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

vpx_lpf_horizontal_4: remove unused count param

Browse Source 
Change-Id: Iec7d8eda343991f7d7d46931dca17af23c821d11
This commit is contained in:
James Zern
2016-02-11 20:26:54 -08:00
parent bd5a5bb561
commit b1e97c6a25
10 changed files with 38 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
test/lpf_8_test.cc
View File

@@ -458,7 +458,8 @@ using std::tr1::make_tuple;
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
MMX, Loop8Test6Param, MMX, Loop8Test6Param,
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_mmx, &vpx_lpf_horizontal_4_c, 8, 1), make_tuple(&wrapper_nc<vpx_lpf_horizontal_4_mmx>,
&wrapper_nc<vpx_lpf_horizontal_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_4_mmx>, make_tuple(&wrapper_nc<vpx_lpf_vertical_4_mmx>,
&wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1))); &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1)));
#endif // HAVE_MMX #endif // HAVE_MMX
@@ -609,8 +610,8 @@ INSTANTIATE_TEST_CASE_P(
&wrapper_nc<vpx_lpf_horizontal_8_c>, 8, 1), &wrapper_nc<vpx_lpf_horizontal_8_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_8_neon>, make_tuple(&wrapper_nc<vpx_lpf_vertical_8_neon>,
&wrapper_nc<vpx_lpf_vertical_8_c>, 8, 1), &wrapper_nc<vpx_lpf_vertical_8_c>, 8, 1),
make_tuple(&vpx_lpf_horizontal_4_neon, make_tuple(&wrapper_nc<vpx_lpf_horizontal_4_neon>,
&vpx_lpf_horizontal_4_c, 8, 1), &wrapper_nc<vpx_lpf_horizontal_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_vertical_4_neon>, make_tuple(&wrapper_nc<vpx_lpf_vertical_4_neon>,
&wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1))); &wrapper_nc<vpx_lpf_vertical_4_c>, 8, 1)));
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
@@ -633,7 +634,8 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
DSPR2, Loop8Test6Param, DSPR2, Loop8Test6Param,
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_dspr2, &vpx_lpf_horizontal_4_c, 8, 1), make_tuple(&wrapper_nc<vpx_lpf_horizontal_4_dspr2>,
&wrapper_nc<vpx_lpf_horizontal_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_horizontal_8_dspr2>, make_tuple(&wrapper_nc<vpx_lpf_horizontal_8_dspr2>,
&wrapper_nc<vpx_lpf_horizontal_8_c>, 8, 1), &wrapper_nc<vpx_lpf_horizontal_8_c>, 8, 1),
make_tuple(&vpx_lpf_horizontal_16_dspr2, make_tuple(&vpx_lpf_horizontal_16_dspr2,
@@ -666,7 +668,8 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P( INSTANTIATE_TEST_CASE_P(
MSA, Loop8Test6Param, MSA, Loop8Test6Param,
::testing::Values( ::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_msa, &vpx_lpf_horizontal_4_c, 8, 1), make_tuple(&wrapper_nc<vpx_lpf_horizontal_4_msa>,
&wrapper_nc<vpx_lpf_horizontal_4_c>, 8, 1),
make_tuple(&wrapper_nc<vpx_lpf_horizontal_8_msa>, make_tuple(&wrapper_nc<vpx_lpf_horizontal_8_msa>,
&wrapper_nc<vpx_lpf_horizontal_8_c>, 8, 1), &wrapper_nc<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, 1),

16
vp10/common/loopfilter.c
View File

@@ -535,10 +535,10 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
} else { } else {
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
else if (mask_4x4_int & 2) else if (mask_4x4_int & 2)
vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim, vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1); lfin->lim, lfin->hev_thr);
} }
count = 2; count = 2;
} else { } else {
@@ -546,7 +546,7 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} else if (mask_4x4 & 1) { } else if (mask_4x4 & 1) {
if ((mask_4x4 & 3) == 3) { if ((mask_4x4 & 3) == 3) {
@@ -563,22 +563,22 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
} else { } else {
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
else if (mask_4x4_int & 2) else if (mask_4x4_int & 2)
vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim, vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1); lfin->lim, lfin->hev_thr);
} }
count = 2; count = 2;
} else { } else {
vpx_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1); vpx_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} else if (mask_4x4_int & 1) { } else if (mask_4x4_int & 1) {
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} }
s += 8 * count; s += 8 * count;

