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

Unify loopfilter function names

Browse Source 
Rename vpx_lpf_horizontal_edge_8() to vpx_lpf_horizontal_16().
Rename vpx_lpf_horizontal_edge_16() to vpx_lpf_horizontal_16_dual().

Change-Id: I798ca8fbbd657d06d3db2bfb0fb3321168f49e52
This commit is contained in:
Linfeng Zhang
2016-09-28 15:20:01 -07:00
parent 691ef20272
commit 7f1f35183a
11 changed files with 97 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

58
test/lpf_test.cc
View File

@@ -402,10 +402,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 8), &vpx_highbd_lpf_vertical_4_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 8), &vpx_highbd_lpf_horizontal_8_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 8), &vpx_highbd_lpf_horizontal_16_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 8), &vpx_highbd_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2, make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 8), &vpx_highbd_lpf_vertical_8_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2, make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@@ -416,10 +416,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 10), &vpx_highbd_lpf_vertical_4_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 10), &vpx_highbd_lpf_horizontal_8_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 10), &vpx_highbd_lpf_horizontal_16_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 10), &vpx_highbd_lpf_horizontal_16_dual_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2, make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 10), &vpx_highbd_lpf_vertical_8_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2, make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@@ -430,10 +430,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 12), &vpx_highbd_lpf_vertical_4_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 12), &vpx_highbd_lpf_horizontal_8_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 12), &vpx_highbd_lpf_horizontal_16_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2, make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 12), &vpx_highbd_lpf_horizontal_16_dual_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2, make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 12), &vpx_highbd_lpf_vertical_8_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2, make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@@ -450,10 +450,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_sse2, &vpx_lpf_horizontal_4_c, 8), make_tuple(&vpx_lpf_horizontal_4_sse2, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_sse2, &vpx_lpf_horizontal_8_c, 8), make_tuple(&vpx_lpf_horizontal_8_sse2, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8_sse2, make_tuple(&vpx_lpf_horizontal_16_sse2, &vpx_lpf_horizontal_16_c, 8),
&vpx_lpf_horizontal_edge_8_c, 8), make_tuple(&vpx_lpf_horizontal_16_dual_sse2,
make_tuple(&vpx_lpf_horizontal_edge_16_sse2, &vpx_lpf_horizontal_16_dual_c, 8),
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_vertical_4_sse2, &vpx_lpf_vertical_4_c, 8), make_tuple(&vpx_lpf_vertical_4_sse2, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_sse2, &vpx_lpf_vertical_8_c, 8), make_tuple(&vpx_lpf_vertical_8_sse2, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_sse2, &vpx_lpf_vertical_16_c, 8), make_tuple(&vpx_lpf_vertical_16_sse2, &vpx_lpf_vertical_16_c, 8),
