rustsynth-sys 0.5.6

Low level bindings to VapourSynth
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282

#include <assert.h>
#include <stdint.h>
#include <emmintrin.h>
#include "VSHelper4.h"
#include "../average.h"

static void load_int_srcs(const uint8_t **srcs, const void * const *srcs_, unsigned num_srcs)
{
	unsigned i;

	assert(num_srcs <= 32);

	for (i = 0; i < num_srcs; ++i) {
		srcs[i] = srcs_[i];
	}
	if (num_srcs % 2)
		srcs[num_srcs] = srcs[num_srcs - 1];
}

static void load_int_weights(__m128i mm_weights[16], const int *iweights, unsigned num_weights)
{
	unsigned i;

	for (i = 0; i < (num_weights & ~1); i += 2) {
		int16_t lo = iweights[i + 0];
		int16_t hi = iweights[i + 1];
		uint32_t coeff = ((uint32_t)(uint16_t)hi) << 16 | (uint16_t)lo;
		mm_weights[i / 2] = _mm_set1_epi32(coeff);
	}
	if (num_weights % 2)
		mm_weights[num_weights / 2] = _mm_set1_epi32((uint16_t)iweights[num_weights - 1]);
}

void vs_average_plane_byte_luma_sse2(const void *weights_, const void * const *srcs_, unsigned num_srcs, void *dst_, const void *scale_, unsigned depth, unsigned w, unsigned h, ptrdiff_t stride)
{
	const uint8_t *srcs[32];
	__m128i weights[16];
	__m128 scale = _mm_set_ps1(1.0f / *(const int *)scale_);
	ptrdiff_t offset = 0;
	unsigned i, j, k;

	load_int_srcs(srcs, srcs_, num_srcs);
	load_int_weights(weights, weights_, num_srcs);

	for (i = 0; i < h; ++i) {
		uint8_t *dst = (uint8_t *)dst_ + offset;

		for (j = 0; j < w; j += 16) {
			__m128i lolo = _mm_setzero_si128();
			__m128i lohi = _mm_setzero_si128();
			__m128i hilo = _mm_setzero_si128();
			__m128i hihi = _mm_setzero_si128();

			for (k = 0; k < num_srcs; k += 2) {
				const uint8_t *ptr1 = srcs[k + 0] + offset;
				const uint8_t *ptr2 = srcs[k + 1] + offset;

				__m128i coeffs = weights[k / 2];
				__m128i v1 = _mm_load_si128((const __m128i *)(ptr1 + j));
				__m128i v2 = _mm_load_si128((const __m128i *)(ptr2 + j));

				__m128i v1_lo = _mm_unpacklo_epi8(v1, _mm_setzero_si128());
				__m128i v1_hi = _mm_unpackhi_epi8(v1, _mm_setzero_si128());
				__m128i v2_lo = _mm_unpacklo_epi8(v2, _mm_setzero_si128());
				__m128i v2_hi = _mm_unpackhi_epi8(v2, _mm_setzero_si128());

				lolo = _mm_add_epi32(lolo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1_lo, v2_lo)));
				lohi = _mm_add_epi32(lohi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1_lo, v2_lo)));
				hilo = _mm_add_epi32(hilo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1_hi, v2_hi)));
				hihi = _mm_add_epi32(hihi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1_hi, v2_hi)));
			}

			lolo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lolo), scale));
			lohi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lohi), scale));
			hilo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hilo), scale));
			hihi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hihi), scale));

			lolo = _mm_packs_epi32(lolo, lohi);
			hilo = _mm_packs_epi32(hilo, hihi);
			lolo = _mm_packus_epi16(lolo, hilo);

			_mm_store_si128((__m128i *)(dst + j), lolo);
		}

		offset += stride;
	}
}

void vs_average_plane_byte_chroma_sse2(const void *weights_, const void * const *srcs_, unsigned num_srcs, void *dst_, const void *scale_, unsigned depth, unsigned w, unsigned h, ptrdiff_t stride)
{
	const uint8_t *srcs[32];
	__m128i weights[16];
	__m128 scale = _mm_set_ps1(1.0f / *(const int *)scale_);
	__m128i bias_i16 = _mm_set1_epi16(128);
	__m128i bias_i8 = _mm_set1_epi8(128);
	ptrdiff_t offset = 0;
	unsigned i, j, k;

