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