Merge branch 'master' of https://github.com/odin-lang/Odin

Partial pseudo-revert
Merge pull request #2839 from Tetralux/aprintf-ally
2026-07-05 03:01:38 -07:00 · 2023-10-02 20:59:49 +01:00 · 2023-10-02 20:59:43 +01:00 · 2023-10-02 11:23:26 +02:00 · 2023-10-02 02:46:26 +00:00 · 2023-09-30 22:47:49 +02:00
1066 changed files with 289702 additions and 71939 deletions
@@ -1,3 +1,4 @@
 # These are supported funding model platforms

+github: odin-lang
 patreon: gingerbill
@@ -11,8 +11,8 @@ assignees: ''

 Please provide any relevant information about your setup. This is important in case the issue is not reproducible except for under certain conditions.

-* Operating System:
-* Please paste `odin version` output:
+* Operating System & Odin Version:
+* Please paste `odin report` output:

 ## Expected Behavior

@@ -1,85 +1,185 @@
 name: CI
-on: [push, pull_request]
+on: [push, pull_request, workflow_dispatch]

 jobs:
  build_linux:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v1
-      - name: Download LLVM
-        run: sudo apt-get install llvm-11 clang-11 llvm
+      - name: Download LLVM, botan
+        run: sudo apt-get install llvm-11 clang-11 libbotan-2-dev botan
      - name: build odin
-        run: make release
+        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/demo.odin -vet
+        run: ./odin check examples/demo -vet
        timeout-minutes: 10
      - name: Odin run
-        run: ./odin run examples/demo/demo.odin
+        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: Vendor library tests
+        run: |
+          cd tests/vendor
+          make
+        timeout-minutes: 10
+      - name: Odin internals tests
+        run: |
+          cd tests/internal
+          make
+        timeout-minutes: 10
+      - 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 FreeBSD amd64
+        run: ./odin check examples/all -vet -strict-style -target:freebsd_amd64
+        timeout-minutes: 10
+      - name: Odin check examples/all for OpenBSD amd64
+        run: ./odin check examples/all -vet -strict-style -target:openbsd_amd64
+        timeout-minutes: 10
  build_macOS:
    runs-on: macos-latest
    steps:
      - uses: actions/checkout@v1
-      - name: Download LLVM and setup PATH
+      - name: Download LLVM, botan and setup PATH
        run: |
-          brew install llvm@11
-          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
-        run: make release
+        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/demo.odin -vet
+        run: ./odin check examples/demo -vet
        timeout-minutes: 10
      - name: Odin run
-        run: ./odin run examples/demo/demo.odin
+        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: Vendor library tests
+        run: |
+          cd tests/vendor
+          make
+        timeout-minutes: 10
+      - name: Odin internals tests
+        run: |
+          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:
-    runs-on: windows-latest
+    runs-on: windows-2022
    steps:
      - uses: actions/checkout@v1
      - name: build Odin
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
          ./build.bat 1
      - name: Odin version
        run: ./odin version
        timeout-minutes: 1
+      - name: Odin report
+        run: ./odin report
+        timeout-minutes: 1
      - name: Odin check
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
-          odin check examples/demo/demo.odin -vet
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin check examples/demo -vet
        timeout-minutes: 10
      - name: Odin run
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
-          odin run examples/demo/demo.odin
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin run examples/demo
+        timeout-minutes: 10
+      - name: Odin run -debug
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin run examples/demo -debug
+        timeout-minutes: 10
+      - name: Odin check examples/all
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin check examples/all -strict-style
        timeout-minutes: 10
      - name: Core library tests
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
          cd tests\core
          call build.bat
        timeout-minutes: 10
+      - name: Vendor library tests
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          cd tests\vendor
+          call build.bat
+        timeout-minutes: 10
+      - name: Odin internals tests
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          cd tests\internal
+          call build.bat
+        timeout-minutes: 10
+      - name: Odin documentation tests
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          cd tests\documentation
+          call build.bat
+        timeout-minutes: 10
      - name: core:math/big tests
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
          cd tests\core\math\big
          call build.bat
        timeout-minutes: 10
+      - name: Odin check examples/all for Windows 32bits
+        shell: cmd
+        run: |
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin check examples/all -strict-style -target:windows_i386
+        timeout-minutes: 10
@@ -7,19 +7,20 @@ on:

 jobs:
  build_windows:
-    runs-on: windows-latest
+    if: github.repository == 'odin-lang/Odin'
+    runs-on: windows-2022
    steps:
      - uses: actions/checkout@v1
      - name: build Odin
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
          ./build.bat 1 1
      - name: Odin run
        shell: cmd
        run: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvars64.bat
-          odin run examples/demo/demo.odin
+          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat
+          odin run examples/demo
      - name: Copy artifacts
        run: |
          rm bin/llvm/windows/LLVM-C.lib
@@ -37,19 +38,21 @@ jobs:
          name: windows_artifacts
          path: dist
  build_ubuntu:
+    if: github.repository == 'odin-lang/Odin'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v1
      - name: (Linux) Download LLVM
-        run: sudo apt-get install llvm-11 clang-11 llvm
+        run: sudo apt-get install llvm-11 clang-11
      - name: build odin
        run: make nightly
      - name: Odin run
-        run: ./odin run examples/demo/demo.odin
+        run: ./odin run examples/demo
      - name: Copy artifacts
        run: |
          mkdir dist
          cp odin dist
+          cp libLLVM* dist
          cp -r shared dist
          cp -r core dist
          cp -r vendor dist
@@ -60,19 +63,20 @@ jobs:
          name: ubuntu_artifacts
          path: dist
  build_macos:
+    if: github.repository == 'odin-lang/Odin'
    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
+          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/demo.odin
+        run: ./odin run examples/demo
      - name: Copy artifacts
        run: |
          mkdir dist
@@ -93,7 +97,7 @@ jobs:
      - uses: actions/checkout@v1
      - uses: actions/setup-python@v2
        with:
-          python-version: '3.x'
+          python-version: '3.8.x'

      - name: Install B2 CLI
        shell: bash
@@ -127,16 +131,23 @@ jobs:
          BUCKET: ${{ secrets.B2_BUCKET }}
          DAYS_TO_KEEP: ${{ secrets.B2_DAYS_TO_KEEP }}
        run: |
+          echo Authorizing B2 account
          b2 authorize-account "$APPID" "$APPKEY"

+          echo Uploading artifcates to B2
          chmod +x ./ci/upload_create_nightly.sh
          ./ci/upload_create_nightly.sh "$BUCKET" windows-amd64 windows_artifacts/
          ./ci/upload_create_nightly.sh "$BUCKET" ubuntu-amd64 ubuntu_artifacts/
          ./ci/upload_create_nightly.sh "$BUCKET" macos-amd64 macos_artifacts/

+          echo Deleting old artifacts in B2
          python3 ci/delete_old_binaries.py "$BUCKET" "$DAYS_TO_KEEP"

+          echo Creating nightly.json
          python3 ci/create_nightly_json.py "$BUCKET" > nightly.json
+
+          echo Uploading nightly.json
          b2 upload-file "$BUCKET" nightly.json nightly.json

+          echo Clear B2 account info
          b2 clear-account
@@ -0,0 +1,43 @@
+name: "Close Stale Issues & PRs"
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 21 * * *"
+    
+permissions:
+  issues: write
+  pull-requests: write
+
+jobs:
+  stale:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Close Stale Issues
+        uses: actions/stale@v7.0.0
+        with:
+#          stale-issue-message: |
+#            Hello!
+#            
+#            I am marking this issue as stale as it has not received any engagement from the community or maintainers 120 days. That does not imply that the issue has no merit! If you feel strongly about this issue
+#            - open a PR referencing and resolving the issue;
+#            - leave a comment on it and discuss ideas how you could contribute towards resolving it;
+#            - leave a comment and describe in detail why this issue is critical for your use case;
+#            - open a new issue with updated details and a plan on resolving the issue.
+#
+#            The motivation for this automation is to help prioritize issues in the backlog and not ignore, reject, or belittle anyone..
+#
+#          stale-pr-message: |
+#            Hello!
+#            
+#            I am marking this PR as stale as it has not received any engagement from the community or maintainers 120 days. That does not imply that the issue has no merit! If you feel strongly about this issue
+#            - leave a comment on it and discuss ideas how you could contribute towards resolving it;
+#            - leave a comment and describe in detail why this issue is critical for your use case;
+#
+#           The motivation for this automation is to help prioritize issues in the backlog and not ignore, reject, or belittle anyone..
+
+          days-before-stale: 120
+          days-before-close: -1
+          exempt-draft-pr: true
+          ascending: true
+          operations-per-run: 1000
+          exempt-issue-labels: "ignore"
@@ -7,6 +7,9 @@
 # User-specific files (MonoDevelop/Xamarin Studio)
 *.userprefs

+# For macOS
+.DS_Store
+
 # Build results
 [Dd]ebug/
 [Dd]ebugPublic/
@@ -19,6 +22,8 @@ bld/
 [Oo]bj/
 [Ll]og/
 ![Cc]ore/[Ll]og/
+tests/documentation/verify/
+tests/documentation/all.odin-doc
 # Visual Studio 2015 cache/options directory
 .vs/
 # Visual Studio Code options directory
@@ -265,7 +270,11 @@ bin/

 # - Linux/MacOS
 odin
+!odin/
 odin.dSYM
+*.bin
+demo.bin
+libLLVM*.so*

 # shared collection
 shared/
@@ -276,3 +285,6 @@ shared/
 *.ll

 *.sublime-workspace
+examples/bug/
+build.sh
+!core/debug/
@@ -1,4 +1,4 @@
-Copyright (c) 2016-2021 Ginger Bill. All rights reserved.
+Copyright (c) 2016-2022 Ginger Bill. All rights reserved.

 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
@@ -1,53 +1,19 @@
-GIT_SHA=$(shell git rev-parse --short HEAD)
-DISABLED_WARNINGS=-Wno-switch -Wno-macro-redefined -Wno-unused-value
-LDFLAGS=-pthread -ldl -lm -lstdc++
-CFLAGS=-std=c++14 -DGIT_SHA=\"$(GIT_SHA)\"
-CFLAGS:=$(CFLAGS) -DODIN_VERSION_RAW=\"dev-$(shell date +"%Y-%m")\"
-CC=clang
-
-OS=$(shell uname)
-
-ifeq ($(OS), Darwin)
-    LLVM_CONFIG=llvm-config
-    ifneq ($(shell llvm-config --version | grep '^11\.'),)
-        LLVM_CONFIG=llvm-config
-    else
-        $(error "Requirement: llvm-config must be version 11")
-    endif
-
-    LDFLAGS:=$(LDFLAGS) -liconv
-    CFLAGS:=$(CFLAGS) $(shell $(LLVM_CONFIG) --cxxflags --ldflags)
-    LDFLAGS:=$(LDFLAGS) -lLLVM-C
-endif
-ifeq ($(OS), Linux)
-    LLVM_CONFIG=llvm-config-11
-    ifneq ($(shell which llvm-config-11 2>/dev/null),)
-        LLVM_CONFIG=llvm-config-11
-    else
-        ifneq ($(shell llvm-config --version | grep '^11\.'),)
-            LLVM_CONFIG=llvm-config
-        else
-            $(error "Requirement: llvm-config must be version 11")
-        endif
-    endif
-
-    CFLAGS:=$(CFLAGS) $(shell $(LLVM_CONFIG) --cxxflags --ldflags)
-    LDFLAGS:=$(LDFLAGS) $(shell $(LLVM_CONFIG) --libs core native --system-libs)
-endif
-
-all: debug demo
+all: debug

 demo:
-	./odin run examples/demo/demo.odin
+	./odin run examples/demo/demo.odin -file
+
+report:
+	./odin report

 debug:
-	$(CC) src/main.cpp src/libtommath.cpp $(DISABLED_WARNINGS) $(CFLAGS) -g $(LDFLAGS) -o odin
+	./build_odin.sh debug

 release:
-	$(CC) src/main.cpp src/libtommath.cpp $(DISABLED_WARNINGS) $(CFLAGS) -O3 $(LDFLAGS) -o odin
+	./build_odin.sh release

 release_native:
-	$(CC) src/main.cpp src/libtommath.cpp $(DISABLED_WARNINGS) $(CFLAGS) -O3 -march=native $(LDFLAGS) -o odin
+	./build_odin.sh release-native

 nightly:
-	$(CC) src/main.cpp src/libtommath.cpp $(DISABLED_WARNINGS) $(CFLAGS) -DNIGHTLY -O3 $(LDFLAGS) -o odin
+	./build_odin.sh nightly
@@ -11,7 +11,7 @@
        <img src="https://img.shields.io/badge/platforms-Windows%20|%20Linux%20|%20macOS-green.svg">
    </a>
    <br>
-    <a href="https://discord.gg/hnwN2Rj">
+    <a href="https://discord.com/invite/sVBPHEv">
        <img src="https://img.shields.io/discord/568138951836172421?logo=discord">
    </a>
    <a href="https://github.com/odin-lang/odin/actions">
@@ -58,6 +58,10 @@ main :: proc() {

 Instructions for downloading and installing the Odin compiler and libraries.

+#### [Nightly Builds](https://odin-lang.org/docs/nightly/)
+
+Get the latest nightly builds of Odin.
+
 ### Learning Odin

 #### [Overview of Odin](https://odin-lang.org/docs/overview)
@@ -68,23 +72,21 @@ An overview of the Odin programming language.

 Answers to common questions about Odin.

+#### [Packages](https://pkg.odin-lang.org/)
+
+Documentation for all the official packages part of the [core](https://pkg.odin-lang.org/core/) and [vendor](https://pkg.odin-lang.org/vendor/) library collections.
+
 #### [The Odin Wiki](https://github.com/odin-lang/Odin/wiki)

 A wiki maintained by the Odin community.

 #### [Odin Discord](https://discord.gg/sVBPHEv)

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

 ### Articles

-#### [The Odin Blog](https://odin-lang.org/blog)
+#### [The Odin Blog](https://odin-lang.org/news/)

 The official blog of the Odin programming language, featuring announcements, news, and in-depth articles by the Odin team and guests.

@@ -0,0 +1,29 @@
+NASM is now licensed under the 2-clause BSD license, also known as the
+simplified BSD license.
+
+    Copyright 1996-2010 the NASM Authors - All rights reserved.
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following
+    conditions are met:
+
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+      copyright notice, this list of conditions and the following
+      disclaimer in the documentation and/or other materials provided
+      with the distribution.
+      
+      THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
+      CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
+      INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+      MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+      DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+      CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+      SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+      NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+      LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+      HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+      CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+      OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
+      EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
@@ -2,6 +2,23 @@

 setlocal EnableDelayedExpansion

+where /Q cl.exe || (
+	set __VSCMD_ARG_NO_LOGO=1
+	for /f "tokens=*" %%i in ('"C:\Program Files (x86)\Microsoft Visual Studio\Installer\vswhere.exe" -latest -requires Microsoft.VisualStudio.Workload.NativeDesktop -property installationPath') do set VS=%%i
+	if "!VS!" equ "" (
+		echo ERROR: Visual Studio installation not found
+		exit /b 1
+	)
+	call "!VS!\VC\Auxiliary\Build\vcvarsall.bat" amd64 || exit /b 1
+)
+
+if "%VSCMD_ARG_TGT_ARCH%" neq "x64" (
+	if "%ODIN_IGNORE_MSVC_CHECK%" == "" (
+		echo ERROR: please run this from MSVC x64 native tools command prompt, 32-bit target is not supported!
+		exit /b 1
+	)
+)
+
 for /f "usebackq tokens=1,2 delims=,=- " %%i in (`wmic os get LocalDateTime /value`) do @if %%i==LocalDateTime (
 	set CURR_DATE_TIME=%%j
 )
@@ -33,8 +50,14 @@ set odin_version_raw="dev-%curr_year%-%curr_month%"
 set compiler_flags= -nologo -Oi -TP -fp:precise -Gm- -MP -FC -EHsc- -GR- -GF
 set compiler_defines= -DODIN_VERSION_RAW=\"%odin_version_raw%\"

-for /f %%i in ('git rev-parse --short HEAD') do set GIT_SHA=%%i
+if not exist .git\ goto skip_git_hash
+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
+
 if %nightly% equ 1 set compiler_defines=%compiler_defines% -DNIGHTLY

 if %release_mode% EQU 0 ( rem Debug
@@ -47,18 +70,30 @@ if %release_mode% EQU 0 ( rem Debug
 set compiler_warnings= ^
 	-W4 -WX ^
 	-wd4100 -wd4101 -wd4127 -wd4146 ^
+	-wd4505 ^
 	-wd4456 -wd4457

 set compiler_includes= ^
 	/Isrc\
 set libs= ^
 	kernel32.lib ^
+	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
-	set linker_flags=%linker_flags% -debug
+	set linker_flags=%linker_flags% -debug /NATVIS:src\odin_compiler.natvis
 ) else ( rem Release
 	set linker_flags=%linker_flags% -debug
 )
@@ -79,4 +114,4 @@ if %release_mode% EQU 0 odin run examples/demo

 del *.obj > NUL 2> NUL

-:end_of_build
+:end_of_build
@@ -0,0 +1,218 @@
+#!/usr/bin/env bash
+set -eu
+
+: ${CXX=clang++}
+: ${CPPFLAGS=}
+: ${CXXFLAGS=}
+: ${LDFLAGS=}
+: ${ODIN_VERSION=dev-$(date +"%Y-%m")}
+: ${GIT_SHA=}
+
+CXXFLAGS="$CXXFLAGS -std=c++14"
+LDFLAGS="$LDFLAGS -pthread -lm -lstdc++"
+
+if [ -d ".git" ] && [ $(which git) ]; then
+	versionTag=( $(git show --pretty='%cd %h' --date=format:%Y-%m --no-patch --no-notes HEAD) )
+	if [ $? -eq 0 ]; then
+		ODIN_VERSION="dev-${versionTag[0]}"
+		GIT_SHA="${versionTag[1]}"
+		CPPFLAGS="$CPPFLAGS -DGIT_SHA=\"$GIT_SHA\""
+	fi
+fi
+
+CPPFLAGS="$CPPFLAGS -DODIN_VERSION_RAW=\"$ODIN_VERSION\""
+
+DISABLED_WARNINGS="-Wno-switch -Wno-macro-redefined -Wno-unused-value"
+OS=$(uname)
+
+panic() {
+	printf "%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")
+	else
+		# allow for x86 / amd64 all llvm versions beginning from 11
+		MIN_LLVM_VERSION=("11.1.0")
+	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"
+		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 "$(command -v llvm-config13)" ]; then
+			LLVM_CONFIG=llvm-config13
+		else
+			panic "Unable to find LLVM-config"
+		fi
+	fi
+
+	CXXFLAGS="$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"
+	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
+	cp $(readlink -f $($LLVM_CONFIG --libfiles)) ./
+}
+
+build_odin() {
+	case $1 in
+	debug)
+		EXTRAFLAGS="-g"
+		;;
+	release)
+		EXTRAFLAGS="-O3"
+		;;
+	release-native)
+		local ARCH=$(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
+		;;
+	nightly)
+		EXTRAFLAGS="-DNIGHTLY -O3"
+		;;
+	*)
+		panic "Build mode unsupported!"
+		;;
+	esac
+
+	set -x
+	$CXX src/main.cpp src/libtommath.cpp $DISABLED_WARNINGS $CPPFLAGS $CXXFLAGS $EXTRAFLAGS $LDFLAGS -o odin
+	set +x
+}
+
+run_demo() {
+	./odin run examples/demo/demo.odin -file
+}
+
+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
+	build_odin debug
+	run_demo
+	exit 0
+fi
+
+if [[ $# -eq 1 ]]; then
+	case $1 in
+	report)
+		if [[ ! -f "./odin" ]]; then
+			build_odin debug
+		fi
+
+		./odin report
+		exit 0
+		;;
+	*)
+		build_odin $1
+		;;
+	esac
+
+	run_demo
+	exit 0
+else
+	panic "Too many arguments!"
+fi
@@ -15,3 +15,10 @@ if not exist "vendor\miniaudio\lib\*.lib" (
 		call build.bat
 	popd
 )
+
+
+if not exist "vendor\cgltf\lib\*.lib" (
+	pushd vendor\cgltf\src
+		call build.bat
+	popd
+)
@@ -2,25 +2,25 @@ package bufio

 import "core:io"

-// Loadahead_Reader provides io lookahead.
+// Lookahead_Reader provides io lookahead.
 // This is useful for tokenizers/parsers.
-// Loadahead_Reader is similar to bufio.Reader, but unlike bufio.Reader, Loadahead_Reader's buffer size
+// Lookahead_Reader is similar to bufio.Reader, but unlike bufio.Reader, Lookahead_Reader's buffer size
 // will EXACTLY match the specified size, whereas bufio.Reader's buffer size may differ from the specified size.
 // This makes sure that the buffer will not be accidentally read beyond the expected size.
-Loadahead_Reader :: struct {
+Lookahead_Reader :: struct {
 	r:   io.Reader,
 	buf: []byte,
 	n:   int,
 }

-lookahead_reader_init :: proc(lr: ^Loadahead_Reader, r: io.Reader, buf: []byte) -> ^Loadahead_Reader {
+lookahead_reader_init :: proc(lr: ^Lookahead_Reader, r: io.Reader, buf: []byte) -> ^Lookahead_Reader {
 	lr.r = r
 	lr.buf = buf
 	lr.n = 0
 	return lr
 }

-lookahead_reader_buffer :: proc(lr: ^Loadahead_Reader) -> []byte {
+lookahead_reader_buffer :: proc(lr: ^Lookahead_Reader) -> []byte {
 	return lr.buf[:lr.n]
 }

@@ -28,7 +28,7 @@ lookahead_reader_buffer :: proc(lr: ^Loadahead_Reader) -> []byte {
 // lookahead_reader_peek returns a slice of the Lookahead_Reader which holds n bytes
 // If the Lookahead_Reader cannot hold enough bytes, it will read from the underlying reader to populate the rest.
 // NOTE: The returned buffer is not a copy of the underlying buffer
-lookahead_reader_peek :: proc(lr: ^Loadahead_Reader, n: int) -> ([]byte, io.Error) {
+lookahead_reader_peek :: proc(lr: ^Lookahead_Reader, n: int) -> ([]byte, io.Error) {
 	switch {
 	case n < 0:
 		return nil, .Negative_Read
@@ -58,13 +58,13 @@ lookahead_reader_peek :: proc(lr: ^Loadahead_Reader, n: int) -> ([]byte, io.Erro
 // lookahead_reader_peek_all returns a slice of the Lookahead_Reader populating the full buffer
 // If the Lookahead_Reader cannot hold enough bytes, it will read from the underlying reader to populate the rest.
 // NOTE: The returned buffer is not a copy of the underlying buffer
-lookahead_reader_peek_all :: proc(lr: ^Loadahead_Reader) -> ([]byte, io.Error) {
+lookahead_reader_peek_all :: proc(lr: ^Lookahead_Reader) -> ([]byte, io.Error) {
 	return lookahead_reader_peek(lr, len(lr.buf))
 }


 // lookahead_reader_consume drops the first n populated bytes from the Lookahead_Reader.
-lookahead_reader_consume :: proc(lr: ^Loadahead_Reader, n: int) -> io.Error {
+lookahead_reader_consume :: proc(lr: ^Lookahead_Reader, n: int) -> io.Error {
 	switch {
 	case n == 0:
 		return nil
@@ -78,6 +78,6 @@ lookahead_reader_consume :: proc(lr: ^Loadahead_Reader, n: int) -> io.Error {
 	return nil
 }

-lookahead_reader_consume_all :: proc(lr: ^Loadahead_Reader) -> io.Error {
+lookahead_reader_consume_all :: proc(lr: ^Lookahead_Reader) -> io.Error {
 	return lookahead_reader_consume(lr, lr.n)
 }
@@ -14,55 +14,29 @@ read_writer_init :: proc(rw: ^Read_Writer, r: ^Reader, w: ^Writer) {
 }

 read_writer_to_stream :: proc(rw: ^Read_Writer) -> (s: io.Stream) {
-	s.stream_data = rw
-	s.stream_vtable = _read_writer_vtable
+	s.procedure = _read_writer_procedure
+	s.data = rw
 	return
 }

@(private)
-_read_writer_vtable := &io.Stream_VTable{
-	impl_read = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_read(b, p)
-	},
-	impl_read_byte = proc(s: io.Stream) -> (c: byte, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_read_byte(b)
-	},
-	impl_unread_byte = proc(s: io.Stream) -> io.Error {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_unread_byte(b)
-	},
-	impl_read_rune = proc(s: io.Stream) -> (r: rune, size: int, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_read_rune(b)
-	},
-	impl_unread_rune = proc(s: io.Stream) -> io.Error {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_unread_rune(b)
-	},
-	impl_write_to = proc(s: io.Stream, w: io.Writer) -> (n: i64, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).r
-		return reader_write_to(b, w)
-	},
-	impl_flush = proc(s: io.Stream)  -> io.Error {
-		b := (^Read_Writer)(s.stream_data).w
-		return writer_flush(b)
-	},
-	impl_write = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).w
-		return writer_write(b, p)
-	},
-	impl_write_byte = proc(s: io.Stream, c: byte) -> io.Error {
-		b := (^Read_Writer)(s.stream_data).w
-		return writer_write_byte(b, c)
-	},
-	impl_write_rune = proc(s: io.Stream, r: rune) -> (int, io.Error) {
-		b := (^Read_Writer)(s.stream_data).w
-		return writer_write_rune(b, r)
-	},
-	impl_read_from = proc(s: io.Stream, r: io.Reader) -> (n: i64, err: io.Error) {
-		b := (^Read_Writer)(s.stream_data).w
-		return writer_read_from(b, r)
-	},
-}
+_read_writer_procedure := proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+	rw := (^Read_Writer)(stream_data)
+	n_int: int
+	#partial switch mode {
+	case .Flush:
+		err = writer_flush(rw.w)
+		return
+	case .Read:
+		n_int, err = reader_read(rw.r, p)
+		n = i64(n_int)
+		return
+	case .Write:
+		n_int, err = writer_write(rw.w, p)
+		n = i64(n_int)
+		return
+	case .Query:
+		return io.query_utility({.Flush, .Read, .Write, .Query})
+	}
+	return 0, .Empty
+}
@@ -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 {
@@ -305,39 +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 nr, ok := io.to_writer_to(b.rd); ok {
-		m, err = io.write_to(nr, w)
-		n += m
-		return n, err
-	}
-
-	if nw, ok := io.to_reader_from(w); ok {
-		m, err = io.read_from(nw, b.rd)
-		n += m
-		return n, err
-	}
-
 	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 {
@@ -352,48 +324,28 @@ reader_write_to :: proc(b: ^Reader, w: io.Writer) -> (n: i64, err: io.Error) {

 // reader_to_stream converts a Reader into an io.Stream
 reader_to_stream :: proc(b: ^Reader) -> (s: io.Stream) {
-	s.stream_data = b
-	s.stream_vtable = _reader_vtable
+	s.data = b
+	s.procedure = _reader_proc
 	return
 }



@(private)
-_reader_vtable := &io.Stream_VTable{
-	impl_destroy = proc(s: io.Stream) -> io.Error {
-		b := (^Reader)(s.stream_data)
+_reader_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+	b := (^Reader)(stream_data)
+	#partial switch mode {
+	case .Read:
+		return io._i64_err(reader_read(b, p))
+	case .Destroy:
 		reader_destroy(b)
-		return nil
-	},
-	impl_read = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Reader)(s.stream_data)
-		return reader_read(b, p)
-	},
-	impl_read_byte = proc(s: io.Stream) -> (c: byte, err: io.Error) {
-		b := (^Reader)(s.stream_data)
-		return reader_read_byte(b)
-	},
-	impl_unread_byte = proc(s: io.Stream) -> io.Error {
-		b := (^Reader)(s.stream_data)
-		return reader_unread_byte(b)
-	},
-	impl_read_rune = proc(s: io.Stream) -> (r: rune, size: int, err: io.Error) {
-		b := (^Reader)(s.stream_data)
-		return reader_read_rune(b)
-	},
-	impl_unread_rune = proc(s: io.Stream) -> io.Error {
-		b := (^Reader)(s.stream_data)
-		return reader_unread_rune(b)
-	},
-	impl_write_to = proc(s: io.Stream, w: io.Writer) -> (n: i64, err: io.Error) {
-		b := (^Reader)(s.stream_data)
-		return reader_write_to(b, w)
-	},
+		return
+	case .Query:
+		return io.query_utility({.Read, .Destroy, .Query})
+	}
+	return 0, .Empty
 }

-
-
 //
 // Utility procedures
 //
@@ -435,9 +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 {
@@ -8,6 +8,7 @@ import "core:intrinsics"

 // Extra errors returns by scanning procedures
 Scanner_Extra_Error :: enum i32 {
+	None,
 	Negative_Advance,
 	Advanced_Too_Far,
 	Bad_Read_Count,
@@ -15,7 +16,7 @@ Scanner_Extra_Error :: enum i32 {
 	Too_Short,
 }

-Scanner_Error :: union {
+Scanner_Error :: union #shared_nil {
 	io.Error,
 	Scanner_Extra_Error,
 }
@@ -65,10 +66,10 @@ scanner_destroy :: proc(s: ^Scanner) {
 }


-// Returns the first non-EOF error that was encounted by the scanner
+// Returns the first non-EOF error that was encountered by the scanner
 scanner_error :: proc(s: ^Scanner) -> Scanner_Error {
 	switch s._err {
-	case .EOF, .None:
+	case .EOF, nil:
 		return nil
 	}
 	return s._err
@@ -93,10 +94,6 @@ scanner_text :: proc(s: ^Scanner) -> string {
 // scanner_scan advances the scanner
 scanner_scan :: proc(s: ^Scanner) -> bool {
 	set_err :: proc(s: ^Scanner, err: Scanner_Error) {
-		err := err
-		if err == .None {
-			err = nil
-		}
 		switch s._err {
 		case nil, .EOF:
 			s._err = err
@@ -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) {
@@ -173,14 +173,6 @@ writer_read_from :: proc(b: ^Writer, r: io.Reader) -> (n: i64, err: io.Error) {
 	if b.err != nil {
 		return 0, b.err
 	}
-	if writer_buffered(b) == 0 {
-		if w, ok := io.to_reader_from(b.wr); !ok {
-			n, err = io.read_from(w, r)
-			b.err = err
-			return
-		}
-	}
-
 	for {
 		if writer_available(b) == 0 {
 			writer_flush(b) or_return
@@ -222,38 +214,35 @@ writer_read_from :: proc(b: ^Writer, r: io.Reader) -> (n: i64, err: io.Error) {

 // writer_to_stream converts a Writer into an io.Stream
 writer_to_stream :: proc(b: ^Writer) -> (s: io.Stream) {
-	s.stream_data = b
-	s.stream_vtable = _writer_vtable
+	s.data = b
+	s.procedure = _writer_proc
 	return
 }

+// writer_to_stream converts a Writer into an io.Stream
+writer_to_writer :: proc(b: ^Writer) -> (s: io.Writer) {
+	return writer_to_stream(b)
+}
+


@(private)
-_writer_vtable := &io.Stream_VTable{
-	impl_destroy = proc(s: io.Stream) -> io.Error {
-		b := (^Writer)(s.stream_data)
+_writer_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+	b := (^Writer)(stream_data)
+	#partial switch mode {
+	case .Flush:
+		err = writer_flush(b)
+		return
+	case .Write:
+		n_int: int
+		n_int, err = writer_write(b, p)
+		n = i64(n_int)
+		return
+	case .Destroy:
 		writer_destroy(b)
-		return nil
-	},
-	impl_flush = proc(s: io.Stream)  -> io.Error {
-		b := (^Writer)(s.stream_data)
-		return writer_flush(b)
-	},
-	impl_write = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Writer)(s.stream_data)
-		return writer_write(b, p)
-	},
-	impl_write_byte = proc(s: io.Stream, c: byte) -> io.Error {
-		b := (^Writer)(s.stream_data)
-		return writer_write_byte(b, c)
-	},
-	impl_write_rune = proc(s: io.Stream, r: rune) -> (int, io.Error) {
-		b := (^Writer)(s.stream_data)
-		return writer_write_rune(b, r)
-	},
-	impl_read_from = proc(s: io.Stream, r: io.Reader) -> (n: i64, err: io.Error) {
-		b := (^Writer)(s.stream_data)
-		return writer_read_from(b, r)
-	},
+		return
+	case .Query:
+		return io.query_utility({.Flush, .Write, .Destroy, .Query})
+	}
+	return 0, .Empty
 }
@@ -1,90 +1,90 @@
 // This is purely for documentation
 package builtin

-nil   :: nil;
-false :: 0!==0;
-true  :: 0==0;
+nil   :: nil
+false :: 0!=0
+true  :: 0==0

-ODIN_OS      :: ODIN_OS;
-ODIN_ARCH    :: ODIN_ARCH;
-ODIN_ENDIAN  :: ODIN_ENDIAN;
-ODIN_VENDOR  :: ODIN_VENDOR;
-ODIN_VERSION :: ODIN_VERSION;
-ODIN_ROOT    :: ODIN_ROOT;
-ODIN_DEBUG   :: ODIN_DEBUG;
+ODIN_OS      :: ODIN_OS
+ODIN_ARCH    :: ODIN_ARCH
+ODIN_ENDIAN  :: ODIN_ENDIAN
+ODIN_VENDOR  :: ODIN_VENDOR
+ODIN_VERSION :: ODIN_VERSION
+ODIN_ROOT    :: ODIN_ROOT
+ODIN_DEBUG   :: ODIN_DEBUG

-byte :: u8; // alias
+byte :: u8 // alias

-bool          :: bool;
-b8            :: b8;
-b16           :: b16;
-b32           :: b32;
-b64           :: b64;
+bool          :: bool
+b8            :: b8
+b16           :: b16
+b32           :: b32
+b64           :: b64

-i8            :: i8;
-u8            :: u8;
-i16           :: i16;
-u16           :: u16;
-i32           :: i32;
-u32           :: u32;
-i64           :: i64;
-u64           :: u64;
+i8            :: i8
+u8            :: u8
+i16           :: i16
+u16           :: u16
+i32           :: i32
+u32           :: u32
+i64           :: i64
+u64           :: u64

-i128          :: i128;
-u128          :: u128;
+i128          :: i128
+u128          :: u128

-rune          :: rune;
+rune          :: rune

-f16           :: f16;
-f32           :: f32;
-f64           :: f64;
+f16           :: f16
+f32           :: f32
+f64           :: f64

-complex32     :: complex32;
-complex64     :: complex64;
-complex128    :: complex128;
+complex32     :: complex32
+complex64     :: complex64
+complex128    :: complex128

-quaternion64  :: quaternion64;
-quaternion128 :: quaternion128;
-quaternion256 :: quaternion256;
+quaternion64  :: quaternion64
+quaternion128 :: quaternion128
+quaternion256 :: quaternion256

-int           :: int;
-uint          :: uint;
-uintptr       :: uintptr;
+int           :: int
+uint          :: uint
+uintptr       :: uintptr

-rawptr        :: rawptr;
-string        :: string;
-cstring       :: cstring;
-any           :: any;
+rawptr        :: rawptr
+string        :: string
+cstring       :: cstring
+any           :: any

-typeid        :: typeid;
+typeid        :: typeid

 // Endian Specific Types
-i16le         :: i16le;
-u16le         :: u16le;
-i32le         :: i32le;
-u32le         :: u32le;
-i64le         :: i64le;
-u64le         :: u64le;
-i128le        :: i128le;
-u128le        :: u128le;
+i16le         :: i16le
+u16le         :: u16le
+i32le         :: i32le
+u32le         :: u32le
+i64le         :: i64le
+u64le         :: u64le
+i128le        :: i128le
+u128le        :: u128le

-i16be         :: i16be;
-u16be         :: u16be;
-i32be         :: i32be;
-u32be         :: u32be;
-i64be         :: i64be;
-u64be         :: u64be;
-i128be        :: i128be;
-u128be        :: u128be;
+i16be         :: i16be
+u16be         :: u16be
+i32be         :: i32be
+u32be         :: u32be
+i64be         :: i64be
+u64be         :: u64be
+i128be        :: i128be
+u128be        :: u128be


-f16le         :: f16le;
-f32le         :: f32le;
-f64le         :: f64le;
+f16le         :: f16le
+f32le         :: f32le
+f64le         :: f64le

-f16be         :: f16be;
-f32be         :: f32be;
-f64be         :: f64be;
+f16be         :: f16be
+f32be         :: f32be
+f64be         :: f64be



@@ -94,7 +94,15 @@ cap :: proc(array: Array_Type) -> int ---

 size_of      :: proc($T: typeid) -> int ---
 align_of     :: proc($T: typeid) -> int ---
-offset_of    :: proc($T: typeid) -> uintptr ---
+
+// e.g. offset_of(t.f), where t is an instance of the type T
+offset_of_selector :: proc(selector: $T) -> uintptr ---
+// e.g. offset_of(T, f), where T can be the type instead of a variable
+offset_of_member   :: proc($T: typeid, member: $M) -> uintptr ---
+offset_of :: proc{offset_of_selector, offset_of_member}
+// e.g. offset_of(T, "f"), where T can be the type instead of a variable
+offset_of_by_string :: proc($T: typeid, member: string) -> uintptr ---
+
 type_of      :: proc(x: expr) -> type ---
 type_info_of :: proc($T: typeid) -> ^runtime.Type_Info ---
 typeid_of    :: proc($T: typeid) -> typeid ---
@@ -109,7 +117,7 @@ jmag       :: proc(value: Quaternion) -> Float ---
 kmag       :: proc(value: Quaternion) -> Float ---
 conj       :: proc(value: Complex_Or_Quaternion) -> Complex_Or_Quaternion ---

-expand_to_tuple :: proc(value: Struct_Or_Array) -> (A, B, C, ...) ---
+expand_values :: proc(value: Struct_Or_Array) -> (A, B, C, ...) ---

 min   :: proc(values: ..T) -> T ---
 max   :: proc(values: ..T) -> T ---
@@ -38,6 +38,11 @@ buffer_init_string :: proc(b: ^Buffer, s: string) {
 }

 buffer_init_allocator :: proc(b: ^Buffer, len, cap: int, allocator := context.allocator) {
+	if b.buf == nil {
+		b.buf = make([dynamic]byte, len, cap, allocator)
+		return
+	}
+
 	b.buf.allocator = allocator
 	reserve(&b.buf, cap)
 	resize(&b.buf, len)
@@ -108,8 +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
 	}

@@ -161,6 +169,10 @@ buffer_write :: proc(b: ^Buffer, p: []byte) -> (n: int, err: io.Error) {
 	return copy(b.buf[m:], p), nil
 }

+buffer_write_ptr :: proc(b: ^Buffer, ptr: rawptr, size: int) -> (n: int, err: io.Error) {
+	return buffer_write(b, ([^]byte)(ptr)[:size])
+}
+
 buffer_write_string :: proc(b: ^Buffer, s: string) -> (n: int, err: io.Error) {
 	b.last_read = .Invalid
 	m, ok := _buffer_try_grow(b, len(s))
@@ -229,17 +241,18 @@ buffer_read :: proc(b: ^Buffer, p: []byte) -> (n: int, err: io.Error) {
 	return
 }

+buffer_read_ptr :: proc(b: ^Buffer, ptr: rawptr, size: int) -> (n: int, err: io.Error) {
+	return buffer_read(b, ([^]byte)(ptr)[:size])
+}
+
 buffer_read_at :: proc(b: ^Buffer, p: []byte, offset: int) -> (n: int, err: io.Error) {
 	b.last_read = .Invalid

-	if offset < 0 || offset >= len(b.buf) {
+	if uint(offset) >= len(b.buf) {
 		err = .Invalid_Offset
 		return
 	}
-
-	if 0 <= offset && offset < len(b.buf) {
-		n = copy(p, b.buf[offset:])
-	}
+	n = copy(p, b.buf[offset:])
 	if n > 0 {
 		b.last_read = .Read
 	}
@@ -365,69 +378,31 @@ buffer_read_from :: proc(b: ^Buffer, r: io.Reader) -> (n: i64, err: io.Error) #n


 buffer_to_stream :: proc(b: ^Buffer) -> (s: io.Stream) {
-	s.stream_data = b
-	s.stream_vtable = _buffer_vtable
+	s.data = b
+	s.procedure = _buffer_proc
 	return
 }

@(private)
-_buffer_vtable := &io.Stream_VTable{
-	impl_size = proc(s: io.Stream) -> i64 {
-		b := (^Buffer)(s.stream_data)
-		return i64(buffer_capacity(b))
-	},
-	impl_read = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_read(b, p)
-	},
-	impl_read_at = proc(s: io.Stream, p: []byte, offset: i64) -> (n: int, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_read_at(b, p, int(offset))
-	},
-	impl_read_byte = proc(s: io.Stream) -> (byte, io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_read_byte(b)
-	},
-	impl_read_rune = proc(s: io.Stream) -> (r: rune, size: int, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_read_rune(b)
-	},
-	impl_write = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_write(b, p)
-	},
-	impl_write_at = proc(s: io.Stream, p: []byte, offset: i64) -> (n: int, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_write_at(b, p, int(offset))
-	},
-	impl_write_byte = proc(s: io.Stream, c: byte) -> io.Error {
-		b := (^Buffer)(s.stream_data)
-		return buffer_write_byte(b, c)
-	},
-	impl_write_rune = proc(s: io.Stream, r: rune) -> (int, io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_write_rune(b, r)
-	},
-	impl_unread_byte = proc(s: io.Stream) -> io.Error {
-		b := (^Buffer)(s.stream_data)
-		return buffer_unread_byte(b)
-	},
-	impl_unread_rune = proc(s: io.Stream) -> io.Error {
-		b := (^Buffer)(s.stream_data)
-		return buffer_unread_rune(b)
-	},
-	impl_destroy = proc(s: io.Stream) -> io.Error {
-		b := (^Buffer)(s.stream_data)
+_buffer_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+	b := (^Buffer)(stream_data)
+	#partial switch mode {
+	case .Read:
+		return io._i64_err(buffer_read(b, p))
+	case .Read_At:
+		return io._i64_err(buffer_read_at(b, p, int(offset)))
+	case .Write:
+		return io._i64_err(buffer_write(b, p))
+	case .Write_At:
+		return io._i64_err(buffer_write_at(b, p, int(offset)))
+	case .Size:
+		n = i64(buffer_capacity(b))
+		return
+	case .Destroy:
 		buffer_destroy(b)
-		return nil
-	},
-	impl_write_to = proc(s: io.Stream, w: io.Writer) -> (n: i64, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_write_to(b, w)
-	},
-	impl_read_from = proc(s: io.Stream, r: io.Reader) -> (n: i64, err: io.Error) {
-		b := (^Buffer)(s.stream_data)
-		return buffer_read_from(b, r)
-	},
+		return
+	case .Query:
+		return io.query_utility({.Read, .Read_At, .Write, .Write_At, .Size, .Destroy})
+	}
+	return 0, .Empty
 }
-
@@ -5,13 +5,19 @@ import "core:unicode"
 import "core:unicode/utf8"

 clone :: proc(s: []byte, allocator := context.allocator, loc := #caller_location) -> []byte {
-	c := make([]byte, len(s)+1, allocator, loc)
+	c := make([]byte, len(s), allocator, loc)
 	copy(c, s)
-	c[len(s)] = 0
 	return c[:len(s)]
 }

-ptr_from_slice :: proc(str: []byte) -> ^byte {
+clone_safe :: proc(s: []byte, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: mem.Allocator_Error) {
+	c := make([]byte, len(s), allocator, loc) or_return
+	copy(c, s)
+	return c[:len(s)], nil
+}
+
+ptr_from_slice :: ptr_from_bytes
+ptr_from_bytes :: proc(str: []byte) -> ^byte {
 	d := transmute(mem.Raw_String)str
 	return d.data
 }
@@ -135,6 +141,25 @@ join :: proc(a: [][]byte, sep: []byte, allocator := context.allocator) -> []byte
 	return b
 }

+join_safe :: proc(a: [][]byte, sep: []byte, allocator := context.allocator) -> (data: []byte, err: mem.Allocator_Error) {
+	if len(a) == 0 {
+		return nil, nil
+	}
+
+	n := len(sep) * (len(a) - 1)
+	for s in a {
+		n += len(s)
+	}
+
+	b := make([]byte, n, allocator) or_return
+	i := copy(b, a[0])
+	for s in a[1:] {
+		i += copy(b[i:], sep)
+		i += copy(b[i:], s)
+	}
+	return b, nil
+}
+
 concatenate :: proc(a: [][]byte, allocator := context.allocator) -> []byte {
 	if len(a) == 0 {
 		return nil
@@ -152,6 +177,24 @@ concatenate :: proc(a: [][]byte, allocator := context.allocator) -> []byte {
 	return b
 }

+concatenate_safe :: proc(a: [][]byte, allocator := context.allocator) -> (data: []byte, err: mem.Allocator_Error) {
+	if len(a) == 0 {
+		return nil, nil
+	}
+
+	n := 0
+	for s in a {
+		n += len(s)
+	}
+	b := make([]byte, n, allocator) or_return
+	i := 0
+	for s in a {
+		i += copy(b[i:], s)
+	}
+	return b, nil
+}
+
+
@private
 _split :: proc(s, sep: []byte, sep_save, n: int, allocator := context.allocator) -> [][]byte {
 	s, n := s, n
@@ -219,61 +262,37 @@ split_after_n :: proc(s, sep: []byte, n: int, allocator := context.allocator) ->


@private
-_split_iterator :: proc(s: ^[]byte, sep: []byte, sep_save, n: int) -> (res: []byte, ok: bool) {
-	s, n := s, n
-
-	if n == 0 {
-		return
-	}
-
-	if sep == nil {
+_split_iterator :: proc(s: ^[]byte, sep: []byte, sep_save: int) -> (res: []byte, ok: bool) {
+	if len(sep) == 0 {
 		res = s[:]
 		ok = true
 		s^ = s[len(s):]
 		return
 	}

-	if n < 0 {
-		n = count(s^, sep) + 1
-	}
-
-	n -= 1
-
-	i := 0
-	for ; i < n; i += 1 {
-		m := index(s^, sep)
-		if m < 0 {
-			break
-		}
+	m := index(s^, sep)
+	if m < 0 {
+		// not found
+		res = s[:]
+		ok = len(res) != 0
+		s^ = s[len(s):]
+	} else {
 		res = s[:m+sep_save]
 		ok = true
 		s^ = s[m+len(sep):]
-		return
 	}
-	res = s[:]
-	ok = res != nil
-	s^ = s[len(s):]
 	return
 }


 split_iterator :: proc(s: ^[]byte, sep: []byte) -> ([]byte, bool) {
-	return _split_iterator(s, sep, 0, -1)
-}
-
-split_n_iterator :: proc(s: ^[]byte, sep: []byte, n: int) -> ([]byte, bool) {
-	return _split_iterator(s, sep, 0, n)
+	return _split_iterator(s, sep, 0)
 }

 split_after_iterator :: proc(s: ^[]byte, sep: []byte) -> ([]byte, bool) {
-	return _split_iterator(s, sep, len(sep), -1)
+	return _split_iterator(s, sep, len(sep))
 }

-split_after_n_iterator :: proc(s: ^[]byte, sep: []byte, n: int) -> ([]byte, bool) {
-	return _split_iterator(s, sep, len(sep), n)
-}
-
-

 index_byte :: proc(s: []byte, c: byte) -> int {
 	for i := 0; i < len(s); i += 1 {
@@ -619,7 +638,7 @@ trim_left_proc :: proc(s: []byte, p: proc(rune) -> bool) -> []byte {

 index_rune :: proc(s: []byte, r: rune) -> int {
 	switch {
-	case 0 <= r && r < utf8.RUNE_SELF:
+	case u32(r) < utf8.RUNE_SELF:
 		return index_byte(s, byte(r))

 	case r == utf8.RUNE_ERROR:
@@ -1143,7 +1162,7 @@ fields_proc :: proc(s: []byte, f: proc(rune) -> bool, allocator := context.alloc
 	}

 	if start >= 0 {
-		append(&subslices, s[start : end])
+		append(&subslices, s[start : len(s)])
 	}

 	return subslices[:]
@@ -16,8 +16,8 @@ reader_init :: proc(r: ^Reader, s: []byte) {
 }

 reader_to_stream :: proc(r: ^Reader) -> (s: io.Stream) {
-	s.stream_data = r
-	s.stream_vtable = _reader_vtable
+	s.data = r
+	s.procedure = _reader_proc
 	return
 }

@@ -137,41 +137,22 @@ reader_write_to :: proc(r: ^Reader, w: io.Writer) -> (n: i64, err: io.Error) {


@(private)
-_reader_vtable := &io.Stream_VTable{
-	impl_size = proc(s: io.Stream) -> i64 {
-		r := (^Reader)(s.stream_data)
-		return reader_size(r)
-	},
-	impl_read = proc(s: io.Stream, p: []byte) -> (n: int, err: io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_read(r, p)
-	},
-	impl_read_at = proc(s: io.Stream, p: []byte, off: i64) -> (n: int, err: io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_read_at(r, p, off)
-	},
-	impl_read_byte = proc(s: io.Stream) -> (byte, io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_read_byte(r)
-	},
-	impl_unread_byte = proc(s: io.Stream) -> io.Error {
-		r := (^Reader)(s.stream_data)
-		return reader_unread_byte(r)
-	},
-	impl_read_rune = proc(s: io.Stream) -> (ch: rune, size: int, err: io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_read_rune(r)
-	},
-	impl_unread_rune = proc(s: io.Stream) -> io.Error {
-		r := (^Reader)(s.stream_data)
-		return reader_unread_rune(r)
-	},
-	impl_seek = proc(s: io.Stream, offset: i64, whence: io.Seek_From) -> (i64, io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_seek(r, offset, whence)
-	},
-	impl_write_to = proc(s: io.Stream, w: io.Writer) -> (n: i64, err: io.Error) {
-		r := (^Reader)(s.stream_data)
-		return reader_write_to(r, w)
-	},
+_reader_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+	r := (^Reader)(stream_data)
+	#partial switch mode {
+	case .Read:
+		return io._i64_err(reader_read(r, p))
+	case .Read_At:
+		return io._i64_err(reader_read_at(r, p, offset))
+	case .Seek:
+		n, err = reader_seek(r, offset, whence)
+		return
+	case .Size:
+		n = reader_size(r)
+		return
+	case .Query:
+		return io.query_utility({.Read, .Read_At, .Seek, .Size, .Query})
+	}
+	return 0, .Empty
 }
+
@@ -3,22 +3,24 @@ package c
 import builtin "core:builtin"

 char           :: builtin.u8  // assuming -funsigned-char
+
+schar          :: builtin.i8
 short          :: builtin.i16
 int            :: builtin.i32
-long           :: builtin.i32 when (ODIN_OS == "windows" || size_of(builtin.rawptr) == 4) else builtin.i64
+long           :: builtin.i32 when (ODIN_OS == .Windows || size_of(builtin.rawptr) == 4) else builtin.i64
 longlong       :: builtin.i64

 uchar          :: builtin.u8
 ushort         :: builtin.u16
 uint           :: builtin.u32
-ulong          :: builtin.u32 when (ODIN_OS == "windows" || size_of(builtin.rawptr) == 4) else builtin.u64
+ulong          :: builtin.u32 when (ODIN_OS == .Windows || size_of(builtin.rawptr) == 4) else builtin.u64
 ulonglong      :: builtin.u64

 bool           :: builtin.bool

 size_t         :: builtin.uint
 ssize_t        :: builtin.int
-wchar_t        :: builtin.u16 when (ODIN_OS == "windows") else builtin.u32
+wchar_t        :: builtin.u16 when (ODIN_OS == .Windows) else builtin.u32

 float          :: builtin.f32
 double         :: builtin.f64
@@ -46,7 +48,7 @@ int_least64_t  :: builtin.i64
 uint_least64_t :: builtin.u64

 // Same on Windows, Linux, and FreeBSD
-when ODIN_ARCH == "386" || ODIN_ARCH == "amd64" {
+when ODIN_ARCH == .i386 || ODIN_ARCH == .amd64 {
 	int_fast8_t    :: builtin.i8
 	uint_fast8_t   :: builtin.u8
 	int_fast16_t   :: builtin.i32
@@ -519,7 +519,7 @@ join_adjacent_string_literals :: proc(cpp: ^Preprocessor, initial_tok: ^Token) {


 quote_string :: proc(s: string) -> []byte {
-	b := strings.make_builder(0, len(s)+2)
+	b := strings.builder_make(0, len(s)+2)
 	io.write_quoted_string(strings.to_writer(&b), s, '"')
 	return b.buf[:]
 }
@@ -956,7 +956,7 @@ substitute_token :: proc(cpp: ^Preprocessor, tok: ^Token, args: ^Macro_Arg) -> ^
 			continue
 		}

