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

media/libvpx/vp9/encoder/vp9_variance_c.c@b8a032363ba2 (annotated)

media/libvpx/vp9/encoder/vp9_variance_c.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 "./vp9_rtcd.h"
michael@0	12
michael@0	13	#include "vpx_ports/mem.h"
michael@0	14	#include "vpx/vpx_integer.h"
michael@0	15
michael@0	16	#include "vp9/common/vp9_common.h"
michael@0	17	#include "vp9/common/vp9_filter.h"
michael@0	18
michael@0	19	#include "vp9/encoder/vp9_variance.h"
michael@0	20
michael@0	21	void variance(const uint8_t *src_ptr,
michael@0	22	int source_stride,
michael@0	23	const uint8_t *ref_ptr,
michael@0	24	int recon_stride,
michael@0	25	int w,
michael@0	26	int h,
michael@0	27	unsigned int *sse,
michael@0	28	int *sum) {
michael@0	29	int i, j;
michael@0	30	int diff;
michael@0	31
michael@0	32	*sum = 0;
michael@0	33	*sse = 0;
michael@0	34
michael@0	35	for (i = 0; i < h; i++) {
michael@0	36	for (j = 0; j < w; j++) {
michael@0	37	diff = src_ptr[j] - ref_ptr[j];
michael@0	38	*sum += diff;
michael@0	39	*sse += diff * diff;
michael@0	40	}
michael@0	41
michael@0	42	src_ptr += source_stride;
michael@0	43	ref_ptr += recon_stride;
michael@0	44	}
michael@0	45	}
michael@0	46
michael@0	47	/****************************************************************************
michael@0	48	*
michael@0	49	* ROUTINE : filter_block2d_bil_first_pass
michael@0	50	*
michael@0	51	* INPUTS : uint8_t *src_ptr : Pointer to source block.
michael@0	52	* uint32_t src_pixels_per_line : Stride of input block.
michael@0	53	* uint32_t pixel_step : Offset between filter input
michael@0	54	* samples (see notes).
michael@0	55	* uint32_t output_height : Input block height.
michael@0	56	* uint32_t output_width : Input block width.
michael@0	57	* int32_t *vp9_filter : Array of 2 bi-linear filter
michael@0	58	* taps.
michael@0	59	*
michael@0	60	* OUTPUTS : int32_t *output_ptr : Pointer to filtered block.
michael@0	61	*
michael@0	62	* RETURNS : void
michael@0	63	*
michael@0	64	* FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
michael@0	65	* either horizontal or vertical direction to produce the
michael@0	66	* filtered output block. Used to implement first-pass
michael@0	67	* of 2-D separable filter.
michael@0	68	*
michael@0	69	* SPECIAL NOTES : Produces int32_t output to retain precision for next pass.
michael@0	70	* Two filter taps should sum to VP9_FILTER_WEIGHT.
michael@0	71	* pixel_step defines whether the filter is applied
michael@0	72	* horizontally (pixel_step=1) or vertically (pixel_step=
michael@0	73	* stride).
michael@0	74	* It defines the offset required to move from one input
michael@0	75	* to the next.
michael@0	76	*
michael@0	77	****************************************************************************/
michael@0	78	static void var_filter_block2d_bil_first_pass(const uint8_t *src_ptr,
michael@0	79	uint16_t *output_ptr,
michael@0	80	unsigned int src_pixels_per_line,
michael@0	81	int pixel_step,
michael@0	82	unsigned int output_height,
michael@0	83	unsigned int output_width,
michael@0	84	const int16_t *vp9_filter) {
michael@0	85	unsigned int i, j;
michael@0	86
michael@0	87	for (i = 0; i < output_height; i++) {
michael@0	88	for (j = 0; j < output_width; j++) {
michael@0	89	output_ptr[j] = ROUND_POWER_OF_TWO((int)src_ptr[0] * vp9_filter[0] +
michael@0	90	(int)src_ptr[pixel_step] * vp9_filter[1],
michael@0	91	FILTER_BITS);
michael@0	92
michael@0	93	src_ptr++;
michael@0	94	}
michael@0	95
michael@0	96	// Next row...
michael@0	97	src_ptr += src_pixels_per_line - output_width;
michael@0	98	output_ptr += output_width;
michael@0	99	}
michael@0	100	}
michael@0	101
michael@0	102	/****************************************************************************
michael@0	103	*
michael@0	104	* ROUTINE : filter_block2d_bil_second_pass
michael@0	105	*
michael@0	106	* INPUTS : int32_t *src_ptr : Pointer to source block.
michael@0	107	* uint32_t src_pixels_per_line : Stride of input block.
michael@0	108	* uint32_t pixel_step : Offset between filter input
michael@0	109	* samples (see notes).
michael@0	110	* uint32_t output_height : Input block height.
michael@0	111	* uint32_t output_width : Input block width.
michael@0	112	* int32_t *vp9_filter : Array of 2 bi-linear filter
michael@0	113	* taps.
michael@0	114	*
michael@0	115	* OUTPUTS : uint16_t *output_ptr : Pointer to filtered block.
michael@0	116	*
michael@0	117	* RETURNS : void
michael@0	118	*
michael@0	119	* FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
michael@0	120	* either horizontal or vertical direction to produce the
michael@0	121	* filtered output block. Used to implement second-pass
michael@0	122	* of 2-D separable filter.
michael@0	123	*
michael@0	124	* SPECIAL NOTES : Requires 32-bit input as produced by
michael@0	125	* filter_block2d_bil_first_pass.
michael@0	126	* Two filter taps should sum to VP9_FILTER_WEIGHT.
michael@0	127	* pixel_step defines whether the filter is applied
michael@0	128	* horizontally (pixel_step=1) or vertically (pixel_step=
michael@0	129	* stride).
michael@0	130	* It defines the offset required to move from one input
michael@0	131	* to the next.
