1. 46fb807 4x8 A53 GEMM, and GEMMINC unpipelined microkernels. by Frank Barchard · 4 years, 10 months ago
  2. cd945c6 Re-enable swizzle GEMM/IGEMM micro-kernels in WAsm SIMD on ARM by Marat Dukhan · 4 years, 10 months ago
  3. f753a7d Rename BUILD to BUILD.bazel by Marat Dukhan · 4 years, 10 months ago
  4. c4ae7de Propagate IGEMM SR argument to weights packing in Deconvolution operator by Marat Dukhan · 4 years, 10 months ago
  5. c6afd9b Add blocked scalar spmm kernels. by Erich Elsen · 4 years, 10 months ago
  6. 8440fde Support TF-style SAME padding via explicit flag by Marat Dukhan · 4 years, 10 months ago
  7. bff791e Use 8x1 SpMM micro-kernel on WebAssembly by Marat Dukhan · 4 years, 10 months ago
  8. 32c74f7 Fix xnn_f32_gavgpool_spchw_ukernel__scalar_x1 test cases by Marat Dukhan · 4 years, 10 months ago
  9. 14fe0b2 Enable sparse MobileNet v1/v2 operators on WebAssembly by Marat Dukhan · 4 years, 10 months ago
  10. a7fb855 6x8 A53 GEMM, GEMMINC and IGEMM unpipelined microkernels. by Frank Barchard · 4 years, 10 months ago
  11. 563df5f Add scalar version of hwc2spchw convolution. by Erich Elsen · 4 years, 10 months ago
  12. 98ba441 Vectorized extexp functions by Marat Dukhan · 4 years, 10 months ago
  13. cb80197 Disable GEMM/IGEMM micro-kernels with swizzle by Marat Dukhan · 4 years, 10 months ago
  14. 4232323 Unify naming of functions in benchmark::utils:: by Marat Dukhan · 4 years, 10 months ago
  15. 31a98d7 Remove warnings about inefficient padding parameters in Convolution by Marat Dukhan · 4 years, 10 months ago
  16. 1756f9e Propagate GEMM/IGEMM SR argument to weights packing in Fully Connected operator by Marat Dukhan · 4 years, 10 months ago
  17. e0df831 Remove trailing whitespace by Marat Dukhan · 4 years, 10 months ago
  18. 07cb676 Refactor initialization of even/odd masks in parameters for SpCHW micro-kernels by Marat Dukhan · 4 years, 10 months ago
  19. 838c8e3 Refactor initialization of masks in parameters for SpCHW micro-kernels by Marat Dukhan · 4 years, 10 months ago
  20. caf8544 LD64/LD128 kernels remove all pushes (d8-d15) Remap d12-d15 to d16-d19 by Frank Barchard · 4 years, 10 months ago
  21. fcfdc0e Automated g4 rollback of changelist 274728310. by Frank Barchard · 4 years, 10 months ago
  22. 05ac8e3 VSCALE microkernel and SoftMax Three-Pass algorithm with Reloading by Marat Dukhan · 4 years, 10 months ago
  23. 4a4a7fa Three-Pass Softargmax benchmark (recomputing version) by Marat Dukhan · 4 years, 10 months ago
  24. 8e3c551 1x8 a53 kernel refactor based on a57. by Frank Barchard · 4 years, 10 months ago
  25. baa9ead Update assembly Copyright notice to // comment by Frank Barchard · 4 years, 10 months ago
  26. 9757953 Refactor and open-source Three-Pass Softmax micro-kernels by Marat Dukhan · 4 years, 10 months ago
  27. 459c9fc 6x8 and a53 kernel comments. by Frank Barchard · 4 years, 10 months ago
  28. 515c977 Refactor and open-source vectorized expminus function by Marat Dukhan · 4 years, 10 months ago
  29. f6839e1 Refactor vectorized exp functions by Marat Dukhan · 4 years, 10 months ago
  30. 2af471b Switch default intrinsics kernel to 6x8 by Frank Barchard · 4 years, 10 months ago
  31. 9cdade3 Add prefetch instructions to 16x1, 16x2, 16x4 kernels. by Erich Elsen · 4 years, 10 months ago
  32. 34dc2c0 Add gavgpool_spchw_scalar__x1 kernel. by Erich Elsen · 4 years, 10 months ago
  33. ac4de80 Add chw 3x3s2p1_scalar kernels. by Erich Elsen · 4 years, 10 months ago
  34. 0cc2c53 add 3x3p1_scalar kernel by Erich Elsen · 4 years, 10 months ago
  35. a5ca10e Neon intrinsics clamping - Replace 2 LD1R with 1 LD2R by Frank Barchard · 4 years, 10 months ago
  36. 6adff4e Vectorized implementations of expf function for AVX2 and AVX512F by Marat Dukhan · 4 years, 10 months ago
  37. bd9e495 Remove 4x12 intrinsics kernels. by Frank Barchard · 4 years, 10 months ago
  38. 7e95597 Pass XNN_ENABLE_ASSEMBLY for all tests and kernel benchmarks by Frank Barchard · 4 years, 10 months ago
  39. 8fe54e4 Extra :xnnpack_operators_nhwc_f32 target with only F32 operators in NHWC layout by Marat Dukhan · 4 years, 10 months ago
  40. 810171d Enable assembly by default. by Frank Barchard · 4 years, 10 months ago
  41. 21be34f 1x8 A53 GEMM, GEMMINC and IGEMM microkernels. by Frank Barchard · 4 years, 10 months ago
