summaryrefslogtreecommitdiffstats
path: root/apps/codecs/libwmapro/wmapro_math.h
blob: b605f275e8bfacfa86d4994e4696d2adb761441f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
#ifndef _WMAPRO_MATH_H_
#define _WMAPRO_MATH_H_

#include <inttypes.h>

/* rockbox: not used
#define fixtof16(x)       (float)((float)(x) / (float)(1 << 16))
#define fixtof31(x)       (float)((float)(x) / (float)(1 << 31))
#define ftofix16(x)       ((int32_t)((x) * (float)(1 << 16) + ((x) < 0 ? -0.5:0.5)))
#define ftofix31(x)       ((int32_t)((x) * (float)(1 << 31) + ((x) < 0 ? -0.5:0.5)))
*/

#if defined(CPU_ARM)
    /* Calculates: result = (X*Y)>>Z */
    #define fixmulshift(X,Y,Z) \
    ({ \
        int32_t lo; \
        int32_t hi; \
        asm volatile ( \
            "smull %[lo], %[hi], %[x], %[y] \n\t"   /* multiply */ \
            "mov   %[lo], %[lo], lsr %[shr] \n\t"   /* lo >>= Z */ \
            "orr   %[lo], %[lo], %[hi], lsl %[shl]" /* lo |= (hi << (32-Z)) */ \
            : [lo]"=&r"(lo), [hi]"=&r"(hi) \
            : [x]"r"(X), [y]"r"(Y), [shr]"r"(Z), [shl]"r"(32-Z)); \
        lo; \
    })
     
    /* Calculates: result = (X*Y)>>16 */
    #define fixmul16(X,Y) \
     ({ \
        int32_t lo; \
        int32_t hi; \
        asm volatile ( \
           "smull %[lo], %[hi], %[x], %[y] \n\t" /* multiply */ \
           "mov   %[lo], %[lo], lsr #16    \n\t" /* lo >>= 16 */ \
           "orr   %[lo], %[lo], %[hi], lsl #16"  /* lo |= (hi << 16) */ \
           : [lo]"=&r"(lo), [hi]"=&r"(hi) \
           : [x]"r"(X), [y]"r"(Y)); \
        lo; \
     })
     
    /* Calculates: result = (X*Y)>>24 */
    #define fixmul24(X,Y) \
     ({ \
        int32_t lo; \
        int32_t hi; \
        asm volatile ( \
           "smull %[lo], %[hi], %[x], %[y] \n\t" /* multiply */ \
           "mov   %[lo], %[lo], lsr #24    \n\t" /* lo >>= 24 */ \
           "orr   %[lo], %[lo], %[hi], lsl #8"   /* lo |= (hi << 8) */ \
           : [lo]"=&r"(lo), [hi]"=&r"(hi) \
           : [x]"r"(X), [y]"r"(Y)); \
        lo; \
     })
     
    /* Calculates: result = (X*Y)>>31 */
    #define fixmul31(X,Y) \
     ({ \
        int32_t lo; \
        int32_t hi; \
        asm volatile ( \
           "smull %[lo], %[hi], %[x], %[y] \n\t" /* multiply */ \
           "mov   %[lo], %[lo], lsr #31    \n\t" /* lo >>= 31 */ \
           "orr   %[lo], %[lo], %[hi], lsl #1"   /* lo |= (hi << 1) */ \
           : [lo]"=&r"(lo), [hi]"=&r"(hi) \
           : [x]"r"(X), [y]"r"(Y)); \
        lo; \
     })
#elif defined(CPU_COLDFIRE)
    /* Calculates: result = (X*Y)>>Z */
    #define fixmulshift(X,Y,Z) \
    ({ \
        int32_t t1; \
        int32_t t2; \
        asm volatile ( \
            "mac.l   %[x],%[y],%%acc0\n\t" /* multiply */ \
            "mulu.l  %[y],%[x]       \n\t" /* get lower half, avoid emac stall */ \
            "movclr.l %%acc0,%[t1]   \n\t" /* get higher half */ \
            "moveq.l #31,%[t2]       \n\t" \
            "sub.l   %[sh],%[t2]     \n\t" /* t2 = 31 - shift */ \
            "ble.s   1f              \n\t" \
            "asl.l   %[t2],%[t1]     \n\t" /* hi <<= 31 - shift */ \
            "lsr.l   %[sh],%[x]      \n\t" /* (unsigned)lo >>= shift */ \
            "or.l    %[x],%[t1]      \n\t" /* combine result */ \
            "bra.s   2f              \n\t" \
         "1:                         \n\t" \
            "neg.l   %[t2]           \n\t" /* t2 = shift - 31 */ \
            "asr.l   %[t2],%[t1]     \n\t" /* hi >>= t2 */ \
         "2:                         \n" \
        : [t1]"=&d"(t1), [t2]"=&d"(t2) \
        : [x] "d"((X)), [y] "d"((Y)), [sh]"d"((Z))); \
        t1; \
    })

