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
|
/*
*
* rgb2rgb.c, Software RGB to RGB convertor
* Written by Nick Kurshev.
*/
#include <inttypes.h>
#include "../config.h"
#include "rgb2rgb.h"
#include "../mmx_defs.h"
#ifdef HAVE_MMX
static const uint64_t mask32 __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
static const uint64_t mask24l __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
static const uint64_t mask24h __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
static const uint64_t mask15b __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
static const uint64_t mask15s __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
#endif
void rgb24to32(const uint8_t *src,uint8_t *dst,uint32_t src_size)
{
uint8_t *dest = dst;
const uint8_t *s = src;
const uint8_t *end;
#ifdef HAVE_MMX
uint8_t *mm_end;
#endif
end = s + src_size;
#ifdef HAVE_MMX
__asm __volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = (uint8_t*)((((unsigned long)end)/(MMREG_SIZE*2))*(MMREG_SIZE*2));
__asm __volatile("movq %0, %%mm7"::"m"(mask32):"memory");
if(mm_end == end) mm_end -= MMREG_SIZE*2;
while(s < mm_end)
{
__asm __volatile(
PREFETCH" 32%1\n\t"
"movd %1, %%mm0\n\t"
"movd 3%1, %%mm1\n\t"
"movd 6%1, %%mm2\n\t"
"movd 9%1, %%mm3\n\t"
"punpckldq %%mm1, %%mm0\n\t"
"punpckldq %%mm3, %%mm2\n\t"
"pand %%mm7, %%mm0\n\t"
"pand %%mm7, %%mm2\n\t"
MOVNTQ" %%mm0, %0\n\t"
MOVNTQ" %%mm2, 8%0"
:"=m"(*dest)
:"m"(*s)
:"memory");
dest += 16;
s += 12;
}
__asm __volatile(SFENCE:::"memory");
__asm __volatile(EMMS:::"memory");
#endif
while(s < end)
{
*dest++ = *s++;
*dest++ = *s++;
*dest++ = *s++;
*dest++ = 0;
}
}
void rgb32to24(const uint8_t *src,uint8_t *dst,uint32_t src_size)
{
uint8_t *dest = dst;
const uint8_t *s = src;
const uint8_t *end;
#ifdef HAVE_MMX
uint8_t *mm_end;
#endif
end = s + src_size;
#ifdef HAVE_MMX
__asm __volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = (uint8_t*)((((unsigned long)end)/(MMREG_SIZE*2))*(MMREG_SIZE*2));
__asm __volatile(
"movq %0, %%mm7\n\t"
"movq %1, %%mm6"
::"m"(mask24l),"m"(mask24h):"memory");
if(mm_end == end) mm_end -= MMREG_SIZE*2;
while(s < mm_end)
{
__asm __volatile(
PREFETCH" 32%1\n\t"
"movq %1, %%mm0\n\t"
"movq 8%1, %%mm1\n\t"
"movq %%mm0, %%mm2\n\t"
"movq %%mm1, %%mm3\n\t"
"psrlq $8, %%mm2\n\t"
"psrlq $8, %%mm3\n\t"
"pand %%mm7, %%mm0\n\t"
"pand %%mm7, %%mm1\n\t"
"pand %%mm6, %%mm2\n\t"
"pand %%mm6, %%mm3\n\t"
"por %%mm2, %%mm0\n\t"
"por %%mm3, %%mm1\n\t"
MOVNTQ" %%mm0, %0\n\t"
MOVNTQ" %%mm1, 6%0"
:"=m"(*dest)
:"m"(*s)
:"memory");
dest += 12;
s += 16;
}
__asm __volatile(SFENCE:::"memory");
__asm __volatile(EMMS:::"memory");
#endif
while(s < end)
{
*dest++ = *s++;
*dest++ = *s++;
*dest++ = *s++;
s++;
}
}
/*
Original by Strepto/Astral
ported to gcc & bugfixed : A'rpi
MMX2, 3DNOW optimization by Nick Kurshev
32bit c version, and and&add trick by Michael Niedermayer
*/
void rgb15to16(const uint8_t *src,uint8_t *dst,uint32_t src_size)