	load_int_srcs(srcs, srcs_, num_srcs);
	load_int_weights(weights, weights_, num_srcs);

	for (i = 0; i < h; ++i) {
		uint8_t *dst = (uint8_t *)dst_ + offset;

		for (j = 0; j < w; j += 16) {
			__m128i lolo = _mm_setzero_si128();
			__m128i lohi = _mm_setzero_si128();
			__m128i hilo = _mm_setzero_si128();
			__m128i hihi = _mm_setzero_si128();

			for (k = 0; k < num_srcs; k += 2) {
				const uint8_t *ptr1 = srcs[k + 0] + offset;
				const uint8_t *ptr2 = srcs[k + 1] + offset;

				__m128i coeffs = weights[k / 2];
				__m128i v1 = _mm_load_si128((const __m128i *)(ptr1 + j));
				__m128i v2 = _mm_load_si128((const __m128i *)(ptr2 + j));

				__m128i v1_lo = _mm_sub_epi16(_mm_unpacklo_epi8(v1, _mm_setzero_si128()), bias_i16);
				__m128i v1_hi = _mm_sub_epi16(_mm_unpackhi_epi8(v1, _mm_setzero_si128()), bias_i16);
				__m128i v2_lo = _mm_sub_epi16(_mm_unpacklo_epi8(v2, _mm_setzero_si128()), bias_i16);
				__m128i v2_hi = _mm_sub_epi16(_mm_unpackhi_epi8(v2, _mm_setzero_si128()), bias_i16);

				lolo = _mm_add_epi32(lolo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1_lo, v2_lo)));
				lohi = _mm_add_epi32(lohi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1_lo, v2_lo)));
				hilo = _mm_add_epi32(hilo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1_hi, v2_hi)));
				hihi = _mm_add_epi32(hihi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1_hi, v2_hi)));
			}

			lolo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lolo), scale));
			lohi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lohi), scale));
			hilo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hilo), scale));
			hihi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hihi), scale));

			lolo = _mm_packs_epi32(lolo, lohi);
			hilo = _mm_packs_epi32(hilo, hihi);
			lolo = _mm_packs_epi16(lolo, hilo);
			lolo = _mm_add_epi8(lolo, bias_i8);

			_mm_store_si128((__m128i *)(dst + j), lolo);
		}

		offset += stride;
	}
}

void vs_average_plane_word_luma_sse2(const void *weights_, const void * const *srcs_, unsigned num_srcs, void *dst_, const void *scale_, unsigned depth, unsigned w, unsigned h, ptrdiff_t stride)
{
	const uint8_t *srcs[32];
	__m128i weights[16];
	__m128 scale = _mm_set_ps1(1.0f / *(const int *)scale_);
	__m128i maxval = _mm_add_epi16(_mm_set1_epi16((1U << depth) - 1), _mm_set1_epi16(INT16_MIN));
	__m128i accum_bias = _mm_setzero_si128();
	ptrdiff_t offset = 0;
	unsigned i, j, k;

	load_int_srcs(srcs, srcs_, num_srcs);
	load_int_weights(weights, weights_, num_srcs);

	/* sum(weights * int16_min) */
	for (unsigned i = 0; i < num_srcs; i += 2) {
		accum_bias = _mm_add_epi32(accum_bias, _mm_madd_epi16(_mm_set1_epi16(INT16_MIN), weights[i / 2]));
	}

	for (i = 0; i < h; ++i) {
		uint16_t *dst = (uint16_t *)((uint8_t *)dst_ + offset);

		for (j = 0; j < w; j += 8) {
			__m128i lo = _mm_setzero_si128();
			__m128i hi = _mm_setzero_si128();

			for (k = 0; k < num_srcs; k += 2) {
				const uint16_t *ptr1 = (const uint16_t *)(srcs[k + 0] + offset);
				const uint16_t *ptr2 = (const uint16_t *)(srcs[k + 1] + offset);

