[x265] [PATCH] asm: avx2 code for add_ps[8x8] for 10bpp -- 24.9x

[sumalatha at multicorewareinc.com]

# HG changeset patch
# User Sumalatha Polureddy<sumalatha at multicorewareinc.com>
# Date 1425286035 -19800
# Node ID 1be088c8bc675752ebfebc4fda3bad41659269a4
# Parent  a9ad4d8202796dfb78e9d180f5fdb7cc0996ea66
asm: avx2 code for add_ps[8x8] for 10bpp -- 24.9x

add_ps[  8x8]  24.97x   275.68          6882.88

diff -r a9ad4d820279 -r 1be088c8bc67 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Mon Mar 02 14:10:07 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Mon Mar 02 14:17:15 2015 +0530
@@ -1069,6 +1069,8 @@
     }
     if (cpuMask & X265_CPU_AVX2)
     {
+        p.cu[BLOCK_8x8].add_ps = x265_pixel_add_ps_8x8_avx2;
+
         p.cu[BLOCK_32x32].ssd_s = x265_pixel_ssd_s_32_avx2;
         p.cu[BLOCK_16x16].sse_ss = x265_pixel_ssd_ss_16x16_avx2;
 
diff -r a9ad4d820279 -r 1be088c8bc67 source/common/x86/pixeladd8.asm
--- a/source/common/x86/pixeladd8.asm	Mon Mar 02 14:10:07 2015 +0530
+++ b/source/common/x86/pixeladd8.asm	Mon Mar 02 14:17:15 2015 +0530
@@ -229,6 +229,53 @@
 
     jnz     .loop
     RET
+
+INIT_YMM avx2
+cglobal pixel_add_ps_8x%2, 6, 7, 6, dest, destride, src0, scr1, srcStride0, srcStride1
+    mova    m5,     [pw_pixel_max]
+    pxor    m4,     m4
+    mov     r6d,    %2/4
+    add     r4,     r4
+    add     r5,     r5
+    add     r1,     r1
+.loop:
+    movu        xm0,        [r2]        ; row 0 of src0
+    movu        xm1,        [r2 + r4]   ; row 1 of src0
+    vinserti128 m0, m0, xm1, 1
+
+    movu    xm1,     [r3]               ; row 0 of src1
+    movu    xm2,     [r3 + r5]          ; row 1 of src1
+    vinserti128 m1, m1, xm2, 1
+    lea     r2,     [r2 + r4 * 2]
+    lea     r3,     [r3 + r5 * 2]
+
+    paddw   m0,     m1
+    CLIPW  m0, m4, m5
+    movu        [r0],       xm0         ; row 0 of dst
+    vextracti128 xm3, m0, 1
+    movu        [r0 + r1],  xm3         ; row 1 of dst
+    lea     r0,     [r0 + r1 * 2]
+
+    movu        xm0,        [r2]        ; row 2 of src0
+    movu        xm1,        [r2 + r4]   ; row 3 of src0
+    vinserti128 m0, m0, xm1, 1
+
+    movu    xm1,     [r3]               ; row 2 of src1
+    movu    xm2,     [r3 + r5]          ; row 3 of src1
+    vinserti128 m1, m1, xm2, 1
+    lea     r2,     [r2 + r4 * 2]
+    lea     r3,     [r3 + r5 * 2]
+
+    paddw   m0,     m1
+    CLIPW  m0, m4, m5
+    movu        [r0],       xm0         ; row 2 of dst
+    vextracti128 xm3, m0, 1
+    movu        [r0 + r1],  xm3         ; row 3 of dst
+    lea     r0,     [r0 + r1 * 2]
+
+    dec     r6d
+    jnz     .loop
+    RET
 %else
 INIT_XMM sse4
 cglobal pixel_add_ps_8x%2, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1