<div dir="ltr">Ignore this patch, Need some clean up. Will send a new one.</div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Oct 21, 2014 at 4:24 PM, <span dir="ltr"><<a href="mailto:murugan@multicorewareinc.com" target="_blank">murugan@multicorewareinc.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"># HG changeset patch<br>
# User Murugan Vairavel <<a href="mailto:murugan@multicorewareinc.com">murugan@multicorewareinc.com</a>><br>
# Date 1413888229 -19800<br>
# Tue Oct 21 16:13:49 2014 +0530<br>
# Node ID b1a86d6e9dc3fab7996a5bc2d653513d1120530c<br>
# Parent e66f78a6df4f7fee71437b6645574b66024574f4<br>
asm: avx2 asm code for 8bpp and 16bpp vesion of scale1D_128to64 module<br>
<br>
diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/asm-primitives.cpp<br>
--- a/source/common/x86/asm-primitives.cpp Tue Oct 21 11:51:19 2014 +0530<br>
+++ b/source/common/x86/asm-primitives.cpp Tue Oct 21 16:13:49 2014 +0530<br>
@@ -1444,6 +1444,7 @@<br>
p.quant = x265_quant_avx2;<br>
p.nquant = x265_nquant_avx2;<br>
p.dequant_normal = x265_dequant_normal_avx2;<br>
+ p.scale1D_128to64 = x265_scale1D_128to64_avx2;<br>
#if X86_64<br>
p.dct[DCT_8x8] = x265_dct8_avx2;<br>
p.dct[DCT_16x16] = x265_dct16_avx2;<br>
@@ -1788,6 +1789,7 @@<br>
p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x24] = x265_blockcopy_ss_16x24_avx;<br>
p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x32] = x265_blockcopy_ss_16x32_avx;<br>
p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x64] = x265_blockcopy_ss_16x64_avx;<br>
+ p.scale1D_128to64 = x265_scale1D_128to64_avx2;<br>
<br>
#if X86_64<br>
p.dct[DCT_8x8] = x265_dct8_avx2;<br>
diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/pixel-util.h<br>
--- a/source/common/x86/pixel-util.h Tue Oct 21 11:51:19 2014 +0530<br>
+++ b/source/common/x86/pixel-util.h Tue Oct 21 16:13:49 2014 +0530<br>
@@ -70,6 +70,7 @@<br>
float x265_pixel_ssim_end4_avx(int sum0[5][4], int sum1[5][4], int width);<br>
<br>
void x265_scale1D_128to64_ssse3(pixel *, pixel *, intptr_t);<br>
+void x265_scale1D_128to64_avx2(pixel *, pixel *, intptr_t);<br>
void x265_scale2D_64to32_ssse3(pixel *, pixel *, intptr_t);<br>
<br>
#define SETUP_CHROMA_PIXELSUB_PS_FUNC(W, H, cpu) \<br>
diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/pixel-util8.asm<br>
--- a/source/common/x86/pixel-util8.asm Tue Oct 21 11:51:19 2014 +0530<br>
+++ b/source/common/x86/pixel-util8.asm Tue Oct 21 16:13:49 2014 +0530<br>
@@ -42,7 +42,7 @@<br>
mask_ff: times 16 db 0xff<br>
times 16 db 0<br>
deinterleave_shuf: db 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15<br>
-deinterleave_word_shuf: db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 15, 15<br>
+deinterleave_word_shuf: db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15<br>
hmul_16p: times 16 db 1<br>
times 8 db 1, -1<br>
hmulw_16p: times 8 dw 1<br>
@@ -53,6 +53,7 @@<br>
SECTION .text<br>
<br>
cextern pw_1<br>
+cextern pb_1<br>
cextern pw_00ff<br>
cextern pw_2000<br>
cextern pw_pixel_max<br>
@@ -3439,6 +3440,69 @@<br>
%endif<br>
RET<br>
<br>
+%if HIGH_BIT_DEPTH == 1<br>
+INIT_YMM avx2<br>
+cglobal scale1D_128to64, 2, 2, 4<br>
+ pxor m2, m2<br>
+ vbroadcasti128 m3, [deinterleave_word_shuf]<br>
+<br>
+ movu m0, [r1]<br>
+ movu m1, [r1 + 32]<br>
+ phaddw m0, m1<br>
+ pavgw m0, m2<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0], m0<br>
+<br>
+ movu m0, [r1 + 64]<br>
+ movu m1, [r1 + 96]<br>
+ phaddw m0, m1<br>
+ pavgw m0, m2<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0 + 32], m0<br>
+<br>
+ movu m0, [r1 + 128]<br>
+ movu m1, [r1 + 160]<br>
+ phaddw m0, m1<br>
+ pavgw m0, m2<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0 + 64], m0<br>
+<br>
+ movu m0, [r1 + 192]<br>
+ movu m1, [r1 + 224]<br>
+ phaddw m0, m1<br>
+ pavgw m0, m2<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0 + 96], m0<br>
+ RET<br>
+%else ; HIGH_BIT_DEPTH == 0<br>
+INIT_YMM avx2<br>
+cglobal scale1D_128to64, 2, 2, 5<br>
+ pxor m2, m2<br>
+ vbroadcasti128 m3, [deinterleave_shuf]<br>
+ mova m4, [pb_1]<br>
+<br>
+ movu m0, [r1]<br>
+ pmaddubsw m0, m0, m4<br>
+ pavgw m0, m2<br>
+ movu m1, [r1 + 32]<br>
+ pmaddubsw m1, m1, m4<br>
+ pavgw m1, m2<br>
+ packuswb m0, m1<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0], m0<br>
+<br>
+ movu m0, [r1 + 64]<br>
+ pmaddubsw m0, m0, m4<br>
+ pavgw m0, m2<br>
+ movu m1, [r1 + 96]<br>
+ pmaddubsw m1, m1, m4<br>
+ pavgw m1, m2<br>
+ packuswb m0, m1<br>
+ vpermq m0, m0, 0xD8<br>
+ movu [r0 + 32], m0<br>
+ RET<br>
+%endif<br>
+<br>
;-----------------------------------------------------------------<br>
; void scale2D_64to32(pixel *dst, pixel *src, intptr_t stride)<br>
;-----------------------------------------------------------------<br>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div dir="ltr">With Regards,<div><br></div><div>Murugan. V</div><div>+919659287478</div></div>
</div>