Shu Chen

@feixuedudiao First of all, it doesn't make sense to compare the performance of primitives on CPU and GPU without considering the GPU hardware capabilities and configurations. For your case, if...

Thanks for reviewing. I have updated the commit messages and also added reference to those examples in the documentation.

Hello @vineel96, the threading backend is common for all oneDNN algorithms/kernels and is the same for x86_64/aarch64 architecture. The source code of threading implementation is in [oneDNN/src/common/dnnl_thread.hpp](https://github.com/oneapi-src/oneDNN/blob/main/src/common/dnnl_thread.hpp). For example, in...

Based on @dzarukin 's comment in https://github.com/oneapi-src/oneDNN/pull/1440#discussion_r1318897877, the documentation for batch normalization shall be updated. Since it hasn't been done yet, I have created an internal Jira ticket to track...

Hi @renato-arantes, do you mean quantized to `s8` or `s32`? The accumulation datatype used during `Op` computation is governed by the [`accumulation_mode`](https://oneapi-src.github.io/oneDNN/dev_guide_attributes_accumulation_mode.html) attribute of the primitive. By default, f32 is...

> By inspect [here](https://github.com/oneapi-src/oneDNN/blob/cbca0823cc5430d5faf8ea8c8423f0c3bf9c86b8/src/cpu/gemm_x8s8s32x_convolution_utils.cpp#L117) I can see that the bias is by default f32 but in the documentation [here](https://oneapi-src.github.io/oneDNN/page_cnn_inference_int8_cpp.html#doxid-cnn-inference-int8-cpp) the quantized bias is s32. Hi @renato-arantes, the second [here](https://oneapi-src.github.io/oneDNN/page_cnn_inference_int8_cpp.html#doxid-cnn-inference-int8-cpp) is...

> llvm-foreach: Segmentation fault (core dumped) > clang-15: error: ptxas command failed with exit code 254 (use -v to see invocation) @wangzy0327 The issue is more likely in the compiler...

@wangzy0327 I meant that the core dump happens in the compiler and for CUDA backend, not in oneDNN. From the log, the compilation of oneDNN has completed, and the compiler...

@nwnk Thank you for reporting the issue. I can reproduce most of the test failures locally. After adding [ONEDNN_VERBOSE=all](https://oneapi-src.github.io/oneDNN/dev_guide_verbose.html) in the environment, the detailed log shows that some problem cases...

Hello @uxlfoundation/onednn-arch, Could you please review the updated example, which now includes backward propagation for the vanilla RNN?