<div style="line-height:1.7;color:#000000;font-size:14px;font-family:arial"><div style="color: rgb(0, 0, 0); line-height: 1.7; font-family: arial; font-size: 14px;"><div>in IACA report, just ~3 cycles<br></div><div></div><div id="divNeteaseMailCard"></div><div><br></div>At 2015-10-07 16:49:24,"Pradeep Ramachandran" <<a href="mailto:pradeep@multicorewareinc.com">pradeep@multicorewareinc.com</a>> wrote:<br> <blockquote id="isReplyContent" style="margin: 0px 0px 0px 0.8ex; padding-left: 1ex; border-left-color: rgb(204, 204, 204); border-left-width: 1px; border-left-style: solid;"><div dir="ltr">Min,<div>Do you see a performance improvement through this sequence of patches? </div></div><div class="gmail_extra"><br clear="all"><div><div class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr">Pradeep Ramachandran, PhD<div>Solution Architect,</div><div><a href="http://pradeeprama.info/" target="_blank">pradeeprama.info/</a></div><div><a href="http://www.multicorewareinc.com/" target="_blank">www.multicorewareinc.com/</a></div><div>Ph: +91 99627 82018</div></div></div></div></div></div></div></div></div>
<br><div class="gmail_quote">On Wed, Oct 7, 2015 at 4:25 AM, Min Chen <span dir="ltr"><<a href="mailto:chenm003@163.com" target="_blank">chenm003@163.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin: 0px 0px 0px 0.8ex; padding-left: 1ex; border-left-color: rgb(204, 204, 204); border-left-width: 1px; border-left-style: solid;"># HG changeset patch<br>
# User Min Chen <<a href="mailto:chenm003@163.com">chenm003@163.com</a>><br>
# Date 1444170873 18000<br>
# Node ID f2158e8c15efde0991cb06033399e9256f54ad7e<br>
# Parent 36a54b2cf7c4c96067bafb67077651d30d83e8e9<br>
asm: improve sao stats compute by SIMD<br>
---<br>
source/common/x86/loopfilter.asm | 124 +++++++++++++++-----------------------<br>
1 files changed, 49 insertions(+), 75 deletions(-)<br>
<br>
diff -r 36a54b2cf7c4 -r f2158e8c15ef source/common/x86/loopfilter.asm<br>
--- a/source/common/x86/loopfilter.asm Tue Oct 06 16:42:06 2015 -0500<br>
+++ b/source/common/x86/loopfilter.asm Tue Oct 06 17:34:33 2015 -0500<br>
@@ -2112,25 +2112,19 @@<br>
mov r0, r6mp<br>
<br>
; s_eoTable = {1, 2, 0, 3, 4}<br>
- movzx r5d, word [rsp + 0 * 2]<br>
- add [r0 + 1 * 4], r5d<br>
- movzx r6d, word [rsp + 1 * 2]<br>
- add [r0 + 2 * 4], r6d<br>
- movzx r5d, word [rsp + 2 * 2]<br>
- add [r0 + 0 * 4], r5d<br>
- movzx r6d, word [rsp + 3 * 2]<br>
- add [r0 + 3 * 4], r6d<br>
+ pmovzxwd m0, [rsp + 0 * 2]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r0]<br>
+ paddd m0, m1<br>
+ movu [r0], m0<br>
movzx r5d, word [rsp + 4 * 2]<br>
add [r0 + 4 * 4], r5d<br>
<br>
- mov r6d, [rsp + 5 * 2 + 0 * 4]<br>
- add [r9 + 1 * 4], r6d<br>
- mov r5d, [rsp + 5 * 2 + 1 * 4]<br>
- add [r9 + 2 * 4], r5d<br>
- mov r6d, [rsp + 5 * 2 + 2 * 4]<br>
- add [r9 + 0 * 4], r6d<br>
- mov r5d, [rsp + 5 * 2 + 3 * 4]<br>
- add [r9 + 3 * 4], r5d<br>
+ movu m0, [rsp + 5 * 2 + 0 * 4]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r9]<br>
+ paddd m0, m1<br>
+ movu [r9], m0<br>
mov r6d, [rsp + 5 * 2 + 4 * 4]<br>
add [r9 + 4 * 4], r6d<br>
RET<br>
@@ -2219,25 +2213,19 @@<br>
mov r0, r7m<br>
<br>
; s_eoTable = {1,2,0,3,4}<br>
- movzx r6d, word [rsp + 0 * 2]<br>
- add [r0 + 1 * 4], r6d<br>
- movzx r6d, word [rsp + 1 * 2]<br>
- add [r0 + 2 * 4], r6d<br>
- movzx r6d, word [rsp + 2 * 2]<br>
- add [r0 + 0 * 4], r6d<br>
- movzx r6d, word [rsp + 3 * 2]<br>
- add [r0 + 3 * 4], r6d<br>
- movzx r6d, word [rsp + 4 * 2]<br>
- add [r0 + 4 * 4], r6d<br>
