ARMv6 optimised pix_abs16

Originally committed as revision 21697 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Måns Rullgård 2010-02-09 16:13:26 +00:00
parent 38e016a7c9
commit e6056a9008
2 changed files with 36 additions and 0 deletions

View File

@ -286,3 +286,32 @@ function ff_add_pixels_clamped_armv6, export=1
bgt 1b
pop {r4-r8,pc}
.endfunc
function ff_pix_abs16_armv6, export=1
ldr r0, [sp]
push {r4-r9, lr}
mov r12, #0
mov lr, #0
ldm r1, {r4-r7}
ldr r8, [r2]
1:
ldr r9, [r2, #4]
pld [r1, r3]
usada8 r12, r4, r8, r12
ldr r8, [r2, #8]
pld [r2, r3]
usada8 lr, r5, r9, lr
ldr r9, [r2, #12]
usada8 r12, r6, r8, r12
subs r0, r0, #1
usada8 lr, r7, r9, lr
beq 2f
add r1, r1, r3
ldm r1, {r4-r7}
add r2, r2, r3
ldr r8, [r2]
b 1b
2:
add r0, r12, lr
pop {r4-r9, pc}
.endfunc

View File

@ -50,6 +50,9 @@ void ff_add_pixels_clamped_armv6(const DCTELEM *block,
uint8_t *restrict pixels,
int line_size);
int ff_pix_abs16_armv6(void *s, uint8_t *blk1, uint8_t *blk2,
int line_size, int h);
void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
{
if (!avctx->lowres && (avctx->idct_algo == FF_IDCT_AUTO ||
@ -82,4 +85,8 @@ void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
c->avg_pixels_tab[1][0] = ff_avg_pixels8_armv6;
c->add_pixels_clamped = ff_add_pixels_clamped_armv6;
c->pix_abs[0][0] = ff_pix_abs16_armv6;
c->sad[0] = ff_pix_abs16_armv6;
}