16
vp9/common/vp9_loopfilter.c
View File

@@ -535,10 +535,10 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
} else { } else {
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
else if (mask_4x4_int & 2) else if (mask_4x4_int & 2)
vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim, vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1); lfin->lim, lfin->hev_thr);
} }
count = 2; count = 2;
} else { } else {
@@ -546,7 +546,7 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} else if (mask_4x4 & 1) { } else if (mask_4x4 & 1) {
if ((mask_4x4 & 3) == 3) { if ((mask_4x4 & 3) == 3) {
@@ -563,22 +563,22 @@ static void filter_selectively_horiz(uint8_t *s, int pitch,
} else { } else {
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
else if (mask_4x4_int & 2) else if (mask_4x4_int & 2)
vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim, vpx_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
lfin->lim, lfin->hev_thr, 1); lfin->lim, lfin->hev_thr);
} }
count = 2; count = 2;
} else { } else {
vpx_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr, 1); vpx_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr);
if (mask_4x4_int & 1) if (mask_4x4_int & 1)
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} else if (mask_4x4_int & 1) { } else if (mask_4x4_int & 1) {
vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim, vpx_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
lfi->hev_thr, 1); lfi->hev_thr);
} }
} }
s += 8 * count; s += 8 * count;

16
vpx_dsp/arm/loopfilter_4_neon.asm
View File

@@ -16,37 +16,28 @@
; Currently vpx only works on iterations 8 at a time. The vp8 loop filter ; Currently vpx only works on iterations 8 at a time. The vp8 loop filter
; works on 16 iterations at a time. ; 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_horizontal_4_neon(uint8_t *s, ; void vpx_lpf_horizontal_4_neon(uint8_t *s,
; int p /* pitch */, ; int p /* 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)
; int count)
; ;
; r0 uint8_t *s, ; r0 uint8_t *s,
; r1 int p, /* pitch */ ; r1 int p, /* pitch */
; 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,
; sp+4 int count
|vpx_lpf_horizontal_4_neon| PROC |vpx_lpf_horizontal_4_neon| PROC
push {lr} push {lr}
vld1.8 {d0[]}, [r2] ; duplicate *blimit vld1.8 {d0[]}, [r2] ; duplicate *blimit
ldr r12, [sp, #8] ; load count
ldr r2, [sp, #4] ; load thresh ldr r2, [sp, #4] ; load thresh
add r1, r1, r1 ; double pitch add r1, r1, r1 ; double pitch
cmp r12, #0
beq end_vpx_lf_h_edge
vld1.8 {d1[]}, [r3] ; duplicate *limit vld1.8 {d1[]}, [r3] ; duplicate *limit
vld1.8 {d2[]}, [r2] ; duplicate *thresh vld1.8 {d2[]}, [r2] ; duplicate *thresh
count_lf_h_loop
sub r2, r0, r1, lsl #1 ; move src pointer down by 4 lines sub r2, r0, r1, lsl #1 ; move src pointer down by 4 lines
add r3, r2, r1, lsr #1 ; set to 3 lines down add r3, r2, r1, lsr #1 ; set to 3 lines down
@@ -69,11 +60,6 @@ count_lf_h_loop
vst1.u8 {d6}, [r2@64], r1 ; store oq0 vst1.u8 {d6}, [r2@64], r1 ; store oq0
vst1.u8 {d7}, [r3@64], r1 ; store oq1 vst1.u8 {d7}, [r3@64], r1 ; store oq1
add r0, r0, #8
subs r12, r12, #1
bne count_lf_h_loop
end_vpx_lf_h_edge
pop {pc} pop {pc}
ENDP ; |vpx_lpf_horizontal_4_neon| ENDP ; |vpx_lpf_horizontal_4_neon|

