vpx_dsp: remove mmx variance implementations

there are sse2 equivalents for all remaining variance implementations

James Zern committed 10 years ago

Commit eea8ea81 parent 7aef9795 files · +8 −516Browse files

Unified · Split

5 files

test/variance_test.cc +0 −10
vpx_dsp/vpx_dsp.mk +0 −2
vpx_dsp/vpx_dsp_rtcd_defs.pl +8 −8
vpx_dsp/x86/variance_impl_mmx.asm +0 −343
vpx_dsp/x86/variance_mmx.c +0 −153

@@ -976,16 +976,6 @@ INSTANTIATE_TEST_CASE_P( · 975 unmodified lines
976	make_tuple(2, 2, &vpx_highbd_12_sub_pixel_avg_variance4x4_c, 12)));	976	make_tuple(2, 2, &vpx_highbd_12_sub_pixel_avg_variance4x4_c, 12)));
977	#endif // CONFIG_VP9_HIGHBITDEPTH	977	#endif // CONFIG_VP9_HIGHBITDEPTH
978		978
979	#if HAVE_MMX
980	INSTANTIATE_TEST_CASE_P(
981	MMX, VpxSubpelVarianceTest,
982	::testing::Values(make_tuple(4, 4, &vpx_sub_pixel_variance16x16_mmx, 0),
983	make_tuple(4, 3, &vpx_sub_pixel_variance16x8_mmx, 0),
984	make_tuple(3, 4, &vpx_sub_pixel_variance8x16_mmx, 0),
985	make_tuple(3, 3, &vpx_sub_pixel_variance8x8_mmx, 0),
986	make_tuple(2, 2, &vpx_sub_pixel_variance4x4_mmx, 0)));
987	#endif // HAVE_MMX
988
989	#if HAVE_SSE2	979	#if HAVE_SSE2
990	INSTANTIATE_TEST_CASE_P(SSE2, SumOfSquaresTest,	980	INSTANTIATE_TEST_CASE_P(SSE2, SumOfSquaresTest,
991	::testing::Values(vpx_get_mb_ss_sse2));	981	::testing::Values(vpx_get_mb_ss_sse2));

vpx_dsp/vpx_dsp.mk+0−2

@@ -322,8 +322,6 @@ DSP_SRCS-$(HAVE_NEON) += arm/variance_neon.c · 321 unmodified lines
322	DSP_SRCS-$(HAVE_MSA) += mips/variance_msa.c	322	DSP_SRCS-$(HAVE_MSA) += mips/variance_msa.c
323	DSP_SRCS-$(HAVE_MSA) += mips/sub_pixel_variance_msa.c	323	DSP_SRCS-$(HAVE_MSA) += mips/sub_pixel_variance_msa.c
324		324
325	DSP_SRCS-$(HAVE_MMX) += x86/variance_mmx.c
326	DSP_SRCS-$(HAVE_MMX) += x86/variance_impl_mmx.asm
327	DSP_SRCS-$(HAVE_SSE) += x86/variance_sse2.c	325	DSP_SRCS-$(HAVE_SSE) += x86/variance_sse2.c
328	DSP_SRCS-$(HAVE_SSE2) += x86/variance_sse2.c # Contains SSE2 and SSSE3	326	DSP_SRCS-$(HAVE_SSE2) += x86/variance_sse2.c # Contains SSE2 and SSSE3
329	DSP_SRCS-$(HAVE_SSE2) += x86/halfpix_variance_sse2.c	327	DSP_SRCS-$(HAVE_SSE2) += x86/halfpix_variance_sse2.c

