nyck33/mlirx-wmma

Attempt at replicating the paper described in this pdf:

• https://mlir.llvm.org/OpenMeetings/2021-08-26-High-Performance-GPU-Tensor-CoreCode-Generation-for-Matmul-Using-MLIR.pdf

MLIRX repo used

• https://github.com/polymage-labs/mlirx/tree/120a9c93a3286c6745f3a824dd73521cd7a18dab

Their benchmark codes

• https://github.com/polymage-labs/mlirx/tree/120a9c93a3286c6745f3a824dd73521cd7a18dab/mlir/benchmark

Cuda Device with no tesnor core local

My machine has a GTX1650 which does not have tensor cores so I need to try to replicate on Google Colab, Kaggle or other cloud service, ie. GCP, Azure, AWS, etc.