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::replaceSColorDefault(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 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
56 |
|
|
const unsigned int data = (*pixels) & 0x00ffffffU; |
57 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
58 |
|
|
|
59 |
|
51 |
const unsigned int data = (*pixels) & 0xffffff00U; |
60 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
61 |
|
|
|
62 |
|
102 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
63 |
✓✓ |
110 |
while (it != it_end) |
64 |
|
|
{ |
65 |
|
99 |
const DyeColor &col = *it; |
66 |
|
99 |
++ it; |
67 |
|
99 |
const DyeColor &col2 = *it; |
68 |
|
|
|
69 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
70 |
|
|
const unsigned int coldata = (col.value[2] << 16U) |
71 |
|
|
| (col.value[1] << 8U) | (col.value[0]); |
72 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
73 |
|
|
|
74 |
|
99 |
const unsigned int coldata = (col.value[2] << 8U) |
75 |
|
99 |
| (col.value[1] << 16U) | (col.value[0] << 24U); |
76 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
77 |
|
|
|
78 |
✓✓ |
99 |
if (data == coldata) |
79 |
|
|
{ |
80 |
|
40 |
p[3] = col2.value[0]; |
81 |
|
40 |
p[2] = col2.value[1]; |
82 |
|
40 |
p[1] = col2.value[2]; |
83 |
|
40 |
break; |
84 |
|
|
} |
85 |
|
|
|
86 |
|
|
++ it; |
87 |
|
|
} |
88 |
|
|
} |
89 |
|
|
} |
90 |
|
|
|
91 |
|
|
#ifdef SIMD_SUPPORTED |
92 |
|
|
/* |
93 |
|
|
static void print256(const char *const text, const __m256i &val); |
94 |
|
|
static void print256(const char *const text, const __m256i &val) |
95 |
|
|
{ |
96 |
|
|
printf("%s 0x%016llx%016llx%016llx%016llx\n", text, val[0], val[1], val[2], val[3]); |
97 |
|
|
} |
98 |
|
|
*/ |
99 |
|
|
|
100 |
|
|
__attribute__ ((target ("sse2"))) |
101 |
|
11 |
void DyePalette::replaceSColorSse2(uint32_t *restrict pixels, |
102 |
|
|
const int bufSize) const restrict2 |
103 |
|
|
{ |
104 |
|
22 |
STD_VECTOR<DyeColor>::const_iterator it_end = mColors.end(); |
105 |
|
22 |
const size_t sz = mColors.size(); |
106 |
✓✗ |
11 |
if (sz == 0U || pixels == nullptr) |
107 |
|
|
return; |
108 |
✗✓ |
11 |
if ((sz % 2) != 0U) |
109 |
|
|
-- it_end; |
110 |
|
11 |
const int mod = bufSize % 8; |
111 |
|
11 |
const int bufEnd = bufSize - mod; |
112 |
|
|
|
113 |
✓✓ |
17 |
for (int ptr = 0; ptr < bufEnd; ptr += 4) |
114 |
|
|
{ |
115 |
|
6 |
__m128i mask = _mm_set1_epi32(0xffffff00U); |
116 |
|
|
// __m128i base = _mm_load_si128(reinterpret_cast<__m128i*>(pixels)); |
117 |
|
|
__m128i base = _mm_loadu_si128(reinterpret_cast<__m128i*>( |
118 |
|
12 |
&pixels[ptr])); |
119 |
|
|
|
120 |
|
12 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
121 |
✓✓ |
18 |
while (it != it_end) |
122 |
|
|
{ |
123 |
|
12 |
const DyeColor &col = *it; |
124 |
|
12 |
++ it; |
125 |
|
12 |
