Google Git
Sign in
chromium/external/github.com/google/XNNPACK/HEAD
  1. d50a82e xnnpack_configurable_defines match xnnpack_kleidiai_defines by Frank Barchard · 11 hours ago upstream/master
  2. e93f337 Merge pull request #11142 from destro4evr-rgb:fix/litert-rank0-crash-fc-embedlookup-gathernd by XNNPACK Team · 11 hours ago
  3. 303dc14 Add ARM SME dot product kernels for int8 activations with int4 and int2 weights. by Volodymyr Kysenko · 13 hours ago
  4. 47fef97 Simplify away provably true checks by Dillon Sharlet · 13 hours ago
  5. 426b707 Merge pull request #10791 from aizu-m:deconv-igemm-indirection-output-size by XNNPACK Team · 13 hours ago
  6. 373f03c Remove internal implementation details from fully-connected subgraph test by Dillon Sharlet · 13 hours ago
  7. 3a32935 litert/tensor: guard FullyConnected, EmbeddingLookup, GatherNd against rank-0 input tensors by destro4evr-rgb · 21 hours ago
  8. 54ea1b2 Merge pull request #11131 from destro4evr-rgb:fix/litert-select-rank0-null-deref by XNNPACK Team · 21 hours ago
  9. 735a993 Merge pull request #10986 from damdoo001-arm:damdoo01/int4_qc4w by XNNPACK Team · 23 hours ago
  10. 175d008 Clean up rebased INT4 integration by Damien Dooley · 26 hours ago
  11. f742e00 Added kernels to build.gn by Damien Dooley · 26 hours ago
  12. cd321f8 Guard SME2 QC4W MR before configuration by Damien Dooley · 26 hours ago
  13. 48e5f8a Addressed feedback re:assert, licence header and yoda condition by Damien Dooley · 26 hours ago
  14. 8a3f46e Keep generated copyright headers generator-owned by Damien Dooley · 26 hours ago
  15. 45eeefd Use Arm copyright notice format 2 by Damien Dooley · 26 hours ago
  16. 510c0d8 Use XNNPACK license for added files by Damien Dooley · 26 hours ago
  17. bd10b21 Add Arm copyright notices by Damien Dooley · 26 hours ago
  18. 168e8c9 Gate packed QC4W rewrite on zero point by Damien Dooley · 26 hours ago
  19. 2b1a0e9 Validate SME2 QC4W packing parameters by Damien Dooley · 26 hours ago
  20. 76fa737 Integrate SME2 INT4 fully connected kernels by Damien Dooley · 26 hours ago
  21. d36e64d Remove DotSchedulingTest.OpaqueSplitFactorBoundsAreProvable by Dillon Sharlet · 35 hours ago
  22. f10c3e9 Align the `m` dimension of `pack_a` outputs by XNNPACK Team · 2 days ago
  23. 2919616 Merge pull request #11105 from destro4evr-rgb:fix/litert-xnnpack-scale-array-oob-read by XNNPACK Team · 2 days ago
  24. d3bfd20 Merge pull request #10877 from damdoo001-arm:damdoo01/int2_qc2w by XNNPACK Team · 2 days ago
  25. 1640a03 Align the `m` dimension of `pack_a` outputs by Dillon Sharlet · 2 days ago
  26. 42992df litert/tensor/arithmetic.h: guard Select() against rank-0 a/b with 1-D condition by destro4evr-rgb · 2 days ago
  27. def8e5b Included bazel dependency by Damien Dooley · 2 days ago
  28. 2f3bec1 Merge pull request #11125 from destro4evr-rgb:fix/litert-split-axis-oob-check by XNNPACK Team · 2 days ago
  29. c0b771a Merge pull request #11056 from destro4evr-rgb:fix/litert-mean-slice-shape-empty-null-deref by XNNPACK Team · 2 days ago
  30. f356f73 Merge pull request #11114 from kgotlinux:freebsd-ppc-fix by XNNPACK Team · 2 days ago
  31. deaf412 conversion: add explicit quantized_dimension < 0 guard per review feedback by destro4evr-rgb · 2 days ago
  32. a9e1d53 put tested value on the left in reshape regression test assertions by Aizal Khan · 2 days ago
  33. 73ef51e litert/tensor/arithmetic.h: bounds-check Split axis against input rank by destro4evr-rgb · 2 days ago
