[llvm] [AMDGPU] Fix SDWA commuting (PR #106920)

[Yaxun Liu via llvm-commits]

================
@@ -0,0 +1,17 @@
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1030 -run-pass=machine-cse -verify-machineinstrs %s -o - 2>&1 | FileCheck --check-prefix=GCN %s
+
+# GCN-LABEL: name: test_machine_cse_op_sel
+# GCN: %2:vgpr_32 = V_ADD_NC_U16_e64 0, %0, 0, %1, 1, 0, implicit $mode, implicit $exec
+# GCN: %3:vgpr_32 = V_ADD_NC_U16_e64 0, %1, 0, %0, 1, 0, implicit $mode, implicit $exec
----------------
yxsamliu wrote:

Can it be done with separate PR? This PR is for SDWA commute issue which causes rocFFT to fail. Whereas VOP3 not commute is missed performance opportunity. The current issue is kind of urgent.

https://github.com/llvm/llvm-project/pull/106920