generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge "rename loop filter functions"

Browse Source
This commit is contained in:
Jim Bankoski 2013-12-17 18:56:09 -08:00 committed by Gerrit Code Review
commit 9d754dcca8
14 changed files with 356 additions and 407 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
vp9/common/arm/neon/vp9_loopfilter_16_neon.asm
View File

@ -8,18 +8,18 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_loop_filter_horizontal_edge_16_neon|
EXPORT |vp9_lpf_horizontal_4_dual_neon|
ARM
AREA ||.text||, CODE, READONLY, ALIGN=2
;void vp9_loop_filter_horizontal_edge_16_neon(uint8_t *s, int p,
; const uint8_t *blimit0,
; const uint8_t *limit0,
; const uint8_t *thresh0,
; const uint8_t *blimit1,
; const uint8_t *limit1,
; const uint8_t *thresh1)
;void vp9_lpf_horizontal_4_dual_neon(uint8_t *s, int p,
; const uint8_t *blimit0,
; const uint8_t *limit0,
; const uint8_t *thresh0,
; const uint8_t *blimit1,
; const uint8_t *limit1,
; const uint8_t *thresh1)
; r0 uint8_t *s,
; r1 int p,
; r2 const uint8_t *blimit0,
@ -29,7 +29,7 @@
; sp+8 const uint8_t *limit1,
; sp+12 const uint8_t *thresh1,
|vp9_loop_filter_horizontal_edge_16_neon| PROC
|vp9_lpf_horizontal_4_dual_neon| PROC
push {lr}
ldr r12, [sp, #4] ; load thresh0
@ -76,7 +76,7 @@
vpop {d8-d15} ; restore neon registers
pop {pc}
ENDP ; |vp9_loop_filter_horizontal_edge_16_neon|
ENDP ; |vp9_lpf_horizontal_4_dual_neon|
; void vp9_loop_filter_neon_16();
; This is a helper function for the loopfilters. The invidual functions do the

67
vp9/common/arm/neon/vp9_loopfilter_16_neon.c
View File

@ -10,44 +10,43 @@
#include "./vp9_rtcd.h"
void vp9_mbloop_filter_horizontal_edge_16_neon(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_horizontal_edge(s, p, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_horizontal_edge(s + 8, p, blimit1, limit1, thresh1, 1);
void vp9_lpf_horizontal_8_dual_neon(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_horizontal_8(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_horizontal_8(s + 8, p, blimit1, limit1, thresh1, 1);
}
void vp9_loop_filter_vertical_edge_16_neon(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_loop_filter_vertical_edge_neon(s, p, blimit0, limit0, thresh0, 1);
vp9_loop_filter_vertical_edge_neon(s + 8 * p, p, blimit1, limit1, thresh1, 1);
void vp9_lpf_vertical_4_dual_neon(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_4_neon(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_4_neon(s + 8 * p, p, blimit1, limit1, thresh1, 1);
}
void vp9_mbloop_filter_vertical_edge_16_neon(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_vertical_edge_neon(s, p, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_vertical_edge_neon(s + 8 * p, p, blimit1, limit1, thresh1,
1);
void vp9_lpf_vertical_8_dual_neon(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_8_neon(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_8_neon(s + 8 * p, p, blimit1, limit1, thresh1, 1);
}
void vp9_mb_lpf_vertical_edge_w_16_neon(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
vp9_mb_lpf_vertical_edge_w_neon(s, p, blimit, limit, thresh);
vp9_mb_lpf_vertical_edge_w_neon(s + 8 * p, p, blimit, limit, thresh);
void vp9_lpf_vertical_16_dual_neon(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
vp9_lpf_vertical_16_neon(s, p, blimit, limit, thresh);
vp9_lpf_vertical_16_neon(s + 8 * p, p, blimit, limit, thresh);
}