  34. a5acbbe Enable memory pool in ynnpack. by Volodymyr Kysenko · 2 days ago
  35. e8a2152 Zero packed weights buffer before packing in dynamic batch matrix multiply by Frank Barchard · 3 days ago
  36. 8974285 Implements "Optimization of a GEMM Implementation using Intel AMX", Shintaro Endo, Hayato Ohshima, and Takeshi Nanri. (HPCAsia 2026). by Marie White · 3 days ago
  37. 6c48650 Update GH actions from gcc9 to gcc10 by Dillon Sharlet · 3 days ago
  38. c550c03 Add gcc10 to the docker images for GH actions by Dillon Sharlet · 3 days ago
  39. 617c23e f32-vexp - improved coefficients for rational polynomial. by Frank Barchard · 4 days ago
  40. 8c02ee3 Merge pull request #11057 from destro4evr-rgb:fix/litert-conv-pool-missing-rank-checks by XNNPACK Team · 4 days ago
  41. 54bc1d6 Merge pull request #10992 from melkap01-Arm:KleidiAI_pf32_sme2_gemm_gemv by XNNPACK Team · 4 days ago
  42. 1483329 Fix PowerPC hardware detection on FreeBSD by Katarzyna Kubaj · 4 days ago
  43. c98efe7 Optimizes split_k lengths depending on shape characteristics: by Dillon Sharlet · 4 days ago
  44. 59081d1 GN error fix by melkap01 · 4 days ago
  45. 8e26fbe Temporarily disable uses of `call_stmt::attrs` by Dillon Sharlet · 4 days ago
  46. c0d7367 Merge branch 'google:master' into KleidiAI_pf32_sme2_gemm_gemv by Melike Kaptan · 4 days ago
  47. b9b1cec [gn] Add missing benchmarks and gemm tests by Richard Townsend · 5 days ago
  48. 3d845dc fix: guard block_size == 0 before division in blockwise path by destro4evr-rgb · 5 days ago
  49. 78574e5 litert/xnnpack: validate scale array size against channel/block count in DefineValue() by destro4evr-rgb · 5 days ago
  50. dc9a4c3 Minor performance improvements and fixes by Dillon Sharlet · 5 days ago
  51. 1291b56 Make it possible to look up quantized embeddings directly. by Quentin Khan · 5 days ago
  52. 6b2a068 Address INT2 review feedback by Damien Dooley · 5 days ago
  53. 9efa9dc Fix QC2W packing-selection test shape by Damien Dooley · 5 days ago
  54. 1e58214 Keep generated copyright headers generator-owned by Damien Dooley · 5 days ago
  55. b51b86d Use Arm copyright notice format 2 by Damien Dooley · 5 days ago
  56. 999b235 Use XNNPACK license for added files by Damien Dooley · 5 days ago
  57. cd4c51a Add Arm copyright notices by Damien Dooley · 5 days ago
  58. 6587c40 Update KleidiAI to v1.30.0 by Damien Dooley · 5 days ago
  59. ddd3e50 Derive QC2W K eligibility from input shape by Damien Dooley · 5 days ago
  60. 18ec556 Fix Bazel dependency for QC2W config test by Damien Dooley · 5 days ago
  61. 8c4a252 Register SME2 QC2W unit-batch microkernel by Damien Dooley · 5 days ago
  62. 53943f0 Integrate SME2 INT2 fully connected kernels by Damien Dooley · 5 days ago
  63. 29a80b0 Merge pull request #11085 from EylonKrause:fix/rsqrt-const-propagation-flag by XNNPACK Team · 5 days ago
  64. 77c0b61 Merge pull request #11091 from destro4evr-rgb:fix/litert-rope-weights-shape-empty-null-deref by XNNPACK Team · 5 days ago
  65. 9fc34eb Merge pull request #10933 from leduclean:qs-rpreluc-rvv by XNNPACK Team · 5 days ago
  66. 5be13a0 Merge pull request #11078 from EylonKrause:fix/fuse-dims-zero-axes-count by XNNPACK Team · 5 days ago
  67. 8d7862d Optimizes split_k lengths depending on shape characteristics: by Marie White · 5 days ago
  68. a2d2195 Merge pull request #10990 from rishuranjanofficial:fix/depth-to-space-int-overflow-32bit by XNNPACK Team · 5 days ago
  69. 4013867 litert/xnnpack: add rank guards to Conv/Pool ToXnnpack() before shape subscripts by destro4evr-rgb · 5 days ago
