1 | // Copyright 2016 The SwiftShader Authors. All Rights Reserved. |
2 | // |
3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
4 | // you may not use this file except in compliance with the License. |
5 | // You may obtain a copy of the License at |
6 | // |
7 | // http://www.apache.org/licenses/LICENSE-2.0 |
8 | // |
9 | // Unless required by applicable law or agreed to in writing, software |
10 | // distributed under the License is distributed on an "AS IS" BASIS, |
11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
12 | // See the License for the specific language governing permissions and |
13 | // limitations under the License. |
14 | |
15 | #ifndef sw_Math_hpp |
16 | #define sw_Math_hpp |
17 | |
18 | #include "Types.hpp" |
19 | |
20 | #include "Vulkan/VkDebug.hpp" |
21 | |
22 | #include <cmath> |
23 | #if defined(_MSC_VER) |
24 | #include <intrin.h> |
25 | #endif |
26 | |
27 | namespace sw |
28 | { |
29 | using std::abs; |
30 | |
31 | #undef min |
32 | #undef max |
33 | |
34 | template<class T> |
35 | inline T constexpr max(T a, T b) |
36 | { |
37 | return a > b ? a : b; |
38 | } |
39 | |
40 | template<class T> |
41 | inline constexpr T min(T a, T b) |
42 | { |
43 | return a < b ? a : b; |
44 | } |
45 | |
46 | template<class T> |
47 | inline constexpr T max(T a, T b, T c) |
48 | { |
49 | return max(max(a, b), c); |
50 | } |
51 | |
52 | template<class T> |
53 | inline constexpr T min(T a, T b, T c) |
54 | { |
55 | return min(min(a, b), c); |
56 | } |
57 | |
58 | template<class T> |
59 | inline constexpr T max(T a, T b, T c, T d) |
60 | { |
61 | return max(max(a, b), max(c, d)); |
62 | } |
63 | |
64 | template<class T> |
65 | inline constexpr T min(T a, T b, T c, T d) |
66 | { |
67 | return min(min(a, b), min(c, d)); |
68 | } |
69 | |
70 | template <typename destType, typename sourceType> |
71 | destType bit_cast(const sourceType &source) |
72 | { |
73 | union |
74 | { |
75 | sourceType s; |
76 | destType d; |
77 | } sd; |
78 | sd.s = source; |
79 | return sd.d; |
80 | } |
81 | |
82 | inline int iround(float x) |
83 | { |
84 | return (int)floor(x + 0.5f); |
85 | // return _mm_cvtss_si32(_mm_load_ss(&x)); // FIXME: Demands SSE support |
86 | } |
87 | |
88 | inline int ifloor(float x) |
89 | { |
90 | return (int)floor(x); |
91 | } |
92 | |
93 | inline int ceilFix4(int x) |
94 | { |
95 | return (x + 0xF) & 0xFFFFFFF0; |
96 | } |
97 | |
98 | inline int ceilInt4(int x) |
99 | { |
100 | return (x + 0xF) >> 4; |
101 | } |
102 | |
103 | #define BITS(x) ( \ |
104 | !!((x) & 0x80000000) + \ |
105 | !!((x) & 0xC0000000) + \ |
106 | !!((x) & 0xE0000000) + \ |
107 | !!((x) & 0xF0000000) + \ |
108 | !!((x) & 0xF8000000) + \ |
109 | !!((x) & 0xFC000000) + \ |
110 | !!((x) & 0xFE000000) + \ |
111 | !!((x) & 0xFF000000) + \ |
112 | !!((x) & 0xFF800000) + \ |
113 | !!((x) & 0xFFC00000) + \ |
114 | !!((x) & 0xFFE00000) + \ |
115 | !!((x) & 0xFFF00000) + \ |
116 | !!((x) & 0xFFF80000) + \ |
117 | !!((x) & 0xFFFC0000) + \ |
118 | !!((x) & 0xFFFE0000) + \ |
119 | !!((x) & 0xFFFF0000) + \ |
120 | !!((x) & 0xFFFF8000) + \ |
121 | !!((x) & 0xFFFFC000) + \ |
122 | !!((x) & 0xFFFFE000) + \ |
123 | !!((x) & 0xFFFFF000) + \ |
124 | !!((x) & 0xFFFFF800) + \ |
125 | !!((x) & 0xFFFFFC00) + \ |
126 | !!((x) & 0xFFFFFE00) + \ |