				__m128i coeffs = weights[k / 2];
				__m128i v1 = _mm_add_epi16(_mm_load_si128((const __m128i *)(ptr1 + j)), _mm_set1_epi16(INT16_MIN));
				__m128i v2 = _mm_add_epi16(_mm_load_si128((const __m128i *)(ptr2 + j)), _mm_set1_epi16(INT16_MIN));

				lo = _mm_add_epi32(lo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1, v2)));
				hi = _mm_add_epi32(hi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1, v2)));
			}
			lo = _mm_sub_epi32(lo, accum_bias);
			hi = _mm_sub_epi32(hi, accum_bias);

			lo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lo), scale));
			hi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hi), scale));

			lo = _mm_add_epi32(lo, _mm_set1_epi32(INT16_MIN));
			hi = _mm_add_epi32(hi, _mm_set1_epi32(INT16_MIN));
			lo = _mm_packs_epi32(lo, hi);

			lo = _mm_min_epi16(lo, maxval);
			lo = _mm_sub_epi16(lo, _mm_set1_epi16(INT16_MIN));

			_mm_store_si128((__m128i *)(dst + j), lo);
		}

		offset += stride;
	}
}

void vs_average_plane_word_chroma_sse2(const void *weights_, const void * const *srcs_, unsigned num_srcs, void *dst_, const void *scale_, unsigned depth, unsigned w, unsigned h, ptrdiff_t stride)
{
	const uint8_t *srcs[32];
	__m128i weights[16];
	__m128 scale = _mm_set_ps1(1.0f / *(const int *)scale_);
	__m128i bias = _mm_set1_epi16(1U << (depth - 1));
	__m128i minval = _mm_sub_epi16(_mm_setzero_si128(), bias);
	__m128i maxval = _mm_sub_epi16(_mm_set1_epi16((1U << depth) - 1), bias);
	ptrdiff_t offset = 0;
	unsigned i, j, k;

	load_int_srcs(srcs, srcs_, num_srcs);
	load_int_weights(weights, weights_, num_srcs);

	for (i = 0; i < h; ++i) {
		uint16_t *dst = (uint16_t *)((uint8_t *)dst_ + offset);

		for (j = 0; j < w; j += 8) {
			__m128i lo = _mm_setzero_si128();
			__m128i hi = _mm_setzero_si128();

			for (k = 0; k < num_srcs; k += 2) {
				const uint16_t *ptr1 = (const uint16_t *)(srcs[k + 0] + offset);
				const uint16_t *ptr2 = (const uint16_t *)(srcs[k + 1] + offset);

				__m128i coeffs = weights[k / 2];
				__m128i v1 = _mm_sub_epi16(_mm_load_si128((const __m128i *)(ptr1 + j)), bias);
				__m128i v2 = _mm_sub_epi16(_mm_load_si128((const __m128i *)(ptr2 + j)), bias);

				lo = _mm_add_epi32(lo, _mm_madd_epi16(coeffs, _mm_unpacklo_epi16(v1, v2)));
				hi = _mm_add_epi32(hi, _mm_madd_epi16(coeffs, _mm_unpackhi_epi16(v1, v2)));
			}

			lo = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(lo), scale));
			hi = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(hi), scale));
			lo = _mm_packs_epi32(lo, hi);

			lo = _mm_max_epi16(lo, minval);
			lo = _mm_min_epi16(lo, maxval);
			lo = _mm_add_epi16(lo, bias);

			_mm_store_si128((__m128i *)(dst + j), lo);
		}

		offset += stride;
	}
}

void vs_average_plane_float_sse2(const void *weights_, const void * const *srcs, unsigned num_srcs, void *dst_, const void *scale_, unsigned depth, unsigned w, unsigned h, ptrdiff_t stride)
{
	__m128 weights[32];
	__m128 scale = _mm_set_ps1(1.0f / *(const float *)scale_);
	ptrdiff_t offset = 0;
	unsigned i, j, k;

	assert(num_srcs <= 32);

	for (i = 0; i < num_srcs; ++i) {
		weights[i] = _mm_set_ps1(((const float *)weights_)[i]);
	}

	for (i = 0; i < h; ++i) {
		float *dst = (float *)((uint8_t *)dst_ + offset);

		for (j = 0; j < w; j += 4) {
			__m128 accum = _mm_setzero_ps();

			for (k = 0; k < num_srcs; ++k) {
				const float *ptr = (const float *)((const uint8_t *)srcs[k] + offset);
				__m128 val = _mm_load_ps(ptr + j);
				accum = _mm_add_ps(accum, _mm_mul_ps(val, weights[k]));
			}

			accum = _mm_mul_ps(accum, scale);
			_mm_store_ps(dst + j, accum);
		}

		offset += stride;
	}
}