Svetlozar Georgiev
Svetlozar Georgiev
# Description Please include a summary of the change. Please also include relevant motivation and context. See [contribution guidelines](https://github.com/oneapi-src/oneDNN/blob/master/CONTRIBUTING.md) for more details. If the change fixes an issue not documented...
# Description This MR slightly refactors the eltwise SYCL kernel to remove some unused data that was stored in the conf struct passed to the kernel. As a result, the...
This PR skips the unsupported examples on generic vendor and fixes an issue in deconvolution where problems post-ops were incorrectly returning unimplemented (and as a result the deconvolution example was...
# Description This PR fixes several gtest issues on generic vendor: - `NDR.LocalSize[0]==0` assertion was being thrown due to problems with 0-dim input/outputs not handled correctly - Several unsupported cases...
# Description This is a proposal for dnncompat - oneDNN compatibility layer. Rendered document can be seen [here](https://github.com/sgeor255/oneDNN/blob/dd7b4f30ac43f0baa0951b9ca30f7e083ad9bc31/rfcs/20240919-dnncompat/README.md).
This PR cleans up and improves some oneDNN-related code: - Use user-scratchpad mode when creating matmul primitives to avoid allocations during execution - Clean up the cmake configuration for finding...
This PR enables reorder optimization for Q4_K layout similarly to https://github.com/ggml-org/llama.cpp/pull/12858 . This branch is based off of @Alcpz 's and before that is merged the easiest way to review...
I would like to get some L1 cache metrics (ideally L1 cache hit rate) of a SYCL application on a lunar lake GPU. Currently I don't think this is possible...