[llvm] CodeGenPrepare: Remove threadlocal_address intrinsic when cheap to recompute. (PR #87844)

[Nikita Popov via llvm-commits]

================
@@ -0,0 +1,201 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -o - %s | FileCheck %s --check-prefix=NOPIC
+; RUN: llc -o - %s -relocation-model=pic | FileCheck %s --check-prefix=PIC
+; RUN: llc -o - %s -relocation-model=pic -enable-tlsdesc | FileCheck %s --check-prefix=TLSDESC
+
+target triple = "x86_64--linux-gnu"
+
+declare void @effect()
+declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull)
+
+ at foo_local = dso_local thread_local(localexec) global i32 0, align 4
+
+define i32 @func_local_tls(i32 %arg0, i32 %arg1) {
----------------
nikic wrote:

Add nounwind to avoid the cfi comments.

https://github.com/llvm/llvm-project/pull/87844