math.big constants were no longer initialized

Fixes #3243

.editorconfig

@@ -0,0 +1,10 @@
+root = true
+
+[*]
+indent_style = tab
+end_of_line = lf
+
+[*.y{,a}ml]
+indent_size = 2
+indent_style = space
+tab_width = 2

.gitattributes

@@ -1 +1,2 @@
 *.odin linguist-language=Odin
+* text=auto

.github/workflows/ci.yml

@@ -3,6 +3,7 @@ on: [push, pull_request, workflow_dispatch]
 
 jobs:
   build_linux:
+    name: Ubuntu Build, Check, and Test
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v1
@@ -46,6 +47,9 @@ jobs:
       - name: Odin check examples/all for Linux i386
         run: ./odin check examples/all -vet -strict-style -target:linux_i386
         timeout-minutes: 10
+      - name: Odin check examples/all for Linux arm64
+        run: ./odin check examples/all -vet -strict-style -target:linux_arm64
+        timeout-minutes: 10
       - name: Odin check examples/all for FreeBSD amd64
         run: ./odin check examples/all -vet -strict-style -target:freebsd_amd64
         timeout-minutes: 10
@@ -53,13 +57,14 @@ jobs:
         run: ./odin check examples/all -vet -strict-style -target:openbsd_amd64
         timeout-minutes: 10
   build_macOS:
+    name: MacOS Build, Check, and Test
     runs-on: macos-latest
     steps:
       - uses: actions/checkout@v1
       - name: Download LLVM, botan and setup PATH
         run: |
-          brew install llvm@11 botan
-          echo "/usr/local/opt/llvm@11/bin" >> $GITHUB_PATH
+          brew install llvm@13 botan
+          echo "/usr/local/opt/llvm@13/bin" >> $GITHUB_PATH
           TMP_PATH=$(xcrun --show-sdk-path)/user/include
           echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
       - name: build odin
@@ -87,9 +92,45 @@ jobs:
           cd tests/core
           make
         timeout-minutes: 10
-      - name: Vendor library tests
+      - name: Odin internals tests
         run: |
-          cd tests/vendor
+          cd tests/internal
+          make
+        timeout-minutes: 10
+  build_macOS_arm:
+    name: MacOS ARM Build, Check, and Test
+    runs-on: macos-14 # This is an arm/m1 runner.
+    steps:
+      - uses: actions/checkout@v1
+      - name: Download LLVM, botan and setup PATH
+        run: |
+          brew install llvm@13 botan
+          echo "/opt/homebrew/opt/llvm@13/bin" >> $GITHUB_PATH
+          TMP_PATH=$(xcrun --show-sdk-path)/user/include
+          echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
+      - name: build odin
+        run: ./build_odin.sh 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 -vet
+        timeout-minutes: 10
+      - name: Odin run
+        run: ./odin run examples/demo
+        timeout-minutes: 10
+      - name: Odin run -debug
+        run: ./odin run examples/demo -debug
+        timeout-minutes: 10
+      - name: Odin check examples/all
+        run: ./odin check examples/all -strict-style
+        timeout-minutes: 10
+      - name: Core library tests
+        run: |
+          cd tests/core
           make
         timeout-minutes: 10
       - name: Odin internals tests
@@ -97,13 +138,8 @@ jobs:
           cd tests/internal
           make
         timeout-minutes: 10
-      - name: Odin check examples/all for Darwin arm64
-        run: ./odin check examples/all -vet -strict-style -target:darwin_arm64
-        timeout-minutes: 10
-      - name: Odin check examples/all for Linux arm64
-        run: ./odin check examples/all -vet -strict-style -target:linux_arm64
-        timeout-minutes: 10
   build_windows:
+    name: Windows Build, Check, and Test
     runs-on: windows-2022
     steps:
       - uses: actions/checkout@v1
@@ -168,7 +204,7 @@ jobs:
         run: |
           call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
           cd tests\documentation
-          call build.bat
+          rem call build.bat
         timeout-minutes: 10
       - name: core:math/big tests
         shell: cmd

.github/workflows/nightly.yml

@@ -7,6 +7,7 @@ on:
 
 jobs:
   build_windows:
+    name: Windows Build
     if: github.repository == 'odin-lang/Odin'
     runs-on: windows-2022
     steps:
@@ -26,8 +27,10 @@ jobs:
           rm bin/llvm/windows/LLVM-C.lib
           mkdir dist
           cp odin.exe dist
+          cp LICENSE dist
           cp LLVM-C.dll dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r bin dist
@@ -38,6 +41,7 @@ jobs:
           name: windows_artifacts
           path: dist
   build_ubuntu:
+    name: Ubuntu Build
     if: github.repository == 'odin-lang/Odin'
     runs-on: ubuntu-latest
     steps:
@@ -52,47 +56,95 @@ jobs:
         run: |
           mkdir dist
           cp odin dist
+          cp LICENSE dist
           cp libLLVM* dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r examples dist
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
       - name: Upload artifact
         uses: actions/upload-artifact@v1
         with:
           name: ubuntu_artifacts
-          path: dist
+          path: dist.zip
   build_macos:
+    name: MacOS Build
     if: github.repository == 'odin-lang/Odin'
-    runs-on: macOS-latest
+    runs-on: macos-latest
     steps:
       - uses: actions/checkout@v1
       - name: Download LLVM and setup PATH
         run: |
-          brew install llvm@11
-          echo "/usr/local/opt/llvm@11/bin" >> $GITHUB_PATH
+          brew install llvm@13 dylibbundler
+          echo "/usr/local/opt/llvm@13/bin" >> $GITHUB_PATH
           TMP_PATH=$(xcrun --show-sdk-path)/user/include
           echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
       - name: build odin
-        run: make nightly
-      - name: Odin run
-        run: ./odin run examples/demo
-      - name: Copy artifacts
+        # These -L makes the linker prioritize system libraries over LLVM libraries, this is mainly to
+        # not link with libunwind bundled with LLVM but link with libunwind on the system.
+        run: CXXFLAGS="-L/usr/lib/system -L/usr/lib" make nightly
+      - name: Bundle
         run: |
           mkdir dist
           cp odin dist
+          cp LICENSE dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r examples dist
+          dylibbundler -b -x dist/odin -d dist/libs -od -p @executable_path/libs
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
+      - name: Odin run
+        run: ./dist/odin run examples/demo
       - name: Upload artifact
         uses: actions/upload-artifact@v1
         with:
           name: macos_artifacts
-          path: dist
+          path: dist.zip
+  build_macos_arm:
+    name: MacOS ARM Build
+    if: github.repository == 'odin-lang/Odin'
+    runs-on: macos-14 # ARM machine
+    steps:
+      - uses: actions/checkout@v1
+      - name: Download LLVM and setup PATH
+        run: |
+          brew install llvm@13 dylibbundler
+          echo "/opt/homebrew/opt/llvm@13/bin" >> $GITHUB_PATH
+          TMP_PATH=$(xcrun --show-sdk-path)/user/include
+          echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
+      - name: build odin
+        # These -L makes the linker prioritize system libraries over LLVM libraries, this is mainly to
+        # not link with libunwind bundled with LLVM but link with libunwind on the system.
+        run: CXXFLAGS="-L/usr/lib/system -L/usr/lib" make nightly
+      - name: Bundle
+        run: |
+          mkdir dist
+          cp odin dist
+          cp LICENSE dist
+          cp -r shared dist
+          cp -r base dist
+          cp -r core dist
+          cp -r vendor dist
+          cp -r examples dist
+          dylibbundler -b -x dist/odin -d dist/libs -od -p @executable_path/libs
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
+      - name: Odin run
+        run: ./dist/odin run examples/demo
+      - name: Upload artifact
+        uses: actions/upload-artifact@v1
+        with:
+          name: macos_arm_artifacts
+          path: dist.zip
   upload_b2:
     runs-on: [ubuntu-latest]
-    needs: [build_windows, build_macos, build_ubuntu]
+    needs: [build_windows, build_macos, build_macos_arm, build_ubuntu]
     steps:
       - uses: actions/checkout@v1
       - uses: actions/setup-python@v2
@@ -123,6 +175,11 @@ jobs:
         with:
           name: macos_artifacts
 
+      - name: Download macOS arm artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: macos_arm_artifacts
+
       - name: Create archives and upload
         shell: bash
         env:
@@ -137,8 +194,9 @@ jobs:
           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 "$BUCKET" macos-amd64 macos_artifacts/
+          ./ci/upload_create_nightly.sh "$BUCKET" ubuntu-amd64 ubuntu_artifacts/dist.zip
+          ./ci/upload_create_nightly.sh "$BUCKET" macos-amd64 macos_artifacts/dist.zip
+          ./ci/upload_create_nightly.sh "$BUCKET" macos-arm64 macos_arm_artifacts/dist.zip
 
           echo Deleting old artifacts in B2
           python3 ci/delete_old_binaries.py "$BUCKET" "$DAYS_TO_KEEP"

.gitignore

@@ -24,6 +24,34 @@ bld/
 ![Cc]ore/[Ll]og/
 tests/documentation/verify/
 tests/documentation/all.odin-doc
+tests/internal/test_map
+tests/internal/test_pow
+tests/internal/test_rtti
+tests/core/test_core_compress
+tests/core/test_core_container
+tests/core/test_core_filepath
+tests/core/test_core_fmt
+tests/core/test_core_i18n
+tests/core/test_core_image
+tests/core/test_core_libc
+tests/core/test_core_match
+tests/core/test_core_math
+tests/core/test_core_net
+tests/core/test_core_os_exit
+tests/core/test_core_reflect
+tests/core/test_core_strings
+tests/core/test_crypto
+tests/core/test_hash
+tests/core/test_hxa
+tests/core/test_json
+tests/core/test_linalg_glsl_math
+tests/core/test_noise
+tests/core/test_varint
+tests/core/test_xml
+tests/core/test_core_slice
+tests/core/test_core_thread
+tests/core/test_core_runtime
+tests/vendor/vendor_botan
 # Visual Studio 2015 cache/options directory
 .vs/
 # Visual Studio Code options directory
@@ -270,6 +298,7 @@ bin/
 
 # - Linux/MacOS
 odin
+!odin/
 odin.dSYM
 *.bin
 demo.bin
@@ -286,4 +315,7 @@ shared/
 *.sublime-workspace
 examples/bug/
 build.sh
-!core/debug/
+!core/debug/
+
+# RAD debugger project file
+*.raddbg

base/builtin/builtin.odin

@@ -110,7 +110,7 @@ typeid_of    :: proc($T: typeid) -> typeid ---
 swizzle :: proc(x: [N]T, indices: ..int) -> [len(indices)]T ---
 
 complex    :: proc(real, imag: Float) -> Complex_Type ---
-quaternion :: proc(real, imag, jmag, kmag: Float) -> Quaternion_Type ---
+quaternion :: proc(imag, jmag, kmag, real: Float) -> Quaternion_Type --- // fields must be named
 real       :: proc(value: Complex_Or_Quaternion) -> Float ---
 imag       :: proc(value: Complex_Or_Quaternion) -> Float ---
 jmag       :: proc(value: Quaternion) -> Float ---

base/intrinsics/intrinsics.odin

@@ -5,6 +5,12 @@ package intrinsics
 // Package-Related
 is_package_imported :: proc(package_name: string) -> bool ---
 
+// Matrix Related Procedures
+transpose        :: proc(m: $T/matrix[$R, $C]$E)    -> matrix[C, R]E ---
+outer_product    :: proc(a: $A/[$X]$E, b: $B/[$Y]E) -> matrix[X, Y]E ---
+hadamard_product :: proc(a, b: $T/matrix[$R, $C]$E) -> T ---
+matrix_flatten   :: proc(m: $T/matrix[$R, $C]$E)    -> [R*C]E ---
+
 // Types
 soa_struct :: proc($N: int, $T: typeid) -> type/#soa[N]T
 
@@ -162,7 +168,14 @@ type_is_matrix           :: proc($T: typeid) -> bool ---
 type_has_nil :: proc($T: typeid) -> bool ---
 
 type_is_specialization_of :: proc($T, $S: typeid) -> bool ---
+
 type_is_variant_of :: proc($U, $V: typeid) -> bool where type_is_union(U) ---
+type_union_tag_type :: proc($T: typeid) -> typeid where type_is_union(T) ---
+type_union_tag_offset :: proc($T: typeid) -> uintptr where type_is_union(T) ---
+type_union_base_tag_value :: proc($T: typeid) -> int where type_is_union(U) ---
+type_union_variant_count :: proc($T: typeid) -> int where type_is_union(T) ---
+type_variant_type_of :: proc($T: typeid, $index: int) -> typeid where type_is_union(T) ---
+type_variant_index_of :: proc($U, $V: typeid) -> int where type_is_union(U) ---
 
 type_has_field :: proc($T: typeid, $name: string) -> bool ---
 type_field_type :: proc($T: typeid, $name: string) -> typeid ---
@@ -215,10 +228,10 @@ simd_shr_masked :: proc(a: #simd[N]T, b: #simd[N]Unsigned_Integer) -> #simd[N]T
 simd_add_sat :: proc(a, b: #simd[N]T) -> #simd[N]T ---
 simd_sub_sat :: proc(a, b: #simd[N]T) -> #simd[N]T ---
 
-simd_and     :: proc(a, b: #simd[N]T) -> #simd[N]T ---
-simd_or      :: proc(a, b: #simd[N]T) -> #simd[N]T ---
-simd_xor     :: proc(a, b: #simd[N]T) -> #simd[N]T ---
-simd_and_not :: proc(a, b: #simd[N]T) -> #simd[N]T ---
+simd_bit_and     :: proc(a, b: #simd[N]T) -> #simd[N]T ---
+simd_bit_or      :: proc(a, b: #simd[N]T) -> #simd[N]T ---
+simd_bit_xor     :: proc(a, b: #simd[N]T) -> #simd[N]T ---
+simd_bit_and_not :: proc(a, b: #simd[N]T) -> #simd[N]T ---
 
 simd_neg  :: proc(a: #simd[N]T) -> #simd[N]T ---
 

base/runtime/core.odin

@@ -18,9 +18,10 @@
 // This could change at a later date if the all these data structures are
 // implemented within the compiler rather than in this "preload" file
 //
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 // NOTE(bill): This must match the compiler's
 Calling_Convention :: enum u8 {
@@ -162,11 +163,11 @@ Type_Info_Simd_Vector :: struct {
 	count:      int,
 }
 Type_Info_Relative_Pointer :: struct {
-	pointer:      ^Type_Info,
+	pointer:      ^Type_Info, // ^T
 	base_integer: ^Type_Info,
 }
-Type_Info_Relative_Slice :: struct {
-	slice:        ^Type_Info,
+Type_Info_Relative_Multi_Pointer :: struct {
+	pointer:      ^Type_Info, // [^]T
 	base_integer: ^Type_Info,
 }
 Type_Info_Matrix :: struct {
@@ -180,6 +181,14 @@ Type_Info_Matrix :: struct {
 Type_Info_Soa_Pointer :: struct {
 	elem: ^Type_Info,
 }
+Type_Info_Bit_Field :: struct {
+	backing_type: ^Type_Info,
+	names:        []string,
+	types:        []^Type_Info,
+	bit_sizes:    []uintptr,
+	bit_offsets:  []uintptr,
+	tags:         []string,
+}
 
 Type_Info_Flag :: enum u8 {
 	Comparable     = 0,
@@ -219,9 +228,10 @@ Type_Info :: struct {
 		Type_Info_Bit_Set,
 		Type_Info_Simd_Vector,
 		Type_Info_Relative_Pointer,
-		Type_Info_Relative_Slice,
+		Type_Info_Relative_Multi_Pointer,
 		Type_Info_Matrix,
 		Type_Info_Soa_Pointer,
+		Type_Info_Bit_Field,
 	},
 }
 
@@ -252,12 +262,14 @@ Typeid_Kind :: enum u8 {
 	Bit_Set,
 	Simd_Vector,
 	Relative_Pointer,
-	Relative_Slice,
+	Relative_Multi_Pointer,
 	Matrix,
+	Soa_Pointer,
+	Bit_Field,
 }
 #assert(len(Typeid_Kind) < 32)
 
-// Typeid_Bit_Field :: bit_field #align align_of(uintptr) {
+// Typeid_Bit_Field :: bit_field #align(align_of(uintptr)) {
 // 	index:    8*size_of(uintptr) - 8,
 // 	kind:     5, // Typeid_Kind
 // 	named:    1,
@@ -268,7 +280,7 @@ Typeid_Kind :: enum u8 {
 
 // NOTE(bill): only the ones that are needed (not all types)
 // This will be set by the compiler
-type_table: []Type_Info
+type_table: []^Type_Info
 
 args__: []cstring
 
@@ -294,6 +306,14 @@ Source_Code_Location :: struct {
 	procedure:    string,
 }
 
+/*
+	Used by the built-in directory `#load_directory(path: string) -> []Load_Directory_File`
+*/
+Load_Directory_File :: struct {
+	name: string,
+	data: []byte, // immutable data
+}
+
 Assertion_Failure_Proc :: #type proc(prefix, message: string, loc: Source_Code_Location) -> !
 
 // Allocation Stuff
@@ -305,6 +325,7 @@ Allocator_Mode :: enum byte {
 	Query_Features,
 	Query_Info,
 	Alloc_Non_Zeroed,
+	Resize_Non_Zeroed,
 }
 
 Allocator_Mode_Set :: distinct bit_set[Allocator_Mode]
@@ -337,6 +358,8 @@ Kilobyte :: 1024 * Byte
 Megabyte :: 1024 * Kilobyte
 Gigabyte :: 1024 * Megabyte
 Terabyte :: 1024 * Gigabyte
+Petabyte :: 1024 * Terabyte
+Exabyte  :: 1024 * Petabyte
 
 // Logging stuff
 
@@ -499,6 +522,29 @@ Odin_Build_Mode_Type :: type_of(ODIN_BUILD_MODE)
 Odin_Endian_Type :: type_of(ODIN_ENDIAN)
 
 
+/*
+	// Defined internally by the compiler
+	Odin_Platform_Subtarget_Type :: enum int {
+		Default,
+		iOS,
+	}
+*/
+Odin_Platform_Subtarget_Type :: type_of(ODIN_PLATFORM_SUBTARGET)
+
+/*
+	// Defined internally by the compiler
+	Odin_Sanitizer_Flag :: enum u32 {
+		Address = 0,
+		Memory  = 1,
+		Thread  = 2,
+	}
+	Odin_Sanitizer_Flags :: distinct bitset[Odin_Sanitizer_Flag; u32]
+
+	ODIN_SANITIZER_FLAGS // is a constant
+*/
+Odin_Sanitizer_Flags :: type_of(ODIN_SANITIZER_FLAGS)
+
+
 /////////////////////////////
 // Init Startup Procedures //
 /////////////////////////////
@@ -563,7 +609,7 @@ __type_info_of :: proc "contextless" (id: typeid) -> ^Type_Info #no_bounds_check
 	if n < 0 || n >= len(type_table) {
 		n = 0
 	}
-	return &type_table[n]
+	return type_table[n]
 }
 
 when !ODIN_NO_RTTI {
@@ -638,8 +684,10 @@ default_assertion_failure_proc :: proc(prefix, message: string, loc: Source_Code
 	when ODIN_OS == .Freestanding {
 		// Do nothing
 	} else {
-		print_caller_location(loc)
-		print_string(" ")
+		when !ODIN_DISABLE_ASSERT {
+			print_caller_location(loc)
+			print_string(" ")
+		}
 		print_string(prefix)
 		if len(message) > 0 {
 			print_string(": ")

base/runtime/core_builtin.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @builtin
 Maybe :: union($T: typeid) {T}
@@ -109,7 +109,7 @@ remove_range :: proc(array: ^$D/[dynamic]$T, lo, hi: int, loc := #caller_locatio
 
 // `pop` will remove and return the end value of dynamic array `array` and reduces the length of `array` by 1.
 //
-// Note: If the dynamic array as no elements (`len(array) == 0`), this procedure will panic.
+// Note: If the dynamic array has no elements (`len(array) == 0`), this procedure will panic.
 @builtin
 pop :: proc(array: ^$T/[dynamic]$E, loc := #caller_location) -> (res: E) #no_bounds_check {
 	assert(len(array) > 0, loc=loc)
@@ -122,7 +122,7 @@ pop :: proc(array: ^$T/[dynamic]$E, loc := #caller_location) -> (res: E) #no_bou
 // `pop_safe` trys to remove and return the end value of dynamic array `array` and reduces the length of `array` by 1.
 // If the operation is not possible, it will return false.
 @builtin
-pop_safe :: proc(array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
+pop_safe :: proc "contextless" (array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
 	if len(array) == 0 {
 		return
 	}
@@ -148,7 +148,7 @@ pop_front :: proc(array: ^$T/[dynamic]$E, loc := #caller_location) -> (res: E) #
 // `pop_front_safe` trys to return and remove the first value of dynamic array `array` and reduces the length of `array` by 1.
 // If the operation is not possible, it will return false.
 @builtin
-pop_front_safe :: proc(array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
+pop_front_safe :: proc "contextless" (array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
 	if len(array) == 0 {
 		return
 	}
@@ -169,10 +169,16 @@ clear :: proc{clear_dynamic_array, clear_map}
 @builtin
 reserve :: proc{reserve_dynamic_array, reserve_map}
 
-// `resize` will try to resize memory of a passed dynamic array or map to the requested element count (setting the `len`, and possibly `cap`).
+@builtin
+non_zero_reserve :: proc{non_zero_reserve_dynamic_array}
+
+// `resize` will try to resize memory of a passed dynamic array to the requested element count (setting the `len`, and possibly `cap`).
 @builtin
 resize :: proc{resize_dynamic_array}
 
+@builtin
+non_zero_resize :: proc{non_zero_resize_dynamic_array}
+
 // Shrinks the capacity of a dynamic array or map down to the current length, or the given capacity.
 @builtin
 shrink :: proc{shrink_dynamic_array, shrink_map}
@@ -234,6 +240,8 @@ delete :: proc{
 	delete_dynamic_array,
 	delete_slice,
 	delete_map,
+	delete_soa_slice,
+	delete_soa_dynamic_array,
 }
 
 
@@ -304,6 +312,7 @@ make_dynamic_array_len :: proc($T: typeid/[dynamic]$E, #any_int len: int, alloca
 @(builtin, require_results)
 make_dynamic_array_len_cap :: proc($T: typeid/[dynamic]$E, #any_int len: int, #any_int cap: int, allocator := context.allocator, loc := #caller_location) -> (array: T, err: Allocator_Error) #optional_allocator_error {
 	make_dynamic_array_error_loc(loc, len, cap)
+	array.allocator = allocator // initialize allocator before just in case it fails to allocate any memory
 	data := mem_alloc_bytes(size_of(E)*cap, align_of(E), allocator, loc) or_return
 	s := Raw_Dynamic_Array{raw_data(data), len, cap, allocator}
 	if data == nil && size_of(E) != 0 {
@@ -346,7 +355,7 @@ make_multi_pointer :: proc($T: typeid/[^]$E, #any_int len: int, allocator := con
 //
 // Similar to `new`, the first argument is a type, not a value. Unlike new, make's return type is the same as the
 // type of its argument, not a pointer to it.
-// Make uses the specified allocator, default is context.allocator, default is context.allocator
+// Make uses the specified allocator, default is context.allocator.
 @builtin
 make :: proc{
 	make_slice,
@@ -404,10 +413,7 @@ delete_key :: proc(m: ^$T/map[$K]$V, key: K) -> (deleted_key: K, deleted_value:
 	return
 }
 
-
-
-@builtin
-append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+_append_elem :: #force_inline proc(array: ^$T/[dynamic]$E, arg: E, should_zero: bool, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
 	if array == nil {
 		return 0, nil
 	}
@@ -418,7 +424,13 @@ append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) ->
 	} else {
 		if cap(array) < len(array)+1 {
 			cap := 2 * cap(array) + max(8, 1)
-			err = reserve(array, cap, loc) // do not 'or_return' here as it could be a partial success
+
+			// do not 'or_return' here as it could be a partial success
+			if should_zero {
+				err = reserve(array, cap, loc)
+			} else {
+				err = non_zero_reserve(array, cap, loc) 
+			}
 		}
 		if cap(array)-len(array) > 0 {
 			a := (^Raw_Dynamic_Array)(array)
@@ -435,7 +447,16 @@ append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) ->
 }
 
 @builtin
-append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem(array, arg, true, loc=loc)
+}
+
+@builtin
+non_zero_append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem(array, arg, false, loc=loc)
+}
+
+_append_elems :: #force_inline proc(array: ^$T/[dynamic]$E, should_zero: bool, loc := #caller_location, args: ..E) -> (n: int, err: Allocator_Error) #optional_allocator_error {
 	if array == nil {
 		return 0, nil
 	}
@@ -452,7 +473,13 @@ append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location)
 	} else {
 		if cap(array) < len(array)+arg_len {
 			cap := 2 * cap(array) + max(8, arg_len)
-			err = reserve(array, cap, loc)  // do not 'or_return' here as it could be a partial success
+
+			// do not 'or_return' here as it could be a partial success
+			if should_zero {
+				err = reserve(array, cap, loc)
+			} else {
+				err = non_zero_reserve(array, cap, loc)
+			}
 		}
 		arg_len = min(cap(array)-len(array), arg_len)
 		if arg_len > 0 {
@@ -468,11 +495,33 @@ append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location)
 	}
 }
 
+@builtin
+append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elems(array, true, loc, ..args)
+}
+
+@builtin
+non_zero_append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elems(array, false, loc, ..args)
+}
+
 // The append_string built-in procedure appends a string to the end of a [dynamic]u8 like type
+_append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, should_zero: bool, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	args := transmute([]E)arg
+	if should_zero { 
+		return append_elems(array, ..args, loc=loc)
+	} else {
+		return non_zero_append_elems(array, ..args, loc=loc)
+	}
+}
+
 @builtin
 append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
-	args := transmute([]E)arg
-	return append_elems(array, ..args, loc=loc)
+	return _append_elem_string(array, arg, true, loc)
+}
+@builtin
+non_zero_append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem_string(array, arg, false, loc)
 }
 
 
@@ -492,6 +541,7 @@ append_string :: proc(array: ^$T/[dynamic]$E/u8, args: ..string, loc := #caller_
 
 // The append built-in procedure appends elements to the end of a dynamic array
 @builtin append :: proc{append_elem, append_elems, append_elem_string}
+@builtin non_zero_append :: proc{non_zero_append_elem, non_zero_append_elems, non_zero_append_elem_string}
 
 
 @builtin
@@ -587,11 +637,14 @@ assign_at_elem :: proc(array: ^$T/[dynamic]$E, index: int, arg: E, loc := #calle
 
 @builtin
 assign_at_elems :: proc(array: ^$T/[dynamic]$E, index: int, args: ..E, loc := #caller_location) -> (ok: bool, err: Allocator_Error) #no_bounds_check #optional_allocator_error {
-	if index+len(args) < len(array) {
+	new_size := index + len(args)
+	if len(args) == 0 {
+		ok = true
+	} else if new_size < len(array) {
 		copy(array[index:], args)
 		ok = true
 	} else {
-		resize(array, index+1+len(args), loc) or_return
+		resize(array, new_size, loc) or_return
 		copy(array[index:], args)
 		ok = true
 	}
@@ -601,14 +654,15 @@ assign_at_elems :: proc(array: ^$T/[dynamic]$E, index: int, args: ..E, loc := #c
 
 @builtin
 assign_at_elem_string :: proc(array: ^$T/[dynamic]$E/u8, index: int, arg: string, loc := #caller_location) -> (ok: bool, err: Allocator_Error) #no_bounds_check #optional_allocator_error {
-	if len(args) == 0 {
+	new_size := index + len(arg)
+	if len(arg) == 0 {
 		ok = true
-	} else if index+len(args) < len(array) {
-		copy(array[index:], args)
+	} else if new_size < len(array) {
+		copy(array[index:], arg)
 		ok = true
 	} else {
-		resize(array, index+1+len(args), loc) or_return
-		copy(array[index:], args)
+		resize(array, new_size, loc) or_return
+		copy(array[index:], arg)
 		ok = true
 	}
 	return
@@ -632,8 +686,7 @@ clear_dynamic_array :: proc "contextless" (array: ^$T/[dynamic]$E) {
 // `reserve_dynamic_array` will try to reserve memory of a passed dynamic array or map to the requested element count (setting the `cap`).
 //
 // Note: Prefer the procedure group `reserve`.
-@builtin
-reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+_reserve_dynamic_array :: #force_inline proc(array: ^$T/[dynamic]$E, capacity: int, should_zero: bool, loc := #caller_location) -> Allocator_Error {
 	if array == nil {
 		return nil
 	}
@@ -652,7 +705,12 @@ reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #cal
 	new_size  := capacity * size_of(E)
 	allocator := a.allocator
 
-	new_data := mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	new_data: []byte
+	if should_zero {
+		new_data = mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	} else {
+		new_data = non_zero_mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	}
 	if new_data == nil && new_size > 0 {
 		return .Out_Of_Memory
 	}
@@ -662,11 +720,20 @@ reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #cal
 	return nil
 }
 
+@builtin
+reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+	return _reserve_dynamic_array(array, capacity, true, loc)
+}
+
+@builtin
+non_zero_reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+	return _reserve_dynamic_array(array, capacity, false, loc)
+}
+
 // `resize_dynamic_array` will try to resize memory of a passed dynamic array or map to the requested element count (setting the `len`, and possibly `cap`).
 //
 // Note: Prefer the procedure group `resize`
-@builtin
-resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+_resize_dynamic_array :: #force_inline proc(array: ^$T/[dynamic]$E, length: int, should_zero: bool, loc := #caller_location) -> Allocator_Error {
 	if array == nil {
 		return nil
 	}
@@ -686,7 +753,12 @@ resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller
 	new_size  := length * size_of(E)
 	allocator := a.allocator
 
-	new_data := mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	new_data : []byte
+	if should_zero {
+		new_data = mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	} else {
+		new_data = non_zero_mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	}
 	if new_data == nil && new_size > 0 {
 		return .Out_Of_Memory
 	}
@@ -697,6 +769,16 @@ resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller
 	return nil
 }
 
+@builtin
+resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+	return _resize_dynamic_array(array, length, true, loc=loc)
+}
+
+@builtin
+non_zero_resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+	return _resize_dynamic_array(array, length, false, loc=loc)
+}
+
 /*
 	Shrinks the capacity of a dynamic array down to the current length, or the given capacity.
 