vpx_dsp/vpx_dsp_rtcd_defs.pl+8−8

vpx_dsp/x86/variance_impl_mmx.asm+0−343

@@ -1,343 +0,0 @@
1	;
2	; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3	;
4	; Use of this source code is governed by a BSD-style license
5	; that can be found in the LICENSE file in the root of the source
6	; tree. An additional intellectual property rights grant can be found
7	; in the file PATENTS. All contributing project authors may
8	; be found in the AUTHORS file in the root of the source tree.
9	;
10
11
12	%include "vpx_ports/x86_abi_support.asm"
13
14	%define mmx_filter_shift 7
15
16	;void vpx_filter_block2d_bil4x4_var_mmx
17	;(
18	; unsigned char *ref_ptr,
19	; int ref_pixels_per_line,
20	; unsigned char *src_ptr,
21	; int src_pixels_per_line,
22	; unsigned short *HFilter,
23	; unsigned short *VFilter,
24	; int *sum,
25	; unsigned int *sumsquared
26	;)
27	global sym(vpx_filter_block2d_bil4x4_var_mmx) PRIVATE
28	sym(vpx_filter_block2d_bil4x4_var_mmx):
29	push rbp
30	mov rbp, rsp
31	SHADOW_ARGS_TO_STACK 8
32	GET_GOT rbx
33	push rsi
34	push rdi
35	sub rsp, 16
36	; end prolog
37
38	pxor mm6, mm6 ;
39	pxor mm7, mm7 ;
40
41	mov rax, arg(4) ;HFilter ;
42	mov rdx, arg(5) ;VFilter ;
43
44	mov rsi, arg(0) ;ref_ptr ;
45	mov rdi, arg(2) ;src_ptr ;
46
47	mov rcx, 4 ;
48	pxor mm0, mm0 ;
49
50	movd mm1, [rsi] ;
51	movd mm3, [rsi+1] ;
52
53	punpcklbw mm1, mm0 ;
54	pmullw mm1, [rax] ;
55
56	punpcklbw mm3, mm0 ;
57	pmullw mm3, [rax+8] ;
58
59	paddw mm1, mm3 ;
60	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
61
62	psraw mm1, mmx_filter_shift ;
63	movq mm5, mm1
64
65	%if ABI_IS_32BIT
66	add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
67	%else
68	movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
69	add rsi, r8
70	%endif
71
72	.filter_block2d_bil4x4_var_mmx_loop:
73
74	movd mm1, [rsi] ;
75	movd mm3, [rsi+1] ;
76
77	punpcklbw mm1, mm0 ;
78	pmullw mm1, [rax] ;
79
80	punpcklbw mm3, mm0 ;
81	pmullw mm3, [rax+8] ;
82
83	paddw mm1, mm3 ;
84	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
85
86	psraw mm1, mmx_filter_shift ;
87	movq mm3, mm5 ;
88
89	movq mm5, mm1 ;
90	pmullw mm3, [rdx] ;
91
92	pmullw mm1, [rdx+8] ;
93	paddw mm1, mm3 ;
94
95	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
96	psraw mm1, mmx_filter_shift ;
97
98	movd mm3, [rdi] ;
99	punpcklbw mm3, mm0 ;
100
101	psubw mm1, mm3 ;
102	paddw mm6, mm1 ;
103
104	pmaddwd mm1, mm1 ;
105	paddd mm7, mm1 ;
106
107	%if ABI_IS_32BIT
108	add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
109	add rdi, dword ptr arg(3) ;src_pixels_per_line ;
110	%else
111	movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
112	movsxd r9, dword ptr arg(3) ;src_pixels_per_line
113	add rsi, r8
114	add rdi, r9
115	%endif
116	sub rcx, 1 ;
117	jnz .filter_block2d_bil4x4_var_mmx_loop ;
118
119	pxor mm3, mm3 ;
120	pxor mm2, mm2 ;
121
122	punpcklwd mm2, mm6 ;
123	punpckhwd mm3, mm6 ;
124
125	paddd mm2, mm3 ;
126	movq mm6, mm2 ;
127
128	psrlq mm6, 32 ;
129	paddd mm2, mm6 ;
130
131	psrad mm2, 16 ;
132	movq mm4, mm7 ;
133
134	psrlq mm4, 32 ;
135	paddd mm4, mm7 ;
136
137	mov rdi, arg(6) ;sum
138	mov rsi, arg(7) ;sumsquared
139
140	movd dword ptr [rdi], mm2 ;
141	movd dword ptr [rsi], mm4 ;
142
143	; begin epilog
144	add rsp, 16
145	pop rdi
146	pop rsi
147	RESTORE_GOT
148	UNSHADOW_ARGS
149	pop rbp
150	ret
151
152	;void vpx_filter_block2d_bil_var_mmx
153	;(
154	; unsigned char *ref_ptr,
155	; int ref_pixels_per_line,
156	; unsigned char *src_ptr,
157	; int src_pixels_per_line,
158	; unsigned int Height,
159	; unsigned short *HFilter,
160	; unsigned short *VFilter,
161	; int *sum,
