• file
• revisions
• annotate
• diff
• comparison
• raw

media/libvpx/vp9/encoder/x86/vp9_variance_sse2.c@b8a032363ba2 (annotated)

media/libvpx/vp9/encoder/x86/vp9_variance_sse2.c

Thu, 22 Jan 2015 13:21:57 +0100

author

Michael Schloh von Bennewitz <michael@schloh.com>

date

Thu, 22 Jan 2015 13:21:57 +0100

branch

TOR_BUG_9701

changeset 15

b8a032363ba2

permissions

-rw-r--r--

Incorporate requested changes from Mozilla in review:
https://bugzilla.mozilla.org/show_bug.cgi?id=1123480#c6

michael@0	1	/*
michael@0	2	* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
michael@0	3	*
michael@0	4	* Use of this source code is governed by a BSD-style license
michael@0	5	* that can be found in the LICENSE file in the root of the source
michael@0	6	* tree. An additional intellectual property rights grant can be found
michael@0	7	* in the file PATENTS. All contributing project authors may
michael@0	8	* be found in the AUTHORS file in the root of the source tree.
michael@0	9	*/
michael@0	10
michael@0	11	#include "./vpx_config.h"
michael@0	12
michael@0	13	#include "vp9/encoder/vp9_variance.h"
michael@0	14	#include "vp9/common/vp9_pragmas.h"
michael@0	15	#include "vpx_ports/mem.h"
michael@0	16
michael@0	17	extern unsigned int vp9_get4x4var_mmx
michael@0	18	(
michael@0	19	const unsigned char *src_ptr,
michael@0	20	int source_stride,
michael@0	21	const unsigned char *ref_ptr,
michael@0	22	int recon_stride,
michael@0	23	unsigned int *SSE,
michael@0	24	int *Sum
michael@0	25	);
michael@0	26
michael@0	27	unsigned int vp9_get_mb_ss_sse2
michael@0	28	(
michael@0	29	const int16_t *src_ptr
michael@0	30	);
michael@0	31	unsigned int vp9_get16x16var_sse2
michael@0	32	(
michael@0	33	const unsigned char *src_ptr,
michael@0	34	int source_stride,
michael@0	35	const unsigned char *ref_ptr,
michael@0	36	int recon_stride,
michael@0	37	unsigned int *SSE,
michael@0	38	int *Sum
michael@0	39	);
michael@0	40	unsigned int vp9_get8x8var_sse2
michael@0	41	(
michael@0	42	const unsigned char *src_ptr,
michael@0	43	int source_stride,
michael@0	44	const unsigned char *ref_ptr,
michael@0	45	int recon_stride,
michael@0	46	unsigned int *SSE,
michael@0	47	int *Sum
michael@0	48	);
michael@0	49	void vp9_half_horiz_vert_variance8x_h_sse2
michael@0	50	(
michael@0	51	const unsigned char *ref_ptr,
michael@0	52	int ref_pixels_per_line,
michael@0	53	const unsigned char *src_ptr,
michael@0	54	int src_pixels_per_line,
michael@0	55	unsigned int Height,
michael@0	56	int *sum,
michael@0	57	unsigned int *sumsquared
michael@0	58	);
michael@0	59	void vp9_half_horiz_vert_variance16x_h_sse2
michael@0	60	(
michael@0	61	const unsigned char *ref_ptr,
michael@0	62	int ref_pixels_per_line,
michael@0	63	const unsigned char *src_ptr,
michael@0	64	int src_pixels_per_line,
michael@0	65	unsigned int Height,
michael@0	66	int *sum,
michael@0	67	unsigned int *sumsquared
michael@0	68	);
michael@0	69	void vp9_half_horiz_variance8x_h_sse2
michael@0	70	(
michael@0	71	const unsigned char *ref_ptr,
michael@0	72	int ref_pixels_per_line,
michael@0	73	const unsigned char *src_ptr,
michael@0	74	int src_pixels_per_line,
michael@0	75	unsigned int Height,
michael@0	76	int *sum,
michael@0	77	unsigned int *sumsquared
michael@0	78	);
michael@0	79	void vp9_half_horiz_variance16x_h_sse2
michael@0	80	(
michael@0	81	const unsigned char *ref_ptr,
michael@0	82	int ref_pixels_per_line,
michael@0	83	const unsigned char *src_ptr,
michael@0	84	int src_pixels_per_line,
michael@0	85	unsigned int Height,
michael@0	86	int *sum,
michael@0	87	unsigned int *sumsquared