@@ -742,42 +824,23 @@ map_insert :: proc(m: ^$T/map[$K]$V, key: K, value: V, loc := #caller_location)
 	return (^V)(__dynamic_map_set_without_hash((^Raw_Map)(m), map_info(T), rawptr(&key), rawptr(&value), loc))
 }
 
-
-@builtin
-incl_elem :: proc(s: ^$S/bit_set[$E; $U], elem: E) {
-	s^ |= {elem}
+// Explicitly inserts a key and value into a map `m`, the same as `map_insert`, but the return values differ.
+// - `prev_key_ptr` will return the previous pointer of a key if it exists, and `nil` otherwise.
+// - `value_ptr` will return the pointer of the memory where the insertion happens, and `nil` if the map failed to resize
+// - `found_previous` will be true if `prev_key_ptr != nil`
+@(require_results)
+map_insert_and_check_for_previous :: proc(m: ^$T/map[$K]$V, key: K, value: V, loc := #caller_location) -> (prev_key_ptr: ^K, value_ptr: ^V, found_previous: bool) {
+	key, value := key, value
+	kp, vp := __dynamic_map_set_extra_without_hash((^Raw_Map)(m), map_info(T), rawptr(&key), rawptr(&value), loc)
+	prev_key_ptr   = (^K)(kp)
+	value_ptr      = (^V)(vp)
+	found_previous = kp != nil
+	return
 }
-@builtin
-incl_elems :: proc(s: ^$S/bit_set[$E; $U], elems: ..E) {
-	for elem in elems {
-		s^ |= {elem}
-	}
-}
-@builtin
-incl_bit_set :: proc(s: ^$S/bit_set[$E; $U], other: S) {
-	s^ |= other
-}
-@builtin
-excl_elem :: proc(s: ^$S/bit_set[$E; $U], elem: E) {
-	s^ &~= {elem}
-}
-@builtin
-excl_elems :: proc(s: ^$S/bit_set[$E; $U], elems: ..E) {
-	for elem in elems {
-		s^ &~= {elem}
-	}
-}
-@builtin
-excl_bit_set :: proc(s: ^$S/bit_set[$E; $U], other: S) {
-	s^ &~= other
-}
-
-@builtin incl :: proc{incl_elem, incl_elems, incl_bit_set}
-@builtin excl :: proc{excl_elem, excl_elems, excl_bit_set}
 
 
 @builtin
-card :: proc(s: $S/bit_set[$E; $U]) -> int {
+card :: proc "contextless" (s: $S/bit_set[$E; $U]) -> int {
 	when size_of(S) == 1 {
 		return int(intrinsics.count_ones(transmute(u8)s))
 	} else when size_of(S) == 2 {
@@ -816,7 +879,6 @@ assert :: proc(condition: bool, message := "", loc := #caller_location) {
 }
 
 @builtin
-@(disabled=ODIN_DISABLE_ASSERT)
 panic :: proc(message: string, loc := #caller_location) -> ! {
 	p := context.assertion_failure_proc
 	if p == nil {
@@ -826,7 +888,6 @@ panic :: proc(message: string, loc := #caller_location) -> ! {
 }
 
 @builtin
-@(disabled=ODIN_DISABLE_ASSERT)
 unimplemented :: proc(message := "", loc := #caller_location) -> ! {
 	p := context.assertion_failure_proc
 	if p == nil {

base/runtime/core_builtin_soa.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 _ :: intrinsics
 
 /*
@@ -287,7 +287,7 @@ append_soa_elem :: proc(array: ^$T/#soa[dynamic]$E, arg: E, loc := #caller_locat
 	footer := raw_soa_footer(array)
 
 	if size_of(E) > 0 && cap(array)-len(array) > 0 {
-		ti := type_info_of(typeid_of(T))
+		ti := type_info_of(T)
 		ti = type_info_base(ti)
 		si := &ti.variant.(Type_Info_Struct)
 		field_count: uintptr
@@ -413,3 +413,16 @@ delete_soa :: proc{
 	delete_soa_slice,
 	delete_soa_dynamic_array,
 }
+
+
+clear_soa_dynamic_array :: proc(array: ^$T/#soa[dynamic]$E) {
+	when intrinsics.type_struct_field_count(E) != 0 {
+		footer := raw_soa_footer(array)
+		footer.len = 0
+	}
+}
+
+@builtin
+clear_soa :: proc{
+	clear_soa_dynamic_array,
+}

base/runtime/default_allocators_arena.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE :: uint(DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE)
 
@@ -28,11 +28,11 @@ safe_add :: #force_inline proc "contextless" (x, y: uint) -> (uint, bool) {
 }
 
 @(require_results)
-memory_block_alloc :: proc(allocator: Allocator, capacity: uint, loc := #caller_location) -> (block: ^Memory_Block, err: Allocator_Error) {
-	total_size := uint(capacity + size_of(Memory_Block))
-	base_offset    := uintptr(size_of(Memory_Block))
+memory_block_alloc :: proc(allocator: Allocator, capacity: uint, alignment: uint, loc := #caller_location) -> (block: ^Memory_Block, err: Allocator_Error) {
+	total_size  := uint(capacity + max(alignment, size_of(Memory_Block)))
+	base_offset := uintptr(max(alignment, size_of(Memory_Block)))
 
-	min_alignment: int = max(16, align_of(Memory_Block))
+	min_alignment: int = max(16, align_of(Memory_Block), int(alignment))
 	data := mem_alloc(int(total_size), min_alignment, allocator, loc) or_return
 	block = (^Memory_Block)(raw_data(data))
 	end := uintptr(raw_data(data)[len(data):])
@@ -102,20 +102,20 @@ arena_alloc :: proc(arena: ^Arena, size, alignment: uint, loc := #caller_locatio
 	if size == 0 {
 		return
 	}
-
-	if arena.curr_block == nil || (safe_add(arena.curr_block.used, size) or_else 0) > arena.curr_block.capacity {
-		size = align_forward_uint(size, alignment)
+	
+	needed := align_forward_uint(size, alignment)
+	if arena.curr_block == nil || (safe_add(arena.curr_block.used, needed) or_else 0) > arena.curr_block.capacity {
 		if arena.minimum_block_size == 0 {
 			arena.minimum_block_size = DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE
 		}
 
-		block_size := max(size, arena.minimum_block_size)
+		block_size := max(needed, arena.minimum_block_size)
 
 		if arena.backing_allocator.procedure == nil {
 			arena.backing_allocator = default_allocator()
 		}
 
-		new_block := memory_block_alloc(arena.backing_allocator, block_size, loc) or_return
+		new_block := memory_block_alloc(arena.backing_allocator, block_size, alignment, loc) or_return
 		new_block.prev = arena.curr_block
 		arena.curr_block = new_block
 		arena.total_capacity += new_block.capacity
@@ -134,7 +134,7 @@ arena_init :: proc(arena: ^Arena, size: uint, backing_allocator: Allocator, loc
 	arena^ = {}
 	arena.backing_allocator = backing_allocator
 	arena.minimum_block_size = max(size, 1<<12) // minimum block size of 4 KiB
-	new_block := memory_block_alloc(arena.backing_allocator, arena.minimum_block_size, loc) or_return
+	new_block := memory_block_alloc(arena.backing_allocator, arena.minimum_block_size, 0, loc) or_return
 	arena.curr_block = new_block
 	arena.total_capacity += new_block.capacity
 	return nil
@@ -195,7 +195,7 @@ arena_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		err = .Mode_Not_Implemented
 	case .Free_All:
 		arena_free_all(arena, location)
-	case .Resize:
+	case .Resize, .Resize_Non_Zeroed:
 		old_data := ([^]byte)(old_memory)
 
 		switch {

base/runtime/default_allocators_general.odin

@@ -0,0 +1,12 @@
+package runtime
+
+when ODIN_DEFAULT_TO_NIL_ALLOCATOR {
+	default_allocator_proc :: nil_allocator_proc
+	default_allocator :: nil_allocator
+} else when ODIN_DEFAULT_TO_PANIC_ALLOCATOR {
+	default_allocator_proc :: panic_allocator_proc
+	default_allocator :: panic_allocator
+} else {
+	default_allocator :: heap_allocator
+	default_allocator_proc :: heap_allocator_proc
+}

base/runtime/default_allocators_nil.odin

@@ -10,7 +10,7 @@ nil_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		return nil, .None
 	case .Free_All:
 		return nil, .Mode_Not_Implemented
-	case .Resize:
+	case .Resize, .Resize_Non_Zeroed:
 		if size == 0 {
 			return nil, .None
 		}
@@ -31,14 +31,6 @@ nil_allocator :: proc() -> Allocator {
 }
 
 
-
-when ODIN_OS == .Freestanding {
-	default_allocator_proc :: nil_allocator_proc
-	default_allocator :: nil_allocator
-} 
-
-
-
 panic_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
                              size, alignment: int,
                              old_memory: rawptr, old_size: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
@@ -55,6 +47,10 @@ panic_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		if size > 0 {
 			panic("panic allocator, .Resize called", loc=loc)
 		}
+	case .Resize_Non_Zeroed:
+		if size > 0 {
+			panic("panic allocator, .Alloc_Non_Zeroed called", loc=loc)
+		}
 	case .Free:
 		if old_memory != nil {
 			panic("panic allocator, .Free called", loc=loc)
@@ -78,9 +74,7 @@ panic_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 
 panic_allocator :: proc() -> Allocator {
 	return Allocator{
-		procedure = nil_allocator_proc,
+		procedure = panic_allocator_proc,
 		data = nil,
 	}
 }
-
-

base/runtime/docs.odin

@@ -0,0 +1,180 @@
+package runtime
+
+/*
+
+package runtime has numerous entities (declarations) which are required by the compiler to function.
+
+
+## Basic types and calls (and anything they rely on)
+
+Source_Code_Location
+Context
+Allocator
+Logger
+
+__init_context
+_cleanup_runtime
+
+
+## cstring calls
+
+cstring_to_string
+cstring_len
+
+
+
+## Required when RTTI is enabled (the vast majority of targets)
+
+Type_Info
+
+type_table
+__type_info_of
+
+
+## Hashing
+
+default_hasher
+default_hasher_cstring
+default_hasher_string
+
+
+## Pseudo-CRT required procedured due to LLVM but useful in general
+memset
+memcpy
+memove
+
+
+## Procedures required by the LLVM backend if u128/i128 is used
+umodti3
+udivti3
+modti3
+divti3
+fixdfti
+fixunsdfti
+fixunsdfdi
+floattidf
+floattidf_unsigned
+truncsfhf2
+truncdfhf2
+gnu_h2f_ieee
+gnu_f2h_ieee
+extendhfsf2
+
+## Procedures required by the LLVM backend if f16 is used
+__ashlti3 // wasm specific
+__multi3  // wasm specific
+
+
+## Required an entry point is defined (i.e. 'main')
+
+args__
+
+
+## When -no-crt is defined (and not a wasm target) (mostly due to LLVM)
+_tls_index
+_fltused
+
+
+## Bounds checking procedures (when not disabled with -no-bounds-check)
+
+bounds_check_error
+matrix_bounds_check_error
+slice_expr_error_hi
+slice_expr_error_lo_hi
+multi_pointer_slice_expr_error
+
+
+## Type assertion check
+
+type_assertion_check
+type_assertion_check2 // takes in typeid
+
+
+## Arithmetic
+
+quo_complex32
+quo_complex64
+quo_complex128
+
+mul_quaternion64
+mul_quaternion128
+mul_quaternion256
+
+quo_quaternion64
+quo_quaternion128
+quo_quaternion256
+
+abs_complex32
+abs_complex64
+abs_complex128
+
+abs_quaternion64
+abs_quaternion128
+abs_quaternion256
+
+
+## Comparison
+
+memory_equal
+memory_compare
+memory_compare_zero
+
+cstring_eq
+cstring_ne
+cstring_lt
+cstring_gt
+cstring_le
+cstring_gt
+
+string_eq
+string_ne
+string_lt
+string_gt
+string_le
+string_gt
+
+complex32_eq
+complex32_ne
+complex64_eq
+complex64_ne
+complex128_eq
+complex128_ne
+
+quaternion64_eq
+quaternion64_ne
+quaternion128_eq
+quaternion128_ne
+quaternion256_eq
+quaternion256_ne
+
+
+## Map specific calls
+
+map_seed_from_map_data
+__dynamic_map_check_grow // static map calls
+map_insert_hash_dynamic  // static map calls
+__dynamic_map_get // dynamic map calls
+__dynamic_map_set // dynamic map calls
+
+
+## Dynamic literals ([dymamic]T and map[K]V) (can be disabled with -no-dynamic-literals)
+
+__dynamic_array_reserve
+__dynamic_array_append
+
+__dynamic_map_reserve
+
+
+## Objective-C specific
+
+objc_lookUpClass
+sel_registerName
+objc_allocateClassPair
+
+
+## for-in `string` type
+
+string_decode_rune
+string_decode_last_rune // #reverse for
+
+*/

base/runtime/dynamic_map_internal.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 _ :: intrinsics
 
 // High performance, cache-friendly, open-addressed Robin Hood hashing hash map
@@ -44,7 +44,7 @@ _ :: intrinsics
 MAP_LOAD_FACTOR :: 75
 
 // Minimum log2 capacity.
-MAP_MIN_LOG2_CAPACITY :: 6 // 64 elements
+MAP_MIN_LOG2_CAPACITY :: 3 // 8 elements
 
 // Has to be less than 100% though.
 #assert(MAP_LOAD_FACTOR < 100)
@@ -87,7 +87,7 @@ MAP_CACHE_LINE_SIZE :: 1 << MAP_CACHE_LINE_LOG2
 //
 // In the optimal case, len(Map_Cell(T){}.data) = 1 so the cell array can be treated
 // as a regular array of T, which is the case for hashes.
-Map_Cell :: struct($T: typeid) #align MAP_CACHE_LINE_SIZE {
+Map_Cell :: struct($T: typeid) #align(MAP_CACHE_LINE_SIZE) {
 	data: [MAP_CACHE_LINE_SIZE / size_of(T) when 0 < size_of(T) && size_of(T) < MAP_CACHE_LINE_SIZE else 1]T,
 }
 
@@ -414,68 +414,21 @@ map_insert_hash_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^
 	tk := map_cell_index_dynamic(sk, info.ks, 1)
 	tv := map_cell_index_dynamic(sv, info.vs, 1)
 
-	for {
-		hp := &hs[pos]
-		element_hash := hp^
+	swap_loop: for {
+		element_hash := hs[pos]
 
 		if map_hash_is_empty(element_hash) {
-			kp := map_cell_index_dynamic(ks, info.ks, pos)
-			vp := map_cell_index_dynamic(vs, info.vs, pos)
-			intrinsics.mem_copy_non_overlapping(rawptr(kp), rawptr(k), size_of_k)
-			intrinsics.mem_copy_non_overlapping(rawptr(vp), rawptr(v), size_of_v)
-			hp^ = h
+			k_dst := map_cell_index_dynamic(ks, info.ks, pos)
+			v_dst := map_cell_index_dynamic(vs, info.vs, pos)
+			intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v), size_of_v)
+			hs[pos] = h
 
-			return result if result != 0 else vp
+			return result if result != 0 else v_dst
 		}
 
 		if map_hash_is_deleted(element_hash) {
-			next_pos := (pos + 1) & mask
-
-			// backward shift
-			for !map_hash_is_empty(hs[next_pos]) {
-				probe_distance := map_probe_distance(m^, hs[next_pos], next_pos)
-				if probe_distance == 0 {
-					break
-				}
-				probe_distance -= 1
-
-				kp := map_cell_index_dynamic(ks, info.ks, pos)
-				vp := map_cell_index_dynamic(vs, info.vs, pos)
-				kn := map_cell_index_dynamic(ks, info.ks, next_pos)
-				vn := map_cell_index_dynamic(vs, info.vs, next_pos)
-
-				if distance > probe_distance {
-					if result == 0 {
-						result = vp
-					}
-					// move stored into pos; store next
-					intrinsics.mem_copy_non_overlapping(rawptr(kp), rawptr(k), size_of_k)
-					intrinsics.mem_copy_non_overlapping(rawptr(vp), rawptr(v), size_of_v)
-					hs[pos] = h
-
-					intrinsics.mem_copy_non_overlapping(rawptr(k), rawptr(kn), size_of_k)
-					intrinsics.mem_copy_non_overlapping(rawptr(v), rawptr(vn), size_of_v)
-					h = hs[next_pos]
-				} else {
-					// move next back 1
-					intrinsics.mem_copy_non_overlapping(rawptr(kp), rawptr(kn), size_of_k)
-					intrinsics.mem_copy_non_overlapping(rawptr(vp), rawptr(vn), size_of_v)
-					hs[pos] = hs[next_pos]
-					distance = probe_distance
-				}
-				hs[next_pos] = 0
-				pos = (pos + 1) & mask
-				next_pos = (next_pos + 1) & mask
-				distance += 1
-			}
-
-			kp := map_cell_index_dynamic(ks, info.ks, pos)
-			vp := map_cell_index_dynamic(vs, info.vs, pos)
-			intrinsics.mem_copy_non_overlapping(rawptr(kp), rawptr(k), size_of_k)
-			intrinsics.mem_copy_non_overlapping(rawptr(vp), rawptr(v), size_of_v)
-			hs[pos] = h
-
-			return result if result != 0 else vp
+			break swap_loop
 		}
 
 		if probe_distance := map_probe_distance(m^, element_hash, pos); distance > probe_distance {
@@ -495,8 +448,8 @@ map_insert_hash_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^
 			intrinsics.mem_copy_non_overlapping(rawptr(vp), rawptr(tv), size_of_v)
 
 			th := h
-			h = hp^
-			hp^ = th
+			h = hs[pos]
+			hs[pos] = th
 
 			distance = probe_distance
 		}
@@ -504,6 +457,103 @@ map_insert_hash_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^
 		pos = (pos + 1) & mask
 		distance += 1
 	}
+
+	// backward shift loop
+	hs[pos] = 0
+	look_ahead: uintptr = 1
+	for {
+		la_pos := (pos + look_ahead) & mask
+		element_hash := hs[la_pos]
+
+		if map_hash_is_deleted(element_hash) {
+			look_ahead += 1
+			hs[la_pos] = 0
+			continue
+		}
+
+		k_dst := map_cell_index_dynamic(ks, info.ks, pos)
+		v_dst := map_cell_index_dynamic(vs, info.vs, pos)
+
+		if map_hash_is_empty(element_hash) {
+			intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v), size_of_v)
+			hs[pos] = h
+
+			return result if result != 0 else v_dst
+		}
+
+		k_src := map_cell_index_dynamic(ks, info.ks, la_pos)
+		v_src := map_cell_index_dynamic(vs, info.vs, la_pos)
+		probe_distance := map_probe_distance(m^, element_hash, la_pos)
+
+		if probe_distance < look_ahead {
+			// probed can be made ideal while placing saved (ending condition)
+			if result == 0 {
+				result = v_dst
+			}
+			intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v), size_of_v)
+			hs[pos] = h
+
+			// This will be an ideal move
+			pos = (la_pos - probe_distance) & mask
+			look_ahead -= probe_distance
+
+			// shift until we hit ideal/empty
+			for probe_distance != 0 {
+				k_dst = map_cell_index_dynamic(ks, info.ks, pos)
+				v_dst = map_cell_index_dynamic(vs, info.vs, pos)
+
+				intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k_src), size_of_k)
+				intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v_src), size_of_v)
+				hs[pos] = element_hash
+				hs[la_pos] = 0
+
+				pos = (pos + 1) & mask
+				la_pos = (la_pos + 1) & mask
+				look_ahead = (la_pos - pos) & mask
+				element_hash = hs[la_pos]
+				if map_hash_is_empty(element_hash) {
+					return
+				}
+
+				probe_distance = map_probe_distance(m^, element_hash, la_pos)
+				if probe_distance == 0 {
+					return
+				}
+				// can be ideal?
+				if probe_distance < look_ahead {
+					pos = (la_pos - probe_distance) & mask
+				}
+				k_src = map_cell_index_dynamic(ks, info.ks, la_pos)
+				v_src = map_cell_index_dynamic(vs, info.vs, la_pos)
+			}
+			return
+		} else if distance < probe_distance - look_ahead {
+			// shift back probed
+			intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k_src), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v_src), size_of_v)
+			hs[pos] = element_hash
+			hs[la_pos] = 0
+		} else {
+			// place saved, save probed
+			if result == 0 {
+				result = v_dst
+			}
+			intrinsics.mem_copy_non_overlapping(rawptr(k_dst), rawptr(k), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v_dst), rawptr(v), size_of_v)
+			hs[pos] = h
+
+			intrinsics.mem_copy_non_overlapping(rawptr(k), rawptr(k_src), size_of_k)
+			intrinsics.mem_copy_non_overlapping(rawptr(v), rawptr(v_src), size_of_v)
+			h = hs[la_pos]
+			hs[la_pos] = 0
+			distance = probe_distance - look_ahead
+		}
+
+		pos = (pos + 1) & mask
+		distance += 1
+	}
 }
 
 @(require_results)
@@ -579,7 +629,7 @@ map_reserve_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_
 
 
 @(require_results)
-map_shrink_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, loc := #caller_location) -> Allocator_Error {
+map_shrink_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, loc := #caller_location) -> (did_shrink: bool, err: Allocator_Error) {
 	if m.allocator.procedure == nil {
 		m.allocator = context.allocator
 	}
@@ -589,7 +639,7 @@ map_shrink_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_I
 	// map needs to be within the max load factor.
 	log2_capacity := map_log2_cap(m^)
 	if uintptr(m.len) >= map_load_factor(log2_capacity - 1) {
-		return nil
+		return false, nil
 	}
 
 	shrunk := map_alloc_dynamic(info, log2_capacity - 1, m.allocator) or_return
@@ -622,7 +672,7 @@ map_shrink_dynamic :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_I
 
 	map_free_dynamic(m^, info, loc) or_return
 	m.data = shrunk.data
-	return nil
+	return true, nil
 }
 
 @(require_results)
@@ -696,49 +746,19 @@ map_erase_dynamic :: #force_inline proc "contextless" (#no_alias m: ^Raw_Map, #n
 	m.len -= 1
 	ok = true
 
-	{ // coalesce tombstones
-		// HACK NOTE(bill): This is an ugly bodge but it is coalescing the tombstone slots
-		mask := (uintptr(1)<<map_log2_cap(m^)) - 1
-		curr_index := uintptr(index)
-
-		// TODO(bill): determine a good value for this empirically
-		// if we do not implement backward shift deletion
-		PROBE_COUNT :: 8
-		for _ in 0..<PROBE_COUNT {
-			next_index := (curr_index + 1) & mask
-			if next_index == index {
-				// looped around
-				break
-			}
-
-			// if the next element is empty or has zero probe distance, then any lookup
-			// will always fail on the next, so we can clear both of them
-			hash := hs[next_index]
-			if map_hash_is_empty(hash) || map_probe_distance(m^, hash, next_index) == 0 {
-				hs[curr_index] = 0
-				return
-			}
-
-			// now the next element will have a probe count of at least one,
-			// so it can use the delete slot instead
-			hs[curr_index] = hs[next_index]
-
-			mem_copy_non_overlapping(
-				rawptr(map_cell_index_dynamic(ks, info.ks, curr_index)),
-				rawptr(map_cell_index_dynamic(ks, info.ks, next_index)),
-				int(info.ks.size_of_type),
-			)
-			mem_copy_non_overlapping(
-				rawptr(map_cell_index_dynamic(vs, info.vs, curr_index)),
-				rawptr(map_cell_index_dynamic(vs, info.vs, next_index)),
-				int(info.vs.size_of_type),
-			)
-
-			curr_index = next_index
-		}
+	mask := (uintptr(1)<<map_log2_cap(m^)) - 1
+	curr_index := uintptr(index)
+	next_index := (curr_index + 1) & mask
 
+	// if the next element is empty or has zero probe distance, then any lookup
+	// will always fail on the next, so we can clear both of them
+	hash := hs[next_index]
+	if map_hash_is_empty(hash) || map_probe_distance(m^, hash, next_index) == 0 {
+		hs[curr_index] = 0
+	} else {
 		hs[curr_index] |= TOMBSTONE_MASK
 	}
+
 	return
 }
 
@@ -821,6 +841,33 @@ __dynamic_map_get :: proc "contextless" (#no_alias m: ^Raw_Map, #no_alias info:
 	}
 }
 
+__dynamic_map_get_key_and_value :: proc "contextless" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, h: Map_Hash, key: rawptr) -> (key_ptr, value_ptr: rawptr) {
+	if m.len == 0 {
+		return nil, nil
+	}
+	pos := map_desired_position(m^, h)
+	distance := uintptr(0)
+	mask := (uintptr(1) << map_log2_cap(m^)) - 1
+	ks, vs, hs, _, _ := map_kvh_data_dynamic(m^, info)
+	for {
+		element_hash := hs[pos]
+		if map_hash_is_empty(element_hash) {
+			return nil, nil
+		} else if distance > map_probe_distance(m^, element_hash, pos) {
+			return nil, nil
+		} else if element_hash == h {
+			other_key := rawptr(map_cell_index_dynamic(ks, info.ks, pos))
+			if info.key_equal(key, other_key) {
+				key_ptr   = other_key
+				value_ptr = rawptr(map_cell_index_dynamic(vs, info.vs, pos))
+				return
+			}
+		}
+		pos = (pos + 1) & mask
+		distance += 1
+	}
+}
+
 // IMPORTANT: USED WITHIN THE COMPILER
 __dynamic_map_check_grow :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, loc := #caller_location) -> (err: Allocator_Error, has_grown: bool) {
 	if m.len >= map_resize_threshold(m^) {
@@ -854,6 +901,30 @@ __dynamic_map_set :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_In
 	m.len += 1
 	return rawptr(result)
 }
+__dynamic_map_set_extra_without_hash :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, key, value: rawptr, loc := #caller_location) -> (prev_key_ptr, value_ptr: rawptr) {
+	return __dynamic_map_set_extra(m, info, info.key_hasher(key, map_seed(m^)), key, value, loc)
+}
+
+__dynamic_map_set_extra :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, hash: Map_Hash, key, value: rawptr, loc := #caller_location) -> (prev_key_ptr, value_ptr: rawptr) {
+	if prev_key_ptr, value_ptr = __dynamic_map_get_key_and_value(m, info, hash, key); value_ptr != nil {
+		intrinsics.mem_copy_non_overlapping(value_ptr, value, info.vs.size_of_type)
+		return
+	}
+
+	hash := hash
+	err, has_grown := __dynamic_map_check_grow(m, info, loc)
+	if err != nil {
+		return nil, nil
+	}
+	if has_grown {
+		hash = info.key_hasher(key, map_seed(m^))
+	}
+
+	result := map_insert_hash_dynamic(m, info, hash, uintptr(key), uintptr(value))
+	m.len += 1
+	return nil, rawptr(result)
+}
+
 
 // IMPORTANT: USED WITHIN THE COMPILER
 @(private)

base/runtime/entry_unix.odin

@@ -0,0 +1,59 @@
+//+private
+//+build linux, darwin, freebsd, openbsd
+//+no-instrumentation
+package runtime
+
+import "base:intrinsics"
+
+when ODIN_BUILD_MODE == .Dynamic {
+	@(link_name="_odin_entry_point", linkage="strong", require/*, link_section=".init"*/)
+	_odin_entry_point :: proc "c" () {
+		context = default_context()
+		#force_no_inline _startup_runtime()
+		intrinsics.__entry_point()
+	}
+	@(link_name="_odin_exit_point", linkage="strong", require/*, link_section=".fini"*/)
+	_odin_exit_point :: proc "c" () {
+		context = default_context()
+		#force_no_inline _cleanup_runtime()
+	}
+	@(link_name="main", linkage="strong", require)
+	main :: proc "c" (argc: i32, argv: [^]cstring) -> i32 {
+		return 0
+	}
+} else when !ODIN_TEST && !ODIN_NO_ENTRY_POINT {
+	when ODIN_NO_CRT {
+		// NOTE(flysand): We need to start from assembly because we need
+		// to retrieve argc and argv from the stack
+		when ODIN_ARCH == .amd64 {
+			@require foreign import entry "entry_unix_no_crt_amd64.asm"
+			SYS_exit :: 60
+		} else when ODIN_ARCH == .i386 {
+			@require foreign import entry "entry_unix_no_crt_i386.asm"
+			SYS_exit :: 1
+		} else when ODIN_OS == .Darwin && ODIN_ARCH == .arm64 {
+			@require foreign import entry "entry_unix_no_crt_darwin_arm64.asm"
+			SYS_exit :: 1
+		}
+		@(link_name="_start_odin", linkage="strong", require)
+		_start_odin :: proc "c" (argc: i32, argv: [^]cstring) -> ! {
+			args__ = argv[:argc]
+			context = default_context()
+			#force_no_inline _startup_runtime()
+			intrinsics.__entry_point()
+			#force_no_inline _cleanup_runtime()
+			intrinsics.syscall(SYS_exit, 0)
+			unreachable()
+		}
+	} else {
+		@(link_name="main", linkage="strong", require)
+		main :: proc "c" (argc: i32, argv: [^]cstring) -> i32 {
+			args__ = argv[:argc]
+			context = default_context()
+			#force_no_inline _startup_runtime()
+			intrinsics.__entry_point()
+			#force_no_inline _cleanup_runtime()
+			return 0
+		}
+	}
+}

base/runtime/entry_unix_no_crt_amd64.asm

@@ -0,0 +1,43 @@
+bits 64
+
+extern _start_odin
+global _start
+
+section .text
+
+;; Entry point for programs that specify -no-crt option
+;; This entry point should be compatible with dynamic loaders on linux
+;; The parameters the dynamic loader passes to the _start function:
+;;    RDX = pointer to atexit function
+;; The stack layout is as follows:
+;;    +-------------------+
+;;            NULL
+;;    +-------------------+
+;;           envp[m]
+;;    +-------------------+
+;;            ...
+;;    +-------------------+
+;;           envp[0]
+;;    +-------------------+
+;;            NULL
+;;    +-------------------+
+;;           argv[n]
+;;    +-------------------+
+;;            ...
+;;    +-------------------+
+;;           argv[0]
+;;    +-------------------+
+;;            argc
+;;    +-------------------+ <------ RSP
+;;
+_start:
+    ;; Mark stack frame as the top of the stack
+    xor rbp, rbp
+    ;; Load argc into 1st param reg, argv into 2nd param reg
+    pop rdi
+    mov rdx, rsi
+    ;; Align stack pointer down to 16-bytes (sysv calling convention)
+    and rsp, -16
+    ;; Call into odin entry point
+    call _start_odin
+    jmp $$

base/runtime/entry_unix_no_crt_darwin_arm64.asm

@@ -0,0 +1,20 @@
+	.section __TEXT,__text
+
+	; NOTE(laytan): this should ideally be the -minimum-os-version flag but there is no nice way of preprocessing assembly in Odin.
+	; 10 seems to be the lowest it goes and I don't see it mess with any targeted os version so this seems fine.
+	.build_version macos, 10, 0
+
+	.extern __start_odin
+
+	.global _main
+	.align 2
+_main:
+	mov x5, sp       ; use x5 as the stack pointer
+
+	str x0, [x5]     ; get argc into x0 (kernel passes 32-bit int argc as 64-bits on stack to keep alignment)
+	str x1, [x5, #8] ; get argv into x1
+
+	and sp, x5, #~15 ; force 16-byte alignment of the stack
+	
+	bl __start_odin  ; call into Odin entry point
+	ret              ; should never get here

base/runtime/entry_unix_no_crt_i386.asm

@@ -0,0 +1,18 @@
+bits 32
+
+extern _start_odin
+global _start
+
+section .text
+
+;; NOTE(flysand): For description see the corresponding *_amd64.asm file
+;; also I didn't test this on x86-32
+_start:
+    xor ebp, rbp
+    pop ecx
+    mov eax, esp
+    and esp, -16
+    push eax
+    push ecx
+    call _start_odin
+    jmp $$

base/runtime/entry_wasm.odin

@@ -1,8 +1,9 @@
 //+private
 //+build wasm32, wasm64p32
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when !ODIN_TEST && !ODIN_NO_ENTRY_POINT {
 	@(link_name="_start", linkage="strong", require, export)

base/runtime/entry_windows.odin

@@ -1,12 +1,13 @@
 //+private
 //+build windows
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when ODIN_BUILD_MODE == .Dynamic {
 	@(link_name="DllMain", linkage="strong", require)
-	DllMain :: proc "stdcall" (hinstDLL: rawptr, fdwReason: u32, lpReserved: rawptr) -> b32 {
+	DllMain :: proc "system" (hinstDLL: rawptr, fdwReason: u32, lpReserved: rawptr) -> b32 {
 		context = default_context()
 