127 | !!((x) & 0xFFFFFF00) + \ |
128 | !!((x) & 0xFFFFFF80) + \ |
129 | !!((x) & 0xFFFFFFC0) + \ |
130 | !!((x) & 0xFFFFFFE0) + \ |
131 | !!((x) & 0xFFFFFFF0) + \ |
132 | !!((x) & 0xFFFFFFF8) + \ |
133 | !!((x) & 0xFFFFFFFC) + \ |
134 | !!((x) & 0xFFFFFFFE) + \ |
135 | !!((x) & 0xFFFFFFFF)) |
136 | |
137 | #define MAX(x, y) ((x) > (y) ? (x) : (y)) |
138 | #define MIN(x, y) ((x) < (y) ? (x) : (y)) |
139 | |
140 | inline unsigned long log2i(int x) |
141 | { |
142 | #if defined(_MSC_VER) |
143 | unsigned long y; |
144 | _BitScanReverse(&y, x); |
145 | return y; |
146 | #else |
147 | return 31 - __builtin_clz(x); |
148 | #endif |
149 | } |
150 | |
151 | inline bool isPow2(int x) |
152 | { |
153 | return (x & -x) == x; |
154 | } |
155 | |
156 | template<class T> |
157 | inline T clamp(T x, T a, T b) |
158 | { |
159 | ASSERT(a <= b); |
160 | if(x < a) x = a; |
161 | if(x > b) x = b; |
162 | |
163 | return x; |
164 | } |
165 | |
166 | inline float clamp01(float x) |
167 | { |
168 | return clamp(x, 0.0f, 1.0f); |
169 | } |
170 | |
171 | // Bit-cast of a floating-point value into a two's complement integer representation. |
172 | // This makes floating-point values comparable as integers. |
173 | inline int32_t float_as_twos_complement(float f) |
174 | { |
175 | // IEEE-754 floating-point numbers are sorted by magnitude in the same way as integers, |
176 | // except negative values are like one's complement integers. Convert them to two's complement. |
177 | int32_t i = bit_cast<int32_t>(f); |
178 | return (i < 0) ? (0x7FFFFFFFu - i) : i; |
179 | } |
180 | |
181 | // 'Safe' clamping operation which always returns a value between min and max (inclusive). |
182 | inline float clamp_s(float x, float min, float max) |
183 | { |
184 | // NaN values can't be compared directly |
185 | if(float_as_twos_complement(x) < float_as_twos_complement(min)) x = min; |
186 | if(float_as_twos_complement(x) > float_as_twos_complement(max)) x = max; |
187 | |
188 | return x; |
189 | } |
190 | |
191 | inline int ceilPow2(int x) |
192 | { |
193 | int i = 1; |
194 | |
195 | while(i < x) |
196 | { |
197 | i <<= 1; |
198 | } |
199 | |
200 | return i; |
201 | } |
202 | |
203 | inline int floorDiv(int a, int b) |
204 | { |
205 | return a / b + ((a % b) >> 31); |
206 | } |
207 | |
208 | inline int floorMod(int a, int b) |
209 | { |
210 | int r = a % b; |
211 | return r + ((r >> 31) & b); |
212 | } |
213 | |
214 | inline int ceilDiv(int a, int b) |
215 | { |
216 | return a / b - (-(a % b) >> 31); |
217 | } |
218 | |
219 | inline int ceilMod(int a, int b) |
220 | { |
221 | int r = a % b; |
222 | return r - ((-r >> 31) & b); |
223 | } |
224 | |
225 | template<const int n> |
226 | inline unsigned int unorm(float x) |
227 | { |
228 | static const unsigned int max = 0xFFFFFFFF >> (32 - n); |
229 | static const float maxf = static_cast<float>(max); |
230 | |
231 | if(x >= 1.0f) |
232 | { |
233 | return max; |
234 | } |
235 | else if(x <= 0.0f) |
236 | { |
237 | return 0; |
238 | } |
239 | else |
240 | { |
241 | return static_cast<unsigned int>(maxf * x + 0.5f); |
242 | } |
243 | } |
244 | |
245 | template<const int n> |
246 | inline int snorm(float x) |
247 | { |
248 | static const unsigned int min = 0x80000000 >> (32 - n); |
