ref: 109a47b3426d302df201295aeff9cf0e40badf69
parent: 37225744dbf30d79711fa9ef182d2007a51b11bd
author: James Zern <[email protected]>
date: Thu Feb 11 14:54:51 EST 2016
vpx_lpf_vertical_4: remove unused count param Change-Id: I43a191cb3d42e51e7bca266adfa11c6239a8064c
--- a/test/lpf_8_test.cc
+++ b/test/lpf_8_test.cc
@@ -459,7 +459,8 @@
MMX, Loop8Test6Param,
::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_mmx, &vpx_lpf_horizontal_4_c, 8, 1),
- make_tuple(&vpx_lpf_vertical_4_mmx, &vpx_lpf_vertical_4_c, 8, 1)));
+ make_tuple(&wrapper_nc<vpx_lpf_vertical_4_mmx>,
+ &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1)));
#endif // HAVE_MMX
#if HAVE_SSE2
@@ -609,8 +610,8 @@
&wrapper_nc<vpx_lpf_vertical_8_c>, 8, 1),
make_tuple(&vpx_lpf_horizontal_4_neon,
&vpx_lpf_horizontal_4_c, 8, 1),
- make_tuple(&vpx_lpf_vertical_4_neon,
- &vpx_lpf_vertical_4_c, 8, 1)));
+ make_tuple(&wrapper_nc<vpx_lpf_vertical_4_neon>,
+ &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1)));
INSTANTIATE_TEST_CASE_P(
NEON, Loop8Test9Param,
::testing::Values(
@@ -637,7 +638,8 @@
&vpx_lpf_horizontal_16_c, 8, 1),
make_tuple(&vpx_lpf_horizontal_16_dspr2,
&vpx_lpf_horizontal_16_c, 8, 2),
- make_tuple(&vpx_lpf_vertical_4_dspr2, &vpx_lpf_vertical_4_c, 8, 1),
+ make_tuple(&wrapper_nc<vpx_lpf_vertical_4_dspr2>,
+ &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_8_dspr2>,
&wrapper_nc<vpx_lpf_vertical_8_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_16_dspr2>,
@@ -666,7 +668,8 @@
make_tuple(&vpx_lpf_horizontal_8_msa, &vpx_lpf_horizontal_8_c, 8, 1),
make_tuple(&vpx_lpf_horizontal_16_msa, &vpx_lpf_horizontal_16_c, 8, 1),
make_tuple(&vpx_lpf_horizontal_16_msa, &vpx_lpf_horizontal_16_c, 8, 2),
- make_tuple(&vpx_lpf_vertical_4_msa, &vpx_lpf_vertical_4_c, 8, 1),
+ make_tuple(&wrapper_nc<vpx_lpf_vertical_4_msa>,
+ &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_8_msa>,
&wrapper_nc<vpx_lpf_vertical_8_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_16_msa>,
--- a/vp10/common/loopfilter.c
+++ b/vp10/common/loopfilter.c
@@ -358,11 +358,10 @@
lfi0->hev_thr, lfi1->mblim, lfi1->lim,
lfi1->hev_thr);
} else if (mask_4x4_0 & 1) {- vpx_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr,
- 1);
+ vpx_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr);
} else {vpx_lpf_vertical_4(s + 8 * pitch, pitch, lfi1->mblim, lfi1->lim,
- lfi1->hev_thr, 1);
+ lfi1->hev_thr);
}
}
@@ -373,10 +372,10 @@
lfi1->hev_thr);
} else if (mask_4x4_int_0 & 1) {vpx_lpf_vertical_4(s + 4, pitch, lfi0->mblim, lfi0->lim,
- lfi0->hev_thr, 1);
+ lfi0->hev_thr);
} else {vpx_lpf_vertical_4(s + 8 * pitch + 4, pitch, lfi1->mblim, lfi1->lim,
- lfi1->hev_thr, 1);
+ lfi1->hev_thr);
}
}
}
@@ -1128,11 +1127,11 @@
} else if (mask_8x8 & 1) {vpx_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
} else if (mask_4x4 & 1) {- vpx_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
+ vpx_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
}
}
if (mask_4x4_int & 1)
- vpx_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
+ vpx_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
s += 8;
lfl += 1;
mask_16x16 >>= 1;
--- a/vp9/common/vp9_loopfilter.c
+++ b/vp9/common/vp9_loopfilter.c
@@ -358,11 +358,10 @@
lfi0->hev_thr, lfi1->mblim, lfi1->lim,
lfi1->hev_thr);
} else if (mask_4x4_0 & 1) {- vpx_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr,
- 1);
+ vpx_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr);
} else {vpx_lpf_vertical_4(s + 8 * pitch, pitch, lfi1->mblim, lfi1->lim,
