File: reduceh_hwy.cpp

package info (click to toggle)
vips 8.16.1-2
  • links: PTS
  • area: main
  • in suites: sid
  • size: 51,940 kB
  • sloc: ansic: 169,179; cpp: 11,890; python: 4,620; xml: 4,353; sh: 732; perl: 40; makefile: 19
file content (236 lines) | stat: -rw-r--r-- 6,253 bytes parent folder | download | duplicates (2)
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
/* 22/07/23 kleisauke
 * 	- from reducev_hwy.cpp
 * 02/12/23 kleisauke
 * 	- prefer use of Dup128VecFromValues
 */

/*

	This file is part of VIPS.

	VIPS is free software; you can redistribute it and/or modify
	it under the terms of the GNU Lesser General Public License as published by
	the Free Software Foundation; either version 2 of the License, or
	(at your option) any later version.

	This program is distributed in the hope that it will be useful,
	but WITHOUT ANY WARRANTY; without even the implied warranty of
	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
	GNU Lesser General Public License for more details.

	You should have received a copy of the GNU Lesser General Public License
	along with this program; if not, write to the Free Software
	Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
	021100301  USA

 */

/*

	These files are distributed with VIPS - http://www.vips.ecs.soton.ac.uk

 */

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif /*HAVE_CONFIG_H*/
#include <glib/gi18n-lib.h>

#include <stdio.h>
#include <stdlib.h>
#include <math.h>

#include <vips/vips.h>
#include <vips/vector.h>
#include <vips/debug.h>
#include <vips/internal.h>

#include "presample.h"

#ifdef HAVE_HWY

#undef HWY_TARGET_INCLUDE
#define HWY_TARGET_INCLUDE "libvips/resample/reduceh_hwy.cpp"
#include <hwy/foreach_target.h>
#include <hwy/highway.h>

namespace HWY_NAMESPACE {

using namespace hwy::HWY_NAMESPACE;

using DI32 = ScalableTag<int32_t>;
using DI16 = ScalableTag<int16_t>;
using DI8 = ScalableTag<int8_t>;
using DU8 = ScalableTag<uint8_t>;
constexpr DU8 du8;
constexpr Rebind<uint8_t, DI32> du8x32;
constexpr DI8 di8;
constexpr DI16 di16;
constexpr DI32 di32;

HWY_ATTR void
vips_reduceh_uchar_hwy(VipsPel *pout, VipsPel *pin,
	int32_t n, int32_t width, int32_t bands,
	int16_t *HWY_RESTRICT cs[VIPS_TRANSFORM_SCALE + 1],
	double X, double hshrink)
{
#if HWY_TARGET != HWY_SCALAR
	const auto initial = Set(di32, VIPS_INTERPOLATE_SCALE >> 1);

#ifdef HAVE_HWY_1_1_0
	/*  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
	 * r0 g0 b0 r1 g1 b1 r2 g2 b2 r3 g3 b3
	 */
	const auto shuf3_lo = Dup128VecFromValues(di8,
		0, -1, 3, -1, 1, -1, 4, -1,
		2, -1, 5, -1, -1, -1, -1, -1);
	const auto shuf3_hi = Dup128VecFromValues(di8,
		6, -1, 9, -1, 7, -1, 10, -1,
		8, -1, 11, -1, -1, -1, -1, -1);

	/*  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
	 * r0 g0 b0 a0 r1 g1 b1 a1 r2 g2 b2 a2 r3 g3 b3 a3
	 */
	const auto shuf4_lo = Dup128VecFromValues(di8,
		0, -1, 4, -1, 1, -1, 5, -1,
		2, -1, 6, -1, 3, -1, 7, -1);
	const auto shuf4_hi = Dup128VecFromValues(di8,
		8, -1, 12, -1, 9, -1, 13, -1,
		10, -1, 14, -1, 11, -1, 15, -1);

	const auto shuf_lo = BitCast(di16, bands == 3 ? shuf3_lo : shuf4_lo);
	const auto shuf_hi = BitCast(di16, bands == 3 ? shuf3_hi : shuf4_hi);
#else
	/*  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
	 * r0 g0 b0 r1 g1 b1 r2 g2 b2 r3 g3 b3
	 */
	alignas(16) static constexpr int8_t tbl3_lo[16] = {
		0, -1, 3, -1, 1, -1, 4, -1,
		2, -1, 5, -1, -1, -1, -1, -1
	};
	alignas(16) static constexpr int8_t tbl3_hi[16] = {
		6, -1, 9, -1, 7, -1, 10, -1,
		8, -1, 11, -1, -1, -1, -1, -1
	};