249 | static const unsigned int max = 0xFFFFFFFF >> (32 - n + 1); |
250 | static const float maxf = static_cast<float>(max); |
251 | static const unsigned int range = 0xFFFFFFFF >> (32 - n); |
252 | |
253 | if(x >= 0.0f) |
254 | { |
255 | if(x >= 1.0f) |
256 | { |
257 | return max; |
258 | } |
259 | else |
260 | { |
261 | return static_cast<int>(maxf * x + 0.5f); |
262 | } |
263 | } |
264 | else |
265 | { |
266 | if(x <= -1.0f) |
267 | { |
268 | return min; |
269 | } |
270 | else |
271 | { |
272 | return static_cast<int>(maxf * x - 0.5f) & range; |
273 | } |
274 | } |
275 | } |
276 | |
277 | template<const int n> |
278 | inline unsigned int ucast(float x) |
279 | { |
280 | static const unsigned int max = 0xFFFFFFFF >> (32 - n); |
281 | static const float maxf = static_cast<float>(max); |
282 | |
283 | if(x >= maxf) |
284 | { |
285 | return max; |
286 | } |
287 | else if(x <= 0.0f) |
288 | { |
289 | return 0; |
290 | } |
291 | else |
292 | { |
293 | return static_cast<unsigned int>(x + 0.5f); |
294 | } |
295 | } |
296 | |
297 | template<const int n> |
298 | inline int scast(float x) |
299 | { |
300 | static const unsigned int min = 0x80000000 >> (32 - n); |
301 | static const unsigned int max = 0xFFFFFFFF >> (32 - n + 1); |
302 | static const float maxf = static_cast<float>(max); |
303 | static const float minf = static_cast<float>(min); |
304 | static const unsigned int range = 0xFFFFFFFF >> (32 - n); |
305 | |
306 | if(x > 0.0f) |
307 | { |
308 | if(x >= maxf) |
309 | { |
310 | return max; |
311 | } |
312 | else |
313 | { |
314 | return static_cast<int>(x + 0.5f); |
315 | } |
316 | } |
317 | else |
318 | { |
319 | if(x <= -minf) |
320 | { |
321 | return min; |
322 | } |
323 | else |
324 | { |
325 | return static_cast<int>(x - 0.5f) & range; |
326 | } |
327 | } |
328 | } |
329 | |
330 | inline float sRGBtoLinear(float c) |
331 | { |
332 | if(c <= 0.04045f) |
333 | { |
334 | return c * 0.07739938f; // 1.0f / 12.92f; |
335 | } |
336 | else |
337 | { |
338 | return powf((c + 0.055f) * 0.9478673f, 2.4f); // 1.0f / 1.055f |
339 | } |
340 | } |
341 | |
342 | inline float linearToSRGB(float c) |
343 | { |
344 | if(c <= 0.0031308f) |
345 | { |
346 | return c * 12.92f; |
347 | } |
348 | else |
349 | { |
350 | return 1.055f * powf(c, 0.4166667f) - 0.055f; // 1.0f / 2.4f |
351 | } |
352 | } |
353 | |
354 | unsigned char sRGB8toLinear8(unsigned char value); |
355 | |
356 | uint64_t FNV_1a(const unsigned char *data, int size); // Fowler-Noll-Vo hash function |
357 | |
358 | // Round up to the next multiple of alignment |
359 | template<typename T> |
360 | inline T align(T value, unsigned int alignment) |
361 | { |
362 | return ((value + alignment - 1) / alignment) * alignment; |
363 | } |
364 | |
365 | template<unsigned int alignment, typename T> |
366 | inline T align(T value) |
367 | { |
368 | return ((value + alignment - 1) / alignment) * alignment; |
369 | } |
370 | |
371 | inline int clampToSignedInt(unsigned int x) |
372 | { |
373 | return static_cast<int>(min(x, 0x7FFFFFFFu)); |
374 | } |
375 | |
376 | // Convert floating value v to fixed point with p digits after the decimal point |
377 | constexpr int toFixedPoint(float v, int p) { |
378 | return static_cast<int>(v * (1 << p)); |
379 | } |
380 | } |
381 | |
382 | #endif // sw_Math_hpp |
383 | |