{
#ifdef HAVE_MMX
register const char* s=src+src_size;
register char* d=dst+src_size;
register int offs=-src_size;
__asm __volatile(PREFETCH" %0"::"m"(*(s+offs)));
__asm __volatile(
"movq %0, %%mm4\n\t"
::"m"(mask15s));
while(offs<0)
{
__asm __volatile(
PREFETCH" 32%1\n\t"
"movq %1, %%mm0\n\t"
"movq 8%1, %%mm2\n\t"
"movq %%mm0, %%mm1\n\t"
"movq %%mm2, %%mm3\n\t"
"pand %%mm4, %%mm0\n\t"
"pand %%mm4, %%mm2\n\t"
"paddw %%mm1, %%mm0\n\t"
"paddw %%mm3, %%mm2\n\t"
MOVNTQ" %%mm0, %0\n\t"
MOVNTQ" %%mm2, 8%0"
:"=m"(*(d+offs))
:"m"(*(s+offs))
);
offs+=16;
}
__asm __volatile(SFENCE:::"memory");
__asm __volatile(EMMS:::"memory");
#else
#if 0
const uint16_t *s1=( uint16_t * )src;
uint16_t *d1=( uint16_t * )dst;
uint16_t *e=((uint8_t *)s1)+src_size;
while( s1<e ){
register int x=*( s1++ );
/* rrrrrggggggbbbbb
0rrrrrgggggbbbbb
0111 1111 1110 0000=0x7FE0
00000000000001 1111=0x001F */
*( d1++ )=( x&0x001F )|( ( x&0x7FE0 )<<1 );
}
#else
const uint32_t *s1=( uint32_t * )src;
uint32_t *d1=( uint32_t * )dst;
int i;
int size= src_size>>2;
for(i=0; i<size; i++)
{
register int x= s1[i];
// d1[i] = x + (x&0x7FE07FE0); //faster but need msbit =0 which might not allways be true
d1[i] = (x&0x7FFF7FFF) + (x&0x7FE07FE0);
}
#endif
#endif
}
/**
* Pallete is assumed to contain bgr32
*/
void palette8torgb32(uint8_t *src, uint8_t *dst, int src_size, uint8_t *palette)
{
int i;
for(i=0; i<src_size; i++)
((uint32_t *)dst)[i] = ((uint32_t *)palette)[ src[i] ];
}
/**
* Pallete is assumed to contain bgr32
*/
void palette8torgb24(uint8_t *src, uint8_t *dst, int src_size, uint8_t *palette)
{
int i;
/*
writes 1 byte o much and might cause alignment issues on some architectures?
for(i=0; i<src_size; i++)
((uint32_t *)(&dst[i*3])) = ((uint32_t *)palette)[ src[i] ];
*/
for(i=0; i<src_size; i++)
{
//FIXME slow?
dst[0]= palette[ src[i]*4+0 ];
dst[1]= palette[ src[i]*4+1 ];
dst[2]= palette[ src[i]*4+2 ];
dst+= 3;
}
}
void rgb32to16(uint8_t *src, uint8_t *dst, int src_size)
{
int i;
for(i=0; i<src_size; i+=4)
{
const int b= src[i+0];
const int g= src[i+1];
const int r= src[i+2];
((uint16_t *)dst)[i]= (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
}
}
void rgb32to15(uint8_t *src, uint8_t *dst, int src_size)
{
int i;
for(i=0; i<src_size; i+=4)
{
const int b= src[i+0];
const int g= src[i+1];
const int r= src[i+2];
((uint16_t *)dst)[i]= (b>>3) | ((g&0xF8)<<3) | ((r&0xF8)<<7);
}
}
/**
* Palette is assumed to contain bgr16, see rgb32to16 to convert the palette
*/
void palette8torgb16(uint8_t *src, uint8_t *dst, int src_size, uint8_t *palette)
{
int i;
for(i=0; i<src_size; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
}
/**
* Pallete is assumed to contain bgr15, see rgb32to15 to convert the palette
*/
void palette8torgb15(uint8_t *src, uint8_t *dst, int src_size, uint8_t *palette)
{
int i;
for(i=0; i<src_size; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
}
|
