[llvm-commits] Patch to use proper calling convention for i386--win32 _ftol2 runtime function

[Joe Groff]

On Fri, Jan 27, 2012 at 1:48 PM, Joe Groff <arcata at gmail.com> wrote:
> Here's a patch that adds a WIN_FTOL_* family of pseudo-instructions
> modelling the unique calling convention used by the Win32 _ftol2
> runtime function. I'm not totally happy with the tests I wrote; even
> using -O0, it appears to depend on the behavior of the scheduler
> somewhat, so any feedback on how to write a more robust test would be
> appreciated, along with any other comments. Thanks Jakob Stoklund
> Olesen for walking me through things.

Here's a revised version of the patch with additional tests to test
the kill and no-kill cases for the new FTOL pseudo-instruction.

-Joe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: llvm-ftol2-4.diff
Type: application/octet-stream
Size: 15995 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120209/079d69ab/attachment.obj>