const DyeColor &col2 = *it; |
126 |
|
|
|
127 |
|
12 |
__m128i base2 = _mm_and_si128(mask, base); |
128 |
|
24 |
__m128i newMask = _mm_set1_epi32(col2.valueS); |
129 |
|
24 |
__m128i cmpMask = _mm_set1_epi32(col.valueS); |
130 |
|
12 |
__m128i cmpRes = _mm_cmpeq_epi32(base2, cmpMask); |
131 |
|
12 |
cmpRes = _mm_and_si128(mask, cmpRes); |
132 |
|
12 |
__m128i srcAnd = _mm_andnot_si128(cmpRes, base); |
133 |
|
12 |
__m128i dstAnd = _mm_and_si128(cmpRes, newMask); |
134 |
|
12 |
base = _mm_or_si128(srcAnd, dstAnd); |
135 |
|
|
++ it; |
136 |
|
|
} |
137 |
|
|
// _mm_store_si128(reinterpret_cast<__m128i*>(pixels), base); |
138 |
|
12 |
_mm_storeu_si128(reinterpret_cast<__m128i*>(&pixels[ptr]), base); |
139 |
|
|
} |
140 |
|
|
|
141 |
|
|
// complete end without simd |
142 |
✓✓ |
65 |
for (int ptr = bufSize - mod; ptr < bufSize; ptr ++) |
143 |
|
|
{ |
144 |
|
27 |
uint8_t *const p = reinterpret_cast<uint8_t *>(&pixels[ptr]); |
145 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
146 |
|
|
const unsigned int data = pixels[ptr] & 0x00ffffffU; |
147 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
148 |
|
|
|
149 |
|
27 |
const unsigned int data = pixels[ptr] & 0xffffff00U; |
150 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
151 |
|
|
|
152 |
|
54 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
153 |
✓✓ |
56 |
while (it != it_end) |
154 |
|
|
{ |
155 |
|
51 |
const DyeColor &col = *it; |
156 |
|
51 |
++ it; |
157 |
|
51 |
const DyeColor &col2 = *it; |
158 |
|
|
|
159 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
160 |
|
|
const unsigned int coldata = (col.value[2] << 16U) |
161 |
|
|
| (col.value[1] << 8U) | (col.value[0]); |
162 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
163 |
|
|
|
164 |
|
51 |
const unsigned int coldata = (col.value[2] << 8U) |
165 |
|
51 |
| (col.value[1] << 16U) | (col.value[0] << 24U); |
166 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
167 |
|
|
|
168 |
✓✓ |
51 |
if (data == coldata) |
169 |
|
|
{ |
170 |
|
22 |
p[3] = col2.value[0]; |
171 |
|
22 |
p[2] = col2.value[1]; |
172 |
|
22 |
p[1] = col2.value[2]; |
173 |
|
22 |
break; |
174 |
|
|
} |
175 |
|
|
|
176 |
|
|
++ it; |
177 |
|
|
} |
178 |
|
|
} |
179 |
|
|
} |
180 |
|
|
|
181 |
|
|
__attribute__ ((target ("avx2"))) |
182 |
|
25 |
void DyePalette::replaceSColorAvx2(uint32_t *restrict pixels, |
183 |
|
|
const int bufSize) const restrict2 |
184 |
|
|
{ |
185 |
|
50 |
STD_VECTOR<DyeColor>::const_iterator it_end = mColors.end(); |
186 |
|
50 |
const size_t sz = mColors.size(); |
187 |
✓✗ |
25 |
if (sz == 0U || pixels == nullptr) |
188 |
|
|
return; |
189 |
✗✓ |
25 |
if ((sz % 2) != 0U) |
190 |
|
|
-- it_end; |
191 |
|
25 |
const int mod = bufSize % 8; |
192 |
|
25 |
const int bufEnd = bufSize - mod; |
193 |
|
|
|
194 |
✓✓ |
415 |
for (int ptr = 0; ptr < bufEnd; ptr += 8) |
