<div dir="ltr">Hello Chen,<div><br></div><div>Thanks for the comments, I will make necessary changes and send the patch again.</div><div><br></div><div><br></div><div>Regards,</div><div><br></div><div>Aasaipriya C</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Nov 12, 2014 at 10:16 PM, 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:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div style="line-height:1.7;color:#000000;font-size:14px;font-family:arial"><div><br> </div><pre><br>At 2014-11-12 18:56:45,<a href="mailto:aasaipriya@multicorewareinc.com" target="_blank">aasaipriya@multicorewareinc.com</a> wrote:
># HG changeset patch
># User Aasaipriya
># Date 1415788047 -19800
>#      Wed Nov 12 15:57:27 2014 +0530
># Node ID f0a17f2b4c22ff8aa05afb40e3360e5ac03590a6
># Parent  98fb658f3229ab10e808204c265a12e18d71638e
>Luma_hpp[16x16] avx2 asm code : improved 2307c->1695c
>
>diff -r 98fb658f3229 -r f0a17f2b4c22 source/common/x86/asm-primitives.cpp
>--- a/source/common/x86/asm-primitives.cpp Tue Nov 11 19:30:19 2014 +0900
>+++ b/source/common/x86/asm-primitives.cpp Wed Nov 12 15:57:27 2014 +0530
>@@ -1799,6 +1799,7 @@
>         p.transpose[BLOCK_64x64] = x265_transpose64_avx2;
> #endif
>         p.luma_hpp[LUMA_4x4] = x265_interp_8tap_horiz_pp_4x4_avx2;
>+        p.luma_hpp[LUMA_16x16] = x265_interp_8tap_horiz_pp_16x16_avx2;
>         p.luma_vpp[LUMA_4x4] = x265_interp_8tap_vert_pp_4x4_avx2;
>     }
> #endif // if HIGH_BIT_DEPTH
>diff -r 98fb658f3229 -r f0a17f2b4c22 source/common/x86/ipfilter8.asm
>--- a/source/common/x86/ipfilter8.asm      Tue Nov 11 19:30:19 2014 +0900
>+++ b/source/common/x86/ipfilter8.asm      Wed Nov 12 15:57:27 2014 +0530
>@@ -854,6 +854,58 @@
>     RET
> 
> 
>+INIT_YMM avx2
>+cglobal interp_8tap_horiz_pp_16x16, 4,6,8
>+    sub             r0, 3
>+    mov             r4d, r4m
>+%ifdef PIC
>+    lea             r5, [tab_LumaCoeff]
>+    vpbroadcastd    m0, [r5 + r4 * 8]
>+    vpbroadcastd    m1, [r5 + r4 * 8 + 4]
>+%else
>+    vpbroadcastd    m0, [tab_LumaCoeff + r4 * 8]
>+    vpbroadcastd    m1, [tab_LumaCoeff + r4 * 8 + 4]
>+%endif
>+    mova            m2, [tab_Tm]
>+    movu            m3, [tab_Tm + 16]
>+    vpbroadcastd    m7, [pw_1]
>+
>+    ; register map
>+    ; m0 , m1 interpolate coeff
>+    ; m2 , m2  shuffle order table
>+    ; m7 - pw_1
>+
>+    mov             r4d,  16
>+.loop:
>+    ; Row 0
>+    vbroadcasti128  m4, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
>+    pshufb          m5, m4, m3
>+    pshufb          m4, m2
>+    pmaddubsw       m4, m0
>+    pmaddubsw       m5, m1
>+    paddw           m4, m5
>+    pmaddwd         m4, m7
>+    vbroadcasti128  m5, [r0 + 8]                    ; second 8 elements in Row0 
>+    pshufb          m6, m5, m3
>+    pshufb          m5, m2
>+    pmaddubsw       m5, m0
>+    pmaddubsw       m6, m1
>+    paddw           m5, m6
>+    pmaddwd         m5, m7
>+    packssdw        m4, m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
>+    pmulhrsw        m4, [pw_512]
</pre><div><div class="h5"><pre>>+    packuswb        m4, m4
>+    vpermq          m4, m4, 11011000b
>+    pshufd          xm4, xm4, 11011000b
>+    movq            [r2], xm4
>+    movhps          [r2 + 8], xm4
</pre></div></div><pre>why split into two write instruction?</pre><span class=""><pre> </pre><pre>>+    lea             r0, [r0 + r1]
>+    lea             r2, [r2 + r3]
</pre></span><pre>add r0, r1</pre><span class=""><pre>>+    dec             r4d
>+    jnz             .loop
>+    RET
</pre></span></div><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">https://mailman.videolan.org/listinfo/x265-devel</a><br>
<br></blockquote></div><br></div>