This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug rtl-optimization/54421] Extra movdqa when accessing quadwords in a 128-bit SSE register
- From: "law at redhat dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 08 Dec 2016 19:51:28 +0000
- Subject: [Bug rtl-optimization/54421] Extra movdqa when accessing quadwords in a 128-bit SSE register
- Auto-submitted: auto-generated
- References: <bug-54421-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54421
Jeffrey A. Law <law at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
CC| |law at redhat dot com
Resolution|--- |FIXED
--- Comment #2 from Jeffrey A. Law <law at redhat dot com> ---
It looks like we went through a series of improvements resulting in the current
compiler generating:
movhlps %xmm0, %xmm1
movq %xmm0, %rdx
movq %xmm1, %rax
orq %rax, %rdx
sete %al
movzbl %al, %eax
ret
ie, it operates solely on registers and thus avoids the unnecessary stack
loads/stores.
Given it was a progression over time, I'm not going to bisect each improvement.
It's just not worth the effort.