70
vp9/common/arm/neon/vp9_loopfilter_neon.asm
View File

@ -8,10 +8,10 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_loop_filter_horizontal_edge_neon|
EXPORT |vp9_loop_filter_vertical_edge_neon|
EXPORT |vp9_mbloop_filter_horizontal_edge_neon|
EXPORT |vp9_mbloop_filter_vertical_edge_neon|
EXPORT |vp9_lpf_horizontal_4_neon|
EXPORT |vp9_lpf_vertical_4_neon|
EXPORT |vp9_lpf_horizontal_8_neon|
EXPORT |vp9_lpf_vertical_8_neon|
ARM
AREA ||.text||, CODE, READONLY, ALIGN=2
@ -21,12 +21,12 @@
; TODO(fgalligan): See about removing the count code as this function is only
; called with a count of 1.
;
; void vp9_loop_filter_horizontal_edge_neon(uint8_t *s,
; int p /* pitch */,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
; void vp9_lpf_horizontal_4_neon(uint8_t *s,
; int p /* pitch */,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
;
; r0 uint8_t *s,
; r1 int p, /* pitch */
@ -34,7 +34,7 @@
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
; sp+4 int count
|vp9_loop_filter_horizontal_edge_neon| PROC
|vp9_lpf_horizontal_4_neon| PROC
push {lr}
vld1.8 {d0[]}, [r2] ; duplicate *blimit
@ -77,19 +77,19 @@ count_lf_h_loop
end_vp9_lf_h_edge
pop {pc}
ENDP ; |vp9_loop_filter_horizontal_edge_neon|
ENDP ; |vp9_lpf_horizontal_4_neon|
; Currently vp9 only works on iterations 8 at a time. The vp8 loop filter
; works on 16 iterations at a time.
; TODO(fgalligan): See about removing the count code as this function is only
; called with a count of 1.
;
; void vp9_loop_filter_vertical_edge_neon(uint8_t *s,
; int p /* pitch */,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
; void vp9_lpf_vertical_4_neon(uint8_t *s,
; int p /* pitch */,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
;
; r0 uint8_t *s,
; r1 int p, /* pitch */
@ -97,7 +97,7 @@ end_vp9_lf_h_edge
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
; sp+4 int count
|vp9_loop_filter_vertical_edge_neon| PROC
|vp9_lpf_vertical_4_neon| PROC
push {lr}
vld1.8 {d0[]}, [r2] ; duplicate *blimit
@ -158,7 +158,7 @@ count_lf_v_loop
end_vp9_lf_v_edge
pop {pc}
ENDP ; |vp9_loop_filter_vertical_edge_neon|
ENDP ; |vp9_lpf_vertical_4_neon|
; void vp9_loop_filter_neon();
; This is a helper function for the loopfilters. The invidual functions do the
@ -276,18 +276,18 @@ end_vp9_lf_v_edge
bx lr
ENDP ; |vp9_loop_filter_neon|
; void vp9_mbloop_filter_horizontal_edge_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
; void vp9_lpf_horizontal_8_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
; r0 uint8_t *s,
; r1 int p, /* pitch */
; r2 const uint8_t *blimit,
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
; sp+4 int count
|vp9_mbloop_filter_horizontal_edge_neon| PROC
|vp9_lpf_horizontal_8_neon| PROC
push {r4-r5, lr}
vld1.8 {d0[]}, [r2] ; duplicate *blimit
@ -333,14 +333,14 @@ count_mblf_h_loop
end_vp9_mblf_h_edge
pop {r4-r5, pc}
ENDP ; |vp9_mbloop_filter_horizontal_edge_neon|
ENDP ; |vp9_lpf_horizontal_8_neon|
; void vp9_mbloop_filter_vertical_edge_neon(uint8_t *s,
; int pitch,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
; void vp9_lpf_vertical_8_neon(uint8_t *s,
; int pitch,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh,
; int count)
;
; r0 uint8_t *s,
; r1 int pitch,
@ -348,7 +348,7 @@ end_vp9_mblf_h_edge
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
; sp+4 int count
|vp9_mbloop_filter_vertical_edge_neon| PROC
|vp9_lpf_vertical_8_neon| PROC
push {r4-r5, lr}
vld1.8 {d0[]}, [r2] ; duplicate *blimit
@ -420,7 +420,7 @@ count_mblf_v_loop
end_vp9_mblf_v_edge
pop {r4-r5, pc}
ENDP ; |vp9_mbloop_filter_vertical_edge_neon|
ENDP ; |vp9_lpf_vertical_8_neon|
; void vp9_mbloop_filter_neon();
; This is a helper function for the loopfilters. The invidual functions do the

30
vp9/common/arm/neon/vp9_mb_lpf_neon.asm
View File

@ -8,23 +8,23 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_mb_lpf_horizontal_edge_w_neon|
EXPORT |vp9_mb_lpf_vertical_edge_w_neon|
EXPORT |vp9_lpf_horizontal_16_neon|
EXPORT |vp9_lpf_vertical_16_neon|
ARM
AREA ||.text||, CODE, READONLY, ALIGN=2
; void vp9_mb_lpf_horizontal_edge_w_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh
; int count)
; void vp9_lpf_horizontal_16_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh
; int count)
; r0 uint8_t *s,
; r1 int p, /* pitch */
; r2 const uint8_t *blimit,
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
|vp9_mb_lpf_horizontal_edge_w_neon| PROC
|vp9_lpf_horizontal_16_neon| PROC
push {r4-r8, lr}
vpush {d8-d15}
ldr r4, [sp, #88] ; load thresh
@ -115,18 +115,18 @@ h_next
vpop {d8-d15}
pop {r4-r8, pc}
ENDP ; |vp9_mb_lpf_horizontal_edge_w_neon|
ENDP ; |vp9_lpf_horizontal_16_neon|
; void vp9_mb_lpf_vertical_edge_w_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh)
; void vp9_lpf_vertical_16_neon(uint8_t *s, int p,
; const uint8_t *blimit,
; const uint8_t *limit,
; const uint8_t *thresh)
; r0 uint8_t *s,
; r1 int p, /* pitch */
; r2 const uint8_t *blimit,
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
|vp9_mb_lpf_vertical_edge_w_neon| PROC
|vp9_lpf_vertical_16_neon| PROC
push {r4-r8, lr}
vpush {d8-d15}
ldr r4, [sp, #88] ; load thresh
@ -279,7 +279,7 @@ v_end
vpop {d8-d15}
pop {r4-r8, pc}
ENDP ; |vp9_mb_lpf_vertical_edge_w_neon|
ENDP ; |vp9_lpf_vertical_16_neon|
; void vp9_wide_mbfilter_neon();
; This is a helper function for the loopfilters. The invidual functions do the