-		if tok.lit == "__VA__OPT__" && tok.next.lit == "(" {
+		if tok.lit == "__VA_OPT__" && tok.next.lit == "(" {
 			opt_arg := read_macro_arg_one(cpp, &tok, tok.next.next, true)
 			if has_varargs(args) {
 				for t := opt_arg.tok; t.kind != .EOF; t = t.next {
@@ -1118,7 +1118,7 @@ expand_macro :: proc(cpp: ^Preprocessor, rest: ^^Token, tok: ^Token) -> bool {

 search_include_next :: proc(cpp: ^Preprocessor, filename: string) -> (path: string, ok: bool) {
 	for ; cpp.include_next_index < len(cpp.include_paths); cpp.include_next_index += 1 {
-		tpath := filepath.join(elems={cpp.include_paths[cpp.include_next_index], filename}, allocator=context.temp_allocator)
+		tpath := filepath.join({cpp.include_paths[cpp.include_next_index], filename}, allocator=context.temp_allocator)
 		if os.exists(tpath) {
 			return strings.clone(tpath), true
 		}
@@ -1136,7 +1136,7 @@ search_include_paths :: proc(cpp: ^Preprocessor, filename: string) -> (path: str
 	}

 	for include_path in cpp.include_paths {
-		tpath := filepath.join(elems={include_path, filename}, allocator=context.temp_allocator)
+		tpath := filepath.join({include_path, filename}, allocator=context.temp_allocator)
 		if os.exists(tpath) {
 			path, ok = strings.clone(tpath), true
 			cpp.filepath_cache[filename] = path
@@ -1276,7 +1276,7 @@ preprocess_internal :: proc(cpp: ^Preprocessor, tok: ^Token) -> ^Token {
 				if start.file != nil {
 					dir = filepath.dir(start.file.name)
 				}
-				path := filepath.join(dir, filename)
+				path := filepath.join({dir, filename})
 				if os.exists(path) {
 					tok = include_file(cpp, tok, path, start.next.next)
 					continue
@@ -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)
 	}
 }
@@ -2,8 +2,10 @@ package libc

 // 7.3 Complex arithmetic

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -47,8 +49,8 @@ foreign libc {
 	// 7.3.8 Power and absolute-value functions
 	cabs    :: proc(z: complex_double) -> complex_double ---
 	cabsf   :: proc(z: complex_float) -> complex_float ---
-	cpow    :: proc(z: complex_double) -> complex_double ---
-	cpowf   :: proc(z: complex_float) -> complex_float ---
+	cpow    :: proc(x, y: complex_double) -> complex_double ---
+	cpowf   :: proc(x, y: complex_float) -> complex_float ---
 	csqrt   :: proc(z: complex_double) -> complex_double ---
 	csqrtf  :: proc(z: complex_float) -> complex_float ---

@@ -1,7 +1,9 @@
 package libc

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -2,8 +2,10 @@ package libc

 // 7.5 Errors

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -12,11 +14,11 @@ when ODIN_OS == "windows" {
 //	EDOM,
 //	EILSEQ
 //	ERANGE
-when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
+when ODIN_OS == .Linux {
 	@(private="file")
 	@(default_calling_convention="c")
 	foreign libc {
-		@(link_name="__libc_errno_location")
+		@(link_name="__errno_location")
 		_get_errno :: proc() -> ^int ---
 	}

@@ -25,7 +27,33 @@ when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
 	ERANGE :: 34
 }

-when ODIN_OS == "windows" {
+when ODIN_OS == .FreeBSD {
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		@(link_name="__error")
+		_get_errno :: proc() -> ^int ---
+	}
+
+	EDOM   :: 33
+	EILSEQ :: 84
+	ERANGE :: 34
+}
+
+when ODIN_OS == .OpenBSD {
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		@(link_name="__errno")
+		_get_errno :: proc() -> ^int ---
+	}
+
+	EDOM   :: 33
+	EILSEQ :: 84
+	ERANGE :: 34
+}
+
+when ODIN_OS == .Windows {
 	@(private="file")
 	@(default_calling_convention="c")
 	foreign libc {
@@ -38,6 +66,20 @@ when ODIN_OS == "windows" {
 	ERANGE :: 34
 }

+when ODIN_OS == .Darwin {
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		@(link_name="__error")
+		_get_errno :: proc() -> ^int ---
+	}
+
+	// Unknown
+	EDOM   :: 33
+	EILSEQ :: 92
+	ERANGE :: 34
+}
+
 // Odin has no way to make an identifier "errno" behave as a function call to
 // read the value, or to produce an lvalue such that you can assign a different
 // error value to errno. To work around this, just expose it as a function like
@@ -4,8 +4,10 @@ package libc

 import "core:intrinsics"

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -209,19 +211,19 @@ _signbitf :: #force_inline proc(x: float) -> int {
 	return int(transmute(uint32_t)x >> 31)
 }

-isfinite :: #force_inline proc(x: $T) where intrinsics.type_is_float(T) {
+isfinite :: #force_inline proc(x: $T) -> bool where intrinsics.type_is_float(T) {
 	return fpclassify(x) == FP_INFINITE
 }

-isinf :: #force_inline proc(x: $T) where intrinsics.type_is_float(T) {
+isinf :: #force_inline proc(x: $T) -> bool where intrinsics.type_is_float(T) {
 	return fpclassify(x) > FP_INFINITE
 }

-isnan :: #force_inline proc(x: $T) where intrinsics.type_is_float(T) {
+isnan :: #force_inline proc(x: $T) -> bool where intrinsics.type_is_float(T) {
 	return fpclassify(x) == FP_NAN
 }

-isnormal :: #force_inline proc(x: $T) where intrinsics.type_is_float(T) {
+isnormal :: #force_inline proc(x: $T) -> bool where intrinsics.type_is_float(T) {
 	return fpclassify(x) == FP_NORMAL
 }

@@ -229,27 +231,27 @@ isnormal :: #force_inline proc(x: $T) where intrinsics.type_is_float(T) {
 // implemented as the relational comparisons, as that would produce an invalid
 // "sticky" state that propagates and affects maths results. These need
 // to be implemented natively in Odin assuming isunordered to prevent that.
-isgreater :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+isgreater :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	return !isunordered(x, y) && x > y
 }

-isgreaterequal :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+isgreaterequal :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	return !isunordered(x, y) && x >= y
 }

-isless :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+isless :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	return !isunordered(x, y) && x < y
 }

-islessequal :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+islessequal :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	return !isunordered(x, y) && x <= y
 }

-islessgreater :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+islessgreater :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	return !isunordered(x, y) && x <= y
 }

-isunordered :: #force_inline proc(x, y: $T) where intrinsics.type_is_float(T) {
+isunordered :: #force_inline proc(x, y: $T) -> bool where intrinsics.type_is_float(T) {
 	if isnan(x) {
 		// Force evaluation of y to propagate exceptions for ordering semantics.
 		// To ensure correct semantics of IEEE 754 this cannot be compiled away.
@@ -329,7 +331,7 @@ fmin       :: proc{libc_fmin, libc_fminf}
 fma        :: proc{libc_fma, libc_fmaf}

 // But retain the 'f' suffix-variant functions as well so they can be used,
-// a trick is used here where we use explicit procedrual overloading of one
+// a trick is used here where we use explicit procedural overloading of one
 // procedure. This is done because the foreign block is marked @(private) and
 // aliasing functions does not remove privateness from the entity.
 acosf      :: proc{libc_acosf}
@@ -2,13 +2,14 @@ package libc

 // 7.13 Nonlocal jumps

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
-
-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	@(default_calling_convention="c")
 	foreign libc {
 		// 7.13.1 Save calling environment
@@ -62,4 +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, }
@@ -2,8 +2,10 @@ package libc

 // 7.14 Signal handling

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -19,7 +21,7 @@ foreign libc {
 	raise  :: proc(sig: int) -> int ---
 }

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	SIG_ERR :: rawptr(~uintptr(0)) 
 	SIG_DFL :: rawptr(uintptr(0))
 	SIG_IGN :: rawptr(uintptr(1))
@@ -32,7 +34,20 @@ when ODIN_OS == "windows" {
 	SIGTERM :: 15
 }

-when ODIN_OS == "linux" || ODIN_OS == "freebsd" || ODIN_OS == "darwin" {
+when ODIN_OS == .Linux || ODIN_OS == .FreeBSD {
+	SIG_ERR  :: rawptr(~uintptr(0))
+	SIG_DFL  :: rawptr(uintptr(0))
+	SIG_IGN  :: rawptr(uintptr(1)) 
+
+	SIGABRT  :: 6
+	SIGFPE   :: 8
+	SIGILL   :: 4
+	SIGINT   :: 2
+	SIGSEGV  :: 11
+	SIGTERM  :: 15
+}
+
+when ODIN_OS == .Darwin {
 	SIG_ERR  :: rawptr(~uintptr(0))
 	SIG_DFL  :: rawptr(uintptr(0))
 	SIG_IGN  :: rawptr(uintptr(1)) 
@@ -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,
 }

@@ -47,29 +47,30 @@ kill_dependency :: #force_inline proc(value: $T) -> T {

 // 7.17.4 Fences
 atomic_thread_fence :: #force_inline proc(order: memory_order) {
-	switch (order) {
-	case .relaxed:
-		return
-	case .consume:
-		intrinsics.atomic_fence_acq()
-	case .acquire:
-		intrinsics.atomic_fence_acq()
-	case .release:
-		intrinsics.atomic_fence_rel()
-	case .acq_rel:
-		intrinsics.atomic_fence_acqrel()
-	case .seq_cst:
-		intrinsics.atomic_fence_acqrel()
+	assert(order != .relaxed)
+	assert(order != .consume)
+	#partial switch order {
+	case .acquire: intrinsics.atomic_thread_fence(.Acquire)
+	case .release: intrinsics.atomic_thread_fence(.Release)
+	case .acq_rel: intrinsics.atomic_thread_fence(.Acq_Rel)
+	case .seq_cst: intrinsics.atomic_thread_fence(.Seq_Cst)
 	}
 }

 atomic_signal_fence :: #force_inline proc(order: memory_order) {
-	atomic_thread_fence(order)
+	assert(order != .relaxed)
+	assert(order != .consume)
+	#partial switch order {
+	case .acquire: intrinsics.atomic_signal_fence(.Acquire)
+	case .release: intrinsics.atomic_signal_fence(.Release)
+	case .acq_rel: intrinsics.atomic_signal_fence(.Acq_Rel)
+	case .seq_cst: intrinsics.atomic_signal_fence(.Seq_Cst)
+	}
 }

 // 7.17.5 Lock-free property
 atomic_is_lock_free :: #force_inline proc(obj: ^$T) -> bool {
-	return size_of(T) <= 8 && (intrinsics.type_is_integer(T) || intrinsics.type_is_pointer(T))
+	return intrinsics.atomic_type_is_lock_free(T)
 }

 // 7.17.6 Atomic integer types
@@ -121,13 +122,10 @@ atomic_store_explicit :: #force_inline proc(object: ^$T, desired: T, order: memo
 	assert(order != .acquire)
 	assert(order != .acq_rel)

-	#partial switch (order) {
-	case .relaxed:
-		intrinsics.atomic_store_relaxed(object, desired)
-	case .release:
-		intrinsics.atomic_store_rel(object, desired)
-	case .seq_cst:
-		intrinsics.atomic_store(object, desired)
+	#partial switch order {
+	case .relaxed: intrinsics.atomic_store_explicit(object, desired, .Relaxed)
+	case .release: intrinsics.atomic_store_explicit(object, desired, .Release)
+	case .seq_cst: intrinsics.atomic_store_explicit(object, desired, .Seq_Cst)
 	}
 }

@@ -139,36 +137,26 @@ atomic_load_explicit :: #force_inline proc(object: ^$T, order: memory_order) {
 	assert(order != .release)
 	assert(order != .acq_rel)

-	#partial switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_load_relaxed(object)
-	case .consume:
-		return intrinsics.atomic_load_acq(object)
-	case .acquire:
-		return intrinsics.atomic_load_acq(object)
-	case .seq_cst:
-		return intrinsics.atomic_load(object)
+	#partial switch order {
+	case .relaxed: return intrinsics.atomic_load_explicit(object, .Relaxed)
+	case .consume: return intrinsics.atomic_load_explicit(object, .Consume)
+	case .acquire: return intrinsics.atomic_load_explicit(object, .Acquire)
+	case .seq_cst: return intrinsics.atomic_load_explicit(object, .Seq_Cst)
 	}
 }

 atomic_exchange :: #force_inline proc(object: ^$T, desired: T) -> T {
-	return intrinsics.atomic_xchg(object, desired)
+	return intrinsics.atomic_exchange(object, desired)
 }

 atomic_exchange_explicit :: #force_inline proc(object: ^$T, desired: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_xchg_relaxed(object, desired)
-	case .consume:
-		return intrinsics.atomic_xchg_acq(object, desired)
-	case .acquire:
-		return intrinsics.atomic_xchg_acq(object, desired)
-	case .release:
-		return intrinsics.atomic_xchg_rel(object, desired)
-	case .acq_rel:
-		return intrinsics.atomic_xchg_acqrel(object, desired)
-	case .seq_cst:
-		return intrinsics.atomic_xchg(object, desired)
+	switch order {
+	case .relaxed: return intrinsics.atomic_exchange_explicit(object, desired, .Relaxed)
+	case .consume: return intrinsics.atomic_exchange_explicit(object, desired, .Consume)
+	case .acquire: return intrinsics.atomic_exchange_explicit(object, desired, .Acquire)
+	case .release: return intrinsics.atomic_exchange_explicit(object, desired, .Release)
+	case .acq_rel: return intrinsics.atomic_exchange_explicit(object, desired, .Acq_Rel)
+	case .seq_cst: return intrinsics.atomic_exchange_explicit(object, desired, .Seq_Cst)
 	}
 	return false
 }
@@ -189,102 +177,104 @@ atomic_exchange_explicit :: #force_inline proc(object: ^$T, desired: T, order: m
 // 	[success = seq_cst, failure = acquire] => failacq
 // 	[success = acquire, failure = relaxed] => acq_failrelaxed
 // 	[success = acq_rel, failure = relaxed] => acqrel_failrelaxed
-atomic_compare_exchange_strong :: #force_inline proc(object, expected: ^$T, desired: T) {
-	value, ok := intrinsics.atomic_cxchg(object, expected^, desired)
+atomic_compare_exchange_strong :: #force_inline proc(object, expected: ^$T, desired: T) -> bool {
+	value, ok := intrinsics.atomic_compare_exchange_strong(object, expected^, desired)
 	if !ok { expected^ = value } 
 	return ok
 }

-atomic_compare_exchange_strong_explicit :: #force_inline proc(object, expected: ^$T, desired: T, success, failure: memory_order) {
+atomic_compare_exchange_strong_explicit :: #force_inline proc(object, expected: ^$T, desired: T, success, failure: memory_order) -> bool {
 	assert(failure != .release)
 	assert(failure != .acq_rel)

 	value: T; ok: bool
-	#partial switch (failure) {
+	#partial switch failure {
 	case .seq_cst:
 		assert(success != .relaxed)
-		#partial switch (success) {
+		#partial switch success {
 		case .seq_cst:
-			value, ok := intrinsics.atomic_cxchg(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Seq_Cst, .Seq_Cst)
 		case .acquire:
-			value, ok := intrinsics.atomic_cxchg_acq(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Acquire, .Seq_Cst)
 		case .consume:
-			value, ok := intrinsics.atomic_cxchg_acq(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Consume, .Seq_Cst)
 		case .release:
-			value, ok := intrinsics.atomic_cxchg_rel(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Release, .Seq_Cst)
 		case .acq_rel:
-			value, ok := intrinsics.atomic_cxchg_acqrel(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Acq_Rel, .Seq_Cst)
 		}
 	case .relaxed:
 		assert(success != .release)
-		#partial switch (success) {
+		#partial switch success {
 		case .relaxed:
-			value, ok := intrinsics.atomic_cxchg_relaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Relaxed, .Relaxed)
 		case .seq_cst:
-			value, ok := intrinsics.atomic_cxchg_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Seq_Cst, .Relaxed)
 		case .acquire:
-			value, ok := intrinsics.atomic_cxchg_acq_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Acquire, .Relaxed)
 		case .consume:
-			value, ok := intrinsics.atomic_cxchg_acq_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Consume, .Relaxed)
 		case .acq_rel:
-			value, ok := intrinsics.atomic_cxchg_acqrel_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Acq_Rel, .Relaxed)
 		}
 	case .consume:
-		fallthrough
+		assert(success == .seq_cst)
+		value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Seq_Cst, .Consume)
 	case .acquire:
 		assert(success == .seq_cst)
-		value, ok := intrinsics.atomic_cxchg_failacq(object, expected^, desired)
+		value, ok = intrinsics.atomic_compare_exchange_strong_explicit(object, expected^, desired, .Seq_Cst, .Acquire)

 	}
 	if !ok { expected^ = value }
 	return ok
 }

-atomic_compare_exchange_weak :: #force_inline proc(object, expected: ^$T, desired: T) {
-	value, ok := intrinsics.atomic_cxchgweak(object, expected^, desired)
+atomic_compare_exchange_weak :: #force_inline proc(object, expected: ^$T, desired: T) -> bool {
+	value, ok := intrinsics.atomic_compare_exchange_weak(object, expected^, desired)
 	if !ok { expected^ = value }
 	return ok
 }

-atomic_compare_exchange_weak_explicit :: #force_inline proc(object, expected: ^$T, desited: T, success, failure: memory_order) {
+atomic_compare_exchange_weak_explicit :: #force_inline proc(object, expected: ^$T, desited: T, success, failure: memory_order) -> bool {
 	assert(failure != .release)
 	assert(failure != .acq_rel)

 	value: T; ok: bool
-	#partial switch (failure) {
+	#partial switch failure {
 	case .seq_cst:
 		assert(success != .relaxed)
-		#partial switch (success) {
+		#partial switch success {
 		case .seq_cst:
-			value, ok := intrinsics.atomic_cxchgweak(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Seq_Cst, .Seq_Cst)
 		case .acquire:
-			value, ok := intrinsics.atomic_cxchgweak_acq(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Acquire, .Seq_Cst)
 		case .consume:
-			value, ok := intrinsics.atomic_cxchgweak_acq(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Consume, .Seq_Cst)
 		case .release:
-			value, ok := intrinsics.atomic_cxchgweak_rel(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Release, .Seq_Cst)
 		case .acq_rel:
-			value, ok := intrinsics.atomic_cxchgweak_acqrel(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Acq_Rel, .Seq_Cst)
 		}
 	case .relaxed:
 		assert(success != .release)
-		#partial switch (success) {
+		#partial switch success {
 		case .relaxed:
-			value, ok := intrinsics.atomic_cxchgweak_relaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Relaxed, .Relaxed)
 		case .seq_cst:
-			value, ok := intrinsics.atomic_cxchgweak_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Seq_Cst, .Relaxed)
 		case .acquire:
-			value, ok := intrinsics.atomic_cxchgweak_acq_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Acquire, .Relaxed)
 		case .consume:
-			value, ok := intrinsics.atomic_cxchgweak_acq_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Consume, .Relaxed)
 		case .acq_rel:
-			value, ok := intrinsics.atomic_cxchgweak_acqrel_failrelaxed(object, expected^, desired)
+			value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Acq_Rel, .Relaxed)
 		}
 	case .consume:
-		fallthrough
+		assert(success == .seq_cst)
+		value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Seq_Cst, .Consume)
 	case .acquire:
 		assert(success == .seq_cst)
-		value, ok := intrinsics.atomic_cxchgweak_failacq(object, expected^, desired)
+		value, ok = intrinsics.atomic_compare_exchange_weak_explicit(object, expected^, desired, .Seq_Cst, .Acquire)

 	}
 	if !ok { expected^ = value }
@@ -297,19 +287,14 @@ atomic_fetch_add :: #force_inline proc(object: ^$T, operand: T) -> T {
 }

 atomic_fetch_add_explicit :: #force_inline proc(object: ^$T, operand: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_add_relaxed(object, operand)
-	case .consume:
-		return intrinsics.atomic_add_acq(object, operand)
-	case .acquire:
-		return intrinsics.atomic_add_acq(object, operand)
-	case .release:
-		return intrinsics.atomic_add_rel(object, operand)
-	case .acq_rel:
-		return intrinsics.atomic_add_acqrel(object, operand)
-	case .seq_cst:
-		return intrinsics.atomic_add(object, operand)
+	switch order {
+	case .relaxed: return intrinsics.atomic_add_explicit(object, operand, .Relaxed)
+	case .consume: return intrinsics.atomic_add_explicit(object, operand, .Consume)
+	case .acquire: return intrinsics.atomic_add_explicit(object, operand, .Acquire)
+	case .release: return intrinsics.atomic_add_explicit(object, operand, .Release)
+	case .acq_rel: return intrinsics.atomic_add_explicit(object, operand, .Acq_Rel)
+	case: fallthrough
+	case .seq_cst: return intrinsics.atomic_add_explicit(object, operand, .Seq_Cst)
 	}
 }

@@ -318,19 +303,14 @@ atomic_fetch_sub :: #force_inline proc(object: ^$T, operand: T) -> T {
 }

 atomic_fetch_sub_explicit :: #force_inline proc(object: ^$T, operand: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_sub_relaxed(object, operand)
-	case .consume:
-		return intrinsics.atomic_sub_acq(object, operand)
-	case .acquire:
-		return intrinsics.atomic_sub_acq(object, operand)
-	case .release:
-		return intrinsics.atomic_sub_rel(object, operand)
-	case .acq_rel:
-		return intrinsics.atomic_sub_acqrel(object, operand)
-	case .seq_cst:
-		return intrinsics.atomic_sub(object, operand)
+	switch order {
+	case .relaxed: return intrinsics.atomic_sub_explicit(object, operand, .Relaxed)
+	case .consume: return intrinsics.atomic_sub_explicit(object, operand, .Consume)
+	case .acquire: return intrinsics.atomic_sub_explicit(object, operand, .Acquire)
+	case .release: return intrinsics.atomic_sub_explicit(object, operand, .Release)
+	case .acq_rel: return intrinsics.atomic_sub_explicit(object, operand, .Acq_Rel)
+	case: fallthrough
+	case .seq_cst: return intrinsics.atomic_sub_explicit(object, operand, .Seq_Cst)
 	}
 }

@@ -339,19 +319,14 @@ atomic_fetch_or :: #force_inline proc(object: ^$T, operand: T) -> T {
 }

 atomic_fetch_or_explicit :: #force_inline proc(object: ^$T, operand: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_or_relaxed(object, operand)
-	case .consume:
-		return intrinsics.atomic_or_acq(object, operand)
-	case .acquire:
-		return intrinsics.atomic_or_acq(object, operand)
-	case .release:
-		return intrinsics.atomic_or_rel(object, operand)
-	case .acq_rel:
-		return intrinsics.atomic_or_acqrel(object, operand)
-	case .seq_cst:
-		return intrinsics.atomic_or(object, operand)
+	switch order {
+	case .relaxed: return intrinsics.atomic_or_explicit(object, operand, .Relaxed)
+	case .consume: return intrinsics.atomic_or_explicit(object, operand, .Consume)
+	case .acquire: return intrinsics.atomic_or_explicit(object, operand, .Acquire)
+	case .release: return intrinsics.atomic_or_explicit(object, operand, .Release)
+	case .acq_rel: return intrinsics.atomic_or_explicit(object, operand, .Acq_Rel)
+	case: fallthrough
+	case .seq_cst: return intrinsics.atomic_or_explicit(object, operand, .Seq_Cst)
 	}
 }

@@ -360,19 +335,14 @@ atomic_fetch_xor :: #force_inline proc(object: ^$T, operand: T) -> T {
 }

 atomic_fetch_xor_explicit :: #force_inline proc(object: ^$T, operand: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_xor_relaxed(object, operand)
-	case .consume:
-		return intrinsics.atomic_xor_acq(object, operand)
-	case .acquire:
-		return intrinsics.atomic_xor_acq(object, operand)
-	case .release:
-		return intrinsics.atomic_xor_rel(object, operand)
-	case .acq_rel:
-		return intrinsics.atomic_xor_acqrel(object, operand)
-	case .seq_cst:
-		return intrinsics.atomic_xor(object, operand)
+	switch order {
+	case .relaxed: return intrinsics.atomic_xor_explicit(object, operand, .Relaxed)
+	case .consume: return intrinsics.atomic_xor_explicit(object, operand, .Consume)
+	case .acquire: return intrinsics.atomic_xor_explicit(object, operand, .Acquire)
+	case .release: return intrinsics.atomic_xor_explicit(object, operand, .Release)
+	case .acq_rel: return intrinsics.atomic_xor_explicit(object, operand, .Acq_Rel)
+	case: fallthrough
+	case .seq_cst: return intrinsics.atomic_xor_explicit(object, operand, .Seq_Cst)
 	}
 }

@@ -380,19 +350,14 @@ atomic_fetch_and :: #force_inline proc(object: ^$T, operand: T) -> T {
 	return intrinsics.atomic_and(object, operand)
 }
 atomic_fetch_and_explicit :: #force_inline proc(object: ^$T, operand: T, order: memory_order) -> T {
-	switch (order) {
-	case .relaxed:
-		return intrinsics.atomic_and_relaxed(object, operand)
-	case .consume:
-		return intrinsics.atomic_and_acq(object, operand)
-	case .acquire:
-		return intrinsics.atomic_and_acq(object, operand)
-	case .release:
-		return intrinsics.atomic_and_rel(object, operand)
-	case .acq_rel:
-		return intrinsics.atomic_and_acqrel(object, operand)
-	case .seq_cst:
-		return intrinsics.atomic_and(object, operand)
+	switch order {
+	case .relaxed: return intrinsics.atomic_and_explicit(object, operand, .Relaxed)
+	case .consume: return intrinsics.atomic_and_explicit(object, operand, .Consume)
+	case .acquire: return intrinsics.atomic_and_explicit(object, operand, .Acquire)
+	case .release: return intrinsics.atomic_and_explicit(object, operand, .Release)
+	case .acq_rel: return intrinsics.atomic_and_explicit(object, operand, .Acq_Rel)
+	case: fallthrough
+	case .seq_cst: return intrinsics.atomic_and_explicit(object, operand, .Seq_Cst)
 	}
 }

@@ -1,7 +1,12 @@
 package libc

-when ODIN_OS == "windows" {
-	foreign import libc "system:libucrt.lib"
+when ODIN_OS == .Windows {
+	foreign import libc {
+		"system:libucrt.lib",
+		"system:legacy_stdio_definitions.lib",
+	}
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -11,7 +16,7 @@ when ODIN_OS == "windows" {
 FILE :: struct {}

 // MSVCRT compatible.
-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	_IOFBF       :: 0x0000
 	_IONBF       :: 0x0004
 	_IOLBF       :: 0x0040
@@ -46,7 +51,7 @@ when ODIN_OS == "windows" {
 }

 // GLIBC and MUSL compatible.
-when ODIN_OS == "linux" {
+when ODIN_OS == .Linux {
 	fpos_t        :: struct #raw_union { _: [16]char, _: longlong, _: double, }

 	_IOFBF        :: 0
@@ -67,7 +72,7 @@ when ODIN_OS == "linux" {
 	SEEK_CUR      :: 1
 	SEEK_END      :: 2

-	TMP_MAX       :: 10000
+	TMP_MAX       :: 308915776

 	foreign libc {
 		stderr: ^FILE
@@ -76,6 +81,86 @@ when ODIN_OS == "linux" {
 	}
 }

+when ODIN_OS == .OpenBSD {
+	fpos_t :: distinct i64
+
+	_IOFBF :: 0
+	_IOLBF :: 1
+	_IONBF :: 1
+
+	BUFSIZ :: 1024
+
+	EOF :: int(-1)
+
+	FOPEN_MAX	:: 20
+	FILENAME_MAX	:: 1024
+
+	SEEK_SET :: 0
+	SEEK_CUR :: 1
+	SEEK_END :: 2
+
+	foreign libc {
+		stderr: ^FILE
+		stdin:  ^FILE
+		stdout: ^FILE
+	}
+}
+
+when ODIN_OS == .FreeBSD {
+	fpos_t :: distinct i64
+
+	_IOFBF :: 0
+	_IOLBF :: 1
+	_IONBF :: 1
+
+	BUFSIZ :: 1024
+
+	EOF :: int(-1)
+
+	FOPEN_MAX	:: 20
+	FILENAME_MAX	:: 1024
+
+	SEEK_SET :: 0
+	SEEK_CUR :: 1
+	SEEK_END :: 2
+
+	foreign libc {
+		stderr: ^FILE
+		stdin:  ^FILE
+		stdout: ^FILE
+	}
+}
+
+when ODIN_OS == .Darwin {
+	fpos_t :: distinct i64
+	
+	_IOFBF        :: 0
+	_IOLBF        :: 1
+	_IONBF        :: 2
+
+	BUFSIZ        :: 1024
+
+	EOF           :: int(-1)
+
+	FOPEN_MAX     :: 20
+
+	FILENAME_MAX  :: 1024
+
+	L_tmpnam      :: 1024
+
+	SEEK_SET      :: 0
+	SEEK_CUR      :: 1
+	SEEK_END      :: 2
+
+	TMP_MAX       :: 308915776
+
+	foreign libc {
+		@(link_name="__stderrp") stderr: ^FILE
+		@(link_name="__stdinp")  stdin:  ^FILE
+		@(link_name="__stdoutp") stdout: ^FILE
+	}
+}
+
@(default_calling_convention="c")
 foreign libc {
 	// 7.21.4 Operations on files
@@ -114,10 +199,10 @@ foreign libc {
 	getc      :: proc(stream: ^FILE) -> int ---
 	getchar   :: proc() -> int ---
 	putc      :: proc(c: int, stream: ^FILE) -> int ---
-	putchar   :: proc() -> int ---
+	putchar   :: proc(c: int) -> int ---
 	puts      :: proc(s: cstring) -> int ---
 	ungetc    :: proc(c: int, stream: ^FILE) -> int ---
-	fread     :: proc(ptr: rawptr, size: size_t, stream: ^FILE) -> size_t ---
+	fread     :: proc(ptr: rawptr, size: size_t, nmemb: size_t, stream: ^FILE) -> size_t ---
 	fwrite    :: proc(ptr: rawptr, size: size_t, nmemb: size_t, stream: ^FILE) -> size_t ---

 	// 7.21.9 File positioning functions
@@ -2,13 +2,15 @@ package libc

 // 7.22 General utilities

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	RAND_MAX :: 0x7fff

 	@(private="file")
@@ -22,7 +24,7 @@ when ODIN_OS == "windows" {
 	}
 }

-when ODIN_OS == "linux" {
+when ODIN_OS == .Linux {
 	RAND_MAX :: 0x7fffffff

 	// GLIBC and MUSL only
@@ -33,7 +35,23 @@ when ODIN_OS == "linux" {
 	}

 	MB_CUR_MAX :: #force_inline proc() -> size_t {
-		return __ctype_get_mb_cur_max()
+		return size_t(__ctype_get_mb_cur_max())
+	}
+}
+
+
+when ODIN_OS == .Darwin {
+	RAND_MAX :: 0x7fffffff
+
+	// GLIBC and MUSL only
+	@(private="file")
+	@(default_calling_convention="c")
+	foreign libc {
+		___mb_cur_max :: proc() -> int ---
+	}
+
+	MB_CUR_MAX :: #force_inline proc() -> size_t {
+		return size_t(___mb_cur_max())
 	}
 }

@@ -70,7 +88,6 @@ foreign libc {
 	srand         :: proc(seed: uint) ---

 	// 7.22.3 Memory management functions
-	aligned_alloc :: proc(aligment, size: size_t) -> rawptr ---
 	calloc        :: proc(nmemb, size: size_t) -> rawptr ---
 	free          :: proc(ptr: rawptr) ---
 	malloc        :: proc(size: size_t) -> rawptr ---
@@ -107,3 +124,30 @@ foreign libc {
 	mbstowcs      :: proc(pwcs: ^wchar_t, s: cstring, n: size_t) -> size_t ---
 	wcstombs      :: proc(s: [^]char, pwcs: ^wchar_t, n: size_t) -> size_t ---
 }
+
+
+aligned_alloc :: #force_inline proc "c" (alignment, size: size_t) -> rawptr {
+	when ODIN_OS == .Windows {
+		foreign libc {
+			_aligned_malloc :: proc(size, alignment: size_t) -> rawptr ---
+		}
+		return _aligned_malloc(size=size, alignment=alignment)
+	} else {
+		foreign libc {
+			aligned_alloc :: proc(alignment, size: size_t) -> rawptr ---
+		}
+		return aligned_alloc(alignment=alignment, size=size)
+	}
+}
+
+
+aligned_free :: #force_inline proc "c" (ptr: rawptr) {
+	when ODIN_OS == .Windows {
+		foreign libc {
+			_aligned_free :: proc(ptr: rawptr) ---
+		}
+		_aligned_free(ptr)
+	} else {
+		free(ptr)
+	}
+}
@@ -4,8 +4,10 @@ import "core:runtime"

 // 7.24 String handling

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -5,10 +5,10 @@ package libc
 thrd_start_t :: proc "c" (rawptr) -> int
 tss_dtor_t   :: proc "c" (rawptr)

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc {
 		"system:libucrt.lib", 
-		"system:msvcprt.lib"
+		"system:msvcprt.lib",
 	}

 	thrd_success        :: 0                             // _Thrd_success
@@ -44,7 +44,7 @@ when ODIN_OS == "windows" {
 		@(link_name="_Cnd_destroy")   cnd_destroy   :: proc(cond: ^cnd_t) ---
 		@(link_name="_Cnd_init")      cnd_init      :: proc(cond: ^cnd_t) -> int ---
 		@(link_name="_Cnd_signal")    cnd_signal    :: proc(cond: ^cnd_t) -> int ---
-		@(link_name="_Cnd_timedwait") cnd_timedwait :: proc(cond: ^cnd_t, ts: ^timespec) -> int ---
+		@(link_name="_Cnd_timedwait") cnd_timedwait :: proc(cond: ^cnd_t, mtx: ^mtx_t, ts: ^timespec) -> int ---
 		@(link_name="_Cnd_wait")      cnd_wait      :: proc(cond: ^cnd_t, mtx: ^mtx_t) -> int ---
 		
 		// 7.26.4 Mutex functions
@@ -74,10 +74,10 @@ when ODIN_OS == "windows" {
 }

 // GLIBC and MUSL compatible constants and types.
-when ODIN_OS == "linux" {
+when ODIN_OS == .Linux {
 	foreign import libc {
 		"system:c",
-		"system:pthread"
+		"system:pthread",
 	}

 	thrd_success        :: 0
@@ -108,7 +108,7 @@ when ODIN_OS == "linux" {
 		cnd_destroy   :: proc(cond: ^cnd_t) ---
 		cnd_init      :: proc(cond: ^cnd_t) -> int ---
 		cnd_signal    :: proc(cond: ^cnd_t) -> int ---
-		cnd_timedwait :: proc(cond: ^cnd_t, ts: ^timespec) -> int ---
+		cnd_timedwait :: proc(cond: ^cnd_t, mtx: ^mtx_t, ts: ^timespec) -> int ---
 		cnd_wait      :: proc(cond: ^cnd_t, mtx: ^mtx_t) -> int ---
 		
 		// 7.26.4 Mutex functions
@@ -136,3 +136,8 @@ when ODIN_OS == "linux" {
 		tss_set       :: proc(key: tss_t, val: rawptr) -> int ---
 	}
 }
+
+
+when ODIN_OS == .Darwin {
+	// TODO: find out what this is meant to be!
+}
@@ -2,15 +2,17 @@ package libc

 // 7.27 Date and time

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

 // We enforce 64-bit time_t and timespec as there is no reason to use 32-bit as
 // we approach the 2038 problem. Windows has defaulted to this since VC8 (2005).
-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign libc {
 		// 7.27.2 Time manipulation functions
 		                               clock        :: proc() -> clock_t ---
@@ -33,17 +35,17 @@ 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,
 	}
 }

-when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
+when ODIN_OS == .Linux || ODIN_OS == .FreeBSD || ODIN_OS == .Darwin || ODIN_OS == .OpenBSD {
 	@(default_calling_convention="c")
 	foreign libc {
 		// 7.27.2 Time manipulation functions
@@ -61,7 +63,12 @@ when ODIN_OS == "linux" || ODIN_OS == "freebsd" {
 		strftime     :: proc(s: [^]char, maxsize: size_t, format: cstring, timeptr: ^tm) -> size_t ---
 	}

-	CLOCKS_PER_SEC :: 1000000
+	when ODIN_OS == .OpenBSD {
+		CLOCKS_PER_SEC :: 100
+	} else {
+		CLOCKS_PER_SEC :: 1000000
+	}
+
 	TIME_UTC       :: 1

 	time_t         :: distinct i64
@@ -75,7 +82,7 @@ when ODIN_OS == "linux" || ODIN_OS == "freebsd" {

 	tm :: struct {
 		tm_sec, tm_min, tm_hour, tm_mday, tm_mon, tm_year, tm_wday, tm_yday, tm_isdst: int,
-		_: long,
-		_: rawptr,
+		tm_gmtoff: long,
+		tm_zone: rawptr,
 	}
 }
@@ -3,6 +3,8 @@ package libc
 import "core:c"

 char           :: c.char // assuming -funsigned-char
+
+schar          :: c.schar
 short          :: c.short
 int            :: c.int
 long           :: c.long
@@ -2,8 +2,10 @@ package libc

 // 7.28 Unicode utilities

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -2,8 +2,10 @@ package libc

 // 7.29 Extended multibyte and wide character utilities

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }
@@ -96,7 +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.
@@ -2,20 +2,34 @@ package libc

 // 7.30 Wide character classification and mapping utilities

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"
+} else when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
 } else {
 	foreign import libc "system:c"
 }

-when ODIN_OS == "windows" {
+when ODIN_OS == .Windows {
 	wctrans_t :: distinct wchar_t
 	wctype_t  :: distinct ushort
-}

-when ODIN_OS == "linux" {
-	wctrans_t :: distinct rawptr
+} else when ODIN_OS == .Linux {
+	wctrans_t :: distinct intptr_t
 	wctype_t  :: distinct ulong
+
+} else when ODIN_OS == .Darwin {
+	wctrans_t :: distinct int
+	wctype_t  :: distinct u32
+
+} else when ODIN_OS == .OpenBSD {
+	wctrans_t :: distinct rawptr
+	wctype_t  :: distinct rawptr
+
+} else when ODIN_OS == .FreeBSD {
+	wctrans_t :: distinct int
+	wctype_t  :: distinct ulong
+	
 }

@(default_calling_convention="c")
@@ -1,5 +1,3 @@
-package compress
-
 /*
 	Copyright 2021 Jeroen van Rijn <nom@duclavier.com>.
 	Made available under Odin's BSD-3 license.
@@ -8,9 +6,13 @@ package compress
 		Jeroen van Rijn: Initial implementation, optimization.
 */

+
+// package compress is a collection of utilities to aid with other compression packages
+package compress
+
 import "core:io"
-import "core:image"
 import "core:bytes"
+import "core:runtime"

 /*
 	These settings bound how much compression algorithms will allocate for their output buffer.
@@ -45,20 +47,18 @@ when size_of(uintptr) == 8 {
 }


-Error :: union {
+Error :: union #shared_nil {
 	General_Error,
 	Deflate_Error,
 	ZLIB_Error,
 	GZIP_Error,
 	ZIP_Error,
-	/*
-		This is here because png.load will return a this type of error union,
-		as it may involve an I/O error, a Deflate error, etc.
-	*/
-	image.Error,
+
+	runtime.Allocator_Error,
 }

 General_Error :: enum {
+	None = 0,
 	File_Not_Found,
 	Cannot_Open_File,
 	File_Too_Short,
@@ -69,7 +69,6 @@ General_Error :: enum {
 	Incompatible_Options,
 	Unimplemented,

-
 	/*
 		Memory errors
 	*/
@@ -78,6 +77,7 @@ General_Error :: enum {
 }

 GZIP_Error :: enum {
+	None = 0,
 	Invalid_GZIP_Signature,
 	Reserved_Flag_Set,
 	Invalid_Extra_Data,
@@ -102,6 +102,7 @@ GZIP_Error :: enum {
 }

 ZIP_Error :: enum {
+	None = 0,
 	Invalid_ZIP_File_Signature,
 	Unexpected_Signature,
 	Insert_Next_Disk,
@@ -109,6 +110,7 @@ ZIP_Error :: enum {
 }

 ZLIB_Error :: enum {
+	None = 0,
 	Unsupported_Window_Size,
 	FDICT_Unsupported,
 	Unsupported_Compression_Level,
@@ -116,6 +118,7 @@ ZLIB_Error :: enum {
 }

 Deflate_Error :: enum {
+	None = 0,
 	Huffman_Bad_Sizes,
 	Huffman_Bad_Code_Lengths,
 	Inflate_Error,
@@ -125,7 +128,6 @@ Deflate_Error :: enum {
 	BType_3,
 }

-
 // General I/O context for ZLIB, LZW, etc.
 Context_Memory_Input :: struct #packed {
 	input_data:        []u8,
@@ -141,7 +143,12 @@ Context_Memory_Input :: struct #packed {
 	size_packed:       i64,
 	size_unpacked:     i64,
 }
-#assert(size_of(Context_Memory_Input) == 64)
+when size_of(rawptr) == 8 {
+	#assert(size_of(Context_Memory_Input) == 64)
+} else {
+	// e.g. `-target:windows_i386`
+	#assert(size_of(Context_Memory_Input) == 52)
+}

 Context_Stream_Input :: struct #packed {
 	input_data:        []u8,
@@ -176,14 +183,13 @@ Context_Stream_Input :: struct #packed {
 	This simplifies end-of-stream handling where bits may be left in the bit buffer.
 */

-// TODO: Make these return compress.Error errors.
-
 input_size_from_memory :: proc(z: ^Context_Memory_Input) -> (res: i64, err: Error) {
 	return i64(len(z.input_data)), nil
 }

 input_size_from_stream :: proc(z: ^Context_Stream_Input) -> (res: i64, err: Error) {
-	return io.size(z.input), nil
+	res, _ = io.size(z.input)
+	return
 }

 input_size :: proc{input_size_from_memory, input_size_from_stream}
@@ -207,25 +213,19 @@ 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
 	b := make([]u8, size, context.temp_allocator)
-	_, e := z.input->impl_read(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")
@@ -242,12 +242,8 @@ 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}
@@ -289,17 +285,32 @@ peek_data_from_memory :: #force_inline proc(z: ^Context_Memory_Input, $T: typeid
 	}
 }

+@(optimization_mode="speed")
+peek_data_at_offset_from_memory :: #force_inline proc(z: ^Context_Memory_Input, $T: typeid, #any_int offset: int) -> (res: T, err: io.Error) {
+	size :: size_of(T)
+
+	#no_bounds_check {
+		if len(z.input_data) >= size + offset {
+			buf := z.input_data[offset:][:size]
+			return (^T)(&buf[0])^, .None
+		}
+	}
+
+	if len(z.input_data) == 0 {
+		return T{}, .EOF
+	} else {
+		return T{}, .Short_Buffer
+	}
+}
+
@(optimization_mode="speed")
 peek_data_from_stream :: #force_inline proc(z: ^Context_Stream_Input, $T: typeid) -> (res: T, err: io.Error) {
 	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 {
@@ -307,8 +318,8 @@ 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
 	}

@@ -316,7 +327,37 @@ peek_data_from_stream :: #force_inline proc(z: ^Context_Stream_Input, $T: typeid
 	return res, .None
 }

-peek_data :: proc{peek_data_from_memory, peek_data_from_stream}
+@(optimization_mode="speed")
+peek_data_at_offset_from_stream :: #force_inline proc(z: ^Context_Stream_Input, $T: typeid, #any_int offset: int) -> (res: T, err: io.Error) {
+	size :: size_of(T)
+
+	// Get current position to return to.
+	cur_pos := z.input->impl_seek(0, .Current) or_return
+	// Seek to offset.
+	pos := z.input->impl_seek(offset, .Start) or_return
+
+	r, e3 := io.to_reader_at(z.input)
+	if !e3 {
+		return T{}, .Empty
+	}
+	when size <= 128 {
+		b: [size]u8
+	} else {
+		b := make([]u8, size, context.temp_allocator)
+	}
+	_, e4 := io.read_at(r, b[:], pos)
+	if e4 != .None {
+		return T{}, .Empty
+	}
+
+	// Return read head to original position.
+	z.input->impl_seek(cur_pos, .Start)
+
+	res = (^T)(&b[0])^
+	return res, .None
+}
+
+peek_data :: proc{peek_data_from_memory, peek_data_from_stream, peek_data_at_offset_from_memory, peek_data_at_offset_from_stream}



@@ -402,7 +443,7 @@ 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")
@@ -410,7 +451,7 @@ 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}
@@ -418,13 +459,13 @@ 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}
@@ -475,4 +516,4 @@ discard_to_next_byte_lsb_from_stream :: proc(z: ^Context_Stream_Input) {
 	consume_bits_lsb(z, discard)
 }

-discard_to_next_byte_lsb :: proc{discard_to_next_byte_lsb_from_memory, discard_to_next_byte_lsb_from_stream};
+discard_to_next_byte_lsb :: proc{discard_to_next_byte_lsb_from_memory, discard_to_next_byte_lsb_from_stream}
@@ -1,4 +1,4 @@
-//+ignore
+//+build ignore
 package gzip

 /*
@@ -45,7 +45,7 @@ main :: proc() {

 	if len(args) < 2 {
 		stderr("No input file specified.\n")
-		err := load(slice=TEST, buf=&buf, known_gzip_size=len(TEST))
+		err := load(data=TEST, buf=&buf, known_gzip_size=len(TEST))
 		if err == nil {
 			stdout("Displaying test vector: ")
 			stdout(bytes.buffer_to_string(&buf))
@@ -66,7 +66,8 @@ OS :: enum u8 {
 	_Unknown     = 14,
 	Unknown      = 255,
 }
-OS_Name :: #partial [OS]string{
+OS_Name :: #sparse[OS]string{
+	._Unknown     = "",
 	.FAT          = "FAT",
 	.Amiga        = "Amiga",
 	.VMS          = "VMS/OpenVMS",
@@ -99,9 +100,9 @@ E_GZIP    :: compress.GZIP_Error
 E_ZLIB    :: compress.ZLIB_Error
 E_Deflate :: compress.Deflate_Error

-GZIP_MAX_PAYLOAD_SIZE :: int(max(u32le))
+GZIP_MAX_PAYLOAD_SIZE :: i64(max(u32le))

-load :: proc{load_from_slice, load_from_file, load_from_context}
+load :: proc{load_from_bytes, load_from_file, load_from_context}

 load_from_file :: proc(filename: string, buf: ^bytes.Buffer, expected_output_size := -1, allocator := context.allocator) -> (err: Error) {
 	context.allocator = allocator
@@ -111,16 +112,16 @@ load_from_file :: proc(filename: string, buf: ^bytes.Buffer, expected_output_siz

 	err = E_General.File_Not_Found
 	if ok {
-		err = load_from_slice(data, buf, len(data), expected_output_size)
+		err = load_from_bytes(data, buf, len(data), expected_output_size)
 	}
 	return
 }

-load_from_slice :: proc(slice: []u8, buf: ^bytes.Buffer, known_gzip_size := -1, expected_output_size := -1, allocator := context.allocator) -> (err: Error) {
+load_from_bytes :: proc(data: []byte, buf: ^bytes.Buffer, known_gzip_size := -1, expected_output_size := -1, allocator := context.allocator) -> (err: Error) {
 	buf := buf

 	z := &compress.Context_Memory_Input{
-		input_data = slice,
+		input_data = data,
 		output = buf,
 	}
 	return load_from_context(z, buf, known_gzip_size, expected_output_size, allocator)
@@ -135,7 +136,7 @@ load_from_context :: proc(z: ^$C, buf: ^bytes.Buffer, known_gzip_size := -1, exp

 	z.output = buf

-	if expected_output_size > GZIP_MAX_PAYLOAD_SIZE {
+	if i64(expected_output_size) > i64(GZIP_MAX_PAYLOAD_SIZE) {
 		return E_GZIP.Payload_Size_Exceeds_Max_Payload
 	}