@@ -465,10 +464,10 @@ INSTANTIATE_TEST_CASE_P(
#if HAVE_AVX2 && (!CONFIG_VP9_HIGHBITDEPTH) #if HAVE_AVX2 && (!CONFIG_VP9_HIGHBITDEPTH)
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
AVX2, Loop8Test6Param, AVX2, Loop8Test6Param,
::testing::Values(make_tuple(&vpx_lpf_horizontal_edge_8_avx2, ::testing::Values(make_tuple(&vpx_lpf_horizontal_16_avx2,
&vpx_lpf_horizontal_edge_8_c, 8), &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_16_avx2, make_tuple(&vpx_lpf_horizontal_16_dual_avx2,
&vpx_lpf_horizontal_edge_16_c, 8))); &vpx_lpf_horizontal_16_dual_c, 8)));
#endif #endif
#if HAVE_SSE2 #if HAVE_SSE2
@@ -520,10 +519,9 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
NEON, Loop8Test6Param, NEON, Loop8Test6Param,
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_edge_8_neon, make_tuple(&vpx_lpf_horizontal_16_neon, &vpx_lpf_horizontal_16_c, 8),
&vpx_lpf_horizontal_edge_8_c, 8), make_tuple(&vpx_lpf_horizontal_16_dual_neon,
make_tuple(&vpx_lpf_horizontal_edge_16_neon, &vpx_lpf_horizontal_16_dual_c, 8),
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_vertical_16_neon, &vpx_lpf_vertical_16_c, 8), make_tuple(&vpx_lpf_vertical_16_neon, &vpx_lpf_vertical_16_c, 8),
make_tuple(&vpx_lpf_vertical_16_dual_neon, &vpx_lpf_vertical_16_dual_c, make_tuple(&vpx_lpf_vertical_16_dual_neon, &vpx_lpf_vertical_16_dual_c,
8), 8),
@@ -550,8 +548,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_dspr2, &vpx_lpf_horizontal_4_c, 8), make_tuple(&vpx_lpf_horizontal_4_dspr2, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_dspr2, &vpx_lpf_horizontal_8_c, 8), make_tuple(&vpx_lpf_horizontal_8_dspr2, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8, &vpx_lpf_horizontal_edge_8, 8), make_tuple(&vpx_lpf_horizontal_16_dspr2, &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_16, &vpx_lpf_horizontal_edge_16, 8), make_tuple(&vpx_lpf_horizontal_16_dual_dspr2,
&vpx_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_lpf_vertical_4_dspr2, &vpx_lpf_vertical_4_c, 8), make_tuple(&vpx_lpf_vertical_4_dspr2, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_dspr2, &vpx_lpf_vertical_8_c, 8), make_tuple(&vpx_lpf_vertical_8_dspr2, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_dspr2, &vpx_lpf_vertical_16_c, 8), make_tuple(&vpx_lpf_vertical_16_dspr2, &vpx_lpf_vertical_16_c, 8),
@@ -576,10 +575,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_msa, &vpx_lpf_horizontal_4_c, 8), make_tuple(&vpx_lpf_horizontal_4_msa, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_msa, &vpx_lpf_horizontal_8_c, 8), make_tuple(&vpx_lpf_horizontal_8_msa, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8_msa, &vpx_lpf_horizontal_edge_8_c, make_tuple(&vpx_lpf_horizontal_16_msa, &vpx_lpf_horizontal_16_c, 8),
8), make_tuple(&vpx_lpf_horizontal_16_dual_msa,
make_tuple(&vpx_lpf_horizontal_edge_16_msa, &vpx_lpf_horizontal_16_dual_c, 8),
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_vertical_4_msa, &vpx_lpf_vertical_4_c, 8), make_tuple(&vpx_lpf_vertical_4_msa, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_msa, &vpx_lpf_vertical_8_c, 8), make_tuple(&vpx_lpf_vertical_8_msa, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_msa, &vpx_lpf_vertical_16_c, 8))); make_tuple(&vpx_lpf_vertical_16_msa, &vpx_lpf_vertical_16_c, 8)));

