[llvm] AMDGPU: Correct cycle counts for f64 mfma on gfx940 (PR #83782)

Stanislav Mekhanoshin via llvm-commits llvm-commits at lists.llvm.org
Mon Mar 4 11:36:02 PST 2024


================
@@ -804,7 +803,7 @@ body:             |
 # GCN-LABEL: name: dmfma16x16_write_vgpr_valu_read
 # GCN:      V_MFMA
 # GCN-NEXT: S_NOP 7
-# GCN-NEXT: S_NOP 2
+# GCN-NEXT: S_NOP 1
----------------
rampitec wrote:

And this is still 11 waitstates.

https://github.com/llvm/llvm-project/pull/83782


More information about the llvm-commits mailing list