8
vpx_dsp/arm/loopfilter_4_neon.c
View File

@@ -115,22 +115,18 @@ void vpx_lpf_horizontal_4_neon(
int pitch, 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) {
int count) {
int i; int i;
uint8_t *s, *psrc; uint8_t *s, *psrc;
uint8x8_t dblimit, dlimit, dthresh; uint8x8_t dblimit, dlimit, dthresh;
uint8x8_t d3u8, d4u8, d5u8, d6u8, d7u8, d16u8, d17u8, d18u8; uint8x8_t d3u8, d4u8, d5u8, d6u8, d7u8, d16u8, d17u8, d18u8;
if (count == 0) // end_vpx_lf_h_edge
return;
dblimit = vld1_u8(blimit); dblimit = vld1_u8(blimit);
dlimit = vld1_u8(limit); dlimit = vld1_u8(limit);
dthresh = vld1_u8(thresh); dthresh = vld1_u8(thresh);
psrc = src - (pitch << 2); psrc = src - (pitch << 2);
for (i = 0; i < count; i++) { for (i = 0; i < 1; i++) {
s = psrc + i * 8; s = psrc + i * 8;
d3u8 = vld1_u8(s); d3u8 = vld1_u8(s);

8
vpx_dsp/loopfilter.c
View File

@@ -119,12 +119,12 @@ static INLINE void filter4(int8_t mask, uint8_t thresh, uint8_t *op1,
void vpx_lpf_horizontal_4_c(uint8_t *s, int p /* pitch */, void vpx_lpf_horizontal_4_c(uint8_t *s, int p /* pitch */,
const uint8_t *blimit, const uint8_t *limit, const uint8_t *blimit, const uint8_t *limit,
const uint8_t *thresh, int count) { const uint8_t *thresh) {
int i; int i;
// loop filter designed to work using chars so that we can make maximum use // loop filter designed to work using chars so that we can make maximum use
// of 8 bit simd instructions. // of 8 bit simd instructions.
for (i = 0; i < 8 * count; ++i) { for (i = 0; i < 8; ++i) {
const uint8_t p3 = s[-4 * p], p2 = s[-3 * p], p1 = s[-2 * p], p0 = s[-p]; const uint8_t p3 = s[-4 * p], p2 = s[-3 * p], p1 = s[-2 * p], p0 = s[-p];
const uint8_t q0 = s[0 * p], q1 = s[1 * p], q2 = s[2 * p], q3 = s[3 * p]; const uint8_t q0 = s[0 * p], q1 = s[1 * p], q2 = s[2 * p], q3 = s[3 * p];
const int8_t mask = filter_mask(*limit, *blimit, const int8_t mask = filter_mask(*limit, *blimit,
@@ -138,8 +138,8 @@ void vpx_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 *limit0, const uint8_t *thresh0,
const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *blimit1, const uint8_t *limit1,
const uint8_t *thresh1) { const uint8_t *thresh1) {
vpx_lpf_horizontal_4_c(s, p, blimit0, limit0, thresh0, 1); vpx_lpf_horizontal_4_c(s, p, blimit0, limit0, thresh0);
vpx_lpf_horizontal_4_c(s + 8, p, blimit1, limit1, thresh1, 1); vpx_lpf_horizontal_4_c(s + 8, p, blimit1, limit1, thresh1);
} }
void vpx_lpf_vertical_4_c(uint8_t *s, int pitch, const uint8_t *blimit, void vpx_lpf_vertical_4_c(uint8_t *s, int pitch, const uint8_t *blimit,

5
vpx_dsp/mips/loopfilter_4_msa.c
View File

@@ -13,14 +13,11 @@
void vpx_lpf_horizontal_4_msa(uint8_t *src, int32_t pitch, void vpx_lpf_horizontal_4_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) {
int32_t count) {
uint64_t p1_d, p0_d, q0_d, q1_d; uint64_t p1_d, p0_d, q0_d, q1_d;
v16u8 mask, hev, flat, thresh, b_limit, limit; v16u8 mask, hev, flat, thresh, b_limit, limit;
v16u8 p3, p2, p1, p0, q3, q2, q1, q0, p1_out, p0_out, q0_out, q1_out; v16u8 p3, p2, p1, p0, q3, q2, q1, q0, p1_out, p0_out, q0_out, q1_out;
(void)count;
/* load vector elements */ /* load vector elements */
LD_UB8((src - 4 * pitch), pitch, p3, p2, p1, p0, q0, q1, q2, q3); LD_UB8((src - 4 * pitch), pitch, p3, p2, p1, p0, q0, q1, q2, q3);

