1 |
schoenebeck |
53 |
/*************************************************************************** |
2 |
|
|
* * |
3 |
|
|
* LinuxSampler - modular, streaming capable sampler * |
4 |
|
|
* * |
5 |
schoenebeck |
56 |
* Copyright (C) 2003, 2004 by Benno Senoner and Christian Schoenebeck * |
6 |
schoenebeck |
554 |
* Copyright (C) 2005 Christian Schoenebeck * |
7 |
schoenebeck |
53 |
* * |
8 |
|
|
* This program is free software; you can redistribute it and/or modify * |
9 |
|
|
* it under the terms of the GNU General Public License as published by * |
10 |
|
|
* the Free Software Foundation; either version 2 of the License, or * |
11 |
|
|
* (at your option) any later version. * |
12 |
|
|
* * |
13 |
|
|
* This program is distributed in the hope that it will be useful, * |
14 |
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of * |
15 |
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * |
16 |
|
|
* GNU General Public License for more details. * |
17 |
|
|
* * |
18 |
|
|
* You should have received a copy of the GNU General Public License * |
19 |
|
|
* along with this program; if not, write to the Free Software * |
20 |
|
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, * |
21 |
|
|
* MA 02111-1307 USA * |
22 |
|
|
***************************************************************************/ |
23 |
|
|
|
24 |
|
|
#ifndef __RT_MATH_H__ |
25 |
|
|
#define __RT_MATH_H__ |
26 |
|
|
|
27 |
|
|
#include <math.h> |
28 |
schoenebeck |
328 |
#include <stdint.h> |
29 |
schoenebeck |
53 |
#include "global.h" |
30 |
|
|
|
31 |
|
|
/// Needed for calculating frequency ratio used to pitch a sample |
32 |
|
|
#define TWELVEHUNDREDTH_ROOT_OF_TWO 1.000577789506555 |
33 |
|
|
|
34 |
schoenebeck |
319 |
enum implementation_t { |
35 |
schoenebeck |
361 |
CPP, |
36 |
|
|
ASM_X86_MMX_SSE |
37 |
schoenebeck |
319 |
}; |
38 |
|
|
|
39 |
|
|
class RTMathBase { |
40 |
schoenebeck |
53 |
public: |
41 |
|
|
/** |
42 |
schoenebeck |
328 |
* Highly accurate time stamp. |
43 |
|
|
*/ |
44 |
|
|
typedef uint32_t time_stamp_t; |
45 |
|
|
|
46 |
|
|
/** |
47 |
|
|
* We read the processor's cycle count register as a reference |
48 |
|
|
* for the real time. These are of course only abstract values |
49 |
|
|
* with arbitrary time entity, but that's not a problem as long |
50 |
|
|
* as we calculate relatively. |
51 |
|
|
*/ |
52 |
|
|
static time_stamp_t CreateTimeStamp(); |
53 |
|
|
|
54 |
|
|
/** |
55 |
schoenebeck |
53 |
* Calculates the frequency ratio for a pitch value given in cents |
56 |
|
|
* (assuming equal tempered scale of course, divided into 12 |
57 |
|
|
* semitones per octave and 100 cents per semitone). |
58 |
|
|
* |
59 |
schoenebeck |
554 |
* Note: CONFIG_MAX_PITCH (defined in config.h) has to be defined to an |
60 |
schoenebeck |
53 |
* appropriate value, otherwise the behavior of this function is |
61 |
schoenebeck |
554 |
* undefined, but most probably if CONFIG_MAX_PITCH is too small, the |
62 |
