255607e7 · Add pointer types to WASM Assembler DSL · Aug 02, 2023

ae311b91 · F32-GEMM avx broadcast use python to remove + 0 from offsets · Aug 20, 2023

e2ab2af7 · Switch the RoPE operator and node to NTC layout · Aug 22, 2023

294f43c1 · QS8 scalar GEMM template support unrolled microkernels · Aug 26, 2023

0d29c8f5 · Rename X8 scalar gemm microkernels with u1 suffix · Aug 29, 2023

7ffb8b60 · Rename xnn_qd8_f32_qc4w_gemm_minmax_ukernel_fn and xnn_qd8_f32_qc8w_gemm_minmax_ukernel_fn · Sep 01, 2023

5b1f4ff3 · Generate neondot qc4w benchmarks. · Sep 14, 2023

0ee6e4d7 · Delegate QU8 average pooling to XNNPACK · Sep 18, 2023

1075bf71 · Average pool subgraph supports QU8 · Sep 18, 2023

c18f1b4e · Call xnnpack transpose from TfLite transpose and remove old optimized implementation · Sep 19, 2023

d808d4fc · QS8 AVX2 broadcast unroll loads before doing cvt · Sep 19, 2023

3c895fa2 · QS8 AVX2 broadcast reorder input and weight loads before conversions · Sep 19, 2023

081c1cef · Fix a crash on internal benchmark with relaxedsimd · Oct 09, 2023

ef97095a · Remove reversed store from QD8 F16 QC4/8W I8MM · Nov 01, 2023

c3d1440c · Enable F32 NR=16 IGEMM NEON microkernels · Nov 07, 2023

5a742d72 · XNNPACK dynamic subgraph first try · Nov 16, 2023

da971e09 · Add QC8 VNNI GEMM microkernels · Nov 26, 2023

2247de64 · vpshufb to shift bytes left by 4 for VNNI microkernels · Dec 01, 2023

bacdf157 · Enable use of dependencies in arbitrary locations · Dec 13, 2023

655378a5 · Shape inference for Add · Dec 13, 2023