11
vp9/common/vp9_loopfilter.c
View File

@@ -465,12 +465,11 @@ static void filter_selectively_horiz(
if (mask_16x16 & 1) { if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) { if ((mask_16x16 & 3) == 3) {
vpx_lpf_horizontal_edge_16(s, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_16_dual(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr); lfi->hev_thr);
count = 2; count = 2;
} else { } else {
vpx_lpf_horizontal_edge_8(s, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
lfi->hev_thr);
} }
} else if (mask_8x8 & 1) { } else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) { if ((mask_8x8 & 3) == 3) {
@@ -559,12 +558,12 @@ static void highbd_filter_selectively_horiz(
if (mask_16x16 & 1) { if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) { if ((mask_16x16 & 3) == 3) {
vpx_highbd_lpf_horizontal_edge_16(s, pitch, lfi->mblim, lfi->lim, vpx_highbd_lpf_horizontal_16_dual(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, bd); lfi->hev_thr, bd);
count = 2; count = 2;
} else { } else {
vpx_highbd_lpf_horizontal_edge_8(s, pitch, lfi->mblim, lfi->lim, vpx_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, bd); lfi->hev_thr, bd);
} }
} else if (mask_8x8 & 1) { } else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) { if ((mask_8x8 & 3) == 3) {

16
vpx_dsp/arm/loopfilter_mb_neon.asm
View File

@@ -8,8 +8,8 @@
; be found in the AUTHORS file in the root of the source tree. ; be found in the AUTHORS file in the root of the source tree.
; ;
EXPORT |vpx_lpf_horizontal_edge_8_neon| EXPORT |vpx_lpf_horizontal_16_neon|
EXPORT |vpx_lpf_horizontal_edge_16_neon| EXPORT |vpx_lpf_horizontal_16_dual_neon|
EXPORT |vpx_lpf_vertical_16_neon| EXPORT |vpx_lpf_vertical_16_neon|
EXPORT |vpx_lpf_vertical_16_dual_neon| EXPORT |vpx_lpf_vertical_16_dual_neon|
ARM ARM
@@ -119,7 +119,7 @@ h_next
ENDP ; |mb_lpf_horizontal_edge| ENDP ; |mb_lpf_horizontal_edge|
; void vpx_lpf_horizontal_edge_8_neon(uint8_t *s, int pitch, ; void vpx_lpf_horizontal_16_neon(uint8_t *s, int pitch,
; const uint8_t *blimit, ; const uint8_t *blimit,
; const uint8_t *limit, ; const uint8_t *limit,
; const uint8_t *thresh) ; const uint8_t *thresh)
@@ -128,12 +128,12 @@ h_next
; r2 const uint8_t *blimit, ; r2 const uint8_t *blimit,
; r3 const uint8_t *limit, ; r3 const uint8_t *limit,
; sp const uint8_t *thresh ; sp const uint8_t *thresh
|vpx_lpf_horizontal_edge_8_neon| PROC |vpx_lpf_horizontal_16_neon| PROC
mov r12, #1 mov r12, #1
b mb_lpf_horizontal_edge b mb_lpf_horizontal_edge
ENDP ; |vpx_lpf_horizontal_edge_8_neon| ENDP ; |vpx_lpf_horizontal_16_neon|
; void vpx_lpf_horizontal_edge_16_neon(uint8_t *s, int pitch, ; void vpx_lpf_horizontal_16_dual_neon(uint8_t *s, int pitch,
; const uint8_t *blimit, ; const uint8_t *blimit,
; const uint8_t *limit, ; const uint8_t *limit,
; const uint8_t *thresh) ; const uint8_t *thresh)
@@ -142,10 +142,10 @@ h_next
; r2 const uint8_t *blimit, ; r2 const uint8_t *blimit,
; r3 const uint8_t *limit, ; r3 const uint8_t *limit,
; sp const uint8_t *thresh ; sp const uint8_t *thresh
|vpx_lpf_horizontal_edge_16_neon| PROC |vpx_lpf_horizontal_16_dual_neon| PROC
mov r12, #2 mov r12, #2
b mb_lpf_horizontal_edge b mb_lpf_horizontal_edge
ENDP ; |vpx_lpf_horizontal_edge_16_neon| ENDP ; |vpx_lpf_horizontal_16_dual_neon|
; void mb_lpf_vertical_edge_w(uint8_t *s, int p, const uint8_t *blimit, ; void mb_lpf_vertical_edge_w(uint8_t *s, int p, const uint8_t *blimit,
; const uint8_t *limit, const uint8_t *thresh, ; const uint8_t *limit, const uint8_t *thresh,

7
vpx_dsp/arm/loopfilter_mb_neon.c
View File

@@ -975,9 +975,8 @@ FUN_LPF_16_KERNEL(_, 8) // lpf_16_kernel
FUN_LPF_16_KERNEL(_dual_, 16) // lpf_16_dual_kernel FUN_LPF_16_KERNEL(_dual_, 16) // lpf_16_dual_kernel
#undef FUN_LPF_16_KERNEL #undef FUN_LPF_16_KERNEL
void vpx_lpf_horizontal_edge_8_neon(uint8_t *s, int p, const uint8_t *blimit, void vpx_lpf_horizontal_16_neon(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *limit, const uint8_t *thresh) {
const uint8_t *thresh) {
uint8x8_t p7, p6, p5, p4, p3, p2, p1, p0, q0, q1, q2, q3, q4, q5, q6, q7, op6, uint8x8_t p7, p6, p5, p4, p3, p2, p1, p0, q0, q1, q2, q3, q4, q5, q6, q7, op6,
op5, op4, op3, op2, op1, op0, oq0, oq1, oq2, oq3, oq4, oq5, oq6; op5, op4, op3, op2, op1, op0, oq0, oq1, oq2, oq3, oq4, oq5, oq6;
uint32_t flat_status, flat2_status; uint32_t flat_status, flat2_status;
@@ -992,7 +991,7 @@ void vpx_lpf_horizontal_edge_8_neon(uint8_t *s, int p, const uint8_t *blimit,
oq5, oq6, flat_status, flat2_status); oq5, oq6, flat_status, flat2_status);
} }
void vpx_lpf_horizontal_edge_16_neon(uint8_t *s, int p, const uint8_t *blimit, void vpx_lpf_horizontal_16_dual_neon(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *limit,
const uint8_t *thresh) { const uint8_t *thresh) {
uint8x16_t p7, p6, p5, p4, p3, p2, p1, p0, q0, q1, q2, q3, q4, q5, q6, q7, uint8x16_t p7, p6, p5, p4, p3, p2, p1, p0, q0, q1, q2, q3, q4, q5, q6, q7,