  70. 7f616c1 litert/xnnpack: guard against 0D scalar axes/begin tensor in Mean and Slice by destro4evr-rgb · 5 days ago
  71. c5ed6a9 Remove custom buffer storage scheduling. by Volodymyr Kysenko · 6 days ago chromium/8038 chromium/8039 chromium/8040 chromium/8041 chromium/8042 chromium/8043
  72. ad20a76 Add Mean Generator to LiteRT ATS by Gerardo Carranza · 8 days ago
  73. 0363fd2 - Enable avx512bf16 by default on x86 (both 32-bit and 64-bit) in Bazel BUILD and tools/xnncommon.py for parity with avx512fp16. by Frank Barchard · 8 days ago
  74. 1eb7302 Use zero-initialized instead of undef partial loads by Dillon Sharlet · 8 days ago chromium/8032 chromium/8033 chromium/8034 chromium/8035 chromium/8036 chromium/8037
  75. 5de8644 Fix cmake build for WASM Relaxed SIMD FP16 by Frank Barchard · 8 days ago
  76. 09117fc Add blockwise product benchmark by Volodymyr Kysenko · 8 days ago
  77. 2b763bd Make `type_size_bits` and `type_size_bytes` more helpful by Dillon Sharlet · 8 days ago
  78. 9f5e4f9 litert/xnnpack: guard against 0D scalar weights tensor in RopeOperation::ToXnnpack() by destro4evr-rgb · 8 days ago
  79. d24998f subgraph: modulus preserves only zeros, not ones, in constant propagation by Eylon Krause · 9 days ago
  80. f7f2d30 Merge pull request #11084 from rahulchaphalkar:f16-avgpool by XNNPACK Team · 9 days ago
  81. 78d4717 Add accurate range reduction for sin/cos/tan for extreme values by Dillon Sharlet · 9 days ago
  82. 851b461 subgraph: reciprocal_square_root preserves ONE, not ZERO, in constant propagation by Eylon Krause · 9 days ago
  83. 34225f2 Add direct Wasm FP16 avgpool microkernel by rahulchaphalkar · 9 days ago
  84. b571de8 Add Wasm f16-f32acc avgpool microkernel by rahulchaphalkar · 9 days ago
  85. eca8e9b Add scalar f16-f32acc avgpool fallback by rahulchaphalkar · 9 days ago
  86. 787c7cb F32/16 AVX2/AVX-512 packw use permi2w by Frank Barchard · 9 days ago chromium/8030 chromium/8031
  87. 2f4b995 Change LockedBufferSpan to make it possible to create multiple spans from the same locked object. by Quentin Khan · 9 days ago
  88. 073f3b0 Merge pull request #11058 from destro4evr-rgb:fix/litert-split-empty-axis-buffer-null-deref by XNNPACK Team · 10 days ago
  89. c02573b [gn] Add a comparison script for compile_commands.json files by Richard Townsend · 10 days ago
  90. e4a13c2 subgraph: reject zero axes_count in fuse_dims (out-of-bounds shape write) by Eylon Krause · 10 days ago
  91. 16f3409 Improve scheduling of reduction-rooted fused chains. by Volodymyr Kysenko · 10 days ago
  92. eb452a7 Merge pull request #11071 from rahulchaphalkar:wasm-f16 by XNNPACK Team · 11 days ago
  93. 5e9b9d4 Avoid setting conflicting "noinline" and "always_inline" attributes on functions in XNNPACK: by Dillon Sharlet · 11 days ago
  94. f682460 Refactor loop fusion and step reconciliation logic in ynnpack runtime. by Volodymyr Kysenko · 11 days ago
  95. 233d90d wasm f16 for dwconv by rahulchaphalkar · 11 days ago
  96. dd2da73 reallocate deconv igemm indirection buffer on output-size change by Aizal Khan · 11 days ago
  97. 2b0b175 Allow using `transpose_a` kernels with rank 1 `a` by Dillon Sharlet · 11 days ago
  98. 2092a22 Add avx512 symmetric int8 dot kernels in a few cases by Dillon Sharlet · 11 days ago
  99. f526f78 Fix f32-vsqrt and f16-vsqrt inf handling by Frank Barchard · 11 days ago
  100. 6c9c509 Guard fully connected tests if kleidi disabled by Frank Barchard · 11 days ago