 		// Populate Windows DLL-specific global
@@ -28,7 +29,7 @@ when ODIN_BUILD_MODE == .Dynamic {
 } else when !ODIN_TEST && !ODIN_NO_ENTRY_POINT {
 	when ODIN_ARCH == .i386 || ODIN_NO_CRT {
 		@(link_name="mainCRTStartup", linkage="strong", require)
-		mainCRTStartup :: proc "stdcall" () -> i32 {
+		mainCRTStartup :: proc "system" () -> i32 {
 			context = default_context()
 			#force_no_inline _startup_runtime()
 			intrinsics.__entry_point()

base/runtime/error_checks.odin

@@ -1,5 +1,6 @@
 package runtime
 
+@(no_instrumentation)
 bounds_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_array_bounds()
@@ -8,6 +9,7 @@ bounds_trap :: proc "contextless" () -> ! {
 	}
 }
 
+@(no_instrumentation)
 type_assertion_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_type_assertion()
@@ -21,7 +23,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	if uint(index) < uint(count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, index, count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Index ")
@@ -34,6 +36,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	handle_error(file, line, column, index, count)
 }
 
+@(no_instrumentation)
 slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int, len: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -46,6 +49,7 @@ slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, h
 	bounds_trap()
 }
 
+@(no_instrumentation)
 multi_pointer_slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -82,7 +86,7 @@ dynamic_array_expr_error :: proc "contextless" (file: string, line, column: i32,
 	if 0 <= low && low <= high && high <= max {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, low, high, max: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Invalid dynamic array indices ")
@@ -103,7 +107,7 @@ matrix_bounds_check_error :: proc "contextless" (file: string, line, column: i32
 	   uint(column_index) < uint(column_count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, row_index, column_index, row_count, column_count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Matrix indices [")
@@ -127,7 +131,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -140,7 +144,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -153,7 +157,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion from ")
@@ -198,7 +202,7 @@ when ODIN_NO_RTTI {
 			return id
 		}
 
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid, from_data: rawptr) -> ! {
 
 			actual := variant_type(from, from_data)
@@ -224,7 +228,7 @@ make_slice_error_loc :: #force_inline proc "contextless" (loc := #caller_locatio
 	if 0 <= len {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len: int) -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid slice length for make: ")
@@ -235,11 +239,11 @@ make_slice_error_loc :: #force_inline proc "contextless" (loc := #caller_locatio
 	handle_error(loc, len)
 }
 
-make_dynamic_array_error_loc :: #force_inline proc "contextless" (using loc := #caller_location, len, cap: int) {
+make_dynamic_array_error_loc :: #force_inline proc "contextless" (loc := #caller_location, len, cap: int) {
 	if 0 <= len && len <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid dynamic array parameters for make: ")
@@ -256,7 +260,7 @@ make_map_expr_error_loc :: #force_inline proc "contextless" (loc := #caller_loca
 	if 0 <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid map capacity for make: ")
@@ -271,18 +275,18 @@ make_map_expr_error_loc :: #force_inline proc "contextless" (loc := #caller_loca
 
 
 
-bounds_check_error_loc :: #force_inline proc "contextless" (using loc := #caller_location, index, count: int) {
-	bounds_check_error(file_path, line, column, index, count)
+bounds_check_error_loc :: #force_inline proc "contextless" (loc := #caller_location, index, count: int) {
+	bounds_check_error(loc.file_path, loc.line, loc.column, index, count)
 }
 
-slice_expr_error_hi_loc :: #force_inline proc "contextless" (using loc := #caller_location, hi: int, len: int) {
-	slice_expr_error_hi(file_path, line, column, hi, len)
+slice_expr_error_hi_loc :: #force_inline proc "contextless" (loc := #caller_location, hi: int, len: int) {
+	slice_expr_error_hi(loc.file_path, loc.line, loc.column, hi, len)
 }
 
-slice_expr_error_lo_hi_loc :: #force_inline proc "contextless" (using loc := #caller_location, lo, hi: int, len: int) {
-	slice_expr_error_lo_hi(file_path, line, column, lo, hi, len)
+slice_expr_error_lo_hi_loc :: #force_inline proc "contextless" (loc := #caller_location, lo, hi: int, len: int) {
+	slice_expr_error_lo_hi(loc.file_path, loc.line, loc.column, lo, hi, len)
 }
 
-dynamic_array_expr_error_loc :: #force_inline proc "contextless" (using loc := #caller_location, low, high, max: int) {
-	dynamic_array_expr_error(file_path, line, column, low, high, max)
+dynamic_array_expr_error_loc :: #force_inline proc "contextless" (loc := #caller_location, low, high, max: int) {
+	dynamic_array_expr_error(loc.file_path, loc.line, loc.column, low, high, max)
 }

base/runtime/heap_allocator.odin

@@ -0,0 +1,110 @@
+package runtime
+
+import "base:intrinsics"
+
+heap_allocator :: proc() -> Allocator {
+	return Allocator{
+		procedure = heap_allocator_proc,
+		data = nil,
+	}
+}
+
+heap_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
+                            size, alignment: int,
+                            old_memory: rawptr, old_size: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
+	//
+	// NOTE(tetra, 2020-01-14): The heap doesn't respect alignment.
+	// Instead, we overallocate by `alignment + size_of(rawptr) - 1`, and insert
+	// padding. We also store the original pointer returned by heap_alloc right before
+	// the pointer we return to the user.
+	//
+
+	aligned_alloc :: proc(size, alignment: int, old_ptr: rawptr = nil, zero_memory := true) -> ([]byte, Allocator_Error) {
+		a := max(alignment, align_of(rawptr))
+		space := size + a - 1
+
+		allocated_mem: rawptr
+		if old_ptr != nil {
+			original_old_ptr := ([^]rawptr)(old_ptr)[-1]
+			allocated_mem = heap_resize(original_old_ptr, space+size_of(rawptr))
+		} else {
+			allocated_mem = heap_alloc(space+size_of(rawptr), zero_memory)
+		}
+		aligned_mem := rawptr(([^]u8)(allocated_mem)[size_of(rawptr):])
+
+		ptr := uintptr(aligned_mem)
+		aligned_ptr := (ptr - 1 + uintptr(a)) & -uintptr(a)
+		diff := int(aligned_ptr - ptr)
+		if (size + diff) > space || allocated_mem == nil {
+			return nil, .Out_Of_Memory
+		}
+
+		aligned_mem = rawptr(aligned_ptr)
+		([^]rawptr)(aligned_mem)[-1] = allocated_mem
+
+		return byte_slice(aligned_mem, size), nil
+	}
+
+	aligned_free :: proc(p: rawptr) {
+		if p != nil {
+			heap_free(([^]rawptr)(p)[-1])
+		}
+	}
+
+	aligned_resize :: proc(p: rawptr, old_size: int, new_size: int, new_alignment: int, zero_memory := true) -> (new_memory: []byte, err: Allocator_Error) {
+		if p == nil {
+			return nil, nil
+		}
+
+		new_memory = aligned_alloc(new_size, new_alignment, p, zero_memory) or_return
+
+		// NOTE: heap_resize does not zero the new memory, so we do it
+		if zero_memory && new_size > old_size {
+			new_region := raw_data(new_memory[old_size:])
+			intrinsics.mem_zero(new_region, new_size - old_size)
+		}
+		return
+	}
+
+	switch mode {
+	case .Alloc, .Alloc_Non_Zeroed:
+		return aligned_alloc(size, alignment, nil, mode == .Alloc)
+
+	case .Free:
+		aligned_free(old_memory)
+
+	case .Free_All:
+		return nil, .Mode_Not_Implemented
+
+	case .Resize, .Resize_Non_Zeroed:
+		if old_memory == nil {
+			return aligned_alloc(size, alignment, nil, mode == .Resize)
+		}
+		return aligned_resize(old_memory, old_size, size, alignment, mode == .Resize)
+
+	case .Query_Features:
+		set := (^Allocator_Mode_Set)(old_memory)
+		if set != nil {
+			set^ = {.Alloc, .Alloc_Non_Zeroed, .Free, .Resize, .Resize_Non_Zeroed, .Query_Features}
+		}
+		return nil, nil
+
+	case .Query_Info:
+		return nil, .Mode_Not_Implemented
+	}
+
+	return nil, nil
+}
+
+
+heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	return _heap_alloc(size, zero_memory)
+}
+
+heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	return _heap_resize(ptr, new_size)
+}
+
+heap_free :: proc(ptr: rawptr) {
+	_heap_free(ptr)
+}

base/runtime/heap_allocator_other.odin

@@ -0,0 +1,15 @@
+//+build js, wasi, freestanding, essence
+//+private
+package runtime
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	unimplemented("base:runtime 'heap_alloc' procedure is not supported on this platform")
+}
+
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	unimplemented("base:runtime 'heap_resize' procedure is not supported on this platform")
+}
+
+_heap_free :: proc(ptr: rawptr) {
+	unimplemented("base:runtime 'heap_free' procedure is not supported on this platform")
+}

base/runtime/heap_allocator_unix.odin

@@ -0,0 +1,38 @@
+//+build linux, darwin, freebsd, openbsd
+//+private
+package runtime
+
+when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
+} else {
+	foreign import libc "system:c"
+}
+
+@(default_calling_convention="c")
+foreign libc {
+	@(link_name="malloc")   _unix_malloc   :: proc(size: int) -> rawptr ---
+	@(link_name="calloc")   _unix_calloc   :: proc(num, size: int) -> rawptr ---
+	@(link_name="free")     _unix_free     :: proc(ptr: rawptr) ---
+	@(link_name="realloc")  _unix_realloc  :: proc(ptr: rawptr, size: int) -> rawptr ---
+}
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	if size <= 0 {
+		return nil
+	}
+	if zero_memory {
+		return _unix_calloc(1, size)
+	} else {
+		return _unix_malloc(size)
+	}
+}
+
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	// NOTE: _unix_realloc doesn't guarantee new memory will be zeroed on
+	// POSIX platforms. Ensure your caller takes this into account.
+	return _unix_realloc(ptr, new_size)
+}
+
+_heap_free :: proc(ptr: rawptr) {
+	_unix_free(ptr)
+}

base/runtime/heap_allocator_windows.odin

@@ -0,0 +1,39 @@
+package runtime
+
+foreign import kernel32 "system:Kernel32.lib"
+
+@(private="file")
+@(default_calling_convention="system")
+foreign kernel32 {
+	// NOTE(bill): The types are not using the standard names (e.g. DWORD and LPVOID) to just minimizing the dependency
+
+	// default_allocator
+	GetProcessHeap :: proc() -> rawptr ---
+	HeapAlloc      :: proc(hHeap: rawptr, dwFlags: u32, dwBytes: uint) -> rawptr ---
+	HeapReAlloc    :: proc(hHeap: rawptr, dwFlags: u32, lpMem: rawptr, dwBytes: uint) -> rawptr ---
+	HeapFree       :: proc(hHeap: rawptr, dwFlags: u32, lpMem: rawptr) -> b32 ---
+}
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	HEAP_ZERO_MEMORY :: 0x00000008
+	return HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY if zero_memory else 0, uint(size))
+}
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	if new_size == 0 {
+		_heap_free(ptr)
+		return nil
+	}
+	if ptr == nil {
+		return _heap_alloc(new_size)
+	}
+
+	HEAP_ZERO_MEMORY :: 0x00000008
+	return HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, ptr, uint(new_size))
+}
+_heap_free :: proc(ptr: rawptr) {
+	if ptr == nil {
+		return
+	}
+	HeapFree(GetProcessHeap(), 0, ptr)
+}
+

base/runtime/internal.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @(private="file")
 IS_WASM :: ODIN_ARCH == .wasm32 || ODIN_ARCH == .wasm64p32
@@ -11,7 +11,10 @@ RUNTIME_LINKAGE :: "strong" when (
 	ODIN_BUILD_MODE == .Dynamic ||
 	!ODIN_NO_CRT) &&
 	!IS_WASM) else "internal"
-RUNTIME_REQUIRE :: true
+RUNTIME_REQUIRE :: false // !ODIN_TILDE
+
+@(private)
+__float16 :: f16 when __ODIN_LLVM_F16_SUPPORTED else u16
 
 
 @(private)
@@ -19,51 +22,7 @@ byte_slice :: #force_inline proc "contextless" (data: rawptr, len: int) -> []byt
 	return ([^]byte)(data)[:max(len, 0)]
 }
 
-bswap_16 :: proc "contextless" (x: u16) -> u16 {
-	return x>>8 | x<<8
-}
-
-bswap_32 :: proc "contextless" (x: u32) -> u32 {
-	return x>>24 | (x>>8)&0xff00 | (x<<8)&0xff0000 | x<<24
-}
-
-bswap_64 :: proc "contextless" (x: u64) -> u64 {
-	z := x
-	z = (z & 0x00000000ffffffff) << 32 | (z & 0xffffffff00000000) >> 32
-	z = (z & 0x0000ffff0000ffff) << 16 | (z & 0xffff0000ffff0000) >> 16
-	z = (z & 0x00ff00ff00ff00ff) << 8  | (z & 0xff00ff00ff00ff00) >> 8
-	return z
-}
-
-bswap_128 :: proc "contextless" (x: u128) -> u128 {
-	z := transmute([4]u32)x
-	z[0], z[3] = bswap_32(z[3]), bswap_32(z[0])
-	z[1], z[2] = bswap_32(z[2]), bswap_32(z[1])
-	return transmute(u128)z
-}
-
-bswap_f16 :: proc "contextless" (f: f16) -> f16 {
-	x := transmute(u16)f
-	z := bswap_16(x)
-	return transmute(f16)z
-
-}
-
-bswap_f32 :: proc "contextless" (f: f32) -> f32 {
-	x := transmute(u32)f
-	z := bswap_32(x)
-	return transmute(f32)z
-
-}
-
-bswap_f64 :: proc "contextless" (f: f64) -> f64 {
-	x := transmute(u64)f
-	z := bswap_64(x)
-	return transmute(f64)z
-}
-
-
-is_power_of_two_int :: #force_inline proc(x: int) -> bool {
+is_power_of_two_int :: #force_inline proc "contextless" (x: int) -> bool {
 	if x <= 0 {
 		return false
 	}
@@ -81,7 +40,7 @@ align_forward_int :: #force_inline proc(ptr, align: int) -> int {
 	return p
 }
 
-is_power_of_two_uintptr :: #force_inline proc(x: uintptr) -> bool {
+is_power_of_two_uintptr :: #force_inline proc "contextless" (x: uintptr) -> bool {
 	if x <= 0 {
 		return false
 	}
@@ -184,7 +143,7 @@ mem_free_all :: #force_inline proc(allocator := context.allocator, loc := #calle
 	return
 }
 
-mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+_mem_resize :: #force_inline proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, should_zero: bool, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
 	if allocator.procedure == nil {
 		return nil, nil
 	}
@@ -195,15 +154,27 @@ mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAUL
 		}
 		return
 	} else if ptr == nil {
-		return allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		if should_zero {
+			return allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		} else {
+			return allocator.procedure(allocator.data, .Alloc_Non_Zeroed, new_size, alignment, nil, 0, loc)
+		}
 	} else if old_size == new_size && uintptr(ptr) % uintptr(alignment) == 0 {
 		data = ([^]byte)(ptr)[:old_size]
 		return
 	}
 
-	data, err = allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
+	if should_zero {
+		data, err = allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
+	} else {
+		data, err = allocator.procedure(allocator.data, .Resize_Non_Zeroed, new_size, alignment, ptr, old_size, loc)
+	}
 	if err == .Mode_Not_Implemented {
-		data, err = allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		if should_zero {
+			data, err = allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		} else {
+			data, err = allocator.procedure(allocator.data, .Alloc_Non_Zeroed, new_size, alignment, nil, 0, loc)
+		}
 		if err != nil {
 			return
 		}
@@ -213,15 +184,30 @@ mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAUL
 	return
 }
 
+mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	return _mem_resize(ptr, old_size, new_size, alignment, allocator, true, loc)
+}
+non_zero_mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	return _mem_resize(ptr, old_size, new_size, alignment, allocator, false, loc)
+}
+
 memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
 	switch {
 	case n == 0: return true
 	case x == y: return true
 	}
-	
 	a, b := ([^]byte)(x), ([^]byte)(y)
 	length := uint(n)
+
+	for i := uint(0); i < length; i += 1 {
+		if a[i] != b[i] {
+			return false
+		}
+	}
+	return true
 	
+/*
+
 	when size_of(uint) == 8 {
 		if word_length := length >> 3; word_length != 0 {
 			for _ in 0..<word_length {
@@ -276,6 +262,7 @@ memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
 
 		return true
 	}
+*/
 
 }
 memory_compare :: proc "contextless" (a, b: rawptr, n: int) -> int #no_bounds_check {
@@ -401,6 +388,48 @@ cstring_to_string :: proc "contextless" (s: cstring) -> string {
 }
 
 
+cstring_eq :: proc "contextless" (lhs, rhs: cstring) -> bool {
+	x := ([^]byte)(lhs)
+	y := ([^]byte)(rhs)
+	if x == y {
+		return true
+	}
+	if (x == nil) ~ (y == nil) {
+		return false
+	}
+	xn := cstring_len(lhs)
+	yn := cstring_len(rhs)
+	if xn != yn {
+		return false
+	}
+	return #force_inline memory_equal(x, y, xn)
+}
+
+cstring_cmp :: proc "contextless" (lhs, rhs: cstring) -> int {
+	x := ([^]byte)(lhs)
+	y := ([^]byte)(rhs)
+	if x == y {
+		return 0
+	}
+	if (x == nil) ~ (y == nil) {
+		return -1 if x == nil else +1
+	}
+	xn := cstring_len(lhs)
+	yn := cstring_len(rhs)
+	ret := memory_compare(x, y, min(xn, yn))
+	if ret == 0 && xn != yn {
+		return -1 if xn < yn else +1
+	}
+	return ret
+}
+
+cstring_ne :: #force_inline proc "contextless" (a, b: cstring) -> bool { return !cstring_eq(a, b) }
+cstring_lt :: #force_inline proc "contextless" (a, b: cstring) -> bool { return cstring_cmp(a, b) < 0 }
+cstring_gt :: #force_inline proc "contextless" (a, b: cstring) -> bool { return cstring_cmp(a, b) > 0 }
+cstring_le :: #force_inline proc "contextless" (a, b: cstring) -> bool { return cstring_cmp(a, b) <= 0 }
+cstring_ge :: #force_inline proc "contextless" (a, b: cstring) -> bool { return cstring_cmp(a, b) >= 0 }
+
+
 complex32_eq :: #force_inline proc "contextless"  (a, b: complex32)  -> bool { return real(a) == real(b) && imag(a) == imag(b) }
 complex32_ne :: #force_inline proc "contextless"  (a, b: complex32)  -> bool { return real(a) != real(b) || imag(a) != imag(b) }
 
@@ -535,36 +564,6 @@ string_decode_last_rune :: proc "contextless" (s: string) -> (rune, int) {
 	return r, size
 }
 
-
-abs_f16 :: #force_inline proc "contextless" (x: f16) -> f16 {
-	return -x if x < 0 else x
-}
-abs_f32 :: #force_inline proc "contextless" (x: f32) -> f32 {
-	return -x if x < 0 else x
-}
-abs_f64 :: #force_inline proc "contextless" (x: f64) -> f64 {
-	return -x if x < 0 else x
-}
-
-min_f16 :: #force_inline proc "contextless" (a, b: f16) -> f16 {
-	return a if a < b else b
-}
-min_f32 :: #force_inline proc "contextless" (a, b: f32) -> f32 {
-	return a if a < b else b
-}
-min_f64 :: #force_inline proc "contextless" (a, b: f64) -> f64 {
-	return a if a < b else b
-}
-max_f16 :: #force_inline proc "contextless" (a, b: f16) -> f16 {
-	return a if a > b else b
-}
-max_f32 :: #force_inline proc "contextless" (a, b: f32) -> f32 {
-	return a if a > b else b
-}
-max_f64 :: #force_inline proc "contextless" (a, b: f64) -> f64 {
-	return a if a > b else b
-}
-
 abs_complex32 :: #force_inline proc "contextless" (x: complex32) -> f16 {
 	p, q := abs(real(x)), abs(imag(x))
 	if p < q {
@@ -676,7 +675,7 @@ mul_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 mul_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
@@ -688,7 +687,7 @@ mul_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 mul_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
@@ -700,7 +699,7 @@ mul_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
@@ -714,7 +713,7 @@ quo_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
@@ -728,7 +727,7 @@ quo_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
@@ -742,11 +741,11 @@ quo_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 @(link_name="__truncsfhf2", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
-truncsfhf2 :: proc "c" (value: f32) -> u16 {
+truncsfhf2 :: proc "c" (value: f32) -> __float16 {
 	v: struct #raw_union { i: u32, f: f32 }
 	i, s, e, m: i32
 
@@ -760,7 +759,7 @@ truncsfhf2 :: proc "c" (value: f32) -> u16 {
 
 	if e <= 0 {
 		if e < -10 {
-			return u16(s)
+			return transmute(__float16)u16(s)
 		}
 		m = (m | 0x00800000) >> u32(1 - e)
 
@@ -768,14 +767,14 @@ truncsfhf2 :: proc "c" (value: f32) -> u16 {
 			m += 0x00002000
 		}
 
-		return u16(s | (m >> 13))
+		return transmute(__float16)u16(s | (m >> 13))
 	} else if e == 0xff - (127 - 15) {
 		if m == 0 {
-			return u16(s | 0x7c00) /* NOTE(bill): infinity */
+			return transmute(__float16)u16(s | 0x7c00) /* NOTE(bill): infinity */
 		} else {
 			/* NOTE(bill): NAN */
 			m >>= 13
-			return u16(s | 0x7c00 | m | i32(m == 0))
+			return transmute(__float16)u16(s | 0x7c00 | m | i32(m == 0))
 		}
 	} else {
 		if m & 0x00001000 != 0 {
@@ -795,23 +794,24 @@ truncsfhf2 :: proc "c" (value: f32) -> u16 {
 				intrinsics.volatile_store(&f, g)
 			}
 
-			return u16(s | 0x7c00)
+			return transmute(__float16)u16(s | 0x7c00)
 		}
 
-		return u16(s | (e << 10) | (m >> 13))
+		return transmute(__float16)u16(s | (e << 10) | (m >> 13))
 	}
 }
 
 
 @(link_name="__truncdfhf2", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
-truncdfhf2 :: proc "c" (value: f64) -> u16 {
+truncdfhf2 :: proc "c" (value: f64) -> __float16 {
 	return truncsfhf2(f32(value))
 }
 
 @(link_name="__gnu_h2f_ieee", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
-gnu_h2f_ieee :: proc "c" (value: u16) -> f32 {
+gnu_h2f_ieee :: proc "c" (value_: __float16) -> f32 {
 	fp32 :: struct #raw_union { u: u32, f: f32 }
 
+	value := transmute(u16)value_
 	v: fp32
 	magic, inf_or_nan: fp32
 	magic.u = u32((254 - 15) << 23)
@@ -828,12 +828,12 @@ gnu_h2f_ieee :: proc "c" (value: u16) -> f32 {
 
 
 @(link_name="__gnu_f2h_ieee", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
-gnu_f2h_ieee :: proc "c" (value: f32) -> u16 {
+gnu_f2h_ieee :: proc "c" (value: f32) -> __float16 {
 	return truncsfhf2(value)
 }
 
 @(link_name="__extendhfsf2", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
-extendhfsf2 :: proc "c" (value: u16) -> f32 {
+extendhfsf2 :: proc "c" (value: __float16) -> f32 {
 	return gnu_h2f_ieee(value)
 }
 
@@ -1034,3 +1034,25 @@ fixdfti :: proc(a: u64) -> i128 {
 	}
 
 }
+
+
+
+__write_bits :: proc "contextless" (dst, src: [^]byte, offset: uintptr, size: uintptr) {
+	for i in 0..<size {
+		j := offset+i
+		the_bit := byte((src[i/8]) & (1<<(i&7)) != 0)
+		b := the_bit<<(j&7)
+		dst[j/8] &~= b
+		dst[j/8] |=  b
+	}
+}
+
+__read_bits :: proc "contextless" (dst, src: [^]byte, offset: uintptr, size: uintptr) {
+	for j in 0..<size {
+		i := offset+j
+		the_bit := byte((src[i/8]) & (1<<(i&7)) != 0)
+		b := the_bit<<(j&7)
+		dst[j/8] &~= b
+		dst[j/8] |=  b
+	}
+}

base/runtime/os_specific.odin

@@ -0,0 +1,7 @@
+package runtime
+
+_OS_Errno :: distinct int
+
+stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	return _stderr_write(data)
+}

base/runtime/os_specific_bsd.odin

@@ -0,0 +1,22 @@
+//+build freebsd, openbsd
+//+private
+package runtime
+
+foreign import libc "system:c"
+
+@(default_calling_convention="c")
+foreign libc {
+	@(link_name="write")
+	_unix_write :: proc(fd: i32, buf: rawptr, size: int) -> int ---
+
+	__error :: proc() -> ^i32 ---
+}
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	ret := _unix_write(2, raw_data(data), len(data))
+	if ret < len(data) {
+		err := __error()
+		return int(ret), _OS_Errno(err^ if err != nil else 0)
+	}
+	return int(ret), 0
+}

base/runtime/os_specific_darwin.odin

@@ -0,0 +1,15 @@
+//+build darwin
+//+private
+package runtime
+
+import "base:intrinsics"
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	WRITE  :: 0x20000004
+	STDERR :: 2
+	ret := intrinsics.syscall(WRITE, STDERR, uintptr(raw_data(data)), uintptr(len(data)))
+	if ret < 0 {
+		return 0, _OS_Errno(-ret)
+	}
+	return int(ret), 0
+}

base/runtime/os_specific_freestanding.odin

@@ -1,7 +1,8 @@
 //+build freestanding
+//+private
 package runtime
 
 // TODO(bill): reimplement `os.write`
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	return 0, -1
 }

base/runtime/os_specific_js.odin

@@ -1,9 +1,10 @@
 //+build js
+//+private
 package runtime
 
 foreign import "odin_env"
 
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	foreign odin_env {
 		write :: proc "contextless" (fd: u32, p: []byte) ---
 	}

base/runtime/os_specific_linux.odin

@@ -0,0 +1,24 @@
+//+private
+package runtime
+
+import "base:intrinsics"
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	when ODIN_ARCH == .amd64 {
+		SYS_write :: uintptr(1)
+	} else when ODIN_ARCH == .arm64 {
+		SYS_write :: uintptr(64)
+	} else when ODIN_ARCH == .i386 {
+		SYS_write :: uintptr(4)
+	} else when ODIN_ARCH == .arm32 {
+		SYS_write :: uintptr(4)
+	}
+
+	stderr :: 2
+
+	ret := int(intrinsics.syscall(SYS_write, uintptr(stderr), uintptr(raw_data(data)), uintptr(len(data))))
+	if ret < 0 && ret > -4096 {
+		return 0, _OS_Errno(-ret)
+	}
+	return ret, 0
+}

base/runtime/os_specific_wasi.odin

@@ -1,9 +1,10 @@
 //+build wasi
+//+private
 package runtime
 
 import "core:sys/wasm/wasi"
 
