oneDNN Files

This is a patch release containing the following changes to v3.9.1: * Fixed correctness issue in int8 convolution on processors with Intel AVX2 and Intel DL Boost instruction set support (a7c4079bfe6aacd997cc18631152845dcf3aa631, 78e781f6bbec4cf622a23426830ae4b531432295) * Fixed performance degradation for f32 convolution primitive on processors with Intel AVX-512 instruction set support (74f23b48e0779e5e39ea14ebdb80eca9afdb92aa) * Fixed performance regression for RNN primitive with LBR GRU cell type on Intel Arc GPUs (ae2844e6ffbb71f27d28fd30ea7686acecea8dd6) * Fixed performance degradation for int8 convolution primitive when using zero points (dbb848437d2f93b3e8a5b407a5e9cc2f266fefd7) * Fixed segmentation fault in matmul primitive when using ONEDNN_VERBOSE=all (7310aa249749f82fcfc77df5e30902c37e150bfb) * Fixed correctness issue in multi-dimensional matmul primitive on Intel Xeon processors with Intel AMX instruction set support (formerly Sapphire Rapids and Granite Rapids) (642d18b7fe4648e8aa8390c580d6fa6e2d8771a6) * Reduced problem size in test_sdpa_decomp test (9bff06eaff33c3e3129fa5f577d2a56732113789) * Restricted test_sdpa_decomp and test_mqa_decomp tests to OMP or THREADPOOL CPU runtimes (3cd917075c92d05a95993bfea23f434c9beb1882) * Fixed illegal instruction issue in pooling primitive on processors with Intel SSE4.1 support (d907c47ae77035da420fd6ef37b9184629fecfb0) * Fixed segmentation fault issue in f16 backward convolution primitive on processors with Intel AVX2 with Intel DL Boost with float16 and bfloat16 support (50cc2282324d2527e04c3f0f7b9c2a36e53f5e0a, fcc7e5ebc8dd2274f7a996c70c07e8db3aa685c4) * Restored support for int8 matmul with per_oc scales and zero points on Intel Arc GPUs (1a5a454270d9bcfe580228c67f348a24e2727872, 04c22c96de0776a76d86ef283fcd37eed30dcbb4)