110
vp9/common/mips/dspr2/vp9_loopfilter_filters_dspr2.c
View File

@ -20,12 +20,12 @@
#include "vp9/common/mips/dspr2/vp9_loopfilter_filters_dspr2.h"
#if HAVE_DSPR2
void vp9_loop_filter_horizontal_edge_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_4_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
uint8_t i;
uint32_t mask;
uint32_t hev;
@ -114,12 +114,12 @@ void vp9_loop_filter_horizontal_edge_dspr2(unsigned char *s,
}
}
void vp9_loop_filter_vertical_edge_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_vertical_4_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
uint8_t i;
uint32_t mask, hev;
uint32_t pm1, p0, p1, p2, p3, p4, p5, p6;
@ -307,58 +307,56 @@ void vp9_loop_filter_vertical_edge_dspr2(unsigned char *s,
}
}
void vp9_loop_filter_horizontal_edge_16_dspr2(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_loop_filter_horizontal_edge_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_loop_filter_horizontal_edge_dspr2(s + 8, p, blimit1, limit1, thresh1, 1);
void vp9_lpf_horizontal_4_dual_dspr2(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_horizontal_4_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_horizontal_4_dspr2(s + 8, p, blimit1, limit1, thresh1, 1);
}
void vp9_mbloop_filter_horizontal_edge_16_dspr2(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_horizontal_edge_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_horizontal_edge_dspr2(s + 8, p, blimit1, limit1, thresh1,
1);
void vp9_lpf_horizontal_8_dual_dspr2(uint8_t *s, int p /* pitch */,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_horizontal_8_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_horizontal_8_dspr2(s + 8, p, blimit1, limit1, thresh1, 1);
}
void vp9_loop_filter_vertical_edge_16_dspr2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_loop_filter_vertical_edge_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_loop_filter_vertical_edge_dspr2(s + 8 * p, p, blimit1, limit1, thresh1,
1);
void vp9_lpf_vertical_4_dual_dspr2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_4_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_4_dspr2(s + 8 * p, p, blimit1, limit1, thresh1, 1);
}
void vp9_mbloop_filter_vertical_edge_16_dspr2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_vertical_edge_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_vertical_edge_dspr2(s + 8 * p, p, blimit1, limit1, thresh1,
void vp9_lpf_vertical_8_dual_dspr2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_8_dspr2(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_8_dspr2(s + 8 * p, p, blimit1, limit1, thresh1,
1);
}
void vp9_mb_lpf_vertical_edge_w_16_dspr2(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
vp9_mb_lpf_vertical_edge_w_dspr2(s, p, blimit, limit, thresh);
vp9_mb_lpf_vertical_edge_w_dspr2(s + 8 * p, p, blimit, limit, thresh);
void vp9_lpf_vertical_16_dual_dspr2(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
vp9_lpf_vertical_16_dspr2(s, p, blimit, limit, thresh);
vp9_lpf_vertical_16_dspr2(s + 8 * p, p, blimit, limit, thresh);
}
#endif // #if HAVE_DSPR2

24
vp9/common/mips/dspr2/vp9_mbloop_loopfilter_dspr2.c
View File

@ -20,12 +20,12 @@
#include "vp9/common/mips/dspr2/vp9_loopfilter_filters_dspr2.h"
#if HAVE_DSPR2
void vp9_mbloop_filter_horizontal_edge_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_8_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
uint32_t mask;
uint32_t hev, flat;
uint8_t i;
@ -319,12 +319,12 @@ void vp9_mbloop_filter_horizontal_edge_dspr2(unsigned char *s,
}
}
void vp9_mbloop_filter_vertical_edge_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_vertical_8_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
uint8_t i;
uint32_t mask, hev, flat;
uint8_t *s1, *s2, *s3, *s4;

12
vp9/common/mips/dspr2/vp9_mblpf_horiz_loopfilter_dspr2.c
View File

@ -20,12 +20,12 @@
#include "vp9/common/mips/dspr2/vp9_loopfilter_filters_dspr2.h"
#if HAVE_DSPR2
void vp9_mb_lpf_horizontal_edge_w_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_16_dspr2(unsigned char *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
uint32_t mask;
uint32_t hev, flat, flat2;
uint8_t i;

