[PATCH] D11438: Fix x86_64 fp128 calling convention

[Chih-Hung Hsieh via llvm-commits]

chh added a comment.

Martell, I don't have Windows or mingw-w64 system to test now, specifically I don't know what gcc generates for that system. If you were compiling test code with gcc and clang, and link with libraries compiled by gcc, then you could hit calling convention problem. You could compare first assembly code output from both compilers.
Please take a look and try the test cases in
https://llvm.org/bugs/show_bug.cgi?id=23897
https://llvm.org/bugs/show_bug.cgi?id=24111

http://reviews.llvm.org/D11437 has the clang changes to fix the calling convention for f128, and this patch is for the llvm part. If mingw-w64 does not use 128-bit floating point, then my patches shouldn't affect your problem, but you can find the source locations to fix mingw-w64 issues.


http://reviews.llvm.org/D11438