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	data := (wasi.ciovec_t)(data)
 	n, err := wasi.fd_write(1, {data})
 	return int(n), _OS_Errno(err)

base/runtime/os_specific_windows.odin

@@ -0,0 +1,51 @@
+//+build windows
+//+private
+package runtime
+
+foreign import kernel32 "system:Kernel32.lib"
+
+@(private="file")
+@(default_calling_convention="system")
+foreign kernel32 {
+	// NOTE(bill): The types are not using the standard names (e.g. DWORD and LPVOID) to just minimizing the dependency
+
+	// stderr_write
+	GetStdHandle         :: proc(which: u32) -> rawptr ---
+	SetHandleInformation :: proc(hObject: rawptr, dwMask: u32, dwFlags: u32) -> b32 ---
+	WriteFile            :: proc(hFile: rawptr, lpBuffer: rawptr, nNumberOfBytesToWrite: u32, lpNumberOfBytesWritten: ^u32, lpOverlapped: rawptr) -> b32 ---
+	GetLastError         :: proc() -> u32 ---
+}
+
+_stderr_write :: proc "contextless" (data: []byte) -> (n: int, err: _OS_Errno) #no_bounds_check {
+	if len(data) == 0 {
+		return 0, 0
+	}
+
+	STD_ERROR_HANDLE :: ~u32(0) -12 + 1
+	HANDLE_FLAG_INHERIT :: 0x00000001
+	MAX_RW :: 1<<30
+
+	h := GetStdHandle(STD_ERROR_HANDLE)
+	when size_of(uintptr) == 8 {
+		SetHandleInformation(h, HANDLE_FLAG_INHERIT, 0)
+	}
+
+	single_write_length: u32
+	total_write: i64
+	length := i64(len(data))
+
+	for total_write < length {
+		remaining := length - total_write
+		to_write := u32(min(i32(remaining), MAX_RW))
+
+		e := WriteFile(h, &data[total_write], to_write, &single_write_length, nil)
+		if single_write_length <= 0 || !e {
+			err = _OS_Errno(GetLastError())
+			n = int(total_write)
+			return
+		}
+		total_write += i64(single_write_length)
+	}
+	n = int(total_write)
+	return
+}

base/runtime/print.odin

@@ -8,6 +8,11 @@ _INTEGER_DIGITS_VAR := _INTEGER_DIGITS
 when !ODIN_NO_RTTI {
 	print_any_single :: proc "contextless" (arg: any) {
 		x := arg
+		if x.data == nil {
+			print_string("nil")
+			return
+		}
+
 		if loc, ok := x.(Source_Code_Location); ok {
 			print_caller_location(loc)
 			return
@@ -48,6 +53,7 @@ when !ODIN_NO_RTTI {
 		case int:     print_int(v)
 		case uint:    print_uint(v)
 		case uintptr: print_uintptr(v)
+		case rawptr:  print_uintptr(uintptr(v))
 
 		case bool: print_string("true" if v else "false")
 		case b8:   print_string("true" if v else "false")
@@ -58,7 +64,7 @@ when !ODIN_NO_RTTI {
 		case:
 			ti := type_info_of(x.id)
 			#partial switch v in ti.variant {
-			case Type_Info_Pointer:
+			case Type_Info_Pointer, Type_Info_Multi_Pointer:
 				print_uintptr((^uintptr)(x.data)^)
 				return
 			}
@@ -67,7 +73,9 @@ when !ODIN_NO_RTTI {
 		}
 	}
 	println_any :: proc "contextless" (args: ..any) {
+		context = default_context()
 		loop: for arg, i in args {
+			assert(arg.id != nil)
 			if i != 0 {
 				print_string(" ")
 			}
@@ -115,13 +123,13 @@ encode_rune :: proc "contextless" (c: rune) -> ([4]u8, int) {
 }
 
 print_string :: proc "contextless" (str: string) -> (n: int) {
-	n, _ = os_write(transmute([]byte)str)
+	n, _ = stderr_write(transmute([]byte)str)
 	return
 }
 
 print_strings :: proc "contextless" (args: ..string) -> (n: int) {
 	for str in args {
-		m, err := os_write(transmute([]byte)str)
+		m, err := stderr_write(transmute([]byte)str)
 		n += m
 		if err != 0 {
 			break
@@ -131,7 +139,7 @@ print_strings :: proc "contextless" (args: ..string) -> (n: int) {
 }
 
 print_byte :: proc "contextless" (b: byte) -> (n: int) {
-	n, _ = os_write([]byte{b})
+	n, _ = stderr_write([]byte{b})
 	return
 }
 
@@ -170,7 +178,7 @@ print_rune :: proc "contextless" (r: rune) -> int #no_bounds_check {
 	}
 
 	b, n := encode_rune(r)
-	m, _ := os_write(b[:n])
+	m, _ := stderr_write(b[:n])
 	return m
 }
 
@@ -186,7 +194,7 @@ print_u64 :: proc "contextless" (x: u64) #no_bounds_check {
 	}
 	i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 
-	os_write(a[i:])
+	stderr_write(a[i:])
 }
 
 
@@ -208,26 +216,26 @@ print_i64 :: proc "contextless" (x: i64) #no_bounds_check {
 		i -= 1; a[i] = '-'
 	}
 
-	os_write(a[i:])
+	stderr_write(a[i:])
 }
 
 print_uint    :: proc "contextless" (x: uint)    { print_u64(u64(x)) }
 print_uintptr :: proc "contextless" (x: uintptr) { print_u64(u64(x)) }
 print_int     :: proc "contextless" (x: int)     { print_i64(i64(x)) }
 
-print_caller_location :: proc "contextless" (using loc: Source_Code_Location) {
-	print_string(file_path)
+print_caller_location :: proc "contextless" (loc: Source_Code_Location) {
+	print_string(loc.file_path)
 	when ODIN_ERROR_POS_STYLE == .Default {
 		print_byte('(')
-		print_u64(u64(line))
+		print_u64(u64(loc.line))
 		print_byte(':')
-		print_u64(u64(column))
+		print_u64(u64(loc.column))
 		print_byte(')')
 	} else when ODIN_ERROR_POS_STYLE == .Unix {
 		print_byte(':')
-		print_u64(u64(line))
+		print_u64(u64(loc.line))
 		print_byte(':')
-		print_u64(u64(column))
+		print_u64(u64(loc.column))
 		print_byte(':')
 	} else {
 		#panic("unhandled ODIN_ERROR_POS_STYLE")
@@ -390,9 +398,9 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 		if info.is_packed    { print_string("#packed ") }
 		if info.is_raw_union { print_string("#raw_union ") }
 		if info.custom_align {
-			print_string("#align ")
+			print_string("#align(")
 			print_u64(u64(ti.align))
-			print_byte(' ')
+			print_string(") ")
 		}
 		print_byte('{')
 		for name, i in info.names {
@@ -406,8 +414,9 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 	case Type_Info_Union:
 		print_string("union ")
 		if info.custom_align {
-			print_string("#align ")
+			print_string("#align(")
 			print_u64(u64(ti.align))
+			print_string(") ")
 		}
 		if info.no_nil {
 			print_string("#no_nil ")
@@ -450,6 +459,20 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 		}
 		print_byte(']')
 
+	case Type_Info_Bit_Field:
+		print_string("bit_field ")
+		print_type(info.backing_type)
+		print_string(" {")
+		for name, i in info.names {
+			if i > 0 { print_string(", ") }
+			print_string(name)
+			print_string(": ")
+			print_type(info.types[i])
+			print_string(" | ")
+			print_u64(u64(info.bit_sizes[i]))
+		}
+		print_byte('}')
+
 
 	case Type_Info_Simd_Vector:
 		print_string("#simd[")
@@ -463,11 +486,11 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 		print_string(") ")
 		print_type(info.pointer)
 
-	case Type_Info_Relative_Slice:
+	case Type_Info_Relative_Multi_Pointer:
 		print_string("#relative(")
 		print_type(info.base_integer)
 		print_string(") ")
-		print_type(info.slice)
+		print_type(info.pointer)
 		
 	case Type_Info_Matrix:
 		print_string("matrix[")

base/runtime/procs.odin

@@ -4,7 +4,7 @@ when ODIN_NO_CRT && ODIN_OS == .Windows {
 	foreign import lib "system:NtDll.lib"
 	
 	@(private="file")
-	@(default_calling_convention="stdcall")
+	@(default_calling_convention="system")
 	foreign lib {
 		RtlMoveMemory :: proc(dst, s: rawptr, length: int) ---
 		RtlFillMemory :: proc(dst: rawptr, length: int, fill: i32) ---
@@ -31,13 +31,24 @@ when ODIN_NO_CRT && ODIN_OS == .Windows {
 		if ptr != nil && len != 0 {
 			b := byte(val)
 			p := ([^]byte)(ptr)
-			for i in 0..<len {
+			for i := 0; i < len; i += 1 {
 				p[i] = b
 			}
 		}
 		return ptr
 	}
-	
+
+	@(link_name="bzero", linkage="strong", require)
+	bzero :: proc "c" (ptr: rawptr, len: int) -> rawptr {
+		if ptr != nil && len != 0 {
+			p := ([^]byte)(ptr)
+			for i := 0; i < len; i += 1 {
+				p[i] = 0
+			}
+		}
+		return ptr
+	}
+
 	@(link_name="memmove", linkage="strong", require)
 	memmove :: proc "c" (dst, src: rawptr, len: int) -> rawptr {
 		d, s := ([^]byte)(dst), ([^]byte)(src)
@@ -75,7 +86,7 @@ when ODIN_NO_CRT && ODIN_OS == .Windows {
 		if ptr != nil && len != 0 {
 			b := byte(val)
 			p := ([^]byte)(ptr)
-			for i in 0..<len {
+			for i := 0; i < len; i += 1 {
 				p[i] = b
 			}
 		}

base/runtime/procs_darwin.odin

@@ -3,7 +3,7 @@ package runtime
 
 foreign import "system:Foundation.framework"
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 objc_id :: ^intrinsics.objc_object
 objc_Class :: ^intrinsics.objc_class

base/runtime/procs_windows_amd64.asm

@@ -0,0 +1,79 @@
+bits 64
+
+global __chkstk
+global _tls_index
+global _fltused
+
+section .data
+	_tls_index: dd 0
+	_fltused:   dd 0x9875
+
+section .text
+; NOTE(flysand): The function call to __chkstk is called
+; by the compiler, when we're allocating arrays larger than
+; a page size. The reason is because the OS doesn't map the
+; whole stack into memory all at once, but does so page-by-page.
+; When the next page is touched, the CPU generates a page fault,
+; which *the OS* is handling by allocating the next page in the
+; stack until we reach the limit of stack size.
+;
+; This page is called the guard page, touching it will extend
+; the size of the stack and overwrite the stack limit in the TEB.
+;
+; If we allocate a large enough array and start writing from the
+; bottom of it, it's possible that we may start touching
+; non-contiguous pages which are unmapped. OS only maps the stack
+; page into the memory if the page above it was also mapped.
+;
+; Therefore the compilers insert this routine, the sole purpose
+; of which is to step through the stack starting from the RSP
+; down to the new RSP after allocation, and touch every page
+; of the new allocation so that the stack is fully mapped for
+; the new allocation
+;
+; I've gotten this code by disassembling the output of MSVC long
+; time ago. I don't remember if I've cleaned it up, but it definately
+; stinks.
+;
+; Additional notes:
+;   RAX (passed as parameter) holds the allocation's size
+;   GS:[0x10] references the current stack limit
+;     (i.e. bottom of the stack (i.e. lowest address accessible))
+;
+; Also this stuff is windows-only kind of thing, because linux people
+; didn't think stack that grows is cool enough for them, but the kernel
+; totally supports this kind of stack.
+__chkstk:
+	;; Allocate 16 bytes to store values of r10 and r11
+	sub   rsp, 0x10
+	mov   [rsp], r10
+	mov   [rsp+0x8], r11
+	;; Set r10 to point to the stack as of the moment of the function call
+	lea   r10, [rsp+0x18]
+	;; Subtract r10 til the bottom of the stack allocation, if we overflow
+	;; reset r10 to 0, we'll crash with segfault anyway
+	xor   r11, r11
+	sub   r10, rax
+	cmovb r10, r11
+	;; Load r11 with the bottom of the stack (lowest allocated address)
+	mov   r11, gs:[0x10] ; NOTE(flysand): gs:[0x10] is stack limit
+	;; If the bottom of the allocation is above the bottom of the stack,
+	;; we don't need to probe
+	cmp   r10, r11
+	jnb   .end
+	;; Align the bottom of the allocation down to page size
+	and   r10w, 0xf000
+.loop:
+	;; Move the pointer to the next guard page, and touch it by loading 0
+	;; into that page
+	lea   r11, [r11-0x1000]
+	mov   byte [r11], 0x0
+	;; Did we reach the bottom of the allocation?
+	cmp   r10, r11
+	jnz   .loop
+.end:
+	;; Restore previous r10 and r11 and return
+	mov   r10, [rsp]
+	mov   r11, [rsp+0x8]
+	add   rsp, 0x10
+	ret

base/runtime/procs_windows_amd64.odin

@@ -1,11 +1,12 @@
 //+private
+//+no-instrumentation
 package runtime
 
 foreign import kernel32 "system:Kernel32.lib"
 
 @(private)
 foreign kernel32 {
-	RaiseException :: proc "stdcall" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: u32, lpArguments: ^uint) -> ! ---
+	RaiseException :: proc "system" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: u32, lpArguments: ^uint) -> ! ---
 }
 
 windows_trap_array_bounds :: proc "contextless" () -> ! {

base/runtime/procs_windows_i386.odin

@@ -1,4 +1,5 @@
 //+private
+//+no-instrumentation
 package runtime
 
 @require foreign import "system:int64.lib"
@@ -12,7 +13,7 @@ windows_trap_array_bounds :: proc "contextless" () -> ! {
 	EXCEPTION_ARRAY_BOUNDS_EXCEEDED :: 0xC000008C
 
 	foreign kernel32 {
-		RaiseException :: proc "stdcall" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: DWORD, lpArguments: ^ULONG_PTR) -> ! ---
+		RaiseException :: proc "system" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: DWORD, lpArguments: ^ULONG_PTR) -> ! ---
 	}
 
 	RaiseException(EXCEPTION_ARRAY_BOUNDS_EXCEEDED, 0, 0, nil)

base/runtime/udivmod128.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 udivmod128 :: proc "c" (a, b: u128, rem: ^u128) -> u128 {
 	_ctz :: intrinsics.count_trailing_zeros

build.bat

@@ -51,7 +51,10 @@ set compiler_flags= -nologo -Oi -TP -fp:precise -Gm- -MP -FC -EHsc- -GR- -GF
 set compiler_defines= -DODIN_VERSION_RAW=\"%odin_version_raw%\"
 
 if not exist .git\ goto skip_git_hash
-for /f %%i in ('git rev-parse --short HEAD') do set GIT_SHA=%%i
+for /f "tokens=1,2" %%i IN ('git show "--pretty=%%cd %%h" "--date=format:%%Y-%%m" --no-patch --no-notes HEAD') do (
+	set odin_version_raw=dev-%%i
+	set GIT_SHA=%%j
+)
 if %ERRORLEVEL% equ 0 set compiler_defines=%compiler_defines% -DGIT_SHA=\"%GIT_SHA%\"
 :skip_git_hash
 
@@ -77,6 +80,16 @@ set libs= ^
 	Synchronization.lib ^
 	bin\llvm\windows\LLVM-C.lib
 
+rem DO NOT TOUCH!
+rem THIS TILDE STUFF IS FOR DEVELOPMENT ONLY!
+set tilde_backend=0
+if %tilde_backend% EQU 1 (
+	set libs=%libs% src\tilde\tb.lib
+	set compiler_defines=%compiler_defines% -DODIN_TILDE_BACKEND
+)
+rem DO NOT TOUCH!
+
+
 set linker_flags= -incremental:no -opt:ref -subsystem:console
 
 if %release_mode% EQU 0 ( rem Debug
@@ -97,8 +110,9 @@ if %errorlevel% neq 0 goto end_of_build
 call build_vendor.bat
 if %errorlevel% neq 0 goto end_of_build
 
-if %release_mode% EQU 0 odin run examples/demo
+rem If the demo doesn't run for you and your CPU is more than a decade old, try -microarch:native
+if %release_mode% EQU 0 odin run examples/demo -- Hellope World
 
 del *.obj > NUL 2> NUL
 
-:end_of_build
+:end_of_build

build_odin.sh

@@ -1,130 +1,91 @@
-#!/usr/bin/env bash
+#!/usr/bin/env sh
 set -eu
 
-: ${CXX=clang++}
 : ${CPPFLAGS=}
+: ${CXX=clang++}
 : ${CXXFLAGS=}
 : ${LDFLAGS=}
-: ${ODIN_VERSION=dev-$(date +"%Y-%m")}
-: ${GIT_SHA=}
+: ${LLVM_CONFIG=}
 
-CPPFLAGS="$CPPFLAGS -DODIN_VERSION_RAW=\"$ODIN_VERSION\""
+CPPFLAGS="$CPPFLAGS -DODIN_VERSION_RAW=\"dev-$(date +"%Y-%m")\""
 CXXFLAGS="$CXXFLAGS -std=c++14"
+DISABLED_WARNINGS="-Wno-switch -Wno-macro-redefined -Wno-unused-value"
 LDFLAGS="$LDFLAGS -pthread -lm -lstdc++"
+OS_ARCH="$(uname -m)"
+OS_NAME="$(uname -s)"
 
-if [ -d ".git" ]; then
-	GIT_SHA=$(git rev-parse --short HEAD || :)
-	if [ "$GIT_SHA" ]; then
-		CPPFLAGS="$CPPFLAGS -DGIT_SHA=\"$GIT_SHA\""
-	fi
+if [ -d ".git" ] && [ -n "$(command -v git)" ]; then
+	GIT_SHA=$(git show --pretty='%h' --no-patch --no-notes HEAD)
+	CPPFLAGS="$CPPFLAGS -DGIT_SHA=\"$GIT_SHA\""
 fi
 
-DISABLED_WARNINGS="-Wno-switch -Wno-macro-redefined -Wno-unused-value"
-OS=$(uname)
-
-panic() {
-	printf "%s\n" "$1"
+error() {
+	printf "ERROR: %s\n" "$1"
 	exit 1
 }
 
-version() { echo "$@" | awk -F. '{ printf("%d%03d%03d%03d\n", $1,$2,$3,$4); }'; }
-
-config_darwin() {
-	local ARCH=$(uname -m)
-	: ${LLVM_CONFIG=llvm-config}
-
-	# allow for arm only llvm's with version 13
-	if [ "${ARCH}" == "arm64" ]; then
-		MIN_LLVM_VERSION=("13.0.0")
+if [ -z "$LLVM_CONFIG" ]; then
+	# darwin, linux, openbsd
+	if   [ -n "$(command -v llvm-config-17)" ]; then LLVM_CONFIG="llvm-config-17"
+	elif [ -n "$(command -v llvm-config-14)" ]; then LLVM_CONFIG="llvm-config-14"
+	elif [ -n "$(command -v llvm-config-13)" ]; then LLVM_CONFIG="llvm-config-13"
+	elif [ -n "$(command -v llvm-config-12)" ]; then LLVM_CONFIG="llvm-config-12"
+	elif [ -n "$(command -v llvm-config-11)" ]; then LLVM_CONFIG="llvm-config-11"
+	# freebsd
+	elif [ -n "$(command -v llvm-config17)" ]; then  LLVM_CONFIG="llvm-config-17"
+	elif [ -n "$(command -v llvm-config14)" ]; then  LLVM_CONFIG="llvm-config-14"
+	elif [ -n "$(command -v llvm-config13)" ]; then  LLVM_CONFIG="llvm-config-13"
+	elif [ -n "$(command -v llvm-config12)" ]; then  LLVM_CONFIG="llvm-config-12"
+	elif [ -n "$(command -v llvm-config11)" ]; then  LLVM_CONFIG="llvm-config-11"
+	# fallback
+	elif [ -n "$(command -v llvm-config)" ]; then LLVM_CONFIG="llvm-config"
 	else
-		# allow for x86 / amd64 all llvm versions beginning from 11
-		MIN_LLVM_VERSION=("11.1.0")
+		error "No llvm-config command found. Set LLVM_CONFIG to proceed."
 	fi
+fi
 
-	if [ $(version $($LLVM_CONFIG --version)) -lt $(version $MIN_LLVM_VERSION) ]; then
-		if [ "${ARCH}" == "arm64" ]; then
-			panic "Requirement: llvm-config must be base version 13 for arm64"
-		else
-			panic "Requirement: llvm-config must be base version greater than 11 for amd64/x86"
+LLVM_VERSION="$($LLVM_CONFIG --version)"
+LLVM_VERSION_MAJOR="$(echo $LLVM_VERSION | awk -F. '{print $1}')"
+LLVM_VERSION_MINOR="$(echo $LLVM_VERSION | awk -F. '{print $2}')"
+LLVM_VERSION_PATCH="$(echo $LLVM_VERSION | awk -F. '{print $3}')"
+
+if [ $LLVM_VERSION_MAJOR -lt 11 ] ||
+	([ $LLVM_VERSION_MAJOR -gt 14 ] && [ $LLVM_VERSION_MAJOR -lt 17 ]); then
+	error "Invalid LLVM version $LLVM_VERSION: must be 11, 12, 13, 14 or 17"
+fi
+
+case "$OS_NAME" in
+Darwin)
+	if [ "$OS_ARCH" == "arm64" ]; then
+		if [ $LLVM_VERSION_MAJOR -lt 13 ] || [ $LLVM_VERSION_MAJOR -gt 17 ]; then
+			error "Darwin Arm64 requires LLVM 13, 14 or 17"
 		fi
 	fi
 
-	MAX_LLVM_VERSION=("14.999.999")
-	if [ $(version $($LLVM_CONFIG --version)) -gt $(version $MAX_LLVM_VERSION) ]; then
-		echo "Tried to use " $(which $LLVM_CONFIG) "version" $($LLVM_CONFIG --version)
-		panic "Requirement: llvm-config must be base version smaller than 15"
-	fi
-
-	LDFLAGS="$LDFLAGS -liconv -ldl -framework System"
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
-	LDFLAGS="$LDFLAGS -lLLVM-C"
-}
-
-config_freebsd() {
-	: ${LLVM_CONFIG=}
-
-	if [ ! "$LLVM_CONFIG" ]; then
-		if [ -x "$(command -v llvm-config11)" ]; then
-			LLVM_CONFIG=llvm-config11
-		elif [ -x "$(command -v llvm-config12)" ]; then
-			LLVM_CONFIG=llvm-config12
-		elif [ -x "$(command -v llvm-config13)" ]; then
-			LLVM_CONFIG=llvm-config13
-		else
-			panic "Unable to find LLVM-config"
-		fi
-	fi
-
+	LDFLAGS="$LDFLAGS -liconv -ldl -framework System -lLLVM"
+	;;
+FreeBSD)
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
 	LDFLAGS="$LDFLAGS $($LLVM_CONFIG --libs core native --system-libs)"
-}
-
-config_openbsd() {
-	: ${LLVM_CONFIG=/usr/local/bin/llvm-config}
-
-	LDFLAGS="$LDFLAGS -liconv"
+	;;
+Linux)
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
-	LDFLAGS="$LDFLAGS $($LLVM_CONFIG --libs core native --system-libs)"
-}
-
-config_linux() {
-	: ${LLVM_CONFIG=}
-
-	if [ ! "$LLVM_CONFIG" ]; then
-		if [ -x "$(command -v llvm-config)" ]; then
-			LLVM_CONFIG=llvm-config
-		elif [ -x "$(command -v llvm-config-11)" ]; then
-			LLVM_CONFIG=llvm-config-11
-		elif [ -x "$(command -v llvm-config-11-64)" ]; then
-			LLVM_CONFIG=llvm-config-11-64
-		elif [ -x "$(command -v llvm-config-14)" ]; then
-			LLVM_CONFIG=llvm-config-14
-		else
-			panic "Unable to find LLVM-config"
-		fi
-	fi
-
-	MIN_LLVM_VERSION=("11.0.0")
-	if [ $(version $($LLVM_CONFIG --version)) -lt $(version $MIN_LLVM_VERSION) ]; then
-		echo "Tried to use " $(which $LLVM_CONFIG) "version" $($LLVM_CONFIG --version)
-		panic "Requirement: llvm-config must be base version greater than 11"
-	fi
-
-	MAX_LLVM_VERSION=("14.999.999")
-	if [ $(version $($LLVM_CONFIG --version)) -gt $(version $MAX_LLVM_VERSION) ]; then
-		echo "Tried to use " $(which $LLVM_CONFIG) "version" $($LLVM_CONFIG --version)
-		panic "Requirement: llvm-config must be base version smaller than 15"
-	fi
-
-	LDFLAGS="$LDFLAGS -ldl"
-	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
-	LDFLAGS="$LDFLAGS $($LLVM_CONFIG --libs core native --system-libs --libfiles) -Wl,-rpath=\$ORIGIN"
-
-	# Creates a copy of the llvm library in the build dir, this is meant to support compiler explorer.
-	# The annoyance is that this copy can be cluttering the development folder. TODO: split staging folders
-	# for development and compiler explorer builds
+	LDFLAGS="$LDFLAGS -ldl $($LLVM_CONFIG --libs core native --system-libs --libfiles)"
+	# Copy libLLVM*.so into current directory for linking
+	# NOTE: This is needed by the Linux release pipeline!
 	cp $(readlink -f $($LLVM_CONFIG --libfiles)) ./
-}
+	LDFLAGS="$LDFLAGS -Wl,-rpath=\$ORIGIN"
+	;;
+OpenBSD)
+	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
+	LDFLAGS="$LDFLAGS -liconv"
+	LDFLAGS="$LDFLAGS $($LLVM_CONFIG --libs core native --system-libs)"
+	;;
+*)
+	error "Platform \"$OS_NAME\" unsupported"
+	;;
+esac
 
 build_odin() {
 	case $1 in
@@ -135,20 +96,19 @@ build_odin() {
 		EXTRAFLAGS="-O3"
 		;;
 	release-native)
-		local ARCH=$(uname -m)
-        	if [ "${ARCH}" == "arm64" ]; then
-            		# Use preferred flag for Arm (ie arm64 / aarch64 / etc)
-            		EXTRAFLAGS="-O3 -mcpu=native"
-        	else
-            		# Use preferred flag for x86 / amd64
-            		EXTRAFLAGS="-O3 -march=native"
-        	fi
+		if [ "$OS_ARCH" == "arm64" ]; then
+			# Use preferred flag for Arm (ie arm64 / aarch64 / etc)
+			EXTRAFLAGS="-O3 -mcpu=native"
+		else
+			# Use preferred flag for x86 / amd64
+			EXTRAFLAGS="-O3 -march=native"
+		fi
 		;;
 	nightly)
 		EXTRAFLAGS="-DNIGHTLY -O3"
 		;;
 	*)
-		panic "Build mode unsupported!"
+		error "Build mode \"$1\" unsupported!"
 		;;
 	esac
 
@@ -158,58 +118,23 @@ build_odin() {
 }
 
 run_demo() {
-	./odin run examples/demo/demo.odin -file
+	./odin run examples/demo/demo.odin -file -- Hellope World
 }
 
-have_which() {
-	if ! command -v which > /dev/null 2>&1 ; then
-		panic "Could not find \`which\`"
-	fi
-}
-
-have_which
-
-case $OS in
-Linux)
-	config_linux
-	;;
-Darwin)
-	config_darwin
-	;;
-OpenBSD)
-	config_openbsd
-	;;
-FreeBSD)
-	config_freebsd
-	;;
-*)
-	panic "Platform unsupported!"
-	;;
-esac
-
-if [[ $# -eq 0 ]]; then
+if [ $# -eq 0 ]; then
 	build_odin debug
 	run_demo
-	exit 0
-fi
-
-if [[ $# -eq 1 ]]; then
+elif [ $# -eq 1 ]; then
 	case $1 in
 	report)
-		if [[ ! -f "./odin" ]]; then
-			build_odin debug
-		fi
-
+		[ ! -f "./odin" ] && build_odin debug
 		./odin report
-		exit 0
 		;;
 	*)
 		build_odin $1
 		;;
 	esac
-
 	run_demo
-	exit 0
 else
-	panic "Too many arguments!"
+	error "Too many arguments!"
 fi

ci/upload_create_nightly.sh

@@ -1,5 +1,7 @@
 #!/bin/bash
 
+set -e
+
 bucket=$1
 platform=$2
 artifact=$3
@@ -9,5 +11,15 @@ filename="odin-$platform-nightly+$now.zip"
 
 echo "Creating archive $filename from $artifact and uploading to $bucket"
 
-7z a -bd "output/$filename" -r "$artifact"
-b2 upload-file --noProgress "$bucket" "output/$filename" "nightly/$filename"
+# If this is already zipped up (done before artifact upload to keep permissions in tact), just move it.
+if [ "${artifact: -4}" == ".zip" ]
+then
+	echo "Artifact already a zip"
+	mkdir -p "output"
+	mv "$artifact" "output/$filename"
+else
+	echo "Artifact needs to be zipped"
+	7z a -bd "output/$filename" -r "$artifact"
+fi
+
+b2 upload-file --noProgress "$bucket" "output/$filename" "nightly/$filename"

core/bufio/reader.odin

@@ -118,9 +118,7 @@ reader_peek :: proc(b: ^Reader, n: int) -> (data: []byte, err: io.Error) {
 	b.last_rune_size = -1
 
 	for b.w-b.r < n && b.w-b.r < len(b.buf) && b.err == nil {
-		if fill_err := _reader_read_new_chunk(b); fill_err != nil {
-			return nil, fill_err
-		}
+		_reader_read_new_chunk(b) or_return
 	}
 
 	if n > len(b.buf) {
@@ -156,9 +154,7 @@ reader_discard :: proc(b: ^Reader, n: int) -> (discarded: int, err: io.Error) {
 	for {
 		skip := reader_buffered(b)
 		if skip == 0 {
-			if fill_err := _reader_read_new_chunk(b); fill_err != nil {
-				return 0, fill_err
-			}
+			_reader_read_new_chunk(b) or_return
 			skip = reader_buffered(b)
 		}
 		skip = min(skip, remaining)
@@ -223,20 +219,18 @@ reader_read :: proc(b: ^Reader, p: []byte) -> (n: int, err: io.Error) {
 
 // reader_read_byte reads and returns a single byte
 // If no byte is available, it return an error
-reader_read_byte :: proc(b: ^Reader) -> (byte, io.Error) {
+reader_read_byte :: proc(b: ^Reader) -> (c: byte, err: io.Error) {
 	b.last_rune_size = -1
 	for b.r == b.w {
 		if b.err != nil {
 			return 0, _reader_consume_err(b)
 		}
-		if err := _reader_read_new_chunk(b); err != nil {
-			return 0, err
-		}
+		_reader_read_new_chunk(b) or_return
 	}
-	c := b.buf[b.r]
+	c = b.buf[b.r]
 	b.r += 1
 	b.last_byte = int(c)
-	return c, nil
+	return
 }
 
 // reader_unread_byte unreads the last byte. Only the most recently read byte can be unread
@@ -264,15 +258,12 @@ reader_read_rune :: proc(b: ^Reader) -> (r: rune, size: int, err: io.Error) {
 	    !utf8.full_rune(b.buf[b.r:b.w]) &&
 	    b.err == nil &&
 	    b.w-b.w < len(b.buf) {
-		if err = _reader_read_new_chunk(b); err != nil {
-			return
-		}
+		_reader_read_new_chunk(b) or_return
 	}
 
 	b.last_rune_size = -1
 	if b.r == b.w {
-		err = _reader_consume_err(b)
-		return
+		return 0, 0, _reader_consume_err(b)
 	}
 	r, size = rune(b.buf[b.r]), 1
 	if r >= utf8.RUNE_SELF {
@@ -305,27 +296,20 @@ reader_write_to :: proc(b: ^Reader, w: io.Writer) -> (n: i64, err: io.Error) {
 		return i64(n), err
 	}
 
-	n, err = write_buf(b, w)
-	if err != nil {
-		return
-	}
+	n = write_buf(b, w) or_return
 
 	m: i64
 	if b.w-b.r < len(b.buf) {
-		if err = _reader_read_new_chunk(b); err != nil {
-			return
-		}
+		_reader_read_new_chunk(b) or_return
 	}
 
 	for b.r < b.w {
 		m, err = write_buf(b, w)
-		n += m
+		n += m // this needs to be done before returning
 		if err != nil {
 			return
 		}
-		if err = _reader_read_new_chunk(b); err != nil {
-			return
-		}
+		_reader_read_new_chunk(b) or_return
 	}
 
 	if b.err == .EOF {
@@ -403,9 +387,7 @@ reader_read_slice :: proc(b: ^Reader, delim: byte) -> (line: []byte, err: io.Err
 
 		s = b.w - b.r
 
-		if err = _reader_read_new_chunk(b); err != nil {
-			break
-		}
+		_reader_read_new_chunk(b) or_break
 	}
 
 	if i := len(line)-1; i >= 0 {

core/bufio/scanner.odin

@@ -4,7 +4,7 @@ import "core:bytes"
 import "core:io"
 import "core:mem"
 import "core:unicode/utf8"
-import "core:intrinsics"
+import "base:intrinsics"
 