@@ -334,10 +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=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.
 	*/
@@ -0,0 +1,148 @@
+/*
+	This file was generated, so don't edit this by hand.
+	Transliterated from https://github.com/Ed-von-Schleck/shoco/blob/master/shoco_model.h,
+	which is an English word model.
+*/
+
+// package shoco is an implementation of the shoco short string compressor
+package shoco
+
+DEFAULT_MODEL :: Shoco_Model {
+	min_char = 39,
+	max_char = 122,
+	characters_by_id = {
+		'e', 'a', 'i', 'o', 't', 'h', 'n', 'r', 's', 'l', 'u', 'c', 'w', 'm', 'd', 'b', 'p', 'f', 'g', 'v', 'y', 'k', '-', 'H', 'M', 'T', '\'', 'B', 'x', 'I', 'W', 'L',
+	},
+	ids_by_character = {
+		-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 26, -1, -1, -1, -1, -1, 22, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 27, -1, -1, -1, -1, -1, 23, 29, -1, -1, 31, 24, -1, -1, -1, -1, -1, -1, 25, -1, -1, 30, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 15, 11, 14, 0, 17, 18, 5, 2, -1, 21, 9, 13, 6, 3, 16, -1, 7, 8, 4, 10, 19, 12, 28, 20, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+	},
+	successors_by_bigram = {
+		7, 4, 12, -1, 6, -1, 1, 0, 3, 5, -1, 9, -1, 8, 2, -1, 15, 14, -1, 10, 11, -1, -1, -1, -1, -1, -1, -1, 13, -1, -1, -1,
+		1, -1, 6, -1, 1, -1, 0, 3, 2, 4, 15, 11, -1, 9, 5, 10, 13, -1, 12, 8, 7, 14, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		9, 11, -1, 4, 2, -1, 0, 8, 1, 5, -1, 6, -1, 3, 7, 15, -1, 12, 10, 13, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		-1, -1, 14, 7, 5, -1, 1, 2, 8, 9, 0, 15, 6, 4, 11, -1, 12, 3, -1, 10, -1, 13, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		2, 4, 3, 1, 5, 0, -1, 6, 10, 9, 7, 12, 11, -1, -1, -1, -1, 13, -1, -1, 8, -1, 15, -1, -1, -1, 14, -1, -1, -1, -1, -1,
+		0, 1, 2, 3, 4, -1, -1, 5, 9, 10, 6, -1, -1, 8, 15, 11, -1, 14, -1, -1, 7, -1, 13, -1, -1, -1, 12, -1, -1, -1, -1, -1,
+		2, 8, 7, 4, 3, -1, 9, -1, 6, 11, -1, 5, -1, -1, 0, -1, -1, 14, 1, 15, 10, 12, -1, -1, -1, -1, 13, -1, -1, -1, -1, -1,
+		0, 3, 1, 2, 6, -1, 9, 8, 4, 12, 13, 10, -1, 11, 7, -1, -1, 15, 14, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 6, 3, 4, 1, 2, -1, -1, 5, 10, 7, 9, 11, 12, -1, -1, 8, 14, -1, -1, 15, 13, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 6, 2, 5, 9, -1, -1, -1, 10, 1, 8, -1, 12, 14, 4, -1, 15, 7, -1, 13, 3, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		8, 10, 9, 15, 1, -1, 4, 0, 3, 2, -1, 6, -1, 12, 11, 13, 7, 14, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		1, 3, 6, 0, 4, 2, -1, 7, 13, 8, 9, 11, -1, -1, 15, -1, -1, -1, -1, -1, 10, 5, 14, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		3, 0, 1, 4, -1, 2, 5, 6, 7, 8, -1, 14, -1, -1, 9, 15, -1, 12, -1, -1, -1, 10, 11, -1, -1, -1, 13, -1, -1, -1, -1, -1,
+		0, 1, 3, 2, 15, -1, 12, -1, 7, 14, 4, -1, -1, 9, -1, 8, 5, 10, -1, -1, 6, -1, 13, -1, -1, -1, 11, -1, -1, -1, -1, -1,
+		0, 3, 1, 2, -1, -1, 12, 6, 4, 9, 7, -1, -1, 14, 8, -1, -1, 15, 11, 13, 5, -1, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 5, 7, 2, 10, 13, -1, 6, 8, 1, 3, -1, -1, 14, 15, 11, -1, -1, -1, 12, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 2, 6, 3, 7, 10, -1, 1, 9, 4, 8, -1, -1, 15, -1, 12, 5, -1, -1, -1, 11, -1, 13, -1, -1, -1, 14, -1, -1, -1, -1, -1,
+		1, 3, 4, 0, 7, -1, 12, 2, 11, 8, 6, 13, -1, -1, -1, -1, -1, 5, -1, -1, 10, 15, 9, -1, -1, -1, 14, -1, -1, -1, -1, -1,
+		1, 3, 5, 2, 13, 0, 9, 4, 7, 6, 8, -1, -1, 15, -1, 11, -1, -1, 10, -1, 14, -1, 12, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 2, 1, 3, -1, -1, -1, 6, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		1, 11, 4, 0, 3, -1, 13, 12, 2, 7, -1, -1, 15, 10, 5, 8, 14, -1, -1, -1, -1, -1, 9, -1, -1, -1, 6, -1, -1, -1, -1, -1,
+		0, 9, 2, 14, 15, 4, 1, 13, 3, 5, -1, -1, 10, -1, -1, -1, -1, 6, 12, -1, 7, -1, 8, -1, -1, -1, 11, -1, -1, -1, -1, -1,
+		-1, 2, 14, -1, 1, 5, 8, 7, 4, 12, -1, 6, 9, 11, 13, 3, 10, 15, -1, -1, -1, -1, 0, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		0, 1, 3, 2, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		4, 3, 1, 5, -1, -1, -1, 0, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		2, 8, 4, 1, -1, 0, -1, 6, -1, -1, 5, -1, 7, -1, -1, -1, -1, -1, -1, -1, 10, -1, -1, 9, -1, -1, -1, -1, -1, -1, -1, -1,
+		12, 5, -1, -1, 1, -1, -1, 7, 0, 3, -1, 2, -1, 4, 6, -1, -1, -1, -1, 8, -1, -1, 15, -1, 13, 9, -1, -1, -1, -1, -1, 11,
+		1, 3, 2, 4, -1, -1, -1, 5, -1, 7, 0, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, 8, -1, -1,
+		5, 3, 4, 12, 1, 6, -1, -1, -1, -1, 8, 2, -1, -1, -1, -1, 0, 9, -1, -1, 11, -1, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+		-1, -1, -1, -1, 0, -1, 1, 12, 3, -1, -1, -1, -1, 5, -1, -1, -1, 2, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, -1, 6, -1, 10,
+		2, 3, 1, 4, -1, 0, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 7, -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, -1,
+		5, 1, 3, 0, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, -1, -1, -1, -1, -1, 9, -1, -1, 6, -1, 7,
+	},
+	successors_reversed = {
+		's', 't', 'c', 'l', 'm', 'a', 'd', 'r', 'v', 'T', 'A', 'L', 'e', 'M', 'Y', '-',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'-', 't', 'a', 'b', 's', 'h', 'c', 'r', 'n', 'w', 'p', 'm', 'l', 'd', 'i', 'f',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'u', 'e', 'i', 'a', 'o', 'r', 'y', 'l', 'I', 'E', 'R', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'e', 'a', 'o', 'i', 'u', 'A', 'y', 'E', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		't', 'n', 'f', 's', '\'', 'm', 'I', 'N', 'A', 'E', 'L', 'Z', 'r', 'V', 'R', 'C',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'o', 'a', 'y', 'i', 'u', 'e', 'I', 'L', 'D', '\'', 'E', 'Y', '\x00', '\x00', '\x00', '\x00',
+		'r', 'i', 'y', 'a', 'e', 'o', 'u', 'Y', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'h', 'o', 'e', 'E', 'i', 'u', 'r', 'w', 'a', 'H', 'y', 'R', 'Z', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'h', 'i', 'e', 'a', 'o', 'r', 'I', 'y', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'n', 't', 's', 'r', 'l', 'd', 'i', 'y', 'v', 'm', 'b', 'c', 'g', 'p', 'k', 'u',
+		'e', 'l', 'o', 'u', 'y', 'a', 'r', 'i', 's', 'j', 't', 'b', 'v', 'h', 'm', 'd',
+		'o', 'e', 'h', 'a', 't', 'k', 'i', 'r', 'l', 'u', 'y', 'c', 'q', 's', '-', 'd',
+		'e', 'i', 'o', 'a', 's', 'y', 'r', 'u', 'd', 'l', '-', 'g', 'n', 'v', 'm', 'f',
+		'r', 'n', 'd', 's', 'a', 'l', 't', 'e', 'm', 'c', 'v', 'y', 'i', 'x', 'f', 'p',
+		'o', 'e', 'r', 'a', 'i', 'f', 'u', 't', 'l', '-', 'y', 's', 'n', 'c', '\'', 'k',
+		'h', 'e', 'o', 'a', 'r', 'i', 'l', 's', 'u', 'n', 'g', 'b', '-', 't', 'y', 'm',
+		'e', 'a', 'i', 'o', 't', 'r', 'u', 'y', 'm', 's', 'l', 'b', '\'', '-', 'f', 'd',
+		'n', 's', 't', 'm', 'o', 'l', 'c', 'd', 'r', 'e', 'g', 'a', 'f', 'v', 'z', 'b',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'e', 'n', 'i', 's', 'h', 'l', 'f', 'y', '-', 'a', 'w', '\'', 'g', 'r', 'o', 't',
+		'e', 'l', 'i', 'y', 'd', 'o', 'a', 'f', 'u', 't', 's', 'k', 'w', 'v', 'm', 'p',
+		'e', 'a', 'o', 'i', 'u', 'p', 'y', 's', 'b', 'm', 'f', '\'', 'n', '-', 'l', 't',
+		'd', 'g', 'e', 't', 'o', 'c', 's', 'i', 'a', 'n', 'y', 'l', 'k', '\'', 'f', 'v',
+		'u', 'n', 'r', 'f', 'm', 't', 'w', 'o', 's', 'l', 'v', 'd', 'p', 'k', 'i', 'c',
+		'e', 'r', 'a', 'o', 'l', 'p', 'i', 't', 'u', 's', 'h', 'y', 'b', '-', '\'', 'm',
+		'\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'e', 'i', 'o', 'a', 's', 'y', 't', 'd', 'r', 'n', 'c', 'm', 'l', 'u', 'g', 'f',
+		'e', 't', 'h', 'i', 'o', 's', 'a', 'u', 'p', 'c', 'l', 'w', 'm', 'k', 'f', 'y',
+		'h', 'o', 'e', 'i', 'a', 't', 'r', 'u', 'y', 'l', 's', 'w', 'c', 'f', '\'', '-',
+		'r', 't', 'l', 's', 'n', 'g', 'c', 'p', 'e', 'i', 'a', 'd', 'm', 'b', 'f', 'o',
+		'e', 'i', 'a', 'o', 'y', 'u', 'r', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00', '\x00',
+		'a', 'i', 'h', 'e', 'o', 'n', 'r', 's', 'l', 'd', 'k', '-', 'f', '\'', 'c', 'b',
+		'p', 't', 'c', 'a', 'i', 'e', 'h', 'q', 'u', 'f', '-', 'y', 'o', '\x00', '\x00', '\x00',
+		'o', 'e', 's', 't', 'i', 'd', '\'', 'l', 'b', '-', 'm', 'a', 'r', 'n', 'p', 'w',
+	},
+
+	character_count = 32,
+	successor_count = 16,
+
+	max_successor_n = 7,
+	packs = {
+		{ 0x80000000, 1, 2, { 26, 24, 24, 24, 24, 24, 24, 24 }, { 15,  3,  0,  0, 0, 0, 0, 0 }, 0xc0, 0x80 },
+		{ 0xc0000000, 2, 4, { 25, 22, 19, 16, 16, 16, 16, 16 }, { 15,  7,  7,  7, 0, 0, 0, 0 }, 0xe0, 0xc0 },
+		{ 0xe0000000, 4, 8, { 23, 19, 15, 11,  8,  5,  2,  0 }, { 31, 15, 15, 15, 7, 7, 7, 3 }, 0xf0, 0xe0 },
+	},
+}
@@ -0,0 +1,314 @@
+/*
+	Copyright 2022 Jeroen van Rijn <nom@duclavier.com>.
+	Made available under Odin's BSD-3 license.
+
+	List of contributors:
+		Jeroen van Rijn: Initial implementation.
+
+	An implementation of [shoco](https://github.com/Ed-von-Schleck/shoco) by Christian Schramm.
+*/
+
+// package shoco is an implementation of the shoco short string compressor
+package shoco
+
+import "core:intrinsics"
+import "core:compress"
+
+Shoco_Pack :: struct {
+	word:           u32,
+	bytes_packed:   i8,
+	bytes_unpacked: i8,
+	offsets:        [8]u16,
+	masks:          [8]i16,
+	header_mask:    u8,
+	header:         u8,
+}
+
+Shoco_Model :: struct {
+	min_char:             u8,
+	max_char:             u8,
+	characters_by_id:     []u8,
+	ids_by_character:     [256]i16,
+	successors_by_bigram: []i8,
+	successors_reversed:  []u8,
+
+	character_count:      u8,
+	successor_count:      u8,
+	max_successor_n:      i8,
+	packs:                []Shoco_Pack,
+}
+
+compress_bound :: proc(uncompressed_size: int) -> (worst_case_compressed_size: int) {
+	// Worst case compression happens when input is non-ASCII (128-255)
+	// Encoded as 0x00 + the byte in question.
+	return uncompressed_size * 2
+}
+
+decompress_bound :: proc(compressed_size: int, model := DEFAULT_MODEL) -> (maximum_decompressed_size: int) {
+	// Best case compression is 2:1
+	most: f64
+	for pack in model.packs {
+		val := f64(compressed_size) / f64(pack.bytes_packed) * f64(pack.bytes_unpacked)
+		most = max(most, val)
+	}
+	return int(most)
+}
+
+find_best_encoding :: proc(indices: []i16, n_consecutive: i8, model := DEFAULT_MODEL) -> (res: int) {
+	for p := len(model.packs); p > 0; p -= 1 {
+		pack := model.packs[p - 1]
+		if n_consecutive >= pack.bytes_unpacked {
+			have_index := true
+			for i := 0; i < int(pack.bytes_unpacked); i += 1 {
+				if indices[i] > pack.masks[i] {
+					have_index = false
+					break
+				}
+			}
+			if have_index {
+				return p - 1
+			}
+		}
+	}
+	return -1
+}
+
+validate_model :: proc(model: Shoco_Model) -> (int, compress.Error) {
+	if len(model.characters_by_id) != int(model.character_count) {
+		return 0, .Unknown_Compression_Method
+	}
+
+	if len(model.successors_by_bigram) != int(model.character_count) * int(model.character_count) {
+		return 0, .Unknown_Compression_Method
+	}
+
+	if len(model.successors_reversed) != int(model.successor_count) * int(model.max_char - model.min_char) {
+		return 0, .Unknown_Compression_Method
+	}
+
+	// Model seems legit.
+	return 0, nil
+}
+
+// Decompresses into provided buffer.
+decompress_slice_to_output_buffer :: proc(input: []u8, output: []u8, model := DEFAULT_MODEL) -> (size: int, err: compress.Error) {
+	inp, inp_end := 0, len(input)
+	out, out_end := 0, len(output)
+
+	validate_model(model) or_return
+
+	for inp < inp_end {
+		val  := transmute(i8)input[inp]
+		mark := int(-1)
+
+		for val < 0 {
+			val <<= 1
+			mark += 1
+		}
+
+		if mark > len(model.packs) {
+			return out, .Unknown_Compression_Method
+		}
+
+		if mark < 0 {
+			if out >= out_end {
+				return out, .Output_Too_Short
+			}
+
+			// Ignore the sentinel value for non-ASCII chars
+			if input[inp] == 0x00 {
+				inp += 1
+				if inp >= inp_end {
+					return out, .Stream_Too_Short
+				}
+			}
+			output[out] = input[inp]
+			inp, out = inp + 1, out + 1
+
+		} else {
+			pack := model.packs[mark]
+
+			if out + int(pack.bytes_unpacked) > out_end {
+				return out, .Output_Too_Short
+			} else if inp + int(pack.bytes_packed) > inp_end {
+				return out, .Stream_Too_Short
+			}
+
+			code := intrinsics.unaligned_load((^u32)(&input[inp]))
+			when ODIN_ENDIAN == .Little {
+				code = intrinsics.byte_swap(code)
+			}
+
+			// Unpack the leading char
+			offset := pack.offsets[0]
+			mask   := pack.masks[0]
+
+			last_chr := model.characters_by_id[(code >> offset) & u32(mask)]
+			output[out] = last_chr
+
+			// Unpack the successor chars
+			for i := 1; i < int(pack.bytes_unpacked); i += 1 {
+				offset = pack.offsets[i]
+				mask   = pack.masks[i]
+
+				index_major := u32(last_chr - model.min_char) * u32(model.successor_count)
+				index_minor := (code >> offset) & u32(mask)
+
+				last_chr = model.successors_reversed[index_major + index_minor]
+
+				output[out + i] = last_chr
+			}
+
+			out += int(pack.bytes_unpacked)
+			inp += int(pack.bytes_packed)
+		}
+	}
+
+	return out, nil
+}
+
+decompress_slice_to_string :: proc(input: []u8, model := DEFAULT_MODEL, allocator := context.allocator) -> (res: string, err: compress.Error) {
+	context.allocator = allocator
+
+	if len(input) == 0 {
+		return "", .Stream_Too_Short
+	}
+
+	max_output_size := decompress_bound(len(input), model)
+
+	buf: [dynamic]u8
+	resize(&buf, max_output_size) or_return
+
+	length, result := decompress_slice_to_output_buffer(input, buf[:])
+	resize(&buf, length) or_return
+	return string(buf[:]), result
+}
+decompress :: proc{decompress_slice_to_output_buffer, decompress_slice_to_string}
+
+compress_string_to_buffer :: proc(input: string, output: []u8, model := DEFAULT_MODEL, allocator := context.allocator) -> (size: int, err: compress.Error) {
+	inp, inp_end := 0, len(input)
+	out, out_end := 0, len(output)
+	output := output
+
+	validate_model(model) or_return
+
+	indices := make([]i16, model.max_successor_n + 1)
+	defer delete(indices)
+
+	last_resort := false
+
+	encode: for inp < inp_end {
+		if last_resort {
+			last_resort = false
+
+			if input[inp] & 0x80 == 0x80 {
+				// Non-ASCII case
+				if out + 2 > out_end {
+					return out, .Output_Too_Short
+				}
+
+				// Put in a sentinel byte
+				output[out] = 0x00
+				out += 1
+			} else {
+				// An ASCII byte
+				if out + 1 > out_end {
+					return out, .Output_Too_Short
+				}
+			}
+			output[out] = input[inp]
+			out, inp = out + 1, inp + 1
+		} else {
+			// Find the longest string of known successors
+			indices[0] = model.ids_by_character[input[inp]]
+			last_chr_index := indices[0]
+
+			if last_chr_index < 0 {
+				last_resort = true
+				continue encode
+			}
+
+			rest := inp_end - inp
+			n_consecutive: i8 = 1
+			for ; n_consecutive <= model.max_successor_n; n_consecutive += 1 {
+				if inp_end > 0 && int(n_consecutive) == rest {
+					break
+				}
+
+				current_index := model.ids_by_character[input[inp + int(n_consecutive)]]
+				if current_index < 0 { // '\0' is always -1
+					break
+				}
+
+				successor_index := model.successors_by_bigram[last_chr_index * i16(model.character_count) + current_index]
+				if successor_index < 0 {
+					break
+				}
+
+				indices[n_consecutive] = i16(successor_index)
+				last_chr_index = current_index
+			}
+
+			if n_consecutive < 2 {
+				last_resort = true
+				continue encode
+			}
+
+			pack_n := find_best_encoding(indices, n_consecutive)
+			if pack_n >= 0 {
+				if out + int(model.packs[pack_n].bytes_packed) > out_end {
+					return out, .Output_Too_Short
+				}
+
+				pack := model.packs[pack_n]
+				code := pack.word
+
+				for i := 0; i < int(pack.bytes_unpacked); i += 1 {
+					code |= u32(indices[i]) << pack.offsets[i]
+				}
+
+				// In the little-endian world, we need to swap what's in the register to match the memory representation.
+				when ODIN_ENDIAN == .Little {
+					code = intrinsics.byte_swap(code)
+				}
+				out_ptr := raw_data(output[out:])
+
+				switch pack.bytes_packed {
+				case 4:
+					intrinsics.unaligned_store(transmute(^u32)out_ptr, code)
+				case 2:
+					intrinsics.unaligned_store(transmute(^u16)out_ptr, u16(code))
+				case 1:
+					intrinsics.unaligned_store(transmute(^u8)out_ptr,  u8(code))
+				case:
+					return out, .Unknown_Compression_Method
+				}
+
+				out += int(pack.bytes_packed)
+				inp += int(pack.bytes_unpacked)
+			} else {
+				last_resort = true
+				continue encode
+			}
+		}
+	}
+	return out, nil
+}
+
+compress_string :: proc(input: string, model := DEFAULT_MODEL, allocator := context.allocator) -> (output: []u8, err: compress.Error) {
+	context.allocator = allocator
+
+	if len(input) == 0 {
+		return {}, .Stream_Too_Short
+	}
+
+	max_output_size := compress_bound(len(input))
+
+	buf: [dynamic]u8
+	resize(&buf, max_output_size) or_return
+
+	length, result := compress_string_to_buffer(input, buf[:])
+	resize(&buf, length) or_return
+	return buf[:length], result
+}
+compress :: proc{compress_string_to_buffer, compress_string}
@@ -1,4 +1,4 @@
-//+ignore
+//+build ignore
 package zlib

 /*
@@ -1,3 +1,4 @@
+//+vet !using-param
 package zlib

 /*
@@ -47,10 +48,10 @@ Options :: struct {
 	level: u8,
 }

-Error     :: compress.Error
-E_General :: compress.General_Error
-E_ZLIB    :: compress.ZLIB_Error
-E_Deflate :: compress.Deflate_Error
+Error         :: compress.Error
+General_Error :: compress.General_Error
+ZLIB_Error    :: compress.ZLIB_Error
+Deflate_Error :: compress.Deflate_Error

 DEFLATE_MAX_CHUNK_SIZE   :: 65535
 DEFLATE_MAX_LITERAL_SIZE :: 65535
@@ -111,9 +112,9 @@ ZFAST_MASK :: ((1 << ZFAST_BITS) - 1)
 */
 Huffman_Table :: struct {
 	fast:        [1 << ZFAST_BITS]u16,
-	firstcode:   [16]u16,
+	firstcode:   [17]u16,
 	maxcode:     [17]int,
-	firstsymbol: [16]u16,
+	firstsymbol: [17]u16,
 	size:        [288]u8,
 	value:       [288]u16,
 }
@@ -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
@@ -244,7 +237,7 @@ allocate_huffman_table :: proc(allocator := context.allocator) -> (z: ^Huffman_T
@(optimization_mode="speed")
 build_huffman :: proc(z: ^Huffman_Table, code_lengths: []u8) -> (err: Error) {
 	sizes:     [HUFFMAN_MAX_BITS+1]int
-	next_code: [HUFFMAN_MAX_BITS]int
+	next_code: [HUFFMAN_MAX_BITS+1]int

 	k := int(0)

@@ -256,21 +249,21 @@ build_huffman :: proc(z: ^Huffman_Table, code_lengths: []u8) -> (err: Error) {
 	}
 	sizes[0] = 0

-	for i in 1..<(HUFFMAN_MAX_BITS+1) {
+	for i in 1 ..< HUFFMAN_MAX_BITS {
 		if sizes[i] > (1 << uint(i)) {
-			return E_Deflate.Huffman_Bad_Sizes
+			return .Huffman_Bad_Sizes
 		}
 	}
 	code := int(0)

-	for i in 1..<HUFFMAN_MAX_BITS {
+	for i in 1 ..= HUFFMAN_MAX_BITS {
 		next_code[i]     = code
 		z.firstcode[i]   = u16(code)
 		z.firstsymbol[i] = u16(k)
 		code = code + sizes[i]
 		if sizes[i] != 0 {
 			if code - 1 >= (1 << u16(i)) {
-				return E_Deflate.Huffman_Bad_Code_Lengths
+				return .Huffman_Bad_Code_Lengths
 			}
 		}
 		z.maxcode[i] = code << (HUFFMAN_MAX_BITS - uint(i))
@@ -305,24 +298,24 @@ 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
 	}
 	if s >= 16 {
-		return 0, E_Deflate.Bad_Huffman_Code
+		return 0, .Bad_Huffman_Code
 	}
 	// code size is s, so:
 	b := (k >> (16-s)) - int(t.firstcode[s]) + int(t.firstsymbol[s])
 	if b >= size_of(t.size) {
-		return 0, E_Deflate.Bad_Huffman_Code
+		return 0, .Bad_Huffman_Code
 	}
 	if t.size[b] != s {
-		return 0, E_Deflate.Bad_Huffman_Code
+		return 0, .Bad_Huffman_Code
 	}

 	compress.consume_bits_lsb(z, s)
@@ -335,11 +328,11 @@ decode_huffman_slowpath :: proc(z: ^$C, t: ^Huffman_Table) -> (r: u16, err: Erro
 decode_huffman :: proc(z: ^$C, t: ^Huffman_Table) -> (r: u16, err: Error) #no_bounds_check {
 	if z.num_bits < 16 {
 		if z.num_bits > 63 {
-			return 0, E_ZLIB.Code_Buffer_Malformed
+			return 0, .Code_Buffer_Malformed
 		}
 		compress.refill_lsb(z)
 		if z.num_bits > 63 {
-			return 0, E_General.Stream_Too_Short
+			return 0, .Stream_Too_Short
 		}
 	}
 	#no_bounds_check b := t.fast[z.code_buffer & ZFAST_MASK]
@@ -361,7 +354,7 @@ parse_huffman_block :: proc(z: ^$C, z_repeat, z_offset: ^Huffman_Table) -> (err:
 		if value < 256 {
 			e := write_byte(z, u8(value))
 			if e != .None {
-				return E_General.Output_Too_Short
+				return .Output_Too_Short
 			}
 		} else {
 			if value == 256 {
@@ -377,7 +370,7 @@ parse_huffman_block :: proc(z: ^$C, z_repeat, z_offset: ^Huffman_Table) -> (err:

 			value, e = decode_huffman(z, z_offset)
 			if e != nil {
-				return E_Deflate.Bad_Huffman_Code
+				return .Bad_Huffman_Code
 			}

 			distance := Z_DIST_BASE[value]
@@ -387,7 +380,7 @@ parse_huffman_block :: proc(z: ^$C, z_repeat, z_offset: ^Huffman_Table) -> (err:

 			if z.bytes_written < i64(distance) {
 				// Distance is longer than we've decoded so far.
-				return E_Deflate.Bad_Distance
+				return .Bad_Distance
 			}

 			/*
@@ -405,14 +398,14 @@ parse_huffman_block :: proc(z: ^$C, z_repeat, z_offset: ^Huffman_Table) -> (err:
 					c := z.output.buf[z.bytes_written - i64(distance)]
 					e := repl_byte(z, length, c)
 					if e != .None {
-						return E_General.Output_Too_Short
+						return .Output_Too_Short
 					}
 				}
 			} else {
 				if length > 0 {
 					e := repl_bytes(z, length, distance)
 					if e != .None {
-						return E_General.Output_Too_Short
+						return .Output_Too_Short
 					}
 				}
 			}
@@ -432,25 +425,25 @@ inflate_from_context :: proc(using ctx: ^compress.Context_Memory_Input, raw := f
 	if !raw {
 		size, size_err := compress.input_size(ctx)
 		if size < 6 || size_err != nil {
-			return E_General.Stream_Too_Short
+			return .Stream_Too_Short
 		}

 		cmf, _ := compress.read_u8(ctx)

 		method := Compression_Method(cmf & 0xf)
 		if method != .DEFLATE {
-			return E_General.Unknown_Compression_Method
+			return .Unknown_Compression_Method
 		}

 		if cinfo := (cmf >> 4) & 0xf; cinfo > 7 {
-			return E_ZLIB.Unsupported_Window_Size
+			return .Unsupported_Window_Size
 		}
 		flg, _ := compress.read_u8(ctx)

 		fcheck := flg & 0x1f
 		fcheck_computed := (cmf << 8 | flg) & 0x1f
 		if fcheck != fcheck_computed {
-			return E_General.Checksum_Failed
+			return .Checksum_Failed
 		}

 		/*
@@ -458,7 +451,7 @@ inflate_from_context :: proc(using ctx: ^compress.Context_Memory_Input, raw := f
 			They're application specific and PNG doesn't use them.
 		*/
 		if fdict := (flg >> 5) & 1; fdict != 0 {
-			return E_ZLIB.FDICT_Unsupported
+			return .FDICT_Unsupported
 		}

 		// flevel  := Compression_Level((flg >> 6) & 3);
@@ -471,7 +464,7 @@ inflate_from_context :: proc(using ctx: ^compress.Context_Memory_Input, raw := f
 	}

 	// Parse ZLIB stream without header.
-	inflate_raw(z=ctx, expected_output_size=expected_output_size) or_return
+	inflate_raw(ctx, expected_output_size=expected_output_size) or_return

 	if !raw {
 		compress.discard_to_next_byte_lsb(ctx)
@@ -485,7 +478,7 @@ inflate_from_context :: proc(using ctx: ^compress.Context_Memory_Input, raw := f
 		output_hash := hash.adler32(ctx.output.buf[:])

 		if output_hash != u32(adler) {
-			return E_General.Checksum_Failed
+			return .Checksum_Failed
 		}
 	}
 	return nil
@@ -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 {
@@ -538,23 +531,24 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 		final = compress.read_bits_lsb(z, 1)
 		type  = compress.read_bits_lsb(z, 2)

-		// fmt.printf("Final: %v | Type: %v\n", final, type);
+		// fmt.printf("Final: %v | Type: %v\n", final, type)

 		switch type {
 		case 0:
+			// fmt.printf("Method 0: STORED\n")
 			// Uncompressed block

 			// Discard bits until next byte boundary
 			compress.discard_to_next_byte_lsb(z)

-			uncompressed_len := i16(compress.read_bits_lsb(z, 16))
-			length_check     := i16(compress.read_bits_lsb(z, 16))
+			uncompressed_len := u16(compress.read_bits_lsb(z, 16))
+			length_check     := u16(compress.read_bits_lsb(z, 16))

-			// fmt.printf("LEN: %v, ~LEN: %v, NLEN: %v, ~NLEN: %v\n", uncompressed_len, ~uncompressed_len, length_check, ~length_check);
+			// fmt.printf("LEN: %v, ~LEN: %v, NLEN: %v, ~NLEN: %v\n", uncompressed_len, ~uncompressed_len, length_check, ~length_check)


 			if ~uncompressed_len != length_check {
-				return E_Deflate.Len_Nlen_Mismatch
+				return .Len_Nlen_Mismatch
 			}

 			/*
@@ -567,10 +561,12 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 				write_byte(z, u8(lit))
 				uncompressed_len -= 1
 			}
+			assert(uncompressed_len == 0)
+
 		case 3:
-			return E_Deflate.BType_3
+			return .BType_3
 		case:
-			// log.debugf("Err: %v | Final: %v | Type: %v\n", err, final, type);
+			// fmt.printf("Err: %v | Final: %v | Type: %v\n", err, final, type)
 			if type == 1 {
 				// Use fixed code lengths.
 				build_huffman(z_repeat, Z_FIXED_LENGTH[:]) or_return
@@ -601,7 +597,7 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 					c = decode_huffman(z, codelength_ht) or_return

 					if c < 0 || c >= 19 {
-						return E_Deflate.Huffman_Bad_Code_Lengths
+						return .Huffman_Bad_Code_Lengths
 					}
 					if c < 16 {
 						lencodes[n] = u8(c)
@@ -613,7 +609,7 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 						case 16:
 							c = u16(compress.read_bits_no_refill_lsb(z, 2) + 3)
 							if n == 0 {
-								return E_Deflate.Huffman_Bad_Code_Lengths
+								return .Huffman_Bad_Code_Lengths
 							}
 							fill = lencodes[n - 1]
 						case 17:
@@ -621,11 +617,11 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 						case 18:
 							c = u16(compress.read_bits_no_refill_lsb(z, 7) + 11)
 						case:
-								return E_Deflate.Huffman_Bad_Code_Lengths
+								return .Huffman_Bad_Code_Lengths
 						}

 						if ntot - n < u32(c) {
-							return E_Deflate.Huffman_Bad_Code_Lengths
+							return .Huffman_Bad_Code_Lengths
 						}

 						nc := n + u32(c)
@@ -636,7 +632,7 @@ inflate_raw :: proc(z: ^$C, expected_output_size := -1, allocator := context.all
 				}

 				if n != ntot {
-					return E_Deflate.Huffman_Bad_Code_Lengths
+					return .Huffman_Bad_Code_Lengths
 				}

 				build_huffman(z_repeat, lencodes[:hlit])     or_return
@@ -650,7 +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
@@ -662,7 +658,7 @@ inflate_from_byte_array :: proc(input: []u8, buf: ^bytes.Buffer, raw := false, e
 	ctx.input_data = input
 	ctx.output = buf

-	return inflate_from_context(ctx=&ctx, raw=raw, expected_output_size=expected_output_size)
+	return inflate_from_context(&ctx, raw=raw, expected_output_size=expected_output_size)
 }

 inflate_from_byte_array_raw :: proc(input: []u8, buf: ^bytes.Buffer, raw := false, expected_output_size := -1) -> (err: Error) {
@@ -671,7 +667,7 @@ inflate_from_byte_array_raw :: proc(input: []u8, buf: ^bytes.Buffer, raw := fals
 	ctx.input_data = input
 	ctx.output = buf

-	return inflate_raw(z=&ctx, expected_output_size=expected_output_size)
+	return inflate_raw(&ctx, expected_output_size=expected_output_size)
 }