7
vpx_dsp/mips/loopfilter_filters_dspr2.c
View File

@@ -23,8 +23,7 @@ void vpx_lpf_horizontal_4_dspr2(unsigned char *s,
int pitch, 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) {
int count) {
uint8_t i; uint8_t i;
uint32_t mask; uint32_t mask;
uint32_t hev; uint32_t hev;
@@ -312,8 +311,8 @@ void vpx_lpf_horizontal_4_dual_dspr2(uint8_t *s, int p /* pitch */,
const uint8_t *blimit1, const uint8_t *blimit1,
const uint8_t *limit1, const uint8_t *limit1,
const uint8_t *thresh1) { const uint8_t *thresh1) {
vpx_lpf_horizontal_4_dspr2(s, p, blimit0, limit0, thresh0, 1); vpx_lpf_horizontal_4_dspr2(s, p, blimit0, limit0, thresh0);
vpx_lpf_horizontal_4_dspr2(s + 8, p, blimit1, limit1, thresh1, 1); vpx_lpf_horizontal_4_dspr2(s + 8, p, blimit1, limit1, thresh1);
} }
void vpx_lpf_horizontal_8_dual_dspr2(uint8_t *s, int p /* pitch */, void vpx_lpf_horizontal_8_dual_dspr2(uint8_t *s, int p /* pitch */,

2
vpx_dsp/vpx_dsp_rtcd_defs.pl
View File

@@ -559,7 +559,7 @@ add_proto qw/void vpx_lpf_horizontal_8_dual/, "uint8_t *s, int pitch, const uint
specialize qw/vpx_lpf_horizontal_8_dual sse2 neon_asm dspr2 msa/; specialize qw/vpx_lpf_horizontal_8_dual sse2 neon_asm dspr2 msa/;
$vpx_lpf_horizontal_8_dual_neon_asm=vpx_lpf_horizontal_8_dual_neon; $vpx_lpf_horizontal_8_dual_neon_asm=vpx_lpf_horizontal_8_dual_neon;
add_proto qw/void vpx_lpf_horizontal_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_horizontal_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
specialize qw/vpx_lpf_horizontal_4 mmx neon dspr2 msa/; specialize qw/vpx_lpf_horizontal_4 mmx neon dspr2 msa/;
add_proto qw/void vpx_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"; add_proto qw/void vpx_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";

12
vpx_dsp/x86/loopfilter_mmx.asm
View File

@@ -18,14 +18,13 @@
; int src_pixel_step, ; int src_pixel_step,
; const char *blimit, ; const char *blimit,
; const char *limit, ; const char *limit,
; const char *thresh, ; const char *thresh
; int count
;) ;)
global sym(vpx_lpf_horizontal_4_mmx) PRIVATE global sym(vpx_lpf_horizontal_4_mmx) PRIVATE
sym(vpx_lpf_horizontal_4_mmx): sym(vpx_lpf_horizontal_4_mmx):
push rbp push rbp
mov rbp, rsp mov rbp, rsp
SHADOW_ARGS_TO_STACK 6 SHADOW_ARGS_TO_STACK 5
GET_GOT rbx GET_GOT rbx
push rsi push rsi
push rdi push rdi
@@ -39,8 +38,6 @@ sym(vpx_lpf_horizontal_4_mmx):
mov rsi, arg(0) ;src_ptr mov rsi, arg(0) ;src_ptr
movsxd rax, dword ptr arg(1) ;src_pixel_step ; destination pitch? movsxd rax, dword ptr arg(1) ;src_pixel_step ; destination pitch?
movsxd rcx, dword ptr arg(5) ;count
.next8_h:
mov rdx, arg(3) ;limit mov rdx, arg(3) ;limit
movq mm7, [rdx] movq mm7, [rdx]
mov rdi, rsi ; rdi points to row +1 for indirect addressing mov rdi, rsi ; rdi points to row +1 for indirect addressing
@@ -208,11 +205,6 @@ sym(vpx_lpf_horizontal_4_mmx):
pxor mm7, [GLOBAL(t80)] ; unoffset pxor mm7, [GLOBAL(t80)] ; unoffset
movq [rdi], mm7 ; write back movq [rdi], mm7 ; write back
add rsi,8
neg rax
dec rcx
jnz .next8_h
add rsp, 32 add rsp, 32
pop rsp pop rsp
; begin epilog ; begin epilog