 // Extra errors returns by scanning procedures
 Scanner_Extra_Error :: enum i32 {

core/bufio/writer.odin

@@ -159,7 +159,7 @@ writer_write_rune :: proc(b: ^Writer, r: rune) -> (size: int, err: io.Error) {
 	return
 }
 
-// writer_write writes a string into the buffer
+// writer_write_string writes a string into the buffer
 // It returns the number of bytes written
 // If n < len(p), it will return an error explaining why the write is short
 writer_write_string :: proc(b: ^Writer, s: string) -> (int, io.Error) {

core/bytes/buffer.odin

@@ -113,8 +113,11 @@ _buffer_grow :: proc(b: ^Buffer, n: int) -> int {
 	if i, ok := _buffer_try_grow(b, n); ok {
 		return i
 	}
+
 	if b.buf == nil && n <= SMALL_BUFFER_SIZE {
-		b.buf = make([dynamic]byte, n, SMALL_BUFFER_SIZE)
+		// Fixes #2756 by preserving allocator if already set on Buffer via init_buffer_allocator
+		reserve(&b.buf, SMALL_BUFFER_SIZE)
+		resize(&b.buf, n)
 		return 0
 	}
 

core/bytes/bytes.odin

@@ -895,7 +895,7 @@ split_multi_iterator :: proc(s: ^[]byte, substrs: [][]byte, skip_empty := false)
 
 
 
-// scrub scruvs invalid utf-8 characters and replaces them with the replacement string
+// Scrubs invalid utf-8 characters and replaces them with the replacement string
 // Adjacent invalid bytes are only replaced once
 scrub :: proc(s: []byte, replacement: []byte, allocator := context.allocator) -> []byte {
 	str := s

core/c/c.odin

@@ -1,6 +1,6 @@
 package c
 
-import builtin "core:builtin"
+import builtin "base:builtin"
 
 char           :: builtin.u8  // assuming -funsigned-char
 

core/c/frontend/tokenizer/tokenizer.odin

@@ -127,7 +127,7 @@ advance_rune :: proc(t: ^Tokenizer) {
 }
 
 advance_rune_n :: proc(t: ^Tokenizer, n: int) {
-	for in 0..<n {
+	for _ in 0..<n {
 		advance_rune(t)
 	}
 }

core/c/libc/complex.odin

@@ -67,7 +67,7 @@ foreign libc {
 	crealf  :: proc(z: complex_float) -> float ---
 }
 
-import builtin "core:builtin"
+import builtin "base:builtin"
 
 complex_float  :: distinct builtin.complex64
 complex_double :: distinct builtin.complex128

core/c/libc/math.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.12 Mathematics
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"

core/c/libc/setjmp.odin

@@ -63,4 +63,4 @@ foreign libc {
 // strictly conformant C implementation is 16 on the platforms we care about.
 // The choice of 4096 bytes for storage of this type is more than enough on all
 // relevant platforms.
-jmp_buf :: struct #align 16 { _: [4096]char, }
+jmp_buf :: struct #align(16) { _: [4096]char, }

core/c/libc/stdarg.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.16 Variable arguments
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @(private="file")
 @(default_calling_convention="none")
@@ -18,7 +18,7 @@ foreign _ {
 // strictly conformant C implementation is 16 on the platforms we care about.
 // The choice of 4096 bytes for storage of this type is more than enough on all
 // relevant platforms.
-va_list :: struct #align 16 {
+va_list :: struct #align(16) {
 	_: [4096]u8,
 }
 

core/c/libc/stdatomic.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.17 Atomics
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 ATOMIC_BOOL_LOCK_FREE     :: true
 ATOMIC_CHAR_LOCK_FREE     :: true

core/c/libc/stdio.odin

@@ -1,5 +1,7 @@
 package libc
 
+import "core:io"
+
 when ODIN_OS == .Windows {
 	foreign import libc {
 		"system:libucrt.lib",
@@ -182,7 +184,7 @@ foreign libc {
 	fscanf    :: proc(stream: ^FILE, format: cstring, #c_vararg args: ..any) -> int ---
 	printf    :: proc(format: cstring, #c_vararg args: ..any) -> int ---
 	scanf     :: proc(format: cstring, #c_vararg args: ..any) -> int ---
-	snprintf  :: proc(s: [^]char, format: cstring, #c_vararg args: ..any) -> int ---
+	snprintf  :: proc(s: [^]char, n: size_t, format: cstring, #c_vararg args: ..any) -> int ---
 	sscanf    :: proc(s, format: cstring, #c_vararg args: ..any) -> int ---
 	vfprintf  :: proc(stream: ^FILE, format: cstring, arg: ^va_list) -> int ---
 	vfscanf   :: proc(stream: ^FILE, format: cstring, arg: ^va_list) -> int ---
@@ -218,3 +220,102 @@ foreign libc {
 	ferror    :: proc(stream: ^FILE) -> int ---
 	perror    :: proc(s: cstring) ---
 }
+
+to_stream :: proc(file: ^FILE) -> io.Stream {
+	stream_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+		unknown_or_eof :: proc(f: ^FILE) -> io.Error {
+			switch {
+			case ferror(f) != 0:
+				return .Unknown
+			case feof(f) != 0:
+				return .EOF
+			case:
+				return nil
+			}
+		}
+
+		file := (^FILE)(stream_data)
+		switch mode {
+		case .Close:
+			if fclose(file) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Flush:
+			if fflush(file) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Read:
+			n = i64(fread(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Read_At:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+			if fseek(file, long(offset), SEEK_SET) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			defer fseek(file, long(curr), SEEK_SET)
+
+			n = i64(fread(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+		
+		case .Write:
+			n = i64(fwrite(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Write_At:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+			if fseek(file, long(offset), SEEK_SET) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			defer fseek(file, long(curr), SEEK_SET)
+
+			n = i64(fwrite(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Seek:
+			if fseek(file, long(offset), int(whence)) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+		
+		case .Size:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+			defer fseek(file, curr, SEEK_SET)
+
+			if fseek(file, 0, SEEK_END) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			n = i64(ftell(file))
+			if n == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Destroy:
+			return 0, .Empty
+		
+		case .Query:
+			return io.query_utility({ .Close, .Flush, .Read, .Read_At, .Write, .Write_At, .Seek, .Size })
+		}
+		return
+	}
+
+	return {
+		data      = file,
+		procedure = stream_proc,
+	}
+}

core/c/libc/string.odin

@@ -1,6 +1,6 @@
 package libc
 
-import "core:runtime"
+import "base:runtime"
 
 // 7.24 String handling
 

core/c/libc/time.odin

@@ -35,12 +35,12 @@ when ODIN_OS == .Windows {
 	clock_t        :: distinct long
 	time_t         :: distinct i64
 
-	timespec :: struct #align 8 {
+	timespec :: struct #align(8) {
 		tv_sec:  time_t,
 		tv_nsec: long,
 	}
 
-	tm :: struct #align 8 {
+	tm :: struct #align(8) {
 		tm_sec, tm_min, tm_hour, tm_mday, tm_mon, tm_year, tm_wday, tm_yday, tm_isdst: int,
 	}
 }

core/c/libc/types.odin

@@ -2,6 +2,8 @@ package libc
 
 import "core:c"
 
+#assert(!ODIN_NO_CRT, `"core:c/libc" cannot be imported when '-no-crt' is used`)
+
 char           :: c.char // assuming -funsigned-char
 
 schar          :: c.schar

core/c/libc/wchar.odin

@@ -98,7 +98,7 @@ foreign libc {
 }
 
 // Large enough and aligned enough for any wide-spread in-use libc.
-mbstate_t :: struct #align 16 { _: [32]char, }
+mbstate_t :: struct #align(16) { _: [32]char, }
 
 // Odin does not have default argument promotion so the need for a separate type
 // here isn't necessary, though make it distinct just to be safe.

core/compress/common.odin

@@ -12,7 +12,7 @@ package compress
 
 import "core:io"
 import "core:bytes"
-import "core:runtime"
+import "base:runtime"
 
 /*
 	These settings bound how much compression algorithms will allocate for their output buffer.
@@ -20,10 +20,9 @@ import "core:runtime"
 
 */
 
-/*
-	When a decompression routine doesn't stream its output, but writes to a buffer,
-	we pre-allocate an output buffer to speed up decompression. The default is 1 MiB.
-*/
+
+// When a decompression routine doesn't stream its output, but writes to a buffer,
+// we pre-allocate an output buffer to speed up decompression. The default is 1 MiB.
 COMPRESS_OUTPUT_ALLOCATE_MIN :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MIN, 1 << 20))
 
 /*
@@ -34,16 +33,14 @@ COMPRESS_OUTPUT_ALLOCATE_MIN :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MIN, 1 << 2
 
 */
 when size_of(uintptr) == 8 {
-	/*
-		For 64-bit platforms, we set the default max buffer size to 4 GiB,
-		which is GZIP and PKZIP's max payload size.
-	*/	
+
+        // For 64-bit platforms, we set the default max buffer size to 4 GiB,
+        // which is GZIP and PKZIP's max payload size.
 	COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 32))
 } else {
-	/*
-		For 32-bit platforms, we set the default max buffer size to 512 MiB.
-	*/
-	COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 29))
+	
+	// For 32-bit platforms, we set the default max buffer size to 512 MiB.
+        COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 29))
 }
 
 
@@ -69,9 +66,8 @@ General_Error :: enum {
 	Incompatible_Options,
 	Unimplemented,
 
-	/*
-		Memory errors
-	*/
+	// Memory errors
+
 	Allocation_Failed,
 	Resize_Failed,
 }
@@ -86,17 +82,16 @@ GZIP_Error :: enum {
 	Payload_Length_Invalid,
 	Payload_CRC_Invalid,
 
-	/*
-		GZIP's payload can be a maximum of max(u32le), or 4 GiB.
-		If you tell it you expect it to contain more, that's obviously an error.
-	*/
-	Payload_Size_Exceeds_Max_Payload,
-	/*
-		For buffered instead of streamed output, the payload size can't exceed
-		the max set by the `COMPRESS_OUTPUT_ALLOCATE_MAX` switch in compress/common.odin.
+	// GZIP's payload can be a maximum of max(u32le), or 4 GiB.
+	// If you tell it you expect it to contain more, that's obviously an error.
+
+	Payload_Size_Exceeds_Max_Payload,
+
+	// For buffered instead of streamed output, the payload size can't exceed
+	// the max set by the `COMPRESS_OUTPUT_ALLOCATE_MAX` switch in compress/common.odin.
+	//
+	// You can tweak this setting using `-define:COMPRESS_OUTPUT_ALLOCATE_MAX=size_in_bytes`
 
-		You can tweak this setting using `-define:COMPRESS_OUTPUT_ALLOCATE_MAX=size_in_bytes`
-	*/
 	Output_Exceeds_COMPRESS_OUTPUT_ALLOCATE_MAX,
 
 }
@@ -137,9 +132,8 @@ Context_Memory_Input :: struct #packed {
 	code_buffer:       u64,
 	num_bits:          u64,
 
-	/*
-		If we know the data size, we can optimize the reads and writes.
-	*/
+	// If we know the data size, we can optimize the reads and writes.
+
 	size_packed:       i64,
 	size_unpacked:     i64,
 }
@@ -159,18 +153,16 @@ Context_Stream_Input :: struct #packed {
 	code_buffer:       u64,
 	num_bits:          u64,
 
-	/*
-		If we know the data size, we can optimize the reads and writes.
-	*/
+	// If we know the data size, we can optimize the reads and writes.
+
 	size_packed:       i64,
 	size_unpacked:     i64,
 
-	/*
-		Flags:
-			`input_fully_in_memory`
-				true  = This tells us we read input from `input_data` exclusively. [] = EOF.
-				false = Try to refill `input_data` from the `input` stream.
-	*/
+	// Flags:
+	// `input_fully_in_memory`
+	//   true  = This tells us we read input from `input_data` exclusively. [] = EOF.
+	//   false = Try to refill `input_data` from the `input` stream.
+
 	input_fully_in_memory: b8,
 
 	padding: [1]u8,
@@ -214,26 +206,18 @@ read_slice_from_memory :: #force_inline proc(z: ^Context_Memory_Input, size: int
 @(optimization_mode="speed")
 read_slice_from_stream :: #force_inline proc(z: ^Context_Stream_Input, size: int) -> (res: []u8, err: io.Error) {
 	// TODO: REMOVE ALL USE OF context.temp_allocator here
-	// the is literally no need for it
+	// there is literally no need for it
 	b := make([]u8, size, context.temp_allocator)
-	_, e := io.read(z.input, b[:])
-	if e == .None {
-		return b, .None
-	}
-
-	return []u8{}, e
+	_ = io.read(z.input, b[:]) or_return
+	return b, nil
 }
 
 read_slice :: proc{read_slice_from_memory, read_slice_from_stream}
 
 @(optimization_mode="speed")
 read_data :: #force_inline proc(z: ^$C, $T: typeid) -> (res: T, err: io.Error) {
-	b, e := read_slice(z, size_of(T))
-	if e == .None {
-		return (^T)(&b[0])^, .None
-	}
-
-	return T{}, e
+	b := read_slice(z, size_of(T)) or_return
+	return (^T)(&b[0])^, nil
 }
 
 @(optimization_mode="speed")
@@ -250,20 +234,14 @@ read_u8_from_memory :: #force_inline proc(z: ^Context_Memory_Input) -> (res: u8,
 
 @(optimization_mode="speed")
 read_u8_from_stream :: #force_inline proc(z: ^Context_Stream_Input) -> (res: u8, err: io.Error) {
-	b, e := read_slice_from_stream(z, 1)
-	if e == .None {
-		return b[0], .None
-	}
-
-	return 0, e
+	b := read_slice_from_stream(z, 1) or_return
+	return b[0], nil
 }
 
 read_u8 :: proc{read_u8_from_memory, read_u8_from_stream}
 
-/*
-	You would typically only use this at the end of Inflate, to drain bits from the code buffer
-	preferentially.
-*/
+// You would typically only use this at the end of Inflate, to drain bits from the code buffer
+// preferentially.
 @(optimization_mode="speed")
 read_u8_prefer_code_buffer_lsb :: #force_inline proc(z: ^$C) -> (res: u8, err: io.Error) {
 	if z.num_bits >= 8 {
@@ -320,12 +298,9 @@ peek_data_from_stream :: #force_inline proc(z: ^Context_Stream_Input, $T: typeid
 	size :: size_of(T)
 
 	// Get current position to read from.
-	curr, e1 := z.input->impl_seek(0, .Current)
-	if e1 != .None {
-		return T{}, e1
-	}
-	r, e2 := io.to_reader_at(z.input)
-	if !e2 {
+	curr := z.input->impl_seek(0, .Current) or_return
+	r, e1 := io.to_reader_at(z.input)
+	if !e1 {
 		return T{}, .Empty
 	}
 	when size <= 128 {
@@ -333,8 +308,8 @@ peek_data_from_stream :: #force_inline proc(z: ^Context_Stream_Input, $T: typeid
 	} else {
 		b := make([]u8, size, context.temp_allocator)
 	}
-	_, e3 := io.read_at(r, b[:], curr)
-	if e3 != .None {
+	_, e2 := io.read_at(r, b[:], curr)
+	if e2 != .None {
 		return T{}, .Empty
 	}
 
@@ -347,16 +322,9 @@ peek_data_at_offset_from_stream :: #force_inline proc(z: ^Context_Stream_Input,
 	size :: size_of(T)
 
 	// Get current position to return to.
-	cur_pos, e1 := z.input->impl_seek(0, .Current)
-	if e1 != .None {
-		return T{}, e1
-	}
-
+	cur_pos := z.input->impl_seek(0, .Current) or_return
 	// Seek to offset.
-	pos, e2 := z.input->impl_seek(offset, .Start)
-	if e2 != .None {
-		return T{}, e2
-	}
+	pos := z.input->impl_seek(offset, .Start) or_return
 
 	r, e3 := io.to_reader_at(z.input)
 	if !e3 {
@@ -465,7 +433,7 @@ peek_bits_lsb_from_memory :: #force_inline proc(z: ^Context_Memory_Input, width:
 	if z.num_bits < u64(width) {
 		refill_lsb(z)
 	}
-	return u32(z.code_buffer & ~(~u64(0) << width))
+	return u32(z.code_buffer &~ (~u64(0) << width))
 }
 
 @(optimization_mode="speed")
@@ -473,7 +441,7 @@ peek_bits_lsb_from_stream :: #force_inline proc(z: ^Context_Stream_Input, width:
 	if z.num_bits < u64(width) {
 		refill_lsb(z)
 	}
-	return u32(z.code_buffer & ~(~u64(0) << width))
+	return u32(z.code_buffer &~ (~u64(0) << width))
 }
 
 peek_bits_lsb :: proc{peek_bits_lsb_from_memory, peek_bits_lsb_from_stream}
@@ -481,13 +449,13 @@ peek_bits_lsb :: proc{peek_bits_lsb_from_memory, peek_bits_lsb_from_stream}
 @(optimization_mode="speed")
 peek_bits_no_refill_lsb_from_memory :: #force_inline proc(z: ^Context_Memory_Input, width: u8) -> u32 {
 	assert(z.num_bits >= u64(width))
-	return u32(z.code_buffer & ~(~u64(0) << width))
+	return u32(z.code_buffer &~ (~u64(0) << width))
 }
 
 @(optimization_mode="speed")
 peek_bits_no_refill_lsb_from_stream :: #force_inline proc(z: ^Context_Stream_Input, width: u8) -> u32 {
 	assert(z.num_bits >= u64(width))
-	return u32(z.code_buffer & ~(~u64(0) << width))
+	return u32(z.code_buffer &~ (~u64(0) << width))
 }
 
 peek_bits_no_refill_lsb :: proc{peek_bits_no_refill_lsb_from_memory, peek_bits_no_refill_lsb_from_stream}

core/compress/gzip/gzip.odin

@@ -335,10 +335,8 @@ load_from_context :: proc(z: ^$C, buf: ^bytes.Buffer, known_gzip_size := -1, exp
 
 	// fmt.printf("GZIP: Expected Payload Size: %v\n", expected_output_size);
 
-	zlib_error := zlib.inflate_raw(z, expected_output_size=expected_output_size)
-	if zlib_error != nil {
-		return zlib_error
-	}
+	zlib.inflate_raw(z, expected_output_size=expected_output_size) or_return
+
 	/*
 		Read CRC32 using the ctx bit reader because zlib may leave bytes in there.
 	*/

core/compress/shoco/shoco.odin

@@ -11,7 +11,7 @@
 // package shoco is an implementation of the shoco short string compressor
 package shoco
 
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:compress"
 
 Shoco_Pack :: struct {

core/compress/zlib/zlib.odin

@@ -1,3 +1,4 @@
+//+vet !using-param
 package zlib
 
 /*
@@ -146,15 +147,7 @@ grow_buffer :: proc(buf: ^[dynamic]u8) -> (err: compress.Error) {
 		Double until we reach the maximum allowed.
 	*/
 	new_size := min(len(buf) << 1, compress.COMPRESS_OUTPUT_ALLOCATE_MAX)
-	resize(buf, new_size)
-	if len(buf) != new_size {
-		/*
-			Resize failed.
-		*/
-		return .Resize_Failed
-	}
-
-	return nil
+	return resize(buf, new_size)
 }
 
 /*
@@ -181,7 +174,7 @@ write_byte :: #force_inline proc(z: ^$C, c: u8) -> (err: io.Error) #no_bounds_ch
 }
 
 @(optimization_mode="speed")
-repl_byte :: proc(z: ^$C, count: u16, c: u8) -> (err: io.Error) 	#no_bounds_check {
+repl_byte :: proc(z: ^$C, count: u16, c: u8) -> (err: io.Error) #no_bounds_check {
 	/*
 		TODO(Jeroen): Once we have a magic ring buffer, we can just peek/write into it
 		without having to worry about wrapping, so no need for a temp allocation to give to
@@ -305,10 +298,10 @@ decode_huffman_slowpath :: proc(z: ^$C, t: ^Huffman_Table) -> (r: u16, err: Erro
 	code := u16(compress.peek_bits_lsb(z,16))
 
 	k := int(z_bit_reverse(code, 16))
-	s: u8
 
-	#no_bounds_check for s = HUFFMAN_FAST_BITS+1; ; {
-		if k < t.maxcode[s] {
+	s: u8 = HUFFMAN_FAST_BITS+1
+	for {
+		#no_bounds_check if k < t.maxcode[s] {
 			break
 		}
 		s += 1
@@ -509,8 +502,8 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 		/*
 			Try to pre-allocate the output buffer.
 		*/
-		reserve(&z.output.buf, expected_output_size)
-		resize (&z.output.buf, expected_output_size)
+		reserve(&z.output.buf, expected_output_size) or_return
+		resize (&z.output.buf, expected_output_size) or_return
 	}
 
 	if len(z.output.buf) != expected_output_size {
@@ -653,7 +646,7 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 	}
 
 	if int(z.bytes_written) != len(z.output.buf) {
-		resize(&z.output.buf, int(z.bytes_written))
+		resize(&z.output.buf, int(z.bytes_written)) or_return
 	}
 
 	return nil

core/container/avl/avl.odin

@@ -0,0 +1,678 @@
+/*
+package avl implements an AVL tree.
+
+The implementation is non-intrusive, and non-recursive.
+*/
+package container_avl
+
+import "base:intrinsics"
+import "base:runtime"
+import "core:slice"
+
+_ :: intrinsics
+_ :: runtime
+
+// Originally based on the CC0 implementation by Eric Biggers
+// See: https://github.com/ebiggers/avl_tree/
+
+// Direction specifies the traversal direction for a tree iterator.
+Direction :: enum i8 {
+	// Backward is the in-order backwards direction.
+	Backward = -1,
+	// Forward is the in-order forwards direction.
+	Forward  = 1,
+}
+
+// Ordering specifies order when inserting/finding values into the tree.
+Ordering :: slice.Ordering
+
+// Tree is an AVL tree.
+Tree :: struct($Value: typeid) {
+	// user_data is a parameter that will be passed to the on_remove
+	// callback.
+	user_data: rawptr,
+	// on_remove is an optional callback that can be called immediately
+	// after a node is removed from the tree.
+	on_remove: proc(value: Value, user_data: rawptr),
+
+	_root:           ^Node(Value),
+	_node_allocator: runtime.Allocator,
+	_cmp_fn:         proc(a, b: Value) -> Ordering,
+	_size:           int,
+}
+
+// Node is an AVL tree node.
+//
+// WARNING: It is unsafe to mutate value if the node is part of a tree
+// if doing so will alter the Node's sort position relative to other
+// elements in the tree.
+Node :: struct($Value: typeid) {
+	value: Value,
+
+	_parent:  ^Node(Value),
+	_left:    ^Node(Value),
+	_right:   ^Node(Value),
+	_balance: i8,
+}
+
+// Iterator is a tree iterator.
+//
+// WARNING: It is unsafe to modify the tree while iterating, except via
+// the iterator_remove method.
+Iterator :: struct($Value: typeid) {
+	_tree:        ^Tree(Value),
+	_cur:         ^Node(Value),
+	_next:        ^Node(Value),
+	_direction:   Direction,
+	_called_next: bool,
+}
+
+// init initializes a tree.
+init :: proc {
+	init_ordered,
+	init_cmp,
+}
+
+// init_cmp initializes a tree.
+init_cmp :: proc(
+	t: ^$T/Tree($Value),
+	cmp_fn: proc(a, b: Value) -> Ordering,
+	node_allocator := context.allocator,
+) {
+	t._root = nil
+	t._node_allocator = node_allocator
+	t._cmp_fn = cmp_fn
+	t._size = 0
+}
+
+// init_ordered initializes a tree containing ordered items, with
+// a comparison function that results in an ascending order sort.
+init_ordered :: proc(
+	t: ^$T/Tree($Value),
+	node_allocator := context.allocator,
+) where intrinsics.type_is_ordered_numeric(Value) {
+	init_cmp(t, slice.cmp_proc(Value), node_allocator)
+}
+
+// destroy de-initializes a tree.
+destroy :: proc(t: ^$T/Tree($Value), call_on_remove: bool = true) {
+	iter := iterator(t, Direction.Forward)
+	for _ in iterator_next(&iter) {
+		iterator_remove(&iter, call_on_remove)
+	}
+}
+
+// len returns the number of elements in the tree.
+len :: proc "contextless" (t: ^$T/Tree($Value)) -> int {
+	return t._size
+}
+
+// first returns the first node in the tree (in-order) or nil iff
+// the tree is empty.
+first :: proc "contextless" (t: ^$T/Tree($Value)) -> ^Node(Value) {
+	return tree_first_or_last_in_order(t, Direction.Backward)
+}
+
+// last returns the last element in the tree (in-order) or nil iff
+// the tree is empty.
+last :: proc "contextless" (t: ^$T/Tree($Value)) -> ^Node(Value) {
+	return tree_first_or_last_in_order(t, Direction.Forward)
+}
+
+// find finds the value in the tree, and returns the corresponding
+// node or nil iff the value is not present.
+find :: proc(t: ^$T/Tree($Value), value: Value) -> ^Node(Value) {
+	cur := t._root
+	descend_loop: for cur != nil {
+		switch t._cmp_fn(value, cur.value) {
+		case .Less:
+			cur = cur._left
+		case .Greater:
+			cur = cur._right
+		case .Equal:
+			break descend_loop
+		}
+	}
+
+	return cur
+}
+
+// find_or_insert attempts to insert the value into the tree, and returns
+// the node, a boolean indicating if the value was inserted, and the
+// node allocator error if relevant.  If the value is already
+// present, the existing node is returned un-altered.
+find_or_insert :: proc(
+	t: ^$T/Tree($Value),
+	value: Value,
+) -> (
+	n: ^Node(Value),
+	inserted: bool,
+	err: runtime.Allocator_Error,
+) {
+	n_ptr := &t._root
+	for n_ptr^ != nil {
+		n = n_ptr^
+		switch t._cmp_fn(value, n.value) {
+		case .Less:
+			n_ptr = &n._left
+		case .Greater:
+			n_ptr = &n._right
+		case .Equal:
+			return
+		}
+	}
+
+	parent := n
+	n = new(Node(Value), t._node_allocator) or_return
+	n.value = value
+	n._parent = parent
+	n_ptr^ = n
+	tree_rebalance_after_insert(t, n)
+
+	t._size += 1
+	inserted = true
+
+	return
+}
+
+// remove removes a node or value from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove :: proc {
+	remove_value,
+	remove_node,
+}
+
+// remove_value removes a value from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove_value :: proc(t: ^$T/Tree($Value), value: Value, call_on_remove: bool = true) -> bool {
+	n := find(t, value)
+	if n == nil {
+		return false
+	}
+	return remove_node(t, n, call_on_remove)
+}
+
+// remove_node removes a node from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove_node :: proc(t: ^$T/Tree($Value), node: ^Node(Value), call_on_remove: bool = true) -> bool {
+	if node._parent == node || (node._parent == nil && t._root != node) {
+		return false
+	}
+	defer {
+		if call_on_remove && t.on_remove != nil {
+			t.on_remove(node.value, t.user_data)
+		}
+		free(node, t._node_allocator)
+	}
+
+	parent: ^Node(Value)
+	left_deleted: bool
+
+	t._size -= 1
+	if node._left != nil && node._right != nil {
+		parent, left_deleted = tree_swap_with_successor(t, node)
+	} else {
+		child := node._left
+		if child == nil {
+			child = node._right
+		}
+		parent = node._parent
+		if parent != nil {
+			if node == parent._left {
+				parent._left = child
+				left_deleted = true
+			} else {
+				parent._right = child
+				left_deleted = false
+			}
+			if child != nil {
+				child._parent = parent
+			}
+		} else {
+			if child != nil {
+				child._parent = parent
+			}
+			t._root = child
+			node_reset(node)
+			return true
+		}
+	}
+
+	for {
+		if left_deleted {
+			parent = tree_handle_subtree_shrink(t, parent, +1, &left_deleted)
+		} else {
+			parent = tree_handle_subtree_shrink(t, parent, -1, &left_deleted)
+		}
+		if parent == nil {
+			break
+		}
+	}
+	node_reset(node)
+
+	return true
+}
+
+// iterator returns a tree iterator in the specified direction.
+iterator :: proc "contextless" (t: ^$T/Tree($Value), direction: Direction) -> Iterator(Value) {
+	it: Iterator(Value)
+	it._tree = transmute(^Tree(Value))t
+	it._direction = direction
+
+	iterator_first(&it)
+
+	return it
+}
+
+// iterator_from_pos returns a tree iterator in the specified direction,
+// spanning the range [pos, last] (inclusive).
+iterator_from_pos :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	pos: ^Node(Value),
+	direction: Direction,
+) -> Iterator(Value) {
+	it: Iterator(Value)
+	it._tree = transmute(^Tree(Value))t
+	it._direction = direction
+	it._next = nil
+	it._called_next = false
+
+	if it._cur = pos; pos != nil {
+		it._next = node_next_or_prev_in_order(it._cur, it._direction)
+	}
+
+	return it
+}
+
+// iterator_get returns the node currently pointed to by the iterator,
+// or nil iff the node has been removed, the tree is empty, or the end
+// of the tree has been reached.
+iterator_get :: proc "contextless" (it: ^$I/Iterator($Value)) -> ^Node(Value) {
+	return it._cur
+}
+
+// iterator_remove removes the node currently pointed to by the iterator,
+// and returns true iff the removal was successful.  Semantics are the
+// same as the Tree remove.
+iterator_remove :: proc(it: ^$I/Iterator($Value), call_on_remove: bool = true) -> bool {
+	if it._cur == nil {
+		return false
+	}
+
+	ok := remove_node(it._tree, it._cur, call_on_remove)
+	if ok {
+		it._cur = nil
+	}
+
+	return ok
+}
+
+// iterator_next advances the iterator and returns the (node, true) or
+// or (nil, false) iff the end of the tree has been reached.
+//
+// Note: The first call to iterator_next will return the first node instead
+// of advancing the iterator.
+iterator_next :: proc "contextless" (it: ^$I/Iterator($Value)) -> (^Node(Value), bool) {
+	// This check is needed so that the first element gets returned from
+	// a brand-new iterator, and so that the somewhat contrived case where
+	// iterator_remove is called before the first call to iterator_next
+	// returns the correct value.
+	if !it._called_next {
+		it._called_next = true
+
+		// There can be the contrived case where iterator_remove is
+		// called before ever calling iterator_next, which needs to be
+		// handled as an actual call to next.
+		//
+		// If this happens it._cur will be nil, so only return the
+		// first value, if it._cur is valid.
+		if it._cur != nil {
+			return it._cur, true
+		}
+	}
+
+	if it._next == nil {
+		return nil, false
+	}
+
+	it._cur = it._next
+	it._next = node_next_or_prev_in_order(it._cur, it._direction)
+
+	return it._cur, true
+}
+
+@(private)
+tree_first_or_last_in_order :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	direction: Direction,
+) -> ^Node(Value) {
+	first, sign := t._root, i8(direction)
+	if first != nil {
+		for {
+			tmp := node_get_child(first, +sign)
+			if tmp == nil {
+				break
+			}
+			first = tmp
+		}
+	}
+
+	return first
+}
+
+@(private)
+tree_replace_child :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	parent, old_child, new_child: ^Node(Value),
+) {
+	if parent != nil {
+		if old_child == parent._left {
+			parent._left = new_child
+		} else {
+			parent._right = new_child
+		}
+	} else {
+		t._root = new_child
+	}
+}
+
+@(private)
+tree_rotate :: proc "contextless" (t: ^$T/Tree($Value), a: ^Node(Value), sign: i8) {
+	b := node_get_child(a, -sign)
+	e := node_get_child(b, +sign)
+	p := a._parent
+
+	node_set_child(a, -sign, e)
+	a._parent = b
+
+	node_set_child(b, +sign, a)
+	b._parent = p
+
+	if e != nil {
+		e._parent = a
+	}
+
+	tree_replace_child(t, p, a, b)
+}
+
+@(private)
+tree_double_rotate :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	b, a: ^Node(Value),
+	sign: i8,
+) -> ^Node(Value) {
+	e := node_get_child(b, +sign)
+	f := node_get_child(e, -sign)
+	g := node_get_child(e, +sign)
+	p := a._parent
+	e_bal := e._balance
+
+	node_set_child(a, -sign, g)
+	a_bal := -e_bal
+	if sign * e_bal >= 0 {
+		a_bal = 0
+	}
+	node_set_parent_balance(a, e, a_bal)
+
+	node_set_child(b, +sign, f)
+	b_bal := -e_bal
+	if sign * e_bal <= 0 {
+		b_bal = 0
+	}
+	node_set_parent_balance(b, e, b_bal)
+
+	node_set_child(e, +sign, a)
+	node_set_child(e, -sign, b)
+	node_set_parent_balance(e, p, 0)
+
+	if g != nil {
+		g._parent = a
+	}
+
+	if f != nil {
+		f._parent = b
+	}
+
+	tree_replace_child(t, p, a, e)
+
+	return e
+}
+
+@(private)
+tree_handle_subtree_growth :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	node, parent: ^Node(Value),
+	sign: i8,
+) -> bool {
+	old_balance_factor := parent._balance
+	if old_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return false
+	}
+
+	new_balance_factor := old_balance_factor + sign
+	if new_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return true
+	}
+
+	if sign * node._balance > 0 {
+		tree_rotate(t, parent, -sign)
+		node_adjust_balance_factor(parent, -sign)
+		node_adjust_balance_factor(node, -sign)
+	} else {
+		tree_double_rotate(t, node, parent, -sign)
+	}
+
+	return true
+}
+
+@(private)
+tree_rebalance_after_insert :: proc "contextless" (t: ^$T/Tree($Value), inserted: ^Node(Value)) {
+	node, parent := inserted, inserted._parent
+	switch {
+	case parent == nil:
+		return
+	case node == parent._left:
+		node_adjust_balance_factor(parent, -1)
+	case:
+		node_adjust_balance_factor(parent, +1)
+	}
+
+	if parent._balance == 0 {
+		return
+	}
+
+	for done := false; !done; {
+		node = parent
+		if parent = node._parent; parent == nil {
+			return
+		}
+
+		if node == parent._left {
+			done = tree_handle_subtree_growth(t, node, parent, -1)
+		} else {
+			done = tree_handle_subtree_growth(t, node, parent, +1)
+		}
+	}
+}
+
+@(private)
+tree_swap_with_successor :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	x: ^Node(Value),
+) -> (
+	^Node(Value),
+	bool,
+) {
+	ret: ^Node(Value)
+	left_deleted: bool
+
+	y := x._right
+	if y._left == nil {
+		ret = y
+	} else {
+		q: ^Node(Value)
+
+		for {
+			q = y
+			if y = y._left; y._left == nil {
+				break
+			}
+		}
+
+		if q._left = y._right; q._left != nil {
+			q._left._parent = q
+		}
+		y._right = x._right
+		x._right._parent = y
+		ret = q
+		left_deleted = true
+	}
+
+	y._left = x._left
+	x._left._parent = y
+
+	y._parent = x._parent
+	y._balance = x._balance
+
+	tree_replace_child(t, x._parent, x, y)
+
+	return ret, left_deleted
+}
+
+@(private)
+tree_handle_subtree_shrink :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	parent: ^Node(Value),
+	sign: i8,
+	left_deleted: ^bool,
+) -> ^Node(Value) {
+	old_balance_factor := parent._balance
+	if old_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return nil
+	}
+
+	node: ^Node(Value)
+	new_balance_factor := old_balance_factor + sign
+	if new_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		node = parent
+	} else {
+		node = node_get_child(parent, sign)
+		if sign * node._balance >= 0 {
+			tree_rotate(t, parent, -sign)
+			if node._balance == 0 {
+				node_adjust_balance_factor(node, -sign)
+				return nil
+			}
+			node_adjust_balance_factor(parent, -sign)
+			node_adjust_balance_factor(node, -sign)
+		} else {
+			node = tree_double_rotate(t, node, parent, -sign)
+		}
+	}
+
+	parent := parent
+	if parent = node._parent; parent != nil {
+		left_deleted^ = node == parent._left
+	}
+	return parent
+}
+
+@(private)
+node_reset :: proc "contextless" (n: ^Node($Value)) {
+	// Mostly pointless as n will be deleted after this is called, but
+	// attempt to be able to catch cases of n not being in the tree.
+	n._parent = n
+	n._left = nil
+	n._right = nil
+	n._balance = 0
+}
+
+@(private)
+node_set_parent_balance :: #force_inline proc "contextless" (
+	n, parent: ^Node($Value),
+	balance: i8,
+) {
+	n._parent = parent
+	n._balance = balance
+}
+
+@(private)
+node_get_child :: #force_inline proc "contextless" (n: ^Node($Value), sign: i8) -> ^Node(Value) {
+	if sign < 0 {
+		return n._left
+	}
+	return n._right
+}
+
+@(private)
+node_next_or_prev_in_order :: proc "contextless" (
+	n: ^Node($Value),
+	direction: Direction,
+) -> ^Node(Value) {
+	next, tmp: ^Node(Value)
+	sign := i8(direction)
+
+	if next = node_get_child(n, +sign); next != nil {
+		for {
+			tmp = node_get_child(next, -sign)
+			if tmp == nil {
+				break
+			}
+			next = tmp
+		}
+	} else {
+		tmp, next = n, n._parent
+		for next != nil && tmp == node_get_child(next, +sign) {
+			tmp, next = next, next._parent
+		}
+	}
+	return next
+}
+
+@(private)
+node_set_child :: #force_inline proc "contextless" (
+	n: ^Node($Value),
+	sign: i8,
+	child: ^Node(Value),
+) {
+	if sign < 0 {
+		n._left = child
+	} else {
+		n._right = child
+	}
+}
+
+@(private)
+node_adjust_balance_factor :: #force_inline proc "contextless" (n: ^Node($Value), amount: i8) {
+	n._balance += amount
+}
+
+@(private)
+iterator_first :: proc "contextless" (it: ^Iterator($Value)) {
+	// This is private because behavior when the user manually calls
+	// iterator_first followed by iterator_next is unintuitive, since
+	// the first call to iterator_next MUST return the first node
+	// instead of advancing so that `for node in iterator_next(&next)`
+	// works as expected.
+
+	switch it._direction {
+	case .Forward:
+		it._cur = tree_first_or_last_in_order(it._tree, .Backward)
+	case .Backward:
+		it._cur = tree_first_or_last_in_order(it._tree, .Forward)
+	}
+
+	it._next = nil
+	it._called_next = false
+
+	if it._cur != nil {
+		it._next = node_next_or_prev_in_order(it._cur, it._direction)
+	}
+}

