Code

0.2.0 Release#74

Code

Merged

austinvhuang merged 53 commits intomainAnswerDotAI/gpu.cpp:mainfrom devAnswerDotAI/gpu.cpp:devCopy head branch name to clipboardFeb 10, 2025

Commits

Commits on Sep 27, 2024

• feat: 1. Modified the 'lib' rule in the Makefile file located in the root directory of the project, allowing it to generate gpu.cpp dynamic libraries with different suffixes based on the system.

  Show description for ba32717
  buttfa

  committed
  ba32717

  Copy full SHA for ba32717

Commits on Oct 5, 2024

• Merge pull request #64 from buttfa/feature_library

  Show description for 0217886
  austinvhuang

  authored
  0217886

  Copy full SHA for 0217886

Commits on Oct 13, 2024

• Add kShaderMatmul2DTiling in kernels.h

  junjihashimoto

  committed
  7addf83

  Copy full SHA for 7addf83

• Reduce matmul-kernel creation time

  junjihashimoto

  committed
  da1f32d

  Copy full SHA for da1f32d

Commits on Oct 18, 2024

• Change Kernel to shared_ptr<RawKernel> to support cached kernels

  junjihashimoto

  committed
  dd2a25f

  Copy full SHA for dd2a25f

• Add caches for ops.cpp

  junjihashimoto

  committed
  9fff8cd

  Copy full SHA for 9fff8cd

• Add caches for unittests.cpp

  junjihashimoto

  committed
  b9fb38b

  Copy full SHA for b9fb38b

• Fix bugs

  junjihashimoto

  committed
  efb87ee

  Copy full SHA for efb87ee

• Remove global variables of kernels

  junjihashimoto

  committed
  c474fba

  Copy full SHA for c474fba

• Fix the matmul of version 1 in unittests

  junjihashimoto

  committed
  438be22

  Copy full SHA for 438be22

• Add the duration-time of matmul_forward_dummy to compare GPU's one with CPU's one

  junjihashimoto

  committed
  dd145ac

  Copy full SHA for dd145ac

• Add wgpuBufferRelease for CopyData

  junjihashimoto

  committed
  590f257

  Copy full SHA for 590f257

• Add wgpuCommandBufferRelease after calling wgpuQueueSubmit

  junjihashimoto

  committed
  6c52b98

  Copy full SHA for 6c52b98

• Add wgpuCommandEncoderRelease after calling wgpuCommandEncoderFinish

  junjihashimoto

  committed
  a6140e0

  Copy full SHA for a6140e0

• Add wgpuComputePassEncoderRelease after calling wgpuComputePassEncoderEnd

  junjihashimoto

  committed
  30ed026

  Copy full SHA for 30ed026

Commits on Oct 19, 2024

• Merge branch 'main' into dev

  austinvhuang

  committed
  3c06137

  Copy full SHA for 3c06137

• chore: Set a check-os target and improved the description of how to install dependencies.