162	; unsigned int *sumsquared
163	;)
164	global sym(vpx_filter_block2d_bil_var_mmx) PRIVATE
165	sym(vpx_filter_block2d_bil_var_mmx):
166	push rbp
167	mov rbp, rsp
168	SHADOW_ARGS_TO_STACK 9
169	GET_GOT rbx
170	push rsi
171	push rdi
172	sub rsp, 16
173	; end prolog
174
175	pxor mm6, mm6 ;
176	pxor mm7, mm7 ;
177	mov rax, arg(5) ;HFilter ;
178
179	mov rdx, arg(6) ;VFilter ;
180	mov rsi, arg(0) ;ref_ptr ;
181
182	mov rdi, arg(2) ;src_ptr ;
183	movsxd rcx, dword ptr arg(4) ;Height ;
184
185	pxor mm0, mm0 ;
186	movq mm1, [rsi] ;
187
188	movq mm3, [rsi+1] ;
189	movq mm2, mm1 ;
190
191	movq mm4, mm3 ;
192	punpcklbw mm1, mm0 ;
193
194	punpckhbw mm2, mm0 ;
195	pmullw mm1, [rax] ;
196
197	pmullw mm2, [rax] ;
198	punpcklbw mm3, mm0 ;
199
200	punpckhbw mm4, mm0 ;
201	pmullw mm3, [rax+8] ;
202
203	pmullw mm4, [rax+8] ;
204	paddw mm1, mm3 ;
205
206	paddw mm2, mm4 ;
207	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
208
209	psraw mm1, mmx_filter_shift ;
210	paddw mm2, [GLOBAL(mmx_bi_rd)] ;
211
212	psraw mm2, mmx_filter_shift ;
213	movq mm5, mm1
214
215	packuswb mm5, mm2 ;
216	%if ABI_IS_32BIT
217	add rsi, dword ptr arg(1) ;ref_pixels_per_line
218	%else
219	movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
220	add rsi, r8
221	%endif
222
223	.filter_block2d_bil_var_mmx_loop:
224
225	movq mm1, [rsi] ;
226	movq mm3, [rsi+1] ;
227
228	movq mm2, mm1 ;
229	movq mm4, mm3 ;
230
231	punpcklbw mm1, mm0 ;
232	punpckhbw mm2, mm0 ;
233
234	pmullw mm1, [rax] ;
235	pmullw mm2, [rax] ;
236
237	punpcklbw mm3, mm0 ;
238	punpckhbw mm4, mm0 ;
239
240	pmullw mm3, [rax+8] ;
241	pmullw mm4, [rax+8] ;
242
243	paddw mm1, mm3 ;
244	paddw mm2, mm4 ;
245
246	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
247	psraw mm1, mmx_filter_shift ;
248
249	paddw mm2, [GLOBAL(mmx_bi_rd)] ;
250	psraw mm2, mmx_filter_shift ;
251
252	movq mm3, mm5 ;
253	movq mm4, mm5 ;
254
255	punpcklbw mm3, mm0 ;
256	punpckhbw mm4, mm0 ;
257
258	movq mm5, mm1 ;
259	packuswb mm5, mm2 ;
260
261	pmullw mm3, [rdx] ;
262	pmullw mm4, [rdx] ;
263
264	pmullw mm1, [rdx+8] ;
265	pmullw mm2, [rdx+8] ;
266
267	paddw mm1, mm3 ;
268	paddw mm2, mm4 ;
269
270	paddw mm1, [GLOBAL(mmx_bi_rd)] ;
271	paddw mm2, [GLOBAL(mmx_bi_rd)] ;
272
273	psraw mm1, mmx_filter_shift ;
274	psraw mm2, mmx_filter_shift ;
275
276	movq mm3, [rdi] ;
277	movq mm4, mm3 ;
278
279	punpcklbw mm3, mm0 ;
280	punpckhbw mm4, mm0 ;
281
282	psubw mm1, mm3 ;
283	psubw mm2, mm4 ;
284
285	paddw mm6, mm1 ;
286	pmaddwd mm1, mm1 ;
287
288	paddw mm6, mm2 ;
289	pmaddwd mm2, mm2 ;
290
291	paddd mm7, mm1 ;
292	paddd mm7, mm2 ;
293
294	%if ABI_IS_32BIT
295	add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
296	add rdi, dword ptr arg(3) ;src_pixels_per_line ;
297	%else
298	movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
299	movsxd r9, dword ptr arg(3) ;src_pixels_per_line ;
300	add rsi, r8
301	add rdi, r9
302	%endif
303	sub rcx, 1 ;
304	jnz .filter_block2d_bil_var_mmx_loop ;
305
306	pxor mm3, mm3 ;
307	pxor mm2, mm2 ;
308
309	punpcklwd mm2, mm6 ;
310	punpckhwd mm3, mm6 ;
311
312	paddd mm2, mm3 ;
313	movq mm6, mm2 ;
314
315	psrlq mm6, 32 ;
316	paddd mm2, mm6 ;
317
318	psrad mm2, 16 ;
319	movq mm4, mm7 ;
320
321	psrlq mm4, 32 ;
322	paddd mm4, mm7 ;
323
324	mov rdi, arg(7) ;sum
325	mov rsi, arg(8) ;sumsquared
326
327	movd dword ptr [rdi], mm2 ;
328	movd dword ptr [rsi], mm4 ;
329
330	; begin epilog
331	add rsp, 16
332	pop rdi
333	pop rsi
334	RESTORE_GOT
335	UNSHADOW_ARGS
336	pop rbp
337	ret
338
339	SECTION_RODATA
340	;short mmx_bi_rd[4] = { 64, 64, 64, 64};
341	align 16
342	mmx_bi_rd:
343	times 4 dw 64