10
vp9/common/mips/dspr2/vp9_mblpf_vert_loopfilter_dspr2.c
View File

@ -20,11 +20,11 @@
#include "vp9/common/mips/dspr2/vp9_loopfilter_filters_dspr2.h"
#if HAVE_DSPR2
void vp9_mb_lpf_vertical_edge_w_dspr2(uint8_t *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
void vp9_lpf_vertical_16_dspr2(uint8_t *s,
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
uint8_t i;
uint32_t mask, hev, flat, flat2;
uint8_t *s1, *s2, *s3, *s4;

140
vp9/common/vp9_loopfilter.c
View File

@ -339,56 +339,56 @@ static void filter_selectively_vert_row2(PLANE_TYPE plane_type,
if (mask & 1) {
if ((mask_16x16_0 | mask_16x16_1) & 1) {
if ((mask_16x16_0 & mask_16x16_1) & 1) {
vp9_mb_lpf_vertical_edge_w_16(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr);
vp9_lpf_vertical_16_dual(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr);
} else if (mask_16x16_0 & 1) {
vp9_mb_lpf_vertical_edge_w(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr);
vp9_lpf_vertical_16(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr);
} else {
vp9_mb_lpf_vertical_edge_w(s + 8 *pitch, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr);
vp9_lpf_vertical_16(s + 8 *pitch, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr);
}
}
if ((mask_8x8_0 | mask_8x8_1) & 1) {
if ((mask_8x8_0 & mask_8x8_1) & 1) {
vp9_mbloop_filter_vertical_edge_16(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim,
lfi1->lim, lfi1->hev_thr);
vp9_lpf_vertical_8_dual(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim, lfi1->lim,
lfi1->hev_thr);
} else if (mask_8x8_0 & 1) {
vp9_mbloop_filter_vertical_edge(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, 1);
vp9_lpf_vertical_8(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr,
1);
} else {
vp9_mbloop_filter_vertical_edge(s + 8 *pitch, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr, 1);
vp9_lpf_vertical_8(s + 8 * pitch, pitch, lfi1->mblim, lfi1->lim,
lfi1->hev_thr, 1);
}
}
if ((mask_4x4_0 | mask_4x4_1) & 1) {
if ((mask_4x4_0 & mask_4x4_1) & 1) {
vp9_loop_filter_vertical_edge_16(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim,
lfi1->lim, lfi1->hev_thr);
vp9_lpf_vertical_4_dual(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim, lfi1->lim,
lfi1->hev_thr);
} else if (mask_4x4_0 & 1) {
vp9_loop_filter_vertical_edge(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, 1);
vp9_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr,
1);
} else {
vp9_loop_filter_vertical_edge(s + 8 *pitch, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr, 1);
vp9_lpf_vertical_4(s + 8 * pitch, pitch, lfi1->mblim, lfi1->lim,
lfi1->hev_thr, 1);
}
}
if ((mask_4x4_int_0 | mask_4x4_int_1) & 1) {
if ((mask_4x4_int_0 & mask_4x4_int_1) & 1) {
vp9_loop_filter_vertical_edge_16(s + 4, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim,
lfi1->lim, lfi1->hev_thr);
vp9_lpf_vertical_4_dual(s + 4, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, lfi1->mblim, lfi1->lim,
lfi1->hev_thr);
} else if (mask_4x4_int_0 & 1) {
vp9_loop_filter_vertical_edge(s + 4, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, 1);
vp9_lpf_vertical_4(s + 4, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, 1);
} else {
vp9_loop_filter_vertical_edge(s + 8 *pitch + 4, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr, 1);
vp9_lpf_vertical_4(s + 8 * pitch + 4, pitch, lfi1->mblim, lfi1->lim,
lfi1->hev_thr, 1);
}
}
}
@ -424,81 +424,73 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
if (mask & 1) {
if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) {
vp9_mb_lpf_horizontal_edge_w(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 2);
vp9_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 2);
count = 2;
} else {
vp9_mb_lpf_horizontal_edge_w(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
}
} else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) {
// Next block's thresholds
const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
vp9_mbloop_filter_horizontal_edge_16(s, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr,
lfin->mblim, lfin->lim,
lfin->hev_thr);
vp9_lpf_horizontal_8_dual(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, lfin->mblim, lfin->lim,
lfin->hev_thr);
if ((mask_4x4_int & 3) == 3) {
vp9_loop_filter_horizontal_edge_16(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr,
lfin->mblim, lfin->lim,
lfin->hev_thr);
vp9_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, lfin->mblim,
lfin->lim, lfin->hev_thr);
} else {
if (mask_4x4_int & 1)
vp9_loop_filter_horizontal_edge(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, 1);
vp9_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
else if (mask_4x4_int & 2)
vp9_loop_filter_horizontal_edge(s + 8 + 4 * pitch, pitch,
lfin->mblim, lfin->lim,
lfin->hev_thr, 1);
vp9_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1);
}
count = 2;
} else {
vp9_mbloop_filter_horizontal_edge(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_horizontal_8(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
if (mask_4x4_int & 1)
vp9_loop_filter_horizontal_edge(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, 1);
vp9_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
}
} else if (mask_4x4 & 1) {
if ((mask_4x4 & 3) == 3) {
// Next block's thresholds
const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
vp9_loop_filter_horizontal_edge_16(s, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr,
lfin->mblim, lfin->lim,
lfin->hev_thr);
vp9_lpf_horizontal_4_dual(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, lfin->mblim, lfin->lim,
lfin->hev_thr);
if ((mask_4x4_int & 3) == 3) {
vp9_loop_filter_horizontal_edge_16(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr,
lfin->mblim, lfin->lim,
lfin->hev_thr);
vp9_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, lfin->mblim,
lfin->lim, lfin->hev_thr);
} else {
if (mask_4x4_int & 1)
vp9_loop_filter_horizontal_edge(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, 1);
vp9_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
else if (mask_4x4_int & 2)
vp9_loop_filter_horizontal_edge(s + 8 + 4 * pitch, pitch,
lfin->mblim, lfin->lim,
lfin->hev_thr, 1);
vp9_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1);
}
count = 2;
} else {
vp9_loop_filter_horizontal_edge(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
if (mask_4x4_int & 1)
vp9_loop_filter_horizontal_edge(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, 1);
if (mask_4x4_int & 1)
vp9_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
}
} else if (mask_4x4_int & 1) {
vp9_loop_filter_horizontal_edge(s + 4 * pitch, pitch, lfi->mblim,
lfi->lim, lfi->hev_thr, 1);
vp9_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
}
}
s += 8 * count;
@ -919,19 +911,15 @@ static void filter_selectively_vert(uint8_t *s, int pitch,
if (mask & 1) {
if (mask_16x16 & 1) {
vp9_mb_lpf_vertical_edge_w(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr);
vp9_lpf_vertical_16(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
} else if (mask_8x8 & 1) {
vp9_mbloop_filter_vertical_edge(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
} else if (mask_4x4 & 1) {
vp9_loop_filter_vertical_edge(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
}
}
if (mask_4x4_int & 1)
vp9_loop_filter_vertical_edge(s + 4, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1);
vp9_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
s += 8;
lfl += 1;
mask_16x16 >>= 1;