michael@0	132	*
michael@0	133	****************************************************************************/
michael@0	134	static void var_filter_block2d_bil_second_pass(const uint16_t *src_ptr,
michael@0	135	uint8_t *output_ptr,
michael@0	136	unsigned int src_pixels_per_line,
michael@0	137	unsigned int pixel_step,
michael@0	138	unsigned int output_height,
michael@0	139	unsigned int output_width,
michael@0	140	const int16_t *vp9_filter) {
michael@0	141	unsigned int i, j;
michael@0	142
michael@0	143	for (i = 0; i < output_height; i++) {
michael@0	144	for (j = 0; j < output_width; j++) {
michael@0	145	output_ptr[j] = ROUND_POWER_OF_TWO((int)src_ptr[0] * vp9_filter[0] +
michael@0	146	(int)src_ptr[pixel_step] * vp9_filter[1],
michael@0	147	FILTER_BITS);
michael@0	148	src_ptr++;
michael@0	149	}
michael@0	150
michael@0	151	src_ptr += src_pixels_per_line - output_width;
michael@0	152	output_ptr += output_width;
michael@0	153	}
michael@0	154	}
michael@0	155
michael@0	156	unsigned int vp9_get_mb_ss_c(const int16_t *src_ptr) {
michael@0	157	unsigned int i, sum = 0;
michael@0	158
michael@0	159	for (i = 0; i < 256; i++) {
michael@0	160	sum += (src_ptr[i] * src_ptr[i]);
michael@0	161	}
michael@0	162
michael@0	163	return sum;
michael@0	164	}
michael@0	165
michael@0	166	unsigned int vp9_variance64x32_c(const uint8_t *src_ptr,
michael@0	167	int source_stride,
michael@0	168	const uint8_t *ref_ptr,
michael@0	169	int recon_stride,
michael@0	170	unsigned int *sse) {
michael@0	171	unsigned int var;
michael@0	172	int avg;
michael@0	173
michael@0	174	variance(src_ptr, source_stride, ref_ptr, recon_stride, 64, 32, &var, &avg);
michael@0	175	*sse = var;
michael@0	176	return (var - (((int64_t)avg * avg) >> 11));
michael@0	177	}
michael@0	178
michael@0	179	unsigned int vp9_sub_pixel_variance64x32_c(const uint8_t *src_ptr,
michael@0	180	int src_pixels_per_line,
michael@0	181	int xoffset,
michael@0	182	int yoffset,
michael@0	183	const uint8_t *dst_ptr,
michael@0	184	int dst_pixels_per_line,
michael@0	185	unsigned int *sse) {
michael@0	186	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	187	uint8_t temp2[68 * 64];
michael@0	188	const int16_t *hfilter, *vfilter;
michael@0	189
michael@0	190	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	191	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	192
michael@0	193	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	194	1, 33, 64, hfilter);
michael@0	195	var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 32, 64, vfilter);
michael@0	196
michael@0	197	return vp9_variance64x32(temp2, 64, dst_ptr, dst_pixels_per_line, sse);
michael@0	198	}
michael@0	199
michael@0	200	unsigned int vp9_sub_pixel_avg_variance64x32_c(const uint8_t *src_ptr,
michael@0	201	int src_pixels_per_line,
michael@0	202	int xoffset,
michael@0	203	int yoffset,
michael@0	204	const uint8_t *dst_ptr,
michael@0	205	int dst_pixels_per_line,
michael@0	206	unsigned int *sse,
michael@0	207	const uint8_t *second_pred) {
michael@0	208	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	209	uint8_t temp2[68 * 64];
michael@0	210	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 64 * 64); // compound pred buffer
michael@0	211	const int16_t *hfilter, *vfilter;
michael@0	212
michael@0	213	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	214	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	215
michael@0	216	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	217	1, 33, 64, hfilter);
michael@0	218	var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 32, 64, vfilter);
michael@0	219	comp_avg_pred(temp3, second_pred, 64, 32, temp2, 64);
michael@0	220	return vp9_variance64x32(temp3, 64, dst_ptr, dst_pixels_per_line, sse);
michael@0	221	}
michael@0	222
michael@0	223	unsigned int vp9_variance32x64_c(const uint8_t *src_ptr,
michael@0	224	int source_stride,
michael@0	225	const uint8_t *ref_ptr,
michael@0	226	int recon_stride,
michael@0	227	unsigned int *sse) {
michael@0	228	unsigned int var;
michael@0	229	int avg;
michael@0	230
michael@0	231	variance(src_ptr, source_stride, ref_ptr, recon_stride, 32, 64, &var, &avg);
michael@0	232	*sse = var;
michael@0	233	return (var - (((int64_t)avg * avg) >> 11));
michael@0	234	}
michael@0	235
michael@0	236	unsigned int vp9_sub_pixel_variance32x64_c(const uint8_t *src_ptr,
michael@0	237	int src_pixels_per_line,
michael@0	238	int xoffset,
michael@0	239	int yoffset,
michael@0	240	const uint8_t *dst_ptr,
michael@0	241	int dst_pixels_per_line,
michael@0	242	unsigned int *sse) {
michael@0	243	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	244	uint8_t temp2[68 * 64];
michael@0	245	const int16_t *hfilter, *vfilter;
michael@0	246
michael@0	247	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	248	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	249
michael@0	250	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	251	1, 65, 32, hfilter);
michael@0	252	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 64, 32, vfilter);
michael@0	253
michael@0	254	return vp9_variance32x64(temp2, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	255	}
michael@0	256
michael@0	257	unsigned int vp9_sub_pixel_avg_variance32x64_c(const uint8_t *src_ptr,
michael@0	258	int src_pixels_per_line,
michael@0	259	int xoffset,
michael@0	260	int yoffset,
michael@0	261	const uint8_t *dst_ptr,
michael@0	262	int dst_pixels_per_line,
michael@0	263	unsigned int *sse,
michael@0	264	const uint8_t *second_pred) {
michael@0	265	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	266	uint8_t temp2[68 * 64];
michael@0	267	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 32 * 64); // compound pred buffer
michael@0	268	const int16_t *hfilter, *vfilter;
michael@0	269
michael@0	270	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	271	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	272