michael@0	88	);
michael@0	89	void vp9_half_vert_variance8x_h_sse2
michael@0	90	(
michael@0	91	const unsigned char *ref_ptr,
michael@0	92	int ref_pixels_per_line,
michael@0	93	const unsigned char *src_ptr,
michael@0	94	int src_pixels_per_line,
michael@0	95	unsigned int Height,
michael@0	96	int *sum,
michael@0	97	unsigned int *sumsquared
michael@0	98	);
michael@0	99	void vp9_half_vert_variance16x_h_sse2
michael@0	100	(
michael@0	101	const unsigned char *ref_ptr,
michael@0	102	int ref_pixels_per_line,
michael@0	103	const unsigned char *src_ptr,
michael@0	104	int src_pixels_per_line,
michael@0	105	unsigned int Height,
michael@0	106	int *sum,
michael@0	107	unsigned int *sumsquared
michael@0	108	);
michael@0	109
michael@0	110	typedef unsigned int (*get_var_sse2) (
michael@0	111	const unsigned char *src_ptr,
michael@0	112	int source_stride,
michael@0	113	const unsigned char *ref_ptr,
michael@0	114	int recon_stride,
michael@0	115	unsigned int *SSE,
michael@0	116	int *Sum
michael@0	117	);
michael@0	118
michael@0	119	static void variance_sse2(const unsigned char *src_ptr, int source_stride,
michael@0	120	const unsigned char *ref_ptr, int recon_stride,
michael@0	121	int w, int h, unsigned int *sse, int *sum,
michael@0	122	get_var_sse2 var_fn, int block_size) {
michael@0	123	unsigned int sse0;
michael@0	124	int sum0;
michael@0	125	int i, j;
michael@0	126
michael@0	127	*sse = 0;
michael@0	128	*sum = 0;
michael@0	129
michael@0	130	for (i = 0; i < h; i += block_size) {
michael@0	131	for (j = 0; j < w; j += block_size) {
michael@0	132	var_fn(src_ptr + source_stride * i + j, source_stride,
michael@0	133	ref_ptr + recon_stride * i + j, recon_stride, &sse0, &sum0);
michael@0	134	*sse += sse0;
michael@0	135	*sum += sum0;
michael@0	136	}
michael@0	137	}
michael@0	138	}
michael@0	139
michael@0	140	unsigned int vp9_variance4x4_sse2(
michael@0	141	const unsigned char *src_ptr,
michael@0	142	int source_stride,
michael@0	143	const unsigned char *ref_ptr,
michael@0	144	int recon_stride,
michael@0	145	unsigned int *sse) {
michael@0	146	unsigned int var;
michael@0	147	int avg;
michael@0	148
michael@0	149	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 4, 4,
michael@0	150	&var, &avg, vp9_get4x4var_mmx, 4);
michael@0	151	*sse = var;
michael@0	152	return (var - (((unsigned int)avg * avg) >> 4));
michael@0	153	}
michael@0	154
michael@0	155	unsigned int vp9_variance8x4_sse2(const uint8_t *src_ptr,
michael@0	156	int source_stride,
michael@0	157	const uint8_t *ref_ptr,
michael@0	158	int recon_stride,
michael@0	159	unsigned int *sse) {
michael@0	160	unsigned int var;
michael@0	161	int avg;
michael@0	162
michael@0	163	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 8, 4,
michael@0	164	&var, &avg, vp9_get4x4var_mmx, 4);
michael@0	165	*sse = var;
michael@0	166	return (var - (((unsigned int)avg * avg) >> 5));
michael@0	167	}
michael@0	168
michael@0	169	unsigned int vp9_variance4x8_sse2(const uint8_t *src_ptr,
michael@0	170	int source_stride,
michael@0	171	const uint8_t *ref_ptr,
michael@0	172	int recon_stride,
michael@0	173	unsigned int *sse) {
michael@0	174	unsigned int var;
michael@0	175	int avg;
michael@0	176
michael@0	177	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 4, 8,
michael@0	178	&var, &avg, vp9_get4x4var_mmx, 4);
michael@0	179	*sse = var;
michael@0	180	return (var - (((unsigned int)avg * avg) >> 5));
michael@0	181	}
michael@0	182
michael@0	183	unsigned int vp9_variance8x8_sse2
michael@0	184	(
michael@0	185	const unsigned char *src_ptr,
