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

media/libvpx/vp9/common/x86/vp9_asm_stubs.c@6474c204b198 (annotated)

media/libvpx/vp9/common/x86/vp9_asm_stubs.c

Wed, 31 Dec 2014 06:09:35 +0100

author

Michael Schloh von Bennewitz <michael@schloh.com>

date

Wed, 31 Dec 2014 06:09:35 +0100

changeset 0

6474c204b198

permissions

-rw-r--r--

Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.

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 <assert.h>
michael@0	12
michael@0	13	#include "./vpx_config.h"
michael@0	14	#include "./vp9_rtcd.h"
michael@0	15	#include "vpx_ports/mem.h"
michael@0	16	///////////////////////////////////////////////////////////////////////////
michael@0	17	// the mmx function that does the bilinear filtering and var calculation //
michael@0	18	// int one pass //
michael@0	19	///////////////////////////////////////////////////////////////////////////
michael@0	20	DECLARE_ALIGNED(16, const short, vp9_bilinear_filters_mmx[16][8]) = {
michael@0	21	{ 128, 128, 128, 128, 0, 0, 0, 0 },
michael@0	22	{ 120, 120, 120, 120, 8, 8, 8, 8 },
michael@0	23	{ 112, 112, 112, 112, 16, 16, 16, 16 },
michael@0	24	{ 104, 104, 104, 104, 24, 24, 24, 24 },
michael@0	25	{ 96, 96, 96, 96, 32, 32, 32, 32 },
michael@0	26	{ 88, 88, 88, 88, 40, 40, 40, 40 },
michael@0	27	{ 80, 80, 80, 80, 48, 48, 48, 48 },
michael@0	28	{ 72, 72, 72, 72, 56, 56, 56, 56 },
michael@0	29	{ 64, 64, 64, 64, 64, 64, 64, 64 },
michael@0	30	{ 56, 56, 56, 56, 72, 72, 72, 72 },
michael@0	31	{ 48, 48, 48, 48, 80, 80, 80, 80 },
michael@0	32	{ 40, 40, 40, 40, 88, 88, 88, 88 },
michael@0	33	{ 32, 32, 32, 32, 96, 96, 96, 96 },
michael@0	34	{ 24, 24, 24, 24, 104, 104, 104, 104 },
michael@0	35	{ 16, 16, 16, 16, 112, 112, 112, 112 },
michael@0	36	{ 8, 8, 8, 8, 120, 120, 120, 120 }
michael@0	37	};
michael@0	38
michael@0	39	typedef void filter8_1dfunction (
michael@0	40	const unsigned char *src_ptr,
michael@0	41	const unsigned int src_pitch,
michael@0	42	unsigned char *output_ptr,
michael@0	43	unsigned int out_pitch,
michael@0	44	unsigned int output_height,
michael@0	45	const short *filter
michael@0	46	);
michael@0	47
michael@0	48	#if HAVE_SSSE3
michael@0	49	filter8_1dfunction vp9_filter_block1d16_v8_ssse3;
michael@0	50	filter8_1dfunction vp9_filter_block1d16_h8_ssse3;
michael@0	51	filter8_1dfunction vp9_filter_block1d8_v8_ssse3;
michael@0	52	filter8_1dfunction vp9_filter_block1d8_h8_ssse3;
michael@0	53	filter8_1dfunction vp9_filter_block1d4_v8_ssse3;
michael@0	54	filter8_1dfunction vp9_filter_block1d4_h8_ssse3;
michael@0	55	filter8_1dfunction vp9_filter_block1d16_v8_avg_ssse3;
michael@0	56	filter8_1dfunction vp9_filter_block1d16_h8_avg_ssse3;
michael@0	57	filter8_1dfunction vp9_filter_block1d8_v8_avg_ssse3;
michael@0	58	filter8_1dfunction vp9_filter_block1d8_h8_avg_ssse3;
michael@0	59	filter8_1dfunction vp9_filter_block1d4_v8_avg_ssse3;
michael@0	60	filter8_1dfunction vp9_filter_block1d4_h8_avg_ssse3;
michael@0	61
michael@0	62	void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	63	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	64	const int16_t *filter_x, int x_step_q4,
michael@0	65	const int16_t *filter_y, int y_step_q4,
michael@0	66	int w, int h) {
michael@0	67	/* Ensure the filter can be compressed to int16_t. */
michael@0	68	if (x_step_q4 == 16 && filter_x[3] != 128) {
michael@0	69	while (w >= 16) {
michael@0	70	vp9_filter_block1d16_h8_ssse3(src, src_stride,
michael@0	71	dst, dst_stride,
michael@0	72	h, filter_x);
michael@0	73	src += 16;
michael@0	74	dst += 16;
michael@0	75	w -= 16;
michael@0	76	}
michael@0	77	while (w >= 8) {
michael@0	78	vp9_filter_block1d8_h8_ssse3(src, src_stride,
michael@0	79	dst, dst_stride,
