[llvm] [SDAG] Honor signed arguments in floating point libcalls (PR #109134)

[Nikita Popov via llvm-commits]

================
@@ -0,0 +1,53 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -O1 -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr8 < %s | FileCheck %s
+
+; Test that a negative parameter smaller than 64 bits (e.g., int)
+; is correctly implemented with sign-extension when passed to
+; a floating point libcall.
+
+define double @ldexp_test(ptr %a, ptr %b) {
----------------
nikic wrote:

```suggestion
define double @ldexp_test(ptr %a, ptr %b) nounwind {
```
To avoid cfi noise.

https://github.com/llvm/llvm-project/pull/109134