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
|
/*
* srfftp.h
*
* Copyright (C) Yuqing Deng <Yuqing_Deng@brown.edu> - April 2000
*
* 64 and 128 point split radix fft for ac3dec
*
* The algorithm is desribed in the book:
* "Computational Frameworks of the Fast Fourier Transform".
*
* The ideas and the the organization of code borrowed from djbfft written by
* D. J. Bernstein <djb@cr.py.to>. djbff can be found at
* http://cr.yp.to/djbfft.html.
*
* srfftp.h is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* srfftp.h is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*
* Modified for using AMD's 3DNow! - 3DNowEx(DSP)! SIMD operations
* by Nick Kurshev <nickols_k@mail.ru>
*/
#ifndef SRFFTP_3DNOW_H__
#define SRFFTP_3DNOW_H__
typedef struct
{
unsigned long val[2];
}i_cmplx_t;
#define TRANS_FILL_MM6_MM7_3DNOW()\
__asm__ volatile(\
"movq %1, %%mm7\n\t"\
"movq %0, %%mm6\n\t"\
::"m"(x_plus_minus_3dnow),\
"m"(x_minus_plus_3dnow)\
:"memory");
#if HAVE_AMD3DNOWEXT
#define PSWAP_MM(mm_base,mm_hlp) "pswapd "mm_base","mm_base"\n\t"
#else
#define PSWAP_MM(mm_base,mm_hlp)\
"movq "mm_base","mm_hlp"\n\t"\
"psrlq $32, "mm_base"\n\t"\
"punpckldq "mm_hlp","mm_base"\n\t"
#endif
#if HAVE_AMD3DNOWEXT
#define PFNACC_MM(mm_base,mm_hlp) "pfnacc "mm_base","mm_base"\n\t"
#else
#define PFNACC_MM(mm_base,mm_hlp)\
"movq "mm_base","mm_hlp"\n\t"\
"psrlq $32,"mm_hlp"\n\t"\
"punpckldq "mm_hlp","mm_hlp"\n\t"\
"pfsub "mm_hlp","mm_base"\n\t"
#endif
#define TRANSZERO_3DNOW(A0,A4,A8,A12) \
{ \
__asm__ volatile(\
"movq %4, %%mm0\n\t" /* mm0 = wTB[0]*/\
"movq %5, %%mm1\n\t" /* mm1 = wTB[k*2]*/ \
"movq %%mm0, %%mm5\n\t"/*u.re = wTB[0].re + wTB[k*2].re;*/\
"pfadd %%mm1, %%mm5\n\t"/*u.im = wTB[0].im + wTB[k*2].im; mm5 = u*/\
"pxor %%mm6, %%mm0\n\t"/*mm0 = wTB[0].re | -wTB[0].im */\
"pxor %%mm7, %%mm1\n\t"/*mm1 = -wTB[k*2].re | wTB[k*2].im */\
"pfadd %%mm1, %%mm0\n\t"/*v.im = wTB[0].re - wTB[k*2].re;*/\
"movq %%mm0, %%mm4\n\t"/*v.re =-wTB[0].im + wTB[k*2].im;*/\
PSWAP_MM("%%mm4","%%mm2")/* mm4 = v*/\
"movq %6, %%mm0\n\t" /* a1 = A0;*/\
"movq %7, %%mm2\n\t" /* a1 = A4;*/\
"movq %%mm0, %%mm1\n\t"\
"movq %%mm2, %%mm3\n\t"\
"pfadd %%mm5, %%mm0\n\t" /*A0 = a1 + u;*/\
"pfadd %%mm4, %%mm2\n\t" /*A12 = a1 + v;*/\
"movq %%mm0, %0\n\t"\
"pfsub %%mm5, %%mm1\n\t" /*A1 = a1 - u;*/\
"movq %%mm2, %3\n\t"\
"pfsub %%mm4, %%mm3\n\t" /*A4 = a1 - v;*/\
"movq %%mm1, %1\n\t"\
"movq %%mm3, %2"\
:"=m"(A0), "=m"(A8), "=m"(A4), "=m"(A12)\
:"m"(wTB[0]), "m"(wTB[k*2]), "m"(A0), "m"(A4)\
:"memory");\
}
#define TRANSHALF_16_3DNOW(A2,A6,A10,A14)\
{\
__asm__ volatile(\
"movq %4, %%mm0\n\t"/*u.re = wTB[2].im + wTB[2].re;*/\
"movq %%mm0, %%mm1\n\t"\
"pxor %%mm7, %%mm1\n\t"\