michael@0	186	int source_stride,
michael@0	187	const unsigned char *ref_ptr,
michael@0	188	int recon_stride,
michael@0	189	unsigned int *sse) {
michael@0	190	unsigned int var;
michael@0	191	int avg;
michael@0	192
michael@0	193	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8,
michael@0	194	&var, &avg, vp9_get8x8var_sse2, 8);
michael@0	195	*sse = var;
michael@0	196	return (var - (((unsigned int)avg * avg) >> 6));
michael@0	197	}
michael@0	198
michael@0	199	unsigned int vp9_variance16x8_sse2
michael@0	200	(
michael@0	201	const unsigned char *src_ptr,
michael@0	202	int source_stride,
michael@0	203	const unsigned char *ref_ptr,
michael@0	204	int recon_stride,
michael@0	205	unsigned int *sse) {
michael@0	206	unsigned int var;
michael@0	207	int avg;
michael@0	208
michael@0	209	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8,
michael@0	210	&var, &avg, vp9_get8x8var_sse2, 8);
michael@0	211	*sse = var;
michael@0	212	return (var - (((unsigned int)avg * avg) >> 7));
michael@0	213	}
michael@0	214
michael@0	215	unsigned int vp9_variance8x16_sse2
michael@0	216	(
michael@0	217	const unsigned char *src_ptr,
michael@0	218	int source_stride,
michael@0	219	const unsigned char *ref_ptr,
michael@0	220	int recon_stride,
michael@0	221	unsigned int *sse) {
michael@0	222	unsigned int var;
michael@0	223	int avg;
michael@0	224
michael@0	225	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16,
michael@0	226	&var, &avg, vp9_get8x8var_sse2, 8);
michael@0	227	*sse = var;
michael@0	228	return (var - (((unsigned int)avg * avg) >> 7));
michael@0	229	}
michael@0	230
michael@0	231	unsigned int vp9_variance16x16_sse2
michael@0	232	(
michael@0	233	const unsigned char *src_ptr,
michael@0	234	int source_stride,
michael@0	235	const unsigned char *ref_ptr,
michael@0	236	int recon_stride,
michael@0	237	unsigned int *sse) {
michael@0	238	unsigned int var;
michael@0	239	int avg;
michael@0	240
michael@0	241	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16,
michael@0	242	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	243	*sse = var;
michael@0	244	return (var - (((unsigned int)avg * avg) >> 8));
michael@0	245	}
michael@0	246
michael@0	247	unsigned int vp9_mse16x16_sse2(
michael@0	248	const unsigned char *src_ptr,
michael@0	249	int source_stride,
michael@0	250	const unsigned char *ref_ptr,
michael@0	251	int recon_stride,
michael@0	252	unsigned int *sse) {
michael@0	253	unsigned int sse0;
michael@0	254	int sum0;
michael@0	255	vp9_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
michael@0	256	&sum0);
michael@0	257	*sse = sse0;
michael@0	258	return sse0;
michael@0	259	}
michael@0	260
michael@0	261	unsigned int vp9_variance32x32_sse2(const uint8_t *src_ptr,
michael@0	262	int source_stride,
michael@0	263	const uint8_t *ref_ptr,
michael@0	264	int recon_stride,
michael@0	265	unsigned int *sse) {
michael@0	266	unsigned int var;
michael@0	267	int avg;
michael@0	268
michael@0	269	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 32, 32,
michael@0	270	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	271	*sse = var;
michael@0	272	return (var - (((int64_t)avg * avg) >> 10));
michael@0	273	}
michael@0	274
michael@0	275	unsigned int vp9_variance32x16_sse2(const uint8_t *src_ptr,
michael@0	276	int source_stride,
michael@0	277	const uint8_t *ref_ptr,
michael@0	278	int recon_stride,
michael@0	279	unsigned int *sse) {
michael@0	280	unsigned int var;
michael@0	281	int avg;
michael@0	282
michael@0	283	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 32, 16,
