		MSIZE :: 10
+	} else when FSIZE == 32 {
+		MSIZE :: 23
+	} else when FSIZE == 64 {
+		MSIZE :: 52
+	} else {
+		#panic("unsupported float type")
+	}
+	
+	MSIZE1 :: MSIZE+1
+	MSIZE2 :: MSIZE+2
+	
+	ESIZE :: FSIZE - MSIZE1
+	EBIAS :: 1<<(ESIZE-1) - 1
+	EMIN  :: 1 - EBIAS
+	EMAX  :: EBIAS
+	
+	assert_if_nil(z)
+	a, b := &z.a, &z.b
+	
+	context.allocator = allocator
+	
+	alen := internal_count_bits(a)
+	if alen == 0 {
+		return 0, true, nil
+	}
+	blen := internal_count_bits(b)
+	if blen == 0 {
+		return T(math.nan_f64()), false, .Division_by_Zero
+	}
+	
+	has_sign := a.sign != b.sign
+	defer if has_sign {
+		f = -builtin.abs(f)
+	}
+	
+	exp := alen - blen
+	a2, b2 := &Int{}, &Int{}
+	defer internal_destroy(a2, b2)
+	internal_int_abs(a2, a) or_return
+	internal_int_abs(b2, b) or_return
+	
+	if shift := MSIZE2 - exp; shift > 0 {
+		internal_int_shl_digit(a2, shift) or_return
+	} else {
+		internal_int_shl_digit(b2, -shift) or_return
+	}
+	
+	q, r := &Int{}, &Int{}
+	defer internal_destroy(q, r)
+	
+	internal_int_divmod(q, r, a2, b2) or_return
+	
+	has_rem := !internal_is_zero(r)
+	mantissa := internal_int_get_u64(q) or_return
+	
+	if mantissa>>MSIZE2 == 1 {
+		if mantissa&1 == 1 {
+			has_rem = true
+		}
+		mantissa >>= 1
+		exp += 1
+	}
+	
+	assert(mantissa>>MSIZE1 == 1, "invalid bit result")
+	
+	
+	if EMIN-MSIZE <= exp && exp <= EMIN {
+		shift := uint(EMIN - (exp - 1))
+		lost_bits := mantissa & (1<<shift - 1)
+		has_rem ||= lost_bits != 0
+		mantissa >>= shift
+		exp = 2 - EBIAS // exp + shift
+	}
+	
+	
+	exact = !has_rem
+	if mantissa&1 != 0 {
+		exact = false
+		if has_rem || mantissa&2 != 0 {
+			mantissa += 1
+			if mantissa >= 1<<MSIZE2 {
+				mantissa >>= 1
+				exp += 1
+			}
+		}
+	}
+	
+	mantissa >>= 1
+	
+	f = T(math.ldexp(f64(mantissa), exp-MSIZE1))
+	if math.is_inf(f, 0) {
+		exact = false
+	}
+	return
+}
+
+
+rat_compare :: proc(x, y: ^Rat, allocator := context.allocator) -> (comparison: int, error: Error) {
+	assert_if_nil(x, y)
+	context.allocator = allocator
+	
+	a, b: Int
+	internal_init_multi(&a, &b) or_return
+	defer internal_destroy(&a, &b)
+	internal_int_scale_denom(&a, &x.a, &y.b) or_return
+	internal_int_scale_denom(&b, &y.a, &x.b) or_return
+	return int_compare(&a, &b)
+}
+
+
+
+rat_add_int :: proc(dst, x: ^Rat, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x)
+	assert_if_nil(y)
+	
+	z: Rat
+	rat_set_int(&z, y, allocator) or_return
+	defer internal_destroy(&z)
+	return rat_add_rat(dst, x, &z, allocator)
+}
+
+rat_sub_int :: proc(dst, x: ^Rat, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x)
+	assert_if_nil(y)
+	
+	z: Rat
+	rat_set_int(&z, y, allocator) or_return
+	defer internal_destroy(&z)
+	return rat_sub_rat(dst, x, &z, allocator)
+}
+
+rat_mul_int :: proc(dst, x: ^Rat, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(dst, x)
+	assert_if_nil(y)
+	
+	z: Rat
+	rat_set_int(&z, y, allocator) or_return
+	defer internal_destroy(&z)
+	return rat_mul_rat(dst, x, &z, allocator)
+}
+
+rat_div_int :: proc(dst, x: ^Rat, y: ^Int, allocator := context.allocator) -> (err: Error) {
+	if internal_is_zero(y) {
+		return .Division_by_Zero
+	}
+	z: Rat
+	rat_set_int(&z, y, allocator) or_return
+	defer internal_destroy(&z)
+	return rat_div_rat(dst, x, &z, allocator)
+}
+
+
+int_add_rat :: proc(dst: ^Rat, x: ^Int, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(x)
+	assert_if_nil(dst, y)
+	
+	w: Rat
+	rat_set_int(&w, x, allocator) or_return
+	defer internal_destroy(&w)
+	return rat_add_rat(dst, &w, y, allocator)
+}
+
+int_sub_rat :: proc(dst: ^Rat, x: ^Int, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(x)
+	assert_if_nil(dst, y)
+	
+	w: Rat
+	rat_set_int(&w, x, allocator) or_return
+	defer internal_destroy(&w)
+	return rat_sub_rat(dst, &w, y, allocator)
+}
+
+int_mul_rat :: proc(dst: ^Rat, x: ^Int, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	assert_if_nil(x)
+	assert_if_nil(dst, y)
+	
+	w: Rat
+	rat_set_int(&w, x, allocator) or_return
+	defer internal_destroy(&w)
+	return rat_mul_rat(dst, &w, y, allocator)
+}
+
+int_div_rat :: proc(dst: ^Rat, x: ^Int, y: ^Rat, allocator := context.allocator) -> (err: Error) {
+	if internal_is_zero(y) {
+		return .Division_by_Zero
+	}
+	w: Rat
+	rat_set_int(&w, x, allocator) or_return
+	defer internal_destroy(&w)
+	return rat_div_rat(dst, &w, y, allocator)
+}

core/math/bits/bits.odin

@@ -147,55 +147,53 @@ len :: proc{len_u8, len_u16, len_u32, len_u64, len_uint}
 
 
 add_u32 :: proc(x, y, carry: u32) -> (sum, carry_out: u32) {
-	yc := y + carry
-	sum = x + yc
-	if sum < x || yc < y {
-		carry_out = 1
-	}
+	tmp_carry, tmp_carry2: bool
+	sum, tmp_carry = intrinsics.overflow_add(x, y)
+	sum, tmp_carry2 = intrinsics.overflow_add(sum, carry)
+	carry_out = u32(tmp_carry | tmp_carry2)
 	return
 }
 add_u64 :: proc(x, y, carry: u64) -> (sum, carry_out: u64) {
-	yc := y + carry
-	sum = x + yc
-	if sum < x || yc < y {
-		carry_out = 1
-	}
+	tmp_carry, tmp_carry2: bool
+	sum, tmp_carry = intrinsics.overflow_add(x, y)
+	sum, tmp_carry2 = intrinsics.overflow_add(sum, carry)
+	carry_out = u64(tmp_carry | tmp_carry2)
 	return
 }
 add_uint :: proc(x, y, carry: uint) -> (sum, carry_out: uint) {
-	yc := y + carry
-	sum = x + yc
-	if sum < x || yc < y {
-		carry_out = 1
+	when size_of(uint) == size_of(u64) {
+		a, b := add_u64(u64(x), u64(y), u64(carry))
+	} else {
+		#assert(size_of(uint) == size_of(u32))
+		a, b := add_u32(u32(x), u32(y), u32(carry))
 	}
-	return
+	return uint(a), uint(b)
 }
 add :: proc{add_u32, add_u64, add_uint}
 
 
 sub_u32 :: proc(x, y, borrow: u32) -> (diff, borrow_out: u32) {
-	yb := y + borrow
-	diff = x - yb
-	if diff > x || yb < y {
-		borrow_out = 1
-	}
+	tmp_borrow, tmp_borrow2: bool
+	diff, tmp_borrow = intrinsics.overflow_sub(x, y)
+	diff, tmp_borrow2 = intrinsics.overflow_sub(diff, borrow)
+	borrow_out = u32(tmp_borrow | tmp_borrow2)
 	return
 }
 sub_u64 :: proc(x, y, borrow: u64) -> (diff, borrow_out: u64) {
-	yb := y + borrow
-	diff = x - yb
-	if diff > x || yb < y {
-		borrow_out = 1
-	}
+	tmp_borrow, tmp_borrow2: bool
+	diff, tmp_borrow = intrinsics.overflow_sub(x, y)
+	diff, tmp_borrow2 = intrinsics.overflow_sub(diff, borrow)
+	borrow_out = u64(tmp_borrow | tmp_borrow2)
 	return
 }
 sub_uint :: proc(x, y, borrow: uint) -> (diff, borrow_out: uint) {
-	yb := y + borrow
-	diff = x - yb
-	if diff > x || yb < y {
-		borrow_out = 1
+	when size_of(uint) == size_of(u64) {
+		a, b := sub_u64(u64(x), u64(y), u64(borrow))
+	} else {
+		#assert(size_of(uint) == size_of(u32))
+		a, b := sub_u32(u32(x), u32(y), u32(borrow))
 	}
-	return
+	return uint(a), uint(b)
 }
 sub :: proc{sub_u32, sub_u64, sub_uint}
 
@@ -206,18 +204,8 @@ mul_u32 :: proc(x, y: u32) -> (hi, lo: u32) {
 	return
 }
 mul_u64 :: proc(x, y: u64) -> (hi, lo: u64) {
-	mask :: 1<<32 - 1
-
-	x0, x1 := x & mask, x >> 32
-	y0, y1 := y & mask, y >> 32
-
-	w0 := x0 * y0
-	t := x1*y0 + w0>>32
-
-	w1, w2 := t & mask, t >> 32
-	w1 += x0 * y1
-	hi = x1*y1 + w2 + w1>>32
-	lo = x * y
+	prod_wide := u128(x) * u128(y)
+	hi, lo = u64(prod_wide>>64), u64(prod_wide)
 	return
 }
 