-inflate :: proc{inflate_from_context, inflate_from_byte_array};
+inflate :: proc{inflate_from_context, inflate_from_byte_array}
@@ -1,216 +0,0 @@
-package container
-
-import "core:mem"
-import "core:runtime"
-
-Array :: struct($T: typeid) {
-	data:      ^T,
-	len:       int,
-	cap:       int,
-	allocator: mem.Allocator,
-}
-
-ARRAY_DEFAULT_CAPACITY :: 16
-
-/*
-array_init :: proc {
-	array_init_none,
-	array_init_len,
-	array_init_len_cap,
-}
-array_init
-array_delete
-array_len
-array_cap
-array_space
-array_slice
-array_get
-array_get_ptr
-array_set
-array_reserve
-array_resize
-array_push = array_append :: proc{
-	array_push_back,
-	array_push_back_elems,
-}
-array_push_front
-array_pop_back
-array_pop_front
-array_consume
-array_trim
-array_clear
-array_clone
-array_set_capacity
-array_grow
-*/
-
-
-array_init_none :: proc(a: ^$A/Array, allocator := context.allocator) {
-	array_init_len_cap(a, 0, ARRAY_DEFAULT_CAPACITY, allocator)
-}
-array_init_len :: proc(a: ^$A/Array, len: int, allocator := context.allocator) {
-	array_init_len_cap(a, len, len, allocator)
-}
-array_init_len_cap :: proc(a: ^$A/Array($T), len: int, cap: int, allocator := context.allocator) {
-	a.allocator = allocator
-	a.data = (^T)(mem.alloc(size_of(T)*cap, align_of(T), a.allocator))
-	a.len = len
-	a.cap = cap
-}
-
-array_init :: proc{array_init_none, array_init_len, array_init_len_cap}
-
-array_delete :: proc(a: $A/Array) {
-	mem.free(a.data, a.allocator)
-}
-
-array_len :: proc(a: $A/Array) -> int {
-	return a.len
-}
-
-array_cap :: proc(a: $A/Array) -> int {
-	return a.cap
-}
-
-array_space :: proc(a: $A/Array) -> int {
-	return a.cap - a.len
-}
-
-array_slice :: proc(a: $A/Array($T)) -> []T {
-	s := mem.Raw_Slice{a.data, a.len}
-	return transmute([]T)s
-}
-
-array_cap_slice :: proc(a: $A/Array($T)) -> []T {
-	s := mem.Raw_Slice{a.data, a.cap}
-	return transmute([]T)s
-}
-
-array_get :: proc(a: $A/Array($T), index: int, loc := #caller_location) -> T {
-	runtime.bounds_check_error_loc(loc, index, array_len(a))
-	return (^T)(uintptr(a.data) + size_of(T)*uintptr(index))^
-}
-array_get_ptr :: proc(a: $A/Array($T), index: int, loc := #caller_location) -> ^T {
-	runtime.bounds_check_error_loc(loc, index, array_len(a))
-	return (^T)(uintptr(a.data) + size_of(T)*uintptr(index))
-}
-
-array_set :: proc(a: ^$A/Array($T), index: int, item: T, loc := #caller_location)  {
-	runtime.bounds_check_error_loc(loc, index, array_len(a^))
-	(^T)(uintptr(a.data) + size_of(T)*uintptr(index))^ = item
-}
-
-
-array_reserve :: proc(a: ^$A/Array, capacity: int) {
-	if capacity > a.len {
-		array_set_capacity(a, capacity)
-	}
-}
-
-array_resize :: proc(a: ^$A/Array, length: int) {
-	if length > a.len {
-		array_set_capacity(a, length)
-	}
-	a.len = length
-}
-
-
-
-array_push_back :: proc(a: ^$A/Array($T), item: T) {
-	if array_space(a^) == 0 {
-		array_grow(a)
-	}
-
-	a.len += 1
-	array_set(a, a.len-1, item)
-}
-
-array_push_front :: proc(a: ^$A/Array($T), item: T) {
-	if array_space(a^) == 0 {
-		array_grow(a)
-	}
-
-	a.len += 1
-	data := array_slice(a^)
-	copy(data[1:], data[:])
-	data[0] = item
-}
-
-array_pop_back :: proc(a: ^$A/Array($T), loc := #caller_location) -> T {
-	assert(condition=a.len > 0, loc=loc)
-	item := array_get(a^, a.len-1)
-	a.len -= 1
-	return item
-}
-
-array_pop_front :: proc(a: ^$A/Array($T), loc := #caller_location) -> T {
-	assert(condition=a.len > 0, loc=loc)
-	item := array_get(a^, 0)
-	s := array_slice(a^)
-	copy(s[:], s[1:])
-	a.len -= 1
-	return item
-}
-
-
-array_consume :: proc(a: ^$A/Array($T), count: int, loc := #caller_location) {
-	assert(condition=a.len >= count, loc=loc)
-	a.len -= count
-}
-
-
-array_trim :: proc(a: ^$A/Array($T)) {
-	array_set_capacity(a, a.len)
-}
-
-array_clear :: proc(a: ^$A/Array($T)) {
-	array_resize(a, 0)
-}
-
-array_clone :: proc(a: $A/Array($T), allocator := context.allocator) -> A {
-	res: A
-	array_init(&res, array_len(a), array_len(a), allocator)
-	copy(array_slice(res), array_slice(a))
-	return res
-}
-
-array_push_back_elems :: proc(a: ^$A/Array($T), items: ..T) {
-	if array_space(a^) < len(items) {
-		array_grow(a, a.len + len(items))
-	}
-	offset := a.len
-	data := array_cap_slice(a^)
-	n := copy(data[a.len:], items)
-	a.len += n
-}
-
-array_push   :: proc{array_push_back, array_push_back_elems}
-array_append :: proc{array_push_back, array_push_back_elems}
-
-array_set_capacity :: proc(a: ^$A/Array($T), new_capacity: int) {
-	if new_capacity == a.cap {
-		return
-	}
-
-	if new_capacity < a.len {
-		array_resize(a, new_capacity)
-	}
-
-	new_data: ^T
-	if new_capacity > 0 {
-		if a.allocator.procedure == nil {
-			a.allocator = context.allocator
-		}
-		new_data = (^T)(mem.alloc(size_of(T)*new_capacity, align_of(T), a.allocator))
-		if new_data != nil {
-			mem.copy(new_data, a.data, size_of(T)*a.len)
-		}
-	}
-	mem.free(a.data, a.allocator)
-	a.data = new_data
-	a.cap = new_capacity
-}
-array_grow :: proc(a: ^$A/Array, min_capacity: int = 0) {
-	new_capacity := max(array_len(a^)*2 + 8, min_capacity)
-	array_set_capacity(a, new_capacity)
-}
@@ -0,0 +1,325 @@
+package dynamic_bit_array
+
+import "core:intrinsics"
+import "core:mem"
+
+/*
+	Note that these constants are dependent on the backing being a u64.
+*/
+@(private="file")
+INDEX_SHIFT :: 6
+
+@(private="file")
+INDEX_MASK  :: 63
+
+@(private="file")
+NUM_BITS :: 64
+
+Bit_Array :: struct {
+	bits:         [dynamic]u64,
+	bias:         int,
+	max_index:    int,
+	free_pointer: bool,
+}
+
+Bit_Array_Iterator :: struct {
+	array:    ^Bit_Array,
+	word_idx: int,
+	bit_idx:  uint,
+}
+/*
+Wraps a `Bit_Array` into an Iterator
+
+Inputs:
+- ba: Pointer to the Bit_Array
+
+Returns:
+- it: Iterator struct
+*/
+make_iterator :: proc (ba: ^Bit_Array) -> (it: Bit_Array_Iterator) {
+	return Bit_Array_Iterator { array = ba }
+}
+/*
+Returns the next bit, including its set-state. ok=false once exhausted
+
+Inputs:
+- it: The iterator that holds the state.
+
+Returns:
+- set: `true` if the bit at `index` is set.
+- index: The next bit of the Bit_Array referenced by `it`.
+- ok: `true` if the iterator can continue, `false` if the iterator is done
+*/
+iterate_by_all :: proc (it: ^Bit_Array_Iterator) -> (set: bool, index: int, ok: bool) {
+	index = it.word_idx * NUM_BITS + int(it.bit_idx) + it.array.bias
+	if index > it.array.max_index { return false, 0, false }
+
+	word := it.array.bits[it.word_idx] if len(it.array.bits) > it.word_idx else 0
+	set = (word >> it.bit_idx & 1) == 1
+
+	it.bit_idx += 1
+	if it.bit_idx >= NUM_BITS {
+		it.bit_idx = 0
+		it.word_idx += 1
+	}
+
+	return set, index, true
+}
+/*
+Returns the next Set Bit, for example if `0b1010`, then the iterator will return index={1, 3} over two calls.
+
+Inputs:
+- it: The iterator that holds the state.
+
+Returns:
+- index: The next *set* bit of the Bit_Array referenced by `it`.
+- ok: `true` if the iterator can continue, `false` if the iterator is done
+*/
+iterate_by_set :: proc (it: ^Bit_Array_Iterator) -> (index: int, ok: bool) {
+	return iterate_internal_(it, true)
+}
+/*
+Returns the next Unset Bit, for example if `0b1010`, then the iterator will return index={0, 2} over two calls.
+
+Inputs:
+- it: The iterator that holds the state.
+
+Returns:
+- index: The next *unset* bit of the Bit_Array referenced by `it`.
+- ok: `true` if the iterator can continue, `false` if the iterator is done
+*/
+iterate_by_unset:: proc (it: ^Bit_Array_Iterator) -> (index: int, ok: bool) {
+	return iterate_internal_(it, false)
+}
+/*
+Iterates through set/unset bits
+
+*Private*
+
+Inputs:
+- it: The iterator that holds the state.
+- ITERATE_SET_BITS: `true` for returning only set bits, false for returning only unset bits
+
+Returns:
+- index: The next *unset* bit of the Bit_Array referenced by `it`.
+- ok: `true` if the iterator can continue, `false` if the iterator is done
+*/
+@(private="file")
+iterate_internal_ :: proc (it: ^Bit_Array_Iterator, $ITERATE_SET_BITS: bool) -> (index: int, ok: bool) {
+	word := it.array.bits[it.word_idx] if len(it.array.bits) > it.word_idx else 0
+	when ! ITERATE_SET_BITS { word = ~word }
+
+	// If the word is empty or we have already gone over all the bits in it,
+	// b.bit_idx is greater than the index of any set bit in the word,
+	// meaning that word >> b.bit_idx == 0.
+	for it.word_idx < len(it.array.bits) && word >> it.bit_idx == 0 {
+		it.word_idx += 1
+		it.bit_idx = 0
+		word = it.array.bits[it.word_idx] if len(it.array.bits) > it.word_idx else 0
+		when ! ITERATE_SET_BITS { word = ~word }
+	}
+
+	// If we are iterating the set bits, reaching the end of the array means we have no more bits to check
+	when ITERATE_SET_BITS {
+		if it.word_idx >= len(it.array.bits) {
+			return 0, false
+		}
+	}
+
+	// Reaching here means that the word has some set bits
+	it.bit_idx += uint(intrinsics.count_trailing_zeros(word >> it.bit_idx))
+	index = it.word_idx * NUM_BITS + int(it.bit_idx) + it.array.bias
+
+	it.bit_idx += 1
+	if it.bit_idx >= NUM_BITS {
+		it.bit_idx = 0
+		it.word_idx += 1
+	}
+	return index, index <= it.array.max_index
+}
+/*
+Gets the state of a bit in the bit-array
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+
+Returns:
+- res: `true` if the bit at `index` is set.
+- ok: Whether the index was valid. Returns `false` if the index is smaller than the bias.
+*/
+get :: proc(ba: ^Bit_Array, #any_int index: uint) -> (res: bool, ok: bool) #optional_ok {
+	idx := int(index) - ba.bias
+
+	if ba == nil || int(index) < ba.bias { return false, false }
+
+	leg_index := idx >> INDEX_SHIFT
+	bit_index := idx &  INDEX_MASK
+
+	/*
+		If we `get` a bit that doesn't fit in the Bit Array, it's naturally `false`.
+		This early-out prevents unnecessary resizing.
+	*/
+	if leg_index + 1 > len(ba.bits) { return false, true }
+
+	val := u64(1 << uint(bit_index))
+	res = ba.bits[leg_index] & val == val
+
+	return res, true
+}
+/*
+Gets the state of a bit in the bit-array
+
+*Bypasses all Checks*
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+
+Returns:
+- `true` if bit is set
+*/
+unsafe_get :: #force_inline proc(ba: ^Bit_Array, #any_int index: uint) -> bool #no_bounds_check {
+	return bool((ba.bits[index >> INDEX_SHIFT] >> uint(index & INDEX_MASK)) & 1)
+}
+/*
+Sets the state of a bit in the bit-array
+
+*Conditionally Allocates (Resizes backing data when `index > len(ba.bits)`)*
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+- set_to: `true` sets the bit on, `false` to turn it off
+- allocator: (default is context.allocator)
+
+Returns:
+- ok: Whether the set was successful, `false` on allocation failure or bad index
+*/
+set :: proc(ba: ^Bit_Array, #any_int index: uint, set_to: bool = true, allocator := context.allocator) -> (ok: bool) {
+
+	idx := int(index) - ba.bias
+
+	if ba == nil || int(index) < ba.bias { return false }
+	context.allocator = allocator
+
+	leg_index := idx >> INDEX_SHIFT
+	bit_index := idx &  INDEX_MASK
+
+	resize_if_needed(ba, leg_index) or_return
+
+	ba.max_index = max(idx, ba.max_index)
+
+	if set_to{ ba.bits[leg_index] |= 1 << uint(bit_index) }
+	else { ba.bits[leg_index] &= ~(1 << uint(bit_index)) }
+
+	return true
+}
+/*
+Sets the state of a bit in the bit-array
+
+*Bypasses all checks*
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+*/
+unsafe_set :: proc(ba: ^Bit_Array, bit: int) #no_bounds_check {
+	ba.bits[bit >> INDEX_SHIFT] |= 1 << uint(bit & INDEX_MASK)
+}
+/*
+Unsets the state of a bit in the bit-array. (Convienence wrapper for `set`)
+
+*Conditionally Allocates (Resizes backing data when `index > len(ba.bits)`)*
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+- allocator: (default is context.allocator)
+
+Returns:
+- ok: Whether the unset was successful, `false` on allocation failure or bad index
+*/
+unset :: #force_inline proc(ba: ^Bit_Array, #any_int index: uint, allocator := context.allocator) -> (ok: bool) {
+	return set(ba, index, false, allocator)
+}
+/*
+Unsets the state of a bit in the bit-array
+
+*Bypasses all Checks*
+
+Inputs:
+- ba: Pointer to the Bit_Array
+- index: Which bit in the array
+*/
+unsafe_unset :: proc(b: ^Bit_Array, bit: int) #no_bounds_check {
+	b.bits[bit >> INDEX_SHIFT] &= ~(1 << uint(bit & INDEX_MASK))
+}
+/*
+A helper function to create a Bit Array with optional bias, in case your smallest index is non-zero (including negative).
+
+*Allocates (`new(Bit_Array) & make(ba.bits)`)*
+
+Inputs:
+- max_index: maximum starting index
+- min_index: minimum starting index (used as a bias)
+- allocator: (default is context.allocator)
+
+Returns:
+- ba: Allocates a bit_Array, backing data is set to `max-min / 64` indices, rounded up (eg 65 - 0 allocates for [2]u64).
+*/
+create :: proc(max_index: int, min_index: int = 0, allocator := context.allocator) -> (res: ^Bit_Array, ok: bool) #optional_ok {
+	context.allocator = allocator
+	size_in_bits := max_index - min_index
+
+	if size_in_bits < 1 { return {}, false }
+
+	legs := size_in_bits >> INDEX_SHIFT
+	if size_in_bits & INDEX_MASK > 0 {legs+=1}
+	bits, err := make([dynamic]u64, legs)
+	ok = err == mem.Allocator_Error.None
+	res = new(Bit_Array)
+	res.bits         = bits
+	res.bias         = min_index
+	res.max_index    = max_index
+	res.free_pointer = true
+	return
+}
+/*
+Sets all values in the Bit_Array to zero.
+
+Inputs:
+- ba: The target Bit_Array
+*/
+clear :: proc(ba: ^Bit_Array) {
+	if ba == nil { return }
+	mem.zero_slice(ba.bits[:])
+}
+/*
+Deallocates the Bit_Array and its backing storage
+
+Inputs:
+- ba: The target Bit_Array
+*/
+destroy :: proc(ba: ^Bit_Array) {
+	if ba == nil { return }
+	delete(ba.bits)
+	if ba.free_pointer { // Only free if this Bit_Array was created using `create`, not when on the stack.
+		free(ba)
+	}
+}
+/*
+	Resizes the Bit Array. For internal use. Provisions needed capacity+1
+	If you want to reserve the memory for a given-sized Bit Array up front, you can use `create`.
+*/
+@(private="file")
+resize_if_needed :: proc(ba: ^Bit_Array, legs: int, allocator := context.allocator) -> (ok: bool) {
+	if ba == nil { return false }
+
+	context.allocator = allocator
+
+	if legs + 1 > len(ba.bits) {
+		resize(&ba.bits, legs + 1)
+	}
+	return len(ba.bits) > legs
+}
@@ -0,0 +1,53 @@
+package dynamic_bit_array
+
+/*
+	The Bit Array can be used in several ways:
+
+	-- By default you don't need to instantiate a Bit Array:
+
+		package test
+
+		import "core:fmt"
+		import "core:container/bit_array"
+
+		main :: proc() {
+			using bit_array
+
+			bits: Bit_Array
+
+			// returns `true`
+			fmt.println(set(&bits, 42))
+
+			// returns `false`, `false`, because this Bit Array wasn't created to allow negative indices.
+			was_set, was_retrieved := get(&bits, -1)
+			fmt.println(was_set, was_retrieved) 
+			destroy(&bits)
+		}
+
+	-- A Bit Array can optionally allow for negative indices, if the mininum value was given during creation:
+
+		package test
+
+		import "core:fmt"
+		import "core:container/bit_array"
+
+		main :: proc() {
+			Foo :: enum int {
+				Negative_Test = -42,
+				Bar           = 420,
+				Leaves        = 69105,
+			}
+
+			using bit_array
+
+			bits := create(int(max(Foo)), int(min(Foo)))
+			defer destroy(bits)
+
+			fmt.printf("Set(Bar):           %v\n",     set(bits, Foo.Bar))
+			fmt.printf("Get(Bar):           %v, %v\n", get(bits, Foo.Bar))
+			fmt.printf("Set(Negative_Test): %v\n",     set(bits, Foo.Negative_Test))
+			fmt.printf("Get(Leaves):        %v, %v\n", get(bits, Foo.Leaves))
+			fmt.printf("Get(Negative_Test): %v, %v\n", get(bits, Foo.Negative_Test))
+			fmt.printf("Freed.\n")
+		}
+*/
@@ -1,80 +0,0 @@
-package container
-
-import "core:mem"
-
-Bloom_Hash_Proc :: #type proc(data: []byte) -> u32
-
-Bloom_Hash :: struct {
-	hash_proc: Bloom_Hash_Proc,
-	next:     ^Bloom_Hash,
-}
-
-Bloom_Filter :: struct {
-	allocator: mem.Allocator,
-	hash:      ^Bloom_Hash,
-	bits:      []byte,
-}
-
-bloom_filter_init :: proc(b: ^Bloom_Filter, size: int, allocator := context.allocator) {
-	b.allocator = allocator
-	b.bits = make([]byte, size, allocator)
-}
-
-bloom_filter_destroy :: proc(b: ^Bloom_Filter) {
-	context.allocator = b.allocator
-	delete(b.bits)
-	for b.hash != nil {
-		hash := b.hash
-		b.hash = b.hash.next
-		free(hash)
-	}
-}
-
-bloom_filter_add_hash_proc :: proc(b: ^Bloom_Filter, hash_proc: Bloom_Hash_Proc) {
-	context.allocator = b.allocator
-	h := new(Bloom_Hash)
-	h.hash_proc = hash_proc
-
-	head := &b.hash
-	for head^ != nil {
-		head = &(head^.next)
-	}
-	head^ = h
-}
-
-bloom_filter_add :: proc(b: ^Bloom_Filter, item: []byte) {
-	#no_bounds_check for h := b.hash; h != nil; h = h.next {
-		hash := h.hash_proc(item)
-		hash %= u32(len(b.bits) * 8)
-		b.bits[hash >> 3] |= 1 << (hash & 3)
-	}
-}
-
-bloom_filter_add_string :: proc(b: ^Bloom_Filter, item: string) {
-	bloom_filter_add(b, transmute([]byte)item)
-}
-
-bloom_filter_add_raw :: proc(b: ^Bloom_Filter, data: rawptr, size: int) {
-	item := mem.slice_ptr((^byte)(data), size)
-	bloom_filter_add(b, item)
-}
-
-bloom_filter_test :: proc(b: ^Bloom_Filter, item: []byte) -> bool {
-	#no_bounds_check for h := b.hash; h != nil; h = h.next {
-		hash := h.hash_proc(item)
-		hash %= u32(len(b.bits) * 8)
-		if (b.bits[hash >> 3] & (1 << (hash & 3)) == 0) {
-			return false
-		}
-	}
-	return true
-}
-
-bloom_filter_test_string :: proc(b: ^Bloom_Filter, item: string) -> bool {
-	return bloom_filter_test(b, transmute([]byte)item)
-}
-
-bloom_filter_test_raw :: proc(b: ^Bloom_Filter, data: rawptr, size: int) -> bool {
-	item := mem.slice_ptr((^byte)(data), size)
-	return bloom_filter_test(b, item)
-}
@@ -0,0 +1,195 @@
+package container_intrusive_list
+
+import "core:intrinsics"
+
+// An intrusive doubly-linked list
+//
+// As this is an intrusive container, a `Node` must be embedded in your own
+// structure which is conventionally called a "link". The use of `push_front`
+// and `push_back` take the address of this node. Retrieving the data
+// associated with the node requires finding the relative offset of the node
+// of the parent structure. The parent type and field name are given to
+// `iterator_*` procedures, or to the built-in `container_of` procedure.
+//
+// This data structure is two-pointers in size:
+// 	8 bytes on 32-bit platforms and 16 bytes on 64-bit platforms
+List :: struct {
+	head: ^Node,
+	tail: ^Node,
+}
+
+
+Node :: struct {
+	prev, next: ^Node,
+}
+
+push_front :: proc "contextless" (list: ^List, node: ^Node) {
+	if list.head != nil {
+		list.head.prev = node
+		node.prev, node.next = nil, list.head
+		list.head = node
+	} else {
+		list.head, list.tail = node, node
+		node.prev, node.next = nil, nil
+	}
+}
+
+push_back :: proc "contextless" (list: ^List, node: ^Node) {
+	if list.tail != nil {
+		list.tail.next = node
+		node.prev, node.next = list.tail, nil
+		list.tail = node
+	} else {
+		list.head, list.tail = node, node
+		node.prev, node.next = nil, nil
+	}
+}
+
+remove :: proc "contextless" (list: ^List, node: ^Node) {
+	if node != nil {
+		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
+		}
+	}
+}
+
+remove_by_proc :: proc(list: ^List, to_erase: proc(^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
+	}
+}
+
+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 "contextless" (list: ^List) -> bool {
+	return list.head == nil
+}
+
+pop_front :: proc "contextless" (list: ^List) -> ^Node {
+	link := list.head
+	if link == nil {
+		return nil
+	}
+	if link.next != nil {
+		link.next.prev = link.prev
+	}
+	if link.prev != nil {
+		link.prev.next = link.next
+	}
+	if link == list.head {
+		list.head = link.next
+	}
+	if link == list.tail {
+		list.tail = link.prev
+	}
+	return link
+
+}
+pop_back :: proc "contextless" (list: ^List) -> ^Node {
+	link := list.tail
+	if link == nil {
+		return nil
+	}
+	if link.next != nil {
+		link.next.prev = link.prev
+	}
+	if link.prev != nil {
+		link.prev.next = link.next
+	}
+	if link == list.head {
+		list.head = link.next
+	}
+	if link == list.tail {
+		list.tail = link.prev
+	}
+	return link
+}
+
+
+Iterator :: struct($T: typeid) {
+	curr:   ^Node,
+	offset: uintptr,
+}
+
+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 "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 "contextless" (node: ^Node, $T: typeid, $field_name: string) -> Iterator(T)
+	where intrinsics.type_has_field(T, field_name),
+	      intrinsics.type_field_type(T, field_name) == Node {
+	return {node, offset_of_by_string(T, field_name)}
+}
+
+iterate_next :: proc "contextless" (it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
+	node := it.curr
+	if node == nil {
+		return nil, false
+	}
+	it.curr = node.next
+
+	return (^T)(uintptr(node) - it.offset), true
+}
+
+iterate_prev :: proc "contextless" (it: ^Iterator($T)) -> (ptr: ^T, ok: bool) {
+	node := it.curr
+	if node == nil {
+		return nil, false
+	}
+	it.curr = node.prev
+
+	return (^T)(uintptr(node) - it.offset), true
+}
@@ -0,0 +1,201 @@
+package container_lru
+
+import "core:runtime"
+import "core:intrinsics"
+_ :: runtime
+_ :: intrinsics
+
+Node :: struct($Key, $Value: typeid) where intrinsics.type_is_valid_map_key(Key) {
+	prev, next: ^Node(Key, Value),
+	key:   Key,
+	value: Value,
+}
+
+// Cache is an LRU cache. It automatically removes entries as new entries are
+// added if the capacity is reached. Entries are removed based on how recently
+// they were used where the oldest entries are removed first.
+Cache :: struct($Key, $Value: typeid) where intrinsics.type_is_valid_map_key(Key) {
+	head: ^Node(Key, Value),
+	tail: ^Node(Key, Value),
+
+	entries: map[Key]^Node(Key, Value),
+
+	count:    int,
+	capacity: int,
+
+	node_allocator: runtime.Allocator,
+
+	on_remove: proc(key: Key, value: Value, user_data: rawptr),
+	on_remove_user_data: rawptr,
+}
+
+// init initializes a Cache
+init :: proc(c: ^$C/Cache($Key, $Value), capacity: int, entries_allocator := context.allocator, node_allocator := context.allocator) {
+	c.entries.allocator = entries_allocator
+	c.node_allocator = node_allocator
+	c.capacity = capacity
+}
+
+// destroy deinitializes a Cachem
+destroy :: proc(c: ^$C/Cache($Key, $Value), call_on_remove: bool) {
+	clear(c, call_on_remove)
+	delete(c.entries)
+}
+
+// clear the contents of a Cache
+clear :: proc(c: ^$C/Cache($Key, $Value), call_on_remove: bool) {
+	for _, node in c.entries {
+		if call_on_remove {
+			_call_on_remove(c, node)
+		}
+		free(node, c.node_allocator)
+	}
+	runtime.clear(&c.entries)
+	c.head = nil
+	c.tail = nil
+	c.count = 0
+}
+
+// set the given key value pair. This operation updates the recent usage of the item.
+set :: proc(c: ^$C/Cache($Key, $Value), key: Key, value: Value) -> runtime.Allocator_Error {
+	if e, ok := c.entries[key]; ok {
+		e.value = value
+		_pop_node(c, e)
+		_push_front_node(c, e)
+		return nil
+	}
+
+	e : ^Node(Key, Value) = nil
+	assert(c.count <= c.capacity)
+	if c.count == c.capacity {
+		e = c.tail
+		_remove_node(c, e)
+	}
+	else {
+		c.count += 1
+		e = new(Node(Key, Value), c.node_allocator) or_return
+	}
+
+	e.key = key
+	e.value = value
+	_push_front_node(c, e)
+	c.entries[key] = e
+
+	return nil
+}
+
+// get a value from the cache from a given key. This operation updates the usage of the item.
+get :: proc(c: ^$C/Cache($Key, $Value), key: Key) -> (value: Value, ok: bool) #optional_ok {
+	e: ^Node(Key, Value)
+	e, ok = c.entries[key]
+	if !ok {
+		return
+	}
+	_pop_node(c, e)
+	_push_front_node(c, e)
+	return e.value, true
+}
+
+// get_ptr gets the pointer to a value the cache from a given key. This operation updates the usage of the item.
+get_ptr :: proc(c: ^$C/Cache($Key, $Value), key: Key) -> (value: ^Value, ok: bool) #optional_ok {
+	e: ^Node(Key, Value)
+	e, ok = c.entries[key]
+	if !ok {
+		return
+	}
+	_pop_node(c, e)
+	_push_front_node(c, e)
+	return &e.value, true
+}
+
+// peek gets the value from the cache from a given key without updating the recent usage.
+peek :: proc(c: ^$C/Cache($Key, $Value), key: Key) -> (value: Value, ok: bool) #optional_ok {
+	e: ^Node(Key, Value)
+	e, ok = c.entries[key]
+	if !ok {
+		return
+	}
+	return e.value, true
+}
+
+// exists checks for the existence of a value from a given key without updating the recent usage.
+exists :: proc(c: ^$C/Cache($Key, $Value), key: Key) -> bool {
+	return key in c.entries
+}
+
+// remove removes an item from the cache.
+remove :: proc(c: ^$C/Cache($Key, $Value), key: Key) -> bool {
+	e, ok := c.entries[key]
+	if !ok {
+		return false
+	}
+	_remove_node(c, e)
+	free(node, c.node_allocator)
+	c.count -= 1
+	return true
+}
+
+
+@(private)
+_remove_node :: proc(c: ^$C/Cache($Key, $Value), node: ^Node(Key, Value)) {
+	if c.head == node {
+		c.head = node.next
+	}
+	if c.tail == node {
+		c.tail = node.prev
+	}
+	if node.prev != nil {
+		node.prev.next = node.next
+	}
+	if node.next != nil {
+		node.next.prev = node.prev
+	}
+	node.prev = nil
+	node.next = nil
+
+	delete_key(&c.entries, node.key)
+
+	_call_on_remove(c, node)
+}
+
+@(private)
+_call_on_remove :: proc(c: ^$C/Cache($Key, $Value), node: ^Node(Key, Value)) {
+	if c.on_remove != nil {
+		c.on_remove(node.key, node.value, c.on_remove_user_data)
+	}
+}
+
+@(private)
+_push_front_node :: proc(c: ^$C/Cache($Key, $Value), e: ^Node(Key, Value)) {
+	if c.head != nil {
+		e.next = c.head
+		e.next.prev = e
+	}
+	c.head = e
+	if c.tail == nil {
+		c.tail = e
+	}
+	e.prev = nil
+}
+
+@(private)
+_pop_node :: proc(c: ^$C/Cache($Key, $Value), e: ^Node(Key, Value)) {
+	if e == nil {
+		return
+	}
+	if c.head == e {
+		c.head = e.next
+	}
+	if c.tail == e {
+		c.tail = e.prev
+	}
+	if e.prev != nil {
+		e.prev.next = e.next
+	}
+
+	if e.next != nil {
+		e.next.prev = e.prev
+	}
+	e.prev = nil
+	e.next = nil
+}
@@ -1,377 +0,0 @@
-package container
-
-import "core:intrinsics"
-_ :: intrinsics
-
-
-Map :: struct($Key, $Value: typeid) where intrinsics.type_is_valid_map_key(Key) {
-	hash: Array(int),
-	entries: Array(Map_Entry(Key, Value)),
-}
-
-Map_Entry :: struct($Key, $Value: typeid) where intrinsics.type_is_valid_map_key(Key) {
-	hash:  uintptr,
-	next:  int,
-	key:   Key,
-	value: Value,
-}
-
-
-/*
-map_init :: proc{
-	map_init_none,
-	map_init_cap,
-}
-map_delete
-
-map_has
-map_get
-map_get_default
-map_get_ptr
-map_set
-map_remove
-map_reserve
-map_clear
-
-// Multi Map
-
-multi_map_find_first
-multi_map_find_next
-multi_map_count
-multi_map_get :: proc{
-	multi_map_get_array,
-	multi_map_get_slice,
-};
-multi_map_get_as_slice
-multi_map_insert
-multi_map_remove
-multi_map_remove_all
-
-*/
-
-map_init :: proc{map_init_none, map_init_cap}
-
-map_init_none :: proc(m: ^$M/Map($Key, $Value), allocator := context.allocator) {
-	m.hash.allocator = allocator
-	m.entries.allocator = allocator
-}
-
-map_init_cap :: proc(m: ^$M/Map($Key, $Value), cap: int, allocator := context.allocator) {
-	m.hash.allocator = allocator
-	m.entries.allocator = allocator
-	map_reserve(m, cap)
-}
-
-map_delete :: proc(m: $M/Map($Key, $Value)) {
-	array_delete(m.hash)
-	array_delete(m.entries)
-}
-
-
-map_has :: proc(m: $M/Map($Key, $Value), key: Key) -> bool {
-	return _map_find_or_fail(m, key) >= 0
-}
-
-map_get :: proc(m: $M/Map($Key, $Value), key: Key) -> (res: Value, ok: bool) #optional_ok {
-	i := _map_find_or_fail(m, key)
-	if i < 0 {
-		return {}, false
-	}
-	return array_get(m.entries, i).value, true
-}
-
-map_get_default :: proc(m: $M/Map($Key, $Value), key: Key, default: Value) -> (res: Value, ok: bool) #optional_ok {
-	i := _map_find_or_fail(m, key)
-	if i < 0 {
-		return default, false
-	}
-	return array_get(m.entries, i).value, true
-}
-
-map_get_ptr :: proc(m: $M/Map($Key, $Value), key: Key) -> ^Value {
-	i := _map_find_or_fail(m, key)
-	if i < 0 {
-		return nil
-	}
-	return array_get_ptr(m.entries, i).value
-}
-
-map_set :: proc(m: ^$M/Map($Key, $Value), key: Key, value: Value) {
-	if array_len(m.hash) == 0 {
-		_map_grow(m)
-	}
-
-	i := _map_find_or_make(m, key)
-	array_get_ptr(m.entries, i).value = value
-	if _map_full(m^) {
-		_map_grow(m)
-	}
-}
-
-map_remove :: proc(m: ^$M/Map($Key, $Value), key: Key) {
-	fr := _map_find_key(m^, key)
-	if fr.entry_index >= 0 {
-		_map_erase(m, fr)
-	}
-}
-
-
-map_reserve :: proc(m: ^$M/Map($Key, $Value), new_size: int) {
-	nm: M
-	map_init(&nm, m.hash.allocator)
-	array_resize(&nm.hash, new_size)
-	array_reserve(&nm.entries, array_len(m.entries))
-
-	for i in 0..<new_size {
-		array_set(&nm.hash, i, -1)
-	}
-	for i in 0..<array_len(m.entries) {
-		e := array_get(m.entries, i)
-		multi_map_insert(&nm, e.key, e.value)
-	}
-
-	map_delete(m^)
-	m^ = nm
-}
-
-map_clear :: proc(m: ^$M/Map($Key, $Value)) {
-	array_clear(&m.hash)
-	array_clear(&m.entries)
-}
-
-
-
-multi_map_find_first :: proc(m: $M/Map($Key, $Value), key: Key) -> ^Map_Entry(Key, Value) {
-	i := _map_find_or_fail(m, key)
-	if i < 0 {
-		return nil
-	}
-	return array_get_ptr(m.entries, i)
-}
-
-multi_map_find_next :: proc(m: $M/Map($Key, $Value), e: ^Map_Entry(Key, Value)) -> ^Map_Entry(Key, Value) {
-	i := e.next
-	for i >= 0 {
-		it := array_get_ptr(m.entries, i)
-		if it.hash == e.hash && it.key == e.key {
-			return it
-		}
-		i = it.next
-	}
-	return nil
-}
-
-multi_map_count :: proc(m: $M/Map($Key, $Value), key: Key) -> int {
-	n := 0
-	e := multi_map_find_first(m, key)
-	for e != nil {
-		n += 1
-		e = multi_map_find_next(m, e)
-	}
-	return n
-}
-
-multi_map_get :: proc{multi_map_get_array, multi_map_get_slice}
-
-multi_map_get_array :: proc(m: $M/Map($Key, $Value), key: Key, items: ^Array(Value)) {
-	if items == nil {
-		return
-	}
-	e := multi_map_find_first(m, key)
-	for e != nil {
-		array_append(items, e.value)
-		e = multi_map_find_next(m, e)
-	}
-}
-
-multi_map_get_slice :: proc(m: $M/Map($Key, $Value), key: Key, items: []Value) {
-	e := multi_map_find_first(m, key)
-	i := 0
-	for e != nil && i < len(items) {
-		items[i] = e.value
-		i += 1
-		e = multi_map_find_next(m, e)
-	}
-}
-
-multi_map_get_as_slice :: proc(m: $M/Map($Key, $Value), key: Key) -> []Value {
-	items: Array(Value)
-	array_init(&items, 0)
-
-	e := multi_map_find_first(m, key)
-	for e != nil {
-		array_append(&items, e.value)
-		e = multi_map_find_next(m, e)
-	}
-
-	return array_slice(items)
-}
-
-
-multi_map_insert :: proc(m: ^$M/Map($Key, $Value), key: Key, value: Value) {
-	if array_len(m.hash) == 0 {
-		_map_grow(m)
-	}
-
-	i := _map_make(m, key)
-	array_get_ptr(m.entries, i).value = value
-	if _map_full(m^) {
-		_map_grow(m)
-	}
-}
-
-multi_map_remove :: proc(m: ^$M/Map($Key, $Value), e: ^Map_Entry(Key, Value)) {
-	fr := _map_find_entry(m, e)
-	if fr.entry_index >= 0 {
-		_map_erase(m, fr)
-	}
-}
-
-multi_map_remove_all :: proc(m: ^$M/Map($Key, $Value), key: Key) {
-	for map_exist(m^, key) {
-		map_remove(m, key)
-	}
-}
-
-
-/// Internal
-
-
-Map_Find_Result :: struct {
-	hash_index:  int,
-	entry_prev:  int,
-	entry_index: int,
-}
-
-_map_add_entry :: proc(m: ^$M/Map($Key, $Value), key: Key) -> int where intrinsics.type_is_valid_map_key(Key) {
-	hasher := intrinsics.type_hasher_proc(Key)
-
-	e: Map_Entry(Key, Value)
-	e.key = key
-	e.hash = hasher(&e.key, 0)
-	e.next = -1
-	idx := array_len(m.entries)
-	array_push(&m.entries, e)
-	return idx
-}
-
-_map_erase :: proc(m: ^$M/Map, fr: Map_Find_Result) {
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, array_get(m.entries, fr.entry_index).next)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = array_get(m.entries, fr.entry_index).next
-	}
-
-	if fr.entry_index == array_len(m.entries)-1 {
-		array_pop_back(&m.entries)
-		return
-	}
-
-	array_set(&m.entries, fr.entry_index, array_get(m.entries, array_len(m.entries)-1))
-	last := _map_find_key(m^, array_get(m.entries, fr.entry_index).key)
-
-	if last.entry_prev < 0 {
-		array_get_ptr(m.entries, last.entry_prev).next = fr.entry_index
-	} else {
-		array_set(&m.hash, last.hash_index, fr.entry_index)
-	}
-}
-
-
-_map_find_key :: proc(m: $M/Map($Key, $Value), key: Key) -> Map_Find_Result where intrinsics.type_is_valid_map_key(Key) {
-	fr: Map_Find_Result
-	fr.hash_index = -1
-	fr.entry_prev = -1
-	fr.entry_index = -1
-
-	if array_len(m.hash) == 0 {
-		return fr
-	}
-
-	hasher := intrinsics.type_hasher_proc(Key)
-
-	key := key
-	hash := hasher(&key, 0)
-
-	fr.hash_index = int(hash % uintptr(array_len(m.hash)))
-	fr.entry_index = array_get(m.hash, fr.hash_index)
-	for fr.entry_index >= 0 {
-		it := array_get_ptr(m.entries, fr.entry_index)
-		if it.hash == hash && it.key == key {
-			return fr
-		}
-		fr.entry_prev = fr.entry_index
-		fr.entry_index = it.next
-	}
-	return fr
-}
-
-_map_find_entry :: proc(m: ^$M/Map($Key, $Value), e: ^Map_Entry(Key, Value)) -> Map_Find_Result {
-	fr: Map_Find_Result
-	fr.hash_index = -1
-	fr.entry_prev = -1
-	fr.entry_index = -1
-
-	if array_len(m.hash) == 0 {
-		return fr
-	}
-
-	fr.hash_index = int(e.hash % uintptr(array_len(m.hash)))
-	fr.entry_index = array_get(m.hash, fr.hash_index)
-	for fr.entry_index >= 0 {
-		it := array_get_ptr(m.entries, fr.entry_index)
-		if it == e {
-			return fr
-		}
-		fr.entry_prev = fr.entry_index
-		fr.entry_index = it.next
-	}
-	return fr
-}
-
-_map_find_or_fail :: proc(m: $M/Map($Key, $Value), key: Key) -> int {
-	return _map_find_key(m, key).entry_index
-}
-_map_find_or_make :: proc(m: ^$M/Map($Key, $Value), key: Key) -> int {
-	fr := _map_find_key(m^, key)
-	if fr.entry_index >= 0 {
-		return fr.entry_index
-	}
-
-	i := _map_add_entry(m, key)
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, i)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = i
-	}
-	return i
-}
-
-
-_map_make :: proc(m: ^$M/Map($Key, $Value), key: Key) -> int {
-	fr := _map_find_key(m^, key)
-	i := _map_add_entry(m, key)
-
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, i)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = i
-	}
-
-	array_get_ptr(m.entries, i).next = fr.entry_index
-
-	return i
-}
-
-
-_map_full :: proc(m: $M/Map($Key, $Value)) -> bool {
-	// TODO(bill): Determine good max load factor
-	return array_len(m.entries) >= (array_len(m.hash) / 4)*3
-}
-
-_map_grow :: proc(m: ^$M/Map($Key, $Value)) {
-	new_size := array_len(m.entries) * 4 + 7 // TODO(bill): Determine good grow rate
-	map_reserve(m, new_size)
-}
-
-
@@ -1,121 +0,0 @@
-package container
-
-Priority_Queue :: struct($T: typeid) {
-	data: Array(T),
-	len: int,
-	priority: proc(item: T) -> int,
-}
-
-priority_queue_init_none :: proc(q: ^$Q/Priority_Queue($T), f: proc(item: T) -> int, allocator := context.allocator) {
-	queue_init_len(q, f, 0, allocator)
-}
-priority_queue_init_len :: proc(q: ^$Q/Priority_Queue($T), f: proc(item: T) -> int, len: int, allocator := context.allocator) {
-	queue_init_len_cap(q, f, 0, 16, allocator)
-}
-priority_queue_init_len_cap :: proc(q: ^$Q/Priority_Queue($T), f: proc(item: T) -> int, len: int, cap: int, allocator := context.allocator) {
-	array_init(&q.data, len, cap, allocator)
-	q.len = len
-	q.priority = f
-}
-
-priority_queue_init :: proc{priority_queue_init_none, priority_queue_init_len, priority_queue_init_len_cap}
-
-
-priority_queue_delete :: proc(q: $Q/Priority_Queue($T)) {
-	array_delete(q.data)
-}
-
-priority_queue_clear :: proc(q: ^$Q/Priority_Queue($T)) {
-	q.len = 0
-}
-
-priority_queue_len :: proc(q: $Q/Priority_Queue($T)) -> int {
-	return q.len
-}
-
-priority_queue_cap :: proc(q: $Q/Priority_Queue($T)) -> int {
-	return array_cap(q.data)
-}
-
-priority_queue_space :: proc(q: $Q/Priority_Queue($T)) -> int {
-	return array_len(q.data) - q.len
-}
-
-priority_queue_reserve :: proc(q: ^$Q/Priority_Queue($T), capacity: int) {
-	if capacity > q.len {
-		array_resize(&q.data, new_capacity)
-	}
-}
-
-priority_queue_resize :: proc(q: ^$Q/Priority_Queue($T), length: int) {
-	if length > q.len {
-		array_resize(&q.data, new_capacity)
-	}
-	q.len = length
-}
-
-_priority_queue_grow :: proc(q: ^$Q/Priority_Queue($T), min_capacity: int = 0) {
-	new_capacity := max(array_len(q.data)*2 + 8, min_capacity)
-	array_resize(&q.data, new_capacity)
-}
-
-
-priority_queue_push :: proc(q: ^$Q/Priority_Queue($T), item: T) {
-	if array_len(q.data) - q.len == 0 {
-		_priority_queue_grow(q)
-	}
-
-	s := array_slice(q.data)
-	s[q.len] = item
-
-	i := q.len
-	for i > 0 {
-		p := (i - 1) / 2
-		if q.priority(s[p]) <= q.priority(item) { 
-			break 
-		}
-		s[i] = s[p]
-		i = p
-	}
-
-	q.len += 1
-	if q.len > 0 { 
-		s[i] = item 
-	} 
-}
-
-
-
-priority_queue_pop :: proc(q: ^$Q/Priority_Queue($T)) -> T {
-	assert(q.len > 0)
-
-	s := array_slice(q.data)
-	min := s[0]
-	root := s[q.len-1]
-	q.len -= 1
-
-	i := 0
-	for i * 2 + 1 < q.len {
-		a := i * 2 + 1
-		b := i * 2 + 2
-		c := b < q.len && q.priority(s[b]) < q.priority(s[a]) ? b : a
-
-		if q.priority(s[c]) >= q.priority(root) {
-			break
-		}
-		s[i] = s[c]
-		i = c
-	}
-
-	if q.len > 0 {
-		s[i] = root
-	}
-	return min
-}
-
-priority_queue_peek :: proc(q: ^$Q/Priority_Queue($T)) -> T {
-	assert(q.len > 0)
-
-	s := array_slice(q.data)
-	return s[0]
-}
@@ -0,0 +1,142 @@
+package container_priority_queue
+
+import "core:builtin"
+
+Priority_Queue :: struct($T: typeid) {
+	queue: [dynamic]T,
+	
+	less:  proc(a, b: T) -> bool,
+	swap:  proc(q: []T, i, j: int),
+}
+
+DEFAULT_CAPACITY :: 16
+
+default_swap_proc :: proc($T: typeid) -> proc(q: []T, i, j: int) {
+	return proc(q: []T, i, j: int) {
+		q[i], q[j] = q[j], q[i]
+	}
+}
+
+init :: proc(pq: ^$Q/Priority_Queue($T), less: proc(a, b: T) -> bool, swap: proc(q: []T, i, j: int), capacity := DEFAULT_CAPACITY, allocator := context.allocator) {
+	if pq.queue.allocator.procedure == nil {
+		pq.queue.allocator = allocator
+	}
+	reserve(pq, capacity)
+	pq.less = less
+	pq.swap = swap
+}
+
+init_from_dynamic_array :: proc(pq: ^$Q/Priority_Queue($T), queue: [dynamic]T, less: proc(a, b: T) -> bool, swap: proc(q: []T, i, j: int)) {
+	pq.queue = queue
+	pq.less = less
+	pq.swap = swap
+	n := builtin.len(pq.queue)
+	for i := n/2 - 1; i >= 0; i -= 1 {
+		_shift_down(pq, i, n)
+	}
+}
+
+destroy :: proc(pq: ^$Q/Priority_Queue($T)) {
+	clear(pq)
+	delete(pq.queue)
+}
+
+reserve :: proc(pq: ^$Q/Priority_Queue($T), capacity: int) {
+	builtin.reserve(&pq.queue, capacity)
+}
+clear :: proc(pq: ^$Q/Priority_Queue($T)) {
+	builtin.clear(&pq.queue)
+}
+len :: proc(pq: $Q/Priority_Queue($T)) -> int {
+	return builtin.len(pq.queue)
+}
+cap :: proc(pq: $Q/Priority_Queue($T)) -> int {
+	return builtin.cap(pq.queue)
+}
+
+_shift_down :: proc(pq: ^$Q/Priority_Queue($T), i0, n: int) -> bool {
+	// O(n log n)
+	if 0 > i0 || i0 > n {
+		return false
+	}
+	
+	i := i0
+	queue := pq.queue[:]
+	
+	for {
+		j1 := 2*i + 1
+		if j1 < 0 || j1 >= n {
+			break
+		}
+		j := j1
+		if j2 := j1+1; j2 < n && pq.less(queue[j2], queue[j1]) {
+			j = j2
+		}
+		if !pq.less(queue[j], queue[i]) {
+			break
+		}
+		
+		pq.swap(queue, i, j)
+		i = j
+	}
+	return i > i0
+}
+
+_shift_up :: proc(pq: ^$Q/Priority_Queue($T), j: int) {
+	j := j
+	queue := pq.queue[:]
+	for 0 <= j {
+		i := (j-1)/2
+		if i == j || !pq.less(queue[j], queue[i]) {
+			break
+		}
+		pq.swap(queue, i, j)
+		j = i
+	}
+}
+
+// NOTE(bill): When an element at index 'i' has changed its value, this will fix the
+// the heap ordering. This is using a basic "heapsort" with shift up and a shift down parts.
+fix :: proc(pq: ^$Q/Priority_Queue($T), i: int) {
+	if !_shift_down(pq, i, builtin.len(pq.queue)) {
+		_shift_up(pq, i)
+	}
+}
+
+push :: proc(pq: ^$Q/Priority_Queue($T), value: T) {
+	append(&pq.queue, value)
+	_shift_up(pq, builtin.len(pq.queue)-1)
+}
+
+pop :: proc(pq: ^$Q/Priority_Queue($T), loc := #caller_location) -> (value: T) {
+	assert(condition=builtin.len(pq.queue)>0, loc=loc)
+	
+	n := builtin.len(pq.queue)-1
+	pq.swap(pq.queue[:], 0, n)
+	_shift_down(pq, 0, n)
+	return builtin.pop(&pq.queue)
+}
+
+pop_safe :: proc(pq: ^$Q/Priority_Queue($T), loc := #caller_location) -> (value: T, ok: bool) {
+	if builtin.len(pq.queue) > 0 {
+		n := builtin.len(pq.queue)-1
+		pq.swap(pq.queue[:], 0, n)
+		_shift_down(pq, 0, n)
+		return builtin.pop_safe(&pq.queue)
+	}
+	return
+}
+
+remove :: proc(pq: ^$Q/Priority_Queue($T), i: int) -> (value: T, ok: bool) {
+	n := builtin.len(pq.queue)
+	if 0 <= i && i < n {
+		if n != i {
+			pq.swap(pq.queue[:], i, n)
+			_shift_down(pq, i, n)
+			_shift_up(pq, i)
+		}
+		value, ok = builtin.pop_safe(&pq.queue)
+	}
+	return
+}
+
@@ -1,175 +0,0 @@
-package container
-
-Queue :: struct($T: typeid) {
-	data: Array(T),
-	len: int,
-	offset: int,
-}
-
-/*
-queue_init :: proc{
-	queue_init_none,
-	queue_init_len,
-	queue_init_len_cap,
-}
-queue_delete
-queue_clear
-queue_len
-queue_cap
-queue_space
-queue_get
-queue_set
-queue_reserve
-queue_resize
-queue_push :: proc{
-	queue_push_back, 
-	queue_push_elems,
-};
-queue_push_front
-queue_pop_front
-queue_pop_back
-queue_consume
-*/
-
-queue_init_none :: proc(q: ^$Q/Queue($T), allocator := context.allocator) {
-	queue_init_len(q, 0, allocator)
-}
-queue_init_len :: proc(q: ^$Q/Queue($T), len: int, allocator := context.allocator) {
-	queue_init_len_cap(q, 0, 16, allocator)
-}
-queue_init_len_cap :: proc(q: ^$Q/Queue($T), len: int, cap: int, allocator := context.allocator) {
-	array_init(&q.data, len, cap, allocator)
-	q.len = len
-	q.offset = 0
-}
-
-queue_init :: proc{queue_init_none, queue_init_len, queue_init_len_cap}
-
-queue_delete :: proc(q: $Q/Queue($T)) {
-	array_delete(q.data)
-}
-
-queue_clear :: proc(q: ^$Q/Queue($T)) {
-	q.len = 0
-}
-
-queue_len :: proc(q: $Q/Queue($T)) -> int {
-	return q.len
-}
-
-queue_cap :: proc(q: $Q/Queue($T)) -> int {
-	return array_cap(q.data)
-}
-
-queue_space :: proc(q: $Q/Queue($T)) -> int {
-	return array_len(q.data) - q.len
-}
-
-queue_get :: proc(q: $Q/Queue($T), index: int) -> T {
-	i := (index + q.offset) % array_len(q.data)
-	data := array_slice(q.data)
-	return data[i]
-}
-
-queue_set :: proc(q: ^$Q/Queue($T), index: int, item: T)  {
-	i := (index + q.offset) % array_len(q.data)
-	data := array_slice(q.data)
-	data[i] = item
-}
-
-
-queue_reserve :: proc(q: ^$Q/Queue($T), capacity: int) {
-	if capacity > q.len {
-		_queue_increase_capacity(q, capacity)
-	}
-}
-
-queue_resize :: proc(q: ^$Q/Queue($T), length: int) {
-	if length > q.len {
-		_queue_increase_capacity(q, length)
-	}
-	q.len = length
