1/*
2 * Copyright (C) 2011-2012 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#if !defined(RS_SERVER) && !defined(RS_COMPATIBILITY_LIB)
18#include <cutils/compiler.h>
19#endif
20
21#include "rsContext.h"
22#include "rsScriptC.h"
23#include "rsMatrix4x4.h"
24#include "rsMatrix3x3.h"
25#include "rsMatrix2x2.h"
26
27#include "rsCpuCore.h"
28#include "rsCpuScript.h"
29
30using namespace android;
31using namespace android::renderscript;
32
33#define EXPORT_F32_FN_F32(func)                                 \
34    float __attribute__((overloadable)) SC_##func(float v) {    \
35        return func(v);                                         \
36    }
37
38#define EXPORT_F32_FN_F32_F32(func)                                     \
39    float __attribute__((overloadable)) SC_##func(float t, float v) {   \
40        return func(t, v);                                              \
41    }
42
43//////////////////////////////////////////////////////////////////////////////
44// Float util
45//////////////////////////////////////////////////////////////////////////////
46
47// Handle missing Gingerbread functions like tgammaf.
48float SC_tgammaf(float x) {
49#ifdef RS_COMPATIBILITY_LIB
50    return tgamma(x);
51#else
52    return tgammaf(x);
53#endif
54}
55
56uint32_t SC_abs_i32(int32_t v) {return abs(v);}
57
58static void SC_MatrixLoadRotate(Matrix4x4 *m, float rot, float x, float y, float z) {
59    m->loadRotate(rot, x, y, z);
60}
61static void SC_MatrixLoadScale(Matrix4x4 *m, float x, float y, float z) {
62    m->loadScale(x, y, z);
63}
64static void SC_MatrixLoadTranslate(Matrix4x4 *m, float x, float y, float z) {
65    m->loadTranslate(x, y, z);
66}
67static void SC_MatrixRotate(Matrix4x4 *m, float rot, float x, float y, float z) {
68    m->rotate(rot, x, y, z);
69}
70static void SC_MatrixScale(Matrix4x4 *m, float x, float y, float z) {
71    m->scale(x, y, z);
72}
73static void SC_MatrixTranslate(Matrix4x4 *m, float x, float y, float z) {
74    m->translate(x, y, z);
75}
76
77static void SC_MatrixLoadOrtho(Matrix4x4 *m, float l, float r, float b, float t, float n, float f) {
78    m->loadOrtho(l, r, b, t, n, f);
79}
80static void SC_MatrixLoadFrustum(Matrix4x4 *m, float l, float r, float b, float t, float n, float f) {
81    m->loadFrustum(l, r, b, t, n, f);
82}
83static void SC_MatrixLoadPerspective(Matrix4x4 *m, float fovy, float aspect, float near, float far) {
84    m->loadPerspective(fovy, aspect, near, far);
85}
86
87static bool SC_MatrixInverse_4x4(Matrix4x4 *m) {
88    return m->inverse();
89}
90static bool SC_MatrixInverseTranspose_4x4(Matrix4x4 *m) {
91    return m->inverseTranspose();
92}
93static void SC_MatrixTranspose_4x4(Matrix4x4 *m) {
94    m->transpose();
95}
96static void SC_MatrixTranspose_3x3(Matrix3x3 *m) {
97    m->transpose();
98}
99static void SC_MatrixTranspose_2x2(Matrix2x2 *m) {
100    m->transpose();
101}
102
103float SC_randf2(float min, float max) {
104    float r = (float)rand();
105    r /= RAND_MAX;
106    r = r * (max - min) + min;
107    return r;
108}
109
110static float SC_frac(float v) {