core/math/linalg/glsl/linalg_glsl_math.odin

@@ -0,0 +1,63 @@
+package math_linalg_glsl
+
+import "core:math"
+
+cos_f32         :: proc "c" (x: f32) -> f32 { return math.cos(x) }
+sin_f32         :: proc "c" (x: f32) -> f32 { return math.sin(x) }
+tan_f32         :: proc "c" (x: f32) -> f32 { return math.tan(x) }
+acos_f32        :: proc "c" (x: f32) -> f32 { return math.acos(x) }
+asin_f32        :: proc "c" (x: f32) -> f32 { return math.asin(x) }
+atan_f32        :: proc "c" (x: f32) -> f32 { return math.atan(x) }
+atan2_f32       :: proc "c" (y, x: f32) -> f32 { return math.atan2(y, x) }
+cosh_f32        :: proc "c" (x: f32) -> f32 { return math.cosh(x) }
+sinh_f32        :: proc "c" (x: f32) -> f32 { return math.sinh(x) }
+tanh_f32        :: proc "c" (x: f32) -> f32 { return math.tanh(x) }
+acosh_f32       :: proc "c" (x: f32) -> f32 { return math.acosh(x) }
+asinh_f32       :: proc "c" (x: f32) -> f32 { return math.asinh(x) }
+atanh_f32       :: proc "c" (x: f32) -> f32 { return math.atanh(x) }
+sqrt_f32        :: proc "c" (x: f32) -> f32 { return math.sqrt(x) }
+inversesqrt_f32 :: proc "c" (x: f32) -> f32 { return 1.0/math.sqrt(x) }
+pow_f32         :: proc "c" (x, y: f32) -> f32 { return math.pow(x, y) }
+exp_f32         :: proc "c" (x: f32) -> f32 { return math.exp(x) }
+log_f32         :: proc "c" (x: f32) -> f32 { return math.ln(x) }
+exp2_f32        :: proc "c" (x: f32) -> f32 { return math.pow(f32(2), x) }
+sign_f32        :: proc "c" (x: f32) -> f32 { return math.sign(x) }
+floor_f32       :: proc "c" (x: f32) -> f32 { return math.floor(x) }
+ceil_f32        :: proc "c" (x: f32) -> f32 { return math.ceil(x) }
+mod_f32         :: proc "c" (x, y: f32) -> f32 { return math.mod(x, y) }
+fract_f32 :: proc "c" (x: f32) -> f32 {
+	if x >= 0 {
+		return x - math.trunc(x)
+	}
+	return math.trunc(-x) + x
+}
+
+cos_f64         :: proc "c" (x: f64) -> f64 { return math.cos(x) }
+sin_f64         :: proc "c" (x: f64) -> f64 { return math.sin(x) }
+tan_f64         :: proc "c" (x: f64) -> f64 { return math.tan(x) }
+acos_f64        :: proc "c" (x: f64) -> f64 { return math.acos(x) }
+asin_f64        :: proc "c" (x: f64) -> f64 { return math.asin(x) }
+atan_f64        :: proc "c" (x: f64) -> f64 { return math.atan(x) }
+atan2_f64       :: proc "c" (y, x: f64) -> f64 { return math.atan2(y, x) }
+cosh_f64        :: proc "c" (x: f64) -> f64 { return math.cosh(x) }
+sinh_f64        :: proc "c" (x: f64) -> f64 { return math.sinh(x) }
+tanh_f64        :: proc "c" (x: f64) -> f64 { return math.tanh(x) }
+acosh_f64       :: proc "c" (x: f64) -> f64 { return math.acosh(x) }
+asinh_f64       :: proc "c" (x: f64) -> f64 { return math.asinh(x) }
+atanh_f64       :: proc "c" (x: f64) -> f64 { return math.atanh(x) }
+sqrt_f64        :: proc "c" (x: f64) -> f64 { return math.sqrt(x) }
+inversesqrt_f64 :: proc "c" (x: f64) -> f64 { return 1.0/math.sqrt(x) }
+pow_f64         :: proc "c" (x, y: f64) -> f64 { return math.pow(x, y) }
+exp_f64         :: proc "c" (x: f64) -> f64 { return math.exp(x) }
+log_f64         :: proc "c" (x: f64) -> f64 { return math.ln(x) }
+exp2_f64        :: proc "c" (x: f64) -> f64 { return math.pow(f64(2), x) }
+sign_f64        :: proc "c" (x: f64) -> f64 { return math.sign(x) }
+floor_f64       :: proc "c" (x: f64) -> f64 { return math.floor(x) }
+ceil_f64        :: proc "c" (x: f64) -> f64 { return math.ceil(x) }
+mod_f64         :: proc "c" (x, y: f64) -> f64 { return math.mod(x, y) }
+fract_f64 :: proc "c" (x: f64) -> f64 {
+	if x >= 0 {
+		return x - math.trunc(x)
+	}
+	return math.trunc(-x) + x
+}

core/math/linalg/hlsl/linalg_hlsl_math.odin

@@ -0,0 +1,69 @@
+package math_linalg_hlsl
+
+import "core:math"
+
+cos_float         :: proc "c" (x: float)    -> float { return math.cos(x) }
+sin_float         :: proc "c" (x: float)    -> float { return math.sin(x) }
+tan_float         :: proc "c" (x: float)    -> float { return math.tan(x) }
+acos_float        :: proc "c" (x: float)    -> float { return math.acos(x) }
+asin_float        :: proc "c" (x: float)    -> float { return math.asin(x) }
+atan_float        :: proc "c" (x: float)    -> float { return math.atan(x) }
+atan2_float       :: proc "c" (y, x: float) -> float { return math.atan2(y, x) }
+cosh_float        :: proc "c" (x: float)    -> float { return math.cosh(x) }
+sinh_float        :: proc "c" (x: float)    -> float { return math.sinh(x) }
+tanh_float        :: proc "c" (x: float)    -> float { return math.tanh(x) }
+acosh_float       :: proc "c" (x: float)    -> float { return math.acosh(x) }
+asinh_float       :: proc "c" (x: float)    -> float { return math.asinh(x) }
+atanh_float       :: proc "c" (x: float)    -> float { return math.atanh(x) }
+sqrt_float        :: proc "c" (x: float)    -> float { return math.sqrt(x) }
+rsqrt_float       :: proc "c" (x: float)    -> float { return 1.0/math.sqrt(x) }
+rcp_float         :: proc "c" (x: float)    -> float { return 1.0/x }
+pow_float         :: proc "c" (x, y: float) -> float { return math.pow(x, y) }
+exp_float         :: proc "c" (x: float)    -> float { return math.exp(x) }
+log_float         :: proc "c" (x: float)    -> float { return math.ln(x) }
+log2_float        :: proc "c" (x: float)    -> float { return math.log(x, 2) }
+log10_float       :: proc "c" (x: float)    -> float { return math.log(x, 10) }
+exp2_float        :: proc "c" (x: float)    -> float { return math.pow(float(2), x) }
+sign_float        :: proc "c" (x: float)    -> float { return math.sign(x) }
+floor_float       :: proc "c" (x: float)    -> float { return math.floor(x) }
+ceil_float        :: proc "c" (x: float)    -> float { return math.ceil(x) }
+fmod_float        :: proc "c" (x, y: float) -> float { return math.mod(x, y) }
+frac_float :: proc "c" (x: float) -> float {
+	if x >= 0 {
+		return x - math.trunc(x)
+	}
+	return math.trunc(-x) + x
+}
+
+cos_double         :: proc "c" (x: double)    -> double { return math.cos(x) }
+sin_double         :: proc "c" (x: double)    -> double { return math.sin(x) }
+tan_double         :: proc "c" (x: double)    -> double { return math.tan(x) }
+acos_double        :: proc "c" (x: double)    -> double { return math.acos(x) }
+asin_double        :: proc "c" (x: double)    -> double { return math.asin(x) }
+atan_double        :: proc "c" (x: double)    -> double { return math.atan(x) }
+atan2_double       :: proc "c" (y, x: double) -> double { return math.atan2(y, x) }
+cosh_double        :: proc "c" (x: double)    -> double { return math.cosh(x) }
+sinh_double        :: proc "c" (x: double)    -> double { return math.sinh(x) }
+tanh_double        :: proc "c" (x: double)    -> double { return math.tanh(x) }
+acosh_double       :: proc "c" (x: double)    -> double { return math.acosh(x) }
+asinh_double       :: proc "c" (x: double)    -> double { return math.asinh(x) }
+atanh_double       :: proc "c" (x: double)    -> double { return math.atanh(x) }
+sqrt_double        :: proc "c" (x: double)    -> double { return math.sqrt(x) }
+rsqrt_double       :: proc "c" (x: double)    -> double { return 1.0/math.sqrt(x) }
+rcp_double         :: proc "c" (x: double)    -> double { return 1.0/x }
+pow_double         :: proc "c" (x, y: double) -> double { return math.pow(x, y) }
+exp_double         :: proc "c" (x: double)    -> double { return math.exp(x) }
+log_double         :: proc "c" (x: double)    -> double { return math.ln(x) }
+log2_double        :: proc "c" (x: double)    -> double { return math.log(x, 2) }
+log10_double       :: proc "c" (x: double)    -> double { return math.log(x, 10) }
+exp2_double        :: proc "c" (x: double)    -> double { return math.pow(double(2), x) }
+sign_double        :: proc "c" (x: double)    -> double { return math.sign(x) }
+floor_double       :: proc "c" (x: double)    -> double { return math.floor(x) }
+ceil_double        :: proc "c" (x: double)    -> double { return math.ceil(x) }
+fmod_double        :: proc "c" (x, y: double) -> double { return math.mod(x, y) }
+frac_double :: proc "c" (x: double) -> double {
+	if x >= 0 {
+		return x - math.trunc(x)
+	}
+	return math.trunc(-x) + x
+}

core/math/linalg/specific.odin

@@ -2033,6 +2033,55 @@ matrix4_look_at :: proc{
 }
 
 
+matrix4_look_at_from_fru_f16 :: proc(eye, f, r, u: Vector3f16, flip_z_axis := true) -> (m: Matrix4f16) {
+	f, s, u := f, r, u
+	f = normalize(f)
+	s = normalize(s)
+	u = normalize(u)
+	fe := dot(f, eye)
+
+	return {
+		{+s.x, +u.x, -f.x, 0},
+		{+s.y, +u.y, -f.y, 0},
+		{+s.z, +u.z, -f.z, 0},
+		{-dot(s, eye), -dot(u, eye), +fe if flip_z_axis else -fe, 1},
+	}
+}
+matrix4_look_at_from_fru_f32 :: proc(eye, f, r, u: Vector3f32, flip_z_axis := true) -> (m: Matrix4f32) {
+	f, s, u := f, r, u
+	f = normalize(f)
+	s = normalize(s)
+	u = normalize(u)
+	fe := dot(f, eye)
+
+	return {
+		{+s.x, +u.x, -f.x, 0},
+		{+s.y, +u.y, -f.y, 0},
+		{+s.z, +u.z, -f.z, 0},
+		{-dot(s, eye), -dot(u, eye), +fe if flip_z_axis else -fe, 1},
+	}
+}
+matrix4_look_at_from_fru_f64 :: proc(eye, f, r, u: Vector3f64, flip_z_axis := true) -> (m: Matrix4f64) {
+	f, s, u := f, r, u
+	f = normalize(f)
+	s = normalize(s)
+	u = normalize(u)
+	fe := dot(f, eye)
+
+	return {
+		{+s.x, +u.x, -f.x, 0},
+		{+s.y, +u.y, -f.y, 0},
+		{+s.z, +u.z, -f.z, 0},
+		{-dot(s, eye), -dot(u, eye), +fe if flip_z_axis else -fe, 1},
+	}
+}
+matrix4_look_at_from_fru :: proc{
+	matrix4_look_at_from_fru_f16,
+	matrix4_look_at_from_fru_f32,
+	matrix4_look_at_from_fru_f64,
+}
+
+
 matrix4_perspective_f16 :: proc(fovy, aspect, near, far: f16, flip_z_axis := true) -> (m: Matrix4f16) {
 	tan_half_fovy := math.tan(0.5 * fovy)
 	m[0][0] = 1 / (aspect*tan_half_fovy)