michael@0	80	h, filter_x);
michael@0	81	src += 8;
michael@0	82	dst += 8;
michael@0	83	w -= 8;
michael@0	84	}
michael@0	85	while (w >= 4) {
michael@0	86	vp9_filter_block1d4_h8_ssse3(src, src_stride,
michael@0	87	dst, dst_stride,
michael@0	88	h, filter_x);
michael@0	89	src += 4;
michael@0	90	dst += 4;
michael@0	91	w -= 4;
michael@0	92	}
michael@0	93	}
michael@0	94	if (w) {
michael@0	95	vp9_convolve8_horiz_c(src, src_stride, dst, dst_stride,
michael@0	96	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	97	w, h);
michael@0	98	}
michael@0	99	}
michael@0	100
michael@0	101	void vp9_convolve8_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	102	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	103	const int16_t *filter_x, int x_step_q4,
michael@0	104	const int16_t *filter_y, int y_step_q4,
michael@0	105	int w, int h) {
michael@0	106	if (y_step_q4 == 16 && filter_y[3] != 128) {
michael@0	107	while (w >= 16) {
michael@0	108	vp9_filter_block1d16_v8_ssse3(src - src_stride * 3, src_stride,
michael@0	109	dst, dst_stride,
michael@0	110	h, filter_y);
michael@0	111	src += 16;
michael@0	112	dst += 16;
michael@0	113	w -= 16;
michael@0	114	}
michael@0	115	while (w >= 8) {
michael@0	116	vp9_filter_block1d8_v8_ssse3(src - src_stride * 3, src_stride,
michael@0	117	dst, dst_stride,
michael@0	118	h, filter_y);
michael@0	119	src += 8;
michael@0	120	dst += 8;
michael@0	121	w -= 8;
michael@0	122	}
michael@0	123	while (w >= 4) {
michael@0	124	vp9_filter_block1d4_v8_ssse3(src - src_stride * 3, src_stride,
michael@0	125	dst, dst_stride,
michael@0	126	h, filter_y);
michael@0	127	src += 4;
michael@0	128	dst += 4;
michael@0	129	w -= 4;
michael@0	130	}
michael@0	131	}
michael@0	132	if (w) {
michael@0	133	vp9_convolve8_vert_c(src, src_stride, dst, dst_stride,
michael@0	134	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	135	w, h);
michael@0	136	}
michael@0	137	}
michael@0	138
michael@0	139	void vp9_convolve8_avg_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	140	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	141	const int16_t *filter_x, int x_step_q4,
michael@0	142	const int16_t *filter_y, int y_step_q4,
michael@0	143	int w, int h) {
michael@0	144	if (x_step_q4 == 16 && filter_x[3] != 128) {
michael@0	145	while (w >= 16) {
michael@0	146	vp9_filter_block1d16_h8_avg_ssse3(src, src_stride,
michael@0	147	dst, dst_stride,
michael@0	148	h, filter_x);
michael@0	149	src += 16;
michael@0	150	dst += 16;
michael@0	151	w -= 16;
michael@0	152	}
michael@0	153	while (w >= 8) {
michael@0	154	vp9_filter_block1d8_h8_avg_ssse3(src, src_stride,
michael@0	155	dst, dst_stride,
michael@0	156	h, filter_x);
michael@0	157	src += 8;
michael@0	158	dst += 8;
michael@0	159	w -= 8;
michael@0	160	}
michael@0	161	while (w >= 4) {
michael@0	162	vp9_filter_block1d4_h8_avg_ssse3(src, src_stride,
michael@0	163	dst, dst_stride,
michael@0	164	h, filter_x);
michael@0	165	src += 4;
michael@0	166	dst += 4;
michael@0	167	w -= 4;
michael@0	168	}
michael@0	169	}
michael@0	170	if (w) {
michael@0	171	vp9_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
michael@0	172	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	173	w, h);
michael@0	174	}
michael@0	175	}
michael@0	176
michael@0	177	void vp9_convolve8_avg_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	178	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	179	const int16_t *filter_x, int x_step_q4,
michael@0	180	const int16_t *filter_y, int y_step_q4,
michael@0	181	int w, int h) {
michael@0	182	if (y_step_q4 == 16 && filter_y[3] != 128) {
michael@0	183	while (w >= 16) {
michael@0	184	vp9_filter_block1d16_v8_avg_ssse3(src - src_stride * 3, src_stride,
michael@0	185	dst, dst_stride,