111    int i = (int)floor(v);
112    return fmin(v - i, 0x1.fffffep-1f);
113}
114
115#ifdef RS_COMPATIBILITY_LIB
116EXPORT_F32_FN_F32(acosf)
117EXPORT_F32_FN_F32(acoshf)
118EXPORT_F32_FN_F32(asinf)
119EXPORT_F32_FN_F32(asinhf)
120EXPORT_F32_FN_F32(atanf)
121EXPORT_F32_FN_F32_F32(atan2f)
122EXPORT_F32_FN_F32(atanhf)
123EXPORT_F32_FN_F32(cbrtf)
124EXPORT_F32_FN_F32(ceilf)
125EXPORT_F32_FN_F32_F32(copysignf)
126EXPORT_F32_FN_F32(cosf)
127EXPORT_F32_FN_F32(coshf)
128EXPORT_F32_FN_F32(erfcf)
129EXPORT_F32_FN_F32(erff)
130EXPORT_F32_FN_F32(expf)
131EXPORT_F32_FN_F32(exp2f)
132EXPORT_F32_FN_F32(expm1f)
133EXPORT_F32_FN_F32_F32(fdimf)
134EXPORT_F32_FN_F32(floorf)
135float SC_fmaf(float u, float t, float v) {return fmaf(u, t, v);}
136EXPORT_F32_FN_F32_F32(fmaxf)
137EXPORT_F32_FN_F32_F32(fminf)
138EXPORT_F32_FN_F32_F32(fmodf)
139float SC_frexpf(float v, int* ptr) {return frexpf(v, ptr);}
140EXPORT_F32_FN_F32_F32(hypotf)
141EXPORT_F32_FN_F32(ilogbf)
142float SC_ldexpf(float v, int i) {return ldexpf(v, i);}
143EXPORT_F32_FN_F32(lgammaf)
144float SC_lgammaf_r(float v, int* ptr) {return lgammaf_r(v, ptr);}
145EXPORT_F32_FN_F32(logf)
146EXPORT_F32_FN_F32(log10f)
147EXPORT_F32_FN_F32(log1pf)
148EXPORT_F32_FN_F32(logbf)
149float SC_modff(float v, float* ptr) {return modff(v, ptr);}
150EXPORT_F32_FN_F32_F32(nextafterf)
151EXPORT_F32_FN_F32_F32(powf)
152EXPORT_F32_FN_F32_F32(remainderf)
153float SC_remquof(float t, float v, int* ptr) {return remquof(t, v, ptr);}
154EXPORT_F32_FN_F32(rintf)
155EXPORT_F32_FN_F32(roundf)
156EXPORT_F32_FN_F32(sinf)
157EXPORT_F32_FN_F32(sinhf)
158EXPORT_F32_FN_F32(sqrtf)
159EXPORT_F32_FN_F32(tanf)
160EXPORT_F32_FN_F32(tanhf)
161EXPORT_F32_FN_F32(truncf)
162float __attribute__((overloadable)) rsFrac(float f) {
163    return SC_frac(f);
164}
165void __attribute__((overloadable)) rsMatrixLoadRotate(rs_matrix4x4 *m,
166        float rot, float x, float y, float z) {
167    SC_MatrixLoadRotate((Matrix4x4 *) m, rot, x, y, z);
168}
169void __attribute__((overloadable)) rsMatrixLoadScale(rs_matrix4x4 *m,
170        float x, float y, float z) {
171    SC_MatrixLoadScale((Matrix4x4 *) m, x, y, z);
172}
173void __attribute__((overloadable)) rsMatrixLoadTranslate(rs_matrix4x4 *m,
174        float x, float y, float z) {
175    SC_MatrixLoadTranslate((Matrix4x4 *) m, x, y, z);
176}
177void __attribute__((overloadable)) rsMatrixRotate(rs_matrix4x4 *m, float rot,
178        float x, float y, float z) {
179    SC_MatrixRotate((Matrix4x4 *) m, rot, x, y, z);
180}
181void __attribute__((overloadable)) rsMatrixScale(rs_matrix4x4 *m, float x,
182        float y, float z) {
183    SC_MatrixScale((Matrix4x4 *) m, x, y, z);
184}
185void __attribute__((overloadable)) rsMatrixTranslate(rs_matrix4x4 *m, float x,
186        float y, float z) {
187    SC_MatrixTranslate((Matrix4x4 *) m, x, y, z);
188}
189void __attribute__((overloadable)) rsMatrixLoadOrtho(rs_matrix4x4 *m, float l,
190        float r, float b, float t, float n, float f) {
191    SC_MatrixLoadOrtho((Matrix4x4 *) m, l, r, b, t, n, f);
192}
193void __attribute__((overloadable)) rsMatrixLoadFrustum(rs_matrix4x4 *m,