core/math/math_basic.odin

@@ -0,0 +1,169 @@
+//+build !js
+package math
+
+import "core:intrinsics"
+
+@(default_calling_convention="none")
+foreign _ {
+	@(link_name="llvm.sin.f16")
+	sin_f16 :: proc(θ: f16) -> f16 ---
+	@(link_name="llvm.sin.f32")
+	sin_f32 :: proc(θ: f32) -> f32 ---
+	@(link_name="llvm.sin.f64")
+	sin_f64 :: proc(θ: f64) -> f64 ---
+
+	@(link_name="llvm.cos.f16")
+	cos_f16 :: proc(θ: f16) -> f16 ---
+	@(link_name="llvm.cos.f32")
+	cos_f32 :: proc(θ: f32) -> f32 ---
+	@(link_name="llvm.cos.f64")
+	cos_f64 :: proc(θ: f64) -> f64 ---
+
+	@(link_name="llvm.pow.f16")
+	pow_f16 :: proc(x, power: f16) -> f16 ---
+	@(link_name="llvm.pow.f32")
+	pow_f32 :: proc(x, power: f32) -> f32 ---
+	@(link_name="llvm.pow.f64")
+	pow_f64 :: proc(x, power: f64) -> f64 ---
+
+	@(link_name="llvm.fmuladd.f16")
+	fmuladd_f16 :: proc(a, b, c: f16) -> f16 ---
+	@(link_name="llvm.fmuladd.f32")
+	fmuladd_f32 :: proc(a, b, c: f32) -> f32 ---
+	@(link_name="llvm.fmuladd.f64")
+	fmuladd_f64 :: proc(a, b, c: f64) -> f64 ---
+
+	@(link_name="llvm.exp.f16")
+	exp_f16 :: proc(x: f16) -> f16 ---
+	@(link_name="llvm.exp.f32")
+	exp_f32 :: proc(x: f32) -> f32 ---
+	@(link_name="llvm.exp.f64")
+	exp_f64 :: proc(x: f64) -> f64 ---
+}
+
+sqrt_f16 :: proc "contextless" (x: f16) -> f16 {
+	return intrinsics.sqrt(x)
+}
+sqrt_f32 :: proc "contextless" (x: f32) -> f32 {
+	return intrinsics.sqrt(x)
+}
+sqrt_f64 :: proc "contextless" (x: f64) -> f64 {
+	return intrinsics.sqrt(x)
+}
+
+
+
+ln_f64 :: proc "contextless" (x: f64) -> f64 {
+	// The original C code, the long comment, and the constants
+	// below are from FreeBSD's /usr/src/lib/msun/src/e_log.c
+	// and came with this notice.
+	//
+	// ====================================================
+	// Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+	//
+	// Developed at SunPro, a Sun Microsystems, Inc. business.
+	// Permission to use, copy, modify, and distribute this
+	// software is freely granted, provided that this notice
+	// is preserved.
+	// ====================================================
+	//
+	// __ieee754_log(x)
+	// Return the logarithm of x
+	//
+	// Method :
+	//   1. Argument Reduction: find k and f such that
+	//			x = 2**k * (1+f),
+	//	   where  sqrt(2)/2 < 1+f < sqrt(2) .
+	//
+	//   2. Approximation of log(1+f).
+	//	Let s = f/(2+f) ; based on log(1+f) = log(1+s) - log(1-s)
+	//		 = 2s + 2/3 s**3 + 2/5 s**5 + .....,
+	//	     	 = 2s + s*R
+	//      We use a special Reme algorithm on [0,0.1716] to generate
+	//	a polynomial of degree 14 to approximate R.  The maximum error
+	//	of this polynomial approximation is bounded by 2**-58.45. In
+	//	other words,
+	//		        2      4      6      8      10      12      14
+	//	    R(z) ~ L1*s +L2*s +L3*s +L4*s +L5*s  +L6*s  +L7*s
+	//	(the values of L1 to L7 are listed in the program) and
+	//	    |      2          14          |     -58.45
+	//	    | L1*s +...+L7*s    -  R(z) | <= 2
+	//	    |                             |
+	//	Note that 2s = f - s*f = f - hfsq + s*hfsq, where hfsq = f*f/2.
+	//	In order to guarantee error in log below 1ulp, we compute log by
+	//		log(1+f) = f - s*(f - R)		(if f is not too large)
+	//		log(1+f) = f - (hfsq - s*(hfsq+R)).	(better accuracy)
+	//
+	//	3. Finally,  log(x) = k*Ln2 + log(1+f).
+	//			    = k*Ln2_hi+(f-(hfsq-(s*(hfsq+R)+k*Ln2_lo)))
+	//	   Here Ln2 is split into two floating point number:
+	//			Ln2_hi + Ln2_lo,
+	//	   where n*Ln2_hi is always exact for |n| < 2000.
+	//
+	// Special cases:
+	//	log(x) is NaN with signal if x < 0 (including -INF) ;
+	//	log(+INF) is +INF; log(0) is -INF with signal;
+	//	log(NaN) is that NaN with no signal.
+	//
+	// Accuracy:
+	//	according to an error analysis, the error is always less than
+	//	1 ulp (unit in the last place).
+	//
+	// Constants:
+	// The hexadecimal values are the intended ones for the following
+	// constants. The decimal values may be used, provided that the
+	// compiler will convert from decimal to binary accurately enough
+	// to produce the hexadecimal values shown.
+	
+	LN2_HI :: 0h3fe62e42_fee00000 // 6.93147180369123816490e-01
+	LN2_LO :: 0h3dea39ef_35793c76 // 1.90821492927058770002e-10
+	L1     :: 0h3fe55555_55555593 // 6.666666666666735130e-01
+	L2     :: 0h3fd99999_9997fa04 // 3.999999999940941908e-01
+	L3     :: 0h3fd24924_94229359 // 2.857142874366239149e-01
+	L4     :: 0h3fcc71c5_1d8e78af // 2.222219843214978396e-01
+	L5     :: 0h3fc74664_96cb03de // 1.818357216161805012e-01
+	L6     :: 0h3fc39a09_d078c69f // 1.531383769920937332e-01
+	L7     :: 0h3fc2f112_df3e5244 // 1.479819860511658591e-01
+	
+	switch {
+	case is_nan(x) || is_inf(x, 1):
+		return x
+	case x < 0:
+		return nan_f64()
+	case x == 0:
+		return inf_f64(-1)
+	}
+
+	// reduce
+	f1, ki := frexp(x)
+	if f1 < SQRT_TWO/2 {
+		f1 *= 2
+		ki -= 1
+	}
+	f := f1 - 1
+	k := f64(ki)
+
+	// compute
+	s := f / (2 + f)
+	s2 := s * s
+	s4 := s2 * s2
+	t1 := s2 * (L1 + s4*(L3+s4*(L5+s4*L7)))
+	t2 := s4 * (L2 + s4*(L4+s4*L6))
+	R := t1 + t2
+	hfsq := 0.5 * f * f
+	return k*LN2_HI - ((hfsq - (s*(hfsq+R) + k*LN2_LO)) - f)
+}
+
+ln_f16 :: proc "contextless" (x: f16) -> f16 { return #force_inline f16(ln_f64(f64(x))) }
+ln_f32 :: proc "contextless" (x: f32) -> f32 { return #force_inline f32(ln_f64(f64(x))) }
+ln_f16le :: proc "contextless" (x: f16le) -> f16le { return #force_inline f16le(ln_f64(f64(x))) }
+ln_f16be :: proc "contextless" (x: f16be) -> f16be { return #force_inline f16be(ln_f64(f64(x))) }
+ln_f32le :: proc "contextless" (x: f32le) -> f32le { return #force_inline f32le(ln_f64(f64(x))) }
+ln_f32be :: proc "contextless" (x: f32be) -> f32be { return #force_inline f32be(ln_f64(f64(x))) }
+ln_f64le :: proc "contextless" (x: f64le) -> f64le { return #force_inline f64le(ln_f64(f64(x))) }
+ln_f64be :: proc "contextless" (x: f64be) -> f64be { return #force_inline f64be(ln_f64(f64(x))) }
+ln :: proc{
+	ln_f16, ln_f16le, ln_f16be,
+	ln_f32, ln_f32le, ln_f32be,
+	ln_f64, ln_f64le, ln_f64be,
+}

core/math/math_basic_js.odin

@@ -0,0 +1,54 @@
+//+build js
+package math
+
+import "core:intrinsics"
+
+foreign import "odin_env"
+
+@(default_calling_convention="c")
+foreign odin_env {
+	@(link_name="sin")
+	sin_f64 :: proc(θ: f64) -> f64 ---
+	@(link_name="cos")
+	cos_f64 :: proc(θ: f64) -> f64 ---
+	@(link_name="pow")
+	pow_f64 :: proc(x, power: f64) -> f64 ---
+	@(link_name="fmuladd")
+	fmuladd_f64 :: proc(a, b, c: f64) -> f64 ---
+	@(link_name="ln")
+	ln_f64 :: proc(x: f64) -> f64 ---
+	@(link_name="exp")
+	exp_f64 :: proc(x: f64) -> f64 ---
+}
+
+sqrt_f64 :: proc "contextless" (x: f64) -> f64 {
+	return intrinsics.sqrt(x)
+}
+
+sqrt_f16    :: proc "c" (x: f16) -> f16             { return f16(sqrt_f64(f64(x)))                    }
+sin_f16     :: proc "c" (θ: f16) -> f16             { return f16(sin_f64(f64(θ)))                     }
+cos_f16     :: proc "c" (θ: f16) -> f16             { return f16(cos_f64(f64(θ)))                     }
+pow_f16     :: proc "c" (x, power: f16) -> f16      { return f16(pow_f64(f64(x), f64(power)))         }
+fmuladd_f16 :: proc "c" (a, b, c: f16) -> f16       { return f16(fmuladd_f64(f64(a), f64(a), f64(c))) }
+ln_f16      :: proc "c" (x: f16) -> f16             { return f16(ln_f64(f64(x)))                      }
+exp_f16     :: proc "c" (x: f16) -> f16             { return f16(exp_f64(f64(x)))                     }
+
+sqrt_f32    :: proc "c" (x: f32) -> f32             { return f32(sqrt_f64(f64(x)))                    }
+sin_f32     :: proc "c" (θ: f32) -> f32             { return f32(sin_f64(f64(θ)))                     }
+cos_f32     :: proc "c" (θ: f32) -> f32             { return f32(cos_f64(f64(θ)))                     }
+pow_f32     :: proc "c" (x, power: f32) -> f32      { return f32(pow_f64(f64(x), f64(power)))         }
+fmuladd_f32 :: proc "c" (a, b, c: f32) -> f32       { return f32(fmuladd_f64(f64(a), f64(a), f64(c))) }
+ln_f32      :: proc "c" (x: f32) -> f32             { return f32(ln_f64(f64(x)))                      }
+exp_f32     :: proc "c" (x: f32) -> f32             { return f32(exp_f64(f64(x)))                     }
+
+ln_f16le :: proc "contextless" (x: f16le) -> f16le { return #force_inline f16le(ln_f64(f64(x))) }
+ln_f16be :: proc "contextless" (x: f16be) -> f16be { return #force_inline f16be(ln_f64(f64(x))) }
+ln_f32le :: proc "contextless" (x: f32le) -> f32le { return #force_inline f32le(ln_f64(f64(x))) }
+ln_f32be :: proc "contextless" (x: f32be) -> f32be { return #force_inline f32be(ln_f64(f64(x))) }
+ln_f64le :: proc "contextless" (x: f64le) -> f64le { return #force_inline f64le(ln_f64(f64(x))) }
+ln_f64be :: proc "contextless" (x: f64be) -> f64be { return #force_inline f64be(ln_f64(f64(x))) }
+ln :: proc{
+	ln_f16, ln_f16le, ln_f16be,
+	ln_f32, ln_f32le, ln_f32be,
+	ln_f64, ln_f64le, ln_f64be,
+}