michael@0	273	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	274	1, 65, 32, hfilter);
michael@0	275	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 64, 32, vfilter);
michael@0	276	comp_avg_pred(temp3, second_pred, 32, 64, temp2, 32);
michael@0	277	return vp9_variance32x64(temp3, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	278	}
michael@0	279
michael@0	280	unsigned int vp9_variance32x16_c(const uint8_t *src_ptr,
michael@0	281	int source_stride,
michael@0	282	const uint8_t *ref_ptr,
michael@0	283	int recon_stride,
michael@0	284	unsigned int *sse) {
michael@0	285	unsigned int var;
michael@0	286	int avg;
michael@0	287
michael@0	288	variance(src_ptr, source_stride, ref_ptr, recon_stride, 32, 16, &var, &avg);
michael@0	289	*sse = var;
michael@0	290	return (var - (((int64_t)avg * avg) >> 9));
michael@0	291	}
michael@0	292
michael@0	293	unsigned int vp9_sub_pixel_variance32x16_c(const uint8_t *src_ptr,
michael@0	294	int src_pixels_per_line,
michael@0	295	int xoffset,
michael@0	296	int yoffset,
michael@0	297	const uint8_t *dst_ptr,
michael@0	298	int dst_pixels_per_line,
michael@0	299	unsigned int *sse) {
michael@0	300	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	301	uint8_t temp2[36 * 32];
michael@0	302	const int16_t *hfilter, *vfilter;
michael@0	303
michael@0	304	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	305	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	306
michael@0	307	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	308	1, 17, 32, hfilter);
michael@0	309	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 16, 32, vfilter);
michael@0	310
michael@0	311	return vp9_variance32x16(temp2, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	312	}
michael@0	313
michael@0	314	unsigned int vp9_sub_pixel_avg_variance32x16_c(const uint8_t *src_ptr,
michael@0	315	int src_pixels_per_line,
michael@0	316	int xoffset,
michael@0	317	int yoffset,
michael@0	318	const uint8_t *dst_ptr,
michael@0	319	int dst_pixels_per_line,
michael@0	320	unsigned int *sse,
michael@0	321	const uint8_t *second_pred) {
michael@0	322	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	323	uint8_t temp2[36 * 32];
michael@0	324	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 32 * 16); // compound pred buffer
michael@0	325	const int16_t *hfilter, *vfilter;
michael@0	326
michael@0	327	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	328	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	329
michael@0	330	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	331	1, 17, 32, hfilter);
michael@0	332	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 16, 32, vfilter);
michael@0	333	comp_avg_pred(temp3, second_pred, 32, 16, temp2, 32);
michael@0	334	return vp9_variance32x16(temp3, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	335	}
michael@0	336
michael@0	337	unsigned int vp9_variance16x32_c(const uint8_t *src_ptr,
michael@0	338	int source_stride,
michael@0	339	const uint8_t *ref_ptr,
michael@0	340	int recon_stride,
michael@0	341	unsigned int *sse) {
michael@0	342	unsigned int var;
michael@0	343	int avg;
michael@0	344
michael@0	345	variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 32, &var, &avg);
michael@0	346	*sse = var;
michael@0	347	return (var - (((int64_t)avg * avg) >> 9));
michael@0	348	}
michael@0	349
michael@0	350	unsigned int vp9_sub_pixel_variance16x32_c(const uint8_t *src_ptr,
michael@0	351	int src_pixels_per_line,
michael@0	352	int xoffset,
michael@0	353	int yoffset,
michael@0	354	const uint8_t *dst_ptr,
michael@0	355	int dst_pixels_per_line,
michael@0	356	unsigned int *sse) {
michael@0	357	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	358	uint8_t temp2[36 * 32];
michael@0	359	const int16_t *hfilter, *vfilter;
michael@0	360
michael@0	361	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	362	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	363
michael@0	364	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	365	1, 33, 16, hfilter);
michael@0	366	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 32, 16, vfilter);
michael@0	367
michael@0	368	return vp9_variance16x32(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	369	}
michael@0	370
michael@0	371	unsigned int vp9_sub_pixel_avg_variance16x32_c(const uint8_t *src_ptr,
michael@0	372	int src_pixels_per_line,
michael@0	373	int xoffset,
michael@0	374	int yoffset,
michael@0	375	const uint8_t *dst_ptr,
michael@0	376	int dst_pixels_per_line,
michael@0	377	unsigned int *sse,
michael@0	378	const uint8_t *second_pred) {
michael@0	379	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	380	uint8_t temp2[36 * 32];
michael@0	381	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 16 * 32); // compound pred buffer
michael@0	382	const int16_t *hfilter, *vfilter;
michael@0	383
michael@0	384	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	385	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	386
michael@0	387	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	388	1, 33, 16, hfilter);
michael@0	389	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 32, 16, vfilter);
michael@0	390	comp_avg_pred(temp3, second_pred, 16, 32, temp2, 16);
michael@0	391	return vp9_variance16x32(temp3, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	392	}
michael@0	393
michael@0	394	unsigned int vp9_variance64x64_c(const uint8_t *src_ptr,
michael@0	395	int source_stride,
michael@0	396	const uint8_t *ref_ptr,
michael@0	397	int recon_stride,
michael@0	398	unsigned int *sse) {
michael@0	399	unsigned int var;
michael@0	400	int avg;
michael@0	401
michael@0	402	variance(src_ptr, source_stride, ref_ptr, recon_stride, 64, 64, &var, &avg);
michael@0	403	*sse = var;
michael@0	404	return (var - (((int64_t)avg * avg) >> 12));
michael@0	405	}
michael@0	406
michael@0	407	unsigned int vp9_variance32x32_c(const uint8_t *src_ptr,
