[GPU][NFC] Follow the official convention to define mfma/wmma attributes

[hanhanW]

The LLVM intrinsics and official docs are all using [output_type]_MxNxK_[input_type] format. The revision updates IREE's definitions to follow the convention.

Some examples from official docs:

• https://gpuopen.com/learn/wmma_on_rdna3/
• https://gpuopen.com/learn/amd-lab-notes/amd-lab-notes-matrix-cores-readme/
• https://github.com/ROCm/amd_matrix_instruction_calculator

This patch is generated by the below commands:

cd compiler
sed -i "s/MFMA_F16_16x16x16_F32/MFMA_F32_16x16x16_F16/g" **/*.mlir
sed -i "s/MFMA_F16_16x16x16_F32/MFMA_F32_16x16x16_F16/g" **/*.td
sed -i "s/MFMA_F16_16x16x16_F32/MFMA_F32_16x16x16_F16/g" **/*.cpp
sed -i "s/MFMA_F16_16x16x16_F32/MFMA_F32_16x16x16_F16/g" **/*.h
sed -i "s/MFMA_F16_32x32x8_F32/MFMA_F32_32x32x8_F16/g" **/*.mlir
sed -i "s/MFMA_F16_32x32x8_F32/MFMA_F32_32x32x8_F16/g" **/*.td
sed -i "s/MFMA_F16_32x32x8_F32/MFMA_F32_32x32x8_F16/g" **/*.h
sed -i "s/MFMA_F16_32x32x8_F32/MFMA_F32_32x32x8_F16/g" **/*.cpp
sed -i "s/MFMA_F8E4M3FNUZ_16x16x32_F32/MFMA_F32_16x16x32_F8E4M3FNUZ/g" **/*.mlir
sed -i "s/MFMA_F8E4M3FNUZ_16x16x32_F32/MFMA_F32_16x16x32_F8E4M3FNUZ/g" **/*.td
sed -i "s/MFMA_F8E4M3FNUZ_16x16x32_F32/MFMA_F32_16x16x32_F8E4M3FNUZ/g" **/*.h
sed -i "s/MFMA_F8E4M3FNUZ_16x16x32_F32/MFMA_F32_16x16x32_F8E4M3FNUZ/g" **/*.cpp
sed -i "s/MFMA_I8_16x16x32_I32/MFMA_I32_16x16x32_I8/g" **/*.mlir
sed -i "s/MFMA_I8_16x16x32_I32/MFMA_I32_16x16x32_I8/g" **/*.td
sed -i "s/MFMA_I8_16x16x32_I32/MFMA_I32_16x16x32_I8/g" **/*.h
sed -i "s/MFMA_I8_16x16x32_I32/MFMA_I32_16x16x32_I8/g" **/*.cpp
sed -i "s/MFMA_I8_32x32x16_I32/MFMA_I32_32x32x16_I8/g" **/*.mlir
sed -i "s/MFMA_I8_32x32x16_I32/MFMA_I32_32x32x16_I8/g" **/*.td
sed -i "s/MFMA_I8_32x32x16_I32/MFMA_I32_32x32x16_I8/g" **/*.h
sed -i "s/MFMA_I8_32x32x16_I32/MFMA_I32_32x32x16_I8/g" **/*.cpp

[antiagainst]

thanks @hanhanW! i was a bit wanting to fix inconsistency before..