[x265] [PATCH] pixel_add_ps_2x4, asm code

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari
# Date 1385039242 -19800
# Node ID 4de4f5ea8bb5b84feac68ce057c2614afdb268eb
# Parent  a02029796cb1c99e75af53844b6b18b74fa68d27
pixel_add_ps_2x4, asm code

diff -r a02029796cb1 -r 4de4f5ea8bb5 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Nov 21 18:21:54 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Nov 21 18:37:22 2013 +0530
@@ -635,6 +635,7 @@
         // until all partitions are coded and commit smaller patches, easier to
         // review.
 
+        p.chroma[X265_CSP_I420].add_ps[CHROMA_2x4] = x265_pixel_add_ps_2x4_sse4;
         p.chroma[X265_CSP_I420].add_ps[CHROMA_4x2] = x265_pixel_add_ps_4x2_sse4;
         p.chroma[X265_CSP_I420].add_ps[CHROMA_4x4] = x265_pixel_add_ps_4x4_sse4;
         p.chroma[X265_CSP_I420].add_ps[CHROMA_4x8] = x265_pixel_add_ps_4x8_sse4;
diff -r a02029796cb1 -r 4de4f5ea8bb5 source/common/x86/pixeladd8.asm
--- a/source/common/x86/pixeladd8.asm	Thu Nov 21 18:21:54 2013 +0530
+++ b/source/common/x86/pixeladd8.asm	Thu Nov 21 18:37:22 2013 +0530
@@ -28,6 +28,53 @@
 SECTION .text
 
 
+
+;-----------------------------------------------------------------------------
+; void pixel_add_ps_2x4(pixel *dest, intptr_t destride, pixel *src0, int16_t *scr1, intptr_t srcStride0, intptr_t srcStride1)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal pixel_add_ps_2x4, 6, 6, 2, dest, destride, src0, scr1, srcStride0, srcStride1
+
+add         r5,            r5
+
+pmovzxbw    m0,            [r2]
+movh        m1,            [r3]
+
+paddw       m0,            m1
+packuswb    m0,            m0
+
+pextrw      [r0],          m0,      0
+
+pmovzxbw    m0,            [r2 + r4]
+movh        m1,            [r3 + r5]
+
+paddw       m0,            m1
+packuswb    m0,            m0
+
+pextrw      [r0 + r1],     m0,     0
+
+pmovzxbw    m0,            [r2 + 2 * r4]
+movh        m1,            [r3 + 2 * r5]
+
+paddw       m0,            m1
+packuswb    m0,            m0
+
+pextrw      [r0 + 2 * r1], m0,     0
+
+lea         r0,            [r0 + 2 * r1]
+lea         r2,            [r2 + 2 * r4]
+lea         r3,            [r3 + 2 * r5]
+
+pmovzxbw    m0,            [r2 + r4]
+movh        m1,            [r3 + r5]
+
+paddw       m0,            m1
+packuswb    m0,            m0
+
+pextrw      [r0 + r1],     m0,     0
+
+RET
+
 ;-----------------------------------------------------------------------------
 ; void pixel_add_ps_4x2(pixel *dest, intptr_t destride, pixel *src0, int16_t *scr1, intptr_t srcStride0, intptr_t srcStride1)
 ;-----------------------------------------------------------------------------