This PR adds the benchmark csv files for gemm and gemm_batched_strided for some of the latest transformer based language models. The oneDNN matmul inputs for language models are used as reference for these csv files. The reference files can be found here: https://github.com/oneapi-src/oneDNN/tree/master/tests/benchdnn/inputs/matmul
For all cases, the oneDNN inputs use bias as an extra input to gemm but I have omitted this argument from the csv files as this argument is not yet supported in portBLAS gemm.
This PR adds the benchmark csv files for
gemmandgemm_batched_stridedfor some of the latesttransformerbased language models. The oneDNN matmul inputs for language models are used as reference for these csv files. The reference files can be found here: https://github.com/oneapi-src/oneDNN/tree/master/tests/benchdnn/inputs/matmulFor all cases, the oneDNN inputs use
biasas an extra input togemmbut I have omitted this argument from the csv files as this argument is not yet supported in portBLASgemm.