core/math/math_erf.odin

@@ -0,0 +1,410 @@
+package math
+
+// The original C code and the long comment below are
+// from FreeBSD's /usr/src/lib/msun/src/s_erf.c and
+// came with this notice. 
+//
+// ====================================================
+// Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+//
+// Developed at SunPro, a Sun Microsystems, Inc. business.
+// Permission to use, copy, modify, and distribute this
+// software is freely granted, provided that this notice
+// is preserved.
+// ====================================================
+//
+//
+// double erf(double x)
+// double erfc(double x)
+//                           x
+//                    2      |\
+//     erf(x)  =  ---------  | exp(-t*t)dt
+//                 sqrt(pi) \|
+//                           0
+//
+//     erfc(x) =  1-erf(x)
+//  Note that
+//              erf(-x) = -erf(x)
+//              erfc(-x) = 2 - erfc(x)
+//
+// Method:
+//      1. For |x| in [0, 0.84375]
+//          erf(x)  = x + x*R(x**2)
+//          erfc(x) = 1 - erf(x)           if x in [-.84375,0.25]
+//                  = 0.5 + ((0.5-x)-x*R)  if x in [0.25,0.84375]
+//         where R = P/Q where P is an odd poly of degree 8 and
+//         Q is an odd poly of degree 10.
+//                                               -57.90
+//                      | R - (erf(x)-x)/x | <= 2
+//
+//
+//         Remark. The formula is derived by noting
+//          erf(x) = (2/sqrt(pi))*(x - x**3/3 + x**5/10 - x**7/42 + ....)
+//         and that
+//          2/sqrt(pi) = 1.128379167095512573896158903121545171688
+//         is close to one. The interval is chosen because the fix
+//         point of erf(x) is near 0.6174 (i.e., erf(x)=x when x is
+//         near 0.6174), and by some experiment, 0.84375 is chosen to
+//         guarantee the error is less than one ulp for erf.
+//
+//      2. For |x| in [0.84375,1.25], let s = |x| - 1, and
+//         c = 0.84506291151 rounded to single (24 bits)
+//              erf(x)  = sign(x) * (c  + P1(s)/Q1(s))
+//              erfc(x) = (1-c)  - P1(s)/Q1(s) if x > 0
+//                        1+(c+P1(s)/Q1(s))    if x < 0
+//              |P1/Q1 - (erf(|x|)-c)| <= 2**-59.06
+//         Remark: here we use the taylor series expansion at x=1.
+//              erf(1+s) = erf(1) + s*Poly(s)
+//                       = 0.845.. + P1(s)/Q1(s)
+//         That is, we use rational approximation to approximate
+//                      erf(1+s) - (c = (single)0.84506291151)
+//         Note that |P1/Q1|< 0.078 for x in [0.84375,1.25]
+//         where
+//              P1(s) = degree 6 poly in s
+//              Q1(s) = degree 6 poly in s
+//
+//      3. For x in [1.25,1/0.35(~2.857143)],
+//              erfc(x) = (1/x)*exp(-x*x-0.5625+R1/S1)
+//              erf(x)  = 1 - erfc(x)
+//         where
+//              R1(z) = degree 7 poly in z, (z=1/x**2)
+//              S1(z) = degree 8 poly in z
+//
+//      4. For x in [1/0.35,28]
+//              erfc(x) = (1/x)*exp(-x*x-0.5625+R2/S2) if x > 0
+//                      = 2.0 - (1/x)*exp(-x*x-0.5625+R2/S2) if -6<x<0
+//                      = 2.0 - tiny            (if x <= -6)
+//              erf(x)  = sign(x)*(1.0 - erfc(x)) if x < 6, else
+//              erf(x)  = sign(x)*(1.0 - tiny)
+//         where
+//              R2(z) = degree 6 poly in z, (z=1/x**2)
+//              S2(z) = degree 7 poly in z
+//
+//      Note1:
+//         To compute exp(-x*x-0.5625+R/S), let s be a single
+//         precision number and s := x; then
+//              -x*x = -s*s + (s-x)*(s+x)
+//              exp(-x*x-0.5626+R/S) =
+//                      exp(-s*s-0.5625)*exp((s-x)*(s+x)+R/S);
+//      Note2:
+//         Here 4 and 5 make use of the asymptotic series
+//                        exp(-x*x)
+//              erfc(x) ~ ---------- * ( 1 + Poly(1/x**2) )
+//                        x*sqrt(pi)
+//         We use rational approximation to approximate
+//              g(s)=f(1/x**2) = log(erfc(x)*x) - x*x + 0.5625
+//         Here is the error bound for R1/S1 and R2/S2
+//              |R1/S1 - f(x)|  < 2**(-62.57)
+//              |R2/S2 - f(x)|  < 2**(-61.52)
+//
+//      5. For inf > x >= 28
+//              erf(x)  = sign(x) *(1 - tiny)  (raise inexact)
+//              erfc(x) = tiny*tiny (raise underflow) if x > 0
+//                      = 2 - tiny if x<0
+//
+//      7. Special case:
+//              erf(0)  = 0, erf(inf)  = 1, erf(-inf) = -1,
+//              erfc(0) = 1, erfc(inf) = 0, erfc(-inf) = 2,
+//              erfc/erf(NaN) is NaN
+
+erf :: proc{
+	erf_f16,
+	erf_f16le,
+	erf_f16be,
+	erf_f32,
+	erf_f32le,
+	erf_f32be,
+	erf_f64,
+}
+
+erf_f16   :: proc "contextless" (x: f16)   -> f16   { return f16(erf_f64(f64(x))) }
+erf_f16le :: proc "contextless" (x: f16le) -> f16le { return f16le(erf_f64(f64(x))) }
+erf_f16be :: proc "contextless" (x: f16be) -> f16be { return f16be(erf_f64(f64(x))) }
+erf_f32   :: proc "contextless" (x: f32)   -> f32   { return f32(erf_f64(f64(x))) }
+erf_f32le :: proc "contextless" (x: f32le) -> f32le { return f32le(erf_f64(f64(x))) }
+erf_f32be :: proc "contextless" (x: f32be) -> f32be { return f32be(erf_f64(f64(x))) }
+
+erf_f64 :: proc "contextless" (x: f64) -> f64 {
+	erx :: 0h3FEB0AC160000000
+	// Coefficients for approximation to  erf in [0, 0.84375]
+	efx  :: 0h3FC06EBA8214DB69
+	efx8 :: 0h3FF06EBA8214DB69
+	pp0  :: 0h3FC06EBA8214DB68
+	pp1  :: 0hBFD4CD7D691CB913
+	pp2  :: 0hBF9D2A51DBD7194F
+	pp3  :: 0hBF77A291236668E4
+	pp4  :: 0hBEF8EAD6120016AC
+	qq1  :: 0h3FD97779CDDADC09
+	qq2  :: 0h3FB0A54C5536CEBA
+	qq3  :: 0h3F74D022C4D36B0F
+	qq4  :: 0h3F215DC9221C1A10
+	qq5  :: 0hBED09C4342A26120
+	// Coefficients for approximation to  erf  in [0.84375, 1.25]
+	pa0 :: 0hBF6359B8BEF77538
+	pa1 :: 0h3FDA8D00AD92B34D
+	pa2 :: 0hBFD7D240FBB8C3F1
+	pa3 :: 0h3FD45FCA805120E4
+	pa4 :: 0hBFBC63983D3E28EC
+	pa5 :: 0h3FA22A36599795EB
+	pa6 :: 0hBF61BF380A96073F
+	qa1 :: 0h3FBB3E6618EEE323
+	qa2 :: 0h3FE14AF092EB6F33
+	qa3 :: 0h3FB2635CD99FE9A7
+	qa4 :: 0h3FC02660E763351F
+	qa5 :: 0h3F8BEDC26B51DD1C
+	qa6 :: 0h3F888B545735151D
+	// Coefficients for approximation to  erfc in [1.25, 1/0.35]
+	ra0 :: 0hBF843412600D6435
+	ra1 :: 0hBFE63416E4BA7360
+	ra2 :: 0hC0251E0441B0E726
+	ra3 :: 0hC04F300AE4CBA38D
+	ra4 :: 0hC0644CB184282266
+	ra5 :: 0hC067135CEBCCABB2
+	ra6 :: 0hC054526557E4D2F2
+	ra7 :: 0hC023A0EFC69AC25C
+	sa1 :: 0h4033A6B9BD707687
+	sa2 :: 0h4061350C526AE721
+	sa3 :: 0h407B290DD58A1A71
+	sa4 :: 0h40842B1921EC2868
+	sa5 :: 0h407AD02157700314
+	sa6 :: 0h405B28A3EE48AE2C