    /* Calculates: result = (X*Y)>>16 */
    #define fixmul16(X,Y) \
    ({ \
        int32_t t, x = (X); \
        asm volatile ( \
            "mac.l    %[x],%[y],%%acc0\n\t" /* multiply */ \
            "mulu.l   %[y],%[x]       \n\t" /* get lower half, avoid emac stall */ \
            "movclr.l %%acc0,%[t]     \n\t" /* get higher half */ \
            "lsr.l    #1,%[t]         \n\t" /* hi >>= 1 to compensate emac shift */ \
            "move.w   %[t],%[x]       \n\t" /* combine halfwords */\
            "swap     %[x]            \n\t" \
            : [t]"=&d"(t), [x] "+d" (x) \
            : [y] "d" ((Y))); \
        x; \
    })
    
    /* Calculates: result = (X*Y)>>24 */
    #define fixmul24(X,Y) \
    ({ \
        int32_t t1, t2; \
        asm volatile ( \
            "mac.l    %[x],%[y],%%acc0 \n\t" /* multiply */ \
            "move.l   %%accext01, %[t1]\n\t" /* get lower 8 bits */ \
            "movclr.l %%acc0,%[t2]     \n\t" /* get higher 24 bits */ \
            "asl.l    #7,%[t2]         \n\t" /* hi <<= 7, plus one free */ \
            "move.b   %[t1],%[t2]      \n\t" /* combine result */ \
            : [t1]"=&d"(t1), [t2]"=&d"(t2) \
            : [x] "d" ((X)), [y] "d" ((Y))); \
        t2; \
    })

    /* Calculates: result = (X*Y)>>32 */
    #define fixmul31(X,Y) \
    ({ \
       int32_t t; \
       asm volatile ( \
          "mac.l %[x], %[y], %%acc0\n\t"   /* multiply */ \
          "movclr.l %%acc0, %[t]\n\t"      /* get higher half as result */ \
          : [t] "=d" (t) \
          : [x] "r" ((X)), [y] "r" ((Y))); \
       t; \
    })
#else
    static inline int32_t fixmulshift(int32_t x, int32_t y, int shamt)
    {
        int64_t temp;
        temp = x;
        temp *= y;
    
        temp >>= shamt;
    
        return (int32_t)temp;
    }
    
    static inline int32_t fixmul31(int32_t x, int32_t y)
    {
        int64_t temp;
        temp = x;
        temp *= y;
    
        temp >>= 31;
    
        return (int32_t)temp;
    }
    
    static inline int32_t fixmul24(int32_t x, int32_t y)
    {
        int64_t temp;
        temp = x;
        temp *= y;
    
        temp >>= 24;
    
        return (int32_t)temp;
    }
    
    static inline int32_t fixmul16(int32_t x, int32_t y)
    {
        int64_t temp;
        temp = x;
        temp *= y;
    
        temp >>= 16;
    
        return (int32_t)temp;
    }
#endif /* CPU_COLDFIRE, CPU_ARM */

#if defined(CPU_COLDFIRE)
    #define VECT_MUL_WIN_KERNEL(i, j, s0, s1, wi, wj) \
        asm volatile ( \
            "mac.l    %[s0], %[wj], %%acc0 \n\t" \
            "msac.l   %[s1], %[wi], %%acc0 \n\t" \
            "mac.l    %[s0], %[wi], %%acc1 \n\t" \
            "mac.l    %[s1], %[wj], %%acc1 \n\t" \
            "movclr.l %%acc0, %[s0]        \n\t" \
            "move.l   %[s0], (%[dst_i])    \n\t" \
            "movclr.l %%acc1, %[s0]        \n\t" \
            "move.l   %[s0], (%[dst_j])    \n\t" \
            : [s0] "+r" (s0) /* register is clobbered so specify it as an input */ \
            : [dst_i] "a" (&dst[i]), [dst_j] "a" (&dst[j]), \
              [s1] "r" (s1), [wi] "r" (wi), [wj] "r" (wj) \
            : "cc", "memory");
#else
    #define VECT_MUL_WIN_KERNEL(i, j, s0, s1, wi, wj) \
        dst[i] = fixmul31(s0, wj) - fixmul31(s1, wi); \
        dst[j] = fixmul31(s0, wi) + fixmul31(s1, wj);
#endif /* CPU_COLDFIRE */

static inline void vector_fixmul_window(int32_t *dst, const int32_t *src0, 
                                   const int32_t *src1, const int32_t *win, 
                                   int len)
{
    int i, j;
    dst += len;
    win += len;
    src0+= len;
    for(i=-len, j=len-1; i<0; i++, j--) {
        int32_t s0 = src0[i]; /* s0 = src0[      0 ... len-1] */
        int32_t s1 = src1[j]; /* s1 = src1[2*len-1 ... len]   */
        int32_t wi = -win[i]; /* wi = -win[      0 ... len-1] */
        int32_t wj = -win[j]; /* wj = -win[2*len-1 ... len]   */
        VECT_MUL_WIN_KERNEL(i, j, s0, s1, wi, wj);
    }
}

