[FFmpeg-devel] [PATCH 3/7] x86: sbrdsp: implement SSE qmf_deint_bfly

[Michael Niedermayer]

On Sat, Apr 06, 2013 at 10:52:10AM +0000, Christophe Gisquet wrote:
> From 312 to 89/68 (sse/sse2) cycles on Arrandale and Win64.
> Sandybridge: 68/47 cycles.
> 
> Having a loop counter is a 7 cycle gain.
> Unrolling is another 7 cycle gain.
> Working in reverse scan is another 6 cycles.

following gives 1 cycle speedup on SSE2 SB

@@ -322,6 +322,7 @@ cglobal sbr_qmf_deint_neg, 2,3,3,v,src,vrev
 cglobal sbr_qmf_deint_bfly, 3,5,8, v,src0,src1,vrev,c
     mov        cq, 64*4-2*mmsize
     lea     vrevq, [vq + 64*4]
+    sub     vrevq, src1q
 .loop:
     mova       m0, [src0q+cq]
     mova       m1, [src1q]
@@ -342,12 +343,11 @@ cglobal sbr_qmf_deint_bfly, 3,5,8, v,src0,src1,vrev,c
     subps      m0, m7
     addps      m1, m6
     subps      m4, m3
-    mova  [vrevq], m1
-    mova  [vrevq+mmsize], m5
+    mova  [vrevq+src1q], m1
+    mova  [vrevq+src1q+mmsize], m5
     mova  [vq+cq], m0
     mova  [vq+cq+mmsize], m4
     add     src1q, 2*mmsize
-    add     vrevq, 2*mmsize
     sub        cq, 2*mmsize
     jge     .loop
     REP_RET

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

No snowflake in an avalanche ever feels responsible. -- Voltaire
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20130406/a31d5629/attachment.asc>