[x265] [PATCH] blockcopy_pp[4x16:4x32], sse2 asm code optimization

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari
# Date 1422944898 -19800
# Node ID 059892f65db3e4c70017241ea847717e11be0124
# Parent  b0be54fb45cf1cbd3bf5a0543da34fb1a854b25f
blockcopy_pp[4x16:4x32], sse2 asm code optimization

improved, 222.93c -> 206.77c
          409.49c -> 392.13c

diff -r b0be54fb45cf -r 059892f65db3 source/common/x86/blockcopy8.asm
--- a/source/common/x86/blockcopy8.asm	Tue Feb 03 11:25:35 2015 +0530
+++ b/source/common/x86/blockcopy8.asm	Tue Feb 03 11:58:18 2015 +0530
@@ -181,37 +181,36 @@
 ;-----------------------------------------------------------------------------
 %macro BLOCKCOPY_PP_W4_H8 2
 INIT_XMM sse2
-cglobal blockcopy_pp_%1x%2, 4, 5, 4
+cglobal blockcopy_pp_%1x%2, 4, 7, 4
     mov    r4d,    %2/8
+    lea    r5,     [3 * r1]
+    lea    r6,     [3 * r3]
+
 .loop:
     movd     m0,     [r2]
     movd     m1,     [r2 + r3]
-    lea      r2,     [r2 + 2 * r3]
-    movd     m2,     [r2]
-    movd     m3,     [r2 + r3]
-
-    movd     [r0],                m0
-    movd     [r0 + r1],           m1
-    lea      r0,                  [r0 + 2 * r1]
-    movd     [r0],                m2
-    movd     [r0 + r1],           m3
-
-    lea       r0,     [r0 + 2 * r1]
-    lea       r2,     [r2 + 2 * r3]
+    movd     m2,     [r2 + 2 * r3]
+    movd     m3,     [r2 + r6]
+
+    movd     [r0],          m0
+    movd     [r0 + r1],     m1
+    movd     [r0 + 2 * r1], m2
+    movd     [r0 + r5],     m3
+
+    lea      r2,     [r2 + 4 * r3]
     movd     m0,     [r2]
     movd     m1,     [r2 + r3]
-    lea      r2,     [r2 + 2 * r3]
-    movd     m2,     [r2]
-    movd     m3,     [r2 + r3]
-
-    movd     [r0],                m0
-    movd     [r0 + r1],           m1
-    lea      r0,                  [r0 + 2 * r1]
-    movd     [r0],                m2
-    movd     [r0 + r1],           m3
-
-    lea       r0,                  [r0 + 2 * r1]
-    lea       r2,                  [r2 + 2 * r3]
+    movd     m2,     [r2 + 2 * r3]
+    movd     m3,     [r2 + r6]
+
+    lea      r0,            [r0 + 4 * r1]
+    movd     [r0],          m0
+    movd     [r0 + r1],     m1
+    movd     [r0 + 2 * r1], m2
+    movd     [r0 + r5],     m3
+
+    lea       r0,                  [r0 + 4 * r1]
+    lea       r2,                  [r2 + 4 * r3]
 
     dec       r4d
     jnz       .loop
@@ -219,7 +218,6 @@
 %endmacro
 
 BLOCKCOPY_PP_W4_H8 4, 16
-
 BLOCKCOPY_PP_W4_H8 4, 32
 
 ;-----------------------------------------------------------------------------