core/container/bit_array/bit_array.odin

@@ -1,6 +1,6 @@
 package dynamic_bit_array
 
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:mem"
 
 /*

core/container/intrusive/list/intrusive_list.odin

@@ -1,6 +1,6 @@
 package container_intrusive_list
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 // An intrusive doubly-linked list
 //
@@ -20,10 +20,10 @@ List :: struct {
 
 
 Node :: struct {
-	next, prev: ^Node,
+	prev, next: ^Node,
 }
 
-push_front :: proc(list: ^List, node: ^Node) {
+push_front :: proc "contextless" (list: ^List, node: ^Node) {
 	if list.head != nil {
 		list.head.prev = node
 		node.prev, node.next = nil, list.head
@@ -34,7 +34,7 @@ push_front :: proc(list: ^List, node: ^Node) {
 	}
 }
 
-push_back :: proc(list: ^List, node: ^Node) {
+push_back :: proc "contextless" (list: ^List, node: ^Node) {
 	if list.tail != nil {
 		list.tail.next = node
 		node.prev, node.next = list.tail, nil
@@ -45,7 +45,7 @@ push_back :: proc(list: ^List, node: ^Node) {
 	}
 }
 
-remove :: proc(list: ^List, node: ^Node) {
+remove :: proc "contextless" (list: ^List, node: ^Node) {
 	if node != nil {
 		if node.next != nil {
 			node.next.prev = node.prev
@@ -83,12 +83,34 @@ remove_by_proc :: proc(list: ^List, to_erase: proc(^Node) -> bool) {
 	}
 }
 
+remove_by_proc_contextless :: proc(list: ^List, to_erase: proc "contextless" (^Node) -> bool) {
+	for node := list.head; node != nil; {
+		next := node.next
+		if to_erase(node) {
+			if node.next != nil {
+				node.next.prev = node.prev
+			}
+			if node.prev != nil {
+				node.prev.next = node.next
+			}
+			if list.head == node {
+				list.head = node.next
+			}
+			if list.tail == node {
+				list.tail = node.prev
+			}
+		}
+		node = next
+	}
+}
 
-is_empty :: proc(list: ^List) -> bool {
+
+
+is_empty :: proc "contextless" (list: ^List) -> bool {
 	return list.head == nil
 }
 
-pop_front :: proc(list: ^List) -> ^Node {
+pop_front :: proc "contextless" (list: ^List) -> ^Node {
 	link := list.head
 	if link == nil {
 		return nil
@@ -108,7 +130,7 @@ pop_front :: proc(list: ^List) -> ^Node {
 	return link
 
 }
-pop_back :: proc(list: ^List) -> ^Node {
+pop_back :: proc "contextless" (list: ^List) -> ^Node {
 	link := list.tail
 	if link == nil {
 		return nil
@@ -134,25 +156,25 @@ Iterator :: struct($T: typeid) {
 	offset: uintptr,
 }
 
-iterator_head :: proc(list: List, $T: typeid, $field_name: string) -> Iterator(T)
+iterator_head :: proc "contextless" (list: List, $T: typeid, $field_name: string) -> Iterator(T)
 	where intrinsics.type_has_field(T, field_name),
 	      intrinsics.type_field_type(T, field_name) == Node {
 	return {list.head, offset_of_by_string(T, field_name)}
 }
 
-iterator_tail :: proc(list: List, $T: typeid, $field_name: string) -> Iterator(T)
+iterator_tail :: proc "contextless" (list: List, $T: typeid, $field_name: string) -> Iterator(T)
 	where intrinsics.type_has_field(T, field_name),
 	      intrinsics.type_field_type(T, field_name) == Node {
 	return {list.tail, offset_of_by_string(T, field_name)}
 }
 
-iterator_from_node :: proc(node: ^Node, $T: typeid, $field_name: string) -> Iterator(T)
+iterator_from_node :: proc "contextless" (node: ^Node, $T: typeid, $field_name: string) -> Iterator(T)
 	where intrinsics.type_has_field(T, field_name),
 	      intrinsics.type_field_type(T, field_name) == Node {
 	return {node, offset_of_by_string(T, field_name)}
 }
 
-iterate_next :: proc(it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
+iterate_next :: proc "contextless" (it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
 	node := it.curr
 	if node == nil {
 		return nil, false
@@ -162,7 +184,7 @@ iterate_next :: proc(it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
 	return (^T)(uintptr(node) - it.offset), true
 }
 
-iterate_prev :: proc(it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
+iterate_prev :: proc "contextless" (it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
 	node := it.curr
 	if node == nil {
 		return nil, false

core/container/lru/lru_cache.odin

@@ -1,7 +1,7 @@
 package container_lru
 
-import "core:runtime"
-import "core:intrinsics"
+import "base:runtime"
+import "base:intrinsics"
 _ :: runtime
 _ :: intrinsics
 

core/container/priority_queue/priority_queue.odin

@@ -1,6 +1,6 @@
 package container_priority_queue
 
-import "core:builtin"
+import "base:builtin"
 
 Priority_Queue :: struct($T: typeid) {
 	queue: [dynamic]T,
@@ -140,3 +140,18 @@ remove :: proc(pq: ^$Q/Priority_Queue($T), i: int) -> (value: T, ok: bool) {
 	return
 }
 
+peek_safe :: proc(pq: $Q/Priority_Queue($T), loc := #caller_location) -> (res: T, ok: bool) {
+	if builtin.len(pq.queue) > 0 {
+		return pq.queue[0], true
+	}
+	return
+}
+
+peek :: proc(pq: $Q/Priority_Queue($T), loc := #caller_location) -> (res: T) {
+	assert(condition=builtin.len(pq.queue)>0, loc=loc)
+
+	if builtin.len(pq.queue) > 0 {
+		return pq.queue[0]
+	}
+	return
+}

core/container/queue/queue.odin

@@ -1,7 +1,7 @@
 package container_queue
 
-import "core:builtin"
-import "core:runtime"
+import "base:builtin"
+import "base:runtime"
 _ :: runtime
 
 // Dynamically resizable double-ended queue/ring-buffer
@@ -22,7 +22,9 @@ init :: proc(q: ^$Q/Queue($T), capacity := DEFAULT_CAPACITY, allocator := contex
 	return reserve(q, capacity)
 }
 
-// Procedure to initialize a queue from a fixed backing slice
+// Procedure to initialize a queue from a fixed backing slice.
+// The contents of the `backing` will be overwritten as items are pushed onto the `Queue`.
+// Any previous contents are not available.
 init_from_slice :: proc(q: ^$Q/Queue($T), backing: []T) -> bool {
 	clear(q)
 	q.data = transmute([dynamic]T)runtime.Raw_Dynamic_Array{
@@ -34,6 +36,21 @@ init_from_slice :: proc(q: ^$Q/Queue($T), backing: []T) -> bool {
 	return true
 }
 
+// Procedure to initialize a queue from a fixed backing slice.
+// Existing contents are preserved and available on the queue.
+init_with_contents :: proc(q: ^$Q/Queue($T), backing: []T) -> bool {
+	clear(q)
+	q.data = transmute([dynamic]T)runtime.Raw_Dynamic_Array{
+		data = raw_data(backing),
+		len = builtin.len(backing),
+		cap = builtin.len(backing),
+		allocator = {procedure=runtime.nil_allocator_proc, data=nil},
+	}
+	q.len    = len(backing)
+	q.offset = len(backing)
+	return true
+}
+
 // Procedure to destroy a queue
 destroy :: proc(q: ^$Q/Queue($T)) {
 	delete(q.data)
@@ -56,7 +73,7 @@ space :: proc(q: $Q/Queue($T)) -> int {
 
 // Reserve enough space for at least the specified capacity
 reserve :: proc(q: ^$Q/Queue($T), capacity: int) -> runtime.Allocator_Error {
-	if uint(capacity) > q.len {
+	if capacity > space(q^) {
 		return _grow(q, uint(capacity)) 
 	}
 	return nil

core/container/small_array/small_array.odin

@@ -1,7 +1,7 @@
 package container_small_array
 
-import "core:builtin"
-import "core:runtime"
+import "base:builtin"
+import "base:runtime"
 _ :: runtime
 
 Small_Array :: struct($N: int, $T: typeid) where N >= 0 {

core/container/topological_sort/topological_sort.odin

@@ -3,8 +3,8 @@
 // map type is being used to accelerate lookups.
 package container_topological_sort
 
-import "core:intrinsics"
-import "core:runtime"
+import "base:intrinsics"
+import "base:runtime"
 _ :: intrinsics
 _ :: runtime
 
@@ -80,11 +80,13 @@ sort :: proc(sorter: ^$S/Sorter($K)) -> (sorted, cycled: [dynamic]K) {
 		}
 	}
 
-	for root in sorted do for k, _ in relations[root].dependents {
-		relation := &relations[k]
-		relation.dependencies -= 1
-		if relation.dependencies == 0 {
-			append(&sorted, k)
+	for root in sorted {
+		for k, _ in relations[root].dependents {
+			relation := &relations[k]
+			relation.dependencies -= 1
+			if relation.dependencies == 0 {
+				append(&sorted, k)
+			}
 		}
 	}
 

core/crypto/README.md

@@ -1,95 +1,24 @@
 # crypto
-A crypto library for the Odin language
+
+A cryptography 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)  | ✔️ |
+This package offers various algorithms implemented in Odin, along with
+useful helpers such as access to the system entropy source, and a
+constant-time byte comparison.
 
-#### 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 six 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_string_to_buffer` - Hash a given string and put the computed hash in the second proc parameter. Just calls `hash_bytes_to_buffer` internally
-* `hash_bytes_to_buffer` - Hash a given string and put the computed hash in the second proc parameter. The destination buffer has to be at least as big as the digest size of the 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)
+## Implementation considerations
 
-\* 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(...)`.
+- The crypto packages are not thread-safe.
+- Best-effort is make to mitigate timing side-channels on reasonable
+  architectures.  Architectures that are known to be unreasonable include
+  but are not limited to i386, i486, and WebAssembly.
+- The packages attempt to santize sensitive data, however this is, and
+  will remain a "best-effort" implementation decision.  As Thomas Pornin
+  puts it "In general, such memory cleansing is a fool's quest."
+- All of these packages have not received independent third party review.
 
-#### Low level API
-The above mentioned procedures internally call three procedures: `init`, `update` and `final`.
-You may also directly call them, if you wish.
+## License
 
-#### 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)
-
-    // Variant that takes a destination buffer, instead of returning the computed hash
-    hash := make([]byte, md4.DIGEST_SIZE) // @note: Destination buffer has to be at least as big as the digest size of the hash
-    md4.hash(input, hash[:])
-
-    // 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.
-Wherever 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/_blake2/blake2.odin

@@ -10,12 +10,13 @@ package _blake2
     Implementation of the BLAKE2 hashing algorithm, as defined in <https://datatracker.ietf.org/doc/html/rfc7693> and <https://www.blake2.net/>
 */
 
-import "../util"
+import "core:encoding/endian"
+import "core:mem"
 
-BLAKE2S_BLOCK_SIZE  :: 64
-BLAKE2S_SIZE        :: 32
-BLAKE2B_BLOCK_SIZE  :: 128
-BLAKE2B_SIZE        :: 64
+BLAKE2S_BLOCK_SIZE :: 64
+BLAKE2S_SIZE :: 32
+BLAKE2B_BLOCK_SIZE :: 128
+BLAKE2B_SIZE :: 64
 
 Blake2s_Context :: struct {
 	h:            [8]u32,
@@ -28,7 +29,8 @@ Blake2s_Context :: struct {
 	is_keyed:     bool,
 	size:         byte,
 	is_last_node: bool,
-	cfg:		  Blake2_Config,
+
+	is_initialized: bool,
 }
 
 Blake2b_Context :: struct {
@@ -42,15 +44,18 @@ Blake2b_Context :: struct {
 	is_keyed:     bool,
 	size:         byte,
 	is_last_node: bool,
-	cfg:		  Blake2_Config,
+
+	is_initialized: bool,
 }
 
 Blake2_Config :: struct {
-    size:   byte,
-	key:    []byte, 
-	salt:   []byte, 
+	size:   byte,
+	key:    []byte,
+	salt:   []byte,
 	person: []byte,
-	tree:   union{Blake2_Tree},
+	tree:   union {
+		Blake2_Tree,
+	},
 }
 
 Blake2_Tree :: struct {
@@ -63,11 +68,13 @@ Blake2_Tree :: struct {
 	is_last_node:    bool,
 }
 
+@(private)
 BLAKE2S_IV := [8]u32 {
 	0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
 	0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19,
 }
 
+@(private)
 BLAKE2B_IV := [8]u64 {
 	0x6a09e667f3bcc908, 0xbb67ae8584caa73b,
 	0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
@@ -75,80 +82,94 @@ BLAKE2B_IV := [8]u64 {
 	0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
 }
 
-init :: proc(ctx: ^$T) {
+init :: proc(ctx: ^$T, cfg: ^Blake2_Config) {
 	when T == Blake2s_Context {
-		block_size :: BLAKE2S_BLOCK_SIZE
+		max_size :: BLAKE2S_SIZE
 	} else when T == Blake2b_Context {
-		block_size :: BLAKE2B_BLOCK_SIZE
+		max_size :: BLAKE2B_SIZE
 	}
 
-	p := make([]byte, block_size)
-	defer delete(p)
+	if cfg.size > max_size {
+		panic("blake2: requested output size exceeeds algorithm max")
+	}
 
-	p[0] = ctx.cfg.size
-	p[1] = byte(len(ctx.cfg.key))
+	// To save having to allocate a scratch buffer, use the internal
+	// data buffer (`ctx.x`), as it is exactly the correct size.
+	p := ctx.x[:]
 
-	if ctx.cfg.salt != nil {
+	p[0] = cfg.size
+	p[1] = byte(len(cfg.key))
+
+	if cfg.salt != nil {
 		when T == Blake2s_Context {
-			copy(p[16:], ctx.cfg.salt)
+			copy(p[16:], cfg.salt)
 		} else when T == Blake2b_Context {
-			copy(p[32:], ctx.cfg.salt)
+			copy(p[32:], cfg.salt)
 		}
 	}
-	if ctx.cfg.person != nil {
+	if cfg.person != nil {
 		when T == Blake2s_Context {
-			copy(p[24:], ctx.cfg.person)
+			copy(p[24:], cfg.person)
 		} else when T == Blake2b_Context {
-			copy(p[48:], ctx.cfg.person)
+			copy(p[48:], cfg.person)
 		}
 	}
 
-	if ctx.cfg.tree != nil {
-		p[2] = ctx.cfg.tree.(Blake2_Tree).fanout
-		p[3] = ctx.cfg.tree.(Blake2_Tree).max_depth
-		util.PUT_U32_LE(p[4:], ctx.cfg.tree.(Blake2_Tree).leaf_size)
+	if cfg.tree != nil {
+		p[2] = cfg.tree.(Blake2_Tree).fanout
+		p[3] = cfg.tree.(Blake2_Tree).max_depth
+		endian.unchecked_put_u32le(p[4:], cfg.tree.(Blake2_Tree).leaf_size)
 		when T == Blake2s_Context {
-			p[8]  = byte(ctx.cfg.tree.(Blake2_Tree).node_offset)
-			p[9]  = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 8)
-			p[10] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 16)
-			p[11] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 24)
-			p[12] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 32)
-			p[13] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 40)
-			p[14] = ctx.cfg.tree.(Blake2_Tree).node_depth
-			p[15] = ctx.cfg.tree.(Blake2_Tree).inner_hash_size
+			p[8] = byte(cfg.tree.(Blake2_Tree).node_offset)
+			p[9] = byte(cfg.tree.(Blake2_Tree).node_offset >> 8)
+			p[10] = byte(cfg.tree.(Blake2_Tree).node_offset >> 16)
+			p[11] = byte(cfg.tree.(Blake2_Tree).node_offset >> 24)
+			p[12] = byte(cfg.tree.(Blake2_Tree).node_offset >> 32)
+			p[13] = byte(cfg.tree.(Blake2_Tree).node_offset >> 40)
+			p[14] = cfg.tree.(Blake2_Tree).node_depth
+			p[15] = cfg.tree.(Blake2_Tree).inner_hash_size
 		} else when T == Blake2b_Context {
-			util.PUT_U64_LE(p[8:], ctx.cfg.tree.(Blake2_Tree).node_offset)
-			p[16] = ctx.cfg.tree.(Blake2_Tree).node_depth
-			p[17] = ctx.cfg.tree.(Blake2_Tree).inner_hash_size
+			endian.unchecked_put_u64le(p[8:], cfg.tree.(Blake2_Tree).node_offset)
+			p[16] = cfg.tree.(Blake2_Tree).node_depth
+			p[17] = cfg.tree.(Blake2_Tree).inner_hash_size
 		}
 	} else {
 		p[2], p[3] = 1, 1
 	}
-	ctx.size = ctx.cfg.size
+	ctx.size = cfg.size
 	for i := 0; i < 8; i += 1 {
 		when T == Blake2s_Context {
-			ctx.h[i] = BLAKE2S_IV[i] ~ util.U32_LE(p[i * 4:])
+			ctx.h[i] = BLAKE2S_IV[i] ~ endian.unchecked_get_u32le(p[i * 4:])
 		}
 		when T == Blake2b_Context {
-			ctx.h[i] = BLAKE2B_IV[i] ~ util.U64_LE(p[i * 8:])
+			ctx.h[i] = BLAKE2B_IV[i] ~ endian.unchecked_get_u64le(p[i * 8:])
 		}
 	}
-	if ctx.cfg.tree != nil && ctx.cfg.tree.(Blake2_Tree).is_last_node {
+
+	mem.zero(&ctx.x, size_of(ctx.x)) // Done with the scratch space, no barrier.
+
+	if cfg.tree != nil && cfg.tree.(Blake2_Tree).is_last_node {
 		ctx.is_last_node = true
 	}
-	if len(ctx.cfg.key) > 0 {
-		copy(ctx.padded_key[:], ctx.cfg.key)
+	if len(cfg.key) > 0 {
+		copy(ctx.padded_key[:], cfg.key)
 		update(ctx, ctx.padded_key[:])
 		ctx.is_keyed = true
 	}
 	copy(ctx.ih[:], ctx.h[:])
-	copy(ctx.h[:],  ctx.ih[:])
+	copy(ctx.h[:], ctx.ih[:])
 	if ctx.is_keyed {
 		update(ctx, ctx.padded_key[:])
 	}
+
+	ctx.nx = 0
+
+	ctx.is_initialized = true
 }
 
-update :: proc "contextless" (ctx: ^$T, p: []byte) {
+update :: proc(ctx: ^$T, p: []byte) {
+	assert(ctx.is_initialized)
+
 	p := p
 	when T == Blake2s_Context {
 		block_size :: BLAKE2S_BLOCK_SIZE
@@ -174,15 +195,43 @@ update :: proc "contextless" (ctx: ^$T, p: []byte) {
 	ctx.nx += copy(ctx.x[ctx.nx:], p)
 }
 
-final :: proc "contextless" (ctx: ^$T, hash: []byte) {
-	when T == Blake2s_Context {
-		blake2s_final(ctx, hash)
+final :: proc(ctx: ^$T, hash: []byte, finalize_clone: bool = false) {
+	assert(ctx.is_initialized)
+
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: T
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
 	}
-	when T == Blake2b_Context {
+	defer(reset(ctx))
+
+	when T == Blake2s_Context {
+		if len(hash) < int(ctx.size) {
+			panic("crypto/blake2s: invalid destination digest size")
+		}
+		blake2s_final(ctx, hash)
+	} else when T == Blake2b_Context {
+		if len(hash) < int(ctx.size) {
+			panic("crypto/blake2b: invalid destination digest size")
+		}
 		blake2b_final(ctx, hash)
 	}
 }
 
+clone :: proc(ctx, other: ^$T) {
+	ctx^ = other^
+}
+
+reset :: proc(ctx: ^$T) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
+}
+
+@(private)
 blake2s_final :: proc "contextless" (ctx: ^Blake2s_Context, hash: []byte) {
 	if ctx.is_keyed {
 		for i := 0; i < len(ctx.padded_key); i += 1 {
@@ -203,16 +252,14 @@ blake2s_final :: proc "contextless" (ctx: ^Blake2s_Context, hash: []byte) {
 
 	blocks(ctx, ctx.x[:])
 
-	j := 0
-	for s, _ in ctx.h[:(ctx.size - 1) / 4 + 1] {
-		hash[j + 0] = byte(s >> 0)
-		hash[j + 1] = byte(s >> 8)
-		hash[j + 2] = byte(s >> 16)
-		hash[j + 3] = byte(s >> 24)
-		j += 4
+	dst: [BLAKE2S_SIZE]byte
+	for i := 0; i < BLAKE2S_SIZE / 4; i += 1 {
+		endian.unchecked_put_u32le(dst[i * 4:], ctx.h[i])
 	}
+	copy(hash, dst[:])
 }
 
+@(private)
 blake2b_final :: proc "contextless" (ctx: ^Blake2b_Context, hash: []byte) {
 	if ctx.is_keyed {
 		for i := 0; i < len(ctx.padded_key); i += 1 {
@@ -229,56 +276,52 @@ blake2b_final :: proc "contextless" (ctx: ^Blake2b_Context, hash: []byte) {
 	ctx.f[0] = 0xffffffffffffffff
 	if ctx.is_last_node {
 		ctx.f[1] = 0xffffffffffffffff
-	} 
+	}
 
 	blocks(ctx, ctx.x[:])
 
-	j := 0
-	for s, _ in ctx.h[:(ctx.size - 1) / 8 + 1] {
-		hash[j + 0] = byte(s >> 0)
-		hash[j + 1] = byte(s >> 8)
-		hash[j + 2] = byte(s >> 16)
-		hash[j + 3] = byte(s >> 24)
-		hash[j + 4] = byte(s >> 32)
-		hash[j + 5] = byte(s >> 40)
-		hash[j + 6] = byte(s >> 48)
-		hash[j + 7] = byte(s >> 56)
-		j += 8
+	dst: [BLAKE2B_SIZE]byte
+	for i := 0; i < BLAKE2B_SIZE / 8; i += 1 {
+		endian.unchecked_put_u64le(dst[i * 8:], ctx.h[i])
 	}
+	copy(hash, dst[:])
 }
 