vpx_dsp/x86/variance_mmx.c+0−153

@@ -1,153 +0,0 @@
1	/*
2	* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3	*
4	* Use of this source code is governed by a BSD-style license
5	* that can be found in the LICENSE file in the root of the source
6	* tree. An additional intellectual property rights grant can be found
7	* in the file PATENTS. All contributing project authors may
8	* be found in the AUTHORS file in the root of the source tree.
9	*/
10
11	#include "./vpx_dsp_rtcd.h"
12
13	#include "vpx_ports/mem.h"
14
15	DECLARE_ALIGNED(16, static const int16_t, bilinear_filters_mmx[8][8]) = {
16	{ 128, 128, 128, 128, 0, 0, 0, 0 },
17	{ 112, 112, 112, 112, 16, 16, 16, 16 },
18	{ 96, 96, 96, 96, 32, 32, 32, 32 },
19	{ 80, 80, 80, 80, 48, 48, 48, 48 },
20	{ 64, 64, 64, 64, 64, 64, 64, 64 },
21	{ 48, 48, 48, 48, 80, 80, 80, 80 },
22	{ 32, 32, 32, 32, 96, 96, 96, 96 },
23	{ 16, 16, 16, 16, 112, 112, 112, 112 }
24	};
25
26	extern void vpx_filter_block2d_bil4x4_var_mmx(const unsigned char *ref_ptr,
27	int ref_pixels_per_line,
28	const unsigned char *src_ptr,
29	int src_pixels_per_line,
30	const int16_t *HFilter,
31	const int16_t *VFilter,
32	int *sum,
33	unsigned int *sumsquared);
34
35	extern void vpx_filter_block2d_bil_var_mmx(const unsigned char *ref_ptr,
36	int ref_pixels_per_line,
37	const unsigned char *src_ptr,
38	int src_pixels_per_line,
39	unsigned int Height,
40	const int16_t *HFilter,
41	const int16_t *VFilter,
42	int *sum,
43	unsigned int *sumsquared);
44
45
46	uint32_t vpx_sub_pixel_variance4x4_mmx(const uint8_t *a, int a_stride,
47	int xoffset, int yoffset,
48	const uint8_t *b, int b_stride,
49	uint32_t *sse) {
50	int xsum;
51	unsigned int xxsum;
52	vpx_filter_block2d_bil4x4_var_mmx(a, a_stride, b, b_stride,
53	bilinear_filters_mmx[xoffset],
54	bilinear_filters_mmx[yoffset],
55	&xsum, &xxsum);
56	*sse = xxsum;
57	return (xxsum - (((unsigned int)xsum * xsum) >> 4));
58	}
59
60
61	uint32_t vpx_sub_pixel_variance8x8_mmx(const uint8_t *a, int a_stride,
62	int xoffset, int yoffset,
63	const uint8_t *b, int b_stride,
64	uint32_t *sse) {
65	int xsum;
66	uint32_t xxsum;
67	vpx_filter_block2d_bil_var_mmx(a, a_stride, b, b_stride, 8,
68	bilinear_filters_mmx[xoffset],
69	bilinear_filters_mmx[yoffset],
70	&xsum, &xxsum);
71	*sse = xxsum;
72	return (xxsum - (((uint32_t)xsum * xsum) >> 6));
73	}
74
75	uint32_t vpx_sub_pixel_variance16x16_mmx(const uint8_t *a, int a_stride,