michael@0	408	int source_stride,
michael@0	409	const uint8_t *ref_ptr,
michael@0	410	int recon_stride,
michael@0	411	unsigned int *sse) {
michael@0	412	unsigned int var;
michael@0	413	int avg;
michael@0	414
michael@0	415	variance(src_ptr, source_stride, ref_ptr, recon_stride, 32, 32, &var, &avg);
michael@0	416	*sse = var;
michael@0	417	return (var - (((int64_t)avg * avg) >> 10));
michael@0	418	}
michael@0	419
michael@0	420	unsigned int vp9_variance16x16_c(const uint8_t *src_ptr,
michael@0	421	int source_stride,
michael@0	422	const uint8_t *ref_ptr,
michael@0	423	int recon_stride,
michael@0	424	unsigned int *sse) {
michael@0	425	unsigned int var;
michael@0	426	int avg;
michael@0	427
michael@0	428	variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
michael@0	429	*sse = var;
michael@0	430	return (var - (((unsigned int)avg * avg) >> 8));
michael@0	431	}
michael@0	432
michael@0	433	unsigned int vp9_variance8x16_c(const uint8_t *src_ptr,
michael@0	434	int source_stride,
michael@0	435	const uint8_t *ref_ptr,
michael@0	436	int recon_stride,
michael@0	437	unsigned int *sse) {
michael@0	438	unsigned int var;
michael@0	439	int avg;
michael@0	440
michael@0	441	variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16, &var, &avg);
michael@0	442	*sse = var;
michael@0	443	return (var - (((unsigned int)avg * avg) >> 7));
michael@0	444	}
michael@0	445
michael@0	446	unsigned int vp9_variance16x8_c(const uint8_t *src_ptr,
michael@0	447	int source_stride,
michael@0	448	const uint8_t *ref_ptr,
michael@0	449	int recon_stride,
michael@0	450	unsigned int *sse) {
michael@0	451	unsigned int var;
michael@0	452	int avg;
michael@0	453
michael@0	454	variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8, &var, &avg);
michael@0	455	*sse = var;
michael@0	456	return (var - (((unsigned int)avg * avg) >> 7));
michael@0	457	}
michael@0	458
michael@0	459	void vp9_get_sse_sum_8x8_c(const uint8_t *src_ptr, int source_stride,
michael@0	460	const uint8_t *ref_ptr, int ref_stride,
michael@0	461	unsigned int *sse, int *sum) {
michael@0	462	variance(src_ptr, source_stride, ref_ptr, ref_stride, 8, 8, sse, sum);
michael@0	463	}
michael@0	464
michael@0	465	unsigned int vp9_variance8x8_c(const uint8_t *src_ptr,
michael@0	466	int source_stride,
michael@0	467	const uint8_t *ref_ptr,
michael@0	468	int recon_stride,
michael@0	469	unsigned int *sse) {
michael@0	470	unsigned int var;
michael@0	471	int avg;
michael@0	472
michael@0	473	variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, &var, &avg);
michael@0	474	*sse = var;
michael@0	475	return (var - (((unsigned int)avg * avg) >> 6));
michael@0	476	}
michael@0	477
michael@0	478	unsigned int vp9_variance8x4_c(const uint8_t *src_ptr,
michael@0	479	int source_stride,
michael@0	480	const uint8_t *ref_ptr,
michael@0	481	int recon_stride,
michael@0	482	unsigned int *sse) {
michael@0	483	unsigned int var;
michael@0	484	int avg;
michael@0	485
michael@0	486	variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 4, &var, &avg);
michael@0	487	*sse = var;
michael@0	488	return (var - (((unsigned int)avg * avg) >> 5));
michael@0	489	}
michael@0	490
michael@0	491	unsigned int vp9_variance4x8_c(const uint8_t *src_ptr,
michael@0	492	int source_stride,
michael@0	493	const uint8_t *ref_ptr,
michael@0	494	int recon_stride,
michael@0	495	unsigned int *sse) {
michael@0	496	unsigned int var;
michael@0	497	int avg;
michael@0	498
michael@0	499	variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 8, &var, &avg);
michael@0	500	*sse = var;
michael@0	501	return (var - (((unsigned int)avg * avg) >> 5));
michael@0	502	}
michael@0	503
michael@0	504	unsigned int vp9_variance4x4_c(const uint8_t *src_ptr,
michael@0	505	int source_stride,
michael@0	506	const uint8_t *ref_ptr,
michael@0	507	int recon_stride,
michael@0	508	unsigned int *sse) {
michael@0	509	unsigned int var;
michael@0	510	int avg;
michael@0	511
michael@0	512	variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 4, &var, &avg);
michael@0	513	*sse = var;
michael@0	514	return (var - (((unsigned int)avg * avg) >> 4));
michael@0	515	}
michael@0	516
michael@0	517
michael@0	518	unsigned int vp9_mse16x16_c(const uint8_t *src_ptr,
michael@0	519	int source_stride,
michael@0	520	const uint8_t *ref_ptr,
michael@0	521	int recon_stride,
michael@0	522	unsigned int *sse) {
michael@0	523	unsigned int var;
michael@0	524	int avg;
michael@0	525
michael@0	526	variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
michael@0	527	*sse = var;
michael@0	528	return var;
michael@0	529	}
michael@0	530
michael@0	531	unsigned int vp9_mse16x8_c(const uint8_t *src_ptr,
michael@0	532	int source_stride,
michael@0	533	const uint8_t *ref_ptr,
michael@0	534	int recon_stride,
michael@0	535	unsigned int *sse) {
michael@0	536	unsigned int var;
michael@0	537	int avg;
michael@0	538
michael@0	539	variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8, &var, &avg);
michael@0	540	*sse = var;
michael@0	541	return var;
michael@0	542	}
michael@0	543
michael@0	544	unsigned int vp9_mse8x16_c(const uint8_t *src_ptr,
michael@0	545	int source_stride,
michael@0	546	const uint8_t *ref_ptr,
michael@0	547	int recon_stride,
michael@0	548	unsigned int *sse) {
michael@0	549	unsigned int var;
michael@0	550	int avg;
michael@0	551
michael@0	552	variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16, &var, &avg);
michael@0	553	*sse = var;
michael@0	554	return var;
michael@0	555	}
michael@0	556
michael@0	557	unsigned int vp9_mse8x8_c(const uint8_t *src_ptr,
michael@0	558	int source_stride,
michael@0	559	const uint8_t *ref_ptr,
michael@0	560	int recon_stride,
michael@0	561	unsigned int *sse) {
michael@0	562	unsigned int var;
michael@0	563	int avg;
michael@0	564
michael@0	565	variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, &var, &avg);
michael@0	566	*sse = var;
michael@0	567	return var;