"pfacc %%mm1, %%mm0\n\t"/*u.im = wTB[2].im - wTB[2].re; mm0 = u*/\
"movq %5, %%mm1\n\t" /*a.re = wTB[6].im - wTB[6].re; */\
"movq %%mm1, %%mm2\n\t"\
"pxor %%mm7, %%mm1\n\t"\
"pfacc %%mm2, %%mm1\n\t"/*a.im = wTB[6].im + wTB[6].re; mm1 = a*/\
"movq %%mm1, %%mm2\n\t"\
"pxor %%mm7, %%mm2\n\t"/*v.im = u.re - a.re;*/\
"movq %%mm0, %%mm3\n\t"/*v.re = u.im + a.im;*/\
"pfadd %%mm2, %%mm3\n\t"\
PSWAP_MM("%%mm3","%%mm2")/*mm3 = v*/\
"pxor %%mm6, %%mm1\n\t"/*u.re = u.re + a.re;*/\
"pfadd %%mm1, %%mm0\n\t"/*u.im = u.im - a.im; mm0 = u*/\
"movq %8, %%mm2\n\t"\
"pfmul %%mm2, %%mm3\n\t" /* v *= HSQRT2_3DNOW; */\
"pfmul %%mm2, %%mm0\n\t" /* u *= HSQRT2_3DNOW; */\
"movq %6, %%mm1\n\t" /* a1 = A2;*/\
"movq %7, %%mm5\n\t" /* a1 = A6;*/\
"movq %%mm1, %%mm2\n\t"\
"movq %%mm3, %%mm4\n\t"\
"pfadd %%mm0, %%mm1\n\t" /*A2 = a1 + u;*/\
"pxor %%mm6, %%mm4\n\t"/*A6.re = a1.re + v.re;*/\
"pfsub %%mm0, %%mm2\n\t" /*A2 = a1 - u;*/\
"pxor %%mm7, %%mm3\n\t"/*A14.re = a1.re - v.re;*/\
"movq %%mm1, %0\n\t"\
"movq %%mm2, %1\n\t"\
"movq %%mm5, %%mm2\n\t"\
"pfadd %%mm4, %%mm5\n\t"/*A6.im = a1.im - v.im;*/\
"pfadd %%mm3, %%mm2\n\t"/*A14.im = a1.im + v.im;*/\
"movq %%mm5, %2\n\t"\
"movq %%mm2, %3"\
:"=m"(A2), "=m"(A10), "=m"(A6), "=m"(A14)\
:"m"(wTB[2]), "m"(wTB[6]), "m"(A2), "m"(A6), "m"(HSQRT2_3DNOW)\
:"memory");\
}
#define TRANS_3DNOW(A1,A5,A9,A13,WT,WB,D,D3)\
{ \
__asm__ volatile(\
"movq %1, %%mm4\n\t"\
"movq %%mm4, %%mm5\n\t"\
"punpckldq %%mm4, %%mm4\n\t"/*mm4 = D.re | D.re */\
"punpckhdq %%mm5, %%mm5\n\t"/*mm5 = D.im | D.im */\
"movq %0, %%mm0\n\t"\
"pfmul %%mm0, %%mm4\n\t"/* mm4 =u.re | u.im */\
"pfmul %%mm0, %%mm5\n\t"/* mm5 = a.re | a.im */\
PSWAP_MM("%%mm5","%%mm3")\
"pxor %%mm7, %%mm5\n\t"\
"pfadd %%mm5, %%mm4\n\t"/* mm4 = u*/\
"movq %3, %%mm1\n\t"\
"movq %2, %%mm0\n\t"\
PSWAP_MM("%%mm1","%%mm3")\
"movq %%mm0, %%mm2\n\t"\
"pfmul %%mm1, %%mm0\n\t"/* mm0 = a*/\
"pfmul %3, %%mm2\n\t"/* mm2 = v*/\
PFNACC_MM("%%mm2","%%mm3")\
"pfacc %%mm0, %%mm0\n\t"\
"movq %%mm4, %%mm5\n\t"\
"punpckldq %%mm0,%%mm2\n\t"/*mm2 = v.re | a.re*/\
"pxor %%mm6, %%mm5\n\t"\
"movq %%mm2, %%mm3\n\t"\
"pxor %%mm7, %%mm3\n\t"\
"pfadd %%mm3, %%mm5\n\t"\
PSWAP_MM("%%mm5","%%mm3")/* mm5 = v*/\
"pfadd %%mm2, %%mm4\n\t"\
:\
:"m"(WT), "m"(D), "m"(WB), "m"(D3)\
:"memory");\
__asm__ volatile(\
"movq %4, %%mm0\n\t"/* a1 = A1*/\
"movq %5, %%mm2\n\t"/* a1 = A5*/\
"movq %%mm0, %%mm1\n\t"\
"movq %%mm2, %%mm3\n\t"\
"pfadd %%mm4, %%mm0\n\t"/*A1 = a1 + u*/\
"pfsub %%mm5, %%mm2\n\t"/*A5 = a1 - v*/\
"movq %%mm0, %0\n\t"\
"pfsub %%mm4, %%mm1\n\t"/*A9 = a1 - u*/\
"movq %%mm2, %2\n\t"\
"pfadd %%mm5, %%mm3\n\t"/*A9 = a1 + v*/\
"movq %%mm1, %1\n\t"\
"movq %%mm3, %3"\
:"=m"(A1), "=m"(A9), "=m"(A5), "=m"(A13)\
:"m"(A1), "m"(A5)\
:"memory");\
}
#endif
|
