[llvm] [NVPTX] Add prefetch tensormap variant (PR #146203)

Alex MacLean via llvm-commits llvm-commits at lists.llvm.org
Mon Jul 7 08:33:47 PDT 2025


================
@@ -587,6 +588,9 @@ Value *NVPTXTTIImpl::rewriteIntrinsicWithAddressSpace(IntrinsicInst *II,
       return ConstantInt::get(II->getType(), *R);
     return nullptr;
   }
+  case Intrinsic::nvvm_prefetch_tensormap: {
+    return nullptr;
----------------
AlexMaclean wrote:

In order for this to be useful we need to actually rewrite the intrinsic here. That is, create an `nvvm_prefetch_tensormap` intrinsic with NewV. It would also be good to add a couple tests to confirm this works as expected. 

https://github.com/llvm/llvm-project/pull/146203


More information about the llvm-commits mailing list