michael@0	568	}
michael@0	569
michael@0	570
michael@0	571	unsigned int vp9_sub_pixel_variance4x4_c(const uint8_t *src_ptr,
michael@0	572	int src_pixels_per_line,
michael@0	573	int xoffset,
michael@0	574	int yoffset,
michael@0	575	const uint8_t *dst_ptr,
michael@0	576	int dst_pixels_per_line,
michael@0	577	unsigned int *sse) {
michael@0	578	uint8_t temp2[20 * 16];
michael@0	579	const int16_t *hfilter, *vfilter;
michael@0	580	uint16_t fdata3[5 * 4]; // Temp data buffer used in filtering
michael@0	581
michael@0	582	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	583	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	584
michael@0	585	// First filter 1d Horizontal
michael@0	586	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	587	1, 5, 4, hfilter);
michael@0	588
michael@0	589	// Now filter Verticaly
michael@0	590	var_filter_block2d_bil_second_pass(fdata3, temp2, 4, 4, 4, 4, vfilter);
michael@0	591
michael@0	592	return vp9_variance4x4(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
michael@0	593	}
michael@0	594
michael@0	595	unsigned int vp9_sub_pixel_avg_variance4x4_c(const uint8_t *src_ptr,
michael@0	596	int src_pixels_per_line,
michael@0	597	int xoffset,
michael@0	598	int yoffset,
michael@0	599	const uint8_t *dst_ptr,
michael@0	600	int dst_pixels_per_line,
michael@0	601	unsigned int *sse,
michael@0	602	const uint8_t *second_pred) {
michael@0	603	uint8_t temp2[20 * 16];
michael@0	604	const int16_t *hfilter, *vfilter;
michael@0	605	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 4 * 4); // compound pred buffer
michael@0	606	uint16_t fdata3[5 * 4]; // Temp data buffer used in filtering
michael@0	607
michael@0	608	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	609	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	610
michael@0	611	// First filter 1d Horizontal
michael@0	612	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	613	1, 5, 4, hfilter);
michael@0	614
michael@0	615	// Now filter Verticaly
michael@0	616	var_filter_block2d_bil_second_pass(fdata3, temp2, 4, 4, 4, 4, vfilter);
michael@0	617	comp_avg_pred(temp3, second_pred, 4, 4, temp2, 4);
michael@0	618	return vp9_variance4x4(temp3, 4, dst_ptr, dst_pixels_per_line, sse);
michael@0	619	}
michael@0	620
michael@0	621	unsigned int vp9_sub_pixel_variance8x8_c(const uint8_t *src_ptr,
michael@0	622	int src_pixels_per_line,
michael@0	623	int xoffset,
michael@0	624	int yoffset,
michael@0	625	const uint8_t *dst_ptr,
michael@0	626	int dst_pixels_per_line,
michael@0	627	unsigned int *sse) {
michael@0	628	uint16_t fdata3[9 * 8]; // Temp data buffer used in filtering
michael@0	629	uint8_t temp2[20 * 16];
michael@0	630	const int16_t *hfilter, *vfilter;
michael@0	631
michael@0	632	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	633	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	634
michael@0	635	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	636	1, 9, 8, hfilter);
michael@0	637	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 8, 8, vfilter);
michael@0	638
michael@0	639	return vp9_variance8x8(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	640	}
michael@0	641
michael@0	642	unsigned int vp9_sub_pixel_avg_variance8x8_c(const uint8_t *src_ptr,
michael@0	643	int src_pixels_per_line,
michael@0	644	int xoffset,
michael@0	645	int yoffset,
michael@0	646	const uint8_t *dst_ptr,
michael@0	647	int dst_pixels_per_line,
michael@0	648	unsigned int *sse,
michael@0	649	const uint8_t *second_pred) {
michael@0	650	uint16_t fdata3[9 * 8]; // Temp data buffer used in filtering
michael@0	651	uint8_t temp2[20 * 16];
michael@0	652	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 8 * 8); // compound pred buffer
michael@0	653	const int16_t *hfilter, *vfilter;
michael@0	654
michael@0	655	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	656	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	657
michael@0	658	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	659	1, 9, 8, hfilter);
michael@0	660	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 8, 8, vfilter);
michael@0	661	comp_avg_pred(temp3, second_pred, 8, 8, temp2, 8);
michael@0	662	return vp9_variance8x8(temp3, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	663	}
michael@0	664
michael@0	665	unsigned int vp9_sub_pixel_variance16x16_c(const uint8_t *src_ptr,
michael@0	666	int src_pixels_per_line,
michael@0	667	int xoffset,
michael@0	668	int yoffset,
michael@0	669	const uint8_t *dst_ptr,
michael@0	670	int dst_pixels_per_line,
michael@0	671	unsigned int *sse) {
michael@0	672	uint16_t fdata3[17 * 16]; // Temp data buffer used in filtering
michael@0	673	uint8_t temp2[20 * 16];
michael@0	674	const int16_t *hfilter, *vfilter;
michael@0	675
michael@0	676	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	677	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	678
michael@0	679	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	680	1, 17, 16, hfilter);
michael@0	681	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 16, 16, vfilter);
michael@0	682
michael@0	683	return vp9_variance16x16(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	684	}
michael@0	685
michael@0	686	unsigned int vp9_sub_pixel_avg_variance16x16_c(const uint8_t *src_ptr,
michael@0	687	int src_pixels_per_line,
michael@0	688	int xoffset,
michael@0	689	int yoffset,
michael@0	690	const uint8_t *dst_ptr,
michael@0	691	int dst_pixels_per_line,
michael@0	692	unsigned int *sse,
michael@0	693	const uint8_t *second_pred) {
michael@0	694	uint16_t fdata3[17 * 16];
michael@0	695	uint8_t temp2[20 * 16];
michael@0	696	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 16 * 16); // compound pred buffer
michael@0	697	const int16_t *hfilter, *vfilter;
michael@0	698
michael@0	699	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	700	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	701