+	sa7 :: 0h401A47EF8E484A93
+	sa8 :: 0hBFAEEFF2EE749A62
+	// Coefficients for approximation to  erfc in [1/.35, 28]
+	rb0 :: 0hBF84341239E86F4A
+	rb1 :: 0hBFE993BA70C285DE
+	rb2 :: 0hC031C209555F995A
+	rb3 :: 0hC064145D43C5ED98
+	rb4 :: 0hC083EC881375F228
+	rb5 :: 0hC09004616A2E5992
+	rb6 :: 0hC07E384E9BDC383F
+	sb1 :: 0h403E568B261D5190
+	sb2 :: 0h40745CAE221B9F0A
+	sb3 :: 0h409802EB189D5118
+	sb4 :: 0h40A8FFB7688C246A
+	sb5 :: 0h40A3F219CEDF3BE6
+	sb6 :: 0h407DA874E79FE763
+	sb7 :: 0hC03670E242712D62
+	
+	
+	VERY_TINY :: 0h0080000000000000
+	SMALL     :: 1.0 / (1 << 28)        // 2**-28
+
+	// special cases
+	switch {
+	case is_nan(x):
+		return nan_f64()
+	case is_inf(x, 1):
+		return 1
+	case is_inf(x, -1):
+		return -1
+	}
+	x := x
+	sign := false
+	if x < 0 {
+		x = -x
+		sign = true
+	}
+	if x < 0.84375 { // |x| < 0.84375
+		temp: f64
+		if x < SMALL { // |x| < 2**-28
+			if x < VERY_TINY {
+				temp = 0.125 * (8.0*x + efx8*x) // avoid underflow
+			} else {
+				temp = x + efx*x
+			}
+		} else {
+			z := x * x
+			r := pp0 + z*(pp1+z*(pp2+z*(pp3+z*pp4)))
+			s := 1 + z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))))
+			y := r / s
+			temp = x + x*y
+		}
+		if sign {
+			return -temp
+		}
+		return temp
+	}
+	if x < 1.25 { // 0.84375 <= |x| < 1.25
+		s := x - 1
+		P := pa0 + s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))))
+		Q := 1 + s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))))
+		if sign {
+			return -erx - P/Q
+		}
+		return erx + P/Q
+	}
+	if x >= 6 { // inf > |x| >= 6
+		if sign {
+			return -1
+		}
+		return 1
+	}
+	s := 1 / (x * x)
+	R, S: f64
+	if x < 1/0.35 { // |x| < 1 / 0.35  ~ 2.857143
+		R = ra0 + s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(ra5+s*(ra6+s*ra7))))))
+		S = 1 + s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(sa5+s*(sa6+s*(sa7+s*sa8)))))))
+	} else { // |x| >= 1 / 0.35  ~ 2.857143
+		R = rb0 + s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(rb5+s*rb6)))))
+		S = 1 + s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(sb5+s*(sb6+s*sb7))))))
+	}
+	z := transmute(f64)(0xffffffff00000000 & transmute(u64)x) // pseudo-single (20-bit) precision x
+	r := exp(-z*z-0.5625) * exp((z-x)*(z+x)+R/S)
+	if sign {
+		return r/x - 1
+	}
+	return 1 - r/x
+}
+
+
+erfc :: proc{
+	erfc_f16,
+	erfc_f16le,
+	erfc_f16be,
+	erfc_f32,
+	erfc_f32le,
+	erfc_f32be,
+	erfc_f64,
+}
+
+erfc_f16   :: proc "contextless" (x: f16)   -> f16   { return f16(erfc_f64(f64(x))) }
+erfc_f16le :: proc "contextless" (x: f16le) -> f16le { return f16le(erfc_f64(f64(x))) }
+erfc_f16be :: proc "contextless" (x: f16be) -> f16be { return f16be(erfc_f64(f64(x))) }
+erfc_f32   :: proc "contextless" (x: f32)   -> f32   { return f32(erfc_f64(f64(x))) }
+erfc_f32le :: proc "contextless" (x: f32le) -> f32le { return f32le(erfc_f64(f64(x))) }
+erfc_f32be :: proc "contextless" (x: f32be) -> f32be { return f32be(erfc_f64(f64(x))) }
+
+erfc_f64 :: proc "contextless" (x: f64) -> f64 {
+	erx :: 0h3FEB0AC160000000
+	// Coefficients for approximation to  erf in [0, 0.84375]
+	efx  :: 0h3FC06EBA8214DB69
+	efx8 :: 0h3FF06EBA8214DB69
+	pp0  :: 0h3FC06EBA8214DB68
+	pp1  :: 0hBFD4CD7D691CB913
+	pp2  :: 0hBF9D2A51DBD7194F
+	pp3  :: 0hBF77A291236668E4
+	pp4  :: 0hBEF8EAD6120016AC
+	qq1  :: 0h3FD97779CDDADC09
+	qq2  :: 0h3FB0A54C5536CEBA
+	qq3  :: 0h3F74D022C4D36B0F
+	qq4  :: 0h3F215DC9221C1A10
+	qq5  :: 0hBED09C4342A26120
+	// Coefficients for approximation to  erf  in [0.84375, 1.25]
+	pa0 :: 0hBF6359B8BEF77538
+	pa1 :: 0h3FDA8D00AD92B34D
+	pa2 :: 0hBFD7D240FBB8C3F1
+	pa3 :: 0h3FD45FCA805120E4
+	pa4 :: 0hBFBC63983D3E28EC
+	pa5 :: 0h3FA22A36599795EB
+	pa6 :: 0hBF61BF380A96073F
+	qa1 :: 0h3FBB3E6618EEE323
+	qa2 :: 0h3FE14AF092EB6F33
+	qa3 :: 0h3FB2635CD99FE9A7
+	qa4 :: 0h3FC02660E763351F
+	qa5 :: 0h3F8BEDC26B51DD1C
+	qa6 :: 0h3F888B545735151D
+	// Coefficients for approximation to  erfc in [1.25, 1/0.35]
+	ra0 :: 0hBF843412600D6435
+	ra1 :: 0hBFE63416E4BA7360
+	ra2 :: 0hC0251E0441B0E726
+	ra3 :: 0hC04F300AE4CBA38D
+	ra4 :: 0hC0644CB184282266
+	ra5 :: 0hC067135CEBCCABB2
+	ra6 :: 0hC054526557E4D2F2
+	ra7 :: 0hC023A0EFC69AC25C
+	sa1 :: 0h4033A6B9BD707687
+	sa2 :: 0h4061350C526AE721
+	sa3 :: 0h407B290DD58A1A71
+	sa4 :: 0h40842B1921EC2868
+	sa5 :: 0h407AD02157700314
+	sa6 :: 0h405B28A3EE48AE2C
+	sa7 :: 0h401A47EF8E484A93
+	sa8 :: 0hBFAEEFF2EE749A62
+	// Coefficients for approximation to  erfc in [1/.35, 28]
+	rb0 :: 0hBF84341239E86F4A
+	rb1 :: 0hBFE993BA70C285DE
+	rb2 :: 0hC031C209555F995A
+	rb3 :: 0hC064145D43C5ED98
+	rb4 :: 0hC083EC881375F228
+	rb5 :: 0hC09004616A2E5992
+	rb6 :: 0hC07E384E9BDC383F
+	sb1 :: 0h403E568B261D5190
+	sb2 :: 0h40745CAE221B9F0A
+	sb3 :: 0h409802EB189D5118
+	sb4 :: 0h40A8FFB7688C246A
+	sb5 :: 0h40A3F219CEDF3BE6
+	sb6 :: 0h407DA874E79FE763
+	sb7 :: 0hC03670E242712D62
+	
+	TINY :: 1.0 / (1 << 56) // 2**-56
+	// special cases
+	switch {
+	case is_nan(x):
+		return nan_f64()
+	case is_inf(x, 1):
+		return 0
+	case is_inf(x, -1):
+		return 2
+	}
+	x := x
+	sign := false
+	if x < 0 {
+		x = -x
+		sign = true
+	}
+	if x < 0.84375 { // |x| < 0.84375
+		temp: f64
+		if x < TINY { // |x| < 2**-56
+			temp = x
+		} else {
+			z := x * x
+			r := pp0 + z*(pp1+z*(pp2+z*(pp3+z*pp4)))
+			s := 1 + z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))))
+			y := r / s
+			if x < 0.25 { // |x| < 1/4
+				temp = x + x*y
+			} else {
+				temp = 0.5 + (x*y + (x - 0.5))
+			}
+		}
+		if sign {
+			return 1 + temp
+		}
+		return 1 - temp
+	}
+	if x < 1.25 { // 0.84375 <= |x| < 1.25
+		s := x - 1
+		P := pa0 + s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))))
+		Q := 1 + s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))))
+		if sign {
+			return 1 + erx + P/Q
+		}
+		return 1 - erx - P/Q
+
+	}
+	if x < 28 { // |x| < 28
+		s := 1 / (x * x)
+		R, S: f64
+		if x < 1/0.35 { // |x| < 1 / 0.35 ~ 2.857143
+			R = ra0 + s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(ra5+s*(ra6+s*ra7))))))
+			S = 1 + s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(sa5+s*(sa6+s*(sa7+s*sa8)))))))
+		} else { // |x| >= 1 / 0.35 ~ 2.857143
+			if sign && x > 6 {
+				return 2 // x < -6
+			}
+			R = rb0 + s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(rb5+s*rb6)))))
+			S = 1 + s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(sb5+s*(sb6+s*sb7))))))
+		}
+		z := transmute(f64)(0xffffffff00000000 & transmute(u64)x) // pseudo-single (20-bit) precision x
+		r := exp(-z*z-0.5625) * exp((z-x)*(z+x)+R/S)
+		if sign {
+			return 2 - r/x
+		}
+		return r / x
+	}
+	if sign {
+		return 2
+	}
+	return 0
+}