schoenebeck |
53 |
* application will crash due to segmentation fault here. |
63 |
|
|
* |
64 |
|
|
* @param cents - pitch value in cents (+1200 cents means +1 octave) |
65 |
|
|
* @returns frequency ratio (e.g. +2.0 for +1 octave) |
66 |
|
|
*/ |
67 |
|
|
inline static double CentsToFreqRatio(double Cents) { |
68 |
schoenebeck |
319 |
int index_int = (int) (Cents); // integer index |
69 |
schoenebeck |
53 |
float index_fract = Cents - index_int; // fractional part of index |
70 |
|
|
return pCentsToFreqTable[index_int] + index_fract * (pCentsToFreqTable[index_int+1] - pCentsToFreqTable[index_int]); |
71 |
|
|
} |
72 |
|
|
|
73 |
schoenebeck |
319 |
private: |
74 |
schoenebeck |
554 |
static float CentsToFreqTable[CONFIG_MAX_PITCH * 1200 * 2 + 1]; |
75 |
schoenebeck |
319 |
static float* pCentsToFreqTable; |
76 |
|
|
|
77 |
|
|
static float* InitCentsToFreqTable(); |
78 |
|
|
}; |
79 |
|
|
|
80 |
|
|
/** Real Time Math |
81 |
|
|
* |
82 |
|
|
* Math functions for real time operation. |
83 |
|
|
*/ |
84 |
|
|
template<implementation_t IMPL = CPP> |
85 |
|
|
class __RTMath : public RTMathBase { |
86 |
|
|
public: |
87 |
|
|
// conversion using truncate |
88 |
|
|
inline static int Int(const float a) { |
89 |
|
|
switch (IMPL) { |
90 |
|
|
case CPP: { |
91 |
|
|
return (int) a; |
92 |
|
|
} |
93 |
schoenebeck |
328 |
#if ARCH_X86 |
94 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
95 |
|
|
int ret; |
96 |
|
|
asm ( |
97 |
|
|
"cvttss2si %1, %0 # convert to int\n\t" |
98 |
|
|
: "=r" (ret) |
99 |
|
|
: "m" (a) |
100 |
|
|
); |
101 |
|
|
return ret; |
102 |
|
|
} |
103 |
schoenebeck |
328 |
#endif // ARCH_X86 |
104 |
schoenebeck |
319 |
} |
105 |
|
|
} |
106 |
|
|
|
107 |
|
|
//for doubles and everything else except floats |
108 |
|
|
template<class T_a> inline static int Int(const T_a a) { |
109 |
|
|
return (int) a; |
110 |
|
|
} |
111 |
|
|
|
112 |
|
|
inline static float Float(const int a) { |
113 |
|
|
switch (IMPL) { |
114 |
|
|
case CPP: { |
115 |
|
|
return (float) a; |
116 |
|
|
} |
117 |
schoenebeck |
328 |
#if ARCH_X86 |
118 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
119 |
|
|
float ret; |
120 |
|
|
asm ( |
121 |
|
|
"cvtsi2ss %1, %%xmm0 # convert to float\n\t" |
122 |
|
|
"movss %%xmm0,%0 # output\n\t" |
123 |
|
|
: "=m" (ret) |
124 |
|
|
: "r" (a) |
125 |
|
|
); |
126 |
|
|
return ret; |
127 |
|
|
} |
128 |
schoenebeck |
328 |
#endif // ARCH_X86 |
129 |
schoenebeck |
319 |
} |
130 |
|
|
} |
131 |
|
|
|
132 |
|
|
#if 0 |
133 |
|
|
//for everything except ints |
134 |
|
|
template<class T_a> inline static float Float(T_a a) { |
135 |
|
|
return (float) a; |
136 |
|
|
} |
137 |
|
|
#endif |
138 |
|
|
|
139 |
|
|
inline static float Sum(const float& a, const float& b) { |
140 |
|
|
switch (IMPL) { |
141 |
|
|
case CPP: { |
142 |
|
|
return (a + b); |
143 |
|
|
} |
144 |
schoenebeck |
328 |
#if ARCH_X86 |
145 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
146 |
|
|
