[llvm] [AMDGPU] Optimize LDS DMA soft waitcnt (PR #138802)

[Jay Foad via llvm-commits]

================
@@ -1608,6 +1608,7 @@ let OtherPredicates = [HasImageInsts] in {
   def S_WAIT_DSCNT_soft : SOPP_Pseudo <"s_soft_wait_dscnt", (ins s16imm:$simm16), "$simm16">;
   def S_WAIT_KMCNT_soft : SOPP_Pseudo <"s_soft_wait_kmcnt", (ins s16imm:$simm16), "$simm16">;
 }
+def S_WAITCNT_VMCNT_LDS_DMA_soft : SOPP_Pseudo <"s_soft_waitcnt" , (ins SWaitCnt:$simm16), "$simm16">;
----------------
jayfoad wrote:

Can we avoid baking "DMA" into the name? I don't think that terminology has been used since GFX9. Later documents just talk about "load to LDS".

https://github.com/llvm/llvm-project/pull/138802