1 |
|
|
/* |
2 |
|
|
* The ManaPlus Client |
3 |
|
|
* Copyright (C) 2007-2009 The Mana World Development Team |
4 |
|
|
* Copyright (C) 2009-2010 The Mana Developers |
5 |
|
|
* Copyright (C) 2011-2019 The ManaPlus Developers |
6 |
|
|
* Copyright (C) 2019-2021 Andrei Karas |
7 |
|
|
* |
8 |
|
|
* This file is part of The ManaPlus Client. |
9 |
|
|
* |
10 |
|
|
* This program is free software; you can redistribute it and/or modify |
11 |
|
|
* it under the terms of the GNU General Public License as published by |
12 |
|
|
* the Free Software Foundation; either version 2 of the License, or |
13 |
|
|
* any later version. |
14 |
|
|
* |
15 |
|
|
* This program is distributed in the hope that it will be useful, |
16 |
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
17 |
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
18 |
|
|
* GNU General Public License for more details. |
19 |
|
|
* |
20 |
|
|
* You should have received a copy of the GNU General Public License |
21 |
|
|
* along with this program. If not, see <http://www.gnu.org/licenses/>. |
22 |
|
|
*/ |
23 |
|
|
|
24 |
|
|
#include "resources/dye/dyepalette.h" |
25 |
|
|
|
26 |
|
|
PRAGMA48(GCC diagnostic push) |
27 |
|
|
PRAGMA48(GCC diagnostic ignored "-Wshadow") |
28 |
|
|
#ifndef SDL_BIG_ENDIAN |
29 |
|
|
#include <SDL_endian.h> |
30 |
|
|
#endif // SDL_BYTEORDER |
31 |
|
|
PRAGMA48(GCC diagnostic pop) |
32 |
|
|
|
33 |
|
|
#ifdef SIMD_SUPPORTED |
34 |
|
|
// avx2 |
35 |
|
|
#include <immintrin.h> |
36 |
|
|
#endif // SIMD_SUPPORTED |
37 |
|
|
|
38 |
|
|
#include "debug.h" |
39 |
|
|
|
40 |
|
11 |
void DyePalette::replaceAColorDefault(uint32_t *restrict pixels, |
41 |
|
|
const int bufSize) const restrict2 |
42 |
|
|
{ |
43 |
|
22 |
STD_VECTOR<DyeColor>::const_iterator it_end = mColors.end(); |
44 |
|
22 |
const size_t sz = mColors.size(); |
45 |
✓✗ |
11 |
if ((sz == 0U) || (pixels == nullptr)) |
46 |
|
|
return; |
47 |
✗✓ |
11 |
if ((sz % 2) != 0U) |
48 |
|
|
-- it_end; |
49 |
|
|
|
50 |
|
62 |
for (const uint32_t *const p_end = pixels + CAST_SIZE(bufSize); |
51 |
✓✓ |
62 |
pixels != p_end; |
52 |
|
|
++pixels) |
53 |
|
|
{ |
54 |
|
51 |
uint8_t *const p = reinterpret_cast<uint8_t *>(pixels); |
55 |
|
51 |
const unsigned int data = *pixels; |
56 |
|
|
|
57 |
|
102 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
58 |
✓✓ |
102 |
while (it != it_end) |
59 |
|
|
{ |
60 |
|
83 |
const DyeColor &col = *it; |
61 |
|
83 |
++ it; |
62 |
|
83 |
const DyeColor &col2 = *it; |
63 |
|
|
|
64 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
65 |
|
|
const unsigned int coldata = (col.value[3] << 24U) |
66 |
|
|
| (col.value[2] << 16U) |
67 |
|
|
| (col.value[1] << 8U) |
68 |
|
|
| (col.value[0]); |
69 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
70 |
|
83 |
const unsigned int coldata = (col.value[3]) |
71 |
|
83 |
| (col.value[2] << 8U) |
72 |
|
166 |
| (col.value[1] << 16U) | |
73 |
|
166 |
(col.value[0] << 24U); |
74 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
75 |
|
|
|
76 |
✓✓ |
83 |
