[llvm] [NFC][CodeGen] Add Register guard to ARMMaterializeFP. (PR #182559)

via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 20 10:18:45 PST 2026


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-arm

Author: Paul Walker (paulwalker-arm)

<details>
<summary>Changes</summary>

This does not directly fix any issue because the implementation indirectly ensures the correct behaviour. However, all the other "<Tgt>Materialize" functions (Int and FP across all targets, including ARMMaterializeInt) have explicit Register guards so for peace of mind I figured it's worth added them.

---
Full diff: https://github.com/llvm/llvm-project/pull/182559.diff


1 Files Affected:

- (modified) llvm/lib/Target/ARM/ARMFastISel.cpp (+3) 


``````````diff
diff --git a/llvm/lib/Target/ARM/ARMFastISel.cpp b/llvm/lib/Target/ARM/ARMFastISel.cpp
index b5244bfb9a80d..80eb0f5f810ff 100644
--- a/llvm/lib/Target/ARM/ARMFastISel.cpp
+++ b/llvm/lib/Target/ARM/ARMFastISel.cpp
@@ -440,6 +440,9 @@ Register ARMFastISel::ARMMoveToIntReg(MVT VT, Register SrcReg) {
 // (the high and the low) into integer registers then use a move to get
 // the combined constant into an FP reg.
 Register ARMFastISel::ARMMaterializeFP(const ConstantFP *CFP, MVT VT) {
+  if (VT != MVT::f32 && VT != MVT::f64)
+    return Register();
+
   const APFloat Val = CFP->getValueAPF();
   bool is64bit = VT == MVT::f64;
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/182559


More information about the llvm-commits mailing list