[Mlir-commits] [clang] [llvm] [mlir] [AMDGPU] Improve handling of numRecords for make.buffer.rsrc (PR #213180)

[Matt Arsenault]

================
@@ -560,6 +561,7 @@ static void fillAMDGCNFeatureMap(StringRef GPU, const Triple &T,
     Features["gfx11-insts"] = true;
     Features["gfx12-insts"] = true;
     Features["gfx1250-insts"] = true;
+    Features["45-bit-num-records-buffer-resource"] = true;
----------------
arsenm wrote:

Don't add a new unstructured feature here: https://github.com/llvm/llvm-project/pull/212945

https://github.com/llvm/llvm-project/pull/213180