Google Git
Sign in
chromium/external/github.com/google/XNNPACK/HEAD
  1. f08e7e8 Merge pull request #10781 from destro4evr-rgb:fix/dwconv-step-height-indirection-size-overflow by XNNPACK Team · 5 hours ago upstream/master
  2. 8def0d9 Implement parsing for CUSTOM and STABLEHLO_COMPOSITE operations. by Ping Yu · 10 hours ago
  3. ec5a9e0 Add a couple more shapes into XNNPACK attention benchmark. by Volodymyr Kysenko · 14 hours ago
  4. b977245 Fix schedule of pack_b for dot by Dillon Sharlet · 17 hours ago
  5. 9bc10e6 Compute loop workers globally on the assembled loop nest by Volodymyr Kysenko · 19 hours ago
  6. d9d3702 Merge pull request #10779 from aizu-m:reduce-compute-dim-narrowing by XNNPACK Team · 21 hours ago
  7. f825eff Adjust thread counts in YNNPACK benchmarks and runtime. by Volodymyr Kysenko · 3 days ago
  8. 32cad9d Disable fp8 on Apple by default by Dillon Sharlet · 4 days ago
  9. bd7f032 operators: guard dwconv step_height and indirection buffer size against overflow by destro4evr-rgb · 4 days ago
  10. 9175dd9 use size_t for reduced input dims in reduce compute kernels by Aizal Khan · 4 days ago
  11. fd976d9 Use cpuinfo support for ARM fp8 by Dillon Sharlet · 4 days ago
  12. c2a632d Fix validation tests when running in YNNPACK by Dillon Sharlet · 4 days ago
  13. aaecd5d Update cpuinfo in XNNPACK by Dillon Sharlet · 4 days ago
  14. 53a1797 Fix crash due to not allowing internal extra dimensions. by Dillon Sharlet · 4 days ago
  15. 5251854 Merge pull request #10702 from destro4evr-rgb:fix/litert-arithmetic-axis-bounds-checks by XNNPACK Team · 4 days ago
  16. 9fe71df litert/tensor/arithmetic.h: bounds-check axis and index values before vector subscript by destro4evr-rgb · 4 days ago
  17. 9b0e561 Merge pull request #10765 from destro4evr-rgb:fix/litert-unpack-axis-bounds-check by XNNPACK Team · 4 days ago
  18. 0e92559 Merge pull request #10756 from sachinhambar:fix/lut-fusion-invalid-unary-node by XNNPACK Team · 4 days ago
  19. 1dfe86e Merge pull request #10748 from aizu-m:conv-reshape-input-channels by XNNPACK Team · 4 days ago
  20. 218bb2e Change x86 uint8 x int2 dot kernels to use tile_k = 8 by Dillon Sharlet · 4 days ago
  21. 11f8dc8 Move num_inputs assert past the LUT-fused-node check in is_pure_unary_elementwise by Sachin Hambar · 4 days ago
  22. ff8907d Actually provide splits for the pack. by Volodymyr Kysenko · 4 days ago
  23. 1336d27 Add `YNN_NODE_FLAG_KEEP_SHAPE` flag and implement it in `ynn_define_slice_like` by Dillon Sharlet · 4 days ago
  24. f6d310e Fixed static tensor behavior with ynnpack to match XNNPACK. Old behavior didn't allow for external tensors with static weights. by Misha Gutman · 5 days ago
  25. c8306ce litert/arithmetic: add axis bounds check in Unpack() by destro4evr-rgb · 5 days ago
  26. 6d553e8 Add xnn_pack_qs8_to_qu8_qc4uw_gemm_gio_w by Frank Barchard · 5 days ago
  27. b1858fa Re-enable packw optimized kernels and refine ISA guards in GEMM configs. by Frank Barchard · 5 days ago
  28. dff8cb5 Force `LockedBufferSpan` data access through an lvalue. by Quentin Khan · 5 days ago chromium/7957 chromium/7958 chromium/7959 chromium/7960 chromium/7961 chromium/7962 chromium/7963
  29. 464e9d3 Remove deprecated quantization parameters from `ynn_define_convert` by Dillon Sharlet · 6 days ago
  30. 66b5f2b Try keeping buffer and tensor type coherent. by Quentin Khan · 6 days ago
  31. a635548 Add a loop fusion test for combining dot steps with LCM. by Volodymyr Kysenko · 6 days ago