core/math/math_gamma.odin

@@ -0,0 +1,226 @@
+package math
+
+// The original C code, the long comment, and the constants
+// below are from http://netlib.sandia.gov/cephes/cprob/gamma.c.
+//
+//      tgamma.c
+//
+//      Gamma function
+//
+// SYNOPSIS:
+//
+// double x, y, tgamma();
+// extern int signgam;
+//
+// y = tgamma( x );
+//
+// DESCRIPTION:
+//
+// Returns gamma function of the argument. The result is
+// correctly signed, and the sign (+1 or -1) is also
+// returned in a global (extern) variable named signgam.
+// This variable is also filled in by the logarithmic gamma
+// function lgamma().
+//
+// Arguments |x| <= 34 are reduced by recurrence and the function
+// approximated by a rational function of degree 6/7 in the
+// interval (2,3).  Large arguments are handled by Stirling's
+// formula. Large negative arguments are made positive using
+// a reflection formula.
+//
+// ACCURACY:
+//
+//                      Relative error:
+// arithmetic   domain     # trials      peak         rms
+//    DEC      -34, 34      10000       1.3e-16     2.5e-17
+//    IEEE    -170,-33      20000       2.3e-15     3.3e-16
+//    IEEE     -33,  33     20000       9.4e-16     2.2e-16
+//    IEEE      33, 171.6   20000       2.3e-15     3.2e-16
+//
+// Error for arguments outside the test range will be larger
+// owing to error amplification by the exponential function.
+//
+// Cephes Math Library Release 2.8:  June, 2000
+// Copyright 1984, 1987, 1989, 1992, 2000 by Stephen L. Moshier
+//
+// The readme file at http://netlib.sandia.gov/cephes/ says:
+//    Some software in this archive may be from the book _Methods and
+// Programs for Mathematical Functions_ (Prentice-Hall or Simon & Schuster
+// International, 1989) or from the Cephes Mathematical Library, a
+// commercial product. In either event, it is copyrighted by the author.
+// What you see here may be used freely but it comes with no support or
+// guarantee.
+//
+//   The two known misprints in the book are repaired here in the
+// source listings for the gamma function and the incomplete beta
+// integral.
+//
+//   Stephen L. Moshier
+//   moshier@na-net.ornl.gov
+
+// Gamma function computed by Stirling's formula.
+// The pair of results must be multiplied together to get the actual answer.
+// The multiplication is left to the caller so that, if careful, the caller can avoid
+// infinity for 172 <= x <= 180.
+// The polynomial is valid for 33 <= x <= 172; larger values are only used
+// in reciprocal and produce denormalized floats. The lower precision there
+// masks any imprecision in the polynomial.
+@(private="file")
+stirling :: proc "contextless" (x: f64) -> (f64, f64) {
+	@(static) gamS := [?]f64{
+		+7.87311395793093628397e-04,
+		-2.29549961613378126380e-04,
+		-2.68132617805781232825e-03,
+		+3.47222221605458667310e-03,
+		+8.33333333333482257126e-02,
+	}
+	
+	if x > 200 {
+		return inf_f64(1), 1
+	}
+	SQRT_TWO_PI :: 0h40040d931ff62706 // 2.506628274631000502417
+	MAX_STIRLING :: 143.01608
+	w := 1 / x
+	w = 1 + w*((((gamS[0]*w+gamS[1])*w+gamS[2])*w+gamS[3])*w+gamS[4])
+	y1 := exp(x)
+	y2 := 1.0
+	if x > MAX_STIRLING { // avoid pow() overflow
+		v := pow(x, 0.5*x-0.25)
+		y1, y2 = v, v/y1
+	} else {
+		y1 = pow(x, x-0.5) / y1
+	}
+	return y1, SQRT_TWO_PI * w * y2
+}
+
+gamma_f64 :: proc "contextless" (x: f64) -> f64 {
+	is_neg_int :: proc "contextless" (x: f64) -> bool {
+		if x < 0 {
+			_, xf := modf(x)
+			return xf == 0
+		}
+		return false
+	}
+	
+	@(static) gamP := [?]f64{
+		1.60119522476751861407e-04,
+		1.19135147006586384913e-03,
+		1.04213797561761569935e-02,
+		4.76367800457137231464e-02,
+		2.07448227648435975150e-01,
+		4.94214826801497100753e-01,
+		9.99999999999999996796e-01,
+	}
+	@(static) gamQ := [?]f64{
+		-2.31581873324120129819e-05,
+		+5.39605580493303397842e-04,
+		-4.45641913851797240494e-03,
+		+1.18139785222060435552e-02,
+		+3.58236398605498653373e-02,
+		-2.34591795718243348568e-01,
+		+7.14304917030273074085e-02,
+		+1.00000000000000000320e+00,
+	}
+
+	
+	EULER :: 0.57721566490153286060651209008240243104215933593992 // A001620
+	
+	switch {
+	case is_neg_int(x) || is_inf(x, -1) || is_nan(x):
+		return nan_f64()
+	case is_inf(x, 1):
+		return inf_f64(1)
+	case x == 0:
+		if signbit(x) {
+			return inf_f64(-1)
+		}
+		return inf_f64(1)
+	}
+	
+	x := x
+	q := abs(x)
+	p := floor(q)
+	if q > 33 {
+		if x >= 0 {
+			y1, y2 := stirling(x)
+			return y1 * y2
+		}
+		// Note: x is negative but (checked above) not a negative integer,
+		// so x must be small enough to be in range for conversion to i64.
+		// If |x| were >= 2⁶³ it would have to be an integer.
+		signgam := 1
+		if ip := i64(p); ip&1 == 0 {
+			signgam = -1
+		}
+		z := q - p
+		if z > 0.5 {
+			p = p + 1
+			z = q - p
+		}
+		z = q * sin(PI*z)
+		if z == 0 {
+			return inf_f64(signgam)
+		}
+		sq1, sq2 := stirling(q)
+		absz := abs(z)
+		d := absz * sq1 * sq2
+		if is_inf(d, 0) {
+			z = PI / absz / sq1 / sq2
+		} else {
+			z = PI / d
+		}
+		return f64(signgam) * z
+	}
+
+	// Reduce argument
+	z := 1.0
+	for x >= 3 {
+		x = x - 1
+		z = z * x
+	}
+	for x < 0 {
+		if x > -1e-09 {
+			if x == 0 {
+				return inf_f64(1)
+			}
+			return z / ((1 + EULER*x) * x)
+		}
+		z = z / x
+		x = x + 1
+	}
+	for x < 2 {
+		if x < 1e-09 {
+			if x == 0 {
+				return inf_f64(1)
+			}
+			return z / ((1 + EULER*x) * x)
+		}
+		z = z / x
+		x = x + 1
+	}
+
+	if x == 2 {
+		return z
+	}
+
+	x = x - 2
+	p = (((((x*gamP[0]+gamP[1])*x+gamP[2])*x+gamP[3])*x+gamP[4])*x+gamP[5])*x + gamP[6]
+	q = ((((((x*gamQ[0]+gamQ[1])*x+gamQ[2])*x+gamQ[3])*x+gamQ[4])*x+gamQ[5])*x+gamQ[6])*x + gamQ[7]
+	return z * p / q
+}
+
+
+gamma_f16   :: proc "contextless" (x: f16)   -> f16   { return f16(gamma_f64(f64(x))) }
+gamma_f16le :: proc "contextless" (x: f16le) -> f16le { return f16le(gamma_f64(f64(x))) }
+gamma_f16be :: proc "contextless" (x: f16be) -> f16be { return f16be(gamma_f64(f64(x))) }
+gamma_f32   :: proc "contextless" (x: f32)   -> f32   { return f32(gamma_f64(f64(x))) }
+gamma_f32le :: proc "contextless" (x: f32le) -> f32le { return f32le(gamma_f64(f64(x))) }
+gamma_f32be :: proc "contextless" (x: f32be) -> f32be { return f32be(gamma_f64(f64(x))) }
+gamma_f64le :: proc "contextless" (x: f64le) -> f64le { return f64le(gamma_f64(f64(x))) }
+gamma_f64be :: proc "contextless" (x: f64be) -> f64be { return f64be(gamma_f64(f64(x))) }
+
+gamma :: proc{
+	gamma_f16, gamma_f16le, gamma_f16be,
+	gamma_f32, gamma_f32le, gamma_f32be,
+	gamma_f64, gamma_f64le, gamma_f64be,
+}

