[llvm] [Transforms] Expand optimizeTan to fold more inverse trig pairs (PR #77799)

[via llvm-commits]

================
@@ -0,0 +1,59 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+
+define float @mytan(float %x) {
+; CHECK-LABEL: define float @mytan(
+; CHECK-SAME: float [[X:%.*]]) {
+; CHECK-NEXT:    [[CALL:%.*]] = call float @atanf(float [[X]])
+; CHECK-NEXT:    [[CALL1:%.*]] = call float @tanf(float [[CALL]])
----------------
AtariDreams wrote:

I think this is what the nofastmath path is for but okay.

https://github.com/llvm/llvm-project/pull/77799