15
vpx_dsp/loopfilter.c
View File

@@ -308,12 +308,12 @@ static void mb_lpf_horizontal_edge_w(uint8_t *s, int p, const uint8_t *blimit,
} }
} }
void vpx_lpf_horizontal_edge_8_c(uint8_t *s, int p, const uint8_t *blimit, void vpx_lpf_horizontal_16_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh) { const uint8_t *limit, const uint8_t *thresh) {
mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 1); mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 1);
} }
void vpx_lpf_horizontal_edge_16_c(uint8_t *s, int p, const uint8_t *blimit, void vpx_lpf_horizontal_16_dual_c(uint8_t *s, int p, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *thresh) { const uint8_t *limit, const uint8_t *thresh) {
mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 2); mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 2);
} }
@@ -673,14 +673,13 @@ static void highbd_mb_lpf_horizontal_edge_w(uint16_t *s, int p,
} }
} }
void vpx_highbd_lpf_horizontal_edge_8_c(uint16_t *s, int p, void vpx_highbd_lpf_horizontal_16_c(uint16_t *s, int p, const uint8_t *blimit,
const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh,
const uint8_t *limit, int bd) {
const uint8_t *thresh, int bd) {
highbd_mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 1, bd); highbd_mb_lpf_horizontal_edge_w(s, p, blimit, limit, thresh, 1, bd);
} }
void vpx_highbd_lpf_horizontal_edge_16_c(uint16_t *s, int p, void vpx_highbd_lpf_horizontal_16_dual_c(uint16_t *s, int p,
const uint8_t *blimit, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *limit,
const uint8_t *thresh, int bd) { const uint8_t *thresh, int bd) {

23
vpx_dsp/mips/loopfilter_16_msa.c
View File

@@ -403,10 +403,11 @@ void vpx_hz_lpf_t16_16w(uint8_t *src, int32_t pitch, uint8_t *filter48) {
} }
} }
void vpx_lpf_horizontal_16_dual_msa(uint8_t *src, int32_t pitch, static void mb_lpf_horizontal_edge_dual(uint8_t *src, int32_t pitch,
const uint8_t *b_limit_ptr, const uint8_t *b_limit_ptr,
const uint8_t *limit_ptr, const uint8_t *limit_ptr,
const uint8_t *thresh_ptr, int32_t count) { const uint8_t *thresh_ptr,
int32_t count) {
DECLARE_ALIGNED(32, uint8_t, filter48[16 * 8]); DECLARE_ALIGNED(32, uint8_t, filter48[16 * 8]);
uint8_t early_exit = 0; uint8_t early_exit = 0;
@@ -638,19 +639,19 @@ static void mb_lpf_horizontal_edge(uint8_t *src, int32_t pitch,
} }
} }
} else { } else {
vpx_lpf_horizontal_16_dual_msa(src, pitch, b_limit_ptr, limit_ptr, mb_lpf_horizontal_edge_dual(src, pitch, b_limit_ptr, limit_ptr, thresh_ptr,
thresh_ptr, count); count);
} }
} }
void vpx_lpf_horizontal_edge_8_msa(uint8_t *src, int32_t pitch, void vpx_lpf_horizontal_16_msa(uint8_t *src, int32_t pitch,
const uint8_t *b_limit_ptr, const uint8_t *b_limit_ptr,
const uint8_t *limit_ptr, const uint8_t *limit_ptr,
const uint8_t *thresh_ptr) { const uint8_t *thresh_ptr) {
mb_lpf_horizontal_edge(src, pitch, b_limit_ptr, limit_ptr, thresh_ptr, 1); mb_lpf_horizontal_edge(src, pitch, b_limit_ptr, limit_ptr, thresh_ptr, 1);
} }
void vpx_lpf_horizontal_edge_16_msa(uint8_t *src, int32_t pitch, void vpx_lpf_horizontal_16_dual_msa(uint8_t *src, int32_t pitch,
const uint8_t *b_limit_ptr, const uint8_t *b_limit_ptr,
const uint8_t *limit_ptr, const uint8_t *limit_ptr,
const uint8_t *thresh_ptr) { const uint8_t *thresh_ptr) {

