43 Pull requests merged by 2 people

• Fix uninitialized memory in some reference packing kernels

  #7988 merged Mar 11, 2025

• Remove debugging leftover

  #7989 merged Mar 11, 2025

• Simplify packing.cc and improve its failure message.

  #7987 merged Mar 10, 2025

• Add support for bf16 * bf16 -> f32 batch_matrix_multiply

  #7933 merged Mar 10, 2025

• Relax tolerance for tanh

  #7986 merged Mar 10, 2025

• Fix stale generated files

  #7983 merged Mar 10, 2025

• Fix flaky ibilinear test failures

  #7961 merged Mar 10, 2025

• F16-GEMM for WAsm Relaxed SIMD

  #7985 merged Mar 10, 2025

• Test wider range of float values

  #7971 merged Mar 10, 2025

• Regenerate stale code

  #7974 merged Mar 10, 2025

• Remove final reference to XNN_ENABLE_GEMM_M_SPECIALIZATION

  #7965 merged Mar 10, 2025

• Set timeout = "moderate" for //test:softmax_test, which sometimes times out on wasm.

  #7980 merged Mar 10, 2025

• Allocate workspace closer to where we'll use it.

  #7981 merged Mar 10, 2025

• Rollback of "Fix missing RVV gemm kernel descriptions and regenerate tests/benchmarks"

  #7982 merged Mar 10, 2025

• Zero initialize memory for weights.

  #7979 merged Mar 10, 2025

• Fix missing RVV gemm kernel descriptions and regenerate tests/benchmarks

  #7888 merged Mar 10, 2025

• Update bf16 microkernel in gemm-config.c

  #7978 merged Mar 10, 2025

• Restored old test tolerances after the proper fix for flakyness.

  #7973 merged Mar 8, 2025

• Add clang-format off to generated code

  #7972 merged Mar 8, 2025

• Fix warnings passing 0 to std::fill

  #7968 merged Mar 7, 2025

• Fix a different source of flakiness in rdsum-microkernel-tester.h

  #7969 merged Mar 7, 2025

• Fix flaky test regression

  #7967 merged Mar 7, 2025

• Make the reference softmax more robust

  #7960 merged Mar 7, 2025

• Move operator benchmarks to bench/operators

  #7959 merged Mar 7, 2025

• Fix warning due to implicit conversion

  #7963 merged Mar 7, 2025

• Cleaned up rdsum, refactored testing.

  #7964 merged Mar 7, 2025

• Update stale generated code

  #7957 merged Mar 7, 2025

• Move operator tests to subfolder

  #7954 merged Mar 7, 2025

• Rewrite all header include paths to be from the root of the repo

  #7958 merged Mar 6, 2025

• Remove XNN_ENABLE_GEMM_M_SPECIALIZATION option

  #7956 merged Mar 6, 2025

• Minor test cleanups

  #7955 merged Mar 6, 2025

• Move simd tests to subfolder

  #7953 merged Mar 6, 2025

• Added u8-rminmax microkernels.

  #7938 merged Mar 5, 2025

• Added s8-rminmax microkernels.

  #7929 merged Mar 5, 2025

• Add qd8-f32-qc4w AVX512 gemm microkernels

  #7935 merged Mar 5, 2025

• Remove tests for deprecated APIs

  #7948 merged Mar 5, 2025

• Explicitly subtract the stack pointer instead of subtracting from it.

  #7943 merged Mar 5, 2025

• Add c8 variants of qd8-f32-qc8w gemm microkernels

  #7825 merged Mar 5, 2025

• Add testing of reshaped shapes

  #7945 merged Mar 5, 2025

• Refactor static-reduce test

  #7944 merged Mar 5, 2025

• Copybara import of the project:

  #7942 merged Mar 4, 2025

• Update KleidiAI to use GitHub

  #7941 merged Mar 4, 2025

• Add sharding to slow test

  #7937 merged Mar 4, 2025

9 Pull requests opened by 3 people

• Implement a tile dispatch in FC layer

  #7940 opened Mar 4, 2025

• Add `4d_tile_2d_dynamic` to `pthreadpool`.

  #7950 opened Mar 5, 2025

• Use dynamic scheduling for `3d_tile_2d` and `2d_tile_2d` in the iGEMM ops.

  #7951 opened Mar 5, 2025

• Use the `4d_tile_2d_dynamic` in iGEMMs.

  #7952 opened Mar 5, 2025

• Rewrite unary and binary subgraph tests to be independent of operator API

  #7962 opened Mar 7, 2025

• Add clang-format off to header tables

  #7975 opened Mar 9, 2025

• Add clang-format off to generated tests and benchmarks

  #7976 opened Mar 9, 2025

• [RVV] add rvv optimization for qs8/qu8-rsum and qs8/qu8-rdsum

  #7977 opened Mar 9, 2025

• Implemented s8-rdminmax via SIMD wrappers.

  #7984 opened Mar 10, 2025

2 Issues opened by 2 people

• Running XNNPACK on constrained environment

  #7966 opened Mar 7, 2025

• Investigation of inactive tests for GEMM kernels

  #7949 opened Mar 5, 2025

11 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Support RVV f32-dwconv-unipass

  #7402 commented on Mar 11, 2025 • 7 new comments

• RVV support qd8-f32-qc8w-gemm/igemm and qd8-f32-qc4w-gemm

  #7932 commented on Mar 11, 2025 • 5 new comments

• Replace conv-hwc yaml with table header

  #7445 commented on Mar 5, 2025 • 3 new comments

• Modify f32-gemm for HVX

  #7918 commented on Mar 5, 2025 • 2 new comments

• Enable AVXVNNI microkernels for mobile platforms

  #7828 commented on Mar 5, 2025 • 1 new comment

• F32-GEMM for HVX use Vsf for clamping instead of Vw

  #7928 commented on Mar 11, 2025 • 1 new comment

• tags please

  #7930 commented on Mar 6, 2025 • 0 new comments

• HVX x32 transposec

  #7395 commented on Mar 5, 2025 • 0 new comments

• QS8 / QU8 PReLU microkernels

  #7738 commented on Mar 6, 2025 • 0 new comments

• Add QNX to recognized systems

  #7798 commented on Mar 6, 2025 • 0 new comments

• Use `xnn_gemm_best_tile_size` and flip the order of `mr`/`nr` in the iGEMM-based ops, analogously to what we did for the GEMM ops.

  #7921 commented on Mar 7, 2025 • 0 new comments