virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-21 10:55:51 +02:00
Code Issues Releases Activity

adler32: rewrite using integer SIMD.

Browse Source 
about twice as fast as before.
the not CONFIG_SMALL case is also droped as it is not faster than the
CONFIG_SMALL case.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2012-02-04 07:52:31 +01:00
parent 5b0103590c
commit 26585d2a7f
1 changed files with 32 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
libavutil/adler32.c
View File

@ -24,6 +24,7 @@
#include "config.h"
#include "adler32.h"
#include "common.h"
#include "intreadwrite.h"
#define BASE 65521L /* largest prime smaller than 65536 */
@ -38,22 +39,42 @@ unsigned long av_adler32_update(unsigned long adler, const uint8_t * buf,
unsigned long s2 = adler >> 16;
while (len > 0) {
unsigned len2 = FFMIN((len-1) & ~15, 2048);
#if HAVE_FAST_64BIT && HAVE_FAST_UNALIGNED && !CONFIG_SMALL
unsigned len2 = FFMIN((len-1) & ~7, 23*8);
if (len2) {
uint64_t a1= 0;
uint64_t a2= 0;
uint64_t b1= 0;
uint64_t b2= 0;
len -= len2;
#if CONFIG_SMALL
while (len2 >= 4) {
DO4(buf);
len2 -= 4;
}
s2 += s1*len2;
while (len2 >= 8) {
uint64_t v = AV_RN64(buf);
a2 += a1;
b2 += b1;
a1 += v &0x00FF00FF00FF00FF;
b1 += (v>>8)&0x00FF00FF00FF00FF;
len2 -= 8;
buf+=8;
}
s1 += ((a1+b1)*0x1000100010001)>>48;
s2 += ((((a2&0xFFFF0000FFFF)+(b2&0xFFFF0000FFFF)+((a2>>16)&0xFFFF0000FFFF)+((b2>>16)&0xFFFF0000FFFF))*0x800000008)>>32)
#if HAVE_BIGENDIAN
+ 2*((b1*0x1000200030004)>>48)
+ ((a1*0x1000100010001)>>48)
+ 2*((a1*0x0000100020003)>>48);
#else
while (len2 >= 16) {
DO16(buf);
len2 -= 16;
}
+ 2*((a1*0x4000300020001)>>48)
+ ((b1*0x1000100010001)>>48)
+ 2*((b1*0x3000200010000)>>48);
#endif
}
#else
while (len > 4 && s2 < (1U << 31)) {
DO4(buf);
len -= 4;
}
#endif
DO1(buf); len--;
s1 %= BASE;
s2 %= BASE;