if (data == coldata) |
77 |
|
|
{ |
78 |
|
32 |
p[3] = col2.value[0]; |
79 |
|
32 |
p[2] = col2.value[1]; |
80 |
|
32 |
p[1] = col2.value[2]; |
81 |
|
32 |
p[0] = col2.value[3]; |
82 |
|
32 |
break; |
83 |
|
|
} |
84 |
|
|
|
85 |
|
|
++ it; |
86 |
|
|
} |
87 |
|
|
} |
88 |
|
|
} |
89 |
|
|
|
90 |
|
|
#ifdef SIMD_SUPPORTED |
91 |
|
|
/* |
92 |
|
|
static void print256(const char *const text, const __m256i &val); |
93 |
|
|
static void print256(const char *const text, const __m256i &val) |
94 |
|
|
{ |
95 |
|
|
printf("%s 0x%016llx%016llx%016llx%016llx\n", text, val[0], val[1], val[2], val[3]); |
96 |
|
|
} |
97 |
|
|
*/ |
98 |
|
|
|
99 |
|
|
__attribute__ ((target ("sse2"))) |
100 |
|
11 |
void DyePalette::replaceAColorSse2(uint32_t *restrict pixels, |
101 |
|
|
const int bufSize) const restrict2 |
102 |
|
|
{ |
103 |
|
22 |
STD_VECTOR<DyeColor>::const_iterator it_end = mColors.end(); |
104 |
|
22 |
const size_t sz = mColors.size(); |
105 |
✓✗ |
11 |
if ((sz == 0U) || (pixels == nullptr)) |
106 |
|
|
return; |
107 |
✗✓ |
11 |
if ((sz % 2) != 0U) |
108 |
|
|
-- it_end; |
109 |
|
11 |
const int mod = bufSize % 4; |
110 |
|
11 |
const int bufEnd = bufSize - mod; |
111 |
|
|
|
112 |
✓✓ |
20 |
for (int ptr = 0; ptr < bufEnd; ptr += 4) |
113 |
|
|
{ |
114 |
|
|
// __m128i base = _mm_load_si128(reinterpret_cast<__m128i*>(pixels)); |
115 |
|
|
__m128i base = _mm_loadu_si128(reinterpret_cast<__m128i*>( |
116 |
|
18 |
&pixels[ptr])); |
117 |
|
|
|
118 |
|
18 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
119 |
✓✓ |
27 |
while (it != it_end) |
120 |
|
|
{ |
121 |
|
18 |
const DyeColor &col = *it; |
122 |
|
18 |
++ it; |
123 |
|
18 |
const DyeColor &col2 = *it; |
124 |
|
|
|
125 |
|
36 |
__m128i newMask = _mm_set1_epi32(col2.valueA); |
126 |
|
36 |
__m128i cmpMask = _mm_set1_epi32(col.valueA); |
127 |
|
18 |
__m128i cmpRes = _mm_cmpeq_epi32(base, cmpMask); |
128 |
|
18 |
__m128i srcAnd = _mm_andnot_si128(cmpRes, base); |
129 |
|
18 |
__m128i dstAnd = _mm_and_si128(cmpRes, newMask); |
130 |
|
18 |
base = _mm_or_si128(srcAnd, dstAnd); |
131 |
|
|
|
132 |
|
|
++ it; |
133 |
|
|
} |
134 |
|
|
// _mm_store_si128(reinterpret_cast<__m128i*>(pixels), base); |
135 |
|
18 |
_mm_storeu_si128(reinterpret_cast<__m128i*>(&pixels[ptr]), base); |
136 |
|
|
} |
137 |
|
|
|
138 |
|
|
// complete end without simd |
139 |
✓✓ |
41 |
for (int ptr = bufSize - mod; ptr < bufSize; ptr ++) |
140 |
|
|
{ |
141 |
|
15 |
uint8_t *const p = reinterpret_cast<uint8_t *>(&pixels[ptr]); |
142 |
|
15 |
const unsigned int data = pixels[ptr]; |
143 |
|
|
|
144 |
|
30 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
145 |
✓✓ |
27 |
while (it != it_end) |
146 |
|
|
{ |
147 |
|
23 |
const DyeColor &col = *it; |
148 |
|
23 |
++ it; |
149 |
|
23 |
const DyeColor &col2 = *it; |
150 |
|
|
|
151 |
|
46 |
const unsigned int coldata = (col.value[3]) | |
152 |
|
46 |
(col.value[2] << 8U) | |
153 |
|
46 |
(col.value[1] << 16U) | |
154 |
|
46 |
(col.value[0] << 24U); |
155 |
|
|
|
156 |
✓✓ |
23 |
if (data == coldata) |
157 |
|