float ret; |
147 |
|
|
asm ( |
148 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
149 |
|
|
"addss %2, %%xmm0 # a + b\n\t" |
150 |
|
|
"movss %%xmm0, %0 # output\n\t" |
151 |
|
|
: "=m" (ret) |
152 |
|
|
: "m" (a), "m" (b) |
153 |
|
|
); |
154 |
|
|
return ret; |
155 |
|
|
} |
156 |
schoenebeck |
328 |
#endif // ARCH_X86 |
157 |
schoenebeck |
319 |
} |
158 |
|
|
} |
159 |
|
|
|
160 |
|
|
template<class T_a, class T_b> inline static T_a Sum(const T_a a, const T_b b) { |
161 |
|
|
return (a + b); |
162 |
|
|
} |
163 |
|
|
|
164 |
|
|
inline static float Sub(const float& a, const float& b) { |
165 |
|
|
switch (IMPL) { |
166 |
|
|
case CPP: { |
167 |
|
|
return (a - b); |
168 |
|
|
} |
169 |
schoenebeck |
328 |
#if ARCH_X86 |
170 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
171 |
|
|
float ret; |
172 |
|
|
asm ( |
173 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
174 |
|
|
"subss %2, %%xmm0 # a - b\n\t" |
175 |
|
|
"movss %%xmm0, %0 # output\n\t" |
176 |
|
|
: "=m" (ret) |
177 |
|
|
: "m" (a), "m" (b) |
178 |
|
|
); |
179 |
|
|
return ret; |
180 |
|
|
} |
181 |
schoenebeck |
328 |
#endif // ARCH_X86 |
182 |
schoenebeck |
319 |
} |
183 |
|
|
} |
184 |
|
|
|
185 |
|
|
template<class T_a, class T_b> inline static T_a Sub(const T_a a, const T_b b) { |
186 |
|
|
return (a - b); |
187 |
|
|
} |
188 |
|
|
|
189 |
|
|
inline static float Mul(const float a, const float b) { |
190 |
|
|
switch (IMPL) { |
191 |
|
|
case CPP: { |
192 |
|
|
return (a * b); |
193 |
|
|
} |
194 |
schoenebeck |
328 |
#if ARCH_X86 |
195 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
196 |
|
|
float ret; |
197 |
|
|
asm ( |
198 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
199 |
|
|
"mulss %2, %%xmm0 # a * b\n\t" |
200 |
|
|
"movss %%xmm0, %0 # output\n\t" |
201 |
|
|
: "=m" (ret) |
202 |
|
|
: "m" (a), "m" (b) |
203 |
|
|
); |
204 |
|
|
return ret; |
205 |
|
|
} |
206 |
schoenebeck |
328 |
#endif // ARCH_X86 |
207 |
schoenebeck |
319 |
} |
208 |
|
|
} |
209 |
|
|
|
210 |
|
|
template<class T_a, class T_b> inline static T_a Mul(const T_a a, const T_b b) { |
211 |
|
|
return (a * b); |
212 |
|
|
} |
213 |
|
|
|
214 |
|
|
inline static float Div(const float a, const float b) { |
215 |
|
|
switch (IMPL) { |
216 |
|
|
case CPP: { |
217 |
|
|
return (a / b); |
218 |
|
|
} |
219 |
schoenebeck |
328 |
#if ARCH_X86 |
220 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
221 |
|
|
float ret; |
222 |
|
|
asm ( |
223 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
224 |
|
|
"divss %2, %%xmm0 # a / b\n\t" |
225 |
|
|
"movss %%xmm0, %0 # output\n\t" |
226 |
|
|
: "=m" (ret) |
227 |
|
|
: "m" (a), "m" (b) |
228 |
|
|
); |
229 |
|
|
return ret; |
230 |
|
|
} |
231 |
schoenebeck |
328 |
#endif // ARCH_X86 |
232 |
schoenebeck |
319 |
} |
233 |
|
|
} |
234 |
|
|
|
235 |
|
|
template<class T_a, class T_b> inline static T_a Div(const T_a a, const T_b b) { |
236 |
|
|
return (a / b); |
237 |
|
|
} |
238 |
|
|
