[x265] [PATCH] asm: assembly code for pixel_sad_12x16

[Praveen Tiwari]

---------- Forwarded message ----------
From: <dnyaneshwar at multicorewareinc.com>
Date: Wed, Oct 30, 2013 at 7:47 PM
Subject: [x265] [PATCH] asm: assembly code for pixel_sad_12x16
To: x265-devel at videolan.org


# HG changeset patch
# User Dnyaneshwar Gorade <dnyaneshwar at multicorewareinc.com>
# Date 1383142575 -19800
#      Wed Oct 30 19:46:15 2013 +0530
# Node ID 5037cc891114619e32ceeff332884d0abfd138fd
# Parent  62a51fe2fcbfd76fc8476a6f714f961b3f3f23ef
asm: assembly code for pixel_sad_12x16

diff -r 62a51fe2fcbf -r 5037cc891114 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp      Wed Oct 30 18:11:01 2013
+0530
+++ b/source/common/x86/asm-primitives.cpp      Wed Oct 30 19:46:15 2013
+0530
@@ -253,6 +253,7 @@

         p.sad[LUMA_48x64]  = x265_pixel_sad_48x64_sse2;
         p.sad[LUMA_24x32]  = x265_pixel_sad_24x32_sse2;
+        p.sad[LUMA_12x16]  = x265_pixel_sad_12x16_sse2;

         ASSGN_SSE(sse2);
         INIT2(sad, _sse2);
diff -r 62a51fe2fcbf -r 5037cc891114 source/common/x86/pixel.h
--- a/source/common/x86/pixel.h Wed Oct 30 18:11:01 2013 +0530
+++ b/source/common/x86/pixel.h Wed Oct 30 19:46:15 2013 +0530
@@ -53,6 +53,7 @@
     ret x265_pixel_ ## name ## _64x64_ ## suffix args; \
     ret x265_pixel_ ## name ## _48x64_ ## suffix args; \
     ret x265_pixel_ ## name ## _24x32_ ## suffix args; \
+    ret x265_pixel_ ## name ## _12x16_ ## suffix args; \

 #define DECL_X1(name, suffix) \
     DECL_PIXELS(int, name, suffix, (pixel *, intptr_t, pixel *, intptr_t))
diff -r 62a51fe2fcbf -r 5037cc891114 source/common/x86/sad-a.asm
--- a/source/common/x86/sad-a.asm       Wed Oct 30 18:11:01 2013 +0530
+++ b/source/common/x86/sad-a.asm       Wed Oct 30 19:46:15 2013 +0530
@@ -31,8 +31,9 @@

 SECTION_RODATA 32

+MSK:                  db
255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0
 pb_shuf8x8c2: times 2 db 0,0,0,0,8,8,8,8,-1,-1,-1,-1,-1,-1,-1,-1
-hpred_shuf: db 0,0,2,2,8,8,10,10,1,1,3,3,9,9,11,11
+hpred_shuf:           db 0,0,2,2,8,8,10,10,1,1,3,3,9,9,11,11

 SECTION .text

@@ -119,6 +120,39 @@
     RET
 %endmacro

+%macro PROCESS_SAD_12x4 0
+    movu    m1,  [r2]
+    movu    m2,  [r0]
+    pand    m1,  m4
+    pand    m2,  m4
+    psadbw  m1,  m2
+    paddd   m0,  m1
+    lea     r2,  [r2 + r3]
+    lea     r0,  [r0 + r1]
+    movu    m1,  [r2]
+    movu    m2,  [r0]
+    pand    m1,  m4
+    pand    m2,  m4
+    psadbw  m1,  m2
+    paddd   m0,  m1
>>+    lea     r2,  [r2 + r3]
>>+    lea     r0,  [r0 + r1]
>>+    movu    m1,  [r2]
>>+    movu    m2,  [r0]

we don't need to load address every time when we are adding stride to it.
we should try to calculate address first using multiply by 1, 2, 4, or 8 if
it not the case then we should load it.
 like above four instruction can be replaced with these two only.

movu    m1,  [r2 + 2 * r3]
movu    m2,  [r0 + 2 * r1]

+    pand    m1,  m4
+    pand    m2,  m4
+    psadbw  m1,  m2
+    paddd   m0,  m1
+    lea     r2,  [r2 + r3]
+    lea     r0,  [r0 + r1]
+    movu    m1,  [r2]
+    movu    m2,  [r0]
+    pand    m1,  m4
+    pand    m2,  m4
+    psadbw  m1,  m2
+    paddd   m0,  m1
+%endmacro
+
 %macro PROCESS_SAD_16x4 0
     movu    m1,  [r2]
     movu    m2,  [r2 + r3]
@@ -1007,6 +1041,29 @@
     movd    eax, m0
     RET

+;-----------------------------------------------------------------------------
+; int pixel_sad_12x16( uint8_t *, intptr_t, uint8_t *, intptr_t )
+;-----------------------------------------------------------------------------
+cglobal pixel_sad_12x16, 4,4,4
+    mova  m4,  [MSK]
+    pxor  m0,  m0
+
+    PROCESS_SAD_12x4
+    lea         r2,  [r2 + r3]
+    lea         r0,  [r0 + r1]
+    PROCESS_SAD_12x4
+    lea         r2,  [r2 + r3]
+    lea         r0,  [r0 + r1]
+    PROCESS_SAD_12x4
+    lea         r2,  [r2 + r3]
+    lea         r0,  [r0 + r1]
+    PROCESS_SAD_12x4
+
+    movhlps m1,  m0
+    paddd   m0,  m1
+    movd    eax, m0
+    RET
+
 %endmacro

overuse of lea  instruction please eliminate them, use available registers
to save loads operations.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131030/42d3b5bf/attachment.html>