194        float l, float r, float b, float t, float n, float f) {
195    SC_MatrixLoadFrustum((Matrix4x4 *) m, l, r, b, t, n, f);
196}
197void __attribute__((overloadable)) rsMatrixLoadPerspective(rs_matrix4x4 *m,
198        float fovy, float aspect, float near, float far) {
199    SC_MatrixLoadPerspective((Matrix4x4 *) m, fovy, aspect, near, far);
200}
201bool __attribute__((overloadable)) rsMatrixInverse(rs_matrix4x4 *m) {
202    return SC_MatrixInverse_4x4((Matrix4x4 *) m);
203}
204bool __attribute__((overloadable)) rsMatrixInverseTranspose(rs_matrix4x4 *m) {
205    return SC_MatrixInverseTranspose_4x4((Matrix4x4 *) m);
206}
207void __attribute__((overloadable)) rsMatrixTranspose(rs_matrix4x4 *m) {
208    SC_MatrixTranspose_4x4((Matrix4x4 *) m);
209}
210void __attribute__((overloadable)) rsMatrixTranspose(rs_matrix3x3 *m) {
211    SC_MatrixTranspose_3x3((Matrix3x3 *) m);
212}
213void __attribute__((overloadable)) rsMatrixTranspose(rs_matrix2x2 *m) {
214    SC_MatrixTranspose_2x2((Matrix2x2 *) m);
215}
216#endif
217
218//////////////////////////////////////////////////////////////////////////////
219// Class implementation
220//////////////////////////////////////////////////////////////////////////////
221
222// llvm name mangling ref
223//  <builtin-type> ::= v  # void
224//                 ::= b  # bool
225//                 ::= c  # char
226//                 ::= a  # signed char
227//                 ::= h  # unsigned char
228//                 ::= s  # short
229//                 ::= t  # unsigned short
230//                 ::= i  # int
231//                 ::= j  # unsigned int
232//                 ::= l  # long
233//                 ::= m  # unsigned long
234//                 ::= x  # long long, __int64
235//                 ::= y  # unsigned long long, __int64
236//                 ::= f  # float
237//                 ::= d  # double
238
239static RsdCpuReference::CpuSymbol gSyms[] = {
240    { "_Z4acosf", (void *)&acosf, true },
241    { "_Z5acoshf", (void *)&acoshf, true },
242    { "_Z4asinf", (void *)&asinf, true },
243    { "_Z5asinhf", (void *)&asinhf, true },
244    { "_Z4atanf", (void *)&atanf, true },
245    { "_Z5atan2ff", (void *)&atan2f, true },
246    { "_Z5atanhf", (void *)&atanhf, true },
247    { "_Z4cbrtf", (void *)&cbrtf, true },
248    { "_Z4ceilf", (void *)&ceilf, true },
249    { "_Z8copysignff", (void *)&copysignf, true },
250    { "_Z3cosf", (void *)&cosf, true },
251    { "_Z4coshf", (void *)&coshf, true },
252    { "_Z4erfcf", (void *)&erfcf, true },
253    { "_Z3erff", (void *)&erff, true },
254    { "_Z3expf", (void *)&expf, true },
255    { "_Z4exp2f", (void *)&exp2f, true },
256    { "exp2f", (void *)&exp2f, true },
257    { "_Z5expm1f", (void *)&expm1f, true },
258    { "_Z4fdimff", (void *)&fdimf, true },
259    { "_Z5floorf", (void *)&floorf, true },
260    { "_Z3fmafff", (void *)&fmaf, true },
261    { "_Z4fmaxff", (void *)&fmaxf, true },
262    { "_Z4fminff", (void *)&fminf, true },  // float fmin(float, float)
263    { "_Z4fmodff", (void *)&fmodf, true },
