oneDNN v3.8 release notes

RELEASE_NOTES.md

@@ -0,0 +1,72 @@
+# Performance Optimizations
+## Intel Architecture Processors
+* Improved matmul and inner product primitives performance on Intel Xeon processors with support for the Intel AMX instruction set.
+* Improved performance of convolution and inner product primitives on processors with Intel AVX2 instruction set support.
+* Improved performance of `int8` convolution support with zero points.
+* Improved `fp32` convolution performance with `fp16` and `bf16` compressed weights on processors with Intel AVX2 and AVX512 instruction set support.
+* Improved `fp16`/`bf16` depthwise convolution performance with `fp32` bias or `sum` post-ops or dilation.
+* Improved `bf16` pooling backpropagation performance.
+* Improved binary post-ops performance with `per_w` broadcast.
+
+## Intel Graphics Products
+* Improved performance on Intel GPUs based on Xe3 architecture.
+* Improved convolution performance on:
+ * Intel Arc Graphics for Intel Core Ultra (Series 2, formerly Lunar Lake).
+ * Intel Arc B-series discrete graphics (formerly Battlemage).
+* Improved `int8` matmul performance with zero-points support for source and weight tensors.
+* Improved matmul and reorder performance for 4-bit floating-point data types `f4_e2m1` and `f4_e3m0`. Compute primitives provide support through internal converison into f16 as current Intel GPUs lack native support.
+* Improved performance of the following subgraphs with Graph API:
+ * Scaled Dot Product Attention (SDPA) with `int4` and `int8` KV cache.
+ * SDPA with bottom-right implicit causal mask.
+ * SDPA with head size 512 and 576.
+ * Grouped Query Attention (GQA) with 5D input tensors.
+
+## AArch64-based Processors
+* Improved `fp16` reorder performance.
+* Improved `int8` matmul performance.
+* Improved `bf16` inner product forward propagation performance with Arm Compute Library (ACL).
+* Improved convolution performance on processors with SVE support with Arm Compute Library (ACL).
+
+# Functionality
+## Intel Graphics Products
+* Introduced support for the [GenIndex](https://oneapi-src.github.io/oneDNN/v3.8/dev_guide_op_genindex.html) operation in Graph API.
+* Introduced select algorithm support in [binary primitive](https://uxlfoundation.github.io/oneDNN/v3.8/dev_guide_binary.html). The functionality is optimized for Intel GPUs.
+* Introduced optimized support for 4-bit floating-point data types `f4_e2m1` and `f4_e3m0` in convolution on Intel(R) Data Center GPU Max Series or newer Intel GPUs.
+* Extended support for 4-bit floating-point data types in matmul and reorder.
+
+## Intel Architecture Processors
+* Introduced support for `f32` convolution with `fp16` compressed weights.
+* Enabled `int8`/`int4` compressed weight support in matmul primitive on Intel(R) CPUs. 
+
+## Generic GPU Vendor
+* Introduced support for:
+ * Vanilla RNN forward propagation
+ * Inner product backpropagation
+ * Group normalization
+* Improved accuracy of inner product primitive with sum post-ops for large shapes.
+
+## NVIDIA GPUs
+* Introduced Graph API support.
+
+# Usability
+
+* Added support for Group Normalization primitive with [`ONEDNN_ENABLE_PRIMITIVE`](https://uxlfoundation.github.io/oneDNN/dev_guide_build_options.html#onednn-enable-primitive) build option.
+* Enabled support for ROCm 6 on AMD GPUs.
+* Improved CMake integration for oneDNN installation with Nvidia backend enabled.
+
+## AArch64-based Processors
+ * Default thread count to maxin `acl_threadpool` to prevent crashes in Tensorflow.
+ * Fixed scratchpad being ignored for some GEMMs. Reduces memory and speeds up execution.
+ * Fixed a bug in `fp32` reorders where ACL returned incorrect results.
+
+# Validation
+* Added benchdnn option [`--execution-mode`](https://github.com/uxlfoundation/oneDNN/blob/rls-v3.8/tests/benchdnn/doc/knobs_common.md#--execution-mode) to test oneDNN functionality with SYCL Graph record/execute mode.
+* Extended benchdnn option [`--cold-cache`](https://github.com/uxlfoundation/oneDNN/blob/main/tests/benchdnn/doc/knob_cold_cache.md) with support for cold TLB mode.
+* Added benchdnn option `--bia-dt` to control bias data type for matmul, inner product, convolution, and deconvolution.
+* Extended syntax of benchdnn `--dt` option in [Graph API driver](https://github.com/uxlfoundation/oneDNN/blob/main/tests/benchdnn/doc/driver_graph.md) to manage data types of individual tensors in a pattern.
+
+# Breaking Changes
+* Removed the experimental [Graph Compiler](https://uxlfoundation.github.io/oneDNN/v3.7/dev_guide_graph_compiler.html) backend.
+
+# Thanks to Contributors
+This release contains contributions from the project core team as well as Alexander Simonov @asimonov1, Denis @redradist, Dmitriy Ovchinnikov @inteldimitrius, Eliezer Weissmann @eliezerweissmann, Hubert Maciak @hmaciak, Ilya Lavrenov @ilya-lavrenov, James McGregor @Jmc18134, Marek Michalowski @michalowski-arm, Maria Zhukova @mzhukova, Orel Yehuda @yehudaorel, Ravi Pushkar @rpushkarr, Renato Barros Arantes @renato-arantes, Shreyas-fuj @Shreyas-fuj, Shu Chen @shu1chen, Viktoriia Gvozdeva @vgvozdeva, Yair Obodovsky @yair-obodovsky, hmaciak @hmaciak, jstachowintel @jstachowintel, zhangfei @zhangfeiv0, James McGregor @Jmc18134, Marek Michalowski @michalowski-arm, Renato Barros Arantes @renato-arantes.