112
vp9/common/vp9_loopfilter_filters.c
View File

@ -101,11 +101,9 @@ static INLINE void filter4(int8_t mask, uint8_t hev, uint8_t *op1,
*op1 = signed_char_clamp(ps1 + filter) ^ 0x80;
}
void vp9_loop_filter_horizontal_edge_c(uint8_t *s, int p /* pitch */,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_4_c(uint8_t *s, int p /* pitch */,
const uint8_t *blimit, const uint8_t *limit,
const uint8_t *thresh, int count) {
int i;
// loop filter designed to work using chars so that we can make maximum use
@ -121,22 +119,17 @@ void vp9_loop_filter_horizontal_edge_c(uint8_t *s, int p /* pitch */,
}
}
void vp9_loop_filter_horizontal_edge_16_c(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_loop_filter_horizontal_edge_c(s, p, blimit0, limit0, thresh0, 1);
vp9_loop_filter_horizontal_edge_c(s + 8, p, blimit1, limit1, thresh1, 1);
void vp9_lpf_horizontal_4_dual_c(uint8_t *s, int p, const uint8_t *blimit0,
const uint8_t *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_horizontal_4_c(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_horizontal_4_c(s + 8, p, blimit1, limit1, thresh1, 1);
}
void vp9_loop_filter_vertical_edge_c(uint8_t *s, int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_vertical_4_c(uint8_t *s, int pitch, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh,
int count) {
int i;
// loop filter designed to work using chars so that we can make maximum use
@ -152,15 +145,12 @@ void vp9_loop_filter_vertical_edge_c(uint8_t *s, int pitch,
}
}
void vp9_loop_filter_vertical_edge_16_c(uint8_t *s, int pitch,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_loop_filter_vertical_edge_c(s, pitch, blimit0, limit0, thresh0, 1);
vp9_loop_filter_vertical_edge_c(s + 8 * pitch, pitch, blimit1, limit1,
void vp9_lpf_vertical_4_dual_c(uint8_t *s, int pitch, const uint8_t *blimit0,
const uint8_t *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_4_c(s, pitch, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_4_c(s + 8 * pitch, pitch, blimit1, limit1,
thresh1, 1);
}
@ -185,11 +175,9 @@ static INLINE void filter8(int8_t mask, uint8_t hev, uint8_t flat,
}
}
void vp9_mbloop_filter_horizontal_edge_c(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_8_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh,
int count) {
int i;
// loop filter designed to work using chars so that we can make maximum use
@ -208,22 +196,17 @@ void vp9_mbloop_filter_horizontal_edge_c(uint8_t *s, int p,
}
}
void vp9_mbloop_filter_horizontal_edge_16_c(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_horizontal_edge_c(s, p, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_horizontal_edge_c(s + 8, p, blimit1, limit1, thresh1, 1);
void vp9_lpf_horizontal_8_dual_c(uint8_t *s, int p, const uint8_t *blimit0,
const uint8_t *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_horizontal_8_c(s, p, blimit0, limit0, thresh0, 1);
vp9_lpf_horizontal_8_c(s + 8, p, blimit1, limit1, thresh1, 1);
}
void vp9_mbloop_filter_vertical_edge_c(uint8_t *s, int pitch,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_vertical_8_c(uint8_t *s, int pitch, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh,
int count) {
int i;
for (i = 0; i < 8 * count; ++i) {
@ -239,15 +222,12 @@ void vp9_mbloop_filter_vertical_edge_c(uint8_t *s, int pitch,
}
}
void vp9_mbloop_filter_vertical_edge_16_c(uint8_t *s, int pitch,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_mbloop_filter_vertical_edge_c(s, pitch, blimit0, limit0, thresh0, 1);
vp9_mbloop_filter_vertical_edge_c(s + 8 * pitch, pitch, blimit1, limit1,
void vp9_lpf_vertical_8_dual_c(uint8_t *s, int pitch, const uint8_t *blimit0,
const uint8_t *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) {
vp9_lpf_vertical_8_c(s, pitch, blimit0, limit0, thresh0, 1);
vp9_lpf_vertical_8_c(s + 8 * pitch, pitch, blimit1, limit1,
thresh1, 1);
}
@ -302,11 +282,9 @@ static INLINE void filter16(int8_t mask, uint8_t hev,
}
}
void vp9_mb_lpf_horizontal_edge_w_c(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh,
