From a3ff5727b78f1beeeee7484174ef86e48d35bd5c Mon Sep 17 00:00:00 2001 From: Nico Weber Date: Wed, 8 May 2019 15:50:21 +0000 Subject: [PATCH] [compiler-rt] Make builtins test pass when using i386 gcc as host compiler Just-built-clang is used to compile the test, but the library is built with gcc, so the usual 80-bit FPU vs 32-bit SSE mismatch makes the floating computations not bitwise identical. Fixes PR32910, see there for details. This uses the same technique used in all the other *c3* tests, see in particular mulsc3_test.c. (It might be cleaner to add compareResultCF to fp_test.h to force the floats into 32-bit in memory, but this is the less invasive fix.) Differential Revision: https://reviews.llvm.org/D61684 llvm-svn: 360264 --- compiler-rt/test/builtins/Unit/divsc3_test.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/compiler-rt/test/builtins/Unit/divsc3_test.c b/compiler-rt/test/builtins/Unit/divsc3_test.c index f9ad42abcdf9e..5781cff477300 100644 --- a/compiler-rt/test/builtins/Unit/divsc3_test.c +++ b/compiler-rt/test/builtins/Unit/divsc3_test.c @@ -102,7 +102,7 @@ int test__divsc3(float a, float b, float c, float d) { float _Complex z = (a * c + b * d) / (c * c + d * d) + (b * c - a * d) / (c * c + d * d) * _Complex_I; - if (r != z) + if (cabsf((r-z)/r) > 1.e-6) return 1; } break;