michael@0	702	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	703	1, 17, 16, hfilter);
michael@0	704	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 16, 16, vfilter);
michael@0	705
michael@0	706	comp_avg_pred(temp3, second_pred, 16, 16, temp2, 16);
michael@0	707	return vp9_variance16x16(temp3, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	708	}
michael@0	709
michael@0	710	unsigned int vp9_sub_pixel_variance64x64_c(const uint8_t *src_ptr,
michael@0	711	int src_pixels_per_line,
michael@0	712	int xoffset,
michael@0	713	int yoffset,
michael@0	714	const uint8_t *dst_ptr,
michael@0	715	int dst_pixels_per_line,
michael@0	716	unsigned int *sse) {
michael@0	717	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	718	uint8_t temp2[68 * 64];
michael@0	719	const int16_t *hfilter, *vfilter;
michael@0	720
michael@0	721	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	722	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	723
michael@0	724	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	725	1, 65, 64, hfilter);
michael@0	726	var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 64, 64, vfilter);
michael@0	727
michael@0	728	return vp9_variance64x64(temp2, 64, dst_ptr, dst_pixels_per_line, sse);
michael@0	729	}
michael@0	730
michael@0	731	unsigned int vp9_sub_pixel_avg_variance64x64_c(const uint8_t *src_ptr,
michael@0	732	int src_pixels_per_line,
michael@0	733	int xoffset,
michael@0	734	int yoffset,
michael@0	735	const uint8_t *dst_ptr,
michael@0	736	int dst_pixels_per_line,
michael@0	737	unsigned int *sse,
michael@0	738	const uint8_t *second_pred) {
michael@0	739	uint16_t fdata3[65 * 64]; // Temp data buffer used in filtering
michael@0	740	uint8_t temp2[68 * 64];
michael@0	741	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 64 * 64); // compound pred buffer
michael@0	742	const int16_t *hfilter, *vfilter;
michael@0	743
michael@0	744	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	745	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	746
michael@0	747	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	748	1, 65, 64, hfilter);
michael@0	749	var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 64, 64, vfilter);
michael@0	750	comp_avg_pred(temp3, second_pred, 64, 64, temp2, 64);
michael@0	751	return vp9_variance64x64(temp3, 64, dst_ptr, dst_pixels_per_line, sse);
michael@0	752	}
michael@0	753
michael@0	754	unsigned int vp9_sub_pixel_variance32x32_c(const uint8_t *src_ptr,
michael@0	755	int src_pixels_per_line,
michael@0	756	int xoffset,
michael@0	757	int yoffset,
michael@0	758	const uint8_t *dst_ptr,
michael@0	759	int dst_pixels_per_line,
michael@0	760	unsigned int *sse) {
michael@0	761	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	762	uint8_t temp2[36 * 32];
michael@0	763	const int16_t *hfilter, *vfilter;
michael@0	764
michael@0	765	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	766	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	767
michael@0	768	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	769	1, 33, 32, hfilter);
michael@0	770	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 32, 32, vfilter);
michael@0	771
michael@0	772	return vp9_variance32x32(temp2, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	773	}
michael@0	774
michael@0	775	unsigned int vp9_sub_pixel_avg_variance32x32_c(const uint8_t *src_ptr,
michael@0	776	int src_pixels_per_line,
michael@0	777	int xoffset,
michael@0	778	int yoffset,
michael@0	779	const uint8_t *dst_ptr,
michael@0	780	int dst_pixels_per_line,
michael@0	781	unsigned int *sse,
michael@0	782	const uint8_t *second_pred) {
michael@0	783	uint16_t fdata3[33 * 32]; // Temp data buffer used in filtering
michael@0	784	uint8_t temp2[36 * 32];
michael@0	785	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 32 * 32); // compound pred buffer
michael@0	786	const int16_t *hfilter, *vfilter;
michael@0	787
michael@0	788	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	789	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	790
michael@0	791	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	792	1, 33, 32, hfilter);
michael@0	793	var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 32, 32, vfilter);
michael@0	794	comp_avg_pred(temp3, second_pred, 32, 32, temp2, 32);
michael@0	795	return vp9_variance32x32(temp3, 32, dst_ptr, dst_pixels_per_line, sse);
michael@0	796	}
michael@0	797
michael@0	798	unsigned int vp9_variance_halfpixvar16x16_h_c(const uint8_t *src_ptr,
michael@0	799	int source_stride,
michael@0	800	const uint8_t *ref_ptr,
michael@0	801	int recon_stride,
michael@0	802	unsigned int *sse) {
michael@0	803	return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 8, 0,
michael@0	804	ref_ptr, recon_stride, sse);
michael@0	805	}
michael@0	806
michael@0	807	unsigned int vp9_variance_halfpixvar32x32_h_c(const uint8_t *src_ptr,
michael@0	808	int source_stride,
michael@0	809	const uint8_t *ref_ptr,
michael@0	810	int recon_stride,
michael@0	811	unsigned int *sse) {
michael@0	812	return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 8, 0,
michael@0	813	ref_ptr, recon_stride, sse);
michael@0	814	}
michael@0	815
michael@0	816	unsigned int vp9_variance_halfpixvar64x64_h_c(const uint8_t *src_ptr,
michael@0	817	int source_stride,
michael@0	818	const uint8_t *ref_ptr,
michael@0	819	int recon_stride,
michael@0	820	unsigned int *sse) {
michael@0	821	return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 8, 0,
michael@0	822	ref_ptr, recon_stride, sse);
michael@0	823	}
michael@0	824
michael@0	825	unsigned int vp9_variance_halfpixvar16x16_v_c(const uint8_t *src_ptr,
michael@0	826	int source_stride,
michael@0	827	const uint8_t *ref_ptr,
michael@0	828	int recon_stride,
michael@0	829	unsigned int *sse) {
michael@0	830	return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 0, 8,