  32. 7647747 Add a transposed decode1 attention benchmark. by Volodymyr Kysenko · 6 days ago
  33. b372ff1 Allow changing the order of the loops when looking for the match to fuse with. by Volodymyr Kysenko · 6 days ago
  34. 0976700 Fix LUT fusion treating an already-fused LUT node as a plain unary op by Sachin Hambar · 6 days ago
  35. b692f91 Allow passing a scalar value as the `TensorInit` buffer. by Quentin Khan · 6 days ago
  36. 79cbaa9 Match channel-mismatch reshape error to batch-matrix-multiply wording by Aizal Khan · 6 days ago
  37. f0fd32b Make B a dynamic value in concatenated_mixed_dot_bench by Dillon Sharlet · 6 days ago
  38. 0de517a Fix quantization export to allow explicit q/ dq nodes in the graph by XNNPACK Team · 6 days ago
  39. 5fe544d Support static FP32 bias for FP16 depthwise convolution 2D in XNNPACK subgraph. by Frank Barchard · 7 days ago
  40. e74ee67 Fix for fingerprint test by Frank Barchard · 7 days ago
  41. 3102091 Add dynamic shapes to `concatenated_mixed_dot_bench` by Dillon Sharlet · 7 days ago
  42. 2c59699 Remove exception for `pack_b` from the constant folding heuristic by Dillon Sharlet · 7 days ago
  43. f303466 Update KleidiAI and Slinky in XNNPACK by Dillon Sharlet · 7 days ago
  44. 599a1be validate runtime input channels in convolution reshape paths by Aizal Khan · 7 days ago
  45. 44db232 Fix presubmit checks about unchecked absl::StatusOr access. by Quentin Khan · 7 days ago
  46. 11d93dd Move scheduler bounds from producer loop splits to consumer inputs. by Volodymyr Kysenko · 7 days ago
  47. a1f8b25 Update XNNPACK dependencies by Dillon Sharlet · 8 days ago
  48. fdd1e56 Added option to enable ynnpack integration with litert by XNNPACK Team · 8 days ago
  49. 5d9f5ae xx-pad AVX512SKX microkernel with p64_u128 by Frank Barchard · 8 days ago
  50. bbc8bb6 Add concatenated_mixed_dot_bench by Dillon Sharlet · 8 days ago
  51. cdbda03 Update Slinky in XNNPACK by Dillon Sharlet · 8 days ago
  52. 79dc6c4 Fix warnings about control reaching the end of a non-void function. by Quentin Khan · 8 days ago
  53. 894aab8 Add `convert(ternary(...))` and `ternary(convert(...))` fusion rules by Dillon Sharlet · 8 days ago
  54. d3f0715 Merge pull request #10690 from velonica0:rvv-ppmm-new by XNNPACK Team · 8 days ago
  55. d5b1ae6 Merge pull request #10629 from GregoryComer:bf16-rewrite by XNNPACK Team · 8 days ago
  56. 64c959c Implement conversion over ranges of elements. by Quentin Khan · 8 days ago
  57. 7365892 Add equality operator to int4_t and int2_t. by Quentin Khan · 8 days ago
  58. 225cae5 Merge pull request #10700 from aizu-m:fully-connected-input-channel-mismatch by XNNPACK Team · 8 days ago
  59. 1890ca2 Merge pull request #10677 from JakeStevens:avx2_bf16_qb4w_gemm by XNNPACK Team · 8 days ago
  60. 5b5b4a3 Relax tolerance of sum of squares in YNNPACK by Dillon Sharlet · 8 days ago
  61. f45b3ea Fix int8 -> uint8 rewrite for dots by Dillon Sharlet · 9 days ago
  62. 4366480 Optimize SIMD sub-byte interleave wrappers. by Dillon Sharlet · 9 days ago
  63. 7e739a4 Add 2- and 4-bit 8-way interleaving kernels by Dillon Sharlet · 10 days ago
  64. ba90c65 Add benchmarks for various attention cases using ynnpack API to define the graphs. by Volodymyr Kysenko · 11 days ago
  65. d0133af Add x86 AVX2/FMA3 FP32 GEMV-M dot kernels. by Volodymyr Kysenko · 11 days ago
  66. be502dd Fix allocation behaviour in OwningCpuBuffer::(Copy|Transform). by Quentin Khan · 11 days ago
