------- Comment #6 from ubizjak at gmail dot com 2008-05-16 19:09 -------
There is still underlying RA problem left. Compiling the testcase from Comment
0 on 32bit i686 (-O2 -msse2), we get:
_mm_set_epi32:
pushl %ebp
movl %esp, %ebp
movd 8(%ebp), %xmm0
movd 12(%ebp), %xmm2
punpckldq %xmm0, %xmm2
movd 16(%ebp), %xmm0
>>> movq %xmm2, %xmm1
movd 20(%ebp), %xmm2
popl %ebp
punpckldq %xmm0, %xmm2
>>> movq %xmm2, %xmm0
punpcklqdq %xmm1, %xmm0
ret
Two movq insns can be avoided by rearranging of registers.
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Ever Confirmed|0 |1
Keywords| |ra
Last reconfirmed|0000-00-00 00:00:00 |2008-05-16 19:09:08
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36222