	/*  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
	 * r0 g0 b0 a0 r1 g1 b1 a1 r2 g2 b2 a2 r3 g3 b3 a3
	 */
	alignas(16) static constexpr int8_t tbl4_lo[16] = {
		0, -1, 4, -1, 1, -1, 5, -1,
		2, -1, 6, -1, 3, -1, 7, -1
	};
	alignas(16) static constexpr int8_t tbl4_hi[16] = {
		8, -1, 12, -1, 9, -1, 13, -1,
		10, -1, 14, -1, 11, -1, 15, -1
	};

	const auto shuf_lo = BitCast(di16,
		LoadDup128(di8, bands == 3 ? tbl3_lo : tbl4_lo));
	const auto shuf_hi = BitCast(di16,
		LoadDup128(di8, bands == 3 ? tbl3_hi : tbl4_hi));
#endif

	for (int32_t x = 0; x < width; ++x) {
		const int ix = (int) X;
		const int sx = X * VIPS_TRANSFORM_SCALE * 2;
		const int six = sx & (VIPS_TRANSFORM_SCALE * 2 - 1);
		const int tx = (six + 1) >> 1;
		const int16_t *k = cs[tx];

		auto *HWY_RESTRICT p = (uint8_t *) pin + ix * bands;
		auto *HWY_RESTRICT q = (uint8_t *) pout + x * bands;

#if HWY_ARCH_X86 || HWY_ARCH_WASM || HWY_TARGET == HWY_EMU128
		/* Initialize the sum with the addition on x86 and Wasm,
		 * avoids an extra add instruction. Should be safe given
		 * that only one accumulator is used.
		 */
		auto sum0 = initial;
#else
		auto sum0 = Zero(di32);
#endif
		auto sum1 = Zero(di32); /* unused on x86 and Wasm */

		int32_t i = 0;
		for (; i + 4 <= n; i += 4) {
			/* Load four coefficients.
			 */
			auto mmk_lo = BitCast(di16, Set(di32, *(int32_t *) &k[i]));
			auto mmk_hi = BitCast(di16, Set(di32, *(int32_t *) &k[i + 2]));

			auto source = LoadU(du8, p);
			p += bands * 4;

			auto pix = TableLookupBytesOr0(source, shuf_lo);

			sum0 = ReorderWidenMulAccumulate(di32, pix, mmk_lo, sum0,
				/* byref */ sum1);

			pix = TableLookupBytesOr0(source, shuf_hi);

			sum0 = ReorderWidenMulAccumulate(di32, pix, mmk_hi, sum0,
				/* byref */ sum1);
		}
		for (; i + 2 <= n; i += 2) {
			/* Load two coefficients at once.
			 */
			auto mmk_lo = BitCast(di16, Set(di32, *(int32_t *) &k[i]));

			auto source = LoadU(du8, p);
			p += bands * 2;

			auto pix = TableLookupBytesOr0(source, shuf_lo);

			sum0 = ReorderWidenMulAccumulate(di32, pix, mmk_lo, sum0,
				/* byref */ sum1);
		}
		for (; i < n; ++i) {
			auto mmk = Set(di16, k[i]);

			auto source = LoadU(du8x32, p);
			p += bands;

			auto pix = BitCast(di16, PromoteTo(di32, source));

			sum0 = ReorderWidenMulAccumulate(di32, pix, mmk, sum0,
				/* byref */ sum1);
		}

		sum0 = RearrangeToOddPlusEven(sum0, sum1);

#if !(HWY_ARCH_X86 || HWY_ARCH_WASM || HWY_TARGET == HWY_EMU128)
		sum0 = Add(sum0, initial);
#endif

		/* The final 32->8 conversion.
		 */
		sum0 = ShiftRight<VIPS_INTERPOLATE_SHIFT>(sum0);

		auto demoted = DemoteTo(du8x32, sum0);
		StoreU(demoted, du8x32, q);

		X += hshrink;
	}
#endif
}

} /*namespace HWY_NAMESPACE*/

#if HWY_ONCE
HWY_EXPORT(vips_reduceh_uchar_hwy);

void
vips_reduceh_uchar_hwy(VipsPel *pout, VipsPel *pin,
	int n, int width, int bands,
	short *restrict cs[VIPS_TRANSFORM_SCALE + 1],
	double X, double hshrink)
{
	/* clang-format off */
	HWY_DYNAMIC_DISPATCH(vips_reduceh_uchar_hwy)(pout, pin,
		n, width, bands, cs, X, hshrink);
	/* clang-format on */
}
#endif /*HWY_ONCE*/

#endif /*HAVE_HWY*/