[Mlir-commits] [mlir] [MLIR][AMDGPU] Add a wrapper for global LDS load intrinsics in AMDGPU (PR #133498)

Alan Li llvmlistbot at llvm.org
Wed Apr 2 17:53:19 PDT 2025


================
@@ -112,21 +114,30 @@ LogicalResult FatRawBufferCastOp::verify() {
   return success();
 }
 
+static bool hasGlobalMemorySpace(Attribute memorySpace) {
+  if (!memorySpace)
+    return true;
+  else if (auto intMemorySpace = llvm::dyn_cast<IntegerAttr>(memorySpace))
+    return intMemorySpace.getInt() == 0 || intMemorySpace.getInt() == 1;
+  else if (auto gpuMemorySpace =
+               llvm::dyn_cast<gpu::AddressSpaceAttr>(memorySpace))
+    return gpuMemorySpace.getValue() == gpu::AddressSpace::Global;
+  return false;
+}
+
+static bool hasWorkgroupMemorySpace(Attribute memorySpace) {
+  if (auto intMemorySpace = llvm::dyn_cast<IntegerAttr>(memorySpace))
+    return intMemorySpace.getInt() == 3;
----------------
lialan wrote:

done. 

https://github.com/llvm/llvm-project/pull/133498


More information about the Mlir-commits mailing list