michael@0	831	ref_ptr, recon_stride, sse);
michael@0	832	}
michael@0	833
michael@0	834	unsigned int vp9_variance_halfpixvar32x32_v_c(const uint8_t *src_ptr,
michael@0	835	int source_stride,
michael@0	836	const uint8_t *ref_ptr,
michael@0	837	int recon_stride,
michael@0	838	unsigned int *sse) {
michael@0	839	return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 0, 8,
michael@0	840	ref_ptr, recon_stride, sse);
michael@0	841	}
michael@0	842
michael@0	843	unsigned int vp9_variance_halfpixvar64x64_v_c(const uint8_t *src_ptr,
michael@0	844	int source_stride,
michael@0	845	const uint8_t *ref_ptr,
michael@0	846	int recon_stride,
michael@0	847	unsigned int *sse) {
michael@0	848	return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 0, 8,
michael@0	849	ref_ptr, recon_stride, sse);
michael@0	850	}
michael@0	851
michael@0	852	unsigned int vp9_variance_halfpixvar16x16_hv_c(const uint8_t *src_ptr,
michael@0	853	int source_stride,
michael@0	854	const uint8_t *ref_ptr,
michael@0	855	int recon_stride,
michael@0	856	unsigned int *sse) {
michael@0	857	return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 8, 8,
michael@0	858	ref_ptr, recon_stride, sse);
michael@0	859	}
michael@0	860
michael@0	861	unsigned int vp9_variance_halfpixvar32x32_hv_c(const uint8_t *src_ptr,
michael@0	862	int source_stride,
michael@0	863	const uint8_t *ref_ptr,
michael@0	864	int recon_stride,
michael@0	865	unsigned int *sse) {
michael@0	866	return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 8, 8,
michael@0	867	ref_ptr, recon_stride, sse);
michael@0	868	}
michael@0	869
michael@0	870	unsigned int vp9_variance_halfpixvar64x64_hv_c(const uint8_t *src_ptr,
michael@0	871	int source_stride,
michael@0	872	const uint8_t *ref_ptr,
michael@0	873	int recon_stride,
michael@0	874	unsigned int *sse) {
michael@0	875	return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 8, 8,
michael@0	876	ref_ptr, recon_stride, sse);
michael@0	877	}
michael@0	878
michael@0	879	unsigned int vp9_sub_pixel_mse16x16_c(const uint8_t *src_ptr,
michael@0	880	int src_pixels_per_line,
michael@0	881	int xoffset,
michael@0	882	int yoffset,
michael@0	883	const uint8_t *dst_ptr,
michael@0	884	int dst_pixels_per_line,
michael@0	885	unsigned int *sse) {
michael@0	886	vp9_sub_pixel_variance16x16_c(src_ptr, src_pixels_per_line,
michael@0	887	xoffset, yoffset, dst_ptr,
michael@0	888	dst_pixels_per_line, sse);
michael@0	889	return *sse;
michael@0	890	}
michael@0	891
michael@0	892	unsigned int vp9_sub_pixel_mse32x32_c(const uint8_t *src_ptr,
michael@0	893	int src_pixels_per_line,
michael@0	894	int xoffset,
michael@0	895	int yoffset,
michael@0	896	const uint8_t *dst_ptr,
michael@0	897	int dst_pixels_per_line,
michael@0	898	unsigned int *sse) {
michael@0	899	vp9_sub_pixel_variance32x32_c(src_ptr, src_pixels_per_line,
michael@0	900	xoffset, yoffset, dst_ptr,
michael@0	901	dst_pixels_per_line, sse);
michael@0	902	return *sse;
michael@0	903	}
michael@0	904
michael@0	905	unsigned int vp9_sub_pixel_mse64x64_c(const uint8_t *src_ptr,
michael@0	906	int src_pixels_per_line,
michael@0	907	int xoffset,
michael@0	908	int yoffset,
michael@0	909	const uint8_t *dst_ptr,
michael@0	910	int dst_pixels_per_line,
michael@0	911	unsigned int *sse) {
michael@0	912	vp9_sub_pixel_variance64x64_c(src_ptr, src_pixels_per_line,
michael@0	913	xoffset, yoffset, dst_ptr,
michael@0	914	dst_pixels_per_line, sse);
michael@0	915	return *sse;
michael@0	916	}
michael@0	917
michael@0	918	unsigned int vp9_sub_pixel_variance16x8_c(const uint8_t *src_ptr,
michael@0	919	int src_pixels_per_line,
michael@0	920	int xoffset,
michael@0	921	int yoffset,
michael@0	922	const uint8_t *dst_ptr,
michael@0	923	int dst_pixels_per_line,
michael@0	924	unsigned int *sse) {
michael@0	925	uint16_t fdata3[16 * 9]; // Temp data buffer used in filtering
michael@0	926	uint8_t temp2[20 * 16];
michael@0	927	const int16_t *hfilter, *vfilter;
michael@0	928
michael@0	929	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	930	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	931
michael@0	932	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	933	1, 9, 16, hfilter);
michael@0	934	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 8, 16, vfilter);
michael@0	935
michael@0	936	return vp9_variance16x8(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	937	}
michael@0	938
michael@0	939	unsigned int vp9_sub_pixel_avg_variance16x8_c(const uint8_t *src_ptr,
michael@0	940	int src_pixels_per_line,
michael@0	941	int xoffset,
michael@0	942	int yoffset,
michael@0	943	const uint8_t *dst_ptr,
michael@0	944	int dst_pixels_per_line,
michael@0	945	unsigned int *sse,
michael@0	946	const uint8_t *second_pred) {
michael@0	947	uint16_t fdata3[16 * 9]; // Temp data buffer used in filtering
michael@0	948	uint8_t temp2[20 * 16];
michael@0	949	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 16 * 8); // compound pred buffer
michael@0	950	const int16_t *hfilter, *vfilter;
michael@0	951
michael@0	952	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	953	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	954
michael@0	955	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	956	1, 9, 16, hfilter);
michael@0	957	var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 8, 16, vfilter);
michael@0	958	comp_avg_pred(temp3, second_pred, 16, 8, temp2, 16);
michael@0	959	return vp9_variance16x8(temp3, 16, dst_ptr, dst_pixels_per_line, sse);
michael@0	960	}
michael@0	961
michael@0	962	unsigned int vp9_sub_pixel_variance8x16_c(const uint8_t *src_ptr,
michael@0	963	int src_pixels_per_line,
michael@0	964	int xoffset,
michael@0	965	int yoffset,
michael@0	966	const uint8_t *dst_ptr,
michael@0	967	int dst_pixels_per_line,
michael@0	968	unsigned int *sse) {
michael@0	969	uint16_t fdata3[9 * 16]; // Temp data buffer used in filtering
michael@0	970	uint8_t temp2[20 * 16];
