[llvm] [RISCV][TableGen] Correct vTtoGetLlvmTyString for RISC-V tuples. (PR #162152)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-tablegen

Author: Craig Topper (topperc)

<details>
<summary>Changes</summary>

RISC-V tuples use "NF" not "nElem" to store the number of fields in the segment.

This fixes a crash when lowering a function with tuple return. getReturnInfo in CallLowering.cpp does Type*->EVT->Type* and we were incorrectly converting EVT to Type*.

---
Full diff: https://github.com/llvm/llvm-project/pull/162152.diff


1 Files Affected:

- (modified) llvm/utils/TableGen/Basic/VTEmitter.cpp (+2-2) 


``````````diff
diff --git a/llvm/utils/TableGen/Basic/VTEmitter.cpp b/llvm/utils/TableGen/Basic/VTEmitter.cpp
index c6b4d0b56c02b..301b27d2ebc5c 100644
--- a/llvm/utils/TableGen/Basic/VTEmitter.cpp
+++ b/llvm/utils/TableGen/Basic/VTEmitter.cpp
@@ -33,11 +33,11 @@ static void vTtoGetLlvmTyString(raw_ostream &OS, const Record *VT) {
   bool IsRISCVVecTuple = VT->getValueAsBit("isRISCVVecTuple");
 
   if (IsRISCVVecTuple) {
-    unsigned NElem = VT->getValueAsInt("nElem");
+    unsigned NF = VT->getValueAsInt("NF");
     unsigned Sz = VT->getValueAsInt("Size");
     OS << "TargetExtType::get(Context, \"riscv.vector.tuple\", "
           "ScalableVectorType::get(Type::getInt8Ty(Context), "
-       << (Sz / (NElem * 8)) << "), " << NElem << ")";
+       << (Sz / (NF * 8)) << "), " << NF << ")";
     return;
   }
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/162152