#if defined(CPU_ARM)
    #define VECT_MUL_SCALAR_KERNEL(dst, src, mul) \
        asm volatile ( \
            "ldmia %[src]!, {r1-r4}   \n\t" \
            "smull r0, r5, r1, %[mul] \n\t" \
            "mov   r0, r0, lsr #24    \n\t" \
            "orr   r0, r0, r5, lsl #8 \n\t" \
            "smull r1, r5, r2, %[mul] \n\t" \
            "mov   r1, r1, lsr #24    \n\t" \
            "orr   r1, r1, r5, lsl #8 \n\t" \
            "smull r2, r5, r3, %[mul] \n\t" \
            "mov   r2, r2, lsr #24    \n\t" \
            "orr   r2, r2, r5, lsl #8 \n\t" \
            "smull r3, r5, r4, %[mul] \n\t" \
            "mov   r3, r3, lsr #24    \n\t" \
            "orr   r3, r3, r5, lsl #8 \n\t" \
            "stmia %[dst]!, {r0-r3}   \n"   \
            : [dst]"+r"(dst), [src]"+r"(src) \
            : [mul]"r"(mul) \
            : "r0", "r1", "r2", "r3", "r4", "r5", "memory");
#elif defined (CPU_COLDFIRE)
    #define VECT_MUL_SCALAR_KERNEL(dst, src, mul) \
        int32_t tmp; \
        asm volatile ( \
            "movem.l  (%[src]), %%d0-%%d3  \n\t" \
            "mac.l    %[mul], %%d0, %%acc0 \n\t" \
            "mac.l    %[mul], %%d1, %%acc1 \n\t" \
            "mac.l    %[mul], %%d2, %%acc2 \n\t" \
            "mac.l    %[mul], %%d3, %%acc3 \n\t" \
            "move.l   %%accext01, %[tmp]   \n\t" \
            "movclr.l %%acc0, %%d0         \n\t" \
            "movclr.l %%acc1, %%d1         \n\t" \
            "lsl.l    #7, %%d0             \n\t" \
            "move.b   %[tmp], %%d0         \n\t" \
            "swap     %[tmp]               \n\t" \
            "lsl.l    #7, %%d1             \n\t" \
            "move.b   %[tmp], %%d1         \n\t" \
            "move.l   %%accext23, %[tmp]   \n\t" \
            "movclr.l %%acc2, %%d2         \n\t" \
            "movclr.l %%acc3, %%d3         \n\t" \
            "lsl.l    #7, %%d2             \n\t" \
            "move.b   %[tmp], %%d2         \n\t" \
            "swap     %[tmp]               \n\t" \
            "lsl.l    #7, %%d3             \n\t" \
            "move.b   %[tmp], %%d3         \n\t" \
            "movem.l  %%d0-%%d3, (%[dst])  \n\t" \
            "lea.l    (4*4, %[src]), %[src]\n\t" \
            "lea.l    (4*4, %[dst]), %[dst]\n\t" \
            : [dst] "+a" (dst), [src] "+a" (src),\
              [tmp] "=d" (tmp)                   \
            : [mul] "r" (mul)                    \
            : "d0", "d1", "d2", "d3", "memory", "cc");
#else
    #define VECT_MUL_SCALAR_KERNEL(dst, src, mul) \
        dst[i  ] = fixmul24(src[i  ], mul); \
        dst[i+1] = fixmul24(src[i+1], mul); \
        dst[i+2] = fixmul24(src[i+2], mul); \
        dst[i+3] = fixmul24(src[i+3], mul);
#endif /* CPU_ARM, CPU_COLDFIRE */

static inline void vector_fixmul_scalar(int32_t *dst, const int32_t *src, 
                                        int32_t mul, int len)
{
    /* len is _always_ a multiple of 4, because len is the difference of sfb's
     * which themselves are always a multiple of 4. */
    int i;
    for (i=0; i<len; i+=4) {
        VECT_MUL_SCALAR_KERNEL(dst, src, mul);
    }
}

static inline int av_clip(int a, int amin, int amax)
{
    if      (a < amin) return amin;
    else if (a > amax) return amax;
    else               return a;
}
#endif /* _WMAPRO_MATH_H_ */