- lfi1->hev_thr, 1);
+ lfi1->hev_thr);
}
}
@@ -373,10 +372,10 @@
lfi1->hev_thr);
} else if (mask_4x4_int_0 & 1) {vpx_lpf_vertical_4(s + 4, pitch, lfi0->mblim, lfi0->lim,
- lfi0->hev_thr, 1);
+ lfi0->hev_thr);
} else {vpx_lpf_vertical_4(s + 8 * pitch + 4, pitch, lfi1->mblim, lfi1->lim,
- lfi1->hev_thr, 1);
+ lfi1->hev_thr);
}
}
}
@@ -1103,11 +1102,11 @@
} else if (mask_8x8 & 1) {vpx_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
} else if (mask_4x4 & 1) {- vpx_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
+ vpx_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
}
}
if (mask_4x4_int & 1)
- vpx_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
+ vpx_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
s += 8;
lfl += 1;
mask_16x16 >>= 1;
--- a/vpx_dsp/arm/loopfilter_4_neon.asm
+++ b/vpx_dsp/arm/loopfilter_4_neon.asm
@@ -79,15 +79,12 @@
; Currently vpx 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 vpx_lpf_vertical_4_neon(uint8_t *s,
; int p /* pitch */,
; const uint8_t *blimit,
; const uint8_t *limit,
-; const uint8_t *thresh,
-; int count)
+; const uint8_t *thresh)
;
; r0 uint8_t *s,
; r1 int p, /* pitch */
@@ -94,22 +91,17 @@
; r2 const uint8_t *blimit,
; r3 const uint8_t *limit,
; sp const uint8_t *thresh,
-; sp+4 int count
|vpx_lpf_vertical_4_neon| PROC
push {lr} vld1.8 {d0[]}, [r2] ; duplicate *blimit- ldr r12, [sp, #8] ; load count
vld1.8 {d1[]}, [r3] ; duplicate *limitldr r3, [sp, #4] ; load thresh
sub r2, r0, #4 ; move s pointer down by 4 columns
- cmp r12, #0
- beq end_vpx_lf_v_edge
vld1.8 {d2[]}, [r3] ; duplicate *thresh-count_lf_v_loop
vld1.u8 {d3}, [r2], r1 ; load s data vld1.u8 {d4}, [r2], r1 vld1.u8 {d5}, [r2], r1@@ -149,12 +141,6 @@
vst4.8 {d4[6], d5[6], d6[6], d7[6]}, [r0], r1 vst4.8 {d4[7], d5[7], d6[7], d7[7]}, [r0]- add r0, r0, r1, lsl #3 ; s += pitch * 8
- subs r12, r12, #1
- subne r2, r0, #4 ; move s pointer down by 4 columns
- bne count_lf_v_loop
-
-end_vpx_lf_v_edge
pop {pc}ENDP ; |vpx_lpf_vertical_4_neon|
--- a/vpx_dsp/arm/loopfilter_4_neon.c
+++ b/vpx_dsp/arm/loopfilter_4_neon.c
@@ -170,8 +170,7 @@
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
- const uint8_t *thresh,
- int count) {+ const uint8_t *thresh) {int i, pitch8;
uint8_t *s;
uint8x8_t dblimit, dlimit, dthresh;
@@ -181,15 +180,12 @@
uint8x8x2_t d2tmp8, d2tmp9, d2tmp10, d2tmp11;
uint8x8x4_t d4Result;
- if (count == 0) // end_vpx_lf_h_edge
- return;
-
dblimit = vld1_u8(blimit);
dlimit = vld1_u8(limit);
dthresh = vld1_u8(thresh);
pitch8 = pitch * 8;
- for (i = 0; i < count; i++, src += pitch8) {+ for (i = 0; i < 1; i++, src += pitch8) {s = src - (i + 1) * 4;
d3u8 = vld1_u8(s);
--- a/vpx_dsp/arm/loopfilter_neon.c
+++ b/vpx_dsp/arm/loopfilter_neon.c
@@ -21,8 +21,8 @@
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {- vpx_lpf_vertical_4_neon(s, p, blimit0, limit0, thresh0, 1);
- vpx_lpf_vertical_4_neon(s + 8 * p, p, blimit1, limit1, thresh1, 1);
+ vpx_lpf_vertical_4_neon(s, p, blimit0, limit0, thresh0);
+ vpx_lpf_vertical_4_neon(s + 8 * p, p, blimit1, limit1, thresh1);
}
#if HAVE_NEON_ASM
--- a/vpx_dsp/loopfilter.c
+++ b/vpx_dsp/loopfilter.c
@@ -143,13 +143,12 @@
}
void vpx_lpf_vertical_4_c(uint8_t *s, int pitch, const uint8_t *blimit,
- const uint8_t *limit, const uint8_t *thresh,
- int count) {+ const uint8_t *limit, const uint8_t *thresh) {int i;
// loop filter designed to work using chars so that we can make maximum use
// of 8 bit simd instructions.