76	int xoffset, int yoffset,
77	const uint8_t *b, int b_stride,
78	uint32_t *sse) {
79	int xsum0, xsum1;
80	unsigned int xxsum0, xxsum1;
81
82	vpx_filter_block2d_bil_var_mmx(a, a_stride, b, b_stride, 16,
83	bilinear_filters_mmx[xoffset],
84	bilinear_filters_mmx[yoffset],
85	&xsum0, &xxsum0);
86
87	vpx_filter_block2d_bil_var_mmx(a + 8, a_stride, b + 8, b_stride, 16,
88	bilinear_filters_mmx[xoffset],
89	bilinear_filters_mmx[yoffset],
90	&xsum1, &xxsum1);
91
92	xsum0 += xsum1;
93	xxsum0 += xxsum1;
94
95	*sse = xxsum0;
96	return (xxsum0 - (((uint32_t)xsum0 * xsum0) >> 8));
97	}
98
99	uint32_t vpx_sub_pixel_variance16x8_mmx(const uint8_t *a, int a_stride,
100	int xoffset, int yoffset,
101	const uint8_t *b, int b_stride,
102	uint32_t *sse) {
103	int xsum0, xsum1;
104	unsigned int xxsum0, xxsum1;
105
106	vpx_filter_block2d_bil_var_mmx(a, a_stride, b, b_stride, 8,
107	bilinear_filters_mmx[xoffset],
108	bilinear_filters_mmx[yoffset],
109	&xsum0, &xxsum0);
110
111	vpx_filter_block2d_bil_var_mmx(a + 8, a_stride, b + 8, b_stride, 8,
112	bilinear_filters_mmx[xoffset],
113	bilinear_filters_mmx[yoffset],
114	&xsum1, &xxsum1);
115
116	xsum0 += xsum1;
117	xxsum0 += xxsum1;
118
119	*sse = xxsum0;
120	return (xxsum0 - (((uint32_t)xsum0 * xsum0) >> 7));
121	}
122
123	uint32_t vpx_sub_pixel_variance8x16_mmx(const uint8_t *a, int a_stride,
124	int xoffset, int yoffset,
125	const uint8_t *b, int b_stride,
126	uint32_t *sse) {
127	int xsum;
128	unsigned int xxsum;
129	vpx_filter_block2d_bil_var_mmx(a, a_stride, b, b_stride, 16,
130	bilinear_filters_mmx[xoffset],
131	bilinear_filters_mmx[yoffset],
132	&xsum, &xxsum);
133	*sse = xxsum;
134	return (xxsum - (((uint32_t)xsum * xsum) >> 7));
135	}
136
137	uint32_t vpx_variance_halfpixvar16x16_h_mmx(const uint8_t *a, int a_stride,
138	const uint8_t *b, int b_stride,
139	uint32_t *sse) {
140	return vpx_sub_pixel_variance16x16_mmx(a, a_stride, 4, 0, b, b_stride, sse);
141	}
142
143	uint32_t vpx_variance_halfpixvar16x16_v_mmx(const uint8_t *a, int a_stride,
144	const uint8_t *b, int b_stride,
145	uint32_t *sse) {
146	return vpx_sub_pixel_variance16x16_mmx(a, a_stride, 0, 4, b, b_stride, sse);
147	}
148
149	uint32_t vpx_variance_halfpixvar16x16_hv_mmx(const uint8_t *a, int a_stride,
150	const uint8_t *b, int b_stride,
151	uint32_t *sse) {
152	return vpx_sub_pixel_variance16x16_mmx(a, a_stride, 4, 4, b, b_stride, sse);
153	}

