James Newling
James Newling
WIP -- will be split into 2 parts. There are no significant functional change to the lowering in this PR. There are some extremely minor improvements in the generated tensor...
In particular I'd like to load an .onnx file in C++. Thanks.
iree-amd-aie would like to run large matmul tests in CI. To make this feasible (not too slow), the calculation of the correct values on CPU needs to be significantly faster....
I am working on improving the performance of the baseline matmul for testing (it's currently unreasonably slow) and just want to check that the current logic is definitely what you...