195 |
|
|
{ |
196 |
|
390 |
__m256i mask = _mm256_set1_epi32(0xffffff00U); |
197 |
|
|
// __m256i base = _mm256_load_si256(reinterpret_cast<__m256i*>(pixels)); |
198 |
|
|
__m256i base = _mm256_loadu_si256(reinterpret_cast<__m256i*>( |
199 |
|
780 |
&pixels[ptr])); |
200 |
|
|
|
201 |
|
780 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
202 |
✓✓ |
2066 |
while (it != it_end) |
203 |
|
|
{ |
204 |
|
1676 |
const DyeColor &col = *it; |
205 |
|
1676 |
++ it; |
206 |
|
1676 |
const DyeColor &col2 = *it; |
207 |
|
|
|
208 |
|
1676 |
__m256i base2 = _mm256_and_si256(mask, base); |
209 |
|
3352 |
__m256i newMask = _mm256_set1_epi32(col2.valueS); |
210 |
|
3352 |
__m256i cmpMask = _mm256_set1_epi32(col.valueS); |
211 |
|
1676 |
__m256i cmpRes = _mm256_cmpeq_epi32(base2, cmpMask); |
212 |
|
1676 |
cmpRes = _mm256_and_si256(mask, cmpRes); |
213 |
|
1676 |
__m256i srcAnd = _mm256_andnot_si256(cmpRes, base); |
214 |
|
1676 |
__m256i dstAnd = _mm256_and_si256(cmpRes, newMask); |
215 |
|
1676 |
base = _mm256_or_si256(srcAnd, dstAnd); |
216 |
|
|
++ it; |
217 |
|
|
} |
218 |
|
|
// _mm256_store_si256(reinterpret_cast<__m256i*>(pixels), base); |
219 |
|
780 |
_mm256_storeu_si256(reinterpret_cast<__m256i*>(&pixels[ptr]), base); |
220 |
|
|
} |
221 |
|
|
|
222 |
|
|
// complete end without simd |
223 |
✓✓ |
133 |
for (int ptr = bufSize - mod; ptr < bufSize; ptr ++) |
224 |
|
|
{ |
225 |
|
54 |
uint8_t *const p = reinterpret_cast<uint8_t *>(&pixels[ptr]); |
226 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
227 |
|
|
const unsigned int data = pixels[ptr] & 0x00ffffffU; |
228 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
229 |
|
|
|
230 |
|
54 |
const unsigned int data = pixels[ptr] & 0xffffff00U; |
231 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
232 |
|
|
|
233 |
|
108 |
STD_VECTOR<DyeColor>::const_iterator it = mColors.begin(); |
234 |
✓✓ |
112 |
while (it != it_end) |
235 |
|
|
{ |
236 |
|
102 |
const DyeColor &col = *it; |
237 |
|
102 |
++ it; |
238 |
|
102 |
const DyeColor &col2 = *it; |
239 |
|
|
|
240 |
|
|
#if SDL_BYTEORDER == SDL_BIG_ENDIAN |
241 |
|
|
const unsigned int coldata = (col.value[2] << 16U) |
242 |
|
|
| (col.value[1] << 8U) | (col.value[0]); |
243 |
|
|
#else // SDL_BYTEORDER == SDL_BIG_ENDIAN |
244 |
|
|
|
245 |
|
102 |
const unsigned int coldata = (col.value[2] << 8U) |
246 |
|
102 |
| (col.value[1] << 16U) | (col.value[0] << 24U); |
247 |
|
|
#endif // SDL_BYTEORDER == SDL_BIG_ENDIAN |
248 |
|
|
|
249 |
✓✓ |
102 |
if (data == coldata) |
250 |
|
|
{ |
251 |
|
44 |
p[3] = col2.value[0]; |
252 |
|
44 |
p[2] = col2.value[1]; |
253 |
|
44 |
p[1] = col2.value[2]; |
254 |
|
44 |
break; |
255 |
|
|
} |
256 |
|
|
|
257 |
|
|
++ it; |
258 |
|
|
} |
259 |
|
|
} |
260 |
|
|
} |
261 |
|
|
|
262 |
|
|
#endif // SIMD_SUPPORTED |