michael@0	186	h, filter_y);
michael@0	187	src += 16;
michael@0	188	dst += 16;
michael@0	189	w -= 16;
michael@0	190	}
michael@0	191	while (w >= 8) {
michael@0	192	vp9_filter_block1d8_v8_avg_ssse3(src - src_stride * 3, src_stride,
michael@0	193	dst, dst_stride,
michael@0	194	h, filter_y);
michael@0	195	src += 8;
michael@0	196	dst += 8;
michael@0	197	w -= 8;
michael@0	198	}
michael@0	199	while (w >= 4) {
michael@0	200	vp9_filter_block1d4_v8_avg_ssse3(src - src_stride * 3, src_stride,
michael@0	201	dst, dst_stride,
michael@0	202	h, filter_y);
michael@0	203	src += 4;
michael@0	204	dst += 4;
michael@0	205	w -= 4;
michael@0	206	}
michael@0	207	}
michael@0	208	if (w) {
michael@0	209	vp9_convolve8_avg_vert_c(src, src_stride, dst, dst_stride,
michael@0	210	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	211	w, h);
michael@0	212	}
michael@0	213	}
michael@0	214
michael@0	215	void vp9_convolve8_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	216	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	217	const int16_t *filter_x, int x_step_q4,
michael@0	218	const int16_t *filter_y, int y_step_q4,
michael@0	219	int w, int h) {
michael@0	220	DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71);
michael@0	221
michael@0	222	assert(w <= 64);
michael@0	223	assert(h <= 64);
michael@0	224	if (x_step_q4 == 16 && y_step_q4 == 16) {
michael@0	225	vp9_convolve8_horiz_ssse3(src - 3 * src_stride, src_stride, fdata2, 64,
michael@0	226	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	227	w, h + 7);
michael@0	228	vp9_convolve8_vert_ssse3(fdata2 + 3 * 64, 64, dst, dst_stride,
michael@0	229	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	230	} else {
michael@0	231	vp9_convolve8_c(src, src_stride, dst, dst_stride,
michael@0	232	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	233	}
michael@0	234	}
michael@0	235
michael@0	236	void vp9_convolve8_avg_ssse3(const uint8_t *src, ptrdiff_t src_stride,
michael@0	237	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	238	const int16_t *filter_x, int x_step_q4,
michael@0	239	const int16_t *filter_y, int y_step_q4,
michael@0	240	int w, int h) {
michael@0	241	DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71);
michael@0	242
michael@0	243	assert(w <= 64);
michael@0	244	assert(h <= 64);
michael@0	245	if (x_step_q4 == 16 && y_step_q4 == 16) {
michael@0	246	vp9_convolve8_horiz_ssse3(src - 3 * src_stride, src_stride, fdata2, 64,
michael@0	247	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	248	w, h + 7);
michael@0	249	vp9_convolve8_avg_vert_ssse3(fdata2 + 3 * 64, 64, dst, dst_stride,
michael@0	250	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	251	w, h);
michael@0	252	} else {
michael@0	253	vp9_convolve8_avg_c(src, src_stride, dst, dst_stride,
michael@0	254	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	255	}
michael@0	256	}
michael@0	257	#endif
michael@0	258
michael@0	259	#if HAVE_SSE2
michael@0	260	filter8_1dfunction vp9_filter_block1d16_v8_sse2;
michael@0	261	filter8_1dfunction vp9_filter_block1d16_h8_sse2;
michael@0	262	filter8_1dfunction vp9_filter_block1d8_v8_sse2;
michael@0	263	filter8_1dfunction vp9_filter_block1d8_h8_sse2;
michael@0	264	filter8_1dfunction vp9_filter_block1d4_v8_sse2;
michael@0	265	filter8_1dfunction vp9_filter_block1d4_h8_sse2;
michael@0	266	filter8_1dfunction vp9_filter_block1d16_v8_avg_sse2;
michael@0	267	filter8_1dfunction vp9_filter_block1d16_h8_avg_sse2;
michael@0	268	filter8_1dfunction vp9_filter_block1d8_v8_avg_sse2;
michael@0	269	filter8_1dfunction vp9_filter_block1d8_h8_avg_sse2;
michael@0	270	filter8_1dfunction vp9_filter_block1d4_v8_avg_sse2;
michael@0	271	filter8_1dfunction vp9_filter_block1d4_h8_avg_sse2;