+@(private)
 blocks :: proc "contextless" (ctx: ^$T, p: []byte) {
 	when T == Blake2s_Context {
 		blake2s_blocks(ctx, p)
-	}
-	when T == Blake2b_Context {
+	} else when T == Blake2b_Context {
 		blake2b_blocks(ctx, p)
 	}
 }
 
+@(private)
 blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []byte) {
-	h0, h1, h2, h3, h4, h5, h6, h7 := ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7]
+	h0, h1, h2, h3, h4, h5, h6, h7 :=
+		ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7]
 	p := p
 	for len(p) >= BLAKE2S_BLOCK_SIZE {
 		ctx.t[0] += BLAKE2S_BLOCK_SIZE
 		if ctx.t[0] < BLAKE2S_BLOCK_SIZE {
 			ctx.t[1] += 1
-		} 
+		}
 		v0, v1, v2, v3, v4, v5, v6, v7 := h0, h1, h2, h3, h4, h5, h6, h7
-		v8  := BLAKE2S_IV[0]
-		v9  := BLAKE2S_IV[1]
+		v8 := BLAKE2S_IV[0]
+		v9 := BLAKE2S_IV[1]
 		v10 := BLAKE2S_IV[2]
 		v11 := BLAKE2S_IV[3]
 		v12 := BLAKE2S_IV[4] ~ ctx.t[0]
 		v13 := BLAKE2S_IV[5] ~ ctx.t[1]
 		v14 := BLAKE2S_IV[6] ~ ctx.f[0]
 		v15 := BLAKE2S_IV[7] ~ ctx.f[1]
-		m: [16]u32
-		j := 0
+
+		m: [16]u32 = ---
 		for i := 0; i < 16; i += 1 {
-			m[i] = u32(p[j]) | u32(p[j + 1]) << 8 | u32(p[j + 2]) << 16 | u32(p[j + 3]) << 24
-			j += 4
+			m[i] = endian.unchecked_get_u32le(p[i * 4:])
 		}
+
+		// Round 1
 		v0 += m[0]
 		v0 += v4
 		v12 ~= v0
@@ -391,6 +434,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 2
 		v0 += m[14]
 		v0 += v4
 		v12 ~= v0
@@ -503,6 +548,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 3
 		v0 += m[11]
 		v0 += v4
 		v12 ~= v0
@@ -615,6 +662,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 4
 		v0 += m[7]
 		v0 += v4
 		v12 ~= v0
@@ -727,6 +776,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 5
 		v0 += m[9]
 		v0 += v4
 		v12 ~= v0
@@ -839,6 +890,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 6
 		v0 += m[2]
 		v0 += v4
 		v12 ~= v0
@@ -951,6 +1004,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 7
 		v0 += m[12]
 		v0 += v4
 		v12 ~= v0
@@ -1063,6 +1118,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 8
 		v0 += m[13]
 		v0 += v4
 		v12 ~= v0
@@ -1175,6 +1232,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 9
 		v0 += m[6]
 		v0 += v4
 		v12 ~= v0
@@ -1287,6 +1346,8 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
+		// Round 10
 		v0 += m[10]
 		v0 += v4
 		v12 ~= v0
@@ -1399,6 +1460,7 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (32 - 7) | v5 >> 7
+
 		h0 ~= v0 ~ v8
 		h1 ~= v1 ~ v9
 		h2 ~= v2 ~ v10
@@ -1407,19 +1469,23 @@ blake2s_blocks :: #force_inline proc "contextless" (ctx: ^Blake2s_Context, p: []
 		h5 ~= v5 ~ v13
 		h6 ~= v6 ~ v14
 		h7 ~= v7 ~ v15
+
 		p = p[BLAKE2S_BLOCK_SIZE:]
 	}
-	ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7] = h0, h1, h2, h3, h4, h5, h6, h7
+	ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7] =
+		h0, h1, h2, h3, h4, h5, h6, h7
 }
 
+@(private)
 blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []byte) {
-	h0, h1, h2, h3, h4, h5, h6, h7 := ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7]
+	h0, h1, h2, h3, h4, h5, h6, h7 :=
+		ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7]
 	p := p
 	for len(p) >= BLAKE2B_BLOCK_SIZE {
 		ctx.t[0] += BLAKE2B_BLOCK_SIZE
 		if ctx.t[0] < BLAKE2B_BLOCK_SIZE {
-			ctx.t[1]+=1
-		} 
+			ctx.t[1] += 1
+		}
 		v0, v1, v2, v3, v4, v5, v6, v7 := h0, h1, h2, h3, h4, h5, h6, h7
 		v8 := BLAKE2B_IV[0]
 		v9 := BLAKE2B_IV[1]
@@ -1429,13 +1495,13 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v13 := BLAKE2B_IV[5] ~ ctx.t[1]
 		v14 := BLAKE2B_IV[6] ~ ctx.f[0]
 		v15 := BLAKE2B_IV[7] ~ ctx.f[1]
+
 		m: [16]u64 = ---
-		j := 0
-		for i := 0; i < 16; i+=1 {
-			m[i] = u64(p[j]) 		   | u64(p[j + 1]) << 8  | u64(p[j + 2]) << 16 | u64(p[j + 3]) << 24 |
-				   u64(p[j + 4]) << 32 | u64(p[j + 5]) << 40 | u64(p[j + 6]) << 48 | u64(p[j + 7]) << 56
-			j += 8
+		for i := 0; i < 16; i += 1 {
+			m[i] = endian.unchecked_get_u64le(p[i * 8:])
 		}
+
+		// Round 1
 		v0 += m[0]
 		v0 += v4
 		v12 ~= v0
@@ -1548,6 +1614,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 2
 		v0 += m[14]
 		v0 += v4
 		v12 ~= v0
@@ -1660,6 +1728,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 3
 		v0 += m[11]
 		v0 += v4
 		v12 ~= v0
@@ -1772,6 +1842,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 4
 		v0 += m[7]
 		v0 += v4
 		v12 ~= v0
@@ -1884,6 +1956,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 5
 		v0 += m[9]
 		v0 += v4
 		v12 ~= v0
@@ -1996,6 +2070,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 6
 		v0 += m[2]
 		v0 += v4
 		v12 ~= v0
@@ -2108,6 +2184,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 7
 		v0 += m[12]
 		v0 += v4
 		v12 ~= v0
@@ -2220,6 +2298,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 8
 		v0 += m[13]
 		v0 += v4
 		v12 ~= v0
@@ -2332,6 +2412,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 9
 		v0 += m[6]
 		v0 += v4
 		v12 ~= v0
@@ -2444,6 +2526,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 10
 		v0 += m[10]
 		v0 += v4
 		v12 ~= v0
@@ -2556,6 +2640,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 11
 		v0 += m[0]
 		v0 += v4
 		v12 ~= v0
@@ -2668,6 +2754,8 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
+		// Round 12
 		v0 += m[14]
 		v0 += v4
 		v12 ~= v0
@@ -2780,6 +2868,7 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		v10 += v15
 		v5 ~= v10
 		v5 = v5 << (64 - 63) | v5 >> 63
+
 		h0 ~= v0 ~ v8
 		h1 ~= v1 ~ v9
 		h2 ~= v2 ~ v10
@@ -2788,7 +2877,9 @@ blake2b_blocks :: #force_inline proc "contextless" (ctx: ^Blake2b_Context, p: []
 		h5 ~= v5 ~ v13
 		h6 ~= v6 ~ v14
 		h7 ~= v7 ~ v15
+
 		p = p[BLAKE2B_BLOCK_SIZE:]
 	}
-	ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7] = h0, h1, h2, h3, h4, h5, h6, h7
-}
+	ctx.h[0], ctx.h[1], ctx.h[2], ctx.h[3], ctx.h[4], ctx.h[5], ctx.h[6], ctx.h[7] =
+		h0, h1, h2, h3, h4, h5, h6, h7
+}

core/crypto/_fiat/field_poly1305/field.odin

@@ -1,6 +1,6 @@
 package field_poly1305
 
-import "core:crypto/util"
+import "core:encoding/endian"
 import "core:mem"
 
 fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
@@ -11,7 +11,7 @@ fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_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) {
+fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, arg2: byte) {
 	// 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.
@@ -22,42 +22,29 @@ fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, a
 
 	assert(len(arg1) == 16)
 
-	when ODIN_ARCH == .i386 || 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]
+	// 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.
+	lo := endian.unchecked_get_u64le(arg1[0:])
+	hi := endian.unchecked_get_u64le(arg1[8:])
 
-		// 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))
-		}
-	}
+	// 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
 }
 
 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)
+	endian.unchecked_put_u64le(tmp[0:], lo)
+	endian.unchecked_put_u64le(tmp[8:], hi)
 
 	_fe_from_bytes(out1, &tmp)
 

core/crypto/_sha3/_sha3.odin

@@ -1,169 +0,0 @@
-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/_sha3/sha3.odin

@@ -0,0 +1,208 @@
+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 "core:math/bits"
+import "core:mem"
+
+ROUNDS :: 24
+
+RATE_224 :: 1152 / 8
+RATE_256 :: 1088 / 8
+RATE_384 :: 832 / 8
+RATE_512 :: 576 / 8
+
+Context :: struct {
+	st:        struct #raw_union {
+		b: [200]u8,
+		q: [25]u64,
+	},
+	pt:        int,
+	rsiz:      int,
+	mdlen:     int,
+	is_keccak: bool,
+
+	is_initialized: bool,
+	is_finalized:   bool, // For SHAKE (unlimited squeeze is allowed)
+}
+
+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 := [?]int {
+		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 {
+		for i = 0; i < 25; i += 1 {
+			st[i] = bits.byte_swap(st[i])
+		}
+	}
+
+	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] ~ bits.rotate_left64(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] = bits.rotate_left64(t, 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 {
+			st[i] = bits.byte_swap(st[i])
+		}
+	}
+}
+
+init :: proc(ctx: ^Context) {
+	for i := 0; i < 25; i += 1 {
+		ctx.st.q[i] = 0
+	}
+	ctx.rsiz = 200 - 2 * ctx.mdlen
+	ctx.pt = 0
+
+	ctx.is_initialized = true
+	ctx.is_finalized = false
+}
+
+update :: proc(ctx: ^Context, data: []byte) {
+	assert(ctx.is_initialized)
+	assert(!ctx.is_finalized)
+
+	j := ctx.pt
+	for i := 0; i < len(data); i += 1 {
+		ctx.st.b[j] ~= data[i]
+		j += 1
+		if j >= ctx.rsiz {
+			keccakf(&ctx.st.q)
+			j = 0
+		}
+	}
+	ctx.pt = j
+}
+
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	assert(ctx.is_initialized)
+
+	if len(hash) < ctx.mdlen {
+		if ctx.is_keccak {
+			panic("crypto/keccac: invalid destination digest size")
+		}
+		panic("crypto/sha3: invalid destination digest size")
+	}
+
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: Context
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
+	if ctx.is_keccak {
+		ctx.st.b[ctx.pt] ~= 0x01
+	} else {
+		ctx.st.b[ctx.pt] ~= 0x06
+	}
+
+	ctx.st.b[ctx.rsiz - 1] ~= 0x80
+	keccakf(&ctx.st.q)
+	for i := 0; i < ctx.mdlen; i += 1 {
+		hash[i] = ctx.st.b[i]
+	}
+}
+
+clone :: proc(ctx, other: ^Context) {
+	ctx^ = other^
+}
+
+reset :: proc(ctx: ^Context) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
+}
+
+shake_xof :: proc(ctx: ^Context) {
+	assert(ctx.is_initialized)
+	assert(!ctx.is_finalized)
+
+	ctx.st.b[ctx.pt] ~= 0x1F
+	ctx.st.b[ctx.rsiz - 1] ~= 0x80
+	keccakf(&ctx.st.q)
+	ctx.pt = 0
+
+	ctx.is_finalized = true // No more absorb, unlimited squeeze.
+}
+
+shake_out :: proc(ctx: ^Context, hash: []byte) {
+	assert(ctx.is_initialized)
+	assert(ctx.is_finalized)
+
+	j := ctx.pt
+	for i := 0; i < len(hash); i += 1 {
+		if j >= ctx.rsiz {
+			keccakf(&ctx.st.q)
+			j = 0
+		}
+		hash[i] = ctx.st.b[j]
+		j += 1
+	}
+	ctx.pt = j
+}

core/crypto/_tiger/_tiger.odin

@@ -1,410 +0,0 @@
-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

@@ -1,726 +0,0 @@
-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
-*/
-
-DIGEST_SIZE_224 :: 28
-DIGEST_SIZE_256 :: 32
-DIGEST_SIZE_384 :: 48
-DIGEST_SIZE_512 :: 64
-
-// hash_string_224 will hash the given input and return the
-// computed hash
-hash_string_224 :: proc "contextless" (data: string) -> [DIGEST_SIZE_224]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) -> [DIGEST_SIZE_224]byte {
-    hash: [DIGEST_SIZE_224]byte
-    ctx: Blake256_Context
-    ctx.is224 = true
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_224 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_224 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_224(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_224 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_224 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_224, "Size of destination buffer is smaller than the digest size")
-    ctx: Blake256_Context
-    ctx.is224 = true
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_224 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_224 :: proc(s: io.Stream) -> ([DIGEST_SIZE_224]byte, bool) {
-    hash: [DIGEST_SIZE_224]byte
-    ctx: Blake256_Context
-    ctx.is224 = true
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_224]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 [DIGEST_SIZE_224]byte{}, false
-}
-
-hash_224 :: proc {
-    hash_stream_224,
-    hash_file_224,
-    hash_bytes_224,
-    hash_string_224,
-    hash_bytes_to_buffer_224,
-    hash_string_to_buffer_224,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc "contextless" (data: string) -> [DIGEST_SIZE_256]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) -> [DIGEST_SIZE_256]byte {
-    hash: [DIGEST_SIZE_256]byte
-    ctx: Blake256_Context
-    ctx.is224 = false
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_256, "Size of destination buffer is smaller than the digest size")
-    ctx: Blake256_Context
-    ctx.is224 = false
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-    hash: [DIGEST_SIZE_256]byte
-    ctx: Blake256_Context
-    ctx.is224 = false
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_256]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 [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-    hash_stream_256,
-    hash_file_256,
-    hash_bytes_256,
-    hash_string_256,
-    hash_bytes_to_buffer_256,
-    hash_string_to_buffer_256,
-}
-
-// hash_string_384 will hash the given input and return the
-// computed hash
-hash_string_384 :: proc "contextless" (data: string) -> [DIGEST_SIZE_384]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) -> [DIGEST_SIZE_384]byte {
-    hash: [DIGEST_SIZE_384]byte
-    ctx: Blake512_Context
-    ctx.is384 = true
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_384 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_384 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_384(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_384 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_384 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_384, "Size of destination buffer is smaller than the digest size")
-    ctx: Blake512_Context
-    ctx.is384 = true
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_384 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_384 :: proc(s: io.Stream) -> ([DIGEST_SIZE_384]byte, bool) {
-    hash: [DIGEST_SIZE_384]byte
-    ctx: Blake512_Context
-    ctx.is384 = true
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_384]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 [DIGEST_SIZE_384]byte{}, false
-}
-
-hash_384 :: proc {
-    hash_stream_384,
-    hash_file_384,
-    hash_bytes_384,
-    hash_string_384,
-    hash_bytes_to_buffer_384,
-    hash_string_to_buffer_384,
-}
-
-// hash_string_512 will hash the given input and return the
-// computed hash
-hash_string_512 :: proc "contextless" (data: string) -> [DIGEST_SIZE_512]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) -> [DIGEST_SIZE_512]byte {
-    hash: [DIGEST_SIZE_512]byte
-    ctx: Blake512_Context
-    ctx.is384 = false
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_512 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_512(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_512, "Size of destination buffer is smaller than the digest size")
-    ctx: Blake512_Context
-    ctx.is384 = false
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_512 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512]byte, bool) {
-    hash: [DIGEST_SIZE_512]byte
-    ctx: Blake512_Context
-    ctx.is384 = false
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_512]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 [DIGEST_SIZE_512]byte{}, false
-}
-
-hash_512 :: proc {
-    hash_stream_512,
-    hash_file_512,
-    hash_bytes_512,
-    hash_string_512,
-    hash_bytes_to_buffer_512,
-    hash_string_to_buffer_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

@@ -1,3 +1,10 @@
+/*
+package blake2b implements the BLAKE2b hash algorithm.
+
+See:
+- https://datatracker.ietf.org/doc/html/rfc7693
+- https://www.blake2.net
+*/
 package blake2b
 
 /*
@@ -6,122 +13,47 @@ package blake2b
 
     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
-*/
-
+// DIGEST_SIZE is the BLAKE2b digest size in bytes.
 DIGEST_SIZE :: 64
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-    return hash_bytes(transmute([]byte)(data))
+// BLOCK_SIZE is the BLAKE2b block size in bytes.
+BLOCK_SIZE :: _blake2.BLAKE2B_BLOCK_SIZE
+
+// Context is a BLAKE2b instance.
+Context :: _blake2.Blake2b_Context
+
+// init initializes a Context with the default BLAKE2b config.
+init :: proc(ctx: ^Context) {
+	cfg: _blake2.Blake2_Config
+	cfg.size = _blake2.BLAKE2B_SIZE
+	_blake2.init(ctx, &cfg)
 }
 
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-    hash: [DIGEST_SIZE]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
+// update adds more data to the Context.
+update :: proc(ctx: ^Context, data: []byte) {
+	_blake2.update(ctx, data)
 }
 
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer(transmute([]byte)(data), hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_blake2.final(ctx, hash, finalize_clone)
 }
 
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE, "Size of destination buffer is smaller than the digest size")
-    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)
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_blake2.clone(ctx, other)
 }
 
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-    hash: [DIGEST_SIZE]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, _ = io.read(s, 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) -> ([DIGEST_SIZE]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 [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-    hash_stream,
-    hash_file,
-    hash_bytes,
-    hash_string,
-    hash_bytes_to_buffer,
-    hash_string_to_buffer,
-}
-
-/*
-    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)
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_blake2.reset(ctx)
 }

core/crypto/blake2s/blake2s.odin

@@ -1,3 +1,10 @@
+/*
+package blake2s implements the BLAKE2s hash algorithm.
+
+See:
+- https://datatracker.ietf.org/doc/html/rfc7693
+- https://www.blake2.net/
+*/
 package blake2s
 
 /*
@@ -6,122 +13,47 @@ package blake2s
 
     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
-*/
-
+// DIGEST_SIZE is the BLAKE2s digest size in bytes.
 DIGEST_SIZE :: 32
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-    return hash_bytes(transmute([]byte)(data))
+// BLOCK_SIZE is the BLAKE2s block size in bytes.
+BLOCK_SIZE :: _blake2.BLAKE2S_BLOCK_SIZE
+
+// Context is a BLAKE2s instance.
+Context :: _blake2.Blake2s_Context
+
+// init initializes a Context with the default BLAKE2s config.
+init :: proc(ctx: ^Context) {
+	cfg: _blake2.Blake2_Config
+	cfg.size = _blake2.BLAKE2S_SIZE
+	_blake2.init(ctx, &cfg)
 }
 
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-    hash: [DIGEST_SIZE]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
+// update adds more data to the Context.
+update :: proc(ctx: ^Context, data: []byte) {
+	_blake2.update(ctx, data)
 }
 
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer(transmute([]byte)(data), hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_blake2.final(ctx, hash, finalize_clone)
 }
 
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE, "Size of destination buffer is smaller than the digest size")
-    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)
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_blake2.clone(ctx, other)
 }
 
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-    hash: [DIGEST_SIZE]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, _ = io.read(s, 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) -> ([DIGEST_SIZE]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 [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-    hash_stream,
-    hash_file,
-    hash_bytes,
-    hash_string,
-    hash_bytes_to_buffer,
-    hash_string_to_buffer,
-}
-
-/*
-    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)
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_blake2.reset(ctx)
 }

core/crypto/chacha20/chacha20.odin

@@ -1,6 +1,6 @@
 package chacha20
 
-import "core:crypto/util"
+import "core:encoding/endian"
 import "core:math/bits"
 import "core:mem"
 
@@ -60,23 +60,23 @@ init :: proc (ctx: ^Context, key, nonce: []byte) {
 	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[4] = endian.unchecked_get_u32le(k[0:4])
+	ctx._s[5] = endian.unchecked_get_u32le(k[4:8])
+	ctx._s[6] = endian.unchecked_get_u32le(k[8:12])
+	ctx._s[7] = endian.unchecked_get_u32le(k[12:16])
+	ctx._s[8] = endian.unchecked_get_u32le(k[16:20])
+	ctx._s[9] = endian.unchecked_get_u32le(k[20:24])
+	ctx._s[10] = endian.unchecked_get_u32le(k[24:28])
+	ctx._s[11] = endian.unchecked_get_u32le(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])
+		ctx._s[13] = endian.unchecked_get_u32le(n[0:4])
+		ctx._s[14] = endian.unchecked_get_u32le(n[4:8])
+		ctx._s[15] = endian.unchecked_get_u32le(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])
+		ctx._s[14] = endian.unchecked_get_u32le(n[0:4])
+		ctx._s[15] = endian.unchecked_get_u32le(n[4:8])
 
 		// The sub-key is stored in the keystream buffer.  While
 		// this will be overwritten in most circumstances, explicitly
@@ -221,114 +221,114 @@ _do_blocks :: proc (ctx: ^Context, dst, src: []byte, nr_blocks: int) {
 			// quarterround(x, 0, 4, 8, 12)
 			x0 += x4
 			x12 ~= x0
-			x12 = util.ROTL32(x12, 16)
+			x12 = bits.rotate_left32(x12, 16)
 			x8 += x12
 			x4 ~= x8
-			x4 = util.ROTL32(x4, 12)
+			x4 = bits.rotate_left32(x4, 12)
 			x0 += x4
 			x12 ~= x0
-			x12 = util.ROTL32(x12, 8)
+			x12 = bits.rotate_left32(x12, 8)
 			x8 += x12
 			x4 ~= x8
-			x4 = util.ROTL32(x4, 7)
+			x4 = bits.rotate_left32(x4, 7)
 
 			// quarterround(x, 1, 5, 9, 13)
 			x1 += x5
 			x13 ~= x1
-			x13 = util.ROTL32(x13, 16)
+			x13 = bits.rotate_left32(x13, 16)
 			x9 += x13
 			x5 ~= x9
-			x5 = util.ROTL32(x5, 12)
+			x5 = bits.rotate_left32(x5, 12)
 			x1 += x5
 			x13 ~= x1
-			x13 = util.ROTL32(x13, 8)
+			x13 = bits.rotate_left32(x13, 8)
 			x9 += x13
 			x5 ~= x9
-			x5 = util.ROTL32(x5, 7)
+			x5 = bits.rotate_left32(x5, 7)
 
 			// quarterround(x, 2, 6, 10, 14)
 			x2 += x6
 			x14 ~= x2
-			x14 = util.ROTL32(x14, 16)
+			x14 = bits.rotate_left32(x14, 16)
 			x10 += x14
 			x6 ~= x10
-			x6 = util.ROTL32(x6, 12)
+			x6 = bits.rotate_left32(x6, 12)
 			x2 += x6
 			x14 ~= x2
-			x14 = util.ROTL32(x14, 8)
+			x14 = bits.rotate_left32(x14, 8)
 			x10 += x14
 			x6 ~= x10
-			x6 = util.ROTL32(x6, 7)
+			x6 = bits.rotate_left32(x6, 7)
 
 			// quarterround(x, 3, 7, 11, 15)
 			x3 += x7
 			x15 ~= x3
-			x15 = util.ROTL32(x15, 16)
+			x15 = bits.rotate_left32(x15, 16)
 			x11 += x15
 			x7 ~= x11
-			x7 = util.ROTL32(x7, 12)
+			x7 = bits.rotate_left32(x7, 12)
 			x3 += x7
 			x15 ~= x3
-			x15 = util.ROTL32(x15, 8)
+			x15 = bits.rotate_left32(x15, 8)
 			x11 += x15
 			x7 ~= x11
-			x7 = util.ROTL32(x7, 7)
+			x7 = bits.rotate_left32(x7, 7)
 
 			// quarterround(x, 0, 5, 10, 15)
 			x0 += x5
 			x15 ~= x0
-			x15 = util.ROTL32(x15, 16)
+			x15 = bits.rotate_left32(x15, 16)
 			x10 += x15
 			x5 ~= x10
-			x5 = util.ROTL32(x5, 12)
+			x5 = bits.rotate_left32(x5, 12)
 			x0 += x5
 			x15 ~= x0
-			x15 = util.ROTL32(x15, 8)
+			x15 = bits.rotate_left32(x15, 8)
 			x10 += x15
 			x5 ~= x10
-			x5 = util.ROTL32(x5, 7)
+			x5 = bits.rotate_left32(x5, 7)
 
 			// quarterround(x, 1, 6, 11, 12)
 			x1 += x6
 			x12 ~= x1
-			x12 = util.ROTL32(x12, 16)
+			x12 = bits.rotate_left32(x12, 16)
 			x11 += x12
 			x6 ~= x11
-			x6 = util.ROTL32(x6, 12)
+			x6 = bits.rotate_left32(x6, 12)
 			x1 += x6
 			x12 ~= x1
-			x12 = util.ROTL32(x12, 8)
+			x12 = bits.rotate_left32(x12, 8)
 			x11 += x12
 			x6 ~= x11
-			x6 = util.ROTL32(x6, 7)
+			x6 = bits.rotate_left32(x6, 7)
 
 			// quarterround(x, 2, 7, 8, 13)
 			x2 += x7
 			x13 ~= x2
-			x13 = util.ROTL32(x13, 16)
+			x13 = bits.rotate_left32(x13, 16)
 			x8 += x13
 			x7 ~= x8
-			x7 = util.ROTL32(x7, 12)
+			x7 = bits.rotate_left32(x7, 12)
 			x2 += x7
 			x13 ~= x2
-			x13 = util.ROTL32(x13, 8)
+			x13 = bits.rotate_left32(x13, 8)
 			x8 += x13
 			x7 ~= x8
-			x7 = util.ROTL32(x7, 7)
+			x7 = bits.rotate_left32(x7, 7)
 
 			// quarterround(x, 3, 4, 9, 14)
 			x3 += x4
 			x14 ~= x3
-			x14 = util.ROTL32(x14, 16)
+			x14 = bits.rotate_left32(x14, 16)
 			x9 += x14
 			x4 ~= x9
-			x4 = util.ROTL32(x4, 12)
+			x4 = bits.rotate_left32(x4, 12)
 			x3 += x4
 			x14 ~= x3
-			x14 = util.ROTL32(x14, 8)
+			x14 = bits.rotate_left32(x14, 8)
 			x9 += x14
 			x4 ~= x9
-			x4 = util.ROTL32(x4, 7)
+			x4 = bits.rotate_left32(x4, 7)
 		}
 
 		x0 += _SIGMA_0
@@ -352,93 +352,48 @@ _do_blocks :: proc (ctx: ^Context, dst, src: []byte, nr_blocks: int) {
 		// 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.
+		// In the meantime:
+		// - The caller(s) ensure that src/dst are valid.
+		// - The compiler knows if the target is picky about alignment.
 
