ryan/mozjpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Use intrinsics for loading aligned data in the IDCT functions. This has no effect on performance, but it makes it more obvious what that code is doing.

Browse Source 
git-svn-id: svn+ssh://svn.code.sf.net/p/libjpeg-turbo/code/trunk@1491 632fc199-4ca6-4c93-a231-07263d6284db
This commit is contained in:
DRC
2015-01-11 06:34:47 +00:00
parent f0abd46c3b
commit d71a6e0c25
2 changed files with 26 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
simd/jidctfst-altivec.c
View File

@@ -1,7 +1,7 @@
/* /*
* AltiVec optimizations for libjpeg-turbo * AltiVec optimizations for libjpeg-turbo
* *
* Copyright (C) 2014, D. R. Commander. * Copyright (C) 2014-2015, D. R. Commander.
* All rights reserved. * All rights reserved.
* This software is provided 'as-is', without any express or implied * This software is provided 'as-is', without any express or implied
* warranty. In no event will the authors be held liable for any damages * warranty. In no event will the authors be held liable for any damages
@@ -153,7 +153,7 @@ jsimd_idct_ifast_altivec (void * dct_table_, JCOEFPTR coef_block,
tmp3 = vec_or(tmp3, col7); tmp3 = vec_or(tmp3, col7);
tmp1 = vec_or(tmp1, tmp3); tmp1 = vec_or(tmp1, tmp3);
quant0 = *(__vector short *)&dct_table[0]; quant0 = vec_ld(0, dct_table);
col0 = vec_mladd(col0, quant0, zero); col0 = vec_mladd(col0, quant0, zero);
if (vec_all_eq(tmp1, zero)) { if (vec_all_eq(tmp1, zero)) {
@@ -170,13 +170,13 @@ jsimd_idct_ifast_altivec (void * dct_table_, JCOEFPTR coef_block,
} else { } else {
quant1 = *(__vector short *)&dct_table[8]; quant1 = vec_ld(16, dct_table);
quant2 = *(__vector short *)&dct_table[16]; quant2 = vec_ld(32, dct_table);
quant3 = *(__vector short *)&dct_table[24]; quant3 = vec_ld(48, dct_table);
quant4 = *(__vector short *)&dct_table[32]; quant4 = vec_ld(64, dct_table);
quant5 = *(__vector short *)&dct_table[40]; quant5 = vec_ld(80, dct_table);
quant6 = *(__vector short *)&dct_table[48]; quant6 = vec_ld(96, dct_table);
quant7 = *(__vector short *)&dct_table[56]; quant7 = vec_ld(112, dct_table);
col1 = vec_mladd(col1, quant1, zero); col1 = vec_mladd(col1, quant1, zero);
col2 = vec_mladd(col2, quant2, zero); col2 = vec_mladd(col2, quant2, zero);

34
simd/jidctint-altivec.c
View File

@@ -1,7 +1,7 @@
/* /*
* AltiVec optimizations for libjpeg-turbo * AltiVec optimizations for libjpeg-turbo
* *
* Copyright (C) 2014, D. R. Commander. * Copyright (C) 2014-2015, D. R. Commander.
* All rights reserved. * All rights reserved.
* This software is provided 'as-is', without any express or implied * This software is provided 'as-is', without any express or implied
* warranty. In no event will the authors be held liable for any damages * warranty. In no event will the authors be held liable for any damages
@@ -246,14 +246,14 @@ jsimd_idct_islow_altivec (void * dct_table_, JCOEFPTR coef_block,
/* Pass 1: process columns */ /* Pass 1: process columns */
col0 = *(__vector short *)&coef_block[0]; col0 = vec_ld(0, coef_block);
col1 = *(__vector short *)&coef_block[8]; col1 = vec_ld(16, coef_block);
col2 = *(__vector short *)&coef_block[16]; col2 = vec_ld(32, coef_block);
col3 = *(__vector short *)&coef_block[24]; col3 = vec_ld(48, coef_block);
col4 = *(__vector short *)&coef_block[32]; col4 = vec_ld(64, coef_block);
col5 = *(__vector short *)&coef_block[40]; col5 = vec_ld(80, coef_block);
col6 = *(__vector short *)&coef_block[48]; col6 = vec_ld(96, coef_block);
col7 = *(__vector short *)&coef_block[56]; col7 = vec_ld(112, coef_block);
tmp1 = vec_or(col1, col2); tmp1 = vec_or(col1, col2);
tmp2 = vec_or(col3, col4); tmp2 = vec_or(col3, col4);
@@ -262,7 +262,7 @@ jsimd_idct_islow_altivec (void * dct_table_, JCOEFPTR coef_block,
tmp3 = vec_or(tmp3, col7); tmp3 = vec_or(tmp3, col7);
tmp1 = vec_or(tmp1, tmp3); tmp1 = vec_or(tmp1, tmp3);
quant0 = *(__vector short *)&dct_table[0]; quant0 = vec_ld(0, dct_table);
col0 = vec_mladd(col0, quant0, zero16); col0 = vec_mladd(col0, quant0, zero16);
if (vec_all_eq(tmp1, zero16)) { if (vec_all_eq(tmp1, zero16)) {
@@ -281,13 +281,13 @@ jsimd_idct_islow_altivec (void * dct_table_, JCOEFPTR coef_block,
} else { } else {
quant1 = *(__vector short *)&dct_table[8]; quant1 = vec_ld(16, dct_table);
quant2 = *(__vector short *)&dct_table[16]; quant2 = vec_ld(32, dct_table);
quant3 = *(__vector short *)&dct_table[24]; quant3 = vec_ld(48, dct_table);
quant4 = *(__vector short *)&dct_table[32]; quant4 = vec_ld(64, dct_table);
quant5 = *(__vector short *)&dct_table[40]; quant5 = vec_ld(80, dct_table);
quant6 = *(__vector short *)&dct_table[48]; quant6 = vec_ld(96, dct_table);
quant7 = *(__vector short *)&dct_table[56]; quant7 = vec_ld(112, dct_table);
col1 = vec_mladd(col1, quant1, zero16); col1 = vec_mladd(col1, quant1, zero16);
col2 = vec_mladd(col2, quant2, zero16); col2 = vec_mladd(col2, quant2, zero16);