[llvm-branch-commits] [llvm] [AMDGPU][SILoadStoreOptimizer] Merge constrained sloads (PR #96162)

[Christudasan Devadasan via llvm-branch-commits]

================
@@ -1701,17 +1732,33 @@ unsigned SILoadStoreOptimizer::getNewOpcode(const CombineInfo &CI,
       return AMDGPU::S_BUFFER_LOAD_DWORDX8_SGPR_IMM;
     }
   case S_LOAD_IMM:
-    switch (Width) {
-    default:
-      return 0;
-    case 2:
-      return AMDGPU::S_LOAD_DWORDX2_IMM;
-    case 3:
-      return AMDGPU::S_LOAD_DWORDX3_IMM;
-    case 4:
-      return AMDGPU::S_LOAD_DWORDX4_IMM;
-    case 8:
-      return AMDGPU::S_LOAD_DWORDX8_IMM;
+    // For targets that support XNACK replay, use the constrained load opcode.
+    if (STI && STI->hasXnackReplay()) {
+      switch (Width) {
----------------
cdevadas wrote:

I guess currently the merged load is always under-aligned. While combining the MMOs, currently the alignment is picked from the first MMO and that'd definitely be smaller than the natural align requirement for the new load. That's one reason I conservatively want to emit _ec equivalent when XNACK is enabled.

https://github.com/llvm/llvm-project/pull/96162