  42. db45b6a 1x8 neonfma IGEMM microkernel and 1x8 benchmarks. by Frank Barchard · 4 years, 10 months ago
  43. 174706e Fix misleading comments for debug_build/optimized_build Bazel configs by Marat Dukhan · 4 years, 10 months ago
  44. dbafc58 extend build flag --define=xnn_enable_assembly=true to GEMM and IGEMM benchmarks. by Frank Barchard · 4 years, 10 months ago
  45. 4e0249a Add performance results on MobileNets & Pixel phones by Marat Dukhan · 4 years, 10 months ago
  46. 466b523 Use GEMM/IGEMM micro-kernels with Swizzle on WAsm SIMD by Marat Dukhan · 4 years, 10 months ago
  47. f633c2c Fix Bazel symblinks in .gitignore by Marat Dukhan · 4 years, 10 months ago
  48. 523448a Add .gitignore file by Marat Dukhan · 4 years, 10 months ago
  49. 2dbdc2f CMake build configurations by Marat Dukhan · 4 years, 10 months ago
  50. afbca9a Remove unused x21 and switch x20 to x8 to avoid push. by Frank Barchard · 4 years, 10 months ago
  51. d620972 Group operators in public header by data type by Marat Dukhan · 4 years, 10 months ago
  52. 5609a08 Document xnn_initialize and xnn_deinitialize functions by Marat Dukhan · 4 years, 10 months ago
  53. cf056b2 Hide all emscripten-specific sources behind xnnpack_cc_library rule by Marat Dukhan · 4 years, 10 months ago
  54. 0c57d2a Fix Bazel build for XNNPACK when using emscripten toolchain. by Daniel Smilkov · 4 years, 10 months ago
  55. 1a729ec Internal change by Marat Dukhan · 4 years, 10 months ago
  56. d343c22 Avoid cpuinfo dependency in Emscripten builds by Marat Dukhan · 4 years, 10 months ago
  57. 1aaabb6 Remove unused field _np in GemmMicrokernelTester by Marat Dukhan · 4 years, 10 months ago
  58. 885ca24 Support MacOS build with Bazel by Marat Dukhan · 4 years, 10 months ago
  59. dd69f0b Rename XNN_CONVOLUTION_FLAG_DEPTHWISE to XNN_FLAG_DEPTHWISE_CONVOLUTION by Marat Dukhan · 4 years, 10 months ago
  60. c8e00eb Disable logging in optimized builds, limit logging in fastbuild by Marat Dukhan · 4 years, 10 months ago
  61. 12f1dea Increase static memory in Emscripten benchmarks to 128MB by Marat Dukhan · 4 years, 10 months ago
  62. c068bb6 End-to-end benchmarks on MobileNet v1 and MobileNet v2 models by Marat Dukhan · 4 years, 10 months ago
  63. 4efb351 Document availability of Fully Connected operator by Marat Dukhan · 4 years, 10 months ago
  64. 1b0421b 1x12 and 1x8 assembly kernel cleanup by Frank Barchard · 4 years, 10 months ago
  65. 4e45e66 Support Linux/AArch64 platform by Marat Dukhan · 4 years, 10 months ago
  66. b8ab4cb Fix signed/unsigned mismatch warning in SpMM benchmarks by Marat Dukhan · 4 years, 10 months ago
  67. c6edf92 Fix bug in Fully Connected F32 unit tests by Marat Dukhan · 4 years, 10 months ago
  68. 9d056a4 Compatibility with TF.js WAsm build by Marat Dukhan · 4 years, 10 months ago
  69. 3bcea2b Final touch before open-source release by Marat Dukhan · 4 years, 10 months ago
  70. 08c4a43 Bazel BUILD file for XNNPACK by Marat Dukhan · 4 years, 10 months ago
  71. 452662b Fix XNNPACK build on Mac by Marat Dukhan · 4 years, 10 months ago
  72. f8c8046 Cleanup redundant #includes by Marat Dukhan · 4 years, 11 months ago
  73. 341c321 Remove unused XNN_INTERNAL_EXTRA_BYTES constant by Marat Dukhan · 4 years, 11 months ago
  74. df6985f Map Exynos-M[1-4] micro-architectures to equivalents in OSS cpuinfo by Marat Dukhan · 4 years, 11 months ago
  75. 33f0c7a Guard Ruy and GemmLowp benchmarks by Marat Dukhan · 4 years, 11 months ago
  76. d62f3cc Avoid using cpuinfo_get_max_cache_size() function by Marat Dukhan · 4 years, 11 months ago
  77. 22f38e4 Refactor architecture identification macros by Marat Dukhan · 4 years, 11 months ago
  78. 1dadbf7 Limit direct dependencies on cpuinfo by Marat Dukhan · 4 years, 11 months ago
  79. 629a33e Fix incompatibilities with open-source Bazel-based build by Marat Dukhan · 4 years, 11 months ago
  80. 4c2637d LD2R for loading clamp parameters by Frank Barchard · 4 years, 11 months ago
  81. bb4c18b Report Freq in additional benchmarks by Frank Barchard · 4 years, 11 months ago
  82. 80fc932 Unify comments style by Marat Dukhan · 4 years, 11 months ago
  83. b455b12 Initial open-source release by XNNPACK Team · 4 years, 11 months ago