9
vpx_dsp/mips/loopfilter_mb_horiz_dspr2.c
View File

@@ -717,14 +717,13 @@ static void mb_lpf_horizontal_edge(unsigned char *s, int pitch,
} }
} }
void vpx_lpf_horizontal_edge_8_dspr2(unsigned char *s, int pitch, void vpx_lpf_horizontal_16_dspr2(unsigned char *s, int pitch,
const uint8_t *blimit, const uint8_t *blimit, const uint8_t *limit,
const uint8_t *limit, const uint8_t *thresh) {
const uint8_t *thresh) {
mb_lpf_horizontal_edge(s, pitch, blimit, limit, thresh, 1); mb_lpf_horizontal_edge(s, pitch, blimit, limit, thresh, 1);
} }
void vpx_lpf_horizontal_edge_16_dspr2(unsigned char *s, int pitch, void vpx_lpf_horizontal_16_dual_dspr2(unsigned char *s, int pitch,
const uint8_t *blimit, const uint8_t *blimit,
const uint8_t *limit, const uint8_t *limit,
const uint8_t *thresh) { const uint8_t *thresh) {

16
vpx_dsp/vpx_dsp_rtcd_defs.pl
View File

@@ -522,11 +522,11 @@ specialize qw/vpx_lpf_vertical_4 sse2 neon dspr2 msa/;
add_proto qw/void vpx_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"; add_proto qw/void vpx_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 qw/vpx_lpf_vertical_4_dual sse2 neon dspr2 msa/; specialize qw/vpx_lpf_vertical_4_dual sse2 neon dspr2 msa/;
add_proto qw/void vpx_lpf_horizontal_edge_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"; add_proto qw/void vpx_lpf_horizontal_16/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
specialize qw/vpx_lpf_horizontal_edge_8 sse2 avx2 neon dspr2 msa/; specialize qw/vpx_lpf_horizontal_16 sse2 avx2 neon dspr2 msa/;
add_proto qw/void vpx_lpf_horizontal_edge_16/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"; add_proto qw/void vpx_lpf_horizontal_16_dual/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
specialize qw/vpx_lpf_horizontal_edge_16 sse2 avx2 neon dspr2 msa/; specialize qw/vpx_lpf_horizontal_16_dual sse2 avx2 neon dspr2 msa/;
add_proto qw/void vpx_lpf_horizontal_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"; add_proto qw/void vpx_lpf_horizontal_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
specialize qw/vpx_lpf_horizontal_8 sse2 neon dspr2 msa/; specialize qw/vpx_lpf_horizontal_8 sse2 neon dspr2 msa/;
@@ -559,11 +559,11 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
add_proto qw/void vpx_highbd_lpf_vertical_4_dual/, "uint16_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, int bd"; add_proto qw/void vpx_highbd_lpf_vertical_4_dual/, "uint16_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, int bd";
specialize qw/vpx_highbd_lpf_vertical_4_dual sse2/; specialize qw/vpx_highbd_lpf_vertical_4_dual sse2/;
add_proto qw/void vpx_highbd_lpf_horizontal_edge_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd"; add_proto qw/void vpx_highbd_lpf_horizontal_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
specialize qw/vpx_highbd_lpf_horizontal_edge_8 sse2/; specialize qw/vpx_highbd_lpf_horizontal_16 sse2/;
add_proto qw/void vpx_highbd_lpf_horizontal_edge_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd"; add_proto qw/void vpx_highbd_lpf_horizontal_16_dual/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
specialize qw/vpx_highbd_lpf_horizontal_edge_16 sse2/; specialize qw/vpx_highbd_lpf_horizontal_16_dual sse2/;
add_proto qw/void vpx_highbd_lpf_horizontal_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd"; add_proto qw/void vpx_highbd_lpf_horizontal_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
specialize qw/vpx_highbd_lpf_horizontal_8 sse2/; specialize qw/vpx_highbd_lpf_horizontal_8 sse2/;