-		when ODIN_ARCH == .i386 || 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])
+		#no_bounds_check {
 			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
+				endian.unchecked_put_u32le(dst[0:4], endian.unchecked_get_u32le(src[0:4]) ~ x0)
+				endian.unchecked_put_u32le(dst[4:8], endian.unchecked_get_u32le(src[4:8]) ~ x1)
+				endian.unchecked_put_u32le(dst[8:12], endian.unchecked_get_u32le(src[8:12]) ~ x2)
+				endian.unchecked_put_u32le(dst[12:16], endian.unchecked_get_u32le(src[12:16]) ~ x3)
+				endian.unchecked_put_u32le(dst[16:20], endian.unchecked_get_u32le(src[16:20]) ~ x4)
+				endian.unchecked_put_u32le(dst[20:24], endian.unchecked_get_u32le(src[20:24]) ~ x5)
+				endian.unchecked_put_u32le(dst[24:28], endian.unchecked_get_u32le(src[24:28]) ~ x6)
+				endian.unchecked_put_u32le(dst[28:32], endian.unchecked_get_u32le(src[28:32]) ~ x7)
+				endian.unchecked_put_u32le(dst[32:36], endian.unchecked_get_u32le(src[32:36]) ~ x8)
+				endian.unchecked_put_u32le(dst[36:40], endian.unchecked_get_u32le(src[36:40]) ~ x9)
+				endian.unchecked_put_u32le(dst[40:44], endian.unchecked_get_u32le(src[40:44]) ~ x10)
+				endian.unchecked_put_u32le(dst[44:48], endian.unchecked_get_u32le(src[44:48]) ~ x11)
+				endian.unchecked_put_u32le(dst[48:52], endian.unchecked_get_u32le(src[48:52]) ~ x12)
+				endian.unchecked_put_u32le(dst[52:56], endian.unchecked_get_u32le(src[52:56]) ~ x13)
+				endian.unchecked_put_u32le(dst[56:60], endian.unchecked_get_u32le(src[56:60]) ~ x14)
+				endian.unchecked_put_u32le(dst[60:64], endian.unchecked_get_u32le(src[60:64]) ~ 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
+				endian.unchecked_put_u32le(dst[0:4], x0)
+				endian.unchecked_put_u32le(dst[4:8], x1)
+				endian.unchecked_put_u32le(dst[8:12], x2)
+				endian.unchecked_put_u32le(dst[12:16], x3)
+				endian.unchecked_put_u32le(dst[16:20], x4)
+				endian.unchecked_put_u32le(dst[20:24], x5)
+				endian.unchecked_put_u32le(dst[24:28], x6)
+				endian.unchecked_put_u32le(dst[28:32], x7)
+				endian.unchecked_put_u32le(dst[32:36], x8)
+				endian.unchecked_put_u32le(dst[36:40], x9)
+				endian.unchecked_put_u32le(dst[40:44], x10)
+				endian.unchecked_put_u32le(dst[44:48], x11)
+				endian.unchecked_put_u32le(dst[48:52], x12)
+				endian.unchecked_put_u32le(dst[52:56], x13)
+				endian.unchecked_put_u32le(dst[56:60], x14)
+				endian.unchecked_put_u32le(dst[60:64], 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
@@ -451,141 +406,141 @@ _do_blocks :: proc (ctx: ^Context, dst, src: []byte, nr_blocks: int) {
 }
 
 @(private)
-_hchacha20 :: proc (dst, key, nonce: []byte) {
+_hchacha20 :: proc "contextless" (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])
+	x4 := endian.unchecked_get_u32le(key[0:4])
+	x5 := endian.unchecked_get_u32le(key[4:8])
+	x6 := endian.unchecked_get_u32le(key[8:12])
+	x7 := endian.unchecked_get_u32le(key[12:16])
+	x8 := endian.unchecked_get_u32le(key[16:20])
+	x9 := endian.unchecked_get_u32le(key[20:24])
+	x10 := endian.unchecked_get_u32le(key[24:28])
+	x11 := endian.unchecked_get_u32le(key[28:32])
+	x12 := endian.unchecked_get_u32le(nonce[0:4])
+	x13 := endian.unchecked_get_u32le(nonce[4:8])
+	x14 := endian.unchecked_get_u32le(nonce[8:12])
+	x15 := endian.unchecked_get_u32le(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)
+		x12 = bits.rotate_left32(x12, 16)
 		x8 += x12
 		x4 ~= x8
-		x4 = util.ROTL32(x4, 12)
+		x4 = bits.rotate_left32(x4, 12)
 		x0 += x4
 		x12 ~= x0
-		x12 = util.ROTL32(x12, 8)
+		x12 = bits.rotate_left32(x12, 8)
 		x8 += x12
 		x4 ~= x8
-		x4 = util.ROTL32(x4, 7)
+		x4 = bits.rotate_left32(x4, 7)
 
 		// quarterround(x, 1, 5, 9, 13)
 		x1 += x5
 		x13 ~= x1
-		x13 = util.ROTL32(x13, 16)
+		x13 = bits.rotate_left32(x13, 16)
 		x9 += x13
 		x5 ~= x9
-		x5 = util.ROTL32(x5, 12)
+		x5 = bits.rotate_left32(x5, 12)
 		x1 += x5
 		x13 ~= x1
-		x13 = util.ROTL32(x13, 8)
+		x13 = bits.rotate_left32(x13, 8)
 		x9 += x13
 		x5 ~= x9
-		x5 = util.ROTL32(x5, 7)
+		x5 = bits.rotate_left32(x5, 7)
 
 		// quarterround(x, 2, 6, 10, 14)
 		x2 += x6
 		x14 ~= x2
-		x14 = util.ROTL32(x14, 16)
+		x14 = bits.rotate_left32(x14, 16)
 		x10 += x14
 		x6 ~= x10
-		x6 = util.ROTL32(x6, 12)
+		x6 = bits.rotate_left32(x6, 12)
 		x2 += x6
 		x14 ~= x2
-		x14 = util.ROTL32(x14, 8)
+		x14 = bits.rotate_left32(x14, 8)
 		x10 += x14
 		x6 ~= x10
-		x6 = util.ROTL32(x6, 7)
+		x6 = bits.rotate_left32(x6, 7)
 
 		// quarterround(x, 3, 7, 11, 15)
 		x3 += x7
 		x15 ~= x3
-		x15 = util.ROTL32(x15, 16)
+		x15 = bits.rotate_left32(x15, 16)
 		x11 += x15
 		x7 ~= x11
-		x7 = util.ROTL32(x7, 12)
+		x7 = bits.rotate_left32(x7, 12)
 		x3 += x7
 		x15 ~= x3
-		x15 = util.ROTL32(x15, 8)
+		x15 = bits.rotate_left32(x15, 8)
 		x11 += x15
 		x7 ~= x11
-		x7 = util.ROTL32(x7, 7)
+		x7 = bits.rotate_left32(x7, 7)
 
 		// quarterround(x, 0, 5, 10, 15)
 		x0 += x5
 		x15 ~= x0
-		x15 = util.ROTL32(x15, 16)
+		x15 = bits.rotate_left32(x15, 16)
 		x10 += x15
 		x5 ~= x10
-		x5 = util.ROTL32(x5, 12)
+		x5 = bits.rotate_left32(x5, 12)
 		x0 += x5
 		x15 ~= x0
-		x15 = util.ROTL32(x15, 8)
+		x15 = bits.rotate_left32(x15, 8)
 		x10 += x15
 		x5 ~= x10
-		x5 = util.ROTL32(x5, 7)
+		x5 = bits.rotate_left32(x5, 7)
 
 		// quarterround(x, 1, 6, 11, 12)
 		x1 += x6
 		x12 ~= x1
-		x12 = util.ROTL32(x12, 16)
+		x12 = bits.rotate_left32(x12, 16)
 		x11 += x12
 		x6 ~= x11
-		x6 = util.ROTL32(x6, 12)
+		x6 = bits.rotate_left32(x6, 12)
 		x1 += x6
 		x12 ~= x1
-		x12 = util.ROTL32(x12, 8)
+		x12 = bits.rotate_left32(x12, 8)
 		x11 += x12
 		x6 ~= x11
-		x6 = util.ROTL32(x6, 7)
+		x6 = bits.rotate_left32(x6, 7)
 
 		// quarterround(x, 2, 7, 8, 13)
 		x2 += x7
 		x13 ~= x2
-		x13 = util.ROTL32(x13, 16)
+		x13 = bits.rotate_left32(x13, 16)
 		x8 += x13
 		x7 ~= x8
-		x7 = util.ROTL32(x7, 12)
+		x7 = bits.rotate_left32(x7, 12)
 		x2 += x7
 		x13 ~= x2
-		x13 = util.ROTL32(x13, 8)
+		x13 = bits.rotate_left32(x13, 8)
 		x8 += x13
 		x7 ~= x8
-		x7 = util.ROTL32(x7, 7)
+		x7 = bits.rotate_left32(x7, 7)
 
 		// quarterround(x, 3, 4, 9, 14)
 		x3 += x4
 		x14 ~= x3
-		x14 = util.ROTL32(x14, 16)
+		x14 = bits.rotate_left32(x14, 16)
 		x9 += x14
 		x4 ~= x9
-		x4 = util.ROTL32(x4, 12)
+		x4 = bits.rotate_left32(x4, 12)
 		x3 += x4
 		x14 ~= x3
-		x14 = util.ROTL32(x14, 8)
+		x14 = bits.rotate_left32(x14, 8)
 		x9 += x14
 		x4 ~= x9
-		x4 = util.ROTL32(x4, 7)
+		x4 = bits.rotate_left32(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)
+	endian.unchecked_put_u32le(dst[0:4], x0)
+	endian.unchecked_put_u32le(dst[4:8], x1)
+	endian.unchecked_put_u32le(dst[8:12], x2)
+	endian.unchecked_put_u32le(dst[12:16], x3)
+	endian.unchecked_put_u32le(dst[16:20], x12)
+	endian.unchecked_put_u32le(dst[20:24], x13)
+	endian.unchecked_put_u32le(dst[24:28], x14)
+	endian.unchecked_put_u32le(dst[28:32], x15)
 }

core/crypto/chacha20poly1305/chacha20poly1305.odin

@@ -3,7 +3,7 @@ package chacha20poly1305
 import "core:crypto"
 import "core:crypto/chacha20"
 import "core:crypto/poly1305"
-import "core:crypto/util"
+import "core:encoding/endian"
 import "core:mem"
 
 KEY_SIZE :: chacha20.KEY_SIZE
@@ -87,8 +87,8 @@ encrypt :: proc (ciphertext, tag, key, nonce, aad, plaintext: []byte) {
 	// 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))
+	endian.unchecked_put_u64le(l_buf[0:8], u64(aad_len))
+	endian.unchecked_put_u64le(l_buf[8:16], u64(ciphertext_len))
 	poly1305.update(&mac_ctx, l_buf)
 
 	// tag = poly1305_mac(mac_data, otk)
@@ -128,8 +128,8 @@ decrypt :: proc (plaintext, tag, key, nonce, aad, ciphertext: []byte) -> bool {
 	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))
+	endian.unchecked_put_u64le(l_buf[0:8], u64(aad_len))
+	endian.unchecked_put_u64le(l_buf[8:16], u64(ciphertext_len))
 	poly1305.update(&mac_ctx, l_buf)
 
 	// tag = poly1305_mac(mac_data, otk)

core/crypto/gost/gost.odin

@@ -1,382 +0,0 @@
-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
-*/
-
-DIGEST_SIZE :: 32
-
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-    return hash_bytes(transmute([]byte)(data))
-}
-
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-    hash: [DIGEST_SIZE]byte
-    ctx: Gost_Context
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE, "Size of destination buffer is smaller than the digest size")
-    ctx: Gost_Context
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-    hash: [DIGEST_SIZE]byte
-    ctx: Gost_Context
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE]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 [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-    hash_stream,
-    hash_file,
-    hash_bytes,
-    hash_string,
-    hash_bytes_to_buffer,
-    hash_string_to_buffer,
-}
-
-/*
-    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

@@ -1,653 +0,0 @@
-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
-*/
-
-DIGEST_SIZE_224 :: 28
-DIGEST_SIZE_256 :: 32
-DIGEST_SIZE_384 :: 48
-DIGEST_SIZE_512 :: 64
-
-// hash_string_224 will hash the given input and return the
-// computed hash
-hash_string_224 :: proc(data: string) -> [DIGEST_SIZE_224]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) -> [DIGEST_SIZE_224]byte {
-    hash: [DIGEST_SIZE_224]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 224
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_224 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_224 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_224(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_224 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_224 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_224, "Size of destination buffer is smaller than the digest size")
-    ctx: Groestl_Context
-    ctx.hashbitlen = 224
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_224 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_224 :: proc(s: io.Stream) -> ([DIGEST_SIZE_224]byte, bool) {
-    hash: [DIGEST_SIZE_224]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 224
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_224]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 [DIGEST_SIZE_224]byte{}, false
-}
-
-hash_224 :: proc {
-    hash_stream_224,
-    hash_file_224,
-    hash_bytes_224,
-    hash_string_224,
-    hash_bytes_to_buffer_224,
-    hash_string_to_buffer_224,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc(data: string) -> [DIGEST_SIZE_256]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) -> [DIGEST_SIZE_256]byte {
-    hash: [DIGEST_SIZE_256]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 256
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_256, "Size of destination buffer is smaller than the digest size")
-    ctx: Groestl_Context
-    ctx.hashbitlen = 256
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-    hash: [DIGEST_SIZE_256]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 256
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_256]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 [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-    hash_stream_256,
-    hash_file_256,
-    hash_bytes_256,
-    hash_string_256,
-    hash_bytes_to_buffer_256,
-    hash_string_to_buffer_256,
-}
-
-// hash_string_384 will hash the given input and return the
-// computed hash
-hash_string_384 :: proc(data: string) -> [DIGEST_SIZE_384]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) -> [DIGEST_SIZE_384]byte {
-    hash: [DIGEST_SIZE_384]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 384
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_384 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_384 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_384(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_384 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_384 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_384, "Size of destination buffer is smaller than the digest size")
-    ctx: Groestl_Context
-    ctx.hashbitlen = 384
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_384 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_384 :: proc(s: io.Stream) -> ([DIGEST_SIZE_384]byte, bool) {
-    hash: [DIGEST_SIZE_384]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 384
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_384]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 [DIGEST_SIZE_384]byte{}, false
-}
-
-hash_384 :: proc {
-    hash_stream_384,
-    hash_file_384,
-    hash_bytes_384,
-    hash_string_384,
-    hash_bytes_to_buffer_384,
-    hash_string_to_buffer_384,
-}
-
-// hash_string_512 will hash the given input and return the
-// computed hash
-hash_string_512 :: proc(data: string) -> [DIGEST_SIZE_512]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) -> [DIGEST_SIZE_512]byte {
-    hash: [DIGEST_SIZE_512]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 512
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash[:])
-    return hash
-}
-
-// hash_string_to_buffer_512 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512 :: proc(data: string, hash: []byte) {
-    hash_bytes_to_buffer_512(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512 :: proc(data, hash: []byte) {
-    assert(len(hash) >= DIGEST_SIZE_512, "Size of destination buffer is smaller than the digest size")
-    ctx: Groestl_Context
-    ctx.hashbitlen = 512
-    init(&ctx)
-    update(&ctx, data)
-    final(&ctx, hash)
-}
-
-// hash_stream_512 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512]byte, bool) {
-    hash: [DIGEST_SIZE_512]byte
-    ctx: Groestl_Context
-    ctx.hashbitlen = 512
-    init(&ctx)
-    buf := make([]byte, 512)
-    defer delete(buf)
-    read := 1
-    for read > 0 {
-        read, _ = io.read(s, 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) -> ([DIGEST_SIZE_512]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 [DIGEST_SIZE_512]byte{}, false
-}
-
-hash_512 :: proc {
-    hash_stream_512,
-    hash_file_512,
-    hash_bytes_512,
-    hash_string_512,
-    hash_bytes_to_buffer_512,
-    hash_string_to_buffer_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/hash/doc.odin

@@ -0,0 +1,62 @@
+/*
+package hash provides a generic interface to the supported hash algorithms.
+
+A high-level convenience procedure group `hash` is provided to easily
+accomplish common tasks.
+- `hash_string` - Hash a given string and return the digest.
+- `hash_bytes` - Hash a given byte slice and return the digest.
+- `hash_string_to_buffer` - Hash a given string and put the digest in
+  the third parameter.  It requires that the destination buffer
+  is at least as big as the digest size.
+- `hash_bytes_to_buffer` - Hash a given string and put the computed
+  digest in the third parameter.  It requires that the destination
+  buffer is at least as big as the digest size.
+- `hash_stream` - Incrementally fully consume a `io.Stream`, and return
+  the computed digest.
+- `hash_file` - Takes a file handle and returns the computed digest.
+  A third optional boolean parameter controls if the file is streamed
+  (default), or or read at once.
+
+```odin
+package hash_example
+
+import "core:crypto/hash"
+
+main :: proc() {
+	input := "Feed the fire."
+
+	// Compute the digest, using the high level API.
+	returned_digest := hash.hash(hash.Algorithm.SHA512_256, input)
+	defer delete(returned_digest)
+
+	// Variant that takes a destination buffer, instead of returning
+	// the digest.
+	digest := make([]byte, hash.DIGEST_SIZES[hash.Algorithm.BLAKE2B]) // @note: Destination buffer has to be at least as big as the digest size of the hash.
+	defer delete(digest)
+	hash.hash(hash.Algorithm.BLAKE2B, input, digest)
+}
+```
+
+A generic low level API is provided supporting the init/update/final interface
+that is typical with cryptographic hash function implementations.
+
+```odin
+package hash_example
+
+import "core:crypto/hash"
+
+main :: proc() {
+    input := "Let the cinders burn."
+
+    // Compute the digest, using the low level API.
+    ctx: hash.Context
+    digest := make([]byte, hash.DIGEST_SIZES[hash.Algorithm.SHA3_512])
+    defer delete(digest)
+
+    hash.init(&ctx, hash.Algorithm.SHA3_512)
+    hash.update(&ctx, transmute([]byte)input)
+    hash.final(&ctx, digest)
+}
+```
+*/
+package crypto_hash

core/crypto/hash/hash.odin

@@ -0,0 +1,116 @@
+package crypto_hash
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+*/
+
+import "core:io"
+import "core:mem"
+import "core:os"
+
+// hash_bytes will hash the given input and return the computed digest
+// in a newly allocated slice.
+hash_string :: proc(algorithm: Algorithm, data: string, allocator := context.allocator) -> []byte {
+	return hash_bytes(algorithm, transmute([]byte)(data), allocator)
+}
+
+// hash_bytes will hash the given input and return the computed digest
+// in a newly allocated slice.
+hash_bytes :: proc(algorithm: Algorithm, data: []byte, allocator := context.allocator) -> []byte {
+	dst := make([]byte, DIGEST_SIZES[algorithm], allocator)
+	hash_bytes_to_buffer(algorithm, data, dst)
+	return dst
+}
+
+// hash_string_to_buffer will hash the given input and assign the
+// computed digest to the third parameter.  It requires that the
+// destination buffer is at least as big as the digest size.
+hash_string_to_buffer :: proc(algorithm: Algorithm, data: string, hash: []byte) {
+	hash_bytes_to_buffer(algorithm, transmute([]byte)(data), hash)
+}
+
+// hash_bytes_to_buffer will hash the given input and write the
+// computed digest into the third parameter.  It requires that the
+// destination buffer is at least as big as the digest size.
+hash_bytes_to_buffer :: proc(algorithm: Algorithm, data, hash: []byte) {
+	ctx: Context
+
+	init(&ctx, algorithm)
+	update(&ctx, data)
+	final(&ctx, hash)
+}
+
+// hash_stream will incrementally fully consume a stream, and return the
+// computed digest in a newly allocated slice.
+hash_stream :: proc(
+	algorithm: Algorithm,
+	s: io.Stream,
+	allocator := context.allocator,
+) -> (
+	[]byte,
+	io.Error,
+) {
+	ctx: Context
+
+	buf: [MAX_BLOCK_SIZE * 4]byte
+	defer mem.zero_explicit(&buf, size_of(buf))
+
+	init(&ctx, algorithm)
+
+	loop: for {
+		n, err := io.read(s, buf[:])
+		if n > 0 {
+			// XXX/yawning: Can io.read return n > 0 and EOF?
+			update(&ctx, buf[:n])
+		}
+		#partial switch err {
+		case .None:
+		case .EOF:
+			break loop
+		case:
+			return nil, err
+		}
+	}
+
+	dst := make([]byte, DIGEST_SIZES[algorithm], allocator)
+	final(&ctx, dst)
+
+	return dst, io.Error.None
+}
+
+// hash_file will read the file provided by the given handle and return the
+// computed digest in a newly allocated slice.
+hash_file :: proc(
+	algorithm: Algorithm,
+	hd: os.Handle,
+	load_at_once := false,
+	allocator := context.allocator,
+) -> (
+	[]byte,
+	io.Error,
+) {
+	if !load_at_once {
+		return hash_stream(algorithm, os.stream_from_handle(hd), allocator)
+	}
+
+	buf, ok := os.read_entire_file(hd, allocator)
+	if !ok {
+		return nil, io.Error.Unknown
+	}
+	defer delete(buf, allocator)
+
+	return hash_bytes(algorithm, buf, allocator), io.Error.None
+}
+
+hash :: proc {
+	hash_stream,
+	hash_file,
+	hash_bytes,
+	hash_string,
+	hash_bytes_to_buffer,
+	hash_string_to_buffer,
+}

core/crypto/hash/low_level.odin

@@ -0,0 +1,353 @@
+package crypto_hash
+
+import "core:crypto/blake2b"
+import "core:crypto/blake2s"
+import "core:crypto/sha2"
+import "core:crypto/sha3"
+import "core:crypto/sm3"
+import "core:crypto/legacy/keccak"
+import "core:crypto/legacy/md5"
+import "core:crypto/legacy/sha1"
+
+import "core:reflect"
+
+// MAX_DIGEST_SIZE is the maximum size digest that can be returned by any
+// of the Algorithms supported via this package.
+MAX_DIGEST_SIZE :: 64
+// MAX_BLOCK_SIZE is the maximum block size used by any of Algorithms
+// supported by this package.
+MAX_BLOCK_SIZE :: sha3.BLOCK_SIZE_224
+
+// Algorithm is the algorithm identifier associated with a given Context.
+Algorithm :: enum {
+	Invalid,
+	BLAKE2B,
+	BLAKE2S,
+	SHA224,
+	SHA256,
+	SHA384,
+	SHA512,
+	SHA512_256,
+	SHA3_224,
+	SHA3_256,
+	SHA3_384,
+	SHA3_512,
+	SM3,
+	Legacy_KECCAK_224,
+	Legacy_KECCAK_256,
+	Legacy_KECCAK_384,
+	Legacy_KECCAK_512,
+	Insecure_MD5,
+	Insecure_SHA1,
+}
+
+// ALGORITHM_NAMES is the Algorithm to algorithm name string.
+ALGORITHM_NAMES := [Algorithm]string {
+	.Invalid           = "Invalid",
+	.BLAKE2B           = "BLAKE2b",
+	.BLAKE2S           = "BLAKE2s",
+	.SHA224            = "SHA-224",
+	.SHA256            = "SHA-256",
+	.SHA384            = "SHA-384",
+	.SHA512            = "SHA-512",
+	.SHA512_256        = "SHA-512/256",
+	.SHA3_224          = "SHA3-224",
+	.SHA3_256          = "SHA3-256",
+	.SHA3_384          = "SHA3-384",
+	.SHA3_512          = "SHA3-512",
+	.SM3               = "SM3",
+	.Legacy_KECCAK_224 = "Keccak-224",
+	.Legacy_KECCAK_256 = "Keccak-256",
+	.Legacy_KECCAK_384 = "Keccak-384",
+	.Legacy_KECCAK_512 = "Keccak-512",
+	.Insecure_MD5      = "MD5",
+	.Insecure_SHA1     = "SHA-1",
+}
+
+// DIGEST_SIZES is the Algorithm to digest size in bytes.
+DIGEST_SIZES := [Algorithm]int {
+	.Invalid           = 0,
+	.BLAKE2B           = blake2b.DIGEST_SIZE,
+	.BLAKE2S           = blake2s.DIGEST_SIZE,
+	.SHA224            = sha2.DIGEST_SIZE_224,
+	.SHA256            = sha2.DIGEST_SIZE_256,
+	.SHA384            = sha2.DIGEST_SIZE_384,
+	.SHA512            = sha2.DIGEST_SIZE_512,
+	.SHA512_256        = sha2.DIGEST_SIZE_512_256,
+	.SHA3_224          = sha3.DIGEST_SIZE_224,
+	.SHA3_256          = sha3.DIGEST_SIZE_256,
+	.SHA3_384          = sha3.DIGEST_SIZE_384,
+	.SHA3_512          = sha3.DIGEST_SIZE_512,
+	.SM3               = sm3.DIGEST_SIZE,
+	.Legacy_KECCAK_224 = keccak.DIGEST_SIZE_224,
+	.Legacy_KECCAK_256 = keccak.DIGEST_SIZE_256,
+	.Legacy_KECCAK_384 = keccak.DIGEST_SIZE_384,
+	.Legacy_KECCAK_512 = keccak.DIGEST_SIZE_512,
+	.Insecure_MD5      = md5.DIGEST_SIZE,
+	.Insecure_SHA1     = sha1.DIGEST_SIZE,
+}
+
+// BLOCK_SIZES is the Algoritm to block size in bytes.
+BLOCK_SIZES := [Algorithm]int {
+	.Invalid           = 0,
+	.BLAKE2B           = blake2b.BLOCK_SIZE,
+	.BLAKE2S           = blake2s.BLOCK_SIZE,
+	.SHA224            = sha2.BLOCK_SIZE_256,
+	.SHA256            = sha2.BLOCK_SIZE_256,
+	.SHA384            = sha2.BLOCK_SIZE_512,
+	.SHA512            = sha2.BLOCK_SIZE_512,
+	.SHA512_256        = sha2.BLOCK_SIZE_512,
+	.SHA3_224          = sha3.BLOCK_SIZE_224,
+	.SHA3_256          = sha3.BLOCK_SIZE_256,
+	.SHA3_384          = sha3.BLOCK_SIZE_384,
+	.SHA3_512          = sha3.BLOCK_SIZE_512,
+	.SM3               = sm3.BLOCK_SIZE,
+	.Legacy_KECCAK_224 = keccak.BLOCK_SIZE_224,
+	.Legacy_KECCAK_256 = keccak.BLOCK_SIZE_256,
+	.Legacy_KECCAK_384 = keccak.BLOCK_SIZE_384,
+	.Legacy_KECCAK_512 = keccak.BLOCK_SIZE_512,
+	.Insecure_MD5      = md5.BLOCK_SIZE,
+	.Insecure_SHA1     = sha1.BLOCK_SIZE,
+}
+
+// Context is a concrete instantiation of a specific hash algorithm.
+Context :: struct {
+	_algo: Algorithm,
+	_impl: union {
+		blake2b.Context,
+		blake2s.Context,
+		sha2.Context_256,
+		sha2.Context_512,
+		sha3.Context,
+		sm3.Context,
+		keccak.Context,
+		md5.Context,
+		sha1.Context,
+	},
+}
+
+@(private)
+_IMPL_IDS := [Algorithm]typeid {
+	.Invalid           = nil,
+	.BLAKE2B           = typeid_of(blake2b.Context),
+	.BLAKE2S           = typeid_of(blake2s.Context),
+	.SHA224            = typeid_of(sha2.Context_256),
+	.SHA256            = typeid_of(sha2.Context_256),
+	.SHA384            = typeid_of(sha2.Context_512),
+	.SHA512            = typeid_of(sha2.Context_512),
+	.SHA512_256        = typeid_of(sha2.Context_512),
+	.SHA3_224          = typeid_of(sha3.Context),
+	.SHA3_256          = typeid_of(sha3.Context),
+	.SHA3_384          = typeid_of(sha3.Context),
+	.SHA3_512          = typeid_of(sha3.Context),
+	.SM3               = typeid_of(sm3.Context),
+	.Legacy_KECCAK_224 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_256 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_384 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_512 = typeid_of(keccak.Context),
+	.Insecure_MD5      = typeid_of(md5.Context),
+	.Insecure_SHA1     = typeid_of(sha1.Context),
+}
+
+// init initializes a Context with a specific hash Algorithm.
+init :: proc(ctx: ^Context, algorithm: Algorithm) {
+	if ctx._impl != nil {
+		reset(ctx)
+	}
+
+	// Directly specialize the union by setting the type ID (save a copy).
+	reflect.set_union_variant_typeid(
+		ctx._impl,
+		_IMPL_IDS[algorithm],
+	)
+	switch algorithm {
+	case .BLAKE2B:
+		blake2b.init(&ctx._impl.(blake2b.Context))
+	case .BLAKE2S:
+		blake2s.init(&ctx._impl.(blake2s.Context))
+	case .SHA224:
+		sha2.init_224(&ctx._impl.(sha2.Context_256))
+	case .SHA256:
+		sha2.init_256(&ctx._impl.(sha2.Context_256))
+	case .SHA384:
+		sha2.init_384(&ctx._impl.(sha2.Context_512))
+	case .SHA512:
+		sha2.init_512(&ctx._impl.(sha2.Context_512))
+	case .SHA512_256:
+		sha2.init_512_256(&ctx._impl.(sha2.Context_512))
+	case .SHA3_224:
+		sha3.init_224(&ctx._impl.(sha3.Context))
+	case .SHA3_256:
+		sha3.init_256(&ctx._impl.(sha3.Context))
+	case .SHA3_384:
+		sha3.init_384(&ctx._impl.(sha3.Context))
+	case .SHA3_512:
+		sha3.init_512(&ctx._impl.(sha3.Context))
+	case .SM3:
+		sm3.init(&ctx._impl.(sm3.Context))
+	case .Legacy_KECCAK_224:
+		keccak.init_224(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_256:
+		keccak.init_256(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_384:
+		keccak.init_384(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_512:
+		keccak.init_512(&ctx._impl.(keccak.Context))
+	case .Insecure_MD5:
+		md5.init(&ctx._impl.(md5.Context))
+	case .Insecure_SHA1:
+		sha1.init(&ctx._impl.(sha1.Context))
+	case .Invalid:
+		panic("crypto/hash: uninitialized algorithm")
+	case:
+		panic("crypto/hash: invalid algorithm")
+	}
+
+	ctx._algo = algorithm
+}
+
+// update adds more data to the Context.
+update :: proc(ctx: ^Context, data: []byte) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.update(&impl, data)
+	case blake2s.Context:
+		blake2s.update(&impl, data)
+	case sha2.Context_256:
+		sha2.update(&impl, data)
+	case sha2.Context_512:
+		sha2.update(&impl, data)
+	case sha3.Context:
+		sha3.update(&impl, data)
+	case sm3.Context:
+		sm3.update(&impl, data)
+	case keccak.Context:
+		keccak.update(&impl, data)
+	case md5.Context:
+		md5.update(&impl, data)
+	case sha1.Context:
+		sha1.update(&impl, data)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+}
+
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.final(&impl, hash, finalize_clone)
+	case blake2s.Context:
+		blake2s.final(&impl, hash, finalize_clone)
+	case sha2.Context_256:
+		sha2.final(&impl, hash, finalize_clone)
+	case sha2.Context_512:
+		sha2.final(&impl, hash, finalize_clone)
+	case sha3.Context:
+		sha3.final(&impl, hash, finalize_clone)
+	case sm3.Context:
+		sm3.final(&impl, hash, finalize_clone)
+	case keccak.Context:
+		keccak.final(&impl, hash, finalize_clone)
+	case md5.Context:
+		md5.final(&impl, hash, finalize_clone)
+	case sha1.Context:
+		sha1.final(&impl, hash, finalize_clone)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+
+	if !finalize_clone {
+		reset(ctx)
+	}
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	// XXX/yawning: Maybe these cases should panic, because both cases,
+	// are probably bugs.
+	if ctx == other {
+		return
+	}
+	if ctx._impl != nil {
+		reset(ctx)
+	}
+
+	ctx._algo = other._algo
+
+	reflect.set_union_variant_typeid(
+		ctx._impl,
+		reflect.union_variant_typeid(other._impl),
+	)
+	switch &src_impl in other._impl {
+	case blake2b.Context:
+		blake2b.clone(&ctx._impl.(blake2b.Context), &src_impl)
+	case blake2s.Context:
+		blake2s.clone(&ctx._impl.(blake2s.Context), &src_impl)
+	case sha2.Context_256:
+		sha2.clone(&ctx._impl.(sha2.Context_256), &src_impl)
+	case sha2.Context_512:
+		sha2.clone(&ctx._impl.(sha2.Context_512), &src_impl)
+	case sha3.Context:
+		sha3.clone(&ctx._impl.(sha3.Context), &src_impl)
+	case sm3.Context:
+		sm3.clone(&ctx._impl.(sm3.Context), &src_impl)
+	case keccak.Context:
+		keccak.clone(&ctx._impl.(keccak.Context), &src_impl)
+	case md5.Context:
+		md5.clone(&ctx._impl.(md5.Context), &src_impl)
+	case sha1.Context:
+		sha1.clone(&ctx._impl.(sha1.Context), &src_impl)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.reset(&impl)
+	case blake2s.Context:
+		blake2s.reset(&impl)
+	case sha2.Context_256:
+		sha2.reset(&impl)
+	case sha2.Context_512:
+		sha2.reset(&impl)
+	case sha3.Context:
+		sha3.reset(&impl)
+	case sm3.Context:
+		sm3.reset(&impl)
+	case keccak.Context:
+		keccak.reset(&impl)
+	case md5.Context:
+		md5.reset(&impl)
+	case sha1.Context:
+		sha1.reset(&impl)
+	case:
+	// Unlike clone, calling reset repeatedly is fine.
+	}
+
+	ctx._algo = .Invalid
+	ctx._impl = nil
+}
+
+// algorithm returns the Algorithm used by a Context instance.
+algorithm :: proc(ctx: ^Context) -> Algorithm {
+	return ctx._algo
+}
+
+// digest_size returns the digest size of a Context instance in bytes.
+digest_size :: proc(ctx: ^Context) -> int {
+	return DIGEST_SIZES[ctx._algo]
+}
+
+// block_size returns the block size of a Context instance in bytes.
+block_size :: proc(ctx: ^Context) -> int {
+	return BLOCK_SIZES[ctx._algo]
+}