michael@0	971	const int16_t *hfilter, *vfilter;
michael@0	972
michael@0	973	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	974	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	975
michael@0	976	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	977	1, 17, 8, hfilter);
michael@0	978	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 16, 8, vfilter);
michael@0	979
michael@0	980	return vp9_variance8x16(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	981	}
michael@0	982
michael@0	983	unsigned int vp9_sub_pixel_avg_variance8x16_c(const uint8_t *src_ptr,
michael@0	984	int src_pixels_per_line,
michael@0	985	int xoffset,
michael@0	986	int yoffset,
michael@0	987	const uint8_t *dst_ptr,
michael@0	988	int dst_pixels_per_line,
michael@0	989	unsigned int *sse,
michael@0	990	const uint8_t *second_pred) {
michael@0	991	uint16_t fdata3[9 * 16]; // Temp data buffer used in filtering
michael@0	992	uint8_t temp2[20 * 16];
michael@0	993	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 8 * 16); // compound pred buffer
michael@0	994	const int16_t *hfilter, *vfilter;
michael@0	995
michael@0	996	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	997	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	998
michael@0	999	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	1000	1, 17, 8, hfilter);
michael@0	1001	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 16, 8, vfilter);
michael@0	1002	comp_avg_pred(temp3, second_pred, 8, 16, temp2, 8);
michael@0	1003	return vp9_variance8x16(temp3, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	1004	}
michael@0	1005
michael@0	1006	unsigned int vp9_sub_pixel_variance8x4_c(const uint8_t *src_ptr,
michael@0	1007	int src_pixels_per_line,
michael@0	1008	int xoffset,
michael@0	1009	int yoffset,
michael@0	1010	const uint8_t *dst_ptr,
michael@0	1011	int dst_pixels_per_line,
michael@0	1012	unsigned int *sse) {
michael@0	1013	uint16_t fdata3[8 * 5]; // Temp data buffer used in filtering
michael@0	1014	uint8_t temp2[20 * 16];
michael@0	1015	const int16_t *hfilter, *vfilter;
michael@0	1016
michael@0	1017	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	1018	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	1019
michael@0	1020	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	1021	1, 5, 8, hfilter);
michael@0	1022	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 4, 8, vfilter);
michael@0	1023
michael@0	1024	return vp9_variance8x4(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	1025	}
michael@0	1026
michael@0	1027	unsigned int vp9_sub_pixel_avg_variance8x4_c(const uint8_t *src_ptr,
michael@0	1028	int src_pixels_per_line,
michael@0	1029	int xoffset,
michael@0	1030	int yoffset,
michael@0	1031	const uint8_t *dst_ptr,
michael@0	1032	int dst_pixels_per_line,
michael@0	1033	unsigned int *sse,
michael@0	1034	const uint8_t *second_pred) {
michael@0	1035	uint16_t fdata3[8 * 5]; // Temp data buffer used in filtering
michael@0	1036	uint8_t temp2[20 * 16];
michael@0	1037	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 8 * 4); // compound pred buffer
michael@0	1038	const int16_t *hfilter, *vfilter;
michael@0	1039
michael@0	1040	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	1041	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	1042
michael@0	1043	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	1044	1, 5, 8, hfilter);
michael@0	1045	var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 4, 8, vfilter);
michael@0	1046	comp_avg_pred(temp3, second_pred, 8, 4, temp2, 8);
michael@0	1047	return vp9_variance8x4(temp3, 8, dst_ptr, dst_pixels_per_line, sse);
michael@0	1048	}
michael@0	1049
michael@0	1050	unsigned int vp9_sub_pixel_variance4x8_c(const uint8_t *src_ptr,
michael@0	1051	int src_pixels_per_line,
michael@0	1052	int xoffset,
michael@0	1053	int yoffset,
michael@0	1054	const uint8_t *dst_ptr,
michael@0	1055	int dst_pixels_per_line,
michael@0	1056	unsigned int *sse) {
michael@0	1057	uint16_t fdata3[5 * 8]; // Temp data buffer used in filtering
michael@0	1058	// FIXME(jingning,rbultje): this temp2 buffer probably doesn't need to be
michael@0	1059	// of this big? same issue appears in all other block size settings.
michael@0	1060	uint8_t temp2[20 * 16];
michael@0	1061	const int16_t *hfilter, *vfilter;
michael@0	1062
michael@0	1063	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	1064	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	1065
michael@0	1066	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	1067	1, 9, 4, hfilter);
michael@0	1068	var_filter_block2d_bil_second_pass(fdata3, temp2, 4, 4, 8, 4, vfilter);
michael@0	1069
michael@0	1070	return vp9_variance4x8(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
michael@0	1071	}
michael@0	1072
michael@0	1073	unsigned int vp9_sub_pixel_avg_variance4x8_c(const uint8_t *src_ptr,
michael@0	1074	int src_pixels_per_line,
michael@0	1075	int xoffset,
michael@0	1076	int yoffset,
michael@0	1077	const uint8_t *dst_ptr,
michael@0	1078	int dst_pixels_per_line,
michael@0	1079	unsigned int *sse,
michael@0	1080	const uint8_t *second_pred) {
michael@0	1081	uint16_t fdata3[5 * 8]; // Temp data buffer used in filtering
michael@0	1082	uint8_t temp2[20 * 16];
michael@0	1083	DECLARE_ALIGNED_ARRAY(16, uint8_t, temp3, 4 * 8); // compound pred buffer
michael@0	1084	const int16_t *hfilter, *vfilter;
michael@0	1085
michael@0	1086	hfilter = BILINEAR_FILTERS_2TAP(xoffset);
michael@0	1087	vfilter = BILINEAR_FILTERS_2TAP(yoffset);
michael@0	1088
michael@0	1089	var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
michael@0	1090	1, 9, 4, hfilter);
michael@0	1091	var_filter_block2d_bil_second_pass(fdata3, temp2, 4, 4, 8, 4, vfilter);
michael@0	1092	comp_avg_pred(temp3, second_pred, 4, 8, temp2, 4);
michael@0	1093	return vp9_variance4x8(temp3, 4, dst_ptr, dst_pixels_per_line, sse);
michael@0	1094	}