- f9fdaa7 Disable applying XNNPACK delegate by default when benchmarking tflite inside XNNPACK/bench because the intention of those benchmarks is to compare the XNNPACK execution path against the original tflite CPU one. by Chao Mei · 3 years, 1 month ago
- d713e8a Refactor microbenchmarks by Marat Dukhan · 3 years, 6 months ago
- 16f1e1a QS8 version of NHWC Convolution operator by Marat Dukhan · 3 years, 10 months ago
- ecd8311 Rename s8rng/s32rng -> i8rng/i32rng by Marat Dukhan · 3 years, 10 months ago
- 44f0ca7 Bind RNG by reference in microbenchmarks by Marat Dukhan · 3 years, 11 months ago
- c664027 Use proper macros to guard code segment to prevent compilation errors. by Chao Mei · 3 years, 11 months ago
- 08b7a97 Rename Q8 microkernels and operators to QU8 by Marat Dukhan · 4 years ago
- 49b4dcc FP16 Convolution NHWC operator by Frank Barchard · 4 years ago
- d986004 Remove cpuinfo as dependency for operator benchmarks by Frank Barchard · 4 years ago
- 5ce30d9 Work around non-standard std::uniform_int_distribution<uint8_t> by Marat Dukhan · 4 years, 2 months ago
- 04f03be Support overriding memory allocation functions by Marat Dukhan · 4 years, 7 months ago
- 4232323 Unify naming of functions in benchmark::utils:: by Marat Dukhan · 4 years, 8 months ago
- 08c4a43 Bazel BUILD file for XNNPACK by Marat Dukhan · 4 years, 9 months ago
- d62f3cc Avoid using cpuinfo_get_max_cache_size() function by Marat Dukhan · 4 years, 9 months ago
- bb4c18b Report Freq in additional benchmarks by Frank Barchard · 4 years, 9 months ago
- b455b12 Initial open-source release by XNNPACK Team · 4 years, 9 months ago