Lixun Zhang 02a2f24dd5 [Backend] Refactor mfma selection (#441) ...

* Select mfma dimensions and instruction from static table

* Extend mfmaLayout to include version and instrShape

* Simplify generateMFMAOp by searching the mfma instruction in the table

* Fix getNonKDim() and non_k_dim

* Break instrShape into MDim and NDim

2024-01-16 21:05:35 -06:00

..

SharedToDotOperandFMA.cpp

[BACKEND] Fix BF16 dot operand type mismatch (#2162)

2023-09-05 20:46:31 +00:00

SharedToDotOperandMFMA.cpp

[Backend] Refactor mfma selection (#441)

2024-01-16 21:05:35 -06:00

SharedToDotOperandMMAv1.cpp

[BACKEND] cleaner V100 tensor core packing (#2222)

2023-08-31 14:00:51 -07:00

SharedToDotOperandMMAv2.cpp

[BACKEND] Fix matmul downcast path (#2528)

2023-10-26 09:43:49 -04:00