-}
-
-queue_push_back :: proc(q: ^$Q/Queue($T), item: T) {
-	if queue_space(q^) == 0 {
-		_queue_grow(q)
-	}
-
-	queue_set(q, q.len, item)
-	q.len += 1
-}
-
-queue_push_front :: proc(q: ^$Q/Queue($T), item: T) {
-	if queue_space(q^) == 0 {
-		_queue_grow(q)
-	}
-
-	q.offset = (q.offset - 1 + array_len(q.data)) % array_len(q.data)
-	q.len += 1
-	queue_set(q, 0, item)
-}
-
-queue_pop_front :: proc(q: ^$Q/Queue($T)) -> T {
-	assert(q.len > 0)
-	item := queue_get(q^, 0)
-	q.offset = (q.offset + 1) % array_len(q.data)
-	q.len -= 1
-	if q.len == 0 {
-		q.offset = 0
-	}
-	return item
-}
-
-queue_pop_back :: proc(q: ^$Q/Queue($T)) -> T {
-	assert(q.len > 0)
-	item := queue_get(q^, q.len-1)
-	q.len -= 1
-	return item
-}
-
-queue_consume :: proc(q: ^$Q/Queue($T), count: int) {
-	q.offset = (q.offset + count) & array_len(q.data)
-	q.len -= count
-}
-
-
-queue_push_elems :: proc(q: ^$Q/Queue($T), items: ..T) {
-	if queue_space(q^) < len(items) {
-		_queue_grow(q, q.len + len(items))
-	}
-	size := array_len(q.data)
-	insert := (q.offset + q.len) % size
-
-	to_insert := len(items)
-	if insert + to_insert > size {
-		to_insert = size - insert
-	}
-
-	the_items := items[:]
-
-	data := array_slice(q.data)
-
-	q.len += copy(data[insert:][:to_insert], the_items)
-	the_items = the_items[to_insert:]
-	q.len += copy(data[:], the_items)
-}
-
-queue_push :: proc{queue_push_back, queue_push_elems}
-
-
-
-_queue_increase_capacity :: proc(q: ^$Q/Queue($T), new_capacity: int) {
-	end := array_len(q.data)
-	array_resize(&q.data, new_capacity)
-	if q.offset + q.len > end {
-		end_items := q.len + end
-		data := array_slice(q.data)
-		copy(data[new_capacity-end_items:][:end_items], data[q.offset:][:end_items])
-		q.offset += new_capacity - end
-	}
-}
-_queue_grow :: proc(q: ^$Q/Queue($T), min_capacity: int = 0) {
-	new_capacity := max(array_len(q.data)*2 + 8, min_capacity)
-	_queue_increase_capacity(q, new_capacity)
-}
@@ -0,0 +1,238 @@
+package container_queue
+
+import "core:builtin"
+import "core:runtime"
+_ :: runtime
+
+// Dynamically resizable double-ended queue/ring-buffer
+Queue :: struct($T: typeid) {
+	data:   [dynamic]T,
+	len:    uint,
+	offset: uint,
+}
+
+DEFAULT_CAPACITY :: 16
+
+// Procedure to initialize a queue
+init :: proc(q: ^$Q/Queue($T), capacity := DEFAULT_CAPACITY, allocator := context.allocator) -> runtime.Allocator_Error {
+	if q.data.allocator.procedure == nil {
+		q.data.allocator = allocator
+	}
+	clear(q)
+	return reserve(q, capacity)
+}
+
+// Procedure to initialize a queue from a fixed backing slice
+init_from_slice :: proc(q: ^$Q/Queue($T), backing: []T) -> bool {
+	clear(q)
+	q.data = transmute([dynamic]T)runtime.Raw_Dynamic_Array{
+		data = raw_data(backing),
+		len = builtin.len(backing),
+		cap = builtin.len(backing),
+		allocator = {procedure=runtime.nil_allocator_proc, data=nil},
+	}
+	return true
+}
+
+// Procedure to destroy a queue
+destroy :: proc(q: ^$Q/Queue($T)) {
+	delete(q.data)
+}
+
+// The length of the queue
+len :: proc(q: $Q/Queue($T)) -> int {
+	return int(q.len)
+}
+
+// The current capacity of the queue
+cap :: proc(q: $Q/Queue($T)) -> int {
+	return builtin.len(q.data)
+}
+
+// Remaining space in the queue (cap-len)
+space :: proc(q: $Q/Queue($T)) -> int {
+	return builtin.len(q.data) - int(q.len)
+}
+
+// Reserve enough space for at least the specified capacity
+reserve :: proc(q: ^$Q/Queue($T), capacity: int) -> runtime.Allocator_Error {
+	if capacity > space(q^) {
+		return _grow(q, uint(capacity)) 
+	}
+	return nil
+}
+
+
+get :: proc(q: ^$Q/Queue($T), #any_int i: int, loc := #caller_location) -> T {
+	runtime.bounds_check_error_loc(loc, i, builtin.len(q.data))
+
+	idx := (uint(i)+q.offset)%builtin.len(q.data)
+	return q.data[idx]
+}
+
+front :: proc(q: ^$Q/Queue($T)) -> T {
+	return q.data[q.offset]
+}
+front_ptr :: proc(q: ^$Q/Queue($T)) -> ^T {
+	return &q.data[q.offset]
+}
+
+back :: proc(q: ^$Q/Queue($T)) -> T {
+	idx := (q.offset+uint(q.len))%builtin.len(q.data)
+	return q.data[idx]
+}
+back_ptr :: proc(q: ^$Q/Queue($T)) -> ^T {
+	idx := (q.offset+uint(q.len))%builtin.len(q.data)
+	return &q.data[idx]
+}
+
+set :: proc(q: ^$Q/Queue($T), #any_int i: int, val: T, loc := #caller_location) {
+	runtime.bounds_check_error_loc(loc, i, builtin.len(q.data))
+	
+	idx := (uint(i)+q.offset)%builtin.len(q.data)
+	q.data[idx] = val
+}
+get_ptr :: proc(q: ^$Q/Queue($T), #any_int i: int, loc := #caller_location) -> ^T {
+	runtime.bounds_check_error_loc(loc, i, builtin.len(q.data))
+	
+	idx := (uint(i)+q.offset)%builtin.len(q.data)
+	return &q.data[idx]
+}
+
+peek_front :: proc(q: ^$Q/Queue($T), loc := #caller_location) -> ^T {
+	runtime.bounds_check_error_loc(loc, 0, builtin.len(q.data))
+	idx := q.offset%builtin.len(q.data)
+	return &q.data[idx]
+}
+
+peek_back :: proc(q: ^$Q/Queue($T), loc := #caller_location) -> ^T {
+	runtime.bounds_check_error_loc(loc, int(q.len - 1), builtin.len(q.data))
+	idx := (uint(q.len - 1)+q.offset)%builtin.len(q.data)
+	return &q.data[idx]
+}
+
+// Push an element to the back of the queue
+push_back :: proc(q: ^$Q/Queue($T), elem: T) -> (ok: bool, err: runtime.Allocator_Error) {
+	if space(q^) == 0 {
+		_grow(q) or_return
+	}
+	idx := (q.offset+uint(q.len))%builtin.len(q.data)
+	q.data[idx] = elem
+	q.len += 1
+	return true, nil
+}
+
+// Push an element to the front of the queue
+push_front :: proc(q: ^$Q/Queue($T), elem: T) -> (ok: bool, err: runtime.Allocator_Error)  {
+	if space(q^) == 0 {
+		_grow(q) or_return
+	}	
+	q.offset = uint(q.offset - 1 + builtin.len(q.data)) % builtin.len(q.data)
+	q.len += 1
+	q.data[q.offset] = elem
+	return true, nil
+}
+
+
+// Pop an element from the back of the queue
+pop_back :: proc(q: ^$Q/Queue($T), loc := #caller_location) -> (elem: T) {
+	assert(condition=q.len > 0, loc=loc)
+	q.len -= 1
+	idx := (q.offset+uint(q.len))%builtin.len(q.data)
+	elem = q.data[idx]
+	return
+}
+// Safely pop an element from the back of the queue
+pop_back_safe :: proc(q: ^$Q/Queue($T)) -> (elem: T, ok: bool) {
+	if q.len > 0 {
+		q.len -= 1
+		idx := (q.offset+uint(q.len))%builtin.len(q.data)
+		elem = q.data[idx]
+		ok = true
+	}
+	return
+}
+
+// Pop an element from the front of the queue
+pop_front :: proc(q: ^$Q/Queue($T), loc := #caller_location) -> (elem: T) {
+	assert(condition=q.len > 0, loc=loc)
+	elem = q.data[q.offset]
+	q.offset = (q.offset+1)%builtin.len(q.data)
+	q.len -= 1
+	return
+}
+// Safely pop an element from the front of the queue
+pop_front_safe :: proc(q: ^$Q/Queue($T)) -> (elem: T, ok: bool) {
+	if q.len > 0 {
+		elem = q.data[q.offset]
+		q.offset = (q.offset+1)%builtin.len(q.data)
+		q.len -= 1
+		ok = true
+	}
+	return
+}
+
+// Push multiple elements to the front of the queue
+push_back_elems :: proc(q: ^$Q/Queue($T), elems: ..T) -> (ok: bool, err: runtime.Allocator_Error)  {
+	n := uint(builtin.len(elems))
+	if space(q^) < int(n) {
+		_grow(q, q.len + n) or_return
+	}
+	
+	sz := uint(builtin.len(q.data))
+	insert_from := (q.offset + q.len) % sz
+	insert_to := n
+	if insert_from + insert_to > sz {
+		insert_to = sz - insert_from
+	}
+	copy(q.data[insert_from:], elems[:insert_to])
+	copy(q.data[:insert_from], elems[insert_to:])
+	q.len += n
+	return true, nil
+}
+
+// Consume `n` elements from the front of the queue
+consume_front :: proc(q: ^$Q/Queue($T), n: int, loc := #caller_location) {
+	assert(condition=int(q.len) >= n, loc=loc)
+	if n > 0 {
+		nu := uint(n)
+		q.offset = (q.offset + nu) % builtin.len(q.data)
+		q.len -= nu	
+	}
+}
+
+// Consume `n` elements from the back of the queue
+consume_back :: proc(q: ^$Q/Queue($T), n: int, loc := #caller_location) {
+	assert(condition=int(q.len) >= n, loc=loc)
+	if n > 0 {
+		q.len -= uint(n)
+	}
+}
+
+
+
+append_elem  :: push_back
+append_elems :: push_back_elems
+push   :: proc{push_back, push_back_elems}
+append :: proc{push_back, push_back_elems}
+
+
+// Clear the contents of the queue
+clear :: proc(q: ^$Q/Queue($T)) {
+	q.len = 0
+	q.offset = 0
+}
+
+
+// Internal growinh procedure
+_grow :: proc(q: ^$Q/Queue($T), min_capacity: uint = 0) -> runtime.Allocator_Error {
+	new_capacity := max(min_capacity, uint(8), uint(builtin.len(q.data))*2)
+	n := uint(builtin.len(q.data))
+	builtin.resize(&q.data, int(new_capacity)) or_return
+	if q.offset + q.len > n {
+		diff := n - q.offset
+		copy(q.data[new_capacity-diff:], q.data[q.offset:][:diff])
+		q.offset += new_capacity - n
+	}
+	return nil
+}
@@ -1,74 +0,0 @@
-package container
-
-
-Ring :: struct($T: typeid) {
-	next, prev: ^Ring(T),
-	value: T,
-}
-
-ring_init :: proc(r: ^$R/Ring) -> ^R {
-	r.prev, r.next = r, r
-	return r
-}
-
-ring_next :: proc(r: ^$R/Ring) -> ^R {
-	if r.next == nil {
-		return ring_init(r)
-	}
-	return r.next
-}
-ring_prev :: proc(r: ^$R/Ring) -> ^R {
-	if r.prev == nil {
-		return ring_init(r)
-	}
-	return r.prev
-}
-
-
-ring_move :: proc(r: ^$R/Ring, n: int) -> ^R {
-  r := r
-	if r.next == nil {
-		return ring_init(r)
-	}
-
-	switch {
-	case n < 0:
-		for _ in n..<0 {
-			r = r.prev
-		}
-	case n > 0:
-		for _ in 0..<n {
-			r = r.next
-		}
-	}
-	return r
-}
-
-ring_link :: proc(r, s: ^$R/Ring) -> ^R {
-	n := ring_next(r)
-	if s != nil {
-		p := ring_prev(s)
-		r.next = s
-		s.prev = r
-		n.prev = p
-		p.next = n
-	}
-	return n
-}
-ring_unlink :: proc(r: ^$R/Ring, n: int) -> ^R {
-	if n <= 0 {
-		return nil
-	}
-	return ring_link(r, ring_move(r, n+1))
-}
-ring_len :: proc(r: ^$R/Ring) -> int {
-	n := 0
-	if r != nil {
-		n = 1
-		for p := ring_next(r); p != r; p = p.next {
-			n += 1
-		}
-	}
-	return n
-}
-
@@ -1,240 +0,0 @@
-package container
-
-Set :: struct {
-	hash:    Array(int),
-	entries: Array(Set_Entry),
-}
-
-Set_Entry :: struct {
-	key:   u64,
-	next:  int,
-}
-
-
-/*
-set_init :: proc{
-	set_init_none,
-	set_init_cap,
-}
-set_delete
-
-set_in
-set_not_in
-set_add
-set_remove
-set_reserve
-set_clear
-*/
-
-set_init :: proc{set_init_none, set_init_cap}
-
-set_init_none :: proc(m: ^Set, allocator := context.allocator) {
-	m.hash.allocator = allocator
-	m.entries.allocator = allocator
-}
-
-set_init_cap :: proc(m: ^Set, cap: int, allocator := context.allocator) {
-	m.hash.allocator = allocator
-	m.entries.allocator = allocator
-	set_reserve(m, cap)
-}
-
-set_delete :: proc(m: Set) {
-	array_delete(m.hash)
-	array_delete(m.entries)
-}
-
-
-set_in :: proc(m: Set, key: u64) -> bool {
-	return _set_find_or_fail(m, key) >= 0
-}
-set_not_in :: proc(m: Set, key: u64) -> bool {
-	return _set_find_or_fail(m, key) < 0
-}
-
-set_add :: proc(m: ^Set, key: u64) {
-	if array_len(m.hash) == 0 {
-		_set_grow(m)
-	}
-
-	_ = _set_find_or_make(m, key)
-	if _set_full(m^) {
-		_set_grow(m)
-	}
-}
-
-set_remove :: proc(m: ^Set, key: u64) {
-	fr := _set_find_key(m^, key)
-	if fr.entry_index >= 0 {
-		_set_erase(m, fr)
-	}
-}
-
-
-set_reserve :: proc(m: ^Set, new_size: int) {
-	nm: Set
-	set_init(&nm, m.hash.allocator)
-	array_resize(&nm.hash, new_size)
-	array_reserve(&nm.entries, array_len(m.entries))
-
-	for i in 0..<new_size {
-		array_set(&nm.hash, i, -1)
-	}
-	for i in 0..<array_len(m.entries) {
-		e := array_get(m.entries, i)
-		set_add(&nm, e.key)
-	}
-
-	set_delete(m^)
-	m^ = nm
-}
-
-set_clear :: proc(m: ^Set) {
-	array_clear(&m.hash)
-	array_clear(&m.entries)
-}
-
-
-set_equal :: proc(a, b: Set) -> bool {
-	a_entries := array_slice(a.entries)
-	b_entries := array_slice(b.entries)
-	if len(a_entries) != len(b_entries) {
-		return false
-	}
-	for e in a_entries {
-		if set_not_in(b, e.key) {
-			return false
-		}
-	}
-
-	return true
-}
-
-
-
-/// Internal
-
-_set_add_entry :: proc(m: ^Set, key: u64) -> int {
-	e: Set_Entry
-	e.key = key
-	e.next = -1
-	idx := array_len(m.entries)
-	array_push(&m.entries, e)
-	return idx
-}
-
-_set_erase :: proc(m: ^Set, fr: Map_Find_Result) {
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, array_get(m.entries, fr.entry_index).next)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = array_get(m.entries, fr.entry_index).next
-	}
-
-	if fr.entry_index == array_len(m.entries)-1 {
-		array_pop_back(&m.entries)
-		return
-	}
-
-	array_set(&m.entries, fr.entry_index, array_get(m.entries, array_len(m.entries)-1))
-	last := _set_find_key(m^, array_get(m.entries, fr.entry_index).key)
-
-	if last.entry_prev < 0 {
-		array_get_ptr(m.entries, last.entry_prev).next = fr.entry_index
-	} else {
-		array_set(&m.hash, last.hash_index, fr.entry_index)
-	}
-}
-
-
-_set_find_key :: proc(m: Set, key: u64) -> Map_Find_Result {
-	fr: Map_Find_Result
-	fr.hash_index = -1
-	fr.entry_prev = -1
-	fr.entry_index = -1
-
-	if array_len(m.hash) == 0 {
-		return fr
-	}
-
-	fr.hash_index = int(key % u64(array_len(m.hash)))
-	fr.entry_index = array_get(m.hash, fr.hash_index)
-	for fr.entry_index >= 0 {
-		it := array_get_ptr(m.entries, fr.entry_index)
-		if it.key == key {
-			return fr
-		}
-		fr.entry_prev = fr.entry_index
-		fr.entry_index = it.next
-	}
-	return fr
-}
-
-_set_find_entry :: proc(m: ^Set, e: ^Set_Entry) -> Map_Find_Result {
-	fr: Map_Find_Result
-	fr.hash_index = -1
-	fr.entry_prev = -1
-	fr.entry_index = -1
-
-	if array_len(m.hash) == 0 {
-		return fr
-	}
-
-	fr.hash_index = int(e.key % u64(array_len(m.hash)))
-	fr.entry_index = array_get(m.hash, fr.hash_index)
-	for fr.entry_index >= 0 {
-		it := array_get_ptr(m.entries, fr.entry_index)
-		if it == e {
-			return fr
-		}
-		fr.entry_prev = fr.entry_index
-		fr.entry_index = it.next
-	}
-	return fr
-}
-
-_set_find_or_fail :: proc(m: Set, key: u64) -> int {
-	return _set_find_key(m, key).entry_index
-}
-_set_find_or_make :: proc(m: ^Set, key: u64) -> int {
-	fr := _set_find_key(m^, key)
-	if fr.entry_index >= 0 {
-		return fr.entry_index
-	}
-
-	i := _set_add_entry(m, key)
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, i)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = i
-	}
-	return i
-}
-
-
-_set_make :: proc(m: ^Set, key: u64) -> int {
-	fr := _set_find_key(m^, key)
-	i := _set_add_entry(m, key)
-
-	if fr.entry_prev < 0 {
-		array_set(&m.hash, fr.hash_index, i)
-	} else {
-		array_get_ptr(m.entries, fr.entry_prev).next = i
-	}
-
-	array_get_ptr(m.entries, i).next = fr.entry_index
-
-	return i
-}
-
-
-_set_full :: proc(m: Set) -> bool {
-	// TODO(bill): Determine good max load factor
-	return array_len(m.entries) >= (array_len(m.hash) / 4)*3
-}
-
-_set_grow :: proc(m: ^Set) {
-	new_size := array_len(m.entries) * 4 + 7 // TODO(bill): Determine good grow rate
-	set_reserve(m, new_size)
-}
-
-
@@ -1,95 +0,0 @@
-package container
-
-Small_Array :: struct($N: int, $T: typeid) where N >= 0 {
-	data: [N]T,
-	len:  int,
-}
-
-
-small_array_len :: proc(a: $A/Small_Array) -> int {
-	return a.len
-}
-
-small_array_cap :: proc(a: $A/Small_Array) -> int {
-	return len(a.data)
-}
-
-small_array_space :: proc(a: $A/Small_Array) -> int {
-	return len(a.data) - a.len
-}
-
-small_array_slice :: proc(a: ^$A/Small_Array($N, $T)) -> []T {
-	return a.data[:a.len]
-}
-
-
-small_array_get :: proc(a: $A/Small_Array($N, $T), index: int, loc := #caller_location) -> T {
-	return a.data[index]
-}
-small_array_get_ptr :: proc(a: $A/Small_Array($N, $T), index: int, loc := #caller_location) -> ^T {
-	return &a.data[index]
-}
-
-small_array_set :: proc(a: ^$A/Small_Array($N, $T), index: int, item: T, loc := #caller_location) {
-	a.data[index] = item
-}
-
-small_array_resize :: proc(a: ^$A/Small_Array, length: int) {
-	a.len = min(length, len(a.data))
-}
-
-
-small_array_push_back :: proc(a: ^$A/Small_Array($N, $T), item: T) -> bool {
-	if a.len < len(a.data) {
-		a.len += 1
-		a.data[a.len-1] = item
-		return true
-	}
-	return false
-}
-
-small_array_push_front :: proc(a: ^$A/Small_Array($N, $T), item: T) -> bool {
-	if a.len < len(a.data) {
-		a.len += 1
-		data := small_array_slice(a)
-		copy(data[1:], data[:])
-		data[0] = item
-		return true
-	}
-	return false
-}
-
-small_array_pop_back :: proc(a: ^$A/Small_Array($N, $T), loc := #caller_location) -> T {
-	assert(condition=a.len > 0, loc=loc)
-	item := a.data[a.len-1]
-	a.len -= 1
-	return item
-}
-
-small_array_pop_front :: proc(a: ^$A/Small_Array($N, $T), loc := #caller_location) -> T {
-	assert(condition=a.len > 0, loc=loc)
-	item := a.data[0]
-	s := small_array_slice(a)
-	copy(s[:], s[1:])
-	a.len -= 1
-	return item
-}
-
-
-small_array_consume :: proc(a: ^$A/Small_Array($N, $T), count: int, loc := #caller_location) {
-	assert(condition=a.len >= count, loc=loc)
-	a.len -= count
-}
-
-small_array_clear :: proc(a: ^$A/Small_Array($N, $T)) {
-	small_array_resize(a, 0)
-}
-
-small_array_push_back_elems :: proc(a: ^$A/Small_Array($N, $T), items: ..T) {
-	n := copy(a.data[a.len:], items[:])
-	a.len += n
-}
-
-small_array_push   :: proc{small_array_push_back, small_array_push_back_elems}
-small_array_append :: proc{small_array_push_back, small_array_push_back_elems}
-
@@ -0,0 +1,162 @@
+package container_small_array
+
+import "core:builtin"
+import "core:runtime"
+_ :: runtime
+
+Small_Array :: struct($N: int, $T: typeid) where N >= 0 {
+	data: [N]T,
+	len:  int,
+}
+
+
+len :: proc "contextless" (a: $A/Small_Array) -> int {
+	return a.len
+}
+
+cap :: proc "contextless" (a: $A/Small_Array) -> int {
+	return builtin.len(a.data)
+}
+
+space :: proc "contextless" (a: $A/Small_Array) -> int {
+	return builtin.len(a.data) - a.len
+}
+
+slice :: proc "contextless" (a: ^$A/Small_Array($N, $T)) -> []T {
+	return a.data[:a.len]
+}
+
+
+get :: proc "contextless" (a: $A/Small_Array($N, $T), index: int) -> T {
+	return a.data[index]
+}
+get_ptr :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int) -> ^T {
+	return &a.data[index]
+}
+
+get_safe :: proc(a: $A/Small_Array($N, $T), index: int) -> (T, bool) #no_bounds_check {
+	if index < 0 || index >= a.len {
+		return {}, false
+	}
+	return a.data[index], true
+}
+
+get_ptr_safe :: proc(a: ^$A/Small_Array($N, $T), index: int) -> (^T, bool) #no_bounds_check {
+	if index < 0 || index >= a.len {
+		return {}, false
+	}
+	return &a.data[index], true
+}
+
+set :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, item: T) {
+	a.data[index] = item
+}
+
+resize :: proc "contextless" (a: ^$A/Small_Array, length: int) {
+	a.len = min(length, builtin.len(a.data))
+}
+
+
+push_back :: proc "contextless" (a: ^$A/Small_Array($N, $T), item: T) -> bool {
+	if a.len < cap(a^) {
+		a.data[a.len] = item
+		a.len += 1
+		return true
+	}
+	return false
+}
+
+push_front :: proc "contextless" (a: ^$A/Small_Array($N, $T), item: T) -> bool {
+	if a.len < cap(a^) {
+		a.len += 1
+		data := slice(a)
+		copy(data[1:], data[:])
+		data[0] = item
+		return true
+	}
+	return false
+}
+
+pop_back :: proc "odin" (a: ^$A/Small_Array($N, $T), loc := #caller_location) -> T {
+	assert(condition=(N > 0 && a.len > 0), loc=loc)
+	item := a.data[a.len-1]
+	a.len -= 1
+	return item
+}
+
+pop_front :: proc "odin" (a: ^$A/Small_Array($N, $T), loc := #caller_location) -> T {
+	assert(condition=(N > 0 && a.len > 0), loc=loc)
+	item := a.data[0]
+	s := slice(a)
+	copy(s[:], s[1:])
+	a.len -= 1
+	return item
+}
+
+pop_back_safe :: proc "contextless" (a: ^$A/Small_Array($N, $T)) -> (item: T, ok: bool) {
+	if N > 0 && a.len > 0 {
+		item = a.data[a.len-1]
+		a.len -= 1
+		ok = true
+	}
+	return
+}
+
+pop_front_safe :: proc "contextless" (a: ^$A/Small_Array($N, $T)) -> (item: T, ok: bool) {
+	if N > 0 && a.len > 0 {
+		item = a.data[0]
+		s := slice(a)
+		copy(s[:], s[1:])
+		a.len -= 1
+		ok = true
+	}
+	return
+}
+
+consume :: proc "odin" (a: ^$A/Small_Array($N, $T), count: int, loc := #caller_location) {
+	assert(condition=a.len >= count, loc=loc)
+	a.len -= count
+}
+
+ordered_remove :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, loc := #caller_location) #no_bounds_check {
+    runtime.bounds_check_error_loc(loc, index, a.len)
+    if index+1 < a.len {
+		copy(a.data[index:], a.data[index+1:])
+	}
+	a.len -= 1
+}
+
+unordered_remove :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, loc := #caller_location) #no_bounds_check {
+    runtime.bounds_check_error_loc(loc, index, a.len)
+	n := a.len-1
+	if index != n {
+		a.data[index] = a.data[n]
+	}
+    a.len -= 1
+}
+
+clear :: proc "contextless" (a: ^$A/Small_Array($N, $T)) {
+	resize(a, 0)
+}
+
+push_back_elems :: proc "contextless" (a: ^$A/Small_Array($N, $T), items: ..T) {
+	n := copy(a.data[a.len:], items[:])
+	a.len += n
+}
+
+inject_at :: proc "contextless" (a: ^$A/Small_Array($N, $T), item: T, index: int) -> bool #no_bounds_check {
+	if a.len < cap(a^) && index >= 0 && index <= len(a^) {
+		a.len += 1
+		for i := a.len - 1; i >= index + 1; i -= 1 {
+			a.data[i] = a.data[i - 1]
+		}
+		a.data[index] = item
+		return true
+	}
+	return false
+}
+
+append_elem  :: push_back
+append_elems :: push_back_elems
+push   :: proc{push_back, push_back_elems}
+append :: proc{push_back, push_back_elems}
@@ -0,0 +1,98 @@
+// The following is a generic O(V+E) topological sorter implementation.
+// This is the fastest known method for topological sorting and Odin's
+// map type is being used to accelerate lookups.
+package container_topological_sort
+
+import "core:intrinsics"
+import "core:runtime"
+_ :: intrinsics
+_ :: runtime
+
+
+Relations :: struct($K: typeid) where intrinsics.type_is_valid_map_key(K) {
+	dependents:   map[K]bool,
+	dependencies: int,
+}
+
+Sorter :: struct(K: typeid) where intrinsics.type_is_valid_map_key(K)  {
+	relations: map[K]Relations(K),
+	dependents_allocator: runtime.Allocator,
+}
+
+@(private="file")
+make_relations :: proc(sorter: ^$S/Sorter($K)) -> (r: Relations(K)) {
+	r.dependents.allocator = sorter.dependents_allocator
+	return
+}
+
+
+init :: proc(sorter: ^$S/Sorter($K)) {
+	sorter.relations = make(map[K]Relations(K))
+	sorter.dependents_allocator = context.allocator
+}
+
+destroy :: proc(sorter: ^$S/Sorter($K)) {
+	for _, v in sorter.relations {
+		delete(v.dependents)
+	}
+	delete(sorter.relations)
+}
+
+add_key :: proc(sorter: ^$S/Sorter($K), key: K) -> bool {
+	if key in sorter.relations {
+		return false
+	}
+	sorter.relations[key] = make_relations(sorter)
+	return true
+}
+
+add_dependency :: proc(sorter: ^$S/Sorter($K), key, dependency: K) -> bool {
+	if key == dependency {
+		return false
+	}
+
+	find := &sorter.relations[dependency]
+	if find == nil {
+		find = map_insert(&sorter.relations, dependency, make_relations(sorter))
+	}
+
+	if find.dependents[key] {
+		return true
+	}
+	find.dependents[key] = true
+
+ 	find = &sorter.relations[key]
+	if find == nil {
+		find = map_insert(&sorter.relations, key, make_relations(sorter))
+	}
+
+	find.dependencies += 1
+
+	return true
+}
+
+sort :: proc(sorter: ^$S/Sorter($K)) -> (sorted, cycled: [dynamic]K) {
+	relations := &sorter.relations
+
+	for k, v in relations {
+		if v.dependencies == 0 {
+			append(&sorted, 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 k, v in relations {
+		if v.dependencies != 0 {
+			append(&cycled, k)
+		}
+	}
+
+	return
+}
@@ -0,0 +1,95 @@
+# crypto
+A crypto library for the Odin language
+
+## Supported
+This library offers various algorithms implemented in Odin.
+Please see the chart below for the options.  
+
+## Hashing algorithms
+| Algorithm                                                                                                    |                  |
+|:-------------------------------------------------------------------------------------------------------------|:-----------------|
+| [BLAKE](https://web.archive.org/web/20190915215948/https://131002.net/blake)                                 | &#10004;&#65039; |
+| [BLAKE2B](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | &#10004;&#65039; |
+| [BLAKE2S](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | &#10004;&#65039; |
+| [GOST](https://datatracker.ietf.org/doc/html/rfc5831)                                                        | &#10004;&#65039; |
+| [Grøstl](http://www.groestl.info/Groestl.zip)                                                                | &#10004;&#65039; |
+| [HAVAL](https://web.archive.org/web/20150111210116/http://labs.calyptix.com/haval.php)                       | &#10004;&#65039; |
+| [JH](https://www3.ntu.edu.sg/home/wuhj/research/jh/index.html)                                               | &#10004;&#65039; |
+| [Keccak](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                           | &#10004;&#65039; |
+| [MD2](https://datatracker.ietf.org/doc/html/rfc1319)                                                         | &#10004;&#65039; |
+| [MD4](https://datatracker.ietf.org/doc/html/rfc1320)                                                         | &#10004;&#65039; |
+| [MD5](https://datatracker.ietf.org/doc/html/rfc1321)                                                         | &#10004;&#65039; |
+| [RIPEMD](https://homes.esat.kuleuven.be/~bosselae/ripemd160.html)                                            | &#10004;&#65039; |
+| [SHA-1](https://datatracker.ietf.org/doc/html/rfc3174)                                                       | &#10004;&#65039; |
+| [SHA-2](https://csrc.nist.gov/csrc/media/publications/fips/180/2/archive/2002-08-01/documents/fips180-2.pdf) | &#10004;&#65039; |
+| [SHA-3](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | &#10004;&#65039; |
+| [SHAKE](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | &#10004;&#65039; |
+| [SM3](https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02)                                           | &#10004;&#65039; |
+| [Streebog](https://datatracker.ietf.org/doc/html/rfc6986)                                                    | &#10004;&#65039; |
+| [Tiger](https://www.cs.technion.ac.il/~biham/Reports/Tiger/)                                                 | &#10004;&#65039; |
+| [Tiger2](https://www.cs.technion.ac.il/~biham/Reports/Tiger/)                                                | &#10004;&#65039; |
+| [Whirlpool](https://web.archive.org/web/20171129084214/http://www.larc.usp.br/~pbarreto/WhirlpoolPage.html)  | &#10004;&#65039; |
+
+#### 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)
+
+\* On some algorithms there is another part to the name, since they might offer control about additional parameters.  
+For instance, `HAVAL` offers different sizes as well as three different round amounts.  
+Computing a 256-bit hash with 3 rounds is therefore achieved by calling `haval.hash_256_3(...)`.
+
+#### Low level API
+The above mentioned procedures internally call three procedures: `init`, `update` and `final`.
+You may also directly call them, if you wish.
+
+#### Example
+```odin
+package crypto_example
+
+// Import the desired package
+import "core:crypto/md4"
+
+main :: proc() {
+    input := "foo"
+
+    // Compute the hash, using the high level API
+    computed_hash := md4.hash(input)
+
+    // 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.
@@ -0,0 +1,35 @@
+# fiat
+
+This package contains low level arithmetic required to implement certain
+cryptographic primitives, ported from the [fiat-crypto project][1]
+along with some higher-level helpers.
+
+## Notes
+
+fiat-crypto gives the choice of 3 licenses for derived works.  The 1-Clause
+BSD license is chosen as it is compatible with Odin's existing licensing.
+
+The routines are intended to be timing-safe, as long as the underlying
+integer arithmetic is constant time.  This is true on most systems commonly
+used today, with the notable exception of WASM.
+
+While fiat-crypto provides both output targeting both 32-bit and 64-bit
+architectures, only the 64-bit versions were used, as 32-bit architectures
+are becoming increasingly uncommon and irrelevant.
+
+With the current Odin syntax, the Go output is trivially ported in most
+cases and was used as the basis of the port.
+
+In the future, it would be better to auto-generate Odin either directly
+by adding an appropriate code-gen backend written in Coq, or perhaps by
+parsing the JSON output.
+
+As this is a port rather than autogenerated output, none of fiat-crypto's
+formal verification guarantees apply, unless it is possible to prove binary
+equivalence.
+
+For the most part, alterations to the base fiat-crypto generated code was
+kept to a minimum, to aid auditability.  This results in a somewhat
+idiosyncratic style, and in some cases minor performance penalties.
+
+[1]: https://github.com/mit-plv/fiat-crypto
@@ -0,0 +1,26 @@
+package fiat
+
+// This package provides various helpers and types common to all of the
+// fiat-crypto derived backends.
+
+// This code only works on a two's complement system.
+#assert((-1 & 3) == 3)
+
+u1 :: distinct u8
+i1 :: distinct i8
+
+@(optimization_mode="none")
+cmovznz_u64 :: proc "contextless" (arg1: u1, arg2, arg3: u64) -> (out1: u64) {
+	x1 := (u64(arg1) * 0xffffffffffffffff)
+	x2 := ((x1 & arg3) | ((~x1) & arg2))
+	out1 = x2
+	return
+}
+
+@(optimization_mode="none")
+cmovznz_u32 :: proc "contextless" (arg1: u1, arg2, arg3: u32) -> (out1: u32) {
+	x1 := (u32(arg1) * 0xffffffff)
+	x2 := ((x1 & arg3) | ((~x1) & arg2))
+	out1 = x2
+	return
+}
@@ -0,0 +1,138 @@
+package field_curve25519
+
+import "core:crypto"
+import "core:mem"
+
+fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
+	return transmute(^Loose_Field_Element)(arg1)
+}
+
+fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_Field_Element) -> ^Tight_Field_Element {
+	return transmute(^Tight_Field_Element)(arg1)
+}
+
+fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	// Ignore the unused bit by copying the input and masking the bit off
+	// prior to deserialization.
+	tmp1: [32]byte = ---
+	copy_slice(tmp1[:], arg1[:])
+	tmp1[31] &= 127
+
+	_fe_from_bytes(out1, &tmp1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+}
+
+fe_equal :: proc "contextless" (arg1, arg2: ^Tight_Field_Element) -> int {
+	tmp2: [32]byte = ---
+
+	fe_to_bytes(&tmp2, arg2)
+	ret := fe_equal_bytes(arg1, &tmp2)
+
+	mem.zero_explicit(&tmp2, size_of(tmp2))
+
+	return ret
+}
+
+fe_equal_bytes :: proc "contextless" (arg1: ^Tight_Field_Element, arg2: ^[32]byte) -> int {
+	tmp1: [32]byte = ---
+
+	fe_to_bytes(&tmp1, arg1)
+
+	ret := crypto.compare_constant_time(tmp1[:], arg2[:])
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+
+	return ret
+}
+
+fe_carry_pow2k :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element, arg2: uint) {
+	// Special case: `arg1^(2 * 0) = 1`, though this should never happen.
+	if arg2 == 0 {
+		fe_one(out1)
+		return
+	}
+
+	fe_carry_square(out1, arg1)
+	for _ in 1..<arg2 {
+		fe_carry_square(out1, fe_relax_cast(out1))
+	}
+}
+
+fe_carry_opp :: #force_inline proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	fe_opp(fe_relax_cast(out1), arg1)
+	fe_carry(out1, fe_relax_cast(out1))
+}
+
+fe_carry_invsqrt :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) -> int {
+	// Inverse square root taken from Monocypher.
+
+	tmp1, tmp2, tmp3: Tight_Field_Element = ---, ---, ---
+
+	// t0 = x^((p-5)/8)
+	// Can be achieved with a simple double & add ladder,
+	// but it would be slower.
+	fe_carry_pow2k(&tmp1, arg1, 1)
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 2)
+	fe_carry_mul(&tmp2, arg1, fe_relax_cast(&tmp2))
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp1), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp1, fe_relax_cast(&tmp1), 1)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 5)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 10)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp3, fe_relax_cast(&tmp2), 20)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp3), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp2), 10)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp1), 50)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp3, fe_relax_cast(&tmp2), 100)
+	fe_carry_mul(&tmp2, fe_relax_cast(&tmp3), fe_relax_cast(&tmp2))
+	fe_carry_pow2k(&tmp2, fe_relax_cast(&tmp2), 50)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp2), fe_relax_cast(&tmp1))
+	fe_carry_pow2k(&tmp1, fe_relax_cast(&tmp1), 2)
+	fe_carry_mul(&tmp1, fe_relax_cast(&tmp1), arg1)
+
+	// quartic = x^((p-1)/4)
+	quartic := &tmp2
+	fe_carry_square(quartic, fe_relax_cast(&tmp1))
+	fe_carry_mul(quartic, fe_relax_cast(quartic), arg1)
+
+	// Serialize quartic once to save on repeated serialization/sanitization.
+	quartic_buf: [32]byte = ---
+	fe_to_bytes(&quartic_buf, quartic)
+	check := &tmp3
+
+	fe_one(check)
+	p1 := fe_equal_bytes(check, &quartic_buf)
+	fe_carry_opp(check, check)
+	m1 := fe_equal_bytes(check, &quartic_buf)
+	fe_carry_opp(check, &SQRT_M1)
+	ms := fe_equal_bytes(check, &quartic_buf)
+
+	// if quartic == -1 or sqrt(-1)
+	// then  isr = x^((p-1)/4) * sqrt(-1)
+	// else  isr = x^((p-1)/4)
+	fe_carry_mul(out1, fe_relax_cast(&tmp1), fe_relax_cast(&SQRT_M1))
+	fe_cond_assign(out1, &tmp1, (m1|ms) ~ 1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+	mem.zero_explicit(&tmp2, size_of(tmp2))
+	mem.zero_explicit(&tmp3, size_of(tmp3))
+	mem.zero_explicit(&quartic_buf, size_of(quartic_buf))
+
+	return p1 | m1
+}
+
+fe_carry_inv :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	tmp1: Tight_Field_Element
+
+	fe_carry_square(&tmp1, arg1)
+	_ = fe_carry_invsqrt(&tmp1, fe_relax_cast(&tmp1))
+	fe_carry_square(&tmp1, fe_relax_cast(&tmp1))
+	fe_carry_mul(out1, fe_relax_cast(&tmp1), arg1)
+
+	mem.zero_explicit(&tmp1, size_of(tmp1))
+}
@@ -0,0 +1,618 @@
+// The BSD 1-Clause License (BSD-1-Clause)
+//
+// Copyright (c) 2015-2020 the fiat-crypto authors (see the AUTHORS file)
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     1. Redistributions of source code must retain the above copyright
+//        notice, this list of conditions and the following disclaimer.
+//
+// THIS SOFTWARE IS PROVIDED BY the fiat-crypto authors "AS IS"
+// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL Berkeley Software Design,
+// Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+package field_curve25519
+
+// The file provides arithmetic on the field Z/(2^255-19) using
+// unsaturated 64-bit integer arithmetic.  It is derived primarily
+// from the machine generated Golang output from the fiat-crypto project.
+//
+// While the base implementation is provably correct, this implementation
+// makes no such claims as the port and optimizations were done by hand.
+// At some point, it may be worth adding support to fiat-crypto for
+// generating Odin output.
+//
+// TODO:
+//  * When fiat-crypto supports it, using a saturated 64-bit limbs
+//    instead of 51-bit limbs will be faster, though the gains are
+//    minimal unless adcx/adox/mulx are used.
+
+import fiat "core:crypto/_fiat"
+import "core:math/bits"
+
+Loose_Field_Element :: distinct [5]u64
+Tight_Field_Element :: distinct [5]u64
+
+SQRT_M1 := Tight_Field_Element{
+	1718705420411056,
+	234908883556509,
+	2233514472574048,
+	2117202627021982,
+	765476049583133,
+}
+
+_addcarryx_u51 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0x7ffffffffffff)
+	x3 := fiat.u1((x1 >> 51))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u51 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 51))
+	x3 := (u64(x1) & 0x7ffffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+fe_carry_mul :: proc (out1: ^Tight_Field_Element, arg1, arg2: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(arg1[4], (arg2[4] * 0x13))
+	x4, x3 := bits.mul_u64(arg1[4], (arg2[3] * 0x13))
+	x6, x5 := bits.mul_u64(arg1[4], (arg2[2] * 0x13))
+	x8, x7 := bits.mul_u64(arg1[4], (arg2[1] * 0x13))
+	x10, x9 := bits.mul_u64(arg1[3], (arg2[4] * 0x13))
+	x12, x11 := bits.mul_u64(arg1[3], (arg2[3] * 0x13))
+	x14, x13 := bits.mul_u64(arg1[3], (arg2[2] * 0x13))
+	x16, x15 := bits.mul_u64(arg1[2], (arg2[4] * 0x13))
+	x18, x17 := bits.mul_u64(arg1[2], (arg2[3] * 0x13))
+	x20, x19 := bits.mul_u64(arg1[1], (arg2[4] * 0x13))
+	x22, x21 := bits.mul_u64(arg1[4], arg2[0])
+	x24, x23 := bits.mul_u64(arg1[3], arg2[1])
+	x26, x25 := bits.mul_u64(arg1[3], arg2[0])
+	x28, x27 := bits.mul_u64(arg1[2], arg2[2])
+	x30, x29 := bits.mul_u64(arg1[2], arg2[1])
+	x32, x31 := bits.mul_u64(arg1[2], arg2[0])
+	x34, x33 := bits.mul_u64(arg1[1], arg2[3])
+	x36, x35 := bits.mul_u64(arg1[1], arg2[2])
+	x38, x37 := bits.mul_u64(arg1[1], arg2[1])
+	x40, x39 := bits.mul_u64(arg1[1], arg2[0])
+	x42, x41 := bits.mul_u64(arg1[0], arg2[4])
+	x44, x43 := bits.mul_u64(arg1[0], arg2[3])
+	x46, x45 := bits.mul_u64(arg1[0], arg2[2])
+	x48, x47 := bits.mul_u64(arg1[0], arg2[1])
+	x50, x49 := bits.mul_u64(arg1[0], arg2[0])
+	x51, x52 := bits.add_u64(x13, x7, u64(0x0))
+	x53, _ := bits.add_u64(x14, x8, u64(fiat.u1(x52)))
+	x55, x56 := bits.add_u64(x17, x51, u64(0x0))
+	x57, _ := bits.add_u64(x18, x53, u64(fiat.u1(x56)))
+	x59, x60 := bits.add_u64(x19, x55, u64(0x0))
+	x61, _ := bits.add_u64(x20, x57, u64(fiat.u1(x60)))
+	x63, x64 := bits.add_u64(x49, x59, u64(0x0))
+	x65, _ := bits.add_u64(x50, x61, u64(fiat.u1(x64)))
+	x67 := ((x63 >> 51) | ((x65 << 13) & 0xffffffffffffffff))
+	x68 := (x63 & 0x7ffffffffffff)
+	x69, x70 := bits.add_u64(x23, x21, u64(0x0))
+	x71, _ := bits.add_u64(x24, x22, u64(fiat.u1(x70)))
+	x73, x74 := bits.add_u64(x27, x69, u64(0x0))
+	x75, _ := bits.add_u64(x28, x71, u64(fiat.u1(x74)))
+	x77, x78 := bits.add_u64(x33, x73, u64(0x0))
+	x79, _ := bits.add_u64(x34, x75, u64(fiat.u1(x78)))
+	x81, x82 := bits.add_u64(x41, x77, u64(0x0))
+	x83, _ := bits.add_u64(x42, x79, u64(fiat.u1(x82)))
+	x85, x86 := bits.add_u64(x25, x1, u64(0x0))
+	x87, _ := bits.add_u64(x26, x2, u64(fiat.u1(x86)))
+	x89, x90 := bits.add_u64(x29, x85, u64(0x0))
+	x91, _ := bits.add_u64(x30, x87, u64(fiat.u1(x90)))
+	x93, x94 := bits.add_u64(x35, x89, u64(0x0))
+	x95, _ := bits.add_u64(x36, x91, u64(fiat.u1(x94)))
+	x97, x98 := bits.add_u64(x43, x93, u64(0x0))
+	x99, _ := bits.add_u64(x44, x95, u64(fiat.u1(x98)))
+	x101, x102 := bits.add_u64(x9, x3, u64(0x0))
+	x103, _ := bits.add_u64(x10, x4, u64(fiat.u1(x102)))
+	x105, x106 := bits.add_u64(x31, x101, u64(0x0))
+	x107, _ := bits.add_u64(x32, x103, u64(fiat.u1(x106)))
+	x109, x110 := bits.add_u64(x37, x105, u64(0x0))
+	x111, _ := bits.add_u64(x38, x107, u64(fiat.u1(x110)))
+	x113, x114 := bits.add_u64(x45, x109, u64(0x0))
+	x115, _ := bits.add_u64(x46, x111, u64(fiat.u1(x114)))
+	x117, x118 := bits.add_u64(x11, x5, u64(0x0))
+	x119, _ := bits.add_u64(x12, x6, u64(fiat.u1(x118)))
+	x121, x122 := bits.add_u64(x15, x117, u64(0x0))
+	x123, _ := bits.add_u64(x16, x119, u64(fiat.u1(x122)))
+	x125, x126 := bits.add_u64(x39, x121, u64(0x0))
+	x127, _ := bits.add_u64(x40, x123, u64(fiat.u1(x126)))
+	x129, x130 := bits.add_u64(x47, x125, u64(0x0))
+	x131, _ := bits.add_u64(x48, x127, u64(fiat.u1(x130)))
+	x133, x134 := bits.add_u64(x67, x129, u64(0x0))
+	x135 := (u64(fiat.u1(x134)) + x131)
+	x136 := ((x133 >> 51) | ((x135 << 13) & 0xffffffffffffffff))
+	x137 := (x133 & 0x7ffffffffffff)
+	x138, x139 := bits.add_u64(x136, x113, u64(0x0))
+	x140 := (u64(fiat.u1(x139)) + x115)
+	x141 := ((x138 >> 51) | ((x140 << 13) & 0xffffffffffffffff))
+	x142 := (x138 & 0x7ffffffffffff)
+	x143, x144 := bits.add_u64(x141, x97, u64(0x0))
+	x145 := (u64(fiat.u1(x144)) + x99)
+	x146 := ((x143 >> 51) | ((x145 << 13) & 0xffffffffffffffff))
+	x147 := (x143 & 0x7ffffffffffff)
+	x148, x149 := bits.add_u64(x146, x81, u64(0x0))
+	x150 := (u64(fiat.u1(x149)) + x83)
+	x151 := ((x148 >> 51) | ((x150 << 13) & 0xffffffffffffffff))
+	x152 := (x148 & 0x7ffffffffffff)
+	x153 := (x151 * 0x13)
+	x154 := (x68 + x153)
+	x155 := (x154 >> 51)
+	x156 := (x154 & 0x7ffffffffffff)
+	x157 := (x155 + x137)
+	x158 := fiat.u1((x157 >> 51))
+	x159 := (x157 & 0x7ffffffffffff)
+	x160 := (u64(x158) + x142)
+	out1[0] = x156
+	out1[1] = x159
+	out1[2] = x160
+	out1[3] = x147
+	out1[4] = x152
+}
+
+fe_carry_square :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := (arg1[4] * 0x13)
+	x2 := (x1 * 0x2)
+	x3 := (arg1[4] * 0x2)
+	x4 := (arg1[3] * 0x13)
+	x5 := (x4 * 0x2)
+	x6 := (arg1[3] * 0x2)
+	x7 := (arg1[2] * 0x2)
+	x8 := (arg1[1] * 0x2)
+	x10, x9 := bits.mul_u64(arg1[4], x1)
+	x12, x11 := bits.mul_u64(arg1[3], x2)
+	x14, x13 := bits.mul_u64(arg1[3], x4)
+	x16, x15 := bits.mul_u64(arg1[2], x2)
+	x18, x17 := bits.mul_u64(arg1[2], x5)
+	x20, x19 := bits.mul_u64(arg1[2], arg1[2])
+	x22, x21 := bits.mul_u64(arg1[1], x2)
+	x24, x23 := bits.mul_u64(arg1[1], x6)
+	x26, x25 := bits.mul_u64(arg1[1], x7)
+	x28, x27 := bits.mul_u64(arg1[1], arg1[1])
+	x30, x29 := bits.mul_u64(arg1[0], x3)
+	x32, x31 := bits.mul_u64(arg1[0], x6)
+	x34, x33 := bits.mul_u64(arg1[0], x7)
+	x36, x35 := bits.mul_u64(arg1[0], x8)
+	x38, x37 := bits.mul_u64(arg1[0], arg1[0])
+	x39, x40 := bits.add_u64(x21, x17, u64(0x0))
+	x41, _ := bits.add_u64(x22, x18, u64(fiat.u1(x40)))
+	x43, x44 := bits.add_u64(x37, x39, u64(0x0))
+	x45, _ := bits.add_u64(x38, x41, u64(fiat.u1(x44)))
+	x47 := ((x43 >> 51) | ((x45 << 13) & 0xffffffffffffffff))
+	x48 := (x43 & 0x7ffffffffffff)
+	x49, x50 := bits.add_u64(x23, x19, u64(0x0))
+	x51, _ := bits.add_u64(x24, x20, u64(fiat.u1(x50)))
+	x53, x54 := bits.add_u64(x29, x49, u64(0x0))
+	x55, _ := bits.add_u64(x30, x51, u64(fiat.u1(x54)))
+	x57, x58 := bits.add_u64(x25, x9, u64(0x0))
+	x59, _ := bits.add_u64(x26, x10, u64(fiat.u1(x58)))
+	x61, x62 := bits.add_u64(x31, x57, u64(0x0))
+	x63, _ := bits.add_u64(x32, x59, u64(fiat.u1(x62)))
+	x65, x66 := bits.add_u64(x27, x11, u64(0x0))
+	x67, _ := bits.add_u64(x28, x12, u64(fiat.u1(x66)))
+	x69, x70 := bits.add_u64(x33, x65, u64(0x0))
+	x71, _ := bits.add_u64(x34, x67, u64(fiat.u1(x70)))
+	x73, x74 := bits.add_u64(x15, x13, u64(0x0))
+	x75, _ := bits.add_u64(x16, x14, u64(fiat.u1(x74)))
+	x77, x78 := bits.add_u64(x35, x73, u64(0x0))
+	x79, _ := bits.add_u64(x36, x75, u64(fiat.u1(x78)))
+	x81, x82 := bits.add_u64(x47, x77, u64(0x0))
+	x83 := (u64(fiat.u1(x82)) + x79)
+	x84 := ((x81 >> 51) | ((x83 << 13) & 0xffffffffffffffff))
+	x85 := (x81 & 0x7ffffffffffff)
+	x86, x87 := bits.add_u64(x84, x69, u64(0x0))
+	x88 := (u64(fiat.u1(x87)) + x71)
+	x89 := ((x86 >> 51) | ((x88 << 13) & 0xffffffffffffffff))
+	x90 := (x86 & 0x7ffffffffffff)
+	x91, x92 := bits.add_u64(x89, x61, u64(0x0))
+	x93 := (u64(fiat.u1(x92)) + x63)
+	x94 := ((x91 >> 51) | ((x93 << 13) & 0xffffffffffffffff))
+	x95 := (x91 & 0x7ffffffffffff)
