[compiler-rt] [compiler-rt][ARM] Enable strict mode in divsf3/mulsf3 tests (PR #179918)
Simon Tatham via llvm-commits
llvm-commits at lists.llvm.org
Fri Feb 20 09:06:06 PST 2026
================
@@ -16,31 +16,33 @@
// 0x7fc00000. For the Arm optimized FP implementation, which commits to a more
// detailed handling of NaNs, we tighten up the check and include some extra
// test cases specific to that NaN policy.
-#if (__arm__ && !(__thumb__ && !__thumb2__)) && COMPILER_RT_ARM_OPTIMIZED_FP
+#if (__arm__ || __thumb__) && COMPILER_RT_ARM_OPTIMIZED_FP
----------------
statham-arm wrote:
I've had a look at this now. It's awkward because of the different sets of functions implemented in Thumb1 and Arm/Thumb2: the way I currently have it, each test file is testing `COMPILER_RT_ARM_OPTIMIZED_FP` plus some extra condition indicating whether _that function_ has an optimized implementation. So I'll either have to make two different #defines for the two subsets, or else find a way to reset the definition of `COMPILER_RT_ARM_OPTIMIZED_FP` per test file.
I fear that once I get this working you might well feel the cure was worse than the disease!
https://github.com/llvm/llvm-project/pull/179918
More information about the llvm-commits
mailing list