core/math/math_lgamma.odin

@@ -0,0 +1,361 @@
+package math
+
+// The original C code and the long comment below are
+// from FreeBSD's /usr/src/lib/msun/src/e_lgamma_r.c and
+// came with this notice. 
+//
+// ====================================================
+// Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+//
+// Developed at SunPro, a Sun Microsystems, Inc. business.
+// Permission to use, copy, modify, and distribute this
+// software is freely granted, provided that this notice
+// is preserved.
+// ====================================================
+//
+// __ieee754_lgamma_r(x, signgamp)
+// Reentrant version of the logarithm of the Gamma function
+// with user provided pointer for the sign of Gamma(x).
+//
+// Method:
+//   1. Argument Reduction for 0 < x <= 8
+//      Since gamma(1+s)=s*gamma(s), for x in [0,8], we may
+//      reduce x to a number in [1.5,2.5] by
+//              lgamma(1+s) = log(s) + lgamma(s)
+//      for example,
+//              lgamma(7.3) = log(6.3) + lgamma(6.3)
+//                          = log(6.3*5.3) + lgamma(5.3)
+//                          = log(6.3*5.3*4.3*3.3*2.3) + lgamma(2.3)
+//   2. Polynomial approximation of lgamma around its
+//      minimum (ymin=1.461632144968362245) to maintain monotonicity.
+//      On [ymin-0.23, ymin+0.27] (i.e., [1.23164,1.73163]), use
+//              Let z = x-ymin;
+//              lgamma(x) = -1.214862905358496078218 + z**2*poly(z)
+//              poly(z) is a 14 degree polynomial.
+//   2. Rational approximation in the primary interval [2,3]
+//      We use the following approximation:
+//              s = x-2.0;
+//              lgamma(x) = 0.5*s + s*P(s)/Q(s)
+//      with accuracy
+//              |P/Q - (lgamma(x)-0.5s)| < 2**-61.71
+//      Our algorithms are based on the following observation
+//
+//                             zeta(2)-1    2    zeta(3)-1    3
+// lgamma(2+s) = s*(1-Euler) + --------- * s  -  --------- * s  + ...
+//                                 2                 3
+//
+//      where Euler = 0.5772156649... is the Euler constant, which
+//      is very close to 0.5.
+//
+//   3. For x>=8, we have
+//      lgamma(x)~(x-0.5)log(x)-x+0.5*log(2pi)+1/(12x)-1/(360x**3)+....
+//      (better formula:
+//         lgamma(x)~(x-0.5)*(log(x)-1)-.5*(log(2pi)-1) + ...)
+//      Let z = 1/x, then we approximation
+//              f(z) = lgamma(x) - (x-0.5)(log(x)-1)
+//      by
+//                                  3       5             11
+//              w = w0 + w1*z + w2*z  + w3*z  + ... + w6*z
+//      where
+//              |w - f(z)| < 2**-58.74
+//
+//   4. For negative x, since (G is gamma function)
+//              -x*G(-x)*G(x) = pi/sin(pi*x),
+//      we have
+//              G(x) = pi/(sin(pi*x)*(-x)*G(-x))
+//      since G(-x) is positive, sign(G(x)) = sign(sin(pi*x)) for x<0
+//      Hence, for x<0, signgam = sign(sin(pi*x)) and
+//              lgamma(x) = log(|Gamma(x)|)
+//                        = log(pi/(|x*sin(pi*x)|)) - lgamma(-x);
+//      Note: one should avoid computing pi*(-x) directly in the
+//            computation of sin(pi*(-x)).
+//
+//   5. Special Cases
+//              lgamma(2+s) ~ s*(1-Euler) for tiny s
+//              lgamma(1)=lgamma(2)=0
+//              lgamma(x) ~ -log(x) for tiny x
+//              lgamma(0) = lgamma(inf) = inf
+//              lgamma(-integer) = +-inf
+//
+//
+
+
+lgamma_f64 :: proc "contextless" (x: f64) -> (lgamma: f64, sign: int) {
+	sin_pi :: proc "contextless" (x: f64) -> f64 {
+		if x < 0.25 {
+			return -sin(PI * x)
+		}
+		x := x
+
+		// argument reduction
+		z := floor(x)
+		n: int
+		if z != x { // inexact
+			x = mod(x, 2)
+			n = int(x * 4)
+		} else {
+			if x >= TWO_53 { // x must be even
+				x = 0
+				n = 0
+			} else {
+				if x < TWO_52 {
+					z = x + TWO_52 // exact
+				}
+				n = int(1 & transmute(u64)z)
+				x = f64(n)
+				n <<= 2
+			}
+		}
+		switch n {
+		case 0:
+			x = sin(PI * x)
+		case 1, 2:
+			x = cos(PI * (0.5 - x))
+		case 3, 4:
+			x = sin(PI * (1 - x))
+		case 5, 6:
+			x = -cos(PI * (x - 1.5))
+		case:
+			x = sin(PI * (x - 2))
+		}
+		return -x
+	}
+	
+	@static lgamA := [?]f64{
+		0h3FB3C467E37DB0C8,
+		0h3FD4A34CC4A60FAD,
+		0h3FB13E001A5562A7,
+		0h3F951322AC92547B,
+		0h3F7E404FB68FEFE8,
+		0h3F67ADD8CCB7926B,
+		0h3F538A94116F3F5D,
+		0h3F40B6C689B99C00,
+		0h3F2CF2ECED10E54D,
+		0h3F1C5088987DFB07,
+		0h3EFA7074428CFA52,
+		0h3F07858E90A45837,
+	}
+	@static lgamR := [?]f64{
+		1.0,
+		0h3FF645A762C4AB74,
+		0h3FE71A1893D3DCDC,
+		0h3FC601EDCCFBDF27,
+		0h3F9317EA742ED475,
+		0h3F497DDACA41A95B,
+		0h3EDEBAF7A5B38140,
+	}
+	@static lgamS := [?]f64{
+		0hBFB3C467E37DB0C8,
+		0h3FCB848B36E20878,
+		0h3FD4D98F4F139F59,
+		0h3FC2BB9CBEE5F2F7,
+		0h3F9B481C7E939961,
+		0h3F5E26B67368F239,
+		0h3F00BFECDD17E945,
+	}
+	@static lgamT := [?]f64{
+		0h3FDEF72BC8EE38A2,
+		0hBFC2E4278DC6C509,
+		0h3FB08B4294D5419B,
+		0hBFA0C9A8DF35B713,
+		0h3F9266E7970AF9EC,
+		0hBF851F9FBA91EC6A,
+		0h3F78FCE0E370E344,
+		0hBF6E2EFFB3E914D7,
+		0h3F6282D32E15C915,
+		0hBF56FE8EBF2D1AF1,
+		0h3F4CDF0CEF61A8E9,
+		0hBF41A6109C73E0EC,
+		0h3F34AF6D6C0EBBF7,
+		0hBF347F24ECC38C38,
+		0h3F35FD3EE8C2D3F4,
+	}
+	@static lgamU := [?]f64{
+		0hBFB3C467E37DB0C8,
+		0h3FE4401E8B005DFF,
+		0h3FF7475CD119BD6F,
+		0h3FEF497644EA8450,
+		0h3FCD4EAEF6010924,
+		0h3F8B678BBF2BAB09,
+	}
+	@static lgamV := [?]f64{
+		1.0,
+		0h4003A5D7C2BD619C,
+		0h40010725A42B18F5,
+		0h3FE89DFBE45050AF,
+		0h3FBAAE55D6537C88,
+		0h3F6A5ABB57D0CF61,
+	}
+	@static lgamW := [?]f64{
+		0h3FDACFE390C97D69,
+		0h3FB555555555553B,
+		0hBF66C16C16B02E5C,
+		0h3F4A019F98CF38B6,
+		0hBF4380CB8C0FE741,
+		0h3F4B67BA4CDAD5D1,
+		0hBF5AB89D0B9E43E4,
+	}
+
+	
+	Y_MIN  :: 0h3ff762d86356be3f // 1.461632144968362245
+	TWO_52 :: 0h4330000000000000 // ~4.5036e+15
+	TWO_53 :: 0h4340000000000000 // ~9.0072e+15
+	TWO_58 :: 0h4390000000000000 // ~2.8823e+17
+	TINY   :: 0h3b90000000000000 // ~8.47033e-22
+	Tc     :: 0h3FF762D86356BE3F
+	Tf     :: 0hBFBF19B9BCC38A42
+	Tt     :: 0hBC50C7CAA48A971F
+	
+	// special cases
+	sign = 1
+	switch {
+	case is_nan(x):
+		lgamma = x
+		return
+	case is_inf(x):
+		lgamma = x
+		return
+	case x == 0:
+		lgamma = inf_f64(1)
+		return
+	}
+
+	x := x
+	neg := false
+	if x < 0 {
+		x = -x
+		neg = true
+	}
+
+	if x < TINY { // if |x| < 2**-70, return -log(|x|)
+		if neg {
+			sign = -1
+		}
+		lgamma = -ln(x)
+		return
+	}
+	nadj: f64
+	if neg {
+		if x >= TWO_52 { // |x| >= 2**52, must be -integer
+			lgamma = inf_f64(1)
+			return
+		}
+		t := sin_pi(x)
+		if t == 0 {
+			lgamma = inf_f64(1) // -integer
+			return
+		}
+		nadj = ln(PI / abs(t*x))
+		if t < 0 {
+			sign = -1
+		}
+	}
+
+	switch {
+	case x == 1 || x == 2: // purge off 1 and 2
+		lgamma = 0
+		return
+	case x < 2: // use lgamma(x) = lgamma(x+1) - log(x)
+		y: f64
+		i: int
+		if x <= 0.9 {
+			lgamma = -ln(x)
+			switch {
+			case x >= (Y_MIN - 1 + 0.27): // 0.7316 <= x <=  0.9
+				y = 1 - x
+				i = 0
+			case x >= (Y_MIN - 1 - 0.27): // 0.2316 <= x < 0.7316
+				y = x - (Tc - 1)
+				i = 1
+			case: // 0 < x < 0.2316
+				y = x
+				i = 2
+			}
+		} else {
+			lgamma = 0
+			switch {
+			case x >= (Y_MIN + 0.27): // 1.7316 <= x < 2
+				y = 2 - x
+				i = 0
+			case x >= (Y_MIN - 0.27): // 1.2316 <= x < 1.7316
+				y = x - Tc
+				i = 1
+			case: // 0.9 < x < 1.2316
+				y = x - 1
+				i = 2
+			}
+		}
+		switch i {
+		case 0:
+			z := y * y
+			p1 := lgamA[0] + z*(lgamA[2]+z*(lgamA[4]+z*(lgamA[6]+z*(lgamA[8]+z*lgamA[10]))))
+			p2 := z * (lgamA[1] + z*(+lgamA[3]+z*(lgamA[5]+z*(lgamA[7]+z*(lgamA[9]+z*lgamA[11])))))
+			p := y*p1 + p2
+			lgamma += (p - 0.5*y)
+		case 1:
+			z := y * y
+			w := z * y
+			p1 := lgamT[0] + w*(lgamT[3]+w*(lgamT[6]+w*(lgamT[9]+w*lgamT[12]))) // parallel comp
+			p2 := lgamT[1] + w*(lgamT[4]+w*(lgamT[7]+w*(lgamT[10]+w*lgamT[13])))
+			p3 := lgamT[2] + w*(lgamT[5]+w*(lgamT[8]+w*(lgamT[11]+w*lgamT[14])))
+			p := z*p1 - (Tt - w*(p2+y*p3))
+			lgamma += (Tf + p)
+		case 2:
+			p1 := y * (lgamU[0] + y*(lgamU[1]+y*(lgamU[2]+y*(lgamU[3]+y*(lgamU[4]+y*lgamU[5])))))
+			p2 := 1 + y*(lgamV[1]+y*(lgamV[2]+y*(lgamV[3]+y*(lgamV[4]+y*lgamV[5]))))
+			lgamma += (-0.5*y + p1/p2)
+		}
+	case x < 8: // 2 <= x < 8
+		i := int(x)
+		y := x - f64(i)
+		p := y * (lgamS[0] + y*(lgamS[1]+y*(lgamS[2]+y*(lgamS[3]+y*(lgamS[4]+y*(lgamS[5]+y*lgamS[6]))))))
+		q := 1 + y*(lgamR[1]+y*(lgamR[2]+y*(lgamR[3]+y*(lgamR[4]+y*(lgamR[5]+y*lgamR[6])))))
+		lgamma = 0.5*y + p/q
+		z := 1.0 // lgamma(1+s) = ln(s) + lgamma(s)
+		switch i {
+		case 7:
+			z *= (y + 6)
+			fallthrough
+		case 6:
+			z *= (y + 5)
+			fallthrough
+		case 5:
+			z *= (y + 4)
+			fallthrough
+		case 4:
+			z *= (y + 3)
+			fallthrough
+		case 3:
+			z *= (y + 2)
+			lgamma += ln(z)
+		}
+	case x < TWO_58: // 8 <= x < 2**58
+		t := ln(x)
+		z := 1 / x
+		y := z * z
+		w := lgamW[0] + z*(lgamW[1]+y*(lgamW[2]+y*(lgamW[3]+y*(lgamW[4]+y*(lgamW[5]+y*lgamW[6])))))
+		lgamma = (x-0.5)*(t-1) + w
+	case: // 2**58 <= x <= Inf
+		lgamma = x * (ln(x) - 1)
+	}
+	if neg {
+		lgamma = nadj - lgamma
+	}
+	return
+}
+
+
+lgamma_f16   :: proc "contextless" (x: f16)   -> (lgamma: f16, sign: int)   { r, s := lgamma_f64(f64(x)); return f16(r), s }
+lgamma_f32   :: proc "contextless" (x: f32)   -> (lgamma: f32, sign: int)   { r, s := lgamma_f64(f64(x)); return f32(r), s }
+lgamma_f16le :: proc "contextless" (x: f16le) -> (lgamma: f16le, sign: int) { r, s := lgamma_f64(f64(x)); return f16le(r), s }
+lgamma_f16be :: proc "contextless" (x: f16be) -> (lgamma: f16be, sign: int) { r, s := lgamma_f64(f64(x)); return f16be(r), s }
+lgamma_f32le :: proc "contextless" (x: f32le) -> (lgamma: f32le, sign: int) { r, s := lgamma_f64(f64(x)); return f32le(r), s }
+lgamma_f32be :: proc "contextless" (x: f32be) -> (lgamma: f32be, sign: int) { r, s := lgamma_f64(f64(x)); return f32be(r), s }
+lgamma_f64le :: proc "contextless" (x: f64le) -> (lgamma: f64le, sign: int) { r, s := lgamma_f64(f64(x)); return f64le(r), s }
+lgamma_f64be :: proc "contextless" (x: f64be) -> (lgamma: f64be, sign: int) { r, s := lgamma_f64(f64(x)); return f64be(r), s }
+
+lgamma :: proc{
+	lgamma_f16, lgamma_f16le, lgamma_f16be,
+	lgamma_f32, lgamma_f32le, lgamma_f32be,
+	lgamma_f64, lgamma_f64le, lgamma_f64be,
+}