- for (i = 0; i < 8 * count; ++i) {+ for (i = 0; i < 8; ++i) {const uint8_t p3 = s[-4], p2 = s[-3], p1 = s[-2], p0 = s[-1];
const uint8_t q0 = s[0], q1 = s[1], q2 = s[2], q3 = s[3];
const int8_t mask = filter_mask(*limit, *blimit,
@@ -163,9 +162,8 @@
const uint8_t *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) {- vpx_lpf_vertical_4_c(s, pitch, blimit0, limit0, thresh0, 1);
- vpx_lpf_vertical_4_c(s + 8 * pitch, pitch, blimit1, limit1,
- thresh1, 1);
+ vpx_lpf_vertical_4_c(s, pitch, blimit0, limit0, thresh0);
+ vpx_lpf_vertical_4_c(s + 8 * pitch, pitch, blimit1, limit1, thresh1);
}
static INLINE void filter8(int8_t mask, uint8_t thresh, uint8_t flat,
--- a/vpx_dsp/mips/loopfilter_4_msa.c
+++ b/vpx_dsp/mips/loopfilter_4_msa.c
@@ -74,13 +74,10 @@
void vpx_lpf_vertical_4_msa(uint8_t *src, int32_t pitch,
const uint8_t *b_limit_ptr,
const uint8_t *limit_ptr,
- const uint8_t *thresh_ptr,
- int32_t count) {+ const uint8_t *thresh_ptr) {v16u8 mask, hev, flat, limit, thresh, b_limit;
v16u8 p3, p2, p1, p0, q3, q2, q1, q0;
v8i16 vec0, vec1, vec2, vec3;
-
- (void)count;
LD_UB8((src - 4), pitch, p3, p2, p1, p0, q0, q1, q2, q3);
--- a/vpx_dsp/mips/loopfilter_filters_dspr2.c
+++ b/vpx_dsp/mips/loopfilter_filters_dspr2.c
@@ -117,8 +117,7 @@
int pitch,
const uint8_t *blimit,
const uint8_t *limit,
- const uint8_t *thresh,
- int count) {+ const uint8_t *thresh) {uint8_t i;
uint32_t mask, hev;
uint32_t pm1, p0, p1, p2, p3, p4, p5, p6;
@@ -335,8 +334,8 @@
const uint8_t *blimit1,
const uint8_t *limit1,
const uint8_t *thresh1) {- vpx_lpf_vertical_4_dspr2(s, p, blimit0, limit0, thresh0, 1);
- vpx_lpf_vertical_4_dspr2(s + 8 * p, p, blimit1, limit1, thresh1, 1);
+ vpx_lpf_vertical_4_dspr2(s, p, blimit0, limit0, thresh0);
+ vpx_lpf_vertical_4_dspr2(s + 8 * p, p, blimit1, limit1, thresh1);
}
void vpx_lpf_vertical_8_dual_dspr2(uint8_t *s, int p,
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -542,7 +542,7 @@
specialize qw/vpx_lpf_vertical_8_dual sse2 neon_asm dspr2 msa/;
$vpx_lpf_vertical_8_dual_neon_asm=vpx_lpf_vertical_8_dual_neon;
-add_proto qw/void vpx_lpf_vertical_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+add_proto qw/void vpx_lpf_vertical_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
specialize qw/vpx_lpf_vertical_4 mmx 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";
--- a/vpx_dsp/x86/loopfilter_mmx.asm
+++ b/vpx_dsp/x86/loopfilter_mmx.asm
@@ -230,14 +230,13 @@
; int src_pixel_step,
; const char *blimit,
; const char *limit,
-; const char *thresh,
-; int count
+; const char *thresh
;)
global sym(vpx_lpf_vertical_4_mmx) PRIVATE
sym(vpx_lpf_vertical_4_mmx):
push rbp
mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
+ SHADOW_ARGS_TO_STACK 5
GET_GOT rbx
push rsi
push rdi
@@ -254,8 +253,6 @@
lea rsi, [rsi + rax*4 - 4]
- movsxd rcx, dword ptr arg(5) ;count
-.next8_v:
mov rdi, rsi ; rdi points to row +1 for indirect addressing
add rdi, rax
@@ -578,10 +575,6 @@
psrlq mm5, 32
movd [rdi+rax*2+2], mm5
-
- lea rsi, [rsi+rax*8]
- dec rcx
- jnz .next8_v
add rsp, 64
pop rsp