+	x96, x97 := bits.add_u64(x94, x53, u64(0x0))
+	x98 := (u64(fiat.u1(x97)) + x55)
+	x99 := ((x96 >> 51) | ((x98 << 13) & 0xffffffffffffffff))
+	x100 := (x96 & 0x7ffffffffffff)
+	x101 := (x99 * 0x13)
+	x102 := (x48 + x101)
+	x103 := (x102 >> 51)
+	x104 := (x102 & 0x7ffffffffffff)
+	x105 := (x103 + x85)
+	x106 := fiat.u1((x105 >> 51))
+	x107 := (x105 & 0x7ffffffffffff)
+	x108 := (u64(x106) + x90)
+	out1[0] = x104
+	out1[1] = x107
+	out1[2] = x108
+	out1[3] = x95
+	out1[4] = x100
+}
+
+fe_carry :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := arg1[0]
+	x2 := ((x1 >> 51) + arg1[1])
+	x3 := ((x2 >> 51) + arg1[2])
+	x4 := ((x3 >> 51) + arg1[3])
+	x5 := ((x4 >> 51) + arg1[4])
+	x6 := ((x1 & 0x7ffffffffffff) + ((x5 >> 51) * 0x13))
+	x7 := (u64(fiat.u1((x6 >> 51))) + (x2 & 0x7ffffffffffff))
+	x8 := (x6 & 0x7ffffffffffff)
+	x9 := (x7 & 0x7ffffffffffff)
+	x10 := (u64(fiat.u1((x7 >> 51))) + (x3 & 0x7ffffffffffff))
+	x11 := (x4 & 0x7ffffffffffff)
+	x12 := (x5 & 0x7ffffffffffff)
+	out1[0] = x8
+	out1[1] = x9
+	out1[2] = x10
+	out1[3] = x11
+	out1[4] = x12
+}
+
+fe_add :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := (arg1[0] + arg2[0])
+	x2 := (arg1[1] + arg2[1])
+	x3 := (arg1[2] + arg2[2])
+	x4 := (arg1[3] + arg2[3])
+	x5 := (arg1[4] + arg2[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_sub :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := ((0xfffffffffffda + arg1[0]) - arg2[0])
+	x2 := ((0xffffffffffffe + arg1[1]) - arg2[1])
+	x3 := ((0xffffffffffffe + arg1[2]) - arg2[2])
+	x4 := ((0xffffffffffffe + arg1[3]) - arg2[3])
+	x5 := ((0xffffffffffffe + arg1[4]) - arg2[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_opp :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := (0xfffffffffffda - arg1[0])
+	x2 := (0xffffffffffffe - arg1[1])
+	x3 := (0xffffffffffffe - arg1[2])
+	x4 := (0xffffffffffffe - arg1[3])
+	x5 := (0xffffffffffffe - arg1[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+@(optimization_mode="none")
+fe_cond_assign :: #force_no_inline proc "contextless" (out1, arg1: ^Tight_Field_Element, arg2: int) {
+	x1 := fiat.cmovznz_u64(fiat.u1(arg2), out1[0], arg1[0])
+	x2 := fiat.cmovznz_u64(fiat.u1(arg2), out1[1], arg1[1])
+	x3 := fiat.cmovznz_u64(fiat.u1(arg2), out1[2], arg1[2])
+	x4 := fiat.cmovznz_u64(fiat.u1(arg2), out1[3], arg1[3])
+	x5 := fiat.cmovznz_u64(fiat.u1(arg2), out1[4], arg1[4])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_to_bytes :: proc "contextless" (out1: ^[32]byte, arg1: ^Tight_Field_Element) {
+	x1, x2 := _subborrowx_u51(0x0, arg1[0], 0x7ffffffffffed)
+	x3, x4 := _subborrowx_u51(x2, arg1[1], 0x7ffffffffffff)
+	x5, x6 := _subborrowx_u51(x4, arg1[2], 0x7ffffffffffff)
+	x7, x8 := _subborrowx_u51(x6, arg1[3], 0x7ffffffffffff)
+	x9, x10 := _subborrowx_u51(x8, arg1[4], 0x7ffffffffffff)
+	x11 := fiat.cmovznz_u64(x10, u64(0x0), 0xffffffffffffffff)
+	x12, x13 := _addcarryx_u51(0x0, x1, (x11 & 0x7ffffffffffed))
+	x14, x15 := _addcarryx_u51(x13, x3, (x11 & 0x7ffffffffffff))
+	x16, x17 := _addcarryx_u51(x15, x5, (x11 & 0x7ffffffffffff))
+	x18, x19 := _addcarryx_u51(x17, x7, (x11 & 0x7ffffffffffff))
+	x20, _ := _addcarryx_u51(x19, x9, (x11 & 0x7ffffffffffff))
+	x22 := (x20 << 4)
+	x23 := (x18 * u64(0x2))
+	x24 := (x16 << 6)
+	x25 := (x14 << 3)
+	x26 := (u8(x12) & 0xff)
+	x27 := (x12 >> 8)
+	x28 := (u8(x27) & 0xff)
+	x29 := (x27 >> 8)
+	x30 := (u8(x29) & 0xff)
+	x31 := (x29 >> 8)
+	x32 := (u8(x31) & 0xff)
+	x33 := (x31 >> 8)
+	x34 := (u8(x33) & 0xff)
+	x35 := (x33 >> 8)
+	x36 := (u8(x35) & 0xff)
+	x37 := u8((x35 >> 8))
+	x38 := (x25 + u64(x37))
+	x39 := (u8(x38) & 0xff)
+	x40 := (x38 >> 8)
+	x41 := (u8(x40) & 0xff)
+	x42 := (x40 >> 8)
+	x43 := (u8(x42) & 0xff)
+	x44 := (x42 >> 8)
+	x45 := (u8(x44) & 0xff)
+	x46 := (x44 >> 8)
+	x47 := (u8(x46) & 0xff)
+	x48 := (x46 >> 8)
+	x49 := (u8(x48) & 0xff)
+	x50 := u8((x48 >> 8))
+	x51 := (x24 + u64(x50))
+	x52 := (u8(x51) & 0xff)
+	x53 := (x51 >> 8)
+	x54 := (u8(x53) & 0xff)
+	x55 := (x53 >> 8)
+	x56 := (u8(x55) & 0xff)
+	x57 := (x55 >> 8)
+	x58 := (u8(x57) & 0xff)
+	x59 := (x57 >> 8)
+	x60 := (u8(x59) & 0xff)
+	x61 := (x59 >> 8)
+	x62 := (u8(x61) & 0xff)
+	x63 := (x61 >> 8)
+	x64 := (u8(x63) & 0xff)
+	x65 := fiat.u1((x63 >> 8))
+	x66 := (x23 + u64(x65))
+	x67 := (u8(x66) & 0xff)
+	x68 := (x66 >> 8)
+	x69 := (u8(x68) & 0xff)
+	x70 := (x68 >> 8)
+	x71 := (u8(x70) & 0xff)
+	x72 := (x70 >> 8)
+	x73 := (u8(x72) & 0xff)
+	x74 := (x72 >> 8)
+	x75 := (u8(x74) & 0xff)
+	x76 := (x74 >> 8)
+	x77 := (u8(x76) & 0xff)
+	x78 := u8((x76 >> 8))
+	x79 := (x22 + u64(x78))
+	x80 := (u8(x79) & 0xff)
+	x81 := (x79 >> 8)
+	x82 := (u8(x81) & 0xff)
+	x83 := (x81 >> 8)
+	x84 := (u8(x83) & 0xff)
+	x85 := (x83 >> 8)
+	x86 := (u8(x85) & 0xff)
+	x87 := (x85 >> 8)
+	x88 := (u8(x87) & 0xff)
+	x89 := (x87 >> 8)
+	x90 := (u8(x89) & 0xff)
+	x91 := u8((x89 >> 8))
+	out1[0] = x26
+	out1[1] = x28
+	out1[2] = x30
+	out1[3] = x32
+	out1[4] = x34
+	out1[5] = x36
+	out1[6] = x39
+	out1[7] = x41
+	out1[8] = x43
+	out1[9] = x45
+	out1[10] = x47
+	out1[11] = x49
+	out1[12] = x52
+	out1[13] = x54
+	out1[14] = x56
+	out1[15] = x58
+	out1[16] = x60
+	out1[17] = x62
+	out1[18] = x64
+	out1[19] = x67
+	out1[20] = x69
+	out1[21] = x71
+	out1[22] = x73
+	out1[23] = x75
+	out1[24] = x77
+	out1[25] = x80
+	out1[26] = x82
+	out1[27] = x84
+	out1[28] = x86
+	out1[29] = x88
+	out1[30] = x90
+	out1[31] = x91
+}
+
+_fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	x1 := (u64(arg1[31]) << 44)
+	x2 := (u64(arg1[30]) << 36)
+	x3 := (u64(arg1[29]) << 28)
+	x4 := (u64(arg1[28]) << 20)
+	x5 := (u64(arg1[27]) << 12)
+	x6 := (u64(arg1[26]) << 4)
+	x7 := (u64(arg1[25]) << 47)
+	x8 := (u64(arg1[24]) << 39)
+	x9 := (u64(arg1[23]) << 31)
+	x10 := (u64(arg1[22]) << 23)
+	x11 := (u64(arg1[21]) << 15)
+	x12 := (u64(arg1[20]) << 7)
+	x13 := (u64(arg1[19]) << 50)
+	x14 := (u64(arg1[18]) << 42)
+	x15 := (u64(arg1[17]) << 34)
+	x16 := (u64(arg1[16]) << 26)
+	x17 := (u64(arg1[15]) << 18)
+	x18 := (u64(arg1[14]) << 10)
+	x19 := (u64(arg1[13]) << 2)
+	x20 := (u64(arg1[12]) << 45)
+	x21 := (u64(arg1[11]) << 37)
+	x22 := (u64(arg1[10]) << 29)
+	x23 := (u64(arg1[9]) << 21)
+	x24 := (u64(arg1[8]) << 13)
+	x25 := (u64(arg1[7]) << 5)
+	x26 := (u64(arg1[6]) << 48)
+	x27 := (u64(arg1[5]) << 40)
+	x28 := (u64(arg1[4]) << 32)
+	x29 := (u64(arg1[3]) << 24)
+	x30 := (u64(arg1[2]) << 16)
+	x31 := (u64(arg1[1]) << 8)
+	x32 := arg1[0]
+	x33 := (x31 + u64(x32))
+	x34 := (x30 + x33)
+	x35 := (x29 + x34)
+	x36 := (x28 + x35)
+	x37 := (x27 + x36)
+	x38 := (x26 + x37)
+	x39 := (x38 & 0x7ffffffffffff)
+	x40 := u8((x38 >> 51))
+	x41 := (x25 + u64(x40))
+	x42 := (x24 + x41)
+	x43 := (x23 + x42)
+	x44 := (x22 + x43)
+	x45 := (x21 + x44)
+	x46 := (x20 + x45)
+	x47 := (x46 & 0x7ffffffffffff)
+	x48 := u8((x46 >> 51))
+	x49 := (x19 + u64(x48))
+	x50 := (x18 + x49)
+	x51 := (x17 + x50)
+	x52 := (x16 + x51)
+	x53 := (x15 + x52)
+	x54 := (x14 + x53)
+	x55 := (x13 + x54)
+	x56 := (x55 & 0x7ffffffffffff)
+	x57 := u8((x55 >> 51))
+	x58 := (x12 + u64(x57))
+	x59 := (x11 + x58)
+	x60 := (x10 + x59)
+	x61 := (x9 + x60)
+	x62 := (x8 + x61)
+	x63 := (x7 + x62)
+	x64 := (x63 & 0x7ffffffffffff)
+	x65 := u8((x63 >> 51))
+	x66 := (x6 + u64(x65))
+	x67 := (x5 + x66)
+	x68 := (x4 + x67)
+	x69 := (x3 + x68)
+	x70 := (x2 + x69)
+	x71 := (x1 + x70)
+	out1[0] = x39
+	out1[1] = x47
+	out1[2] = x56
+	out1[3] = x64
+	out1[4] = x71
+}
+
+fe_relax :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	x4 := arg1[3]
+	x5 := arg1[4]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+fe_carry_scmul_121666 :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(0x1db42, arg1[4])
+	x4, x3 := bits.mul_u64(0x1db42, arg1[3])
+	x6, x5 := bits.mul_u64(0x1db42, arg1[2])
+	x8, x7 := bits.mul_u64(0x1db42, arg1[1])
+	x10, x9 := bits.mul_u64(0x1db42, arg1[0])
+	x11 := ((x9 >> 51) | ((x10 << 13) & 0xffffffffffffffff))
+	x12 := (x9 & 0x7ffffffffffff)
+	x13, x14 := bits.add_u64(x11, x7, u64(0x0))
+	x15 := (u64(fiat.u1(x14)) + x8)
+	x16 := ((x13 >> 51) | ((x15 << 13) & 0xffffffffffffffff))
+	x17 := (x13 & 0x7ffffffffffff)
+	x18, x19 := bits.add_u64(x16, x5, u64(0x0))
+	x20 := (u64(fiat.u1(x19)) + x6)
+	x21 := ((x18 >> 51) | ((x20 << 13) & 0xffffffffffffffff))
+	x22 := (x18 & 0x7ffffffffffff)
+	x23, x24 := bits.add_u64(x21, x3, u64(0x0))
+	x25 := (u64(fiat.u1(x24)) + x4)
+	x26 := ((x23 >> 51) | ((x25 << 13) & 0xffffffffffffffff))
+	x27 := (x23 & 0x7ffffffffffff)
+	x28, x29 := bits.add_u64(x26, x1, u64(0x0))
+	x30 := (u64(fiat.u1(x29)) + x2)
+	x31 := ((x28 >> 51) | ((x30 << 13) & 0xffffffffffffffff))
+	x32 := (x28 & 0x7ffffffffffff)
+	x33 := (x31 * 0x13)
+	x34 := (x12 + x33)
+	x35 := fiat.u1((x34 >> 51))
+	x36 := (x34 & 0x7ffffffffffff)
+	x37 := (u64(x35) + x17)
+	x38 := fiat.u1((x37 >> 51))
+	x39 := (x37 & 0x7ffffffffffff)
+	x40 := (u64(x38) + x22)
+	out1[0] = x36
+	out1[1] = x39
+	out1[2] = x40
+	out1[3] = x27
+	out1[4] = x32
+}
+
+// The following routines were added by hand, and do not come from fiat-crypto.
+
+fe_zero :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 0
+	out1[1] = 0
+	out1[2] = 0
+	out1[3] = 0
+	out1[4] = 0
+}
+
+fe_one :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 1
+	out1[1] = 0
+	out1[2] = 0
+	out1[3] = 0
+	out1[4] = 0
+}
+
+fe_set :: proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	x4 := arg1[3]
+	x5 := arg1[4]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+	out1[3] = x4
+	out1[4] = x5
+}
+
+@(optimization_mode="none")
+fe_cond_swap :: #force_no_inline proc "contextless" (out1, out2: ^Tight_Field_Element, arg1: int) {
+	mask := -u64(arg1)
+	x := (out1[0] ~ out2[0]) & mask
+	x1, y1 := out1[0] ~ x, out2[0] ~ x
+	x = (out1[1] ~ out2[1]) & mask
+	x2, y2 := out1[1] ~ x, out2[1] ~ x
+	x = (out1[2] ~ out2[2]) & mask
+	x3, y3 := out1[2] ~ x, out2[2] ~ x
+	x = (out1[3] ~ out2[3]) & mask
+	x4, y4 := out1[3] ~ x, out2[3] ~ x
+	x = (out1[4] ~ out2[4]) & mask
+	x5, y5 := out1[4] ~ x, out2[4] ~ x
+	out1[0], out2[0] = x1, y1
+	out1[1], out2[1] = x2, y2
+	out1[2], out2[2] = x3, y3
+	out1[3], out2[3] = x4, y4
+	out1[4], out2[4] = x5, y5
+}
@@ -0,0 +1,66 @@
+package field_poly1305
+
+import "core:crypto/util"
+import "core:mem"
+
+fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
+	return transmute(^Loose_Field_Element)(arg1)
+}
+
+fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_Field_Element) -> ^Tight_Field_Element {
+	return transmute(^Tight_Field_Element)(arg1)
+}
+
+fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, arg2: byte, sanitize: bool = true) {
+	// fiat-crypto's deserialization routine effectively processes a
+	// single byte at a time, and wants 256-bits of input for a value
+	// that will be 128-bits or 129-bits.
+	//
+	// This is somewhat cumbersome to use, so at a minimum a wrapper
+	// makes implementing the actual MAC block processing considerably
+	// neater.
+
+	assert(len(arg1) == 16)
+
+	when ODIN_ARCH == .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]
+
+		// This is inspired by poly1305-donna, though adjustments were
+		// made since a Tight_Field_Element's limbs are 44-bits, 43-bits,
+		// and 43-bits wide.
+		//
+		// Note: This could be transplated into fe_from_u64s, but that
+		// code is called once per MAC, and is non-criticial path.
+		hibit := u64(arg2) << 41 // arg2 << 128
+		out1[0] = lo & 0xfffffffffff
+		out1[1] = ((lo >> 44) | (hi << 20)) & 0x7ffffffffff
+		out1[2] = ((hi >> 23) & 0x7ffffffffff) | hibit
+	} else {
+		tmp: [32]byte
+		copy_slice(tmp[0:16], arg1[:])
+		tmp[16] = arg2
+
+		_fe_from_bytes(out1, &tmp)
+		if sanitize {
+			// This is used to deserialize `s` which is confidential.
+			mem.zero_explicit(&tmp, size_of(tmp))
+		}
+	}
+}
+
+fe_from_u64s :: proc "contextless" (out1: ^Tight_Field_Element, lo, hi: u64) {
+	tmp: [32]byte
+	util.PUT_U64_LE(tmp[0:8], lo)
+	util.PUT_U64_LE(tmp[8:16], hi)
+
+	_fe_from_bytes(out1, &tmp)
+
+	// This routine is only used to deserialize `r` which is confidential.
+	mem.zero_explicit(&tmp, size_of(tmp))
+}
@@ -0,0 +1,358 @@
+// The BSD 1-Clause License (BSD-1-Clause)
+//
+// Copyright (c) 2015-2020 the fiat-crypto authors (see the AUTHORS file)
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     1. Redistributions of source code must retain the above copyright
+//        notice, this list of conditions and the following disclaimer.
+//
+// THIS SOFTWARE IS PROVIDED BY the fiat-crypto authors "AS IS"
+// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL Berkeley Software Design,
+// Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+package field_poly1305
+
+// This file provides arithmetic on the field Z/(2^130 - 5) using
+// unsaturated 64-bit integer arithmetic.  It is derived primarily
+// from the machine generate Golang output from the fiat-crypto project.
+//
+// While the base implementation is provably correct, this implementation
+// makes no such claims as the port and optimizations were done by hand.
+// At some point, it may be worth adding support to fiat-crypto for
+// generating Odin output.
+
+import fiat "core:crypto/_fiat"
+import "core:math/bits"
+
+Loose_Field_Element :: distinct [3]u64
+Tight_Field_Element :: distinct [3]u64
+
+_addcarryx_u44 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0xfffffffffff)
+	x3 := fiat.u1((x1 >> 44))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u44 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 44))
+	x3 := (u64(x1) & 0xfffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+_addcarryx_u43 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((u64(arg1) + arg2) + arg3)
+	x2 := (x1 & 0x7ffffffffff)
+	x3 := fiat.u1((x1 >> 43))
+	out1 = x2
+	out2 = x3
+	return
+}
+
+_subborrowx_u43 :: #force_inline proc "contextless" (arg1: fiat.u1, arg2, arg3: u64) -> (out1: u64, out2: fiat.u1) {
+	x1 := ((i64(arg2) - i64(arg1)) - i64(arg3))
+	x2 := fiat.i1((x1 >> 43))
+	x3 := (u64(x1) & 0x7ffffffffff)
+	out1 = x3
+	out2 = (0x0 - fiat.u1(x2))
+	return
+}
+
+fe_carry_mul :: proc (out1: ^Tight_Field_Element, arg1, arg2: ^Loose_Field_Element) {
+	x2, x1 := bits.mul_u64(arg1[2], (arg2[2] * 0x5))
+	x4, x3 := bits.mul_u64(arg1[2], (arg2[1] * 0xa))
+	x6, x5 := bits.mul_u64(arg1[1], (arg2[2] * 0xa))
+	x8, x7 := bits.mul_u64(arg1[2], arg2[0])
+	x10, x9 := bits.mul_u64(arg1[1], (arg2[1] * 0x2))
+	x12, x11 := bits.mul_u64(arg1[1], arg2[0])
+	x14, x13 := bits.mul_u64(arg1[0], arg2[2])
+	x16, x15 := bits.mul_u64(arg1[0], arg2[1])
+	x18, x17 := bits.mul_u64(arg1[0], arg2[0])
+	x19, x20 := bits.add_u64(x5, x3, u64(0x0))
+	x21, _ := bits.add_u64(x6, x4, u64(fiat.u1(x20)))
+	x23, x24 := bits.add_u64(x17, x19, u64(0x0))
+	x25, _ := bits.add_u64(x18, x21, u64(fiat.u1(x24)))
+	x27 := ((x23 >> 44) | ((x25 << 20) & 0xffffffffffffffff))
+	x28 := (x23 & 0xfffffffffff)
+	x29, x30 := bits.add_u64(x9, x7, u64(0x0))
+	x31, _ := bits.add_u64(x10, x8, u64(fiat.u1(x30)))
+	x33, x34 := bits.add_u64(x13, x29, u64(0x0))
+	x35, _ := bits.add_u64(x14, x31, u64(fiat.u1(x34)))
+	x37, x38 := bits.add_u64(x11, x1, u64(0x0))
+	x39, _ := bits.add_u64(x12, x2, u64(fiat.u1(x38)))
+	x41, x42 := bits.add_u64(x15, x37, u64(0x0))
+	x43, _ := bits.add_u64(x16, x39, u64(fiat.u1(x42)))
+	x45, x46 := bits.add_u64(x27, x41, u64(0x0))
+	x47 := (u64(fiat.u1(x46)) + x43)
+	x48 := ((x45 >> 43) | ((x47 << 21) & 0xffffffffffffffff))
+	x49 := (x45 & 0x7ffffffffff)
+	x50, x51 := bits.add_u64(x48, x33, u64(0x0))
+	x52 := (u64(fiat.u1(x51)) + x35)
+	x53 := ((x50 >> 43) | ((x52 << 21) & 0xffffffffffffffff))
+	x54 := (x50 & 0x7ffffffffff)
+	x55 := (x53 * 0x5)
+	x56 := (x28 + x55)
+	x57 := (x56 >> 44)
+	x58 := (x56 & 0xfffffffffff)
+	x59 := (x57 + x49)
+	x60 := fiat.u1((x59 >> 43))
+	x61 := (x59 & 0x7ffffffffff)
+	x62 := (u64(x60) + x54)
+	out1[0] = x58
+	out1[1] = x61
+	out1[2] = x62
+}
+
+fe_carry_square :: proc (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := (arg1[2] * 0x5)
+	x2 := (x1 * 0x2)
+	x3 := (arg1[2] * 0x2)
+	x4 := (arg1[1] * 0x2)
+	x6, x5 := bits.mul_u64(arg1[2], x1)
+	x8, x7 := bits.mul_u64(arg1[1], (x2 * 0x2))
+	x10, x9 := bits.mul_u64(arg1[1], (arg1[1] * 0x2))
+	x12, x11 := bits.mul_u64(arg1[0], x3)
+	x14, x13 := bits.mul_u64(arg1[0], x4)
+	x16, x15 := bits.mul_u64(arg1[0], arg1[0])
+	x17, x18 := bits.add_u64(x15, x7, u64(0x0))
+	x19, _ := bits.add_u64(x16, x8, u64(fiat.u1(x18)))
+	x21 := ((x17 >> 44) | ((x19 << 20) & 0xffffffffffffffff))
+	x22 := (x17 & 0xfffffffffff)
+	x23, x24 := bits.add_u64(x11, x9, u64(0x0))
+	x25, _ := bits.add_u64(x12, x10, u64(fiat.u1(x24)))
+	x27, x28 := bits.add_u64(x13, x5, u64(0x0))
+	x29, _ := bits.add_u64(x14, x6, u64(fiat.u1(x28)))
+	x31, x32 := bits.add_u64(x21, x27, u64(0x0))
+	x33 := (u64(fiat.u1(x32)) + x29)
+	x34 := ((x31 >> 43) | ((x33 << 21) & 0xffffffffffffffff))
+	x35 := (x31 & 0x7ffffffffff)
+	x36, x37 := bits.add_u64(x34, x23, u64(0x0))
+	x38 := (u64(fiat.u1(x37)) + x25)
+	x39 := ((x36 >> 43) | ((x38 << 21) & 0xffffffffffffffff))
+	x40 := (x36 & 0x7ffffffffff)
+	x41 := (x39 * 0x5)
+	x42 := (x22 + x41)
+	x43 := (x42 >> 44)
+	x44 := (x42 & 0xfffffffffff)
+	x45 := (x43 + x35)
+	x46 := fiat.u1((x45 >> 43))
+	x47 := (x45 & 0x7ffffffffff)
+	x48 := (u64(x46) + x40)
+	out1[0] = x44
+	out1[1] = x47
+	out1[2] = x48
+}
+
+fe_carry :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^Loose_Field_Element) {
+	x1 := arg1[0]
+	x2 := ((x1 >> 44) + arg1[1])
+	x3 := ((x2 >> 43) + arg1[2])
+	x4 := ((x1 & 0xfffffffffff) + ((x3 >> 43) * 0x5))
+	x5 := (u64(fiat.u1((x4 >> 44))) + (x2 & 0x7ffffffffff))
+	x6 := (x4 & 0xfffffffffff)
+	x7 := (x5 & 0x7ffffffffff)
+	x8 := (u64(fiat.u1((x5 >> 43))) + (x3 & 0x7ffffffffff))
+	out1[0] = x6
+	out1[1] = x7
+	out1[2] = x8
+}
+
+fe_add :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := (arg1[0] + arg2[0])
+	x2 := (arg1[1] + arg2[1])
+	x3 := (arg1[2] + arg2[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_sub :: proc "contextless" (out1: ^Loose_Field_Element, arg1, arg2: ^Tight_Field_Element) {
+	x1 := ((0x1ffffffffff6 + arg1[0]) - arg2[0])
+	x2 := ((0xffffffffffe + arg1[1]) - arg2[1])
+	x3 := ((0xffffffffffe + arg1[2]) - arg2[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_opp :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := (0x1ffffffffff6 - arg1[0])
+	x2 := (0xffffffffffe - arg1[1])
+	x3 := (0xffffffffffe - arg1[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+@(optimization_mode="none")
+fe_cond_assign :: #force_no_inline proc "contextless" (out1, arg1: ^Tight_Field_Element, arg2: bool) {
+	x1 := fiat.cmovznz_u64(fiat.u1(arg2), out1[0], arg1[0])
+	x2 := fiat.cmovznz_u64(fiat.u1(arg2), out1[1], arg1[1])
+	x3 := fiat.cmovznz_u64(fiat.u1(arg2), out1[2], arg1[2])
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+fe_to_bytes :: proc "contextless" (out1: ^[32]byte, arg1: ^Tight_Field_Element) {
+	x1, x2 := _subborrowx_u44(0x0, arg1[0], 0xffffffffffb)
+	x3, x4 := _subborrowx_u43(x2, arg1[1], 0x7ffffffffff)
+	x5, x6 := _subborrowx_u43(x4, arg1[2], 0x7ffffffffff)
+	x7 := fiat.cmovznz_u64(x6, u64(0x0), 0xffffffffffffffff)
+	x8, x9 := _addcarryx_u44(0x0, x1, (x7 & 0xffffffffffb))
+	x10, x11 := _addcarryx_u43(x9, x3, (x7 & 0x7ffffffffff))
+	x12, _ := _addcarryx_u43(x11, x5, (x7 & 0x7ffffffffff))
+	x14 := (x12 << 7)
+	x15 := (x10 << 4)
+	x16 := (u8(x8) & 0xff)
+	x17 := (x8 >> 8)
+	x18 := (u8(x17) & 0xff)
+	x19 := (x17 >> 8)
+	x20 := (u8(x19) & 0xff)
+	x21 := (x19 >> 8)
+	x22 := (u8(x21) & 0xff)
+	x23 := (x21 >> 8)
+	x24 := (u8(x23) & 0xff)
+	x25 := u8((x23 >> 8))
+	x26 := (x15 + u64(x25))
+	x27 := (u8(x26) & 0xff)
+	x28 := (x26 >> 8)
+	x29 := (u8(x28) & 0xff)
+	x30 := (x28 >> 8)
+	x31 := (u8(x30) & 0xff)
+	x32 := (x30 >> 8)
+	x33 := (u8(x32) & 0xff)
+	x34 := (x32 >> 8)
+	x35 := (u8(x34) & 0xff)
+	x36 := u8((x34 >> 8))
+	x37 := (x14 + u64(x36))
+	x38 := (u8(x37) & 0xff)
+	x39 := (x37 >> 8)
+	x40 := (u8(x39) & 0xff)
+	x41 := (x39 >> 8)
+	x42 := (u8(x41) & 0xff)
+	x43 := (x41 >> 8)
+	x44 := (u8(x43) & 0xff)
+	x45 := (x43 >> 8)
+	x46 := (u8(x45) & 0xff)
+	x47 := (x45 >> 8)
+	x48 := (u8(x47) & 0xff)
+	x49 := u8((x47 >> 8))
+	out1[0] = x16
+	out1[1] = x18
+	out1[2] = x20
+	out1[3] = x22
+	out1[4] = x24
+	out1[5] = x27
+	out1[6] = x29
+	out1[7] = x31
+	out1[8] = x33
+	out1[9] = x35
+	out1[10] = x38
+	out1[11] = x40
+	out1[12] = x42
+	out1[13] = x44
+	out1[14] = x46
+	out1[15] = x48
+	out1[16] = x49
+}
+
+_fe_from_bytes :: proc "contextless" (out1: ^Tight_Field_Element, arg1: ^[32]byte) {
+	x1 := (u64(arg1[16]) << 41)
+	x2 := (u64(arg1[15]) << 33)
+	x3 := (u64(arg1[14]) << 25)
+	x4 := (u64(arg1[13]) << 17)
+	x5 := (u64(arg1[12]) << 9)
+	x6 := (u64(arg1[11]) * u64(0x2))
+	x7 := (u64(arg1[10]) << 36)
+	x8 := (u64(arg1[9]) << 28)
+	x9 := (u64(arg1[8]) << 20)
+	x10 := (u64(arg1[7]) << 12)
+	x11 := (u64(arg1[6]) << 4)
+	x12 := (u64(arg1[5]) << 40)
+	x13 := (u64(arg1[4]) << 32)
+	x14 := (u64(arg1[3]) << 24)
+	x15 := (u64(arg1[2]) << 16)
+	x16 := (u64(arg1[1]) << 8)
+	x17 := arg1[0]
+	x18 := (x16 + u64(x17))
+	x19 := (x15 + x18)
+	x20 := (x14 + x19)
+	x21 := (x13 + x20)
+	x22 := (x12 + x21)
+	x23 := (x22 & 0xfffffffffff)
+	x24 := u8((x22 >> 44))
+	x25 := (x11 + u64(x24))
+	x26 := (x10 + x25)
+	x27 := (x9 + x26)
+	x28 := (x8 + x27)
+	x29 := (x7 + x28)
+	x30 := (x29 & 0x7ffffffffff)
+	x31 := fiat.u1((x29 >> 43))
+	x32 := (x6 + u64(x31))
+	x33 := (x5 + x32)
+	x34 := (x4 + x33)
+	x35 := (x3 + x34)
+	x36 := (x2 + x35)
+	x37 := (x1 + x36)
+	out1[0] = x23
+	out1[1] = x30
+	out1[2] = x37
+}
+
+fe_relax :: proc "contextless" (out1: ^Loose_Field_Element, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+// The following routines were added by hand, and do not come from fiat-crypto.
+
+fe_zero :: proc "contextless" (out1: ^Tight_Field_Element) {
+	out1[0] = 0
+	out1[1] = 0
+	out1[2] = 0
+}
+
+fe_set :: #force_inline proc "contextless" (out1, arg1: ^Tight_Field_Element) {
+	x1 := arg1[0]
+	x2 := arg1[1]
+	x3 := arg1[2]
+	out1[0] = x1
+	out1[1] = x2
+	out1[2] = x3
+}
+
+@(optimization_mode="none")
+fe_cond_swap :: #force_no_inline proc "contextless" (out1, out2: ^Tight_Field_Element, arg1: bool) {
+	mask := -u64(arg1)
+	x := (out1[0] ~ out2[0]) & mask
+	x1, y1 := out1[0] ~ x, out2[0] ~ x
+	x = (out1[1] ~ out2[1]) & mask
+	x2, y2 := out1[1] ~ x, out2[1] ~ x
+	x = (out1[2] ~ out2[2]) & mask
+	x3, y3 := out1[2] ~ x, out2[2] ~ x
+	out1[0], out2[0] = x1, y1
+	out1[1], out2[1] = x2, y2
+	out1[2], out2[2] = x3, y3
+}
@@ -0,0 +1,169 @@
+package _sha3
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Keccak hashing algorithm, standardized as SHA3 in <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf>
+    To use the original Keccak padding, set the is_keccak bool to true, otherwise it will use SHA3 padding.
+*/
+
+import "../util"
+
+ROUNDS :: 24
+
+Sha3_Context :: struct {
+    st: struct #raw_union {
+        b: [200]u8,
+        q: [25]u64,
+    },
+    pt:        int,
+    rsiz:      int,
+    mdlen:     int,
+    is_keccak: bool,
+}
+
+keccakf :: proc "contextless" (st: ^[25]u64) {
+    keccakf_rndc := [?]u64 {
+        0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
+        0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
+        0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
+        0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
+        0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
+        0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
+        0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
+        0x8000000000008080, 0x0000000080000001, 0x8000000080008008,
+    }
+
+    keccakf_rotc := [?]i32 {
+        1,  3,  6,  10, 15, 21, 28, 36, 45, 55, 2,  14,
+        27, 41, 56, 8,  25, 43, 62, 18, 39, 61, 20, 44,
+    }
+
+    keccakf_piln := [?]i32 {
+        10, 7,  11, 17, 18, 3, 5,  16, 8,  21, 24, 4,
+        15, 23, 19, 13, 12, 2, 20, 14, 22, 9,  6,  1,
+    }
+
+    i, j, r: i32 = ---, ---, ---
+    t: u64       = ---
+    bc: [5]u64   = ---
+
+    when ODIN_ENDIAN != .Little {
+        v: uintptr = ---
+        for i = 0; i < 25; i += 1 {
+            v := uintptr(&st[i])
+            st[i] = u64((^u8)(v + 0)^ << 0)  | u64((^u8)(v + 1)^ << 8)  |
+                    u64((^u8)(v + 2)^ << 16) | u64((^u8)(v + 3)^ << 24) |
+                    u64((^u8)(v + 4)^ << 32) | u64((^u8)(v + 5)^ << 40) |
+                    u64((^u8)(v + 6)^ << 48) | u64((^u8)(v + 7)^ << 56)
+        }
+    }
+
+    for r = 0; r < ROUNDS; r += 1 {
+        // theta
+        for i = 0; i < 5; i += 1 {
+            bc[i] = st[i] ~ st[i + 5] ~ st[i + 10] ~ st[i + 15] ~ st[i + 20]
+        }
+
+        for i = 0; i < 5; i += 1 {
+            t = bc[(i + 4) % 5] ~ util.ROTL64(bc[(i + 1) % 5], 1)
+            for j = 0; j < 25; j += 5 {
+                st[j + i] ~= t
+            }
+        }
+
+        // rho pi
+        t = st[1]
+        for i = 0; i < 24; i += 1 {
+            j = keccakf_piln[i]
+            bc[0] = st[j]
+            st[j] = util.ROTL64(t, u64(keccakf_rotc[i]))
+            t = bc[0]
+        }
+
+        // chi
+        for j = 0; j < 25; j += 5 {
+            for i = 0; i < 5; i += 1 {
+                bc[i] = st[j + i]
+            }
+            for i = 0; i < 5; i += 1 {
+                st[j + i] ~= ~bc[(i + 1) % 5] & bc[(i + 2) % 5]
+            }
+        }
+
+        st[0] ~= keccakf_rndc[r]
+    }
+
+    when ODIN_ENDIAN != .Little {
+        for i = 0; i < 25; i += 1 {
+            v = uintptr(&st[i])
+            t = st[i]
+            (^u8)(v + 0)^ = (t >> 0)  & 0xff
+            (^u8)(v + 1)^ = (t >> 8)  & 0xff
+            (^u8)(v + 2)^ = (t >> 16) & 0xff
+            (^u8)(v + 3)^ = (t >> 24) & 0xff
+            (^u8)(v + 4)^ = (t >> 32) & 0xff
+            (^u8)(v + 5)^ = (t >> 40) & 0xff
+            (^u8)(v + 6)^ = (t >> 48) & 0xff
+            (^u8)(v + 7)^ = (t >> 56) & 0xff
+        }
+    }
+}
+
+init :: proc "contextless" (c: ^Sha3_Context) {
+    for i := 0; i < 25; i += 1 {
+        c.st.q[i] = 0
+    }
+    c.rsiz = 200 - 2 * c.mdlen
+}
+
+update :: proc "contextless" (c: ^Sha3_Context, data: []byte) {
+    j := c.pt
+    for i := 0; i < len(data); i += 1 {
+        c.st.b[j] ~= data[i]
+        j += 1
+        if j >= c.rsiz {
+            keccakf(&c.st.q)
+            j = 0
+        }
+    }
+    c.pt = j
+}
+
+final :: proc "contextless" (c: ^Sha3_Context, hash: []byte) {
+    if c.is_keccak {
+        c.st.b[c.pt] ~= 0x01
+    } else {
+        c.st.b[c.pt] ~= 0x06
+    }
+    
+    c.st.b[c.rsiz - 1] ~= 0x80
+    keccakf(&c.st.q)
+    for i := 0; i < c.mdlen; i += 1 {
+        hash[i] = c.st.b[i]
+    }
+}
+
+shake_xof :: proc "contextless" (c: ^Sha3_Context) {
+    c.st.b[c.pt]       ~= 0x1F
+    c.st.b[c.rsiz - 1] ~= 0x80
+    keccakf(&c.st.q)
+    c.pt = 0
+}
+
+shake_out :: proc "contextless" (c: ^Sha3_Context, hash: []byte) {
+    j := c.pt
+    for i := 0; i < len(hash); i += 1 {
+        if j >= c.rsiz {
+            keccakf(&c.st.q)
+            j = 0
+        }
+        hash[i] = c.st.b[j]
+        j += 1
+    }
+    c.pt = j
+}
@@ -0,0 +1,410 @@
+package _tiger
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the Tiger hashing algorithm, as defined in <https://www.cs.technion.ac.il/~biham/Reports/Tiger/>
+*/
+
+import "../util"
+
+T1 := [?]u64 {
+	0x02aab17cf7e90c5e, 0xac424b03e243a8ec, 0x72cd5be30dd5fcd3, 0x6d019b93f6f97f3a,
+	0xcd9978ffd21f9193, 0x7573a1c9708029e2, 0xb164326b922a83c3, 0x46883eee04915870,
+	0xeaace3057103ece6, 0xc54169b808a3535c, 0x4ce754918ddec47c, 0x0aa2f4dfdc0df40c,
+	0x10b76f18a74dbefa, 0xc6ccb6235ad1ab6a, 0x13726121572fe2ff, 0x1a488c6f199d921e,
+	0x4bc9f9f4da0007ca, 0x26f5e6f6e85241c7, 0x859079dbea5947b6, 0x4f1885c5c99e8c92,
+	0xd78e761ea96f864b, 0x8e36428c52b5c17d, 0x69cf6827373063c1, 0xb607c93d9bb4c56e,
+	0x7d820e760e76b5ea, 0x645c9cc6f07fdc42, 0xbf38a078243342e0, 0x5f6b343c9d2e7d04,
+	0xf2c28aeb600b0ec6, 0x6c0ed85f7254bcac, 0x71592281a4db4fe5, 0x1967fa69ce0fed9f,
+	0xfd5293f8b96545db, 0xc879e9d7f2a7600b, 0x860248920193194e, 0xa4f9533b2d9cc0b3,
+	0x9053836c15957613, 0xdb6dcf8afc357bf1, 0x18beea7a7a370f57, 0x037117ca50b99066,
+	0x6ab30a9774424a35, 0xf4e92f02e325249b, 0x7739db07061ccae1, 0xd8f3b49ceca42a05,
+	0xbd56be3f51382f73, 0x45faed5843b0bb28, 0x1c813d5c11bf1f83, 0x8af0e4b6d75fa169,
+	0x33ee18a487ad9999, 0x3c26e8eab1c94410, 0xb510102bc0a822f9, 0x141eef310ce6123b,
+	0xfc65b90059ddb154, 0xe0158640c5e0e607, 0x884e079826c3a3cf, 0x930d0d9523c535fd,
+	0x35638d754e9a2b00, 0x4085fccf40469dd5, 0xc4b17ad28be23a4c, 0xcab2f0fc6a3e6a2e,
+	0x2860971a6b943fcd, 0x3dde6ee212e30446, 0x6222f32ae01765ae, 0x5d550bb5478308fe,
+	0xa9efa98da0eda22a, 0xc351a71686c40da7, 0x1105586d9c867c84, 0xdcffee85fda22853,
+	0xccfbd0262c5eef76, 0xbaf294cb8990d201, 0xe69464f52afad975, 0x94b013afdf133e14,
+	0x06a7d1a32823c958, 0x6f95fe5130f61119, 0xd92ab34e462c06c0, 0xed7bde33887c71d2,
+	0x79746d6e6518393e, 0x5ba419385d713329, 0x7c1ba6b948a97564, 0x31987c197bfdac67,
+	0xde6c23c44b053d02, 0x581c49fed002d64d, 0xdd474d6338261571, 0xaa4546c3e473d062,
+	0x928fce349455f860, 0x48161bbacaab94d9, 0x63912430770e6f68, 0x6ec8a5e602c6641c,
+	0x87282515337ddd2b, 0x2cda6b42034b701b, 0xb03d37c181cb096d, 0xe108438266c71c6f,
+	0x2b3180c7eb51b255, 0xdf92b82f96c08bbc, 0x5c68c8c0a632f3ba, 0x5504cc861c3d0556,
+	0xabbfa4e55fb26b8f, 0x41848b0ab3baceb4, 0xb334a273aa445d32, 0xbca696f0a85ad881,
+	0x24f6ec65b528d56c, 0x0ce1512e90f4524a, 0x4e9dd79d5506d35a, 0x258905fac6ce9779,
+	0x2019295b3e109b33, 0xf8a9478b73a054cc, 0x2924f2f934417eb0, 0x3993357d536d1bc4,
+	0x38a81ac21db6ff8b, 0x47c4fbf17d6016bf, 0x1e0faadd7667e3f5, 0x7abcff62938beb96,
+	0xa78dad948fc179c9, 0x8f1f98b72911e50d, 0x61e48eae27121a91, 0x4d62f7ad31859808,
+	0xeceba345ef5ceaeb, 0xf5ceb25ebc9684ce, 0xf633e20cb7f76221, 0xa32cdf06ab8293e4,
+	0x985a202ca5ee2ca4, 0xcf0b8447cc8a8fb1, 0x9f765244979859a3, 0xa8d516b1a1240017,
+	0x0bd7ba3ebb5dc726, 0xe54bca55b86adb39, 0x1d7a3afd6c478063, 0x519ec608e7669edd,
+	0x0e5715a2d149aa23, 0x177d4571848ff194, 0xeeb55f3241014c22, 0x0f5e5ca13a6e2ec2,
+	0x8029927b75f5c361, 0xad139fabc3d6e436, 0x0d5df1a94ccf402f, 0x3e8bd948bea5dfc8,
+	0xa5a0d357bd3ff77e, 0xa2d12e251f74f645, 0x66fd9e525e81a082, 0x2e0c90ce7f687a49,
+	0xc2e8bcbeba973bc5, 0x000001bce509745f, 0x423777bbe6dab3d6, 0xd1661c7eaef06eb5,
+	0xa1781f354daacfd8, 0x2d11284a2b16affc, 0xf1fc4f67fa891d1f, 0x73ecc25dcb920ada,
+	0xae610c22c2a12651, 0x96e0a810d356b78a, 0x5a9a381f2fe7870f, 0xd5ad62ede94e5530,
+	0xd225e5e8368d1427, 0x65977b70c7af4631, 0x99f889b2de39d74f, 0x233f30bf54e1d143,
+	0x9a9675d3d9a63c97, 0x5470554ff334f9a8, 0x166acb744a4f5688, 0x70c74caab2e4aead,
+	0xf0d091646f294d12, 0x57b82a89684031d1, 0xefd95a5a61be0b6b, 0x2fbd12e969f2f29a,
+	0x9bd37013feff9fe8, 0x3f9b0404d6085a06, 0x4940c1f3166cfe15, 0x09542c4dcdf3defb,
+	0xb4c5218385cd5ce3, 0xc935b7dc4462a641, 0x3417f8a68ed3b63f, 0xb80959295b215b40,
+	0xf99cdaef3b8c8572, 0x018c0614f8fcb95d, 0x1b14accd1a3acdf3, 0x84d471f200bb732d,
+	0xc1a3110e95e8da16, 0x430a7220bf1a82b8, 0xb77e090d39df210e, 0x5ef4bd9f3cd05e9d,
+	0x9d4ff6da7e57a444, 0xda1d60e183d4a5f8, 0xb287c38417998e47, 0xfe3edc121bb31886,
+	0xc7fe3ccc980ccbef, 0xe46fb590189bfd03, 0x3732fd469a4c57dc, 0x7ef700a07cf1ad65,
+	0x59c64468a31d8859, 0x762fb0b4d45b61f6, 0x155baed099047718, 0x68755e4c3d50baa6,
+	0xe9214e7f22d8b4df, 0x2addbf532eac95f4, 0x32ae3909b4bd0109, 0x834df537b08e3450,
+	0xfa209da84220728d, 0x9e691d9b9efe23f7, 0x0446d288c4ae8d7f, 0x7b4cc524e169785b,
+	0x21d87f0135ca1385, 0xcebb400f137b8aa5, 0x272e2b66580796be, 0x3612264125c2b0de,
+	0x057702bdad1efbb2, 0xd4babb8eacf84be9, 0x91583139641bc67b, 0x8bdc2de08036e024,
+	0x603c8156f49f68ed, 0xf7d236f7dbef5111, 0x9727c4598ad21e80, 0xa08a0896670a5fd7,
+	0xcb4a8f4309eba9cb, 0x81af564b0f7036a1, 0xc0b99aa778199abd, 0x959f1ec83fc8e952,
+	0x8c505077794a81b9, 0x3acaaf8f056338f0, 0x07b43f50627a6778, 0x4a44ab49f5eccc77,
+	0x3bc3d6e4b679ee98, 0x9cc0d4d1cf14108c, 0x4406c00b206bc8a0, 0x82a18854c8d72d89,
+	0x67e366b35c3c432c, 0xb923dd61102b37f2, 0x56ab2779d884271d, 0xbe83e1b0ff1525af,
+	0xfb7c65d4217e49a9, 0x6bdbe0e76d48e7d4, 0x08df828745d9179e, 0x22ea6a9add53bd34,
+	0xe36e141c5622200a, 0x7f805d1b8cb750ee, 0xafe5c7a59f58e837, 0xe27f996a4fb1c23c,
+	0xd3867dfb0775f0d0, 0xd0e673de6e88891a, 0x123aeb9eafb86c25, 0x30f1d5d5c145b895,
+	0xbb434a2dee7269e7, 0x78cb67ecf931fa38, 0xf33b0372323bbf9c, 0x52d66336fb279c74,
+	0x505f33ac0afb4eaa, 0xe8a5cd99a2cce187, 0x534974801e2d30bb, 0x8d2d5711d5876d90,
+	0x1f1a412891bc038e, 0xd6e2e71d82e56648, 0x74036c3a497732b7, 0x89b67ed96361f5ab,
+	0xffed95d8f1ea02a2, 0xe72b3bd61464d43d, 0xa6300f170bdc4820, 0xebc18760ed78a77a,
+}
+
+T2 := [?]u64 {
+	0xe6a6be5a05a12138, 0xb5a122a5b4f87c98, 0x563c6089140b6990, 0x4c46cb2e391f5dd5,
+	0xd932addbc9b79434, 0x08ea70e42015aff5, 0xd765a6673e478cf1, 0xc4fb757eab278d99,
+	0xdf11c6862d6e0692, 0xddeb84f10d7f3b16, 0x6f2ef604a665ea04, 0x4a8e0f0ff0e0dfb3,
+	0xa5edeef83dbcba51, 0xfc4f0a2a0ea4371e, 0xe83e1da85cb38429, 0xdc8ff882ba1b1ce2,
+	0xcd45505e8353e80d, 0x18d19a00d4db0717, 0x34a0cfeda5f38101, 0x0be77e518887caf2,
+	0x1e341438b3c45136, 0xe05797f49089ccf9, 0xffd23f9df2591d14, 0x543dda228595c5cd,
+	0x661f81fd99052a33, 0x8736e641db0f7b76, 0x15227725418e5307, 0xe25f7f46162eb2fa,
+	0x48a8b2126c13d9fe, 0xafdc541792e76eea, 0x03d912bfc6d1898f, 0x31b1aafa1b83f51b,
+	0xf1ac2796e42ab7d9, 0x40a3a7d7fcd2ebac, 0x1056136d0afbbcc5, 0x7889e1dd9a6d0c85,
+	0xd33525782a7974aa, 0xa7e25d09078ac09b, 0xbd4138b3eac6edd0, 0x920abfbe71eb9e70,
+	0xa2a5d0f54fc2625c, 0xc054e36b0b1290a3, 0xf6dd59ff62fe932b, 0x3537354511a8ac7d,
+	0xca845e9172fadcd4, 0x84f82b60329d20dc, 0x79c62ce1cd672f18, 0x8b09a2add124642c,
+	0xd0c1e96a19d9e726, 0x5a786a9b4ba9500c, 0x0e020336634c43f3, 0xc17b474aeb66d822,
+	0x6a731ae3ec9baac2, 0x8226667ae0840258, 0x67d4567691caeca5, 0x1d94155c4875adb5,
+	0x6d00fd985b813fdf, 0x51286efcb774cd06, 0x5e8834471fa744af, 0xf72ca0aee761ae2e,
+	0xbe40e4cdaee8e09a, 0xe9970bbb5118f665, 0x726e4beb33df1964, 0x703b000729199762,
+	0x4631d816f5ef30a7, 0xb880b5b51504a6be, 0x641793c37ed84b6c, 0x7b21ed77f6e97d96,
+	0x776306312ef96b73, 0xae528948e86ff3f4, 0x53dbd7f286a3f8f8, 0x16cadce74cfc1063,
+	0x005c19bdfa52c6dd, 0x68868f5d64d46ad3, 0x3a9d512ccf1e186a, 0x367e62c2385660ae,
+	0xe359e7ea77dcb1d7, 0x526c0773749abe6e, 0x735ae5f9d09f734b, 0x493fc7cc8a558ba8,
+	0xb0b9c1533041ab45, 0x321958ba470a59bd, 0x852db00b5f46c393, 0x91209b2bd336b0e5,
+	0x6e604f7d659ef19f, 0xb99a8ae2782ccb24, 0xccf52ab6c814c4c7, 0x4727d9afbe11727b,
+	0x7e950d0c0121b34d, 0x756f435670ad471f, 0xf5add442615a6849, 0x4e87e09980b9957a,
+	0x2acfa1df50aee355, 0xd898263afd2fd556, 0xc8f4924dd80c8fd6, 0xcf99ca3d754a173a,
+	0xfe477bacaf91bf3c, 0xed5371f6d690c12d, 0x831a5c285e687094, 0xc5d3c90a3708a0a4,
+	0x0f7f903717d06580, 0x19f9bb13b8fdf27f, 0xb1bd6f1b4d502843, 0x1c761ba38fff4012,
+	0x0d1530c4e2e21f3b, 0x8943ce69a7372c8a, 0xe5184e11feb5ce66, 0x618bdb80bd736621,
+	0x7d29bad68b574d0b, 0x81bb613e25e6fe5b, 0x071c9c10bc07913f, 0xc7beeb7909ac2d97,
+	0xc3e58d353bc5d757, 0xeb017892f38f61e8, 0xd4effb9c9b1cc21a, 0x99727d26f494f7ab,
+	0xa3e063a2956b3e03, 0x9d4a8b9a4aa09c30, 0x3f6ab7d500090fb4, 0x9cc0f2a057268ac0,
+	0x3dee9d2dedbf42d1, 0x330f49c87960a972, 0xc6b2720287421b41, 0x0ac59ec07c00369c,
+	0xef4eac49cb353425, 0xf450244eef0129d8, 0x8acc46e5caf4deb6, 0x2ffeab63989263f7,
+	0x8f7cb9fe5d7a4578, 0x5bd8f7644e634635, 0x427a7315bf2dc900, 0x17d0c4aa2125261c,
+	0x3992486c93518e50, 0xb4cbfee0a2d7d4c3, 0x7c75d6202c5ddd8d, 0xdbc295d8e35b6c61,
+	0x60b369d302032b19, 0xce42685fdce44132, 0x06f3ddb9ddf65610, 0x8ea4d21db5e148f0,
+	0x20b0fce62fcd496f, 0x2c1b912358b0ee31, 0xb28317b818f5a308, 0xa89c1e189ca6d2cf,
+	0x0c6b18576aaadbc8, 0xb65deaa91299fae3, 0xfb2b794b7f1027e7, 0x04e4317f443b5beb,
+	0x4b852d325939d0a6, 0xd5ae6beefb207ffc, 0x309682b281c7d374, 0xbae309a194c3b475,
+	0x8cc3f97b13b49f05, 0x98a9422ff8293967, 0x244b16b01076ff7c, 0xf8bf571c663d67ee,
+	0x1f0d6758eee30da1, 0xc9b611d97adeb9b7, 0xb7afd5887b6c57a2, 0x6290ae846b984fe1,
+	0x94df4cdeacc1a5fd, 0x058a5bd1c5483aff, 0x63166cc142ba3c37, 0x8db8526eb2f76f40,
+	0xe10880036f0d6d4e, 0x9e0523c9971d311d, 0x45ec2824cc7cd691, 0x575b8359e62382c9,
+	0xfa9e400dc4889995, 0xd1823ecb45721568, 0xdafd983b8206082f, 0xaa7d29082386a8cb,
+	0x269fcd4403b87588, 0x1b91f5f728bdd1e0, 0xe4669f39040201f6, 0x7a1d7c218cf04ade,
+	0x65623c29d79ce5ce, 0x2368449096c00bb1, 0xab9bf1879da503ba, 0xbc23ecb1a458058e,
+	0x9a58df01bb401ecc, 0xa070e868a85f143d, 0x4ff188307df2239e, 0x14d565b41a641183,
+	0xee13337452701602, 0x950e3dcf3f285e09, 0x59930254b9c80953, 0x3bf299408930da6d,
+	0xa955943f53691387, 0xa15edecaa9cb8784, 0x29142127352be9a0, 0x76f0371fff4e7afb,
+	0x0239f450274f2228, 0xbb073af01d5e868b, 0xbfc80571c10e96c1, 0xd267088568222e23,
+	0x9671a3d48e80b5b0, 0x55b5d38ae193bb81, 0x693ae2d0a18b04b8, 0x5c48b4ecadd5335f,
+	0xfd743b194916a1ca, 0x2577018134be98c4, 0xe77987e83c54a4ad, 0x28e11014da33e1b9,
+	0x270cc59e226aa213, 0x71495f756d1a5f60, 0x9be853fb60afef77, 0xadc786a7f7443dbf,
+	0x0904456173b29a82, 0x58bc7a66c232bd5e, 0xf306558c673ac8b2, 0x41f639c6b6c9772a,
+	0x216defe99fda35da, 0x11640cc71c7be615, 0x93c43694565c5527, 0xea038e6246777839,
+	0xf9abf3ce5a3e2469, 0x741e768d0fd312d2, 0x0144b883ced652c6, 0xc20b5a5ba33f8552,
+	0x1ae69633c3435a9d, 0x97a28ca4088cfdec, 0x8824a43c1e96f420, 0x37612fa66eeea746,
+	0x6b4cb165f9cf0e5a, 0x43aa1c06a0abfb4a, 0x7f4dc26ff162796b, 0x6cbacc8e54ed9b0f,
+	0xa6b7ffefd2bb253e, 0x2e25bc95b0a29d4f, 0x86d6a58bdef1388c, 0xded74ac576b6f054,
+	0x8030bdbc2b45805d, 0x3c81af70e94d9289, 0x3eff6dda9e3100db, 0xb38dc39fdfcc8847,
+	0x123885528d17b87e, 0xf2da0ed240b1b642, 0x44cefadcd54bf9a9, 0x1312200e433c7ee6,
+	0x9ffcc84f3a78c748, 0xf0cd1f72248576bb, 0xec6974053638cfe4, 0x2ba7b67c0cec4e4c,
+	0xac2f4df3e5ce32ed, 0xcb33d14326ea4c11, 0xa4e9044cc77e58bc, 0x5f513293d934fcef,
+	0x5dc9645506e55444, 0x50de418f317de40a, 0x388cb31a69dde259, 0x2db4a83455820a86,
+	0x9010a91e84711ae9, 0x4df7f0b7b1498371, 0xd62a2eabc0977179, 0x22fac097aa8d5c0e,
+}
+
+T3 := [?]u64 {
+	0xf49fcc2ff1daf39b, 0x487fd5c66ff29281, 0xe8a30667fcdca83f, 0x2c9b4be3d2fcce63,
+	0xda3ff74b93fbbbc2, 0x2fa165d2fe70ba66, 0xa103e279970e93d4, 0xbecdec77b0e45e71,
+	0xcfb41e723985e497, 0xb70aaa025ef75017, 0xd42309f03840b8e0, 0x8efc1ad035898579,
+	0x96c6920be2b2abc5, 0x66af4163375a9172, 0x2174abdcca7127fb, 0xb33ccea64a72ff41,
+	0xf04a4933083066a5, 0x8d970acdd7289af5, 0x8f96e8e031c8c25e, 0xf3fec02276875d47,
+	0xec7bf310056190dd, 0xf5adb0aebb0f1491, 0x9b50f8850fd58892, 0x4975488358b74de8,
+	0xa3354ff691531c61, 0x0702bbe481d2c6ee, 0x89fb24057deded98, 0xac3075138596e902,
+	0x1d2d3580172772ed, 0xeb738fc28e6bc30d, 0x5854ef8f63044326, 0x9e5c52325add3bbe,
+	0x90aa53cf325c4623, 0xc1d24d51349dd067, 0x2051cfeea69ea624, 0x13220f0a862e7e4f,
+	0xce39399404e04864, 0xd9c42ca47086fcb7, 0x685ad2238a03e7cc, 0x066484b2ab2ff1db,
+	0xfe9d5d70efbf79ec, 0x5b13b9dd9c481854, 0x15f0d475ed1509ad, 0x0bebcd060ec79851,
+	0xd58c6791183ab7f8, 0xd1187c5052f3eee4, 0xc95d1192e54e82ff, 0x86eea14cb9ac6ca2,
+	0x3485beb153677d5d, 0xdd191d781f8c492a, 0xf60866baa784ebf9, 0x518f643ba2d08c74,
+	0x8852e956e1087c22, 0xa768cb8dc410ae8d, 0x38047726bfec8e1a, 0xa67738b4cd3b45aa,
+	0xad16691cec0dde19, 0xc6d4319380462e07, 0xc5a5876d0ba61938, 0x16b9fa1fa58fd840,
+	0x188ab1173ca74f18, 0xabda2f98c99c021f, 0x3e0580ab134ae816, 0x5f3b05b773645abb,
+	0x2501a2be5575f2f6, 0x1b2f74004e7e8ba9, 0x1cd7580371e8d953, 0x7f6ed89562764e30,
+	0xb15926ff596f003d, 0x9f65293da8c5d6b9, 0x6ecef04dd690f84c, 0x4782275fff33af88,
+	0xe41433083f820801, 0xfd0dfe409a1af9b5, 0x4325a3342cdb396b, 0x8ae77e62b301b252,
+	0xc36f9e9f6655615a, 0x85455a2d92d32c09, 0xf2c7dea949477485, 0x63cfb4c133a39eba,
+	0x83b040cc6ebc5462, 0x3b9454c8fdb326b0, 0x56f56a9e87ffd78c, 0x2dc2940d99f42bc6,
+	0x98f7df096b096e2d, 0x19a6e01e3ad852bf, 0x42a99ccbdbd4b40b, 0xa59998af45e9c559,
+	0x366295e807d93186, 0x6b48181bfaa1f773, 0x1fec57e2157a0a1d, 0x4667446af6201ad5,
+	0xe615ebcacfb0f075, 0xb8f31f4f68290778, 0x22713ed6ce22d11e, 0x3057c1a72ec3c93b,
+	0xcb46acc37c3f1f2f, 0xdbb893fd02aaf50e, 0x331fd92e600b9fcf, 0xa498f96148ea3ad6,
+	0xa8d8426e8b6a83ea, 0xa089b274b7735cdc, 0x87f6b3731e524a11, 0x118808e5cbc96749,
+	0x9906e4c7b19bd394, 0xafed7f7e9b24a20c, 0x6509eadeeb3644a7, 0x6c1ef1d3e8ef0ede,
+	0xb9c97d43e9798fb4, 0xa2f2d784740c28a3, 0x7b8496476197566f, 0x7a5be3e6b65f069d,
+	0xf96330ed78be6f10, 0xeee60de77a076a15, 0x2b4bee4aa08b9bd0, 0x6a56a63ec7b8894e,
+	0x02121359ba34fef4, 0x4cbf99f8283703fc, 0x398071350caf30c8, 0xd0a77a89f017687a,
+	0xf1c1a9eb9e423569, 0x8c7976282dee8199, 0x5d1737a5dd1f7abd, 0x4f53433c09a9fa80,
+	0xfa8b0c53df7ca1d9, 0x3fd9dcbc886ccb77, 0xc040917ca91b4720, 0x7dd00142f9d1dcdf,
+	0x8476fc1d4f387b58, 0x23f8e7c5f3316503, 0x032a2244e7e37339, 0x5c87a5d750f5a74b,
+	0x082b4cc43698992e, 0xdf917becb858f63c, 0x3270b8fc5bf86dda, 0x10ae72bb29b5dd76,
+	0x576ac94e7700362b, 0x1ad112dac61efb8f, 0x691bc30ec5faa427, 0xff246311cc327143,
+	0x3142368e30e53206, 0x71380e31e02ca396, 0x958d5c960aad76f1, 0xf8d6f430c16da536,
+	0xc8ffd13f1be7e1d2, 0x7578ae66004ddbe1, 0x05833f01067be646, 0xbb34b5ad3bfe586d,
+	0x095f34c9a12b97f0, 0x247ab64525d60ca8, 0xdcdbc6f3017477d1, 0x4a2e14d4decad24d,
+	0xbdb5e6d9be0a1eeb, 0x2a7e70f7794301ab, 0xdef42d8a270540fd, 0x01078ec0a34c22c1,
+	0xe5de511af4c16387, 0x7ebb3a52bd9a330a, 0x77697857aa7d6435, 0x004e831603ae4c32,
+	0xe7a21020ad78e312, 0x9d41a70c6ab420f2, 0x28e06c18ea1141e6, 0xd2b28cbd984f6b28,
+	0x26b75f6c446e9d83, 0xba47568c4d418d7f, 0xd80badbfe6183d8e, 0x0e206d7f5f166044,
+	0xe258a43911cbca3e, 0x723a1746b21dc0bc, 0xc7caa854f5d7cdd3, 0x7cac32883d261d9c,
+	0x7690c26423ba942c, 0x17e55524478042b8, 0xe0be477656a2389f, 0x4d289b5e67ab2da0,
+	0x44862b9c8fbbfd31, 0xb47cc8049d141365, 0x822c1b362b91c793, 0x4eb14655fb13dfd8,
+	0x1ecbba0714e2a97b, 0x6143459d5cde5f14, 0x53a8fbf1d5f0ac89, 0x97ea04d81c5e5b00,
+	0x622181a8d4fdb3f3, 0xe9bcd341572a1208, 0x1411258643cce58a, 0x9144c5fea4c6e0a4,
+	0x0d33d06565cf620f, 0x54a48d489f219ca1, 0xc43e5eac6d63c821, 0xa9728b3a72770daf,
+	0xd7934e7b20df87ef, 0xe35503b61a3e86e5, 0xcae321fbc819d504, 0x129a50b3ac60bfa6,
+	0xcd5e68ea7e9fb6c3, 0xb01c90199483b1c7, 0x3de93cd5c295376c, 0xaed52edf2ab9ad13,
+	0x2e60f512c0a07884, 0xbc3d86a3e36210c9, 0x35269d9b163951ce, 0x0c7d6e2ad0cdb5fa,
+	0x59e86297d87f5733, 0x298ef221898db0e7, 0x55000029d1a5aa7e, 0x8bc08ae1b5061b45,
+	0xc2c31c2b6c92703a, 0x94cc596baf25ef42, 0x0a1d73db22540456, 0x04b6a0f9d9c4179a,
