3d01cf78 · Fix missing `#include`s in `XNNPACK/test` subdirectory. · Apr 16, 2024

c6131ce2 · Rsum ukernels accumulate into output. · Apr 16, 2024

6ff7b9ec · Mean op can handle arbitrary reduction axis · Apr 16, 2024

d427ff73 · F16 GEMM use F16-FP32ACC for improved performance · Apr 16, 2024

15c2312b · Don't attempt to call `cpuinfo_initialize()` unless `XNN_ENABLE_CPUINFO` is enabled. · Apr 15, 2024

52df8e54 · AMX QD8_F32_QC8W GEMM generate all tile sizes · Apr 14, 2024

8a7896ef · Change AMX k-block from 64 to 4 for faster testing. · Apr 13, 2024

ea546fa9 · Add `WAsm SIMD` microkernel for `f32-rsqrt`. · Apr 12, 2024

35c99f21 · When no weight cache is provided to XNNPack, create one to share packed weights between operations. · Apr 12, 2024

887c3d8b · Exported helper functions for transposition normalization. · Apr 11, 2024

240bd607 · Enable AVX512 and AVX2 F32_RADDSTOREEXPMINUSMAX microkernels · Apr 09, 2024

923f4df9 · Add iterative `vsqrt` microkernels for `x86_64`, which computes `x*rsqrt(x)`, i.e. · Apr 05, 2024

a9074a0d · F32 & F16 rsum microkernels return the sum instead of taking it as a parameter. · Apr 04, 2024

d3557f75 · Enable QD8 AMX 16x16c4 GEMM and VNNI 16x16c4 IGEMM microkernel · Mar 31, 2024

3e0c4e96 · Disable GCC version of AMX GEMM microkernel · Mar 31, 2024

4481a4df · QU8 remove udot detect and linux kernel version · Mar 29, 2024

4e35f29a · AMX GEMM support 32 bit x86 · Mar 27, 2024

75350a1e · Rollback of github.com/google/XNNPACK/pull/6214. · Mar 26, 2024

34858da7 · Add `qd8_f32_qc8w` to the `batch_matrix_multiply` operator. · Mar 26, 2024