core/math/math_log1p.odin

@@ -0,0 +1,198 @@
+package math
+
+// The original C code, the long comment, and the constants
+// below are from FreeBSD's /usr/src/lib/msun/src/s_log1p.c
+// and came with this notice. The go code is a simplified
+// version of the original C.
+//
+// ====================================================
+// Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+//
+// Developed at SunPro, a Sun Microsystems, Inc. business.
+// Permission to use, copy, modify, and distribute this
+// software is freely granted, provided that this notice
+// is preserved.
+// ====================================================
+//
+//
+// double log1p(double x)
+//
+// Method :
+//   1. Argument Reduction: find k and f such that
+//                      1+x = 2**k * (1+f),
+//         where  sqrt(2)/2 < 1+f < sqrt(2) .
+//
+//      Note. If k=0, then f=x is exact. However, if k!=0, then f
+//      may not be representable exactly. In that case, a correction
+//      term is need. Let u=1+x rounded. Let c = (1+x)-u, then
+//      log(1+x) - log(u) ~ c/u. Thus, we proceed to compute log(u),
+//      and add back the correction term c/u.
+//      (Note: when x > 2**53, one can simply return log(x))
+//
+//   2. Approximation of log1p(f).
+//      Let s = f/(2+f) ; based on log(1+f) = log(1+s) - log(1-s)
+//               = 2s + 2/3 s**3 + 2/5 s**5 + .....,
+//               = 2s + s*R
+//      We use a special Reme algorithm on [0,0.1716] to generate
+//      a polynomial of degree 14 to approximate R The maximum error
+//      of this polynomial approximation is bounded by 2**-58.45. In
+//      other words,
+//                      2      4      6      8      10      12      14
+//          R(z) ~ Lp1*s +Lp2*s +Lp3*s +Lp4*s +Lp5*s  +Lp6*s  +Lp7*s
+//      (the values of Lp1 to Lp7 are listed in the program)
+//      and
+//          |      2          14          |     -58.45
+//          | Lp1*s +...+Lp7*s    -  R(z) | <= 2
+//          |                             |
+//      Note that 2s = f - s*f = f - hfsq + s*hfsq, where hfsq = f*f/2.
+//      In order to guarantee error in log below 1ulp, we compute log
+//      by
+//              log1p(f) = f - (hfsq - s*(hfsq+R)).
+//
+//   3. Finally, log1p(x) = k*ln2 + log1p(f).
+//                        = k*ln2_hi+(f-(hfsq-(s*(hfsq+R)+k*ln2_lo)))
+//      Here ln2 is split into two floating point number:
+//                   ln2_hi + ln2_lo,
+//      where n*ln2_hi is always exact for |n| < 2000.
+//
+// Special cases:
+//      log1p(x) is NaN with signal if x < -1 (including -INF) ;
+//      log1p(+INF) is +INF; log1p(-1) is -INF with signal;
+//      log1p(NaN) is that NaN with no signal.
+//
+// Accuracy:
+//      according to an error analysis, the error is always less than
+//      1 ulp (unit in the last place).
+//
+// Constants:
+// The hexadecimal values are the intended ones for the following
+// constants. The decimal values may be used, provided that the
+// compiler will convert from decimal to binary accurately enough
+// to produce the hexadecimal values shown.
+//
+// Note: Assuming log() return accurate answer, the following
+//       algorithm can be used to compute log1p(x) to within a few ULP:
+//
+//              u = 1+x;
+//              if(u==1.0) return x ; else
+//                         return log(u)*(x/(u-1.0));
+//
+//       See HP-15C Advanced Functions Handbook, p.193.
+
+log1p :: proc {
+	log1p_f16,
+	log1p_f32,
+	log1p_f64,
+	log1p_f16le,
+	log1p_f16be,
+	log1p_f32le,
+	log1p_f32be,
+	log1p_f64le,
+	log1p_f64be,
+}
+log1p_f16   :: proc "contextless" (x: f16)   -> f16   { return f16(log1p_f64(f64(x))) }
+log1p_f32   :: proc "contextless" (x: f32)   -> f32   { return f32(log1p_f64(f64(x))) }
+log1p_f16le :: proc "contextless" (x: f16le) -> f16le { return f16le(log1p_f64(f64(x))) }
+log1p_f16be :: proc "contextless" (x: f16be) -> f16be { return f16be(log1p_f64(f64(x))) }
+log1p_f32le :: proc "contextless" (x: f32le) -> f32le { return f32le(log1p_f64(f64(x))) }
+log1p_f32be :: proc "contextless" (x: f32be) -> f32be { return f32be(log1p_f64(f64(x))) }
+log1p_f64le :: proc "contextless" (x: f64le) -> f64le { return f64le(log1p_f64(f64(x))) }
+log1p_f64be :: proc "contextless" (x: f64be) -> f64be { return f64be(log1p_f64(f64(x))) }
+
+log1p_f64 :: proc "contextless" (x: f64) -> f64 {
+	SQRT2_M1      :: 0h3fda827999fcef34 // sqrt(2)-1 
+	SQRT2_HALF_M1 :: 0hbfd2bec333018866 // sqrt(2)/2-1
+	SMALL         :: 0h3e20000000000000 // 2**-29
+	TINY          :: 0h3c90000000000000 // 2**-54
+	TWO53         :: 0h4340000000000000 // 2**53
+	LN2HI         :: 0h3fe62e42fee00000
+	LN2LO         :: 0h3dea39ef35793c76
+	LP1           :: 0h3FE5555555555593
+	LP2           :: 0h3FD999999997FA04
+	LP3           :: 0h3FD2492494229359
+	LP4           :: 0h3FCC71C51D8E78AF
+	LP5           :: 0h3FC7466496CB03DE
+	LP6           :: 0h3FC39A09D078C69F
+	LP7           :: 0h3FC2F112DF3E5244
+	
+	switch {
+	case x < -1 || is_nan(x):
+		return nan_f64()
+	case x == -1:
+		return inf_f64(-1)
+	case is_inf(x, 1):
+		return inf_f64(+1)
+	}
+	absx := abs(x)
+	
+	f: f64
+	iu: u64
+	k := 1
+	if absx < SQRT2_M1 { //  |x| < sqrt(2)-1
+		if absx < SMALL { // |x| < 2**-29
+			if absx < TINY { // |x| < 2**-54
+				return x
+			}
+			return x - x*x*0.5
+		}
+		if x > SQRT2_HALF_M1 { // sqrt(2)/2-1 < x
+			// (sqrt(2)/2-1) < x < (sqrt(2)-1)
+			k = 0
+			f = x
+			iu = 1
+		}
+	}
+	c: f64
+	if k != 0 {
+		u: f64
+		if absx < TWO53 { // 1<<53
+			u = 1.0 + x
+			iu = transmute(u64)u
+			k = int((iu >> 52) - 1023)
+			// correction term
+			if k > 0 {
+				c = 1.0 - (u - x)
+			} else {
+				c = x - (u - 1.0)
+			}
+			c /= u
+		} else {
+			u = x
+			iu = transmute(u64)u
+			k = int((iu >> 52) - 1023)
+			c = 0
+		}
+		iu &= 0x000fffffffffffff
+		if iu < 0x0006a09e667f3bcd { // mantissa of sqrt(2)
+			u = transmute(f64)(iu | 0x3ff0000000000000) // normalize u
+		} else {
+			k += 1
+			u = transmute(f64)(iu | 0x3fe0000000000000) // normalize u/2
+			iu = (0x0010000000000000 - iu) >> 2
+		}
+		f = u - 1.0 // sqrt(2)/2 < u < sqrt(2)
+	}
+	hfsq := 0.5 * f * f
+	s, R, z: f64
+	if iu == 0 { // |f| < 2**-20
+		if f == 0 {
+			if k == 0 {
+				return 0
+			}
+			c += f64(k) * LN2LO
+			return f64(k)*LN2HI + c
+		}
+		R = hfsq * (1.0 - 0.66666666666666666*f) // avoid division
+		if k == 0 {
+			return f - R
+		}
+		return f64(k)*LN2HI - ((R - (f64(k)*LN2LO + c)) - f)
+	}
+	s = f / (2.0 + f)
+	z = s * s
+	R = z * (LP1 + z*(LP2+z*(LP3+z*(LP4+z*(LP5+z*(LP6+z*LP7))))))
+	if k == 0 {
+		return f - (hfsq - s*(hfsq+R))
+	}
+	return f64(k)*LN2HI - ((hfsq - (s*(hfsq+R) + (f64(k)*LN2LO + c))) - f)
+}

core/mem/mem.odin

@@ -10,6 +10,15 @@ zero :: proc "contextless" (data: rawptr, len: int) -> rawptr {
 	intrinsics.mem_zero(data, len)
 	return data
 }
+zero_explicit :: proc "contextless" (data: rawptr, len: int) -> rawptr {
+	// This routine tries to avoid the compiler optimizing away the call,
+	// so that it is always executed.  It is intended to provided
+	// equivalent semantics to those provided by the C11 Annex K 3.7.4.1
+	// memset_s call.
+	intrinsics.mem_zero_volatile(data, len) // Use the volatile mem_zero
+	intrinsics.atomic_fence() // Prevent reordering
+	return data
+}
 zero_item :: proc "contextless" (item: $P/^$T) {
 	intrinsics.mem_zero(item, size_of(T))
 }
@@ -128,14 +137,8 @@ compare_ptrs :: proc "contextless" (a, b: rawptr, n: int) -> int {
 	return compare_byte_ptrs((^byte)(a), (^byte)(b), n)
 }
 
-ptr_offset :: proc "contextless" (ptr: $P/^$T, n: int) -> P {
-	new := int(uintptr(ptr)) + size_of(T)*n
-	return P(uintptr(new))
-}
-
-ptr_sub :: proc "contextless" (a, b: $P/^$T) -> int {
-	return (int(uintptr(a)) - int(uintptr(b)))/size_of(T)
-}
+ptr_offset :: intrinsics.ptr_offset
+ptr_sub :: intrinsics.ptr_sub
 
 slice_ptr :: proc "contextless" (ptr: ^$T, len: int) -> []T {
 	return ([^]T)(ptr)[:len]
@@ -179,7 +182,6 @@ buffer_from_slice :: proc "contextless" (backing: $T/[]$E) -> [dynamic]E {
 }
 
 ptr_to_bytes :: proc "contextless" (ptr: ^$T, len := 1) -> []byte {
-	assert(len >= 0)
 	return transmute([]byte)Raw_Slice{ptr, len*size_of(T)}
 }
 

core/mem/virtual/arena.odin

@@ -1,117 +0,0 @@
-package mem_virtual
-
-import "core:mem"
-import sync "core:sync/sync2"
-
-Growing_Arena :: struct {
-	curr_block:      ^Memory_Block,
-	total_used:      int,
-	total_allocated: int,
-	
-	minimum_block_size: int,
-
-	ignore_mutex: bool,	
-	mutex: sync.Mutex,
-}
-
-DEFAULT_MINIMUM_BLOCK_SIZE :: 8*1024*1024
-DEFAULT_PAGE_SIZE := 4096
-
-growing_arena_alloc :: proc(arena: ^Growing_Arena, min_size: int, alignment: int) -> (data: []byte, err: mem.Allocator_Error) {
-	align_forward_offset :: proc(arena: ^Growing_Arena, alignment: int) -> int #no_bounds_check {
-		alignment_offset := 0
-		ptr := uintptr(arena.curr_block.base[arena.curr_block.used:])
-		mask := uintptr(alignment-1)
-		if ptr & mask != 0 {
-			alignment_offset = alignment - int(ptr & mask)
-		}
-		return alignment_offset
-	}
-	
-	assert(mem.is_power_of_two(uintptr(alignment)))
-	
-	mutex := &arena.mutex
-	if !arena.ignore_mutex {
-		sync.mutex_lock(mutex)
-	}
-	
-	size := 0
-	if arena.curr_block != nil {
-		size = min_size + align_forward_offset(arena, alignment)
-	}
-	
-	if arena.curr_block == nil || arena.curr_block.used + size > arena.curr_block.size {
-		size = mem.align_forward_int(min_size, alignment)
-		arena.minimum_block_size = max(DEFAULT_MINIMUM_BLOCK_SIZE, arena.minimum_block_size)
-		
-		block_size := max(size, arena.minimum_block_size)
-		
-		new_block := memory_alloc(block_size) or_return
-		new_block.prev = arena.curr_block
-		arena.curr_block = new_block
-		arena.total_allocated += new_block.size
-	}
-	
-	curr_block := arena.curr_block
-	assert(curr_block.used + size <= curr_block.size)
-	
-	ptr := curr_block.base[curr_block.used:]
-	ptr = ptr[uintptr(align_forward_offset(arena, alignment)):]
-	
-	curr_block.used += size
-	assert(curr_block.used <= curr_block.size)
-	arena.total_used += size
-	
-	if !arena.ignore_mutex {
-		sync.mutex_unlock(mutex)
-	}
-	
-	return ptr[:min_size], nil
-}
-
-growing_arena_free_all :: proc(arena: ^Growing_Arena) {
-	mutex := &arena.mutex
-	if !arena.ignore_mutex {
-		sync.mutex_lock(mutex)
-	}
-	for arena.curr_block != nil {
-		free_block := arena.curr_block
-		arena.curr_block = free_block.prev
-		memory_dealloc(free_block)
-	}
-	arena.total_used = 0
-	if !arena.ignore_mutex {
-		sync.mutex_unlock(mutex)
-	}
-}
-
-growing_arena_allocator :: proc(arena: ^Growing_Arena) -> mem.Allocator {
-	return mem.Allocator{growing_arena_allocator_proc, arena}
-}
-
-growing_arena_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
-                             size, alignment: int,
-                             old_memory: rawptr, old_size: int,
-                             location := #caller_location) -> (data: []byte, err: mem.Allocator_Error) {
-	arena := (^Growing_Arena)(allocator_data)
-	
-	switch mode {
-	case .Alloc:
-		return growing_arena_alloc(arena, size, alignment)
-	case .Free:
-		err = .Mode_Not_Implemented
-		return
-	case .Free_All:
-		growing_arena_free_all(arena)
-		return
-	case .Resize:
-		return mem.default_resize_bytes_align(mem.byte_slice(old_memory, old_size), size, alignment, growing_arena_allocator(arena), location)
-		
-	case .Query_Features, .Query_Info:
-		err = .Mode_Not_Implemented
-		return	
-	}	
-	
-	err = .Mode_Not_Implemented
-	return 
-}

core/mem/virtual/arena_util.odin

@@ -0,0 +1,36 @@
+package mem_virtual
+
+arena_temp_begin :: proc{
+	static_arena_temp_begin,
+	growing_arena_temp_begin,
+}
+
+arena_temp_end :: proc{
+	static_arena_temp_end,
+	growing_arena_temp_end,
+}
+
+arena_check_temp :: proc{
+	static_arena_check_temp,
+	growing_arena_check_temp,
+}
+
+arena_allocator :: proc{
+	static_arena_allocator,
+	growing_arena_allocator,
+}
+
+arena_alloc :: proc{
+	static_arena_alloc,
+	growing_arena_alloc,
+}
+
+arena_free_all :: proc{
+	static_arena_free_all,
+	growing_arena_free_all,
+}
+
+arena_destroy :: proc{
+	static_arena_destroy,
+	growing_arena_destroy,
+}