20
vpx_dsp/x86/highbd_loopfilter_sse2.c
View File

@@ -48,10 +48,10 @@ static INLINE __m128i signed_char_clamp_bd_sse2(__m128i value, int bd) {
// TODO(debargha, peter): Break up large functions into smaller ones // TODO(debargha, peter): Break up large functions into smaller ones
// in this file. // in this file.
void vpx_highbd_lpf_horizontal_edge_8_sse2(uint16_t *s, int p, void vpx_highbd_lpf_horizontal_16_sse2(uint16_t *s, int p,
const uint8_t *_blimit, const uint8_t *_blimit,
const uint8_t *_limit, const uint8_t *_limit,
const uint8_t *_thresh, int bd) { const uint8_t *_thresh, int bd) {
const __m128i zero = _mm_set1_epi16(0); const __m128i zero = _mm_set1_epi16(0);
const __m128i one = _mm_set1_epi16(1); const __m128i one = _mm_set1_epi16(1);
__m128i blimit, limit, thresh; __m128i blimit, limit, thresh;
@@ -475,12 +475,12 @@ void vpx_highbd_lpf_horizontal_edge_8_sse2(uint16_t *s, int p,
_mm_store_si128((__m128i *)(s - 0 * p), q0); _mm_store_si128((__m128i *)(s - 0 * p), q0);
} }
void vpx_highbd_lpf_horizontal_edge_16_sse2(uint16_t *s, int p, void vpx_highbd_lpf_horizontal_16_dual_sse2(uint16_t *s, int p,
const uint8_t *_blimit, const uint8_t *_blimit,
const uint8_t *_limit, const uint8_t *_limit,
const uint8_t *_thresh, int bd) { const uint8_t *_thresh, int bd) {
vpx_highbd_lpf_horizontal_edge_8_sse2(s, p, _blimit, _limit, _thresh, bd); vpx_highbd_lpf_horizontal_16_sse2(s, p, _blimit, _limit, _thresh, bd);
vpx_highbd_lpf_horizontal_edge_8_sse2(s + 8, p, _blimit, _limit, _thresh, bd); vpx_highbd_lpf_horizontal_16_sse2(s + 8, p, _blimit, _limit, _thresh, bd);
} }
void vpx_highbd_lpf_horizontal_8_sse2(uint16_t *s, int p, void vpx_highbd_lpf_horizontal_8_sse2(uint16_t *s, int p,
@@ -1108,8 +1108,8 @@ void vpx_highbd_lpf_vertical_16_sse2(uint16_t *s, int p, const uint8_t *blimit,
highbd_transpose(src, p, dst, 8, 2); highbd_transpose(src, p, dst, 8, 2);
// Loop filtering // Loop filtering
vpx_highbd_lpf_horizontal_edge_8_sse2(t_dst + 8 * 8, 8, blimit, limit, thresh, vpx_highbd_lpf_horizontal_16_sse2(t_dst + 8 * 8, 8, blimit, limit, thresh,
bd); bd);
src[0] = t_dst; src[0] = t_dst;
src[1] = t_dst + 8 * 8; src[1] = t_dst + 8 * 8;
dst[0] = s - 8; dst[0] = s - 8;
@@ -1130,7 +1130,7 @@ void vpx_highbd_lpf_vertical_16_dual_sse2(uint16_t *s, int p,
highbd_transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16); highbd_transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16);
// Loop filtering // Loop filtering
vpx_highbd_lpf_horizontal_edge_16_sse2(t_dst + 8 * 16, 16, blimit, limit, vpx_highbd_lpf_horizontal_16_dual_sse2(t_dst + 8 * 16, 16, blimit, limit,
thresh, bd); thresh, bd);
// Transpose back // Transpose back