vpx_dsp: remove mmx variance implementations

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code

@@ -1478,16 +1478,16 @@ add_proto qw/uint32_t vpx_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int · 1477 unmodified lines
1478	specialize qw/vpx_sub_pixel_variance16x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";	1478	specialize qw/vpx_sub_pixel_variance16x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";
1479		1479
1480	add_proto qw/uint32_t vpx_sub_pixel_variance16x16/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1480	add_proto qw/uint32_t vpx_sub_pixel_variance16x16/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1481	specialize qw/vpx_sub_pixel_variance16x16 mmx media neon msa/, "$sse2_x86inc", "$ssse3_x86inc";	1481	specialize qw/vpx_sub_pixel_variance16x16 media neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
1482		1482
1483	add_proto qw/uint32_t vpx_sub_pixel_variance16x8/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1483	add_proto qw/uint32_t vpx_sub_pixel_variance16x8/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1484	specialize qw/vpx_sub_pixel_variance16x8 mmx msa/, "$sse2_x86inc", "$ssse3_x86inc";	1484	specialize qw/vpx_sub_pixel_variance16x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";
1485		1485
1486	add_proto qw/uint32_t vpx_sub_pixel_variance8x16/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1486	add_proto qw/uint32_t vpx_sub_pixel_variance8x16/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1487	specialize qw/vpx_sub_pixel_variance8x16 mmx msa/, "$sse2_x86inc", "$ssse3_x86inc";	1487	specialize qw/vpx_sub_pixel_variance8x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
1488		1488
1489	add_proto qw/uint32_t vpx_sub_pixel_variance8x8/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1489	add_proto qw/uint32_t vpx_sub_pixel_variance8x8/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1490	specialize qw/vpx_sub_pixel_variance8x8 mmx media neon msa/, "$sse2_x86inc", "$ssse3_x86inc";	1490	specialize qw/vpx_sub_pixel_variance8x8 media neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
1491		1491
1492	add_proto qw/uint32_t vpx_sub_pixel_variance8x4/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1492	add_proto qw/uint32_t vpx_sub_pixel_variance8x4/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1493	specialize qw/vpx_sub_pixel_variance8x4 msa/, "$sse2_x86inc", "$ssse3_x86inc";	1493	specialize qw/vpx_sub_pixel_variance8x4 msa/, "$sse2_x86inc", "$ssse3_x86inc";
@@ -1496,7 +1496,7 @@ add_proto qw/uint32_t vpx_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int s · 2 unmodified lines
1496	specialize qw/vpx_sub_pixel_variance4x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";	1496	specialize qw/vpx_sub_pixel_variance4x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";
1497		1497
1498	add_proto qw/uint32_t vpx_sub_pixel_variance4x4/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";	1498	add_proto qw/uint32_t vpx_sub_pixel_variance4x4/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t *sse";
1499	specialize qw/vpx_sub_pixel_variance4x4 mmx msa/, "$sse2_x86inc", "$ssse3_x86inc";	1499	specialize qw/vpx_sub_pixel_variance4x4 msa/, "$sse2_x86inc", "$ssse3_x86inc";
1500		1500
1501	add_proto qw/uint32_t vpx_sub_pixel_avg_variance64x64/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t sse, const uint8_t second_pred";	1501	add_proto qw/uint32_t vpx_sub_pixel_avg_variance64x64/, "const uint8_t src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t ref_ptr, int ref_stride, uint32_t sse, const uint8_t second_pred";
1502	specialize qw/vpx_sub_pixel_avg_variance64x64 avx2 msa/, "$sse2_x86inc", "$ssse3_x86inc";	1502	specialize qw/vpx_sub_pixel_avg_variance64x64 avx2 msa/, "$sse2_x86inc", "$ssse3_x86inc";
@@ -1541,13 +1541,13 @@ add_proto qw/uint32_t vpx_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, i · 38 unmodified lines
1541	# Specialty Subpixel	1541	# Specialty Subpixel
1542	#	1542	#
1543	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_h/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";	1543	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_h/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";
1544	specialize qw/vpx_variance_halfpixvar16x16_h mmx sse2 media/;	1544	specialize qw/vpx_variance_halfpixvar16x16_h sse2 media/;
1545		1545
1546	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_v/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";	1546	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_v/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";
1547	specialize qw/vpx_variance_halfpixvar16x16_v mmx sse2 media/;	1547	specialize qw/vpx_variance_halfpixvar16x16_v sse2 media/;
1548		1548
1549	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_hv/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";	1549	add_proto qw/uint32_t vpx_variance_halfpixvar16x16_hv/, "const unsigned char src_ptr, int source_stride, const unsigned char ref_ptr, int ref_stride, uint32_t *sse";
1550	specialize qw/vpx_variance_halfpixvar16x16_hv mmx sse2 media/;	1550	specialize qw/vpx_variance_halfpixvar16x16_hv sse2 media/;
1551		1551
1552	if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {	1552	if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
1553	add_proto qw/unsigned int vpx_highbd_12_variance64x64/, "const uint8_t src_ptr, int source_stride, const uint8_t ref_ptr, int ref_stride, unsigned int *sse";	1553	add_proto qw/unsigned int vpx_highbd_12_variance64x64/, "const uint8_t src_ptr, int source_stride, const uint8_t ref_ptr, int ref_stride, unsigned int *sse";