[x265] [PATCH] asm: Optimized sad_48x64: +5x and sad_24x32: +2x asm routines

[dnyaneshwar at multicorewareinc.com]

# HG changeset patch
# User Dnyaneshwar Gorade <dnyaneshwar at multicorewareinc.com>
# Date 1383218218 -19800
#      Thu Oct 31 16:46:58 2013 +0530
# Node ID 515b0af5eb805407d40ead87fd29a8c32118d3a2
# Parent  86ff1a3ec89720a73325148e8ac01ec1dbdab3c2
asm: Optimized sad_48x64: +5x and sad_24x32: +2x asm routines

diff -r 86ff1a3ec897 -r 515b0af5eb80 source/common/x86/sad-a.asm
--- a/source/common/x86/sad-a.asm	Thu Oct 31 16:21:35 2013 +0530
+++ b/source/common/x86/sad-a.asm	Thu Oct 31 16:46:58 2013 +0530
@@ -175,39 +175,37 @@
 %macro PROCESS_SAD_24x4 0
     movu        m1,  [r2]
     movq        m2,  [r2 + 16]
-    lea         r2,  [r2 + r3]
-    movu        m3,  [r2]
-    movq        m4,  [r2 + 16]
+    movu        m3,  [r2 + r3]
+    movq        m4,  [r2 + r3 + 16]
     psadbw      m1,  [r0]
     psadbw      m3,  [r0 + r1]
     paddd       m0,  m1
     paddd       m0,  m3
     movq        m1,  [r0 + 16]
-    lea         r0,  [r0 + r1]
-    movq        m3,  [r0 + 16]
+    movq        m3,  [r0 + r1 + 16]
     punpcklqdq  m2,  m4
     punpcklqdq  m1,  m3
     psadbw      m2, m1
     paddd       m0, m2
-    lea         r2,  [r2 + r3]
-    lea         r0,  [r0 + r1]
+    lea         r2,  [r2 + 2 * r3]
+    lea         r0,  [r0 + 2 * r1]
 
     movu        m1,  [r2]
     movq        m2,  [r2 + 16]
-    lea         r2,  [r2 + r3]
-    movu        m3,  [r2]
-    movq        m4,  [r2 + 16]
+    movu        m3,  [r2 + r3]
+    movq        m4,  [r2 + r3 + 16]
     psadbw      m1,  [r0]
     psadbw      m3,  [r0 + r1]
     paddd       m0,  m1
     paddd       m0,  m3
     movq        m1,  [r0 + 16]
-    lea         r0,  [r0 + r1]
-    movq        m3,  [r0 + 16]
+    movq        m3,  [r0 + r1 + 16]
     punpcklqdq  m2,  m4
     punpcklqdq  m1,  m3
     psadbw      m2, m1
     paddd       m0, m2
+    lea         r2,  [r2 + 2 * r3]
+    lea         r0,  [r0 + 2 * r1]
 %endmacro
 
 %macro PROCESS_SAD_32x4 0
@@ -255,8 +253,18 @@
     paddd   m1,  m2
     paddd   m0,  m1
     paddd   m0,  m3
-    lea     r2,  [r2 + r3]
-    lea     r0,  [r0 + r1]
+
+    movu    m1,  [r2 + r3]
+    movu    m2,  [r2 + r3 + 16]
+    movu    m3,  [r2 + r3 + 32]
+    psadbw  m1,  [r0 + r1]
+    psadbw  m2,  [r0 + r1 + 16]
+    psadbw  m3,  [r0 + r1 + 32]
+    paddd   m1,  m2
+    paddd   m0,  m1
+    paddd   m0,  m3
+    lea     r2,  [r2 + 2 * r3]
+    lea     r0,  [r0 + 2 * r1]
 
     movu    m1,  [r2]
     movu    m2,  [r2 + 16]
@@ -267,30 +275,18 @@
     paddd   m1,  m2
     paddd   m0,  m1
     paddd   m0,  m3
-    lea     r2,  [r2 + r3]
-    lea     r0,  [r0 + r1]
 
-    movu    m1,  [r2]
-    movu    m2,  [r2 + 16]
-    movu    m3,  [r2 + 32]
-    psadbw  m1,  [r0]
-    psadbw  m2,  [r0 + 16]
-    psadbw  m3,  [r0 + 32]
+    movu    m1,  [r2 + r3]
+    movu    m2,  [r2 + r3 + 16]
+    movu    m3,  [r2 + r3 + 32]
+    psadbw  m1,  [r0 + r1]
+    psadbw  m2,  [r0 + r1 + 16]
+    psadbw  m3,  [r0 + r1 + 32]
     paddd   m1,  m2
     paddd   m0,  m1
     paddd   m0,  m3
-    lea     r2,  [r2 + r3]
-    lea     r0,  [r0 + r1]
-
-    movu    m1,  [r2]
-    movu    m2,  [r2 + 16]
-    movu    m3,  [r2 + 32]
-    psadbw  m1,  [r0]
-    psadbw  m2,  [r0 + 16]
-    psadbw  m3,  [r0 + 32]
-    paddd   m1,  m2
-    paddd   m0,  m1
-    paddd   m0,  m3
+    lea     r2,  [r2 + 2 * r3]
+    lea     r0,  [r0 + 2 * r1]
 %endmacro
 
 %macro PROCESS_SAD_8x4 0
@@ -725,27 +721,17 @@
 ;-----------------------------------------------------------------------------
 ; int pixel_sad_48x64( uint8_t *, intptr_t, uint8_t *, intptr_t )
 ;-----------------------------------------------------------------------------
-cglobal pixel_sad_48x64, 4,4,5
+cglobal pixel_sad_48x64, 4,5,5
     pxor  m0,  m0
-    mov   r4,  64
+    mov   r4d, 4
 
 .loop
     PROCESS_SAD_48x4
-    lea     r2,  [r2 + r3]
-    lea     r0,  [r0 + r1]
-
     PROCESS_SAD_48x4
-    lea     r2,  [r2 + r3]
-    lea     r0,  [r0 + r1]
-
-    sub   r4,  8
-    cmp   r4,  8
-
-jnz .loop
     PROCESS_SAD_48x4
-    lea   r2,  [r2 + r3]
-    lea   r0,  [r0 + r1]
     PROCESS_SAD_48x4
+    dec  r4d
+    jnz .loop
 
     movhlps m1,  m0
     paddd   m0,  m1
@@ -755,24 +741,17 @@
 ;-----------------------------------------------------------------------------
 ; int pixel_sad_24x32( uint8_t *, intptr_t, uint8_t *, intptr_t )
 ;-----------------------------------------------------------------------------
-cglobal pixel_sad_24x32, 4,4,4
+cglobal pixel_sad_24x32, 4,5,4
     pxor  m0,  m0
-    mov   r4,  32
+    mov   r4d, 2
 
 .loop
     PROCESS_SAD_24x4
-    lea         r2,  [r2 + r3]
-    lea         r0,  [r0 + r1]
     PROCESS_SAD_24x4
-    lea         r2,  [r2 + r3]
-    lea         r0,  [r0 + r1]
-    sub   r4,  8
-    cmp   r4,  8
+    PROCESS_SAD_24x4
+    PROCESS_SAD_24x4
+    dec  r4d
 jnz .loop
-    PROCESS_SAD_24x4
-    lea         r2,  [r2 + r3]
-    lea         r0,  [r0 + r1]
-    PROCESS_SAD_24x4
 
     movhlps m1,  m0
     paddd   m0,  m1