On Thu, Apr 7, 2011 at 11:51 AM, Uros Bizjak <ubiz...@gmail.com> wrote:
> Hello!
>
> This patch converts "Parallel single-precision floating point element
> swizzling" and "Parallel double-precision floating point element
> swizzling" sections of sse.md.
>
> 2011-04-07 Uros Bizjak <ubiz...@gmail.com>
>
> * config/i386/sse.md: Update copyright year.
> (avxcvtvecmode): Remove.
> (sse_movhlps): Merge with *avx_movhlps.
> (sse_movlhps): Merge with *avx_movlhps.
> (vec_interleave_highv4sf): Merge with *avx_interleave_highv4sf.
> (vec_interleave_lowv4sf): Merge with *avx_interleave_lowv4sf.
> (sse_shufps_<mode>): Merge with *avx_shufps_<mode>.
> (sse_loadhps): Merge with *avx_loadhps.
> (sse_storelps): Merge with *avx_storelps.
> (sse_loadlps): Merge with *avx_loadlps.
> (sse_movss): Merge with *avx_movss.
> (*vec_concatv2sf_sse4_1): Merge with *vec_concatv2sf_avx.
> (*vec_concatv4sf_sse): Merge with *vec_concatv4sf_avx.
> (*vec_set<mode>_0_sse4_1): Merge with *vec_set<mode>_0_avx.
> (*vec_set<mode>_0_sse2): Split multi-unit alternative.
> (vec_set<mode>_0): Ditto.
> (*vec_setv4sf_sse4_1): Merge with *vec_setv4sf_avx.
> (sse4_1_insertps): Merge with *avx_insertps.
> (*sse3_interleave_highv2df): Merge with *avx_interleave_highv2df.
> (*sse3_interleave_lowv2df): Merge with *avx_interleave_lowv2df.
> (vec_interleave_highv2di): Merge with *avx_interleave_highv2di.
> (sse2_shufpd_<mode>): Merge with *avx_shufpd_<mode>.
> (sse2_storehpd): Merge with *avx_storehpd.
> (sse2_loadhpd): Merge with *avx_loadhpd.
> (sse2_loadlpd): Merge with *avx_loadlpd.
> (sse2_movsd): Merge with *avx_movsd.
> (*vec_concatv2df): Merge with *vec_concatv2df.
>
> Patch was tested on x86_64-pc-linux-gnu {,-m32}. Committed to mainline SVN.
>
This patch caused:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53759
I think this patch changes the pattern order by accident.
--
H.J.