-<br>
- mov r6d, [rsp + 5 * 2 + 0 * 4]<br>
- add [r1 + 1 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 1 * 4]<br>
- add [r1 + 2 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 2 * 4]<br>
- add [r1 + 0 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 3 * 4]<br>
- add [r1 + 3 * 4], r6d<br>
+ pmovzxwd m0, [rsp + 0 * 2]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r0]<br>
+ paddd m0, m1<br>
+ movu [r0], m0<br>
+ movzx r5d, word [rsp + 4 * 2]<br>
+ add [r0 + 4 * 4], r5d<br>
+<br>
+ movu m0, [rsp + 5 * 2 + 0 * 4]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r1]<br>
+ paddd m0, m1<br>
+ movu [r1], m0<br>
mov r6d, [rsp + 5 * 2 + 4 * 4]<br>
add [r1 + 4 * 4], r6d<br>
RET<br>
@@ -2379,33 +2367,25 @@<br>
mov r0, r8m<br>
<br>
; s_eoTable = {1,2,0,3,4}<br>
- movzx r6d, word [rsp + 0 * 2]<br>
- add [r0 + 1 * 4], r6d<br>
- movzx r6d, word [rsp + 1 * 2]<br>
- add [r0 + 2 * 4], r6d<br>
- movzx r6d, word [rsp + 2 * 2]<br>
- add [r0 + 0 * 4], r6d<br>
- movzx r6d, word [rsp + 3 * 2]<br>
- add [r0 + 3 * 4], r6d<br>
- movzx r6d, word [rsp + 4 * 2]<br>
- add [r0 + 4 * 4], r6d<br>
-<br>
- mov r6d, [rsp + 5 * 2 + 0 * 4]<br>
- add [r1 + 1 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 1 * 4]<br>
- add [r1 + 2 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 2 * 4]<br>
- add [r1 + 0 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 3 * 4]<br>
- add [r1 + 3 * 4], r6d<br>
+ pmovzxwd m0, [rsp + 0 * 2]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r0]<br>
+ paddd m0, m1<br>
+ movu [r0], m0<br>
+ movzx r5d, word [rsp + 4 * 2]<br>
+ add [r0 + 4 * 4], r5d<br>
+<br>
+ movu m0, [rsp + 5 * 2 + 0 * 4]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r1]<br>
+ paddd m0, m1<br>
+ movu [r1], m0<br>
mov r6d, [rsp + 5 * 2 + 4 * 4]<br>
add [r1 + 4 * 4], r6d<br>
RET<br>
%endif ; ARCH_X86_64<br>
<br>
<br>
-<br>
-<br>
;void saoStatE3(const int16_t *diff, const pixel *rec, intptr_t stride, int8_t *upBuff1, int endX, int endY, int32_t *stats, int32_t *count);<br>
;{<br>
; memset(tmp_stats, 0, sizeof(tmp_stats));<br>
@@ -2520,25 +2500,19 @@<br>
mov r0, r7m<br>
<br>
; s_eoTable = {1,2,0,3,4}<br>
- movzx r6d, word [rsp + 0 * 2]<br>
- add [r0 + 1 * 4], r6d<br>
- movzx r6d, word [rsp + 1 * 2]<br>
- add [r0 + 2 * 4], r6d<br>
- movzx r6d, word [rsp + 2 * 2]<br>
- add [r0 + 0 * 4], r6d<br>
- movzx r6d, word [rsp + 3 * 2]<br>
- add [r0 + 3 * 4], r6d<br>
- movzx r6d, word [rsp + 4 * 2]<br>
- add [r0 + 4 * 4], r6d<br>
-<br>
- mov r6d, [rsp + 5 * 2 + 0 * 4]<br>
- add [r1 + 1 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 1 * 4]<br>
- add [r1 + 2 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 2 * 4]<br>
- add [r1 + 0 * 4], r6d<br>
- mov r6d, [rsp + 5 * 2 + 3 * 4]<br>
- add [r1 + 3 * 4], r6d<br>
+ pmovzxwd m0, [rsp + 0 * 2]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r0]<br>
+ paddd m0, m1<br>
+ movu [r0], m0<br>
+ movzx r5d, word [rsp + 4 * 2]<br>
+ add [r0 + 4 * 4], r5d<br>
+<br>
+ movu m0, [rsp + 5 * 2 + 0 * 4]<br>
+ pshufd m0, m0, q3102<br>
+ movu m1, [r1]<br>
+ paddd m0, m1<br>
+ movu [r1], m0<br>
mov r6d, [rsp + 5 * 2 + 4 * 4]<br>
add [r1 + 4 * 4], r6d<br>
RET<br>
<br>
_______________________________________________<br>
x265-devel mailing list<br>
<a href="mailto:x265-devel@videolan.org">x265-devel@videolan.org</a><br>
<a href="https://mailman.videolan.org/listinfo/x265-devel" target="_blank" rel="noreferrer">https://mailman.videolan.org/listinfo/x265-devel</a><br>
</blockquote></div><br></div>
</blockquote></div></div>