int count) {
void vp9_lpf_horizontal_16_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh,
int count) {
int i;
// loop filter designed to work using chars so that we can make maximum use
@ -355,16 +333,12 @@ static void mb_lpf_vertical_edge_w(uint8_t *s, int p,
}
}
void vp9_mb_lpf_vertical_edge_w_c(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
void vp9_lpf_vertical_16_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh) {
mb_lpf_vertical_edge_w(s, p, blimit, limit, thresh, 8);
}
void vp9_mb_lpf_vertical_edge_w_16_c(uint8_t *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
void vp9_lpf_vertical_16_dual_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh) {
mb_lpf_vertical_edge_w(s, p, blimit, limit, thresh, 16);
}

44
vp9/common/vp9_rtcd_defs.sh
View File

@ -191,38 +191,38 @@ specialize vp9_dc_128_predictor_32x32
#
# Loopfilter
#
prototype void vp9_mb_lpf_vertical_edge_w "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
specialize vp9_mb_lpf_vertical_edge_w sse2 neon dspr2
prototype void vp9_lpf_vertical_16 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
specialize vp9_lpf_vertical_16 sse2 neon dspr2
prototype void vp9_mb_lpf_vertical_edge_w_16 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
specialize vp9_mb_lpf_vertical_edge_w_16 sse2 neon dspr2
prototype void vp9_lpf_vertical_16_dual "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
specialize vp9_lpf_vertical_16_dual sse2 neon dspr2
prototype void vp9_mbloop_filter_vertical_edge "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_mbloop_filter_vertical_edge sse2 neon dspr2
prototype void vp9_lpf_vertical_8 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_lpf_vertical_8 sse2 neon dspr2
prototype void vp9_mbloop_filter_vertical_edge_16 "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_mbloop_filter_vertical_edge_16 sse2 neon dspr2
prototype void vp9_lpf_vertical_8_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_lpf_vertical_8_dual sse2 neon dspr2
prototype void vp9_loop_filter_vertical_edge "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_loop_filter_vertical_edge mmx neon dspr2
prototype void vp9_lpf_vertical_4 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_lpf_vertical_4 mmx neon dspr2
prototype void vp9_loop_filter_vertical_edge_16 "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_loop_filter_vertical_edge_16 sse2 neon dspr2
prototype void vp9_lpf_vertical_4_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_lpf_vertical_4_dual sse2 neon dspr2
prototype void vp9_mb_lpf_horizontal_edge_w "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_mb_lpf_horizontal_edge_w sse2 avx2 neon dspr2
prototype void vp9_lpf_horizontal_16 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_lpf_horizontal_16 sse2 avx2 neon dspr2
prototype void vp9_mbloop_filter_horizontal_edge "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_mbloop_filter_horizontal_edge sse2 neon dspr2
prototype void vp9_lpf_horizontal_8 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_lpf_horizontal_8 sse2 neon dspr2
prototype void vp9_mbloop_filter_horizontal_edge_16 "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_mbloop_filter_horizontal_edge_16 sse2 neon dspr2
prototype void vp9_lpf_horizontal_8_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_lpf_horizontal_8_dual sse2 neon dspr2
prototype void vp9_loop_filter_horizontal_edge "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_loop_filter_horizontal_edge mmx neon dspr2
prototype void vp9_lpf_horizontal_4 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
specialize vp9_lpf_horizontal_4 mmx neon dspr2
prototype void vp9_loop_filter_horizontal_edge_16 "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_loop_filter_horizontal_edge_16 sse2 neon dspr2
prototype void vp9_lpf_horizontal_4_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
specialize vp9_lpf_horizontal_4_dual sse2 neon dspr2
#
# post proc