+	0xeffdafa2ae3d3c60, 0xf7c8075bb49496c4, 0x9cc5c7141d1cd4e3, 0x78bd1638218e5534,
+	0xb2f11568f850246a, 0xedfabcfa9502bc29, 0x796ce5f2da23051b, 0xaae128b0dc93537c,
+	0x3a493da0ee4b29ae, 0xb5df6b2c416895d7, 0xfcabbd25122d7f37, 0x70810b58105dc4b1,
+	0xe10fdd37f7882a90, 0x524dcab5518a3f5c, 0x3c9e85878451255b, 0x4029828119bd34e2,
+	0x74a05b6f5d3ceccb, 0xb610021542e13eca, 0x0ff979d12f59e2ac, 0x6037da27e4f9cc50,
+	0x5e92975a0df1847d, 0xd66de190d3e623fe, 0x5032d6b87b568048, 0x9a36b7ce8235216e,
+	0x80272a7a24f64b4a, 0x93efed8b8c6916f7, 0x37ddbff44cce1555, 0x4b95db5d4b99bd25,
+	0x92d3fda169812fc0, 0xfb1a4a9a90660bb6, 0x730c196946a4b9b2, 0x81e289aa7f49da68,
+	0x64669a0f83b1a05f, 0x27b3ff7d9644f48b, 0xcc6b615c8db675b3, 0x674f20b9bcebbe95,
+	0x6f31238275655982, 0x5ae488713e45cf05, 0xbf619f9954c21157, 0xeabac46040a8eae9,
+	0x454c6fe9f2c0c1cd, 0x419cf6496412691c, 0xd3dc3bef265b0f70, 0x6d0e60f5c3578a9e,
+}
+
+T4 := [?]u64 {
+	0x5b0e608526323c55, 0x1a46c1a9fa1b59f5, 0xa9e245a17c4c8ffa, 0x65ca5159db2955d7,
+	0x05db0a76ce35afc2, 0x81eac77ea9113d45, 0x528ef88ab6ac0a0d, 0xa09ea253597be3ff,
+	0x430ddfb3ac48cd56, 0xc4b3a67af45ce46f, 0x4ececfd8fbe2d05e, 0x3ef56f10b39935f0,
+	0x0b22d6829cd619c6, 0x17fd460a74df2069, 0x6cf8cc8e8510ed40, 0xd6c824bf3a6ecaa7,
+	0x61243d581a817049, 0x048bacb6bbc163a2, 0xd9a38ac27d44cc32, 0x7fddff5baaf410ab,
+	0xad6d495aa804824b, 0xe1a6a74f2d8c9f94, 0xd4f7851235dee8e3, 0xfd4b7f886540d893,
+	0x247c20042aa4bfda, 0x096ea1c517d1327c, 0xd56966b4361a6685, 0x277da5c31221057d,
+	0x94d59893a43acff7, 0x64f0c51ccdc02281, 0x3d33bcc4ff6189db, 0xe005cb184ce66af1,
+	0xff5ccd1d1db99bea, 0xb0b854a7fe42980f, 0x7bd46a6a718d4b9f, 0xd10fa8cc22a5fd8c,
+	0xd31484952be4bd31, 0xc7fa975fcb243847, 0x4886ed1e5846c407, 0x28cddb791eb70b04,
+	0xc2b00be2f573417f, 0x5c9590452180f877, 0x7a6bddfff370eb00, 0xce509e38d6d9d6a4,
+	0xebeb0f00647fa702, 0x1dcc06cf76606f06, 0xe4d9f28ba286ff0a, 0xd85a305dc918c262,
+	0x475b1d8732225f54, 0x2d4fb51668ccb5fe, 0xa679b9d9d72bba20, 0x53841c0d912d43a5,
+	0x3b7eaa48bf12a4e8, 0x781e0e47f22f1ddf, 0xeff20ce60ab50973, 0x20d261d19dffb742,
+	0x16a12b03062a2e39, 0x1960eb2239650495, 0x251c16fed50eb8b8, 0x9ac0c330f826016e,
+	0xed152665953e7671, 0x02d63194a6369570, 0x5074f08394b1c987, 0x70ba598c90b25ce1,
+	0x794a15810b9742f6, 0x0d5925e9fcaf8c6c, 0x3067716cd868744e, 0x910ab077e8d7731b,
+	0x6a61bbdb5ac42f61, 0x93513efbf0851567, 0xf494724b9e83e9d5, 0xe887e1985c09648d,
+	0x34b1d3c675370cfd, 0xdc35e433bc0d255d, 0xd0aab84234131be0, 0x08042a50b48b7eaf,
+	0x9997c4ee44a3ab35, 0x829a7b49201799d0, 0x263b8307b7c54441, 0x752f95f4fd6a6ca6,
+	0x927217402c08c6e5, 0x2a8ab754a795d9ee, 0xa442f7552f72943d, 0x2c31334e19781208,
+	0x4fa98d7ceaee6291, 0x55c3862f665db309, 0xbd0610175d53b1f3, 0x46fe6cb840413f27,
+	0x3fe03792df0cfa59, 0xcfe700372eb85e8f, 0xa7be29e7adbce118, 0xe544ee5cde8431dd,
+	0x8a781b1b41f1873e, 0xa5c94c78a0d2f0e7, 0x39412e2877b60728, 0xa1265ef3afc9a62c,
+	0xbcc2770c6a2506c5, 0x3ab66dd5dce1ce12, 0xe65499d04a675b37, 0x7d8f523481bfd216,
+	0x0f6f64fcec15f389, 0x74efbe618b5b13c8, 0xacdc82b714273e1d, 0xdd40bfe003199d17,
+	0x37e99257e7e061f8, 0xfa52626904775aaa, 0x8bbbf63a463d56f9, 0xf0013f1543a26e64,
+	0xa8307e9f879ec898, 0xcc4c27a4150177cc, 0x1b432f2cca1d3348, 0xde1d1f8f9f6fa013,
+	0x606602a047a7ddd6, 0xd237ab64cc1cb2c7, 0x9b938e7225fcd1d3, 0xec4e03708e0ff476,
+	0xfeb2fbda3d03c12d, 0xae0bced2ee43889a, 0x22cb8923ebfb4f43, 0x69360d013cf7396d,
+	0x855e3602d2d4e022, 0x073805bad01f784c, 0x33e17a133852f546, 0xdf4874058ac7b638,
+	0xba92b29c678aa14a, 0x0ce89fc76cfaadcd, 0x5f9d4e0908339e34, 0xf1afe9291f5923b9,
+	0x6e3480f60f4a265f, 0xeebf3a2ab29b841c, 0xe21938a88f91b4ad, 0x57dfeff845c6d3c3,
+	0x2f006b0bf62caaf2, 0x62f479ef6f75ee78, 0x11a55ad41c8916a9, 0xf229d29084fed453,
+	0x42f1c27b16b000e6, 0x2b1f76749823c074, 0x4b76eca3c2745360, 0x8c98f463b91691bd,
+	0x14bcc93cf1ade66a, 0x8885213e6d458397, 0x8e177df0274d4711, 0xb49b73b5503f2951,
+	0x10168168c3f96b6b, 0x0e3d963b63cab0ae, 0x8dfc4b5655a1db14, 0xf789f1356e14de5c,
+	0x683e68af4e51dac1, 0xc9a84f9d8d4b0fd9, 0x3691e03f52a0f9d1, 0x5ed86e46e1878e80,
+	0x3c711a0e99d07150, 0x5a0865b20c4e9310, 0x56fbfc1fe4f0682e, 0xea8d5de3105edf9b,
+	0x71abfdb12379187a, 0x2eb99de1bee77b9c, 0x21ecc0ea33cf4523, 0x59a4d7521805c7a1,
+	0x3896f5eb56ae7c72, 0xaa638f3db18f75dc, 0x9f39358dabe9808e, 0xb7defa91c00b72ac,
+	0x6b5541fd62492d92, 0x6dc6dee8f92e4d5b, 0x353f57abc4beea7e, 0x735769d6da5690ce,
+	0x0a234aa642391484, 0xf6f9508028f80d9d, 0xb8e319a27ab3f215, 0x31ad9c1151341a4d,
+	0x773c22a57bef5805, 0x45c7561a07968633, 0xf913da9e249dbe36, 0xda652d9b78a64c68,
+	0x4c27a97f3bc334ef, 0x76621220e66b17f4, 0x967743899acd7d0b, 0xf3ee5bcae0ed6782,
+	0x409f753600c879fc, 0x06d09a39b5926db6, 0x6f83aeb0317ac588, 0x01e6ca4a86381f21,
+	0x66ff3462d19f3025, 0x72207c24ddfd3bfb, 0x4af6b6d3e2ece2eb, 0x9c994dbec7ea08de,
+	0x49ace597b09a8bc4, 0xb38c4766cf0797ba, 0x131b9373c57c2a75, 0xb1822cce61931e58,
+	0x9d7555b909ba1c0c, 0x127fafdd937d11d2, 0x29da3badc66d92e4, 0xa2c1d57154c2ecbc,
+	0x58c5134d82f6fe24, 0x1c3ae3515b62274f, 0xe907c82e01cb8126, 0xf8ed091913e37fcb,
+	0x3249d8f9c80046c9, 0x80cf9bede388fb63, 0x1881539a116cf19e, 0x5103f3f76bd52457,
+	0x15b7e6f5ae47f7a8, 0xdbd7c6ded47e9ccf, 0x44e55c410228bb1a, 0xb647d4255edb4e99,
+	0x5d11882bb8aafc30, 0xf5098bbb29d3212a, 0x8fb5ea14e90296b3, 0x677b942157dd025a,
+	0xfb58e7c0a390acb5, 0x89d3674c83bd4a01, 0x9e2da4df4bf3b93b, 0xfcc41e328cab4829,
+	0x03f38c96ba582c52, 0xcad1bdbd7fd85db2, 0xbbb442c16082ae83, 0xb95fe86ba5da9ab0,
+	0xb22e04673771a93f, 0x845358c9493152d8, 0xbe2a488697b4541e, 0x95a2dc2dd38e6966,
+	0xc02c11ac923c852b, 0x2388b1990df2a87b, 0x7c8008fa1b4f37be, 0x1f70d0c84d54e503,
+	0x5490adec7ece57d4, 0x002b3c27d9063a3a, 0x7eaea3848030a2bf, 0xc602326ded2003c0,
+	0x83a7287d69a94086, 0xc57a5fcb30f57a8a, 0xb56844e479ebe779, 0xa373b40f05dcbce9,
+	0xd71a786e88570ee2, 0x879cbacdbde8f6a0, 0x976ad1bcc164a32f, 0xab21e25e9666d78b,
+	0x901063aae5e5c33c, 0x9818b34448698d90, 0xe36487ae3e1e8abb, 0xafbdf931893bdcb4,
+	0x6345a0dc5fbbd519, 0x8628fe269b9465ca, 0x1e5d01603f9c51ec, 0x4de44006a15049b7,
+	0xbf6c70e5f776cbb1, 0x411218f2ef552bed, 0xcb0c0708705a36a3, 0xe74d14754f986044,
+	0xcd56d9430ea8280e, 0xc12591d7535f5065, 0xc83223f1720aef96, 0xc3a0396f7363a51f,
+}
+
+Tiger_Context :: struct {
+    a:      u64,
+	b:      u64,
+	c:      u64,
+	x:      [64]byte,
+	nx:     int,
+	length: u64,
+	ver:    int,
+}
+
+round :: #force_inline proc "contextless" (a, b, c, x, mul: u64) -> (u64, u64, u64) {
+	a, b, c := a, b, c
+	c ~= x
+	a -= T1[c & 0xff] ~ T2[(c >> 16) & 0xff] ~ T3[(c >> 32) & 0xff] ~ T4[(c >> 48) & 0xff]
+	b += T4[(c >> 8) & 0xff] ~ T3[(c >> 24) & 0xff] ~ T2[(c >> 40) & 0xff] ~ T1[(c >> 56) & 0xff]
+	b *= mul
+	return a, b, c
+}
+
+pass :: #force_inline proc "contextless" (a, b, c: u64, d: []u64, mul: u64) -> (x, y, z: u64) {
+	x, y, z = round(a, b, c, d[0], mul)
+	y, z, x = round(y, z, x, d[1], mul)
+	z, x, y = round(z, x, y, d[2], mul)
+	x, y, z = round(x, y, z, d[3], mul)
+	y, z, x = round(y, z, x, d[4], mul)
+	z, x, y = round(z, x, y, d[5], mul)
+	x, y, z = round(x, y, z, d[6], mul)
+	y, z, x = round(y, z, x, d[7], mul)
+	return
+}
+
+key_schedule :: #force_inline proc "contextless" (x: []u64) {
+	x[0] -= x[7] ~ 0xa5a5a5a5a5a5a5a5
+	x[1] ~= x[0]
+	x[2] += x[1]
+	x[3] -= x[2] ~ ((~x[1]) << 19)
+	x[4] ~= x[3]
+	x[5] += x[4]
+	x[6] -= x[5] ~ ((~x[4]) >> 23)
+	x[7] ~= x[6]
+	x[0] += x[7]
+	x[1] -= x[0] ~ ((~x[7]) << 19)
+	x[2] ~= x[1]
+	x[3] += x[2]
+	x[4] -= x[3] ~ ((~x[2]) >> 23)
+	x[5] ~= x[4]
+	x[6] += x[5]
+	x[7] -= x[6] ~ 0x0123456789abcdef
+}
+
+compress :: #force_inline proc "contextless" (ctx: ^Tiger_Context, data: []byte) {
+	a := ctx.a
+	b := ctx.b
+	c := ctx.c
+	x := util.cast_slice([]u64, data)
+	ctx.a, ctx.b, ctx.c = pass(ctx.a, ctx.b, ctx.c, x, 5)
+	key_schedule(x)
+	ctx.c, ctx.a, ctx.b = pass(ctx.c, ctx.a, ctx.b, x, 7)
+	key_schedule(x)
+	ctx.b, ctx.c, ctx.a = pass(ctx.b, ctx.c, ctx.a, x, 9)
+	ctx.a ~= a
+	ctx.b -= b
+	ctx.c += c
+}
+
+init :: proc "contextless" (ctx: ^Tiger_Context) {
+	ctx.a = 0x0123456789abcdef
+	ctx.b = 0xfedcba9876543210
+	ctx.c = 0xf096a5b4c3b2e187
+}
+
+update :: proc(ctx: ^Tiger_Context, input: []byte) {
+	p := make([]byte, len(input))
+	copy(p, input)
+
+	length     := len(p)
+	ctx.length += u64(length)
+	if ctx.nx > 0 {
+		n := len(p)
+		if n > 64 - ctx.nx {
+			n = 64 - ctx.nx
+		}
+		copy(ctx.x[ctx.nx:ctx.nx + n], p[:n])
+		ctx.nx += n
+		if ctx.nx == 64 {
+			compress(ctx, ctx.x[:64 - 1])
+			ctx.nx = 0
+		}
+		p = p[n:]
+	}
+	for len(p) >= 64 {
+		compress(ctx, p[:64])
+		p = p[64:]
+	}
+	if len(p) > 0 {
+		ctx.nx = copy(ctx.x[:], p)
+	}
+}
+
+final :: proc(ctx: ^Tiger_Context, hash: []byte) {
+	length := ctx.length
+	tmp: [64]byte
+	if ctx.ver == 1 {
+		tmp[0] = 0x01
+	} else {
+		tmp[0] = 0x80
+	}
+
+	size := length & 0x3f
+	if size < 56 {
+		update(ctx, tmp[:56 - size])
+	} else {
+		update(ctx, tmp[:64 + 56 - size])
+	}
+
+	length <<= 3
+	for i := uint(0); i < 8; i += 1 {
+		tmp[i] = byte(length >> (8 * i))
+	}
+	update(ctx, tmp[:8])
+
+	for i := uint(0); i < 8; i += 1 {
+		tmp[i]      = byte(ctx.a >> (8 * i))
+		tmp[i + 8]  = byte(ctx.b >> (8 * i))
+		tmp[i + 16] = byte(ctx.c >> (8 * i))
+	}
+	copy(hash[:], tmp[:len(hash)])
+}
@@ -0,0 +1,726 @@
+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)
+}
@@ -0,0 +1,127 @@
+package blake2b
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the BLAKE2B hashing algorithm.
+    BLAKE2B and BLAKE2B share the implementation in the _blake2 package.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_blake2"
+
+/*
+    High level API
+*/
+
+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))
+}
+
+// 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
+}
+
+// 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: _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)
+}
+
+
+// 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)
+}
@@ -0,0 +1,127 @@
+package blake2s
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the BLAKE2S hashing algorithm.
+    BLAKE2B and BLAKE2B share the implementation in the _blake2 package.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_blake2"
+
+/*
+    High level API
+*/
+
+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: _blake2.Blake2s_Context
+    cfg: _blake2.Blake2_Config
+    cfg.size = _blake2.BLAKE2S_SIZE
+    ctx.cfg  = cfg
+    _blake2.init(&ctx)
+    _blake2.update(&ctx, data)
+    _blake2.final(&ctx, hash[:])
+    return hash
+}
+
+
+// hash_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: _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)
+}
+
+// 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)
+}
@@ -0,0 +1,591 @@
+package chacha20
+
+import "core:crypto/util"
+import "core:math/bits"
+import "core:mem"
+
+KEY_SIZE :: 32
+NONCE_SIZE :: 12
+XNONCE_SIZE :: 24
+
+@(private)
+_MAX_CTR_IETF :: 0xffffffff
+
+@(private)
+_BLOCK_SIZE :: 64
+@(private)
+_STATE_SIZE_U32 :: 16
+@(private)
+_ROUNDS :: 20
+
+@(private)
+_SIGMA_0 : u32 : 0x61707865
+@(private)
+_SIGMA_1 : u32 : 0x3320646e
+@(private)
+_SIGMA_2 : u32 : 0x79622d32
+@(private)
+_SIGMA_3 : u32 : 0x6b206574
+
+Context :: struct {
+	_s: [_STATE_SIZE_U32]u32,
+
+	_buffer: [_BLOCK_SIZE]byte,
+	_off: int,
+
+	_is_ietf_flavor: bool,
+	_is_initialized: bool,
+}
+
+init :: proc (ctx: ^Context, key, nonce: []byte) {
+	if len(key) != KEY_SIZE {
+		panic("crypto/chacha20: invalid ChaCha20 key size")
+	}
+	if n_len := len(nonce); n_len != NONCE_SIZE && n_len != XNONCE_SIZE {
+		panic("crypto/chacha20: invalid (X)ChaCha20 nonce size")
+	}
+
+	k, n := key, nonce
+
+	// Derive the XChaCha20 subkey and sub-nonce via HChaCha20.
+	is_xchacha := len(nonce) == XNONCE_SIZE
+	if is_xchacha {
+		sub_key := ctx._buffer[:KEY_SIZE]
+		_hchacha20(sub_key, k, n)
+		k = sub_key
+		n = n[16:24]
+	}
+
+	ctx._s[0] = _SIGMA_0
+	ctx._s[1] = _SIGMA_1
+	ctx._s[2] = _SIGMA_2
+	ctx._s[3] = _SIGMA_3
+	ctx._s[4] = util.U32_LE(k[0:4])
+	ctx._s[5] = util.U32_LE(k[4:8])
+	ctx._s[6] = util.U32_LE(k[8:12])
+	ctx._s[7] = util.U32_LE(k[12:16])
+	ctx._s[8] = util.U32_LE(k[16:20])
+	ctx._s[9] = util.U32_LE(k[20:24])
+	ctx._s[10] = util.U32_LE(k[24:28])
+	ctx._s[11] = util.U32_LE(k[28:32])
+	ctx._s[12] = 0
+	if !is_xchacha {
+		ctx._s[13] = util.U32_LE(n[0:4])
+		ctx._s[14] = util.U32_LE(n[4:8])
+		ctx._s[15] = util.U32_LE(n[8:12])
+	} else {
+		ctx._s[13] = 0
+		ctx._s[14] = util.U32_LE(n[0:4])
+		ctx._s[15] = util.U32_LE(n[4:8])
+
+		// The sub-key is stored in the keystream buffer.  While
+		// this will be overwritten in most circumstances, explicitly
+		// clear it out early.
+		mem.zero_explicit(&ctx._buffer, KEY_SIZE)
+	}
+
+	ctx._off = _BLOCK_SIZE
+	ctx._is_ietf_flavor = !is_xchacha
+	ctx._is_initialized = true
+}
+
+seek :: proc (ctx: ^Context, block_nr: u64) {
+	assert(ctx._is_initialized)
+
+	if ctx._is_ietf_flavor {
+		if block_nr > _MAX_CTR_IETF {
+			panic("crypto/chacha20: attempted to seek past maximum counter")
+		}
+	} else {
+		ctx._s[13] = u32(block_nr >> 32)
+	}
+	ctx._s[12] = u32(block_nr)
+	ctx._off = _BLOCK_SIZE
+}
+
+xor_bytes :: proc (ctx: ^Context, dst, src: []byte) {
+	assert(ctx._is_initialized)
+
+	// TODO: Enforcing that dst and src alias exactly or not at all
+	// is a good idea, though odd aliasing should be extremely uncommon.
+
+	src, dst := src, dst
+	if dst_len := len(dst); dst_len < len(src) {
+		src = src[:dst_len]
+	}
+
+	for remaining := len(src); remaining > 0; {
+		// Process multiple blocks at once
+		if ctx._off == _BLOCK_SIZE {
+			if nr_blocks := remaining / _BLOCK_SIZE; nr_blocks > 0 {
+				direct_bytes := nr_blocks * _BLOCK_SIZE
+				_do_blocks(ctx, dst, src, nr_blocks)
+				remaining -= direct_bytes
+				if remaining == 0 {
+					return
+				}
+				dst = dst[direct_bytes:]
+				src = src[direct_bytes:]
+			}
+
+			// If there is a partial block, generate and buffer 1 block
+			// worth of keystream.
+			_do_blocks(ctx, ctx._buffer[:], nil, 1)
+			ctx._off = 0
+		}
+
+		// Process partial blocks from the buffered keystream.
+		to_xor := min(_BLOCK_SIZE - ctx._off, remaining)
+		buffered_keystream := ctx._buffer[ctx._off:]
+		for i := 0; i < to_xor; i = i + 1 {
+			dst[i] = buffered_keystream[i] ~ src[i]
+		}
+		ctx._off += to_xor
+		dst = dst[to_xor:]
+		src = src[to_xor:]
+		remaining -= to_xor
+	}
+}
+
+keystream_bytes :: proc (ctx: ^Context, dst: []byte) {
+	assert(ctx._is_initialized)
+
+	dst := dst
+	for remaining := len(dst); remaining > 0; {
+		// Process multiple blocks at once
+		if ctx._off == _BLOCK_SIZE {
+			if nr_blocks := remaining / _BLOCK_SIZE; nr_blocks > 0 {
+				direct_bytes := nr_blocks * _BLOCK_SIZE
+				_do_blocks(ctx, dst, nil, nr_blocks)
+				remaining -= direct_bytes
+				if remaining == 0 {
+					return
+				}
+				dst = dst[direct_bytes:]
+			}
+
+			// If there is a partial block, generate and buffer 1 block
+			// worth of keystream.
+			_do_blocks(ctx, ctx._buffer[:], nil, 1)
+			ctx._off = 0
+		}
+
+		// Process partial blocks from the buffered keystream.
+		to_copy := min(_BLOCK_SIZE - ctx._off, remaining)
+		buffered_keystream := ctx._buffer[ctx._off:]
+		copy(dst[:to_copy], buffered_keystream[:to_copy])
+		ctx._off += to_copy
+		dst = dst[to_copy:]
+		remaining -= to_copy
+	}
+}
+
+reset :: proc (ctx: ^Context) {
+	mem.zero_explicit(&ctx._s, size_of(ctx._s))
+	mem.zero_explicit(&ctx._buffer, size_of(ctx._buffer))
+
+	ctx._is_initialized = false
+}
+
+@(private)
+_do_blocks :: proc (ctx: ^Context, dst, src: []byte, nr_blocks: int) {
+	// Enforce the maximum consumed keystream per nonce.
+	//
+	// While all modern "standard" definitions of ChaCha20 use
+	// the IETF 32-bit counter, for XChaCha20 most common
+	// implementations allow for a 64-bit counter.
+	//
+	// Honestly, the answer here is "use a MRAE primitive", but
+	// go with common practice in the case of XChaCha20.
+	if ctx._is_ietf_flavor {
+		if u64(ctx._s[12]) + u64(nr_blocks) > 0xffffffff {
+			panic("crypto/chacha20: maximum ChaCha20 keystream per nonce reached")
+		}
+	} else {
+		ctr := (u64(ctx._s[13]) << 32) | u64(ctx._s[12])
+		if _, carry := bits.add_u64(ctr, u64(nr_blocks), 0); carry != 0 {
+			panic("crypto/chacha20: maximum XChaCha20 keystream per nonce reached")
+		}
+	}
+
+	dst, src := dst, src
+	x := &ctx._s
+	for n := 0; n < nr_blocks; n = n + 1 {
+		x0, x1, x2, x3 := _SIGMA_0, _SIGMA_1, _SIGMA_2, _SIGMA_3
+		x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15 := x[4], x[5], x[6], x[7], x[8], x[9], x[10], x[11], x[12], x[13], x[14], x[15]
+
+		for i := _ROUNDS; i > 0; i = i - 2 {
+			// Even when forcing inlining manually inlining all of
+			// these is decently faster.
+
+			// quarterround(x, 0, 4, 8, 12)
+			x0 += x4
+			x12 ~= x0
+			x12 = util.ROTL32(x12, 16)
+			x8 += x12
+			x4 ~= x8
+			x4 = util.ROTL32(x4, 12)
+			x0 += x4
+			x12 ~= x0
+			x12 = util.ROTL32(x12, 8)
+			x8 += x12
+			x4 ~= x8
+			x4 = util.ROTL32(x4, 7)
+
+			// quarterround(x, 1, 5, 9, 13)
+			x1 += x5
+			x13 ~= x1
+			x13 = util.ROTL32(x13, 16)
+			x9 += x13
+			x5 ~= x9
+			x5 = util.ROTL32(x5, 12)
+			x1 += x5
+			x13 ~= x1
+			x13 = util.ROTL32(x13, 8)
+			x9 += x13
+			x5 ~= x9
+			x5 = util.ROTL32(x5, 7)
+
+			// quarterround(x, 2, 6, 10, 14)
+			x2 += x6
+			x14 ~= x2
+			x14 = util.ROTL32(x14, 16)
+			x10 += x14
+			x6 ~= x10
+			x6 = util.ROTL32(x6, 12)
+			x2 += x6
+			x14 ~= x2
+			x14 = util.ROTL32(x14, 8)
+			x10 += x14
+			x6 ~= x10
+			x6 = util.ROTL32(x6, 7)
+
+			// quarterround(x, 3, 7, 11, 15)
+			x3 += x7
+			x15 ~= x3
+			x15 = util.ROTL32(x15, 16)
+			x11 += x15
+			x7 ~= x11
+			x7 = util.ROTL32(x7, 12)
+			x3 += x7
+			x15 ~= x3
+			x15 = util.ROTL32(x15, 8)
+			x11 += x15
+			x7 ~= x11
+			x7 = util.ROTL32(x7, 7)
+
+			// quarterround(x, 0, 5, 10, 15)
+			x0 += x5
+			x15 ~= x0
+			x15 = util.ROTL32(x15, 16)
+			x10 += x15
+			x5 ~= x10
+			x5 = util.ROTL32(x5, 12)
+			x0 += x5
+			x15 ~= x0
+			x15 = util.ROTL32(x15, 8)
+			x10 += x15
+			x5 ~= x10
+			x5 = util.ROTL32(x5, 7)
+
+			// quarterround(x, 1, 6, 11, 12)
+			x1 += x6
+			x12 ~= x1
+			x12 = util.ROTL32(x12, 16)
+			x11 += x12
+			x6 ~= x11
+			x6 = util.ROTL32(x6, 12)
+			x1 += x6
+			x12 ~= x1
+			x12 = util.ROTL32(x12, 8)
+			x11 += x12
+			x6 ~= x11
+			x6 = util.ROTL32(x6, 7)
+
+			// quarterround(x, 2, 7, 8, 13)
+			x2 += x7
+			x13 ~= x2
+			x13 = util.ROTL32(x13, 16)
+			x8 += x13
+			x7 ~= x8
+			x7 = util.ROTL32(x7, 12)
+			x2 += x7
+			x13 ~= x2
+			x13 = util.ROTL32(x13, 8)
+			x8 += x13
+			x7 ~= x8
+			x7 = util.ROTL32(x7, 7)
+
+			// quarterround(x, 3, 4, 9, 14)
+			x3 += x4
+			x14 ~= x3
+			x14 = util.ROTL32(x14, 16)
+			x9 += x14
+			x4 ~= x9
+			x4 = util.ROTL32(x4, 12)
+			x3 += x4
+			x14 ~= x3
+			x14 = util.ROTL32(x14, 8)
+			x9 += x14
+			x4 ~= x9
+			x4 = util.ROTL32(x4, 7)
+		}
+
+		x0 += _SIGMA_0
+		x1 += _SIGMA_1
+		x2 += _SIGMA_2
+		x3 += _SIGMA_3
+		x4 += x[4]
+		x5 += x[5]
+		x6 += x[6]
+		x7 += x[7]
+		x8 += x[8]
+		x9 += x[9]
+		x10 += x[10]
+		x11 += x[11]
+		x12 += x[12]
+		x13 += x[13]
+		x14 += x[14]
+		x15 += x[15]
+
+		// While the "correct" answer to getting more performance out of
+		// this is "use vector operations", support for that is currently
+		// a work in progress/to be designed.
+		//
+		// Until dedicated assembly can be written leverage the fact that
+		// the callers of this routine ensure that src/dst are valid.
+
+		when ODIN_ARCH == .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])
+			if src != nil {
+				src_p := transmute(^[16]u32)(&src[0])
+				dst_p[0] = src_p[0] ~ x0
+				dst_p[1] = src_p[1] ~ x1
+				dst_p[2] = src_p[2] ~ x2
+				dst_p[3] = src_p[3] ~ x3
+				dst_p[4] = src_p[4] ~ x4
+				dst_p[5] = src_p[5] ~ x5
+				dst_p[6] = src_p[6] ~ x6
+				dst_p[7] = src_p[7] ~ x7
+				dst_p[8] = src_p[8] ~ x8
+				dst_p[9] = src_p[9] ~ x9
+				dst_p[10] = src_p[10] ~ x10
+				dst_p[11] = src_p[11] ~ x11
+				dst_p[12] = src_p[12] ~ x12
+				dst_p[13] = src_p[13] ~ x13
+				dst_p[14] = src_p[14] ~ x14
+				dst_p[15] = src_p[15] ~ x15
+				src = src[_BLOCK_SIZE:]
+			} else {
+				dst_p[0] = x0
+				dst_p[1] = x1
+				dst_p[2] = x2
+				dst_p[3] = x3
+				dst_p[4] = x4
+				dst_p[5] = x5
+				dst_p[6] = x6
+				dst_p[7] = x7
+				dst_p[8] = x8
+				dst_p[9] = x9
+				dst_p[10] = x10
+				dst_p[11] = x11
+				dst_p[12] = x12
+				dst_p[13] = x13
+				dst_p[14] = x14
+				dst_p[15] = x15
+			}
+			dst = dst[_BLOCK_SIZE:]
+		} else {
+			#no_bounds_check {
+				if src != nil {
+					util.PUT_U32_LE(dst[0:4], util.U32_LE(src[0:4]) ~ x0)
+					util.PUT_U32_LE(dst[4:8], util.U32_LE(src[4:8]) ~ x1)
+					util.PUT_U32_LE(dst[8:12], util.U32_LE(src[8:12]) ~ x2)
+					util.PUT_U32_LE(dst[12:16], util.U32_LE(src[12:16]) ~ x3)
+					util.PUT_U32_LE(dst[16:20], util.U32_LE(src[16:20]) ~ x4)
+					util.PUT_U32_LE(dst[20:24], util.U32_LE(src[20:24]) ~ x5)
+					util.PUT_U32_LE(dst[24:28], util.U32_LE(src[24:28]) ~ x6)
+					util.PUT_U32_LE(dst[28:32], util.U32_LE(src[28:32]) ~ x7)
+					util.PUT_U32_LE(dst[32:36], util.U32_LE(src[32:36]) ~ x8)
+					util.PUT_U32_LE(dst[36:40], util.U32_LE(src[36:40]) ~ x9)
+					util.PUT_U32_LE(dst[40:44], util.U32_LE(src[40:44]) ~ x10)
+					util.PUT_U32_LE(dst[44:48], util.U32_LE(src[44:48]) ~ x11)
+					util.PUT_U32_LE(dst[48:52], util.U32_LE(src[48:52]) ~ x12)
+					util.PUT_U32_LE(dst[52:56], util.U32_LE(src[52:56]) ~ x13)
+					util.PUT_U32_LE(dst[56:60], util.U32_LE(src[56:60]) ~ x14)
+					util.PUT_U32_LE(dst[60:64], util.U32_LE(src[60:64]) ~ x15)
+					src = src[_BLOCK_SIZE:]
+				} else {
+					util.PUT_U32_LE(dst[0:4], x0)
+					util.PUT_U32_LE(dst[4:8], x1)
+					util.PUT_U32_LE(dst[8:12], x2)
+					util.PUT_U32_LE(dst[12:16], x3)
+					util.PUT_U32_LE(dst[16:20], x4)
+					util.PUT_U32_LE(dst[20:24], x5)
+					util.PUT_U32_LE(dst[24:28], x6)
+					util.PUT_U32_LE(dst[28:32], x7)
+					util.PUT_U32_LE(dst[32:36], x8)
+					util.PUT_U32_LE(dst[36:40], x9)
+					util.PUT_U32_LE(dst[40:44], x10)
+					util.PUT_U32_LE(dst[44:48], x11)
+					util.PUT_U32_LE(dst[48:52], x12)
+					util.PUT_U32_LE(dst[52:56], x13)
+					util.PUT_U32_LE(dst[56:60], x14)
+					util.PUT_U32_LE(dst[60:64], x15)
+				}
+				dst = dst[_BLOCK_SIZE:]
+			}
+		}
+
+		// Increment the counter.  Overflow checking is done upon
+		// entry into the routine, so a 64-bit increment safely
+		// covers both cases.
+		new_ctr := ((u64(ctx._s[13]) << 32) | u64(ctx._s[12])) + 1
+		x[12] = u32(new_ctr)
+		x[13] = u32(new_ctr >> 32)
+	}
+}
+
+@(private)
+_hchacha20 :: proc (dst, key, nonce: []byte) {
+	x0, x1, x2, x3 := _SIGMA_0, _SIGMA_1, _SIGMA_2, _SIGMA_3
+	x4 := util.U32_LE(key[0:4])
+	x5 := util.U32_LE(key[4:8])
+	x6 := util.U32_LE(key[8:12])
+	x7 := util.U32_LE(key[12:16])
+	x8 := util.U32_LE(key[16:20])
+	x9 := util.U32_LE(key[20:24])
+	x10 := util.U32_LE(key[24:28])
+	x11 := util.U32_LE(key[28:32])
+	x12 := util.U32_LE(nonce[0:4])
+	x13 := util.U32_LE(nonce[4:8])
+	x14 := util.U32_LE(nonce[8:12])
+	x15 := util.U32_LE(nonce[12:16])
+
+	for i := _ROUNDS; i > 0; i = i - 2 {
+		// quarterround(x, 0, 4, 8, 12)
+		x0 += x4
+		x12 ~= x0
+		x12 = util.ROTL32(x12, 16)
+		x8 += x12
+		x4 ~= x8
+		x4 = util.ROTL32(x4, 12)
+		x0 += x4
+		x12 ~= x0
+		x12 = util.ROTL32(x12, 8)
+		x8 += x12
+		x4 ~= x8
+		x4 = util.ROTL32(x4, 7)
+
+		// quarterround(x, 1, 5, 9, 13)
+		x1 += x5
+		x13 ~= x1
+		x13 = util.ROTL32(x13, 16)
+		x9 += x13
+		x5 ~= x9
+		x5 = util.ROTL32(x5, 12)
+		x1 += x5
+		x13 ~= x1
+		x13 = util.ROTL32(x13, 8)
+		x9 += x13
+		x5 ~= x9
+		x5 = util.ROTL32(x5, 7)
+
+		// quarterround(x, 2, 6, 10, 14)
+		x2 += x6
+		x14 ~= x2
+		x14 = util.ROTL32(x14, 16)
+		x10 += x14
+		x6 ~= x10
+		x6 = util.ROTL32(x6, 12)
+		x2 += x6
+		x14 ~= x2
+		x14 = util.ROTL32(x14, 8)
+		x10 += x14
+		x6 ~= x10
+		x6 = util.ROTL32(x6, 7)
+
+		// quarterround(x, 3, 7, 11, 15)
+		x3 += x7
+		x15 ~= x3
+		x15 = util.ROTL32(x15, 16)
+		x11 += x15
+		x7 ~= x11
+		x7 = util.ROTL32(x7, 12)
+		x3 += x7
+		x15 ~= x3
+		x15 = util.ROTL32(x15, 8)
+		x11 += x15
+		x7 ~= x11
+		x7 = util.ROTL32(x7, 7)
+
+		// quarterround(x, 0, 5, 10, 15)
+		x0 += x5
+		x15 ~= x0
+		x15 = util.ROTL32(x15, 16)
+		x10 += x15
+		x5 ~= x10
+		x5 = util.ROTL32(x5, 12)
+		x0 += x5
+		x15 ~= x0
+		x15 = util.ROTL32(x15, 8)
+		x10 += x15
+		x5 ~= x10
+		x5 = util.ROTL32(x5, 7)
+
+		// quarterround(x, 1, 6, 11, 12)
+		x1 += x6
+		x12 ~= x1
+		x12 = util.ROTL32(x12, 16)
+		x11 += x12
+		x6 ~= x11
+		x6 = util.ROTL32(x6, 12)
+		x1 += x6
+		x12 ~= x1
+		x12 = util.ROTL32(x12, 8)
+		x11 += x12
+		x6 ~= x11
+		x6 = util.ROTL32(x6, 7)
+
+		// quarterround(x, 2, 7, 8, 13)
+		x2 += x7
+		x13 ~= x2
+		x13 = util.ROTL32(x13, 16)
+		x8 += x13
+		x7 ~= x8
+		x7 = util.ROTL32(x7, 12)
+		x2 += x7
+		x13 ~= x2
+		x13 = util.ROTL32(x13, 8)
+		x8 += x13
+		x7 ~= x8
+		x7 = util.ROTL32(x7, 7)
+
+		// quarterround(x, 3, 4, 9, 14)
+		x3 += x4
+		x14 ~= x3
+		x14 = util.ROTL32(x14, 16)
+		x9 += x14
+		x4 ~= x9
+		x4 = util.ROTL32(x4, 12)
+		x3 += x4
+		x14 ~= x3
+		x14 = util.ROTL32(x14, 8)
+		x9 += x14
+		x4 ~= x9
+		x4 = util.ROTL32(x4, 7)
+	}
+
+	util.PUT_U32_LE(dst[0:4], x0)
+	util.PUT_U32_LE(dst[4:8], x1)
+	util.PUT_U32_LE(dst[8:12], x2)
+	util.PUT_U32_LE(dst[12:16], x3)
+	util.PUT_U32_LE(dst[16:20], x12)
+	util.PUT_U32_LE(dst[20:24], x13)
+	util.PUT_U32_LE(dst[24:28], x14)
+	util.PUT_U32_LE(dst[28:32], x15)
+}
@@ -0,0 +1,151 @@
+package chacha20poly1305
+
+import "core:crypto"
+import "core:crypto/chacha20"
+import "core:crypto/poly1305"
+import "core:crypto/util"
+import "core:mem"
+
+KEY_SIZE :: chacha20.KEY_SIZE
+NONCE_SIZE :: chacha20.NONCE_SIZE
+TAG_SIZE :: poly1305.TAG_SIZE
+
+@(private)
+_P_MAX :: 64 * 0xffffffff // 64 * (2^32-1)
+
+@(private)
+_validate_common_slice_sizes :: proc (tag, key, nonce, aad, text: []byte) {
+	if len(tag) != TAG_SIZE {
+		panic("crypto/chacha20poly1305: invalid destination tag size")
+	}
+	if len(key) != KEY_SIZE {
+		panic("crypto/chacha20poly1305: invalid key size")
+	}
+	if len(nonce) != NONCE_SIZE {
+		panic("crypto/chacha20poly1305: invalid nonce size")
+	}
+
+	#assert(size_of(int) == 8 || size_of(int) <= 4)
+	when size_of(int) == 8 {
+		// A_MAX = 2^64 - 1 due to the length field limit.
+		// P_MAX = 64 * (2^32 - 1) due to the IETF ChaCha20 counter limit.
+		//
+		// A_MAX is limited by size_of(int), so there is no need to
+		// enforce it. P_MAX only needs to be checked on 64-bit targets,
+		// for reasons that should be obvious.
+		if text_len := len(text); text_len > _P_MAX {
+			panic("crypto/chacha20poly1305: oversized src data")
+		}
+	}
+}
+
+@(private)
+_PAD: [16]byte
+
+@(private)
+_update_mac_pad16 :: #force_inline proc (ctx: ^poly1305.Context, x_len: int) {
+	if pad_len := 16 - (x_len & (16-1)); pad_len != 16 {
+		poly1305.update(ctx, _PAD[:pad_len])
+	}
+}
+
+encrypt :: proc (ciphertext, tag, key, nonce, aad, plaintext: []byte) {
+	_validate_common_slice_sizes(tag, key, nonce, aad, plaintext)
+	if len(ciphertext) != len(plaintext) {
+		panic("crypto/chacha20poly1305: invalid destination ciphertext size")
+	}
+
+	stream_ctx: chacha20.Context = ---
+	chacha20.init(&stream_ctx, key, nonce)
+
+	// otk = poly1305_key_gen(key, nonce)
+	otk: [poly1305.KEY_SIZE]byte = ---
+	chacha20.keystream_bytes(&stream_ctx, otk[:])
+	mac_ctx: poly1305.Context = ---
+	poly1305.init(&mac_ctx, otk[:])
+	mem.zero_explicit(&otk, size_of(otk))
+
+	aad_len, ciphertext_len := len(aad), len(ciphertext)
+
+	// There is nothing preventing aad and ciphertext from overlapping
+	// so auth the AAD before encrypting (slightly different from the
+	// RFC, since the RFC encrypts into a new buffer).
+	//
+	// mac_data = aad | pad16(aad)
+	poly1305.update(&mac_ctx, aad)
+	_update_mac_pad16(&mac_ctx, aad_len)
+
+	// ciphertext = chacha20_encrypt(key, 1, nonce, plaintext)
+	chacha20.seek(&stream_ctx, 1)
+	chacha20.xor_bytes(&stream_ctx, ciphertext, plaintext)
+	chacha20.reset(&stream_ctx) // Don't need the stream context anymore.
+
+	// mac_data |= ciphertext | pad16(ciphertext)
+	poly1305.update(&mac_ctx, ciphertext)
+	_update_mac_pad16(&mac_ctx, ciphertext_len)
+
+	// mac_data |= num_to_8_le_bytes(aad.length)
+	// mac_data |= num_to_8_le_bytes(ciphertext.length)
+	l_buf := otk[0:16] // Reuse the scratch buffer.
+	util.PUT_U64_LE(l_buf[0:8], u64(aad_len))
+	util.PUT_U64_LE(l_buf[8:16], u64(ciphertext_len))
+	poly1305.update(&mac_ctx, l_buf)
+
+	// tag = poly1305_mac(mac_data, otk)
+	poly1305.final(&mac_ctx, tag) // Implicitly sanitizes context.
+}
+
+decrypt :: proc (plaintext, tag, key, nonce, aad, ciphertext: []byte) -> bool {
+	_validate_common_slice_sizes(tag, key, nonce, aad, ciphertext)
+	if len(ciphertext) != len(plaintext) {
+		panic("crypto/chacha20poly1305: invalid destination plaintext size")
+	}
+
+	// Note: Unlike encrypt, this can fail early, so use defer for
+	// sanitization rather than assuming control flow reaches certain
+	// points where needed.
+
+	stream_ctx: chacha20.Context = ---
+	chacha20.init(&stream_ctx, key, nonce)
+
+	// otk = poly1305_key_gen(key, nonce)
+	otk: [poly1305.KEY_SIZE]byte = ---
+	chacha20.keystream_bytes(&stream_ctx, otk[:])
+	defer chacha20.reset(&stream_ctx)
+
+	mac_ctx: poly1305.Context = ---
+	poly1305.init(&mac_ctx, otk[:])
+	defer mem.zero_explicit(&otk, size_of(otk))
+
+	aad_len, ciphertext_len := len(aad), len(ciphertext)
+
+	// mac_data = aad | pad16(aad)
+	// mac_data |= ciphertext | pad16(ciphertext)
+	// mac_data |= num_to_8_le_bytes(aad.length)
+	// mac_data |= num_to_8_le_bytes(ciphertext.length)
+	poly1305.update(&mac_ctx, aad)
+	_update_mac_pad16(&mac_ctx, aad_len)
+	poly1305.update(&mac_ctx, ciphertext)
+	_update_mac_pad16(&mac_ctx, ciphertext_len)
+	l_buf := otk[0:16] // Reuse the scratch buffer.
+	util.PUT_U64_LE(l_buf[0:8], u64(aad_len))
+	util.PUT_U64_LE(l_buf[8:16], u64(ciphertext_len))
+	poly1305.update(&mac_ctx, l_buf)
+
+	// tag = poly1305_mac(mac_data, otk)
+	derived_tag := otk[0:poly1305.TAG_SIZE] // Reuse the scratch buffer again.
+	poly1305.final(&mac_ctx, derived_tag) // Implicitly sanitizes context.
+
+	// Validate the tag in constant time.
+	if crypto.compare_constant_time(tag, derived_tag) != 1 {
+		// Zero out the plaintext, as a defense in depth measure.
+		mem.zero_explicit(raw_data(plaintext), ciphertext_len)
+		return false
+	}
+
+	// plaintext = chacha20_decrypt(key, 1, nonce, ciphertext)
+	chacha20.seek(&stream_ctx, 1)
+	chacha20.xor_bytes(&stream_ctx, plaintext, ciphertext)
+
+	return true
+}
@@ -0,0 +1,53 @@
+package crypto
+
+import "core:mem"
+
+// compare_constant_time returns 1 iff a and b are equal, 0 otherwise.
+//
+// The execution time of this routine is constant regardless of the contents
+// of the slices being compared, as long as the length of the slices is equal.
+// If the length of the two slices is different, it will early-return 0.
+compare_constant_time :: proc "contextless" (a, b: []byte) -> int {
+	// If the length of the slices is different, early return.
+	//
+	// This leaks the fact that the slices have a different length,
+	// but the routine is primarily intended for comparing things
+	// like MACS and password digests.
+	n := len(a)
+	if n != len(b) {
+		return 0
+	}
+
+	return compare_byte_ptrs_constant_time(raw_data(a), raw_data(b), n)
+}
+
+// compare_byte_ptrs_constant_time returns 1 iff the bytes pointed to by
+// a and b are equal, 0 otherwise.
+//
+// The execution time of this routine is constant regardless of the
+// contents of the memory being compared.
+@(optimization_mode="none")
+compare_byte_ptrs_constant_time :: proc "contextless" (a, b: ^byte, n: int) -> int {
+	x := mem.slice_ptr(a, n)
+	y := mem.slice_ptr(b, n)
+
+	v: byte
+	for i in 0..<n {
+		v |= x[i] ~ y[i]
+	}
+
+	// After the loop, v == 0 iff a == b.  The subtraction will underflow
+	// iff v == 0, setting the sign-bit, which gets returned.
+	return int((u32(v)-1) >> 31)
+}
+
+// rand_bytes fills the dst buffer with cryptographic entropy taken from
+// the system entropy source.  This routine will block if the system entropy
+// source is not ready yet.  All system entropy source failures are treated
+// as catastrophic, resulting in a panic.
+rand_bytes :: proc (dst: []byte) {
+	// zero-fill the buffer first
+	mem.zero_explicit(raw_data(dst), len(dst))
+
+	_rand_bytes(dst)
+}
@@ -0,0 +1,382 @@
+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]
+}
@@ -0,0 +1,653 @@
+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
+            }
+    }
+}
@@ -0,0 +1,584 @@
+package jh
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the JH hashing algorithm, as defined in <https://www3.ntu.edu.sg/home/wuhj/research/jh/index.html>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: Jh_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: ^Jh_Context) {
+    assert(ctx.hashbitlen == 224 || ctx.hashbitlen == 256 || ctx.hashbitlen == 384 || ctx.hashbitlen == 512, "hashbitlen must be set to 224, 256, 384 or 512")
+    ctx.H[1] = byte(ctx.hashbitlen)      & 0xff
+    ctx.H[0] = byte(ctx.hashbitlen >> 8) & 0xff
+    F8(ctx)
+}
+
+update :: proc(ctx: ^Jh_Context, data: []byte) {
+    databitlen     := u64(len(data)) * 8
+    ctx.databitlen += databitlen
+    i              := u64(0)
+
+    if (ctx.buffer_size > 0) && ((ctx.buffer_size + databitlen) < 512) {
+        if (databitlen & 7) == 0 {
+            copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3)])
+        } else {
+            copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3) + 1])
+        } 
+        ctx.buffer_size += databitlen
+        databitlen = 0
+    }
+
+    if (ctx.buffer_size > 0 ) && ((ctx.buffer_size + databitlen) >= 512) {
+        copy(ctx.buffer[ctx.buffer_size >> 3:], data[:64 - (ctx.buffer_size >> 3)])
+        i      = 64 - (ctx.buffer_size >> 3)
+        databitlen = databitlen - (512 - ctx.buffer_size)
+        F8(ctx)
+        ctx.buffer_size = 0
+    }
+
+    for databitlen >= 512 {
+        copy(ctx.buffer[:], data[i:i + 64])
+        F8(ctx)
+        i += 64
+        databitlen -= 512
+    }
+
+    if databitlen > 0 {
+        if (databitlen & 7) == 0 {
+            copy(ctx.buffer[:], data[i:i + ((databitlen & 0x1ff) >> 3)])
+        } else {
+            copy(ctx.buffer[:], data[i:i + ((databitlen & 0x1ff) >> 3) + 1])
+        }
+        ctx.buffer_size = databitlen
+    }
+}
+
+final :: proc(ctx: ^Jh_Context, hash: []byte) {
+    if ctx.databitlen & 0x1ff == 0 {
+        for i := 0; i < 64; i += 1 {
+            ctx.buffer[i] = 0
+        }
+        ctx.buffer[0]  = 0x80
+        ctx.buffer[63] = byte(ctx.databitlen)       & 0xff
+        ctx.buffer[62] = byte(ctx.databitlen >> 8)  & 0xff
+        ctx.buffer[61] = byte(ctx.databitlen >> 16) & 0xff
+        ctx.buffer[60] = byte(ctx.databitlen >> 24) & 0xff
+        ctx.buffer[59] = byte(ctx.databitlen >> 32) & 0xff
+        ctx.buffer[58] = byte(ctx.databitlen >> 40) & 0xff
+        ctx.buffer[57] = byte(ctx.databitlen >> 48) & 0xff
+        ctx.buffer[56] = byte(ctx.databitlen >> 56) & 0xff
+        F8(ctx)
+    } else {
+        if ctx.buffer_size & 7 == 0 {
+            for i := (ctx.databitlen & 0x1ff) >> 3; i < 64; i += 1 {
+                ctx.buffer[i] = 0
+            }
+        } else {
+            for i := ((ctx.databitlen & 0x1ff) >> 3) + 1; i < 64; i += 1 {
+                ctx.buffer[i] = 0
+            }
+        }
+        ctx.buffer[(ctx.databitlen & 0x1ff) >> 3] |= 1 << (7 - (ctx.databitlen & 7))
+        F8(ctx)
+        for i := 0; i < 64; i += 1 {
+            ctx.buffer[i] = 0
+        }
+        ctx.buffer[63] = byte(ctx.databitlen)       & 0xff
+        ctx.buffer[62] = byte(ctx.databitlen >> 8)  & 0xff
+        ctx.buffer[61] = byte(ctx.databitlen >> 16) & 0xff
+        ctx.buffer[60] = byte(ctx.databitlen >> 24) & 0xff
+        ctx.buffer[59] = byte(ctx.databitlen >> 32) & 0xff
+        ctx.buffer[58] = byte(ctx.databitlen >> 40) & 0xff
+        ctx.buffer[57] = byte(ctx.databitlen >> 48) & 0xff
+        ctx.buffer[56] = byte(ctx.databitlen >> 56) & 0xff
+        F8(ctx)
+    }
+    switch ctx.hashbitlen {
+        case 224: copy(hash[:], ctx.H[100:128])
+        case 256: copy(hash[:], ctx.H[96:128])
+        case 384: copy(hash[:], ctx.H[80:128])
+        case 512: copy(hash[:], ctx.H[64:128])
+    }
+}
+
+/*
+    JH implementation
+*/
+
+ROUNDCONSTANT_ZERO := [64]byte {
+    0x6, 0xa, 0x0, 0x9, 0xe, 0x6, 0x6, 0x7,
+    0xf, 0x3, 0xb, 0xc, 0xc, 0x9, 0x0, 0x8,
+    0xb, 0x2, 0xf, 0xb, 0x1, 0x3, 0x6, 0x6,
+    0xe, 0xa, 0x9, 0x5, 0x7, 0xd, 0x3, 0xe,
+    0x3, 0xa, 0xd, 0xe, 0xc, 0x1, 0x7, 0x5,
+    0x1, 0x2, 0x7, 0x7, 0x5, 0x0, 0x9, 0x9,
+    0xd, 0xa, 0x2, 0xf, 0x5, 0x9, 0x0, 0xb,
+    0x0, 0x6, 0x6, 0x7, 0x3, 0x2, 0x2, 0xa,
+}
+
+SBOX := [2][16]byte {
+    {9, 0,  4, 11, 13, 12, 3, 15, 1,  10, 2, 6, 7,  5,  8,  14},
+    {3, 12, 6, 13, 5,  7,  1, 9,  15, 2,  0, 4, 11, 10, 14, 8},
+}
+
+Jh_Context :: struct {
+    hashbitlen:    int,
+    databitlen:    u64,
+    buffer_size:   u64,
+    H:             [128]byte,
+    A:             [256]byte,
+    roundconstant: [64]byte,
+    buffer:        [64]byte,
+}
+
+E8_finaldegroup :: proc(ctx: ^Jh_Context) {
+    t0,t1,t2,t3: byte
+    tem: [256]byte
+    for i := 0; i < 128; i += 1 {
+        tem[i]       = ctx.A[i << 1]
+        tem[i + 128] = ctx.A[(i << 1) + 1]
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.H[i] = 0
+    }
+    for i := 0; i < 256; i += 1 {
+        t0 = (tem[i] >> 3) & 1
+        t1 = (tem[i] >> 2) & 1
+        t2 = (tem[i] >> 1) & 1
+        t3 = (tem[i] >> 0) & 1
+
+        ctx.H[uint(i) >> 3]         |= t0 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 256) >> 3] |= t1 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 512) >> 3] |= t2 << (7 - (uint(i) & 7))
+        ctx.H[(uint(i) + 768) >> 3] |= t3 << (7 - (uint(i) & 7))
+    }
+}
+
+update_roundconstant :: proc(ctx: ^Jh_Context) {
+    tem: [64]byte
+    t: byte
+    for i := 0; i < 64; i += 1 {
+        tem[i] = SBOX[0][ctx.roundconstant[i]]
+    }
+    for i := 0; i < 64; i += 2 {
+        tem[i + 1] ~= ((tem[i]   << 1)   ~ (tem[i]   >> 3)   ~ ((tem[i]   >> 2) & 2))   & 0xf
+        tem[i]     ~= ((tem[i + 1] << 1) ~ (tem[i + 1] >> 3) ~ ((tem[i + 1] >> 2) & 2)) & 0xf
+    }
+    for i := 0; i < 64; i += 4 {
+        t          = tem[i + 2]
+        tem[i + 2] = tem[i + 3]
+        tem[i + 3] = t
+    }
+    for i := 0; i < 32; i += 1 {
+        ctx.roundconstant[i]      = tem[i << 1]
+        ctx.roundconstant[i + 32] = tem[(i << 1) + 1]
+    }
+    for i := 32; i < 64; i += 2 {
+        t                        = ctx.roundconstant[i]
+        ctx.roundconstant[i]     = ctx.roundconstant[i + 1]
+        ctx.roundconstant[i + 1] = t
+    }
+}
+
+R8 :: proc(ctx: ^Jh_Context) {
+    t: byte
+    tem, roundconstant_expanded: [256]byte
+    for i := u32(0); i < 256; i += 1 {
+        roundconstant_expanded[i] = (ctx.roundconstant[i >> 2] >> (3 - (i & 3)) ) & 1
+    }
+    for i := 0; i < 256; i += 1 {
+        tem[i] = SBOX[roundconstant_expanded[i]][ctx.A[i]]
+    }
+    for i := 0; i < 256; i += 2 {
+        tem[i+1] ~= ((tem[i]   << 1)   ~ (tem[i]   >> 3)   ~ ((tem[i]   >> 2) & 2))   & 0xf
+        tem[i]   ~= ((tem[i + 1] << 1) ~ (tem[i + 1] >> 3) ~ ((tem[i + 1] >> 2) & 2)) & 0xf
+    }
+    for i := 0; i < 256; i += 4 {
+        t        = tem[i + 2]
+        tem[i+2] = tem[i + 3]
+        tem[i+3] = t
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.A[i]       = tem[i << 1]
+        ctx.A[i + 128] = tem[(i << 1) + 1]
+    }
+    for i := 128; i < 256; i += 2 {
+        t            = ctx.A[i]
+        ctx.A[i]     = ctx.A[i + 1]
+        ctx.A[i + 1] = t
+    }
+}
+
+E8_initialgroup :: proc(ctx: ^Jh_Context) {
+    t0, t1, t2, t3: byte
+    tem:            [256]byte
+    for i := u32(0); i < 256; i += 1 {
+        t0     = (ctx.H[i >> 3]   >> (7 - (i & 7)))       & 1
+        t1     = (ctx.H[(i + 256) >> 3] >> (7 - (i & 7))) & 1
+        t2     = (ctx.H[(i + 512) >> 3] >> (7 - (i & 7))) & 1
+        t3     = (ctx.H[(i + 768) >> 3] >> (7 - (i & 7))) & 1
+        tem[i] = (t0 << 3) | (t1 << 2) | (t2 << 1) | (t3 << 0)
+    }
+    for i := 0; i < 128; i += 1 {
+        ctx.A[i << 1]       = tem[i]
+        ctx.A[(i << 1) + 1] = tem[i + 128]
+    }
+}
+
+E8 :: proc(ctx: ^Jh_Context) {
+    for i := 0; i < 64; i += 1 {
+        ctx.roundconstant[i] = ROUNDCONSTANT_ZERO[i]
+    }
+    E8_initialgroup(ctx)
+    for i := 0; i < 42; i += 1 {
+        R8(ctx)
+        update_roundconstant(ctx)
+    }
+    E8_finaldegroup(ctx)
+}
+
+F8 :: proc(ctx: ^Jh_Context) {
+    for i := 0; i < 64; i += 1 {
+        ctx.H[i] ~= ctx.buffer[i]
+    }
+    E8(ctx)
+    for i := 0; i < 64; i += 1 {
+        ctx.H[i + 64] ~= ctx.buffer[i]
+    }
+}
@@ -0,0 +1,374 @@
+package keccak
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Interface for the Keccak hashing algorithm.
+    This is done because the padding in the SHA3 standard was changed by the NIST, resulting in a different output.
+*/
+
+import "core:os"
+import "core:io"
+
+import "../_sha3"
+
+
+/*
+    High level API
+*/
+
+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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_224
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_224
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_224
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = io.read(s, buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_224 will read the file provided by the given handle
+// and compute a hash
+hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_256
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_256
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_256
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = io.read(s, buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_256 will read the file provided by the given handle
+// and compute a hash
+hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_384
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_384
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_384
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = io.read(s, buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_384 will read the file provided by the given handle