|
239 |
|
|
inline static float Min(const float a, const float b) { |
240 |
|
|
switch (IMPL) { |
241 |
|
|
case CPP: { |
242 |
|
|
return (b < a) ? b : a; |
243 |
|
|
} |
244 |
schoenebeck |
328 |
#if ARCH_X86 |
245 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
246 |
|
|
float ret; |
247 |
|
|
asm ( |
248 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
249 |
|
|
"minss %2, %%xmm0 # Minimum(a, b)\n\t" |
250 |
|
|
"movss %%xmm0, %0 # output\n\t" |
251 |
|
|
: "=m" (ret) |
252 |
|
|
: "m" (a), "m" (b) |
253 |
|
|
); |
254 |
|
|
return ret; |
255 |
|
|
} |
256 |
schoenebeck |
328 |
#endif // ARCH_X86 |
257 |
schoenebeck |
319 |
} |
258 |
|
|
} |
259 |
|
|
|
260 |
|
|
template<class T_a, class T_b> inline static T_a Min(const T_a a, const T_b b) { |
261 |
schoenebeck |
53 |
return (b < a) ? b : a; |
262 |
|
|
} |
263 |
|
|
|
264 |
schoenebeck |
319 |
inline static float Max(const float a, const float b) { |
265 |
|
|
switch (IMPL) { |
266 |
|
|
case CPP: { |
267 |
|
|
return (b > a) ? b : a; |
268 |
|
|
} |
269 |
schoenebeck |
328 |
#if ARCH_X86 |
270 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
271 |
|
|
float ret; |
272 |
|
|
asm ( |
273 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
274 |
|
|
"maxss %2, %%xmm0 # Maximum(a, b)\n\t" |
275 |
|
|
"movss %%xmm0, %0 # output\n\t" |
276 |
|
|
: "=m" (ret) |
277 |
|
|
: "m" (a), "m" (b) |
278 |
|
|
); |
279 |
|
|
return ret; |
280 |
|
|
} |
281 |
schoenebeck |
328 |
#endif // ARCH_X86 |
282 |
schoenebeck |
319 |
} |
283 |
|
|
} |
284 |
|
|
|
285 |
|
|
template<class T_a, class T_b> inline static T_a Max(const T_a a, const T_b b) { |
286 |
schoenebeck |
53 |
return (b > a) ? b : a; |
287 |
|
|
} |
288 |
|
|
|
289 |
schoenebeck |
319 |
inline static float Fmodf(const float &a, const float &b) { |
290 |
|
|
switch (IMPL) { |
291 |
|
|
case CPP: { |
292 |
|
|
return fmodf(a, b); |
293 |
|
|
} |
294 |
schoenebeck |
328 |
#if ARCH_X86 |
295 |
schoenebeck |
319 |
case ASM_X86_MMX_SSE: { |
296 |
|
|
float ret; |
297 |
|
|
asm ( |
298 |
|
|
"movss %1, %%xmm0 # load a\n\t" |
299 |
|
|
"movss %2, %%xmm1 # load b\n\t" |
300 |
|
|
"movss %%xmm0,%%xmm2\n\t" |
301 |
|
|
"divss %%xmm1, %%xmm2 # xmm2 = a / b\n\t" |
302 |
|
|
"cvttss2si %%xmm2, %%ecx #convert to int\n\t" |
303 |
|
|
"cvtsi2ss %%ecx, %%xmm2 #convert back to float\n\t" |
304 |
|
|
"mulss %%xmm1, %%xmm2 # xmm2 = b * int(a/b)\n\t" |
305 |
|
|
"subss %%xmm2, %%xmm0 #sub a\n\t" |
306 |
|
|
"movss %%xmm0, %0 # output\n\t" |
307 |
|
|
: "=m" (ret) |
308 |
|
|
: "m" (a), "m" (b) |
309 |
|
|
: "%ecx" |
310 |
|
|
); |
311 |
|
|
return ret; |
312 |
|
|
} |
313 |
schoenebeck |
328 |
#endif // ARCH_X86 |
314 |
schoenebeck |
319 |
} |
315 |
|
|
} |
316 |
schoenebeck |
53 |
}; |
317 |
|
|
|
318 |
schoenebeck |
319 |
/// convenience typedef for using the default implementation (which is CPP) |
319 |
|
|
typedef __RTMath<> RTMath; |
320 |
|
|
|
321 |
schoenebeck |
53 |
#endif // __RT_MATH_H__ |