michael@0	272
michael@0	273	void vp9_convolve8_horiz_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	274	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	275	const int16_t *filter_x, int x_step_q4,
michael@0	276	const int16_t *filter_y, int y_step_q4,
michael@0	277	int w, int h) {
michael@0	278	/* Ensure the filter can be compressed to int16_t. */
michael@0	279	if (x_step_q4 == 16 && filter_x[3] != 128) {
michael@0	280	while (w >= 16) {
michael@0	281	vp9_filter_block1d16_h8_sse2(src, src_stride,
michael@0	282	dst, dst_stride,
michael@0	283	h, filter_x);
michael@0	284	src += 16;
michael@0	285	dst += 16;
michael@0	286	w -= 16;
michael@0	287	}
michael@0	288	while (w >= 8) {
michael@0	289	vp9_filter_block1d8_h8_sse2(src, src_stride,
michael@0	290	dst, dst_stride,
michael@0	291	h, filter_x);
michael@0	292	src += 8;
michael@0	293	dst += 8;
michael@0	294	w -= 8;
michael@0	295	}
michael@0	296	while (w >= 4) {
michael@0	297	vp9_filter_block1d4_h8_sse2(src, src_stride,
michael@0	298	dst, dst_stride,
michael@0	299	h, filter_x);
michael@0	300	src += 4;
michael@0	301	dst += 4;
michael@0	302	w -= 4;
michael@0	303	}
michael@0	304	}
michael@0	305	if (w) {
michael@0	306	vp9_convolve8_horiz_c(src, src_stride, dst, dst_stride,
michael@0	307	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	308	w, h);
michael@0	309	}
michael@0	310	}
michael@0	311
michael@0	312	void vp9_convolve8_vert_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	313	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	314	const int16_t *filter_x, int x_step_q4,
michael@0	315	const int16_t *filter_y, int y_step_q4,
michael@0	316	int w, int h) {
michael@0	317	if (y_step_q4 == 16 && filter_y[3] != 128) {
michael@0	318	while (w >= 16) {
michael@0	319	vp9_filter_block1d16_v8_sse2(src - src_stride * 3, src_stride,
michael@0	320	dst, dst_stride,
michael@0	321	h, filter_y);
michael@0	322	src += 16;
michael@0	323	dst += 16;
michael@0	324	w -= 16;
michael@0	325	}
michael@0	326	while (w >= 8) {
michael@0	327	vp9_filter_block1d8_v8_sse2(src - src_stride * 3, src_stride,
michael@0	328	dst, dst_stride,
michael@0	329	h, filter_y);
michael@0	330	src += 8;
michael@0	331	dst += 8;
michael@0	332	w -= 8;
michael@0	333	}
michael@0	334	while (w >= 4) {
michael@0	335	vp9_filter_block1d4_v8_sse2(src - src_stride * 3, src_stride,
michael@0	336	dst, dst_stride,
michael@0	337	h, filter_y);
michael@0	338	src += 4;
michael@0	339	dst += 4;
michael@0	340	w -= 4;
michael@0	341	}
michael@0	342	}
michael@0	343	if (w) {
michael@0	344	vp9_convolve8_vert_c(src, src_stride, dst, dst_stride,
michael@0	345	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	346	w, h);
michael@0	347	}
michael@0	348	}
michael@0	349
michael@0	350	void vp9_convolve8_avg_horiz_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	351	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	352	const int16_t *filter_x, int x_step_q4,
michael@0	353	const int16_t *filter_y, int y_step_q4,
michael@0	354	int w, int h) {
michael@0	355	if (x_step_q4 == 16 && filter_x[3] != 128) {
michael@0	356	while (w >= 16) {
michael@0	357	vp9_filter_block1d16_h8_avg_sse2(src, src_stride,
michael@0	358	dst, dst_stride,
michael@0	359	h, filter_x);
michael@0	360	src += 16;
michael@0	361	dst += 16;
michael@0	362	w -= 16;
michael@0	363	}
michael@0	364	while (w >= 8) {
michael@0	365	vp9_filter_block1d8_h8_avg_sse2(src, src_stride,
michael@0	366	dst, dst_stride,
michael@0	367	h, filter_x);
michael@0	368	src += 8;
michael@0	369	dst += 8;
michael@0	370	w -= 8;
michael@0	371	}
michael@0	372	while (w >= 4) {
michael@0	373	vp9_filter_block1d4_h8_avg_sse2(src, src_stride,
michael@0	374	dst, dst_stride,
