This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug middle-end/69454] [6 Regression] ix86_expand_prologue internal compiler error: Segmentation fault
- From: "hjl.tools at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 28 Jan 2016 13:59:08 +0000
- Subject: [Bug middle-end/69454] [6 Regression] ix86_expand_prologue internal compiler error: Segmentation fault
- Auto-submitted: auto-generated
- References: <bug-69454-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69454
--- Comment #36 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Richard Biener from comment #35)
> I wonder why LRA cannot spill using unaligned moves?
We keep track precise alignment requirement. RA will use
spill/fill the used part of vector registers. For GPR,
RA spill/fill the whole register since the minimal stack
alignment is the GPR size. In ix86_minimum_alignment,
since we only spill/fill SI registers, which need 32-bit
alignment, we return 32 for DImode in 32-bit mode. Now
STV introduce DImode spill/fill in 32-bit mode. If STV
is used, ix86_minimum_alignment must return 64 for DImode.
Otherwise, we are lying to RA.