264    { "_Z5frexpfPi", (void *)&frexpf, true },
265    { "_Z5hypotff", (void *)&hypotf, true },
266    { "_Z5ilogbf", (void *)&ilogbf, true },
267    { "_Z5ldexpfi", (void *)&ldexpf, true },
268    { "_Z6lgammaf", (void *)&lgammaf, true },
269    { "_Z6lgammafPi", (void *)&lgammaf_r, true },
270    { "_Z3logf", (void *)&logf, true },
271    { "_Z5log10f", (void *)&log10f, true },
272    { "_Z5log1pf", (void *)&log1pf, true },
273    { "_Z4logbf", (void *)&logbf, true },
274    { "_Z4modffPf", (void *)&modff, true },
275    //{ "_Z3nanj", (void *)&SC_nan, true },
276    { "_Z9nextafterff", (void *)&nextafterf, true },
277    { "_Z3powff", (void *)&powf, true },
278    { "powf", (void *)&powf, true },
279    { "_Z9remainderff", (void *)&remainderf, true },
280    { "_Z6remquoffPi", (void *)&remquof, true },
281    { "_Z4rintf", (void *)&rintf, true },
282    { "_Z5roundf", (void *)&roundf, true },
283    { "_Z3sinf", (void *)&sinf, true },
284    { "_Z4sinhf", (void *)&sinhf, true },
285    { "_Z4sqrtf", (void *)&sqrtf, true },
286    { "_Z3tanf", (void *)&tanf, true },
287    { "_Z4tanhf", (void *)&tanhf, true },
288    { "_Z6tgammaf", (void *)&SC_tgammaf, true },
289    { "_Z5truncf", (void *)&truncf, true },
290
291    //{ "smoothstep", (void *)&, true },
292
293    // matrix
294    { "_Z18rsMatrixLoadRotateP12rs_matrix4x4ffff", (void *)&SC_MatrixLoadRotate, true },
295    { "_Z17rsMatrixLoadScaleP12rs_matrix4x4fff", (void *)&SC_MatrixLoadScale, true },
296    { "_Z21rsMatrixLoadTranslateP12rs_matrix4x4fff", (void *)&SC_MatrixLoadTranslate, true },
297    { "_Z14rsMatrixRotateP12rs_matrix4x4ffff", (void *)&SC_MatrixRotate, true },
298    { "_Z13rsMatrixScaleP12rs_matrix4x4fff", (void *)&SC_MatrixScale, true },
299    { "_Z17rsMatrixTranslateP12rs_matrix4x4fff", (void *)&SC_MatrixTranslate, true },
300
301    { "_Z17rsMatrixLoadOrthoP12rs_matrix4x4ffffff", (void *)&SC_MatrixLoadOrtho, true },
302    { "_Z19rsMatrixLoadFrustumP12rs_matrix4x4ffffff", (void *)&SC_MatrixLoadFrustum, true },
303    { "_Z23rsMatrixLoadPerspectiveP12rs_matrix4x4ffff", (void *)&SC_MatrixLoadPerspective, true },
304
305    { "_Z15rsMatrixInverseP12rs_matrix4x4", (void *)&SC_MatrixInverse_4x4, true },
306    { "_Z24rsMatrixInverseTransposeP12rs_matrix4x4", (void *)&SC_MatrixInverseTranspose_4x4, true },
307    { "_Z17rsMatrixTransposeP12rs_matrix4x4", (void *)&SC_MatrixTranspose_4x4, true },
308    { "_Z17rsMatrixTransposeP12rs_matrix3x3", (void *)&SC_MatrixTranspose_3x3, true },
309    { "_Z17rsMatrixTransposeP12rs_matrix2x2", (void *)&SC_MatrixTranspose_2x2, true },
310
311    // RS Math
312    { "_Z6rsRandff", (void *)&SC_randf2, true },
313    { "_Z6rsFracf", (void *)&SC_frac, true },
314
315    { NULL, NULL, false }
316};
317
318const RsdCpuReference::CpuSymbol * RsdCpuScriptImpl::lookupSymbolMath(const char *sym) {
319    const RsdCpuReference::CpuSymbol *syms = gSyms;
320
321    while (syms->fnPtr) {
322        if (!strcmp(syms->name, sym)) {
323            return syms;
324        }
325        syms++;
326    }
327    return NULL;
328}
329
330