michael@0	375	h, filter_x);
michael@0	376	src += 4;
michael@0	377	dst += 4;
michael@0	378	w -= 4;
michael@0	379	}
michael@0	380	}
michael@0	381	if (w) {
michael@0	382	vp9_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
michael@0	383	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	384	w, h);
michael@0	385	}
michael@0	386	}
michael@0	387
michael@0	388	void vp9_convolve8_avg_vert_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	389	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	390	const int16_t *filter_x, int x_step_q4,
michael@0	391	const int16_t *filter_y, int y_step_q4,
michael@0	392	int w, int h) {
michael@0	393	if (y_step_q4 == 16 && filter_y[3] != 128) {
michael@0	394	while (w >= 16) {
michael@0	395	vp9_filter_block1d16_v8_avg_sse2(src - src_stride * 3, src_stride,
michael@0	396	dst, dst_stride,
michael@0	397	h, filter_y);
michael@0	398	src += 16;
michael@0	399	dst += 16;
michael@0	400	w -= 16;
michael@0	401	}
michael@0	402	while (w >= 8) {
michael@0	403	vp9_filter_block1d8_v8_avg_sse2(src - src_stride * 3, src_stride,
michael@0	404	dst, dst_stride,
michael@0	405	h, filter_y);
michael@0	406	src += 8;
michael@0	407	dst += 8;
michael@0	408	w -= 8;
michael@0	409	}
michael@0	410	while (w >= 4) {
michael@0	411	vp9_filter_block1d4_v8_avg_sse2(src - src_stride * 3, src_stride,
michael@0	412	dst, dst_stride,
michael@0	413	h, filter_y);
michael@0	414	src += 4;
michael@0	415	dst += 4;
michael@0	416	w -= 4;
michael@0	417	}
michael@0	418	}
michael@0	419	if (w) {
michael@0	420	vp9_convolve8_avg_vert_c(src, src_stride, dst, dst_stride,
michael@0	421	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	422	w, h);
michael@0	423	}
michael@0	424	}
michael@0	425
michael@0	426	void vp9_convolve8_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	427	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	428	const int16_t *filter_x, int x_step_q4,
michael@0	429	const int16_t *filter_y, int y_step_q4,
michael@0	430	int w, int h) {
michael@0	431	DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71);
michael@0	432
michael@0	433	assert(w <= 64);
michael@0	434	assert(h <= 64);
michael@0	435	if (x_step_q4 == 16 && y_step_q4 == 16) {
michael@0	436	vp9_convolve8_horiz_sse2(src - 3 * src_stride, src_stride, fdata2, 64,
michael@0	437	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	438	w, h + 7);
michael@0	439	vp9_convolve8_vert_sse2(fdata2 + 3 * 64, 64, dst, dst_stride,
michael@0	440	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	441	} else {
michael@0	442	vp9_convolve8_c(src, src_stride, dst, dst_stride,
michael@0	443	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	444	}
michael@0	445	}
michael@0	446
michael@0	447	void vp9_convolve8_avg_sse2(const uint8_t *src, ptrdiff_t src_stride,
michael@0	448	uint8_t *dst, ptrdiff_t dst_stride,
michael@0	449	const int16_t *filter_x, int x_step_q4,
michael@0	450	const int16_t *filter_y, int y_step_q4,
michael@0	451	int w, int h) {
michael@0	452	DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71);
michael@0	453
michael@0	454	assert(w <= 64);
michael@0	455	assert(h <= 64);
michael@0	456	if (x_step_q4 == 16 && y_step_q4 == 16) {
michael@0	457	vp9_convolve8_horiz_sse2(src - 3 * src_stride, src_stride, fdata2, 64,
michael@0	458	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	459	w, h + 7);
michael@0	460	vp9_convolve8_avg_vert_sse2(fdata2 + 3 * 64, 64, dst, dst_stride,
michael@0	461	filter_x, x_step_q4, filter_y, y_step_q4,
michael@0	462	w, h);
michael@0	463	} else {
michael@0	464	vp9_convolve8_avg_c(src, src_stride, dst, dst_stride,
michael@0	465	filter_x, x_step_q4, filter_y, y_step_q4, w, h);
michael@0	466	}
michael@0	467	}
michael@0	468	#endif