michael@0	284	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	285	*sse = var;
michael@0	286	return (var - (((int64_t)avg * avg) >> 9));
michael@0	287	}
michael@0	288
michael@0	289	unsigned int vp9_variance16x32_sse2(const uint8_t *src_ptr,
michael@0	290	int source_stride,
michael@0	291	const uint8_t *ref_ptr,
michael@0	292	int recon_stride,
michael@0	293	unsigned int *sse) {
michael@0	294	unsigned int var;
michael@0	295	int avg;
michael@0	296
michael@0	297	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 16, 32,
michael@0	298	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	299	*sse = var;
michael@0	300	return (var - (((int64_t)avg * avg) >> 9));
michael@0	301	}
michael@0	302
michael@0	303	unsigned int vp9_variance64x64_sse2(const uint8_t *src_ptr,
michael@0	304	int source_stride,
michael@0	305	const uint8_t *ref_ptr,
michael@0	306	int recon_stride,
michael@0	307	unsigned int *sse) {
michael@0	308	unsigned int var;
michael@0	309	int avg;
michael@0	310
michael@0	311	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 64, 64,
michael@0	312	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	313	*sse = var;
michael@0	314	return (var - (((int64_t)avg * avg) >> 12));
michael@0	315	}
michael@0	316
michael@0	317	unsigned int vp9_variance64x32_sse2(const uint8_t *src_ptr,
michael@0	318	int source_stride,
michael@0	319	const uint8_t *ref_ptr,
michael@0	320	int recon_stride,
michael@0	321	unsigned int *sse) {
michael@0	322	unsigned int var;
michael@0	323	int avg;
michael@0	324
michael@0	325	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 64, 32,
michael@0	326	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	327	*sse = var;
michael@0	328	return (var - (((int64_t)avg * avg) >> 11));
michael@0	329	}
michael@0	330
michael@0	331	unsigned int vp9_variance32x64_sse2(const uint8_t *src_ptr,
michael@0	332	int source_stride,
michael@0	333	const uint8_t *ref_ptr,
michael@0	334	int recon_stride,
michael@0	335	unsigned int *sse) {
michael@0	336	unsigned int var;
michael@0	337	int avg;
michael@0	338
michael@0	339	variance_sse2(src_ptr, source_stride, ref_ptr, recon_stride, 32, 64,
michael@0	340	&var, &avg, vp9_get16x16var_sse2, 16);
michael@0	341	*sse = var;
michael@0	342	return (var - (((int64_t)avg * avg) >> 11));
michael@0	343	}
michael@0	344
michael@0	345	#define DECL(w, opt) \
michael@0	346	int vp9_sub_pixel_variance##w##xh_##opt(const uint8_t *src, \
michael@0	347	ptrdiff_t src_stride, \
michael@0	348	int x_offset, int y_offset, \
michael@0	349	const uint8_t *dst, \
michael@0	350	ptrdiff_t dst_stride, \
michael@0	351	int height, unsigned int *sse)
michael@0	352	#define DECLS(opt1, opt2) \
michael@0	353	DECL(4, opt2); \
michael@0	354	DECL(8, opt1); \
michael@0	355	DECL(16, opt1)
michael@0	356
michael@0	357	DECLS(sse2, sse);
michael@0	358	DECLS(ssse3, ssse3);
michael@0	359	#undef DECLS
michael@0	360	#undef DECL
michael@0	361
michael@0	362	#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
michael@0	363	unsigned int vp9_sub_pixel_variance##w##x##h##_##opt(const uint8_t *src, \
michael@0	364	int src_stride, \
michael@0	365	int x_offset, \
michael@0	366	int y_offset, \
michael@0	367	const uint8_t *dst, \
michael@0	368	int dst_stride, \
michael@0	369	unsigned int *sse_ptr) { \
michael@0	370	unsigned int sse; \
michael@0	371	int se = vp9_sub_pixel_variance##wf##xh_##opt(src, src_stride, x_offset, \
michael@0	372	y_offset, dst, dst_stride, \
michael@0	373	h, &sse); \
michael@0	374	if (w > wf) { \
michael@0	375	unsigned int sse2; \
