[Mlir-commits] [mlir] [NVGPU] Add FP8 (e4m3/e5m2) support to nvgpu.mma.sync (PR #207342)

[Kirill Vedernikov]

================
@@ -0,0 +1,12 @@
+// RUN: mlir-opt %s -convert-nvgpu-to-nvvm -split-input-file | FileCheck %s
+
+// Test that FP8 (e4m3) nvgpu.mma.sync lowers to nvvm.mma.sync with the
+// correct multiplicand PTX type.
+func.func @fp8_mma_sync(%arg0: vector<4x4xf8E4M3FN>, %arg1: vector<2x4xf8E4M3FN>, %arg2: vector<2x2xf32>) -> vector<2x2xf32> {
----------------
kvederni wrote:

I like that new mma fp8 tests are added to a separate file. Thanks!
Could you please rename the filename to something like nvgpu-to-nvvm-mma-fp8.mlir and add more combinations here for e5m2 type and shapes?

https://github.com/llvm/llvm-project/pull/207342