From a2dfe8d18d30b686c2f5401d7de8399d7b3fa4eb Mon Sep 17 00:00:00 2001
From: "Ronald S. Bultje" <rsbultje@gmail.com>
Date: Wed, 29 Dec 2010 18:00:26 +0000
Subject: Port pred8x8_dc_mmxext (H.264 intra prediction) from x264 to FFmpeg.
 Original authors: Holger Lubitz <holger lubitz org>, Jason Garrett-Glaser
 <darkshikari gmail com> (approves LGPL relicensing for this code) and Loren
 Merritt <lorenm at u dot washington dot edu> (approves LGPL relicensing for
 this code). Patch by Daniel Kang <daniel dot d dot kang at gmail com>, as
 part of Google's GCI 2010.

Originally committed as revision 26135 to svn://svn.ffmpeg.org/ffmpeg/trunk
---
 libavcodec/x86/h264_intrapred.asm | 62 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 62 insertions(+)

(limited to 'libavcodec/x86/h264_intrapred.asm')

diff --git a/libavcodec/x86/h264_intrapred.asm b/libavcodec/x86/h264_intrapred.asm
index 4f7ba3738a..12193defc7 100644
--- a/libavcodec/x86/h264_intrapred.asm
+++ b/libavcodec/x86/h264_intrapred.asm
@@ -865,6 +865,68 @@ cglobal pred8x8_top_dc_mmxext, 2,5
     RET
 %endif
 
+;-----------------------------------------------------------------------------
+; void pred8x8_dc_mmxext(uint8_t *src, int stride)
+;-----------------------------------------------------------------------------
+%ifdef CONFIG_GPL
+INIT_MMX
+cglobal pred8x8_dc_mmxext, 2,5
+    sub       r0, r1
+    pxor      m7, m7
+    movd      m0, [r0+0]
+    movd      m1, [r0+4]
+    psadbw    m0, m7            ; s0
+    mov       r4, r0
+    psadbw    m1, m7            ; s1
+
+    movzx    r2d, byte [r0+r1*1-1]
+    movzx    r3d, byte [r0+r1*2-1]
+    lea       r0, [r0+r1*2]
+    add      r2d, r3d
+    movzx    r3d, byte [r0+r1*1-1]
+    add      r2d, r3d
+    movzx    r3d, byte [r0+r1*2-1]
+    add      r2d, r3d
+    lea       r0, [r0+r1*2]
+    movd      m2, r2d            ; s2
+    movzx    r2d, byte [r0+r1*1-1]
+    movzx    r3d, byte [r0+r1*2-1]
+    lea       r0, [r0+r1*2]
+    add      r2d, r3d
+    movzx    r3d, byte [r0+r1*1-1]
+    add      r2d, r3d
+    movzx    r3d, byte [r0+r1*2-1]
+    add      r2d, r3d
+    movd      m3, r2d            ; s3
+
+    punpcklwd m0, m1
+    mov       r0, r4
+    punpcklwd m2, m3
+    punpckldq m0, m2            ; s0, s1, s2, s3
+    pshufw    m3, m0, 11110110b ; s2, s1, s3, s3
+    lea       r2, [r0+r1*2]
+    pshufw    m0, m0, 01110100b ; s0, s1, s3, s1
+    paddw     m0, m3
+    lea       r3, [r2+r1*2]
+    psrlw     m0, 2
+    pavgw     m0, m7            ; s0+s2, s1, s3, s1+s3
+    lea       r4, [r3+r1*2]
+    packuswb  m0, m0
+    punpcklbw m0, m0
+    movq      m1, m0
+    punpcklbw m0, m0
+    punpckhbw m1, m1
+    movq [r0+r1*1], m0
+    movq [r0+r1*2], m0
+    movq [r2+r1*1], m0
+    movq [r2+r1*2], m0
+    movq [r3+r1*1], m1
+    movq [r3+r1*2], m1
+    movq [r4+r1*1], m1
+    movq [r4+r1*2], m1
+    RET
+%endif
+
 ;-----------------------------------------------------------------------------
 ; void pred8x8_dc_rv40(uint8_t *src, int stride)
 ;-----------------------------------------------------------------------------
-- 
cgit v1.2.3