michael@0	376	int se2 = vp9_sub_pixel_variance##wf##xh_##opt(src + 16, src_stride, \
michael@0	377	x_offset, y_offset, \
michael@0	378	dst + 16, dst_stride, \
michael@0	379	h, &sse2); \
michael@0	380	se += se2; \
michael@0	381	sse += sse2; \
michael@0	382	if (w > wf * 2) { \
michael@0	383	se2 = vp9_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
michael@0	384	x_offset, y_offset, \
michael@0	385	dst + 32, dst_stride, \
michael@0	386	h, &sse2); \
michael@0	387	se += se2; \
michael@0	388	sse += sse2; \
michael@0	389	se2 = vp9_sub_pixel_variance##wf##xh_##opt(src + 48, src_stride, \
michael@0	390	x_offset, y_offset, \
michael@0	391	dst + 48, dst_stride, \
michael@0	392	h, &sse2); \
michael@0	393	se += se2; \
michael@0	394	sse += sse2; \
michael@0	395	} \
michael@0	396	} \
michael@0	397	*sse_ptr = sse; \
michael@0	398	return sse - ((cast se * se) >> (wlog2 + hlog2)); \
michael@0	399	}
michael@0	400
michael@0	401	#define FNS(opt1, opt2) \
michael@0	402	FN(64, 64, 16, 6, 6, opt1, (int64_t)); \
michael@0	403	FN(64, 32, 16, 6, 5, opt1, (int64_t)); \
michael@0	404	FN(32, 64, 16, 5, 6, opt1, (int64_t)); \
michael@0	405	FN(32, 32, 16, 5, 5, opt1, (int64_t)); \
michael@0	406	FN(32, 16, 16, 5, 4, opt1, (int64_t)); \
michael@0	407	FN(16, 32, 16, 4, 5, opt1, (int64_t)); \
michael@0	408	FN(16, 16, 16, 4, 4, opt1, (unsigned int)); \
michael@0	409	FN(16, 8, 16, 4, 3, opt1, (unsigned int)); \
michael@0	410	FN(8, 16, 8, 3, 4, opt1, (unsigned int)); \
michael@0	411	FN(8, 8, 8, 3, 3, opt1, (unsigned int)); \
michael@0	412	FN(8, 4, 8, 3, 2, opt1, (unsigned int)); \
michael@0	413	FN(4, 8, 4, 2, 3, opt2, (unsigned int)); \
michael@0	414	FN(4, 4, 4, 2, 2, opt2, (unsigned int))
michael@0	415
michael@0	416	FNS(sse2, sse);
michael@0	417	FNS(ssse3, ssse3);
michael@0	418
michael@0	419	#undef FNS
michael@0	420	#undef FN
michael@0	421
michael@0	422	#define DECL(w, opt) \
michael@0	423	int vp9_sub_pixel_avg_variance##w##xh_##opt(const uint8_t *src, \
michael@0	424	ptrdiff_t src_stride, \
michael@0	425	int x_offset, int y_offset, \
michael@0	426	const uint8_t *dst, \
michael@0	427	ptrdiff_t dst_stride, \
michael@0	428	const uint8_t *sec, \
michael@0	429	ptrdiff_t sec_stride, \
michael@0	430	int height, unsigned int *sse)
michael@0	431	#define DECLS(opt1, opt2) \
michael@0	432	DECL(4, opt2); \
michael@0	433	DECL(8, opt1); \
michael@0	434	DECL(16, opt1)
michael@0	435
michael@0	436	DECLS(sse2, sse);
michael@0	437	DECLS(ssse3, ssse3);
michael@0	438	#undef DECL
michael@0	439	#undef DECLS
michael@0	440
michael@0	441	#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
michael@0	442	unsigned int vp9_sub_pixel_avg_variance##w##x##h##_##opt(const uint8_t *src, \
michael@0	443	int src_stride, \
michael@0	444	int x_offset, \
michael@0	445	int y_offset, \
michael@0	446	const uint8_t *dst, \
michael@0	447	int dst_stride, \
michael@0	448	unsigned int *sseptr, \
michael@0	449	const uint8_t *sec) { \
michael@0	450	unsigned int sse; \
michael@0	451	int se = vp9_sub_pixel_avg_variance##wf##xh_##opt(src, src_stride, x_offset, \
michael@0	452	y_offset, dst, dst_stride, \
michael@0	453	sec, w, h, &sse); \
michael@0	454	if (w > wf) { \
michael@0	455	unsigned int sse2; \
michael@0	456	int se2 = vp9_sub_pixel_avg_variance##wf##xh_##opt(src + 16, src_stride, \
michael@0	457	x_offset, y_offset, \
michael@0	458	dst + 16, dst_stride, \
michael@0	459	sec + 16, w, h, &sse2); \
michael@0	460	se += se2; \
michael@0	461	sse += sse2; \
michael@0	462	if (w > wf * 2) { \
michael@0	463	se2 = vp9_sub_pixel_avg_variance##wf##xh_##opt(src + 32, src_stride, \