|
{ |
158 |
|
11 |
p[3] = col2.value[0]; |
159 |
|
11 |
p[2] = col2.value[1]; |
160 |
|
11 |
p[1] = col2.value[2]; |
161 |
|
11 |
p[0] = col2.value[3]; |
162 |
|
11 |
break; |
163 |
|
|
} |
164 |
|
|
|
165 |
|
|
++ it; |
166 |
|
|
} |
167 |
|
|
} |
168 |
|
|
} |
169 |
|
|
|
170 |
|
|
__attribute__ ((target ("avx2"))) |
171 |
|
23 |
void DyePalette::replaceAColorAvx2(uint32_t *restrict pixels, |
172 |
|
|
const int bufSize) const restrict2 |
173 |
|
|
{ |
174 |
|
46 |
STD_VECTOR<DyeColor>::const_iterator it_end = mColors.end(); |
175 |
|
46 |
const size_t sz = mColors.size(); |
176 |
✓✗ |
23 |
if ((sz == 0U) || (pixels == nullptr)) |
177 |
|
|
return; |
178 |
✗✓ |
23 |
if ((sz % 2) != 0U) |
179 |
|
|
-- it_end; |
180 |
|
23 |
const int mod = bufSize % 8; |
181 |
|
23 |
const int bufEnd = bufSize - mod; |
182 |
|
|
|
183 |
✓✓ |
157 |
for (int ptr = 0; ptr < bufEnd; ptr += 8) |
184 |
|
|
{ |
185 |
|
|
// __m256i base = _mm256_load_si256(reinterpret_cast<__m256i*>(pixels)); |
186 |
|
|
__m256i base = _mm256_loadu_si256(reinterpret_cast<__m256i*>( |
187 |
|
268 |
&pixels[ptr])); |
188 |
|
|
|
189 |
|
268 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
190 |
✓✓ |
274 |
while (it != it_end) |
191 |
|
|
{ |
192 |
|
140 |
const DyeColor &col = *it; |
193 |
|
140 |
++ it; |
194 |
|
140 |
const DyeColor &col2 = *it; |
195 |
|
|
|
196 |
|
280 |
__m256i newMask = _mm256_set1_epi32(col2.valueA); |
197 |
|
280 |
__m256i cmpMask = _mm256_set1_epi32(col.valueA); |
198 |
|
140 |
__m256i cmpRes = _mm256_cmpeq_epi32(base, cmpMask); |
199 |
|
140 |
__m256i srcAnd = _mm256_andnot_si256(cmpRes, base); |
200 |
|
140 |
__m256i dstAnd = _mm256_and_si256(cmpRes, newMask); |
201 |
|
140 |
base = _mm256_or_si256(srcAnd, dstAnd); |
202 |
|
|
|
203 |
|
|
++ it; |
204 |
|
|
} |
205 |
|
|
// _mm256_store_si256(reinterpret_cast<__m256i*>(pixels), base); |
206 |
|
268 |
_mm256_storeu_si256(reinterpret_cast<__m256i*>(&pixels[ptr]), base); |
207 |
|
|
} |
208 |
|
|
|
209 |
|
|
// complete end without simd |
210 |
✓✓ |
131 |
for (int ptr = bufSize - mod; ptr < bufSize; ptr ++) |
211 |
|
|
{ |
212 |
|
54 |
uint8_t *const p = reinterpret_cast<uint8_t *>(&pixels[ptr]); |
213 |
|
54 |
const unsigned int data = pixels[ptr]; |
214 |
|
|
|
215 |
|
108 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
216 |
✓✓ |
100 |
while (it != it_end) |
217 |
|
|
{ |
218 |
|
84 |
const DyeColor &col = *it; |
219 |
|
84 |
++ it; |
220 |
|
84 |
const DyeColor &col2 = *it; |
221 |
|
|
|
222 |
|
168 |
const unsigned int coldata = (col.value[3]) | |
223 |
|
168 |
(col.value[2] << 8U) | |
224 |
|
168 |
(col.value[1] << 16U) | |
225 |
|
168 |
(col.value[0] << 24U); |
226 |
|
|
|
227 |
✓✓ |
84 |
if (data == coldata) |
228 |
|
|
{ |
229 |
|
38 |
p[3] = col2.value[0]; |
230 |
|
38 |
p[2] = col2.value[1]; |
231 |
|
38 |
p[1] = col2.value[2]; |
232 |
|
38 |
p[0] = col2.value[3]; |
233 |
|
38 |
break; |
234 |
|
|
} |
235 |
|
|
|
236 |
|
|
++ it; |
237 |
|
|
} |
238 |
|
|
} |
239 |
|
|
} |
240 |
|
|
|
241 |
|
|
#endif // SIMD_SUPPORTED |