2
vp9/common/x86/vp9_loopfilter_intrin_avx2.c
View File

@ -933,7 +933,7 @@ static void mb_lpf_horizontal_edge_w_avx2_16(unsigned char *s, int p,
}
}
void vp9_mb_lpf_horizontal_edge_w_avx2(unsigned char *s, int p,
void vp9_lpf_horizontal_16_avx2(unsigned char *s, int p,
const unsigned char *_blimit, const unsigned char *_limit,
const unsigned char *_thresh, int count) {
if (count == 1)

110
vp9/common/x86/vp9_loopfilter_intrin_sse2.c
View File

@ -846,24 +846,20 @@ static void mb_lpf_horizontal_edge_w_sse2_16(unsigned char *s,
}
// TODO(yunqingwang): remove count and call these 2 functions(8 or 16) directly.
void vp9_mb_lpf_horizontal_edge_w_sse2(unsigned char *s,
int p,
const unsigned char *_blimit,
const unsigned char *_limit,
const unsigned char *_thresh,
int count) {
void vp9_lpf_horizontal_16_sse2(unsigned char *s, int p,
const unsigned char *_blimit,
const unsigned char *_limit,
const unsigned char *_thresh, int count) {
if (count == 1)
mb_lpf_horizontal_edge_w_sse2_8(s, p, _blimit, _limit, _thresh);
else
mb_lpf_horizontal_edge_w_sse2_16(s, p, _blimit, _limit, _thresh);
}
void vp9_mbloop_filter_horizontal_edge_sse2(unsigned char *s,
int p,
const unsigned char *_blimit,
const unsigned char *_limit,
const unsigned char *_thresh,
int count) {
void vp9_lpf_horizontal_8_sse2(unsigned char *s, int p,
const unsigned char *_blimit,
const unsigned char *_limit,
const unsigned char *_thresh, int count) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op2, 16);
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op1, 16);
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op0, 16);
@ -1083,13 +1079,13 @@ void vp9_mbloop_filter_horizontal_edge_sse2(unsigned char *s,
}
}
void vp9_mbloop_filter_horizontal_edge_16_sse2(uint8_t *s, int p,
const uint8_t *_blimit0,
const uint8_t *_limit0,
const uint8_t *_thresh0,
const uint8_t *_blimit1,
const uint8_t *_limit1,
const uint8_t *_thresh1) {
void vp9_lpf_horizontal_8_dual_sse2(uint8_t *s, int p,
const uint8_t *_blimit0,
const uint8_t *_limit0,
const uint8_t *_thresh0,
const uint8_t *_blimit1,
const uint8_t *_limit1,
const uint8_t *_thresh1) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op2, 16);
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op1, 16);
DECLARE_ALIGNED_ARRAY(16, unsigned char, flat_op0, 16);
@ -1330,14 +1326,13 @@ void vp9_mbloop_filter_horizontal_edge_16_sse2(uint8_t *s, int p,
}
}
void vp9_loop_filter_horizontal_edge_16_sse2(unsigned char *s,
int p,
const unsigned char *_blimit0,
const unsigned char *_limit0,
const unsigned char *_thresh0,
const unsigned char *_blimit1,
const unsigned char *_limit1,
const unsigned char *_thresh1) {
void vp9_lpf_horizontal_4_dual_sse2(unsigned char *s, int p,
const unsigned char *_blimit0,
const unsigned char *_limit0,
const unsigned char *_thresh0,
const unsigned char *_blimit1,
const unsigned char *_limit1,
const unsigned char *_thresh1) {
const __m128i blimit =
_mm_unpacklo_epi64(_mm_load_si128((const __m128i *)_blimit0),
_mm_load_si128((const __m128i *)_blimit1));
@ -1598,13 +1593,12 @@ static INLINE void transpose(unsigned char *src[], int in_p,
} while (++idx8x8 < num_8x8_to_transpose);
}
void vp9_loop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
void vp9_lpf_vertical_4_dual_sse2(uint8_t *s, int p, const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, t_dst, 16 * 8);
unsigned char *src[2];
unsigned char *dst[2];
@ -1613,8 +1607,8 @@ void vp9_loop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
transpose8x16(s - 4, s - 4 + p * 8, p, t_dst, 16);
// Loop filtering
vp9_loop_filter_horizontal_edge_16_sse2(t_dst + 4 * 16, 16, blimit0, limit0,
thresh0, blimit1, limit1, thresh1);
vp9_lpf_horizontal_4_dual_sse2(t_dst + 4 * 16, 16, blimit0, limit0, thresh0,
blimit1, limit1, thresh1);
src[0] = t_dst;
src[1] = t_dst + 8;
dst[0] = s - 4;
@ -1624,11 +1618,10 @@ void vp9_loop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
transpose(src, 16, dst, p, 2);
}
void vp9_mbloop_filter_vertical_edge_sse2(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count) {
void vp9_lpf_vertical_8_sse2(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh, int count) {
DECLARE_ALIGNED_ARRAY(8, unsigned char, t_dst, 8 * 8);
unsigned char *src[1];
unsigned char *dst[1];
@ -1641,8 +1634,7 @@ void vp9_mbloop_filter_vertical_edge_sse2(unsigned char *s, int p,
transpose(src, p, dst, 8, 1);
// Loop filtering
vp9_mbloop_filter_horizontal_edge_sse2(t_dst + 4 * 8, 8, blimit, limit,
thresh, 1);
vp9_lpf_horizontal_8_sse2(t_dst + 4 * 8, 8, blimit, limit, thresh, 1);
src[0] = t_dst;
dst[0] = s - 4;
@ -1651,13 +1643,12 @@ void vp9_mbloop_filter_vertical_edge_sse2(unsigned char *s, int p,
transpose(src, 8, dst, p, 1);
}
void vp9_mbloop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
void vp9_lpf_vertical_8_dual_sse2(uint8_t *s, int p, const uint8_t *blimit0,
const uint8_t *limit0,
const uint8_t *thresh0,
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, t_dst, 16 * 8);
unsigned char *src[2];
unsigned char *dst[2];
@ -1666,8 +1657,8 @@ void vp9_mbloop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
transpose8x16(s - 4, s - 4 + p * 8, p, t_dst, 16);
// Loop filtering
vp9_mbloop_filter_horizontal_edge_16_sse2(t_dst + 4 * 16, 16, blimit0, limit0,
thresh0, blimit1, limit1, thresh1);
vp9_lpf_horizontal_8_dual_sse2(t_dst + 4 * 16, 16, blimit0, limit0, thresh0,
blimit1, limit1, thresh1);
src[0] = t_dst;
src[1] = t_dst + 8;
@ -1678,10 +1669,10 @@ void vp9_mbloop_filter_vertical_edge_16_sse2(uint8_t *s, int p,
transpose(src, 16, dst, p, 2);
}
void vp9_mb_lpf_vertical_edge_w_sse2(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh) {
void vp9_lpf_vertical_16_sse2(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh) {
DECLARE_ALIGNED_ARRAY(8, unsigned char, t_dst, 8 * 16);
unsigned char *src[2];
unsigned char *dst[2];
@ -1706,10 +1697,9 @@ void vp9_mb_lpf_vertical_edge_w_sse2(unsigned char *s, int p,
transpose(src, 8, dst, p, 2);
}
void vp9_mb_lpf_vertical_edge_w_16_sse2(unsigned char *s, int p,
const uint8_t *blimit,
const uint8_t *limit,
const uint8_t *thresh) {
void vp9_lpf_vertical_16_dual_sse2(unsigned char *s, int p,
const uint8_t *blimit, const uint8_t *limit,
const uint8_t *thresh) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, t_dst, 256);
// Transpose 16x16