+// and compute a hash
+hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_512
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_512
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    _sha3.update(&ctx, data)
+    _sha3.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: _sha3.Sha3_Context
+    ctx.mdlen = DIGEST_SIZE_512
+    ctx.is_keccak = true
+    _sha3.init(&ctx)
+    buf := make([]byte, 512)
+    defer delete(buf)
+    read := 1
+    for read > 0 {
+        read, _ = io.read(s, buf)
+        if read > 0 {
+            _sha3.update(&ctx, buf[:read])
+        } 
+    }
+    _sha3.final(&ctx, hash[:])
+    return hash, true
+}
+
+// hash_file_512 will read the file provided by the given handle
+// and compute a hash
+hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([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
+*/
+
+Keccak_Context :: _sha3.Sha3_Context
+
+init :: proc(ctx: ^_sha3.Sha3_Context) {
+    ctx.is_keccak = true
+    _sha3.init(ctx)
+}
+
+update :: proc "contextless" (ctx: ^_sha3.Sha3_Context, data: []byte) {
+    _sha3.update(ctx, data)
+}
+
+final :: proc "contextless" (ctx: ^_sha3.Sha3_Context, hash: []byte) {
+    _sha3.final(ctx, hash)
+}
@@ -0,0 +1,182 @@
+package md2
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the MD2 hashing algorithm, as defined in RFC 1319 <https://datatracker.ietf.org/doc/html/rfc1319>
+*/
+
+import "core:os"
+import "core:io"
+
+/*
+    High level API
+*/
+
+DIGEST_SIZE :: 16
+
+// 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: Md2_Context
+    // init(&ctx) No-op
+    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: Md2_Context
+    // init(&ctx) No-op
+    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: Md2_Context
+	// init(&ctx) No-op
+	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
+*/
+
+@(warning="Init is a no-op for MD2")
+init :: proc(ctx: ^Md2_Context) {
+	// No action needed here
+}
+
+update :: proc(ctx: ^Md2_Context, data: []byte) {
+	for i := 0; i < len(data); i += 1 {
+		ctx.data[ctx.datalen] = data[i]
+		ctx.datalen += 1
+		if (ctx.datalen == DIGEST_SIZE) {
+			transform(ctx, ctx.data[:])
+			ctx.datalen = 0
+		}
+	}
+}
+
+final :: proc(ctx: ^Md2_Context, hash: []byte) {
+	to_pad := byte(DIGEST_SIZE - ctx.datalen)
+    for ctx.datalen < DIGEST_SIZE {
+        ctx.data[ctx.datalen] = to_pad
+		ctx.datalen += 1
+    }
+	transform(ctx, ctx.data[:])
+	transform(ctx, ctx.checksum[:])
+    for i := 0; i < DIGEST_SIZE; i += 1 {
+        hash[i] = ctx.state[i]
+    }
+}
+
+/*
+    MD2 implementation
+*/
+
+Md2_Context :: struct {
+    data:     [DIGEST_SIZE]byte,
+    state:    [DIGEST_SIZE * 3]byte,
+    checksum: [DIGEST_SIZE]byte,
+    datalen:  int,
+}
+
+PI_TABLE := [?]byte {
+	41,  46,  67,  201, 162, 216, 124, 1,   61,  54,  84,  161, 236, 240, 6,
+	19,  98,  167, 5,   243, 192, 199, 115, 140, 152, 147, 43,  217, 188, 76,
+	130, 202, 30,  155, 87,  60,  253, 212, 224, 22,  103, 66,  111, 24,  138, 
+	23,  229, 18,  190, 78,  196, 214, 218, 158, 222, 73,  160, 251, 245, 142,
+	187, 47,  238, 122, 169, 104, 121, 145, 21,  178, 7,   63,  148, 194, 16,
+	137, 11,  34,  95,  33,  128, 127, 93,  154, 90,  144, 50,  39,  53,  62, 
+	204, 231, 191, 247, 151, 3,   255, 25,  48,  179, 72,  165, 181, 209, 215,
+	94,  146, 42,  172, 86,  170, 198, 79,  184, 56,  210, 150, 164, 125, 182,
+	118, 252, 107, 226, 156, 116, 4,   241, 69,  157, 112, 89,  100, 113, 135,
+	32,  134, 91,  207, 101, 230, 45,  168, 2,   27,  96,  37,  173, 174, 176,
+	185, 246, 28,  70,  97,  105, 52,  64,  126, 15,  85,  71,  163, 35,  221,
+	81,  175, 58,  195, 92,  249, 206, 186, 197, 234, 38,  44,  83,  13,  110,
+	133, 40,  132, 9,   211, 223, 205, 244, 65,  129, 77,  82,  106, 220, 55,
+	200, 108, 193, 171, 250, 36,  225, 123, 8,   12,  189, 177, 74,  120, 136,
+	149, 139, 227, 99,  232, 109, 233, 203, 213, 254, 59,  0,   29,  57,  242,
+	239, 183, 14,  102, 88,  208, 228, 166, 119, 114, 248, 235, 117, 75,  10,
+	49,  68,  80,  180, 143, 237, 31,  26,  219, 153, 141, 51,  159, 17,  131,
+	20,
+}
+
+transform :: proc(ctx: ^Md2_Context, data: []byte) {
+    j,k,t: byte
+	for j = 0; j < DIGEST_SIZE; j += 1 {
+		ctx.state[j + DIGEST_SIZE] = data[j]
+		ctx.state[j + DIGEST_SIZE * 2] = (ctx.state[j + DIGEST_SIZE] ~ ctx.state[j])
+	}
+	t = 0
+	for j = 0; j < DIGEST_SIZE + 2; j += 1 {
+		for k = 0; k < DIGEST_SIZE * 3; k += 1 {
+			ctx.state[k] ~= PI_TABLE[t]
+			t = ctx.state[k]
+		}
+		t = (t + j) & 0xff
+	}
+	t = ctx.checksum[DIGEST_SIZE - 1]
+	for j = 0; j < DIGEST_SIZE; j += 1 {
+		ctx.checksum[j] ~= PI_TABLE[data[j] ~ t]
+		t = ctx.checksum[j]
+	}
+}
@@ -0,0 +1,263 @@
+package md4
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+        Jeroen van Rijn: Context design to be able to change from Odin implementation to bindings.
+
+    Implementation of the MD4 hashing algorithm, as defined in RFC 1320 <https://datatracker.ietf.org/doc/html/rfc1320>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+DIGEST_SIZE :: 16
+
+// 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: Md4_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: Md4_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: Md4_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(ctx: ^Md4_Context) {
+    ctx.state[0] = 0x67452301
+    ctx.state[1] = 0xefcdab89
+    ctx.state[2] = 0x98badcfe
+    ctx.state[3] = 0x10325476
+}
+
+update :: proc(ctx: ^Md4_Context, data: []byte) {
+    for i := 0; i < len(data); i += 1 {
+        ctx.data[ctx.datalen] = data[i]
+        ctx.datalen += 1
+        if(ctx.datalen == BLOCK_SIZE) {
+            transform(ctx, ctx.data[:])
+            ctx.bitlen += 512
+            ctx.datalen = 0
+        }
+    }
+}
+
+final :: proc(ctx: ^Md4_Context, hash: []byte) {
+    i := ctx.datalen
+    if ctx.datalen < 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < 56 {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+    } else if ctx.datalen >= 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < BLOCK_SIZE {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+        transform(ctx, ctx.data[:])
+        mem.set(&ctx.data, 0, 56)
+    }
+
+    ctx.bitlen  += u64(ctx.datalen * 8)
+    ctx.data[56] = byte(ctx.bitlen)
+    ctx.data[57] = byte(ctx.bitlen >> 8)
+    ctx.data[58] = byte(ctx.bitlen >> 16)
+    ctx.data[59] = byte(ctx.bitlen >> 24)
+    ctx.data[60] = byte(ctx.bitlen >> 32)
+    ctx.data[61] = byte(ctx.bitlen >> 40)
+    ctx.data[62] = byte(ctx.bitlen >> 48)
+    ctx.data[63] = byte(ctx.bitlen >> 56)
+    transform(ctx, ctx.data[:])
+
+    for i = 0; i < 4; i += 1 {
+        hash[i]      = byte(ctx.state[0] >> (i * 8)) & 0x000000ff
+        hash[i + 4]  = byte(ctx.state[1] >> (i * 8)) & 0x000000ff
+        hash[i + 8]  = byte(ctx.state[2] >> (i * 8)) & 0x000000ff
+        hash[i + 12] = byte(ctx.state[3] >> (i * 8)) & 0x000000ff
+    }
+}
+
+/*
+    MD4 implementation
+*/
+
+BLOCK_SIZE  :: 64
+
+Md4_Context :: struct {
+    data:    [64]byte,
+    state:   [4]u32,
+    bitlen:  u64,
+    datalen: u32,
+}
+
+/*
+    @note(zh): F, G and H, as mentioned in the RFC, have been inlined into FF, GG 
+    and HH respectively, instead of declaring them separately.
+*/
+
+FF :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + ((b & c) | (~b & d)) + x, s)
+}
+
+GG :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + ((b & c) | (b & d) | (c & d)) + x + 0x5a827999, s)
+}
+
+HH :: #force_inline proc "contextless"(a, b, c, d, x: u32, s : int) -> u32 {
+    return util.ROTL32(a + (b ~ c ~ d) + x + 0x6ed9eba1, s)
+}
+
+transform :: proc(ctx: ^Md4_Context, data: []byte) {
+    a, b, c, d, i, j: u32
+    m: [DIGEST_SIZE]u32
+
+    for i, j = 0, 0; i < DIGEST_SIZE; i += 1 {
+        m[i] = u32(data[j]) | (u32(data[j + 1]) << 8) | (u32(data[j + 2]) << 16) | (u32(data[j + 3]) << 24)
+        j += 4
+    }
+
+    a = ctx.state[0]
+    b = ctx.state[1]
+    c = ctx.state[2]
+    d = ctx.state[3]
+
+    a = FF(a, b, c, d, m[0],  3)
+    d = FF(d, a, b, c, m[1],  7)
+    c = FF(c, d, a, b, m[2],  11)
+    b = FF(b, c, d, a, m[3],  19)
+    a = FF(a, b, c, d, m[4],  3)
+    d = FF(d, a, b, c, m[5],  7)
+    c = FF(c, d, a, b, m[6],  11)
+    b = FF(b, c, d, a, m[7],  19)
+    a = FF(a, b, c, d, m[8],  3)
+    d = FF(d, a, b, c, m[9],  7)
+    c = FF(c, d, a, b, m[10], 11)
+    b = FF(b, c, d, a, m[11], 19)
+    a = FF(a, b, c, d, m[12], 3)
+    d = FF(d, a, b, c, m[13], 7)
+    c = FF(c, d, a, b, m[14], 11)
+    b = FF(b, c, d, a, m[15], 19)
+
+    a = GG(a, b, c, d, m[0],  3)
+    d = GG(d, a, b, c, m[4],  5)
+    c = GG(c, d, a, b, m[8],  9)
+    b = GG(b, c, d, a, m[12], 13)
+    a = GG(a, b, c, d, m[1],  3)
+    d = GG(d, a, b, c, m[5],  5)
+    c = GG(c, d, a, b, m[9],  9)
+    b = GG(b, c, d, a, m[13], 13)
+    a = GG(a, b, c, d, m[2],  3)
+    d = GG(d, a, b, c, m[6],  5)
+    c = GG(c, d, a, b, m[10], 9)
+    b = GG(b, c, d, a, m[14], 13)
+    a = GG(a, b, c, d, m[3],  3)
+    d = GG(d, a, b, c, m[7],  5)
+    c = GG(c, d, a, b, m[11], 9)
+    b = GG(b, c, d, a, m[15], 13)
+
+    a = HH(a, b, c, d, m[0],  3)
+    d = HH(d, a, b, c, m[8],  9)
+    c = HH(c, d, a, b, m[4],  11)
+    b = HH(b, c, d, a, m[12], 15)
+    a = HH(a, b, c, d, m[2],  3)
+    d = HH(d, a, b, c, m[10], 9)
+    c = HH(c, d, a, b, m[6],  11)
+    b = HH(b, c, d, a, m[14], 15)
+    a = HH(a, b, c, d, m[1],  3)
+    d = HH(d, a, b, c, m[9],  9)
+    c = HH(c, d, a, b, m[5],  11)
+    b = HH(b, c, d, a, m[13], 15)
+    a = HH(a, b, c, d, m[3],  3)
+    d = HH(d, a, b, c, m[11], 9)
+    c = HH(c, d, a, b, m[7],  11)
+    b = HH(b, c, d, a, m[15], 15)
+
+    ctx.state[0] += a
+    ctx.state[1] += b
+    ctx.state[2] += c
+    ctx.state[3] += d
+}
@@ -0,0 +1,285 @@
+package md5
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation of the MD5 hashing algorithm, as defined in RFC 1321 <https://datatracker.ietf.org/doc/html/rfc1321>
+*/
+
+import "core:mem"
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+DIGEST_SIZE :: 16
+
+// 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: Md5_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: Md5_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: Md5_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(ctx: ^Md5_Context) {
+    ctx.state[0] = 0x67452301
+    ctx.state[1] = 0xefcdab89
+    ctx.state[2] = 0x98badcfe
+    ctx.state[3] = 0x10325476
+}
+
+update :: proc(ctx: ^Md5_Context, data: []byte) {
+    for i := 0; i < len(data); i += 1 {
+        ctx.data[ctx.datalen] = data[i]
+        ctx.datalen += 1
+        if(ctx.datalen == BLOCK_SIZE) {
+            transform(ctx, ctx.data[:])
+            ctx.bitlen += 512
+            ctx.datalen = 0
+        }
+    }
+}
+
+final :: proc(ctx: ^Md5_Context, hash: []byte){
+    i : u32
+    i = ctx.datalen
+
+    if ctx.datalen < 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < 56 {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+    } else if ctx.datalen >= 56 {
+        ctx.data[i] = 0x80
+        i += 1
+        for i < BLOCK_SIZE {
+            ctx.data[i] = 0x00
+            i += 1
+        }
+        transform(ctx, ctx.data[:])
+        mem.set(&ctx.data, 0, 56)
+    }
+
+    ctx.bitlen  += u64(ctx.datalen * 8)
+    ctx.data[56] = byte(ctx.bitlen)
+    ctx.data[57] = byte(ctx.bitlen >> 8)
+    ctx.data[58] = byte(ctx.bitlen >> 16)
+    ctx.data[59] = byte(ctx.bitlen >> 24)
+    ctx.data[60] = byte(ctx.bitlen >> 32)
+    ctx.data[61] = byte(ctx.bitlen >> 40)
+    ctx.data[62] = byte(ctx.bitlen >> 48)
+    ctx.data[63] = byte(ctx.bitlen >> 56)
+    transform(ctx, ctx.data[:])
+
+    for i = 0; i < 4; i += 1 {
+        hash[i]      = byte(ctx.state[0] >> (i * 8)) & 0x000000ff
+        hash[i + 4]  = byte(ctx.state[1] >> (i * 8)) & 0x000000ff
+        hash[i + 8]  = byte(ctx.state[2] >> (i * 8)) & 0x000000ff
+        hash[i + 12] = byte(ctx.state[3] >> (i * 8)) & 0x000000ff
+    }
+}
+
+/*
+    MD4 implementation
+*/
+
+BLOCK_SIZE  :: 64
+
+Md5_Context :: struct {
+    data:    [BLOCK_SIZE]byte,
+    state:   [4]u32,
+    bitlen:  u64,
+    datalen: u32,
+}
+
+/*
+    @note(zh): F, G, H and I, as mentioned in the RFC, have been inlined into FF, GG, HH 
+    and II respectively, instead of declaring them separately.
+*/
+
+FF :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + ((b & c) | (~b & d)) + m + t, s)
+}
+
+GG :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + ((b & d) | (c & ~d)) + m + t, s)
+}
+
+HH :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + (b ~ c ~ d) + m + t, s)
+}
+
+II :: #force_inline proc "contextless" (a, b, c, d, m: u32, s: int, t: u32) -> u32 {
+    return b + util.ROTL32(a + (c ~ (b | ~d)) + m + t, s)
+}
+
+transform :: proc(ctx: ^Md5_Context, data: []byte) {
+    i, j: u32
+    m: [DIGEST_SIZE]u32
+
+    for i, j = 0, 0; i < DIGEST_SIZE; i+=1 {
+        m[i] = u32(data[j]) + u32(data[j + 1]) << 8 + u32(data[j + 2]) << 16 + u32(data[j + 3]) << 24
+        j += 4
+    }
+
+    a := ctx.state[0]
+    b := ctx.state[1]
+    c := ctx.state[2]
+    d := ctx.state[3]
+
+    a = FF(a, b, c, d, m[0],   7, 0xd76aa478)
+    d = FF(d, a, b, c, m[1],  12, 0xe8c7b756)
+    c = FF(c, d, a, b, m[2],  17, 0x242070db)
+    b = FF(b, c, d, a, m[3],  22, 0xc1bdceee)
+    a = FF(a, b, c, d, m[4],   7, 0xf57c0faf)
+    d = FF(d, a, b, c, m[5],  12, 0x4787c62a)
+    c = FF(c, d, a, b, m[6],  17, 0xa8304613)
+    b = FF(b, c, d, a, m[7],  22, 0xfd469501)
+    a = FF(a, b, c, d, m[8],   7, 0x698098d8)
+    d = FF(d, a, b, c, m[9],  12, 0x8b44f7af)
+    c = FF(c, d, a, b, m[10], 17, 0xffff5bb1)
+    b = FF(b, c, d, a, m[11], 22, 0x895cd7be)
+    a = FF(a, b, c, d, m[12],  7, 0x6b901122)
+    d = FF(d, a, b, c, m[13], 12, 0xfd987193)
+    c = FF(c, d, a, b, m[14], 17, 0xa679438e)
+    b = FF(b, c, d, a, m[15], 22, 0x49b40821)
+
+    a = GG(a, b, c, d, m[1],   5, 0xf61e2562)
+    d = GG(d, a, b, c, m[6],   9, 0xc040b340)
+    c = GG(c, d, a, b, m[11], 14, 0x265e5a51)
+    b = GG(b, c, d, a, m[0],  20, 0xe9b6c7aa)
+    a = GG(a, b, c, d, m[5],   5, 0xd62f105d)
+    d = GG(d, a, b, c, m[10],  9, 0x02441453)
+    c = GG(c, d, a, b, m[15], 14, 0xd8a1e681)
+    b = GG(b, c, d, a, m[4],  20, 0xe7d3fbc8)
+    a = GG(a, b, c, d, m[9],   5, 0x21e1cde6)
+    d = GG(d, a, b, c, m[14],  9, 0xc33707d6)
+    c = GG(c, d, a, b, m[3],  14, 0xf4d50d87)
+    b = GG(b, c, d, a, m[8],  20, 0x455a14ed)
+    a = GG(a, b, c, d, m[13],  5, 0xa9e3e905)
+    d = GG(d, a, b, c, m[2],   9, 0xfcefa3f8)
+    c = GG(c, d, a, b, m[7],  14, 0x676f02d9)
+    b = GG(b, c, d, a, m[12], 20, 0x8d2a4c8a)
+
+    a = HH(a, b, c, d, m[5],   4, 0xfffa3942)
+    d = HH(d, a, b, c, m[8],  11, 0x8771f681)
+    c = HH(c, d, a, b, m[11], 16, 0x6d9d6122)
+    b = HH(b, c, d, a, m[14], 23, 0xfde5380c)
+    a = HH(a, b, c, d, m[1],   4, 0xa4beea44)
+    d = HH(d, a, b, c, m[4],  11, 0x4bdecfa9)
+    c = HH(c, d, a, b, m[7],  16, 0xf6bb4b60)
+    b = HH(b, c, d, a, m[10], 23, 0xbebfbc70)
+    a = HH(a, b, c, d, m[13],  4, 0x289b7ec6)
+    d = HH(d, a, b, c, m[0],  11, 0xeaa127fa)
+    c = HH(c, d, a, b, m[3],  16, 0xd4ef3085)
+    b = HH(b, c, d, a, m[6],  23, 0x04881d05)
+    a = HH(a, b, c, d, m[9],   4, 0xd9d4d039)
+    d = HH(d, a, b, c, m[12], 11, 0xe6db99e5)
+    c = HH(c, d, a, b, m[15], 16, 0x1fa27cf8)
+    b = HH(b, c, d, a, m[2],  23, 0xc4ac5665)
+
+    a = II(a, b, c, d, m[0],   6, 0xf4292244)
+    d = II(d, a, b, c, m[7],  10, 0x432aff97)
+    c = II(c, d, a, b, m[14], 15, 0xab9423a7)
+    b = II(b, c, d, a, m[5],  21, 0xfc93a039)
+    a = II(a, b, c, d, m[12],  6, 0x655b59c3)
+    d = II(d, a, b, c, m[3],  10, 0x8f0ccc92)
+    c = II(c, d, a, b, m[10], 15, 0xffeff47d)
+    b = II(b, c, d, a, m[1],  21, 0x85845dd1)
+    a = II(a, b, c, d, m[8],   6, 0x6fa87e4f)
+    d = II(d, a, b, c, m[15], 10, 0xfe2ce6e0)
+    c = II(c, d, a, b, m[6],  15, 0xa3014314)
+    b = II(b, c, d, a, m[13], 21, 0x4e0811a1)
+    a = II(a, b, c, d, m[4],   6, 0xf7537e82)
+    d = II(d, a, b, c, m[11], 10, 0xbd3af235)
+    c = II(c, d, a, b, m[2],  15, 0x2ad7d2bb)
+    b = II(b, c, d, a, m[9],  21, 0xeb86d391)
+
+    ctx.state[0] += a
+    ctx.state[1] += b
+    ctx.state[2] += c
+    ctx.state[3] += d
+}
@@ -0,0 +1,165 @@
+package poly1305
+
+import "core:crypto"
+import "core:crypto/util"
+import field "core:crypto/_fiat/field_poly1305"
+import "core:mem"
+
+KEY_SIZE :: 32
+TAG_SIZE :: 16
+
+@(private)
+_BLOCK_SIZE :: 16
+
+sum :: proc (dst, msg, key: []byte) {
+	ctx: Context = ---
+
+	init(&ctx, key)
+	update(&ctx, msg)
+	final(&ctx, dst)
+}
+
+verify :: proc (tag, msg, key: []byte) -> bool {
+	ctx: Context = ---
+	derived_tag: [16]byte = ---
+
+	if len(tag) != TAG_SIZE {
+		panic("crypto/poly1305: invalid tag size")
+	}
+
+	init(&ctx, key)
+	update(&ctx, msg)
+	final(&ctx, derived_tag[:])
+
+	return crypto.compare_constant_time(derived_tag[:], tag) == 1
+}
+
+Context :: struct {
+	_r: field.Tight_Field_Element,
+	_a: field.Tight_Field_Element,
+	_s: field.Tight_Field_Element,
+
+	_buffer: [_BLOCK_SIZE]byte,
+	_leftover: int,
+
+	_is_initialized: bool,
+}
+
+init :: proc (ctx: ^Context, key: []byte) {
+	if len(key) != KEY_SIZE {
+		panic("crypto/poly1305: invalid key size")
+	}
+
+	// r = le_bytes_to_num(key[0..15])
+	// r = clamp(r) (r &= 0xffffffc0ffffffc0ffffffc0fffffff)
+	tmp_lo := util.U64_LE(key[0:8]) & 0x0ffffffc0fffffff
+	tmp_hi := util.U64_LE(key[8:16]) & 0xffffffc0ffffffc
+	field.fe_from_u64s(&ctx._r, tmp_lo, tmp_hi)
+
+	// s = le_bytes_to_num(key[16..31])
+	field.fe_from_bytes(&ctx._s, key[16:32], 0)
+
+	// a = 0
+	field.fe_zero(&ctx._a)
+
+	// No leftover in buffer
+	ctx._leftover = 0
+
+	ctx._is_initialized = true
+}
+
+update :: proc (ctx: ^Context, data: []byte) {
+	assert(ctx._is_initialized)
+
+	msg := data
+	msg_len := len(data)
+
+	// Handle leftover
+	if ctx._leftover > 0 {
+		want := min(_BLOCK_SIZE - ctx._leftover, msg_len)
+		copy_slice(ctx._buffer[ctx._leftover:], msg[:want])
+		msg_len = msg_len - want
+		msg = msg[want:]
+		ctx._leftover = ctx._leftover + want
+		if ctx._leftover < _BLOCK_SIZE {
+			return
+		}
+		_blocks(ctx, ctx._buffer[:])
+		ctx._leftover = 0
+	}
+
+	// Process full blocks
+	if msg_len >= _BLOCK_SIZE {
+		want := msg_len & (~int(_BLOCK_SIZE - 1))
+		_blocks(ctx, msg[:want])
+		msg = msg[want:]
+		msg_len = msg_len - want
+	}
+
+	// Store leftover
+	if msg_len > 0 {
+		// TODO: While -donna does it this way, I'm fairly sure that
+		// `ctx._leftover == 0` is an invariant at this point.
+		copy(ctx._buffer[ctx._leftover:], msg)
+		ctx._leftover = ctx._leftover + msg_len
+	}
+}
+
+final :: proc (ctx: ^Context, dst: []byte) {
+	assert(ctx._is_initialized)
+
+	if len(dst) != TAG_SIZE {
+		panic("poly1305: invalid destination tag size")
+	}
+
+	// Process remaining block
+	if ctx._leftover > 0 {
+		ctx._buffer[ctx._leftover] = 1
+		for i := ctx._leftover + 1; i < _BLOCK_SIZE; i = i + 1 {
+			ctx._buffer[i] = 0
+		}
+		_blocks(ctx, ctx._buffer[:], true)
+	}
+
+	// a += s
+	field.fe_add(field.fe_relax_cast(&ctx._a), &ctx._a, &ctx._s) // _a unreduced
+	field.fe_carry(&ctx._a, field.fe_relax_cast(&ctx._a)) // _a reduced
+
+	// return num_to_16_le_bytes(a)
+	tmp: [32]byte = ---
+	field.fe_to_bytes(&tmp, &ctx._a)
+	copy_slice(dst, tmp[0:16])
+
+	reset(ctx)
+}
+
+reset :: proc (ctx: ^Context) {
+	mem.zero_explicit(&ctx._r, size_of(ctx._r))
+	mem.zero_explicit(&ctx._a, size_of(ctx._a))
+	mem.zero_explicit(&ctx._s, size_of(ctx._s))
+	mem.zero_explicit(&ctx._buffer, size_of(ctx._buffer))
+
+	ctx._is_initialized = false
+}
+
+@(private)
+_blocks :: proc (ctx: ^Context, msg: []byte, final := false) {
+	n: field.Tight_Field_Element = ---
+	final_byte := byte(!final)
+
+	data := msg
+	data_len := len(data)
+	for data_len >= _BLOCK_SIZE {
+		// n = le_bytes_to_num(msg[((i-1)*16)..*i*16] | [0x01])
+		field.fe_from_bytes(&n, data[:_BLOCK_SIZE], final_byte, false)
+
+		// a += n
+		field.fe_add(field.fe_relax_cast(&ctx._a), &ctx._a, &n) // _a unreduced
+
+		// a = (r * a) % p
+		field.fe_carry_mul(&ctx._a, field.fe_relax_cast(&ctx._a), field.fe_relax_cast(&ctx._r)) // _a reduced
+
+		data = data[_BLOCK_SIZE:]
+		data_len = data_len - _BLOCK_SIZE
+	}
+}
@@ -0,0 +1,7 @@
+package crypto
+
+when ODIN_OS != .Linux && ODIN_OS != .OpenBSD && ODIN_OS != .Windows {
+	_rand_bytes :: proc (dst: []byte) {
+		unimplemented("crypto: rand_bytes not supported on this OS")
+	}
+}
@@ -0,0 +1,37 @@
+package crypto
+
+import "core:fmt"
+import "core:os"
+import "core:sys/unix"
+
+_MAX_PER_CALL_BYTES :: 33554431 // 2^25 - 1
+
+_rand_bytes :: proc (dst: []byte) {
+	dst := dst
+	l := len(dst)
+
+	for l > 0 {
+		to_read := min(l, _MAX_PER_CALL_BYTES)
+		ret := unix.sys_getrandom(raw_data(dst), uint(to_read), 0)
+		if ret < 0 {
+			switch os.Errno(-ret) {
+			case os.EINTR:
+				// Call interupted by a signal handler, just retry the
+				// request.
+				continue
+			case os.ENOSYS:
+				// The kernel is apparently prehistoric (< 3.17 circa 2014)
+				// and does not support getrandom.
+				panic("crypto: getrandom not available in kernel")
+			case:
+				// All other failures are things that should NEVER happen
+				// unless the kernel interface changes (ie: the Linux
+				// developers break userland).
+				panic(fmt.tprintf("crypto: getrandom failed: %d", ret))
+			}
+		}
+
+		l -= ret
+		dst = dst[ret:]
+	}
+}
@@ -0,0 +1,12 @@
+package crypto
+
+import "core:c"
+
+foreign import libc "system:c"
+foreign libc {
+	arc4random_buf :: proc "c" (buf: rawptr, nbytes: c.size_t) ---
+}
+
+_rand_bytes :: proc (dst: []byte) {
+	arc4random_buf(raw_data(dst), len(dst))
+}
@@ -0,0 +1,23 @@
+package crypto
+
+import win32 "core:sys/windows"
+import "core:os"
+import "core:fmt"
+
+_rand_bytes :: proc(dst: []byte) {
+	ret := (os.Errno)(win32.BCryptGenRandom(nil, raw_data(dst), u32(len(dst)), win32.BCRYPT_USE_SYSTEM_PREFERRED_RNG))
+	if ret != os.ERROR_NONE {
+		switch ret {
+			case os.ERROR_INVALID_HANDLE:
+				// The handle to the first parameter is invalid.
+				// This should not happen here, since we explicitly pass nil to it
+				panic("crypto: BCryptGenRandom Invalid handle for hAlgorithm")
+			case os.ERROR_INVALID_PARAMETER:
+				// One of the parameters was invalid
+				panic("crypto: BCryptGenRandom Invalid parameter")
+			case:
+				// Unknown error
+				panic(fmt.tprintf("crypto: BCryptGenRandom failed: %d\n", ret))
+		}
+	}
+}
@@ -0,0 +1,919 @@
+package ripemd
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+
+    Implementation for the RIPEMD hashing algorithm as defined in <https://homes.esat.kuleuven.be/~bosselae/ripemd160.html>
+*/
+
+import "core:os"
+import "core:io"
+
+import "../util"
+
+/*
+    High level API
+*/
+
+DIGEST_SIZE_128 :: 16
+DIGEST_SIZE_160 :: 20
+DIGEST_SIZE_256 :: 32
+DIGEST_SIZE_320 :: 40
+
+// hash_string_128 will hash the given input and return the
+// computed hash
+hash_string_128 :: proc(data: string) -> [DIGEST_SIZE_128]byte {
+    return hash_bytes_128(transmute([]byte)(data))
+}
+
+// hash_bytes_128 will hash the given input and return the
+// computed hash
+hash_bytes_128 :: proc(data: []byte) -> [DIGEST_SIZE_128]byte {
+    hash: [DIGEST_SIZE_128]byte
+    ctx: Ripemd128_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_string_to_buffer_128 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_128 :: proc(data: string, hash: []byte) {
+    hash_bytes_to_buffer_128(transmute([]byte)(data), hash)
+}
+
+// hash_bytes_to_buffer_128 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_128 :: proc(data, hash: []byte) {
+    assert(len(hash) >= DIGEST_SIZE_128, "Size of destination buffer is smaller than the digest size")
+    ctx: Ripemd128_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash)
+}
+
+// hash_stream_128 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_128 :: proc(s: io.Stream) -> ([DIGEST_SIZE_128]byte, bool) {
+    hash: [DIGEST_SIZE_128]byte
+    ctx: Ripemd128_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_128 will read the file provided by the given handle
+// and compute a hash
+hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_128]byte, bool) {
+    if !load_at_once {
+        return hash_stream_128(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_128(buf[:]), ok
+        }
+    }
+    return [DIGEST_SIZE_128]byte{}, false
+}
+
+hash_128 :: proc {
+    hash_stream_128,
+    hash_file_128,
+    hash_bytes_128,
+    hash_string_128,
+    hash_bytes_to_buffer_128,
+    hash_string_to_buffer_128,
+}
+
+// hash_string_160 will hash the given input and return the
+// computed hash
+hash_string_160 :: proc(data: string) -> [DIGEST_SIZE_160]byte {
+    return hash_bytes_160(transmute([]byte)(data))
+}
+
+// hash_bytes_160 will hash the given input and return the
+// computed hash
+hash_bytes_160 :: proc(data: []byte) -> [DIGEST_SIZE_160]byte {
+    hash: [DIGEST_SIZE_160]byte
+    ctx: Ripemd160_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_string_to_buffer_160 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_160 :: proc(data: string, hash: []byte) {
+    hash_bytes_to_buffer_160(transmute([]byte)(data), hash)
+}
+
+// hash_bytes_to_buffer_160 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_160 :: proc(data, hash: []byte) {
+    assert(len(hash) >= DIGEST_SIZE_160, "Size of destination buffer is smaller than the digest size")
+    ctx: Ripemd160_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash)
+}
+
+// hash_stream_160 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_160 :: proc(s: io.Stream) -> ([DIGEST_SIZE_160]byte, bool) {
+    hash: [DIGEST_SIZE_160]byte
+    ctx: Ripemd160_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_160 will read the file provided by the given handle
+// and compute a hash
+hash_file_160 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_160]byte, bool) {
+    if !load_at_once {
+        return hash_stream_160(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_160(buf[:]), ok
+        }
+    }
+    return [DIGEST_SIZE_160]byte{}, false
+}
+
+hash_160 :: proc {
+    hash_stream_160,
+    hash_file_160,
+    hash_bytes_160,
+    hash_string_160,
+    hash_bytes_to_buffer_160,
+    hash_string_to_buffer_160,
+}
+
+// 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: Ripemd256_Context
+    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: Ripemd256_Context
+    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: Ripemd256_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_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_320 will hash the given input and return the
+// computed hash
+hash_string_320 :: proc(data: string) -> [DIGEST_SIZE_320]byte {
+    return hash_bytes_320(transmute([]byte)(data))
+}
+
+// hash_bytes_320 will hash the given input and return the
+// computed hash
+hash_bytes_320 :: proc(data: []byte) -> [DIGEST_SIZE_320]byte {
+    hash: [DIGEST_SIZE_320]byte
+    ctx: Ripemd320_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash[:])
+    return hash
+}
+
+// hash_string_to_buffer_320 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_320 :: proc(data: string, hash: []byte) {
+    hash_bytes_to_buffer_320(transmute([]byte)(data), hash)
+}
+
+// hash_bytes_to_buffer_320 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_320 :: proc(data, hash: []byte) {
+    assert(len(hash) >= DIGEST_SIZE_320, "Size of destination buffer is smaller than the digest size")
+    ctx: Ripemd320_Context
+    init(&ctx)
+    update(&ctx, data)
+    final(&ctx, hash)
+}
+
+// hash_stream_320 will read the stream in chunks and compute a
+// hash from its contents
+hash_stream_320 :: proc(s: io.Stream) -> ([DIGEST_SIZE_320]byte, bool) {
+    hash: [DIGEST_SIZE_320]byte
+    ctx: Ripemd320_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_320 will read the file provided by the given handle
+// and compute a hash
+hash_file_320 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_320]byte, bool) {
+    if !load_at_once {
+        return hash_stream_320(os.stream_from_handle(hd))
+    } else {
+        if buf, ok := os.read_entire_file(hd); ok {
+            return hash_bytes_320(buf[:]), ok
+        }
+    }
+    return [DIGEST_SIZE_320]byte{}, false
+}
+
+hash_320 :: proc {
+    hash_stream_320,
+    hash_file_320,
+    hash_bytes_320,
+    hash_string_320,
+    hash_bytes_to_buffer_320,
+    hash_string_to_buffer_320,
+}
+
+/*
+    Low level API
+*/
+
+init :: proc(ctx: ^$T) {
+    when T == Ripemd128_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3] = S0, S1, S2, S3
+    } else when T == Ripemd160_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4] = S0, S1, S2, S3, S4
+    } else when T == Ripemd256_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3] = S0, S1, S2, S3
+        ctx.s[4], ctx.s[5], ctx.s[6], ctx.s[7] = S5, S6, S7, S8
+    } else when T == Ripemd320_Context {
+        ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4] = S0, S1, S2, S3, S4
+        ctx.s[5], ctx.s[6], ctx.s[7], ctx.s[8], ctx.s[9] = S5, S6, S7, S8, S9
+    }
+}
+
+update :: proc(ctx: ^$T, data: []byte) {
+    ctx.tc += u64(len(data))
+    data := data
+    if ctx.nx > 0 {
+        n := len(data)
+
+        when T == Ripemd128_Context {
+            if n > RIPEMD_128_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_128_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd160_Context {
+            if n > RIPEMD_160_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_160_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd256_Context{
+            if n > RIPEMD_256_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_256_BLOCK_SIZE - ctx.nx
+            }
+        } else when T == Ripemd320_Context{
+            if n > RIPEMD_320_BLOCK_SIZE - ctx.nx {
+                n = RIPEMD_320_BLOCK_SIZE - ctx.nx
+            }
+        }
+
+        for i := 0; i < n; i += 1 {
+            ctx.x[ctx.nx + i] = data[i]
+        }
+
+        ctx.nx += n
+        when T == Ripemd128_Context {
+            if ctx.nx == RIPEMD_128_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd160_Context {
+            if ctx.nx == RIPEMD_160_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd256_Context{
+            if ctx.nx == RIPEMD_256_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        } else when T == Ripemd320_Context{
+            if ctx.nx == RIPEMD_320_BLOCK_SIZE {
+                block(ctx, ctx.x[0:])
+                ctx.nx = 0
+            }
+        }
+        data = data[n:]
+    }
+    n := block(ctx, data)
+    data = data[n:]
+    if len(data) > 0 {
+        ctx.nx = copy(ctx.x[:], data)
+    }
+}
+
+final :: proc(ctx: ^$T, hash: []byte) {
+    d := ctx
+    tc := d.tc
+    tmp: [64]byte
+    tmp[0] = 0x80
+
+    if tc % 64 < 56 {
+        update(d, tmp[0:56 - tc % 64])
+    } else {
+        update(d, tmp[0:64 + 56 - tc % 64])
+    }
+
+    tc <<= 3
+    for i : u32 = 0; i < 8; i += 1 {
+        tmp[i] = byte(tc >> (8 * i))
+    }
+
+    update(d, tmp[0:8])
+
+    when T == Ripemd128_Context {
+        size :: RIPEMD_128_SIZE
+    } else when T == Ripemd160_Context {
+        size :: RIPEMD_160_SIZE
+    } else when T == Ripemd256_Context{
+        size :: RIPEMD_256_SIZE
+    } else when T == Ripemd320_Context{
+        size :: RIPEMD_320_SIZE
+    }
+
+    digest: [size]byte
+    for s, i in d.s {
+        digest[i * 4]     = byte(s)
+        digest[i * 4 + 1] = byte(s >> 8)
+        digest[i * 4 + 2] = byte(s >> 16)
+        digest[i * 4 + 3] = byte(s >> 24)
+    }
+    copy(hash[:], digest[:])
+}
+
+
+/*
+    RIPEMD implementation
+*/
+
+Ripemd128_Context :: struct {
+	s:  [4]u32,
+	x:  [RIPEMD_128_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd160_Context :: struct {
+	s:  [5]u32,
+	x:  [RIPEMD_160_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd256_Context :: struct {
+	s:  [8]u32,
+	x:  [RIPEMD_256_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+Ripemd320_Context :: struct {
+	s:  [10]u32,
+	x:  [RIPEMD_320_BLOCK_SIZE]byte,
+	nx: int,
+	tc: u64,
+}
+
+RIPEMD_128_SIZE       :: 16
+RIPEMD_128_BLOCK_SIZE :: 64
+RIPEMD_160_SIZE       :: 20
+RIPEMD_160_BLOCK_SIZE :: 64
+RIPEMD_256_SIZE       :: 32
+RIPEMD_256_BLOCK_SIZE :: 64
+RIPEMD_320_SIZE       :: 40
+RIPEMD_320_BLOCK_SIZE :: 64
+
+S0 :: 0x67452301
+S1 :: 0xefcdab89
+S2 :: 0x98badcfe
+S3 :: 0x10325476
+S4 :: 0xc3d2e1f0
+S5 :: 0x76543210
+S6 :: 0xfedcba98
+S7 :: 0x89abcdef
+S8 :: 0x01234567
+S9 :: 0x3c2d1e0f
+
+RIPEMD_128_N0 := [64]uint {
+	0, 1,  2,  3,  4,  5,  6,  7, 8,  9, 10, 11, 12, 13, 14, 15,
+	7, 4,  13, 1,  10, 6,  15, 3, 12, 0, 9,  5,  2,  14, 11, 8,
+	3, 10, 14, 4,  9,  15, 8,  1, 2,  7, 0,  6,  13, 11, 5,  12,
+	1, 9,  11, 10, 0,  8,  12, 4, 13, 3, 7,  15, 14, 5,  6,  2,
+}
+
+RIPEMD_128_R0 := [64]uint {
+	11, 14, 15, 12, 5,  8,  7,  9,  11, 13, 14, 15, 6,  7,  9,  8,
+	7,  6,  8,  13, 11, 9,  7,  15, 7,  12, 15, 9,  11, 7,  13, 12,
+	11, 13, 6,  7,  14, 9,  13, 15, 14, 8,  13, 6,  5,  12, 7,  5,
+	11, 12, 14, 15, 14, 15, 9,  8,  9,  14, 5,  6,  8,  6,  5,  12,
+}
+
+RIPEMD_128_N1 := [64]uint {
+	5, 14, 7, 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12,
+	6, 11, 3, 7, 0, 13, 5, 10, 14, 15, 8, 12, 4, 9, 1, 2,
+	15, 5, 1, 3, 7, 14, 6, 9, 11, 8, 12, 2, 10, 0, 4, 13,
+	8, 6, 4, 1, 3, 11, 15, 0, 5, 12, 2, 13, 9, 7, 10, 14,
+}
+
+RIPEMD_128_R1 := [64]uint {
+	8, 9, 9, 11, 13, 15, 15, 5, 7, 7, 8, 11, 14, 14, 12, 6,
+	9, 13, 15, 7, 12, 8, 9, 11, 7, 7, 12, 7, 6, 15, 13, 11,
+	9, 7, 15, 11, 8, 6, 6, 14, 12, 13, 5, 14, 13, 13, 7, 5,
+	15, 5, 8, 11, 14, 14, 6, 14, 6, 9, 12, 9, 12, 5, 15, 8,
+}
+
+RIPEMD_160_N0 := [80]uint {
+	0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	7, 4, 13, 1, 10, 6, 15, 3, 12, 0, 9, 5, 2, 14, 11, 8,
+	3, 10, 14, 4, 9, 15, 8, 1, 2, 7, 0, 6, 13, 11, 5, 12,
+	1, 9, 11, 10, 0, 8, 12, 4, 13, 3, 7, 15, 14, 5, 6, 2,
+	4, 0, 5, 9, 7, 12, 2, 10, 14, 1, 3, 8, 11, 6, 15, 13,
+}
+
+RIPEMD_160_R0 := [80]uint {
+	11, 14, 15, 12, 5, 8, 7, 9, 11, 13, 14, 15, 6, 7, 9, 8,
+	7, 6, 8, 13, 11, 9, 7, 15, 7, 12, 15, 9, 11, 7, 13, 12,
+	11, 13, 6, 7, 14, 9, 13, 15, 14, 8, 13, 6, 5, 12, 7, 5,
+	11, 12, 14, 15, 14, 15, 9, 8, 9, 14, 5, 6, 8, 6, 5, 12,
+	9, 15, 5, 11, 6, 8, 13, 12, 5, 12, 13, 14, 11, 8, 5, 6,
+}
+
+RIPEMD_160_N1 := [80]uint {
+	5, 14, 7, 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12,
+	6, 11, 3, 7, 0, 13, 5, 10, 14, 15, 8, 12, 4, 9, 1, 2,
+	15, 5, 1, 3, 7, 14, 6, 9, 11, 8, 12, 2, 10, 0, 4, 13,
+	8, 6, 4, 1, 3, 11, 15, 0, 5, 12, 2, 13, 9, 7, 10, 14,
+	12, 15, 10, 4, 1, 5, 8, 7, 6, 2, 13, 14, 0, 3, 9, 11,
+}
+
+RIPEMD_160_R1 := [80]uint {
+	8, 9, 9, 11, 13, 15, 15, 5, 7, 7, 8, 11, 14, 14, 12, 6,
+	9, 13, 15, 7, 12, 8, 9, 11, 7, 7, 12, 7, 6, 15, 13, 11,
+	9, 7, 15, 11, 8, 6, 6, 14, 12, 13, 5, 14, 13, 13, 7, 5,
+	15, 5, 8, 11, 14, 14, 6, 14, 6, 9, 12, 9, 12, 5, 15, 8,
+	8, 5, 12, 9, 12, 5, 14, 6, 8, 13, 6, 5, 15, 13, 11, 11,
+}
+
+block :: #force_inline proc (ctx: ^$T, p: []byte) -> int {
+    when T == Ripemd128_Context {
+    	return ripemd_128_block(ctx, p)
+    }
+    else when T == Ripemd160_Context {
+    	return ripemd_160_block(ctx, p)
+    }
+    else when T == Ripemd256_Context {
+    	return ripemd_256_block(ctx, p)
+    }
+    else when T == Ripemd320_Context {
+    	return ripemd_320_block(ctx, p)
+    }
+}
+
+ripemd_128_block :: proc(ctx: ^$T, p: []byte) -> int {
+	n := 0
+	x: [16]u32 = ---
+	alpha: u32 = ---
+	p := p
+	for len(p) >= RIPEMD_128_BLOCK_SIZE {
+		a, b, c, d := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3]
+		aa, bb, cc, dd := a, b, c, d
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_128_N0[i]]
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb & dd | cc &~ dd) + x[RIPEMD_128_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd= dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 32 {
+			alpha = a + (d ~ (b & (c~d))) + x[RIPEMD_128_N0[i]] + 0x5a827999
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb | ~cc)) + x[RIPEMD_128_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 48 {
+			alpha = a + (d ~ (b | ~c)) + x[RIPEMD_128_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb & (cc~dd))) + x[RIPEMD_128_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		for i < 64 {
+			alpha = a + (c ~ (d & (b~c))) + x[RIPEMD_128_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_128_N1[i]]
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		c = ctx.s[1] + c + dd
+		ctx.s[1] = ctx.s[2] + d + aa
+		ctx.s[2] = ctx.s[3] + a + bb
+		ctx.s[3] = ctx.s[0] + b + cc
+		ctx.s[0] = c
+		p = p[RIPEMD_128_BLOCK_SIZE:]
+		n += RIPEMD_128_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_160_block :: proc(ctx: ^$T, p: []byte) -> int {
+    n := 0
+	x: [16]u32 = ---
+	alpha, beta: u32 = ---, ---
+	p := p
+	for len(p) >= RIPEMD_160_BLOCK_SIZE {
+		a, b, c, d, e := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4]
+		aa, bb, cc, dd, ee := a, b, c, d, e
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_160_N0[i]]
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ (cc | ~dd)) + x[RIPEMD_160_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 32 {
+			alpha = a + (b&c | ~b&d) + x[RIPEMD_160_N0[i]] + 0x5a827999
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&dd | cc&~dd) + x[RIPEMD_160_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 48 {
+			alpha = a + (b | ~c ~ d) + x[RIPEMD_160_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb | ~cc ~ dd) + x[RIPEMD_160_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 64 {
+			alpha = a + (b&d | c&~d) + x[RIPEMD_160_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&cc | ~bb&dd) + x[RIPEMD_160_N1[i]] + 0x7a6d76e9
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		for i < 80 {
+			alpha = a + (b ~ (c | ~d)) + x[RIPEMD_160_N0[i]] + 0xa953fd4e
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_160_N1[i]]
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		dd += c + ctx.s[1]
+		ctx.s[1] = ctx.s[2] + d + ee
+		ctx.s[2] = ctx.s[3] + e + aa
+		ctx.s[3] = ctx.s[4] + a + bb
+		ctx.s[4] = ctx.s[0] + b + cc
+		ctx.s[0] = dd
+		p = p[RIPEMD_160_BLOCK_SIZE:]
+		n += RIPEMD_160_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_256_block :: proc(ctx: ^$T, p: []byte) -> int {
+	n := 0
+	x: [16]u32 = ---
+	alpha: u32 = ---
+	p := p
+	for len(p) >= RIPEMD_256_BLOCK_SIZE {
+		a, b, c, d := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3]
+		aa, bb, cc, dd := ctx.s[4], ctx.s[5], ctx.s[6], ctx.s[7]
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_128_N0[i]]
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb & dd | cc &~ dd) + x[RIPEMD_128_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd= dd, alpha, bb, cc
+			i += 1
+		}
+		t := a
+		a = aa
+		aa = t
+		for i < 32 {
+			alpha = a + (d ~ (b & (c~d))) + x[RIPEMD_128_N0[i]] + 0x5a827999
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb | ~cc)) + x[RIPEMD_128_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = b
+		b = bb
+		bb = t
+		for i < 48 {
+			alpha = a + (d ~ (b | ~c)) + x[RIPEMD_128_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (dd ~ (bb & (cc~dd))) + x[RIPEMD_128_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = c
+		c = cc
+		cc = t
+		for i < 64 {
+			alpha = a + (c ~ (d & (b~c))) + x[RIPEMD_128_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_128_R0[i])
+			alpha = util.ROTL32(alpha, s)
+			a, b, c, d = d, alpha, b, c
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_128_N1[i]]
+			s = int(RIPEMD_128_R1[i])
+			alpha = util.ROTL32(alpha, s)
+			aa, bb, cc, dd = dd, alpha, bb, cc
+			i += 1
+		}
+		t = d
+		d = dd
+		dd = t
+		ctx.s[0] += a
+		ctx.s[1] += b
+		ctx.s[2] += c
+		ctx.s[3] += d
+		ctx.s[4] += aa
+		ctx.s[5] += bb
+		ctx.s[6] += cc
+		ctx.s[7] += dd
+		p = p[RIPEMD_256_BLOCK_SIZE:]
+		n += RIPEMD_256_BLOCK_SIZE
+	}
+	return n
+}
+
+ripemd_320_block :: proc(ctx: ^$T, p: []byte) -> int {
+    n := 0
+	x: [16]u32 = ---
+	alpha, beta: u32 = ---, ---
+	p := p
+	for len(p) >= RIPEMD_320_BLOCK_SIZE {
+		a, b, c, d, e := ctx.s[0], ctx.s[1], ctx.s[2], ctx.s[3], ctx.s[4]
+		aa, bb, cc, dd, ee := ctx.s[5], ctx.s[6], ctx.s[7], ctx.s[8], ctx.s[9]
+		for i,j := 0, 0; i < 16; i, j = i+1, j+4 {
+			x[i] = u32(p[j]) | u32(p[j+1])<<8 | u32(p[j+2])<<16 | u32(p[j+3])<<24
+		}
+		i := 0
+		for i < 16 {
+			alpha = a + (b ~ c ~ d) + x[RIPEMD_160_N0[i]]
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ (cc | ~dd)) + x[RIPEMD_160_N1[i]] + 0x50a28be6
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t := b
+		b = bb
+		bb = t
+		for i < 32 {
+			alpha = a + (b&c | ~b&d) + x[RIPEMD_160_N0[i]] + 0x5a827999
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&dd | cc&~dd) + x[RIPEMD_160_N1[i]] + 0x5c4dd124
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = d
+		d = dd
+		dd = t
+		for i < 48 {
+			alpha = a + (b | ~c ~ d) + x[RIPEMD_160_N0[i]] + 0x6ed9eba1
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb | ~cc ~ dd) + x[RIPEMD_160_N1[i]] + 0x6d703ef3
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = a
+		a = aa
+		aa = t
+		for i < 64 {
+			alpha = a + (b&d | c&~d) + x[RIPEMD_160_N0[i]] + 0x8f1bbcdc
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb&cc | ~bb&dd) + x[RIPEMD_160_N1[i]] + 0x7a6d76e9
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = c
+		c = cc
+		cc = t
+		for i < 80 {
+			alpha = a + (b ~ (c | ~d)) + x[RIPEMD_160_N0[i]] + 0xa953fd4e
+			s := int(RIPEMD_160_R0[i])
+			alpha = util.ROTL32(alpha, s) + e
+			beta = util.ROTL32(c, 10)
+			a, b, c, d, e = e, alpha, b, beta, d
+			alpha = aa + (bb ~ cc ~ dd) + x[RIPEMD_160_N1[i]]
+			s = int(RIPEMD_160_R1[i])
+			alpha = util.ROTL32(alpha, s) + ee
+			beta = util.ROTL32(cc, 10)
+			aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
+			i += 1
+		}
+		t = e
+		e = ee
+		ee = t
+		ctx.s[0] += a
+		ctx.s[1] += b
+		ctx.s[2] += c
+		ctx.s[3] += d
+		ctx.s[4] += e
+		ctx.s[5] += aa
+		ctx.s[6] += bb
+		ctx.s[7] += cc
+		ctx.s[8] += dd
+		ctx.s[9] += ee
+		p = p[RIPEMD_320_BLOCK_SIZE:]
+		n += RIPEMD_320_BLOCK_SIZE
+	}
+	return n
+}
--- a/Show More
+++ b/Show More