10
vpx_dsp/x86/loopfilter_avx2.c
View File

@@ -13,10 +13,10 @@
#include "./vpx_dsp_rtcd.h" #include "./vpx_dsp_rtcd.h"
#include "vpx_ports/mem.h" #include "vpx_ports/mem.h"
void vpx_lpf_horizontal_edge_8_avx2(unsigned char *s, int p, void vpx_lpf_horizontal_16_avx2(unsigned char *s, int p,
const unsigned char *_blimit, const unsigned char *_blimit,
const unsigned char *_limit, const unsigned char *_limit,
const unsigned char *_thresh) { const unsigned char *_thresh) {
__m128i mask, hev, flat, flat2; __m128i mask, hev, flat, flat2;
const __m128i zero = _mm_set1_epi16(0); const __m128i zero = _mm_set1_epi16(0);
const __m128i one = _mm_set1_epi8(1); const __m128i one = _mm_set1_epi8(1);
@@ -367,7 +367,7 @@ DECLARE_ALIGNED(32, static const uint8_t, filt_loopfilter_avx2[32]) = {
8, 128, 9, 128, 10, 128, 11, 128, 12, 128, 13, 128, 14, 128, 15, 128 8, 128, 9, 128, 10, 128, 11, 128, 12, 128, 13, 128, 14, 128, 15, 128
}; };
void vpx_lpf_horizontal_edge_16_avx2(unsigned char *s, int p, void vpx_lpf_horizontal_16_dual_avx2(unsigned char *s, int p,
const unsigned char *_blimit, const unsigned char *_blimit,
const unsigned char *_limit, const unsigned char *_limit,
const unsigned char *_thresh) { const unsigned char *_thresh) {

14
vpx_dsp/x86/loopfilter_sse2.c
View File

@@ -229,10 +229,10 @@ void vpx_lpf_vertical_4_sse2(uint8_t *s, int p /* pitch */,
*(int *)(s + 7 * p - 2) = _mm_cvtsi128_si32(qs1qs0); *(int *)(s + 7 * p - 2) = _mm_cvtsi128_si32(qs1qs0);
} }
void vpx_lpf_horizontal_edge_8_sse2(unsigned char *s, int p, void vpx_lpf_horizontal_16_sse2(unsigned char *s, int p,
const unsigned char *_blimit, const unsigned char *_blimit,
const unsigned char *_limit, const unsigned char *_limit,
const unsigned char *_thresh) { const unsigned char *_thresh) {
const __m128i zero = _mm_set1_epi16(0); const __m128i zero = _mm_set1_epi16(0);
const __m128i one = _mm_set1_epi8(1); const __m128i one = _mm_set1_epi8(1);
const __m128i blimit = _mm_load_si128((const __m128i *)_blimit); const __m128i blimit = _mm_load_si128((const __m128i *)_blimit);
@@ -591,7 +591,7 @@ static INLINE __m128i filter16_mask(const __m128i *const flat,
return _mm_or_si128(_mm_andnot_si128(*flat, *other_filt), result); return _mm_or_si128(_mm_andnot_si128(*flat, *other_filt), result);
} }
void vpx_lpf_horizontal_edge_16_sse2(unsigned char *s, int p, void vpx_lpf_horizontal_16_dual_sse2(unsigned char *s, int p,
const unsigned char *_blimit, const unsigned char *_blimit,
const unsigned char *_limit, const unsigned char *_limit,
const unsigned char *_thresh) { const unsigned char *_thresh) {
@@ -1745,7 +1745,7 @@ void vpx_lpf_vertical_16_sse2(unsigned char *s, int p,
transpose(src, p, dst, 8, 2); transpose(src, p, dst, 8, 2);
// Loop filtering // Loop filtering
vpx_lpf_horizontal_edge_8_sse2(t_dst + 8 * 8, 8, blimit, limit, thresh); vpx_lpf_horizontal_16_sse2(t_dst + 8 * 8, 8, blimit, limit, thresh);
src[0] = t_dst; src[0] = t_dst;
src[1] = t_dst + 8 * 8; src[1] = t_dst + 8 * 8;
@@ -1766,7 +1766,7 @@ void vpx_lpf_vertical_16_dual_sse2(unsigned char *s, int p,
transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16); transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16);
// Loop filtering // Loop filtering
vpx_lpf_horizontal_edge_16_sse2(t_dst + 8 * 16, 16, blimit, limit, thresh); vpx_lpf_horizontal_16_dual_sse2(t_dst + 8 * 16, 16, blimit, limit, thresh);
// Transpose back // Transpose back
transpose8x16(t_dst, t_dst + 8 * 16, 16, s - 8, p); transpose8x16(t_dst, t_dst + 8 * 16, 16, s - 8, p);