12
vp9/common/x86/vp9_loopfilter_mmx.asm
View File

@ -12,7 +12,7 @@
%include "vpx_ports/x86_abi_support.asm"
;void vp9_loop_filter_horizontal_edge_mmx
;void vp9_lpf_horizontal_4_mmx
;(
; unsigned char *src_ptr,
; int src_pixel_step,
@ -21,8 +21,8 @@
; const char *thresh,
; int count
;)
global sym(vp9_loop_filter_horizontal_edge_mmx) PRIVATE
sym(vp9_loop_filter_horizontal_edge_mmx):
global sym(vp9_lpf_horizontal_4_mmx) PRIVATE
sym(vp9_lpf_horizontal_4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
@ -224,7 +224,7 @@ sym(vp9_loop_filter_horizontal_edge_mmx):
ret
;void vp9_loop_filter_vertical_edge_mmx
;void vp9_lpf_vertical_4_mmx
;(
; unsigned char *src_ptr,
; int src_pixel_step,
@ -233,8 +233,8 @@ sym(vp9_loop_filter_horizontal_edge_mmx):
; const char *thresh,
; int count
;)
global sym(vp9_loop_filter_vertical_edge_mmx) PRIVATE
sym(vp9_loop_filter_vertical_edge_mmx):
global sym(vp9_lpf_vertical_4_mmx) PRIVATE
sym(vp9_lpf_vertical_4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6