  buttfa

  committed
  0a9437f

  Copy full SHA for 0a9437f

Commits on Oct 20, 2024

• Merge pull request #68 from junjihashimoto/feature/cache

  Show description for f4e1683
  austinvhuang

  authored
  f4e1683

  Copy full SHA for f4e1683

• Merge pull request #69 from buttfa/feature_library

  Show description for 0e89e65
  austinvhuang

  authored
  0e89e65

  Copy full SHA for 0e89e65

Commits on Oct 21, 2024

• Add the ops of AoT

  junjihashimoto

  committed
  d4eb571

  Copy full SHA for d4eb571

• Update

  junjihashimoto

  committed
  43e4ac0

  Copy full SHA for 43e4ac0

Commits on Oct 22, 2024

• Update

  junjihashimoto

  committed
  1d8e435

  Copy full SHA for 1d8e435

• Update

  junjihashimoto

  committed
  4985930

  Copy full SHA for 4985930

Commits on Oct 30, 2024

• Add summantion kernels

  junjihashimoto

  committed
  f3e0dbc

  Copy full SHA for f3e0dbc

• Add SumKernel

  junjihashimoto

  committed
  f956f2b

  Copy full SHA for f956f2b

Commits on Nov 3, 2024

• Add SumKernel2d

  junjihashimoto

  committed
  c13833f

  Copy full SHA for c13833f

• Merge branch 'dev' into feature/reduce

  junjihashimoto

  authored
  189375f

  Copy full SHA for 189375f

• Merge pull request #71 from junjihashimoto/feature/reduce

  Show description for e94aa02
  austinvhuang

  authored
  e94aa02

  Copy full SHA for e94aa02

• fix printf format codes

  austinvhuang

  committed
  c9b7018

  Copy full SHA for c9b7018

Commits on Nov 4, 2024

• Add a flag to disable bardward-pass

  junjihashimoto

  committed
  7ef40b0

  Copy full SHA for 7ef40b0

Commits on Nov 16, 2024

• Fix the bug of memory allocation

  junjihashimoto

  committed
  6be7e1e

  Copy full SHA for 6be7e1e

• Remove NUM_PARAMETER_LAYERS

  junjihashimoto

  committed
  f629a33

  Copy full SHA for f629a33

Commits on Nov 18, 2024

• Merge pull request #70 from junjihashimoto/feature/aot

  Show description for 28c7062
  austinvhuang

  authored
  28c7062

  Copy full SHA for 28c7062

Commits on Dec 25, 2024

• Add python bindings

  junjihashimoto

  committed
  6e3a240

  Copy full SHA for 6e3a240

Commits on Dec 28, 2024

• Add haskell bindings

  junjihashimoto

  committed
  3228b1b

  Copy full SHA for 3228b1b

Commits on Dec 29, 2024

• Merge pull request #73 from junjihashimoto/feature/python

  Show description for a7520ce
  austinvhuang

  authored
  a7520ce

  Copy full SHA for a7520ce

Commits on Jan 28, 2025

• migrate to updated dawn commit 556f960f44690b3b808c779c08b44d48d4292925 - main library builds and examples run successfully, TODOs are code cleanup + experimental + non-make builds

  austinvhuang

  committed
  4669791

  Copy full SHA for 4669791

• remove legacy dir from experimental

  austinvhuang

  committed
  254e4ea

  Copy full SHA for 254e4ea

• gpt2_webgpu_aot runs on mac after updating experimental/kernels/Makefile to point to updated dylib artifact, add note to third_party/lib/README.md regarding new dawn shared library build process

  austinvhuang

  committed
  f3f3b27

  Copy full SHA for f3f3b27

• add detailed note regarding dawn modifications to fix linker errors on mac

  austinvhuang

  committed
  b397959

  Copy full SHA for b397959

• correct commit hash

  austinvhuang

  committed
  40fd25d

  Copy full SHA for 40fd25d

• update setup.py auto-downloads to point to updated libwebgpu_dawn.dylib/so shared libraries

  austinvhuang

  committed
  46db79d

  Copy full SHA for 46db79d

• clang-format cleanup

  austinvhuang

  committed
  73f438a

  Copy full SHA for 73f438a

Commits on Jan 30, 2025

• make Context lifetime more robust dont rely on RVO which seems to fail if createContext is called from eg conditional branches. Remove webgpu from scratch tutorial to avoid having to maintain/updat…

  Show description for d8d618a
  austinvhuang

  committed
  d8d618a

  Copy full SHA for d8d618a

• Change priority of internal logging from kInfo to kTrace. Make julia the default shader for shadertui.

  austinvhuang

  committed
  ec68b14

  Copy full SHA for ec68b14

Commits on Feb 1, 2025

• bump dawn version to c469d593ac and remove WebGPU-distribution from third_party

  austinvhuang

  committed
  4589f1f

  Copy full SHA for 4589f1f

Commits on Feb 2, 2025

• move web build example to experimental due to emscriptens webgpu implementation lagging. few readme tweaks

  austinvhuang

  committed
  c3ee69b

  Copy full SHA for c3ee69b

• update artifact link

  austinvhuang

  committed
  3924552

  Copy full SHA for 3924552

• skip float16 targets in CI

  austinvhuang

  committed
  a8a44d3

  Copy full SHA for a8a44d3

• test float16 in CI

  austinvhuang

  committed
  7dc064c

  Copy full SHA for 7dc064c

Commits on Feb 7, 2025

• Fix pybind

  junjihashimoto

  committed
  041d2fd

  Copy full SHA for 041d2fd

• Fix haskell binding

  junjihashimoto

  committed
  89f9097

  Copy full SHA for 89f9097

Commits on Feb 8, 2025

• Merge pull request #75 from junjihashimoto/fix/pybind

  Show description for 6447d85
  austinvhuang

  authored
  6447d85

  Copy full SHA for 6447d85