  67. ae5e2ea Suppress linter warning for TensorHandle error status implicit construction. by Quentin Khan · 11 days ago
  68. efa2e75 Expose native FP16 support query in XNNPACK. by Misha Gutman · 11 days ago
  69. a3fc80e reject input channel mismatch in fully-connected reshape by Aizal Khan · 11 days ago
  70. 9f2ea13 Add :math dep by Gregory Comer · 12 days ago
  71. 17b2392 Fix bug where static tensors are incorrectly deduplicated by Dillon Sharlet · 12 days ago
  72. cb9c05a Update GN build and fix bf16 sqrt tolerance by Gregory Comer · 12 days ago
  73. e6f7ecf [RVV] add rvv f32 kernel for ppmm by velonica0 · 12 days ago
  74. 47f738f Fix wasm undeclared identifier 'XNN_SIMD_NUM_RCP_ITER_F32' by Frank Barchard · 13 days ago
  75. 203a0ad Fix SEH access violation in static broadcast optimization by allocating XNN_EXTRA_BYTES. by Frank Barchard · 13 days ago
  76. 316f4cf Merge pull request #10682 from velonica0:rvv-ibilinear-new by XNNPACK Team · 13 days ago
  77. 303b083 Overload `XnnpackRunner::(Set|Write)Input` for contiguous data containers. by Quentin Khan · 13 days ago
  78. 4906b93 Use transpose kernels for all transpose ops by Dillon Sharlet · 13 days ago
  79. 83b8931 Add transpose subgraph benchmark by Dillon Sharlet · 13 days ago
  80. dfde8dc [RVV] add rvv f32 kernel for ibilinear by velonica0 · 13 days ago
  81. b3d2c9a Merge pull request #10656 from yolanda15:update_f32_gemm by XNNPACK Team · 13 days ago
  82. c7361e4 Add a memory bandwidth measurement to transpose benchmarks by Dillon Sharlet · 13 days ago
  83. 1db304f Refactor LiteRT backend numerical tests into a unified suite. by Ping Yu · 13 days ago
  84. a02371b Fix spurious log messages by Dillon Sharlet · 14 days ago
  85. 49bd3ab Merge branch 'google:master' into avx2_bf16_qb4w_gemm by Jacob Stevens · 14 days ago
  86. 8677691 Add support for avx2 qd8_bf16_qb4w gemm by Jake Stevens · 14 days ago
  87. d9ff5bf Add T=1 case to the attention benchamrk. by Volodymyr Kysenko · 14 days ago chromium/7940 chromium/7941 chromium/7942 chromium/7943 chromium/7944 chromium/7945 chromium/7946 chromium/7947 chromium/7948 chromium/7949 chromium/7950 chromium/7951 chromium/7952 chromium/7953 chromium/7954 chromium/7955 chromium/7956
  88. 8c8ec0c Enable fully connected to work directly with 2/4-bit weights in YNNPACK by Dillon Sharlet · 14 days ago
  89. 440e891 Rewrite int8 dynamic quantization to uint8 in ynn_define_dot. by Dillon Sharlet · 14 days ago
  90. 7db6b89 Fix gn build configuration by Dillon Sharlet · 14 days ago
  91. 13fffc0 Move and generalize bf16 dot rewrite to composites library by Dillon Sharlet · 14 days ago
  92. 5b65190 Merge pull request #10602 from JakeStevens:neon_w_bf16_gemm by XNNPACK Team · 14 days ago
  93. ceb6983 Move and generalize bf16 dot rewrite to composites library by XNNPACK Team · 2 weeks ago
  94. 4a87378 Move and generalize bf16 dot rewrite to composites library by Dillon Sharlet · 2 weeks ago
  95. 59684e5 Add add/subtract kernels for fp32 for hexagon by Alexander Shaposhnikov · 2 weeks ago
  96. 3c8854b Add support for neon dot qd8_bf16_qb4w by Jake Stevens · 2 weeks ago
  97. 16cf671 Add int2 and int4 dot kernels for x86 and ARM by Dillon Sharlet · 2 weeks ago
  98. 1101499 Fix overzealous split_dim test for YNNPACK by Dillon Sharlet · 2 weeks ago
  99. 21b3c02 Merge pull request #10473 from aizu-m:split-dims-axis-bound by XNNPACK Team · 2 weeks ago
  100. 355e0ff Add support for promoting sub-byte input types in reduction subgraph. by Dillon Sharlet · 2 weeks ago