Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit History

Commits on Jul 22, 2024

• increase logging buffer size limit to avoid failing on large logging messages

  austinvhuang

  committed
  15142bd

  Copy full SHA for 15142bd

Commits on Jul 21, 2024

• change toGPU to allow ahead-of-time creation of the staging buffer, fix out-of-date docstrings

  austinvhuang

  committed
  d6a02d5

  Copy full SHA for d6a02d5

Commits on Jul 15, 2024

• fix docstring formatting for doxygen consumption

  austinvhuang

  committed
  356adc7

  Copy full SHA for 356adc7

Commits on Jul 12, 2024

• replace wgpuQueueWriteBuffer for params in examples to an overload of toGPU. Update readme based on internal blog post content

  austinvhuang

  committed
  33cb63a

  Copy full SHA for 33cb63a

Commits on Jul 11, 2024

• shadertui live reloading demo initial implementation

  austinvhuang

  committed
  fb56cf0

  Copy full SHA for fb56cf0

Commits on Jul 10, 2024

• add -stdlib=libc++ to build flags, minor cleanup

  austinvhuang

  committed
  f983bcf

  Copy full SHA for f983bcf

Commits on Jul 9, 2024

• 1D blocktiling matmul works, set as the default

  austinvhuang

  committed
  dda3934

  Copy full SHA for dda3934

• matmul cleanup (kernel v3 partly working for small tile sizes), minor readme edits

  austinvhuang

  committed
  a11c98d

  Copy full SHA for a11c98d

Commits on Jul 4, 2024

• Specify xyz of number of workgroups for createKernel (ala cuda) instead of specifying total # threads (and automatically dividing each dimension by workgroup size). tiled matmul shader now produces…

  Show description for f72c857
  austinvhuang

  committed
  f72c857

  Copy full SHA for f72c857

Commits on Jul 1, 2024

• debugging mm example variations

  austinvhuang

  committed
  3781b21

  Copy full SHA for 3781b21

Commits on Jun 29, 2024

• add arrayutil overloads, debugging mm

  austinvhuang

  committed
  829ddc5

  Copy full SHA for 829ddc5

Commits on Jun 28, 2024

• migrate from google style casing to webgpu camelcasing, qkv projection (wip)

  austinvhuang

  committed
  65ae859

  Copy full SHA for 65ae859

• use wgpu style convention for functions

  austinvhuang

  committed
  fae185f

  Copy full SHA for fae185f

• Merge branch 'main' into fix/duplicated-symbols

  austinvhuang

  authored
  3958b23

  Copy full SHA for 3958b23

Commits on Jun 27, 2024

• Merge pull request #6 from junjihashimoto/fix/logger

  Show description for 63c1ef7
  austinvhuang

  authored
  63c1ef7

  Copy full SHA for 63c1ef7

• typo fix (via PR web ui)

  austinvhuang

  authored
  5137a13

  Copy full SHA for 5137a13

• Add inline to avoid duplicated-symbols-error

  junjihashimoto

  committed
  3affc31

  Copy full SHA for 3affc31

• Update LOG's color with reset

  junjihashimoto

  committed
  a402541

  Copy full SHA for a402541

Commits on Jun 26, 2024

• move tui into experimental.

  austinvhuang

  committed
  e5ca915

  Copy full SHA for e5ca915

• move kernel tests to transformers, tweak readme and builds. move superfluous files out of utils/

  austinvhuang

  committed
  e635ec4

  Copy full SHA for e635ec4

Commits on Jun 25, 2024

• add more docstrings to gpu.h, add sonnet fluid sim to experimental, fix makefile dependency chain for examples

  austinvhuang

  committed
  59c7bf6

  Copy full SHA for 59c7bf6

• fast builds working for linux + examples

  austinvhuang

  committed
  2b6154b

  Copy full SHA for 2b6154b

Commits on Jun 24, 2024

• Migrate to prebuilt libdawn binary. MacOS+apple silicon is working, will check linux next. Add setup.cpp script for downloading the shared library and other setup tasks.

  austinvhuang

  committed
  d285702

  Copy full SHA for d285702

Commits on Jun 23, 2024

• minor cleaup + release prep

  austinvhuang

  committed
  34c17a9

  Copy full SHA for 34c17a9

Commits on Jun 20, 2024

• move tui code to tui.h

  austinvhuang

  committed
  bc91b5e

  Copy full SHA for bc91b5e

Commits on Jun 19, 2024

• Simplify CreateKernel overloads, move shaders.h to utils/

  austinvhuang

  committed
  bf71d53

  Copy full SHA for bf71d53

• Refactor - fold output into one of the TensorList arguments, get rid of MultiKernel (for now), update examples accordingly

  austinvhuang

  committed
  cb77adc

  Copy full SHA for cb77adc

• fix test_kernels build, clean up makefile

  austinvhuang

  committed
  f9d9819

  Copy full SHA for f9d9819

Commits on Jun 18, 2024

• API refinement in preparation for release (GPUTensor -> Tensor, GPUContext -> Context for Create* consistency), fix example builds, start docstrings (WIP)

  austinvhuang

  committed
  32828d5

  Copy full SHA for 32828d5

• fix build (disable error-on-warnings), WIP multikernel support for nThreads specificaiton

  austinvhuang

  committed
  adf4ce3

  Copy full SHA for adf4ce3

Commits on Jun 17, 2024

• handle 3D nthreads -> workgroup mappings for CreateKernel, fix GELU bug for inputs > 10.0, get rid of inferred # threads from output from output dimensions (too easy to introduce bugs), refine rend…

  Show description for 0735132
  austinvhuang

  committed
  0735132

  Copy full SHA for 0735132

Commits on Jun 16, 2024

• Persist WGPUComputePipeline in the kernel struct, generalize # workgroup calculation for X Y Z directions instead of only working in 1D

  austinvhuang

  committed
  ac7743b

  Copy full SHA for ac7743b

Commits on Jun 14, 2024

• resource deallocation (debugging WIP)

  austinvhuang

  committed
  61b92da

  Copy full SHA for 61b92da

Commits on Jun 12, 2024

• Simplify shader API - use Create* convention for CreateShader, start gpu puzzles example, pin dawn build commit, add some checks to Makefile

  austinvhuang

  committed
  a1b31f7

  Copy full SHA for a1b31f7

Commits on Jun 11, 2024

• clean up API naming convention consistency - Prepare -> Create, Launch -> Dispatch

  austinvhuang

  committed
  2223ea4

  Copy full SHA for 2223ea4