michael@0	464	x_offset, y_offset, \
michael@0	465	dst + 32, dst_stride, \
michael@0	466	sec + 32, w, h, &sse2); \
michael@0	467	se += se2; \
michael@0	468	sse += sse2; \
michael@0	469	se2 = vp9_sub_pixel_avg_variance##wf##xh_##opt(src + 48, src_stride, \
michael@0	470	x_offset, y_offset, \
michael@0	471	dst + 48, dst_stride, \
michael@0	472	sec + 48, w, h, &sse2); \
michael@0	473	se += se2; \
michael@0	474	sse += sse2; \
michael@0	475	} \
michael@0	476	} \
michael@0	477	*sseptr = sse; \
michael@0	478	return sse - ((cast se * se) >> (wlog2 + hlog2)); \
michael@0	479	}
michael@0	480
michael@0	481	#define FNS(opt1, opt2) \
michael@0	482	FN(64, 64, 16, 6, 6, opt1, (int64_t)); \
michael@0	483	FN(64, 32, 16, 6, 5, opt1, (int64_t)); \
michael@0	484	FN(32, 64, 16, 5, 6, opt1, (int64_t)); \
michael@0	485	FN(32, 32, 16, 5, 5, opt1, (int64_t)); \
michael@0	486	FN(32, 16, 16, 5, 4, opt1, (int64_t)); \
michael@0	487	FN(16, 32, 16, 4, 5, opt1, (int64_t)); \
michael@0	488	FN(16, 16, 16, 4, 4, opt1, (unsigned int)); \
michael@0	489	FN(16, 8, 16, 4, 3, opt1, (unsigned int)); \
michael@0	490	FN(8, 16, 8, 3, 4, opt1, (unsigned int)); \
michael@0	491	FN(8, 8, 8, 3, 3, opt1, (unsigned int)); \
michael@0	492	FN(8, 4, 8, 3, 2, opt1, (unsigned int)); \
michael@0	493	FN(4, 8, 4, 2, 3, opt2, (unsigned int)); \
michael@0	494	FN(4, 4, 4, 2, 2, opt2, (unsigned int))
michael@0	495
michael@0	496	FNS(sse2, sse);
michael@0	497	FNS(ssse3, ssse3);
michael@0	498
michael@0	499	#undef FNS
michael@0	500	#undef FN
michael@0	501
michael@0	502	unsigned int vp9_variance_halfpixvar16x16_h_sse2(
michael@0	503	const unsigned char *src_ptr,
michael@0	504	int src_pixels_per_line,
michael@0	505	const unsigned char *dst_ptr,
michael@0	506	int dst_pixels_per_line,
michael@0	507	unsigned int *sse) {
michael@0	508	int xsum0;
michael@0	509	unsigned int xxsum0;
michael@0	510
michael@0	511	vp9_half_horiz_variance16x_h_sse2(
michael@0	512	src_ptr, src_pixels_per_line,
michael@0	513	dst_ptr, dst_pixels_per_line, 16,
michael@0	514	&xsum0, &xxsum0);
michael@0	515
michael@0	516	*sse = xxsum0;
michael@0	517	return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
michael@0	518	}
michael@0	519
michael@0	520
michael@0	521	unsigned int vp9_variance_halfpixvar16x16_v_sse2(
michael@0	522	const unsigned char *src_ptr,
michael@0	523	int src_pixels_per_line,
michael@0	524	const unsigned char *dst_ptr,
michael@0	525	int dst_pixels_per_line,
michael@0	526	unsigned int *sse) {
michael@0	527	int xsum0;
michael@0	528	unsigned int xxsum0;
michael@0	529	vp9_half_vert_variance16x_h_sse2(
michael@0	530	src_ptr, src_pixels_per_line,
michael@0	531	dst_ptr, dst_pixels_per_line, 16,
michael@0	532	&xsum0, &xxsum0);
michael@0	533
michael@0	534	*sse = xxsum0;
michael@0	535	return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
michael@0	536	}
michael@0	537
michael@0	538
michael@0	539	unsigned int vp9_variance_halfpixvar16x16_hv_sse2(
michael@0	540	const unsigned char *src_ptr,
michael@0	541	int src_pixels_per_line,
michael@0	542	const unsigned char *dst_ptr,
michael@0	543	int dst_pixels_per_line,
michael@0	544	unsigned int *sse) {
michael@0	545	int xsum0;
michael@0	546	unsigned int xxsum0;
michael@0	547
michael@0	548	vp9_half_horiz_vert_variance16x_h_sse2(
michael@0	549	src_ptr, src_pixels_per_line,
michael@0	550	dst_ptr, dst_pixels_per_line, 16,
michael@0	551	&xsum0, &xxsum0);
michael@0	552
michael@0	553	*sse = xxsum0;
michael@0	554	return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
michael@0	555	}