[x265] [PATCH] asm: avx2 asm code for 8bpp and 16bpp vesion of scale1D_128to64 module

[murugan at multicorewareinc.com]

# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1413888229 -19800
#      Tue Oct 21 16:13:49 2014 +0530
# Node ID b1a86d6e9dc3fab7996a5bc2d653513d1120530c
# Parent  e66f78a6df4f7fee71437b6645574b66024574f4
asm: avx2 asm code for 8bpp and 16bpp vesion of scale1D_128to64 module

diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Oct 21 11:51:19 2014 +0530
+++ b/source/common/x86/asm-primitives.cpp	Tue Oct 21 16:13:49 2014 +0530
@@ -1444,6 +1444,7 @@
         p.quant = x265_quant_avx2;
         p.nquant = x265_nquant_avx2;
         p.dequant_normal = x265_dequant_normal_avx2;
+        p.scale1D_128to64 = x265_scale1D_128to64_avx2;
 #if X86_64
         p.dct[DCT_8x8] = x265_dct8_avx2;
         p.dct[DCT_16x16] = x265_dct16_avx2;
@@ -1788,6 +1789,7 @@
         p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x24] = x265_blockcopy_ss_16x24_avx;
         p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x32] = x265_blockcopy_ss_16x32_avx;
         p.chroma[X265_CSP_I422].copy_ss[CHROMA422_16x64] = x265_blockcopy_ss_16x64_avx;
+        p.scale1D_128to64 = x265_scale1D_128to64_avx2;
 
 #if X86_64
         p.dct[DCT_8x8] = x265_dct8_avx2;
diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/pixel-util.h
--- a/source/common/x86/pixel-util.h	Tue Oct 21 11:51:19 2014 +0530
+++ b/source/common/x86/pixel-util.h	Tue Oct 21 16:13:49 2014 +0530
@@ -70,6 +70,7 @@
 float x265_pixel_ssim_end4_avx(int sum0[5][4], int sum1[5][4], int width);
 
 void x265_scale1D_128to64_ssse3(pixel *, pixel *, intptr_t);
+void x265_scale1D_128to64_avx2(pixel *, pixel *, intptr_t);
 void x265_scale2D_64to32_ssse3(pixel *, pixel *, intptr_t);
 
 #define SETUP_CHROMA_PIXELSUB_PS_FUNC(W, H, cpu) \
diff -r e66f78a6df4f -r b1a86d6e9dc3 source/common/x86/pixel-util8.asm
--- a/source/common/x86/pixel-util8.asm	Tue Oct 21 11:51:19 2014 +0530
+++ b/source/common/x86/pixel-util8.asm	Tue Oct 21 16:13:49 2014 +0530
@@ -42,7 +42,7 @@
 mask_ff:   times 16 db 0xff
            times 16 db 0
 deinterleave_shuf: db 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15
-deinterleave_word_shuf: db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 15, 15
+deinterleave_word_shuf: db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15
 hmul_16p:  times 16 db 1
            times 8 db 1, -1
 hmulw_16p:  times 8 dw 1
@@ -53,6 +53,7 @@
 SECTION .text
 
 cextern pw_1
+cextern pb_1
 cextern pw_00ff
 cextern pw_2000
 cextern pw_pixel_max
@@ -3439,6 +3440,69 @@
 %endif
 RET
 
+%if HIGH_BIT_DEPTH == 1
+INIT_YMM avx2
+cglobal scale1D_128to64, 2, 2, 4
+    pxor            m2, m2
+    vbroadcasti128  m3, [deinterleave_word_shuf]
+
+    movu            m0, [r1]
+    movu            m1, [r1 + 32]
+    phaddw          m0, m1
+    pavgw           m0, m2
+    vpermq          m0, m0, 0xD8
+    movu            [r0], m0
+
+    movu            m0, [r1 + 64]
+    movu            m1, [r1 + 96]
+    phaddw          m0, m1
+    pavgw           m0, m2
+    vpermq          m0, m0, 0xD8
+    movu            [r0 + 32], m0
+
+    movu            m0, [r1 + 128]
+    movu            m1, [r1 + 160]
+    phaddw          m0, m1
+    pavgw           m0, m2
+    vpermq          m0, m0, 0xD8
+    movu            [r0 + 64], m0
+
+    movu            m0, [r1 + 192]
+    movu            m1, [r1 + 224]
+    phaddw          m0, m1
+    pavgw           m0, m2
+    vpermq          m0, m0, 0xD8
+    movu            [r0 + 96], m0
+    RET
+%else ; HIGH_BIT_DEPTH == 0
+INIT_YMM avx2
+cglobal scale1D_128to64, 2, 2, 5
+    pxor            m2, m2
+    vbroadcasti128  m3, [deinterleave_shuf]
+    mova            m4, [pb_1]
+
+    movu            m0, [r1]
+    pmaddubsw       m0, m0, m4
+    pavgw           m0, m2
+    movu            m1, [r1 + 32]
+    pmaddubsw       m1, m1, m4
+    pavgw           m1, m2
+    packuswb        m0, m1
+    vpermq          m0, m0, 0xD8
+    movu            [r0], m0
+
+    movu            m0, [r1 + 64]
+    pmaddubsw       m0, m0, m4
+    pavgw           m0, m2
+    movu            m1, [r1 + 96]
+    pmaddubsw       m1, m1, m4
+    pavgw           m1, m2
+    packuswb        m0, m1
+    vpermq          m0, m0, 0xD8
+    movu            [r0 + 32], m0
+    RET
+%endif
+
 ;-----------------------------------------------------------------
 ; void scale2D_64to32(pixel *dst, pixel *src, intptr_t stride)
 ;-----------------------------------------------------------------