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
|
//altivec version by Chris Clepper
//
static t_int *partconv_perform(t_int *w)
{
t_partconv *x = (t_partconv *)(w[1]);
t_float *in = (t_float *)(w[2]);
t_float *out = (t_float *)(w[3]);
int n = (int)(w[4]);
int i;
int j;
int k; // bin
int p; // partition
int endpart;
fftwf_complex *cursumbuf_fd;
float *sumbuf1ptr;
float *sumbuf2ptr;
union {
unsigned char c[16];
vector unsigned char v;
}permfill;
union {
float f[4];
vector float v;
}floatfill;
vector float *load_input, *load_irpart;
vector float store_multbuf1,store_multbuf2;
vector float vinput_fd0, vinput_fd4; //input vectors
vector float virpart_fd0, virpart_fd4; //ir partition vectors
vector float permtemp1357, permtemp0246;
vector float vzero;// vscale;
vector unsigned char input_0022, input_1133, perm_0246, perm_1357, perm_0123,perm_4567;
vector float vtemp1, vtemp2, vtemp3, vtemp4, vtemp5, vtemp6, vtemp7, vtemp8;
floatfill.f[0] = 0.f;
floatfill.f[1] = 0.f;
floatfill.f[2] = 0.f;
floatfill.f[3] = 0.f;
vzero = floatfill.v;
//store_multbuf = vzero;
floatfill.f[0] = x->scale;
floatfill.f[1] = x->scale;
floatfill.f[2] = x->scale;
floatfill.f[3] = x->scale;
//vscale = floatfill.v;
//fill the permute buffer for the first input_fd multiply
permfill.c[0] = 0; permfill.c[1] = 1; permfill.c[2] = 2; permfill.c[3] = 3; //first float
permfill.c[4] = 0; permfill.c[5] = 1; permfill.c[6] = 2; permfill.c[7] = 3; //second float
permfill.c[8] = 8; permfill.c[9] = 9; permfill.c[10] = 10; permfill.c[11] = 11; //third float
permfill.c[12] = 8; permfill.c[13] = 9; permfill.c[14] = 10; permfill.c[15] = 11; //fourth float
input_0022 = permfill.v;
permfill.c[0] = 4; permfill.c[1] = 5; permfill.c[2] = 6; permfill.c[3] = 7; //first float
permfill.c[4] = 4; permfill.c[5] = 5; permfill.c[6] = 6; permfill.c[7] = 7; //second float
permfill.c[8] = 12; permfill.c[9] = 13; permfill.c[10] = 14; permfill.c[11] = 15; //third float
permfill.c[12] = 12; permfill.c[13] = 13; permfill.c[14] = 14; permfill.c[15] = 15; //fourth float
input_1133 = permfill.v;
//perm_0246
//0,1,2,3, 8,9,10,11, 16,17,18,19, 24,25,26,27
permfill.c[0] = 0; permfill.c[1] = 1; permfill.c[2] = 2; permfill.c[3] = 3; //first float
permfill.c[4] = 8; permfill.c[5] = 9; permfill.c[6] = 10; permfill.c[7] = 11; //second float
permfill.c[8] = 16; permfill.c[9] = 17; permfill.c[10] = 18; permfill.c[11] = 19; //third float
permfill.c[12] = 24; permfill.c[13] = 25; permfill.c[14] = 26; permfill.c[15] = 27; //fourth float
perm_0246 = permfill.v;
// perm_1357
// 4,5,6,7, 12,13,14,15, 20,21,22,23, 28,29,30,31
permfill.c[0] = 4; permfill.c[1] = 5; permfill.c[2] = 6; permfill.c[3] = 7; //first float
permfill.c[4] = 12; permfill.c[5] = 13; permfill.c[6] = 14; permfill.c[7] = 15; //second float
permfill.c[8] = 20; permfill.c[9] = 21; permfill.c[10] = 22; permfill.c[11] = 23; //third float
permfill.c[12] = 28; permfill.c[13] = 29; permfill.c[14] = 30; permfill.c[15] = 31; //fourth float
perm_1357 = permfill.v;
// perm_0123 from [0,2,4,6] and [1,3,5,7]
// 0,1,2,3 16,17,18,19 4,5,6,7 20,21,22,23
permfill.c[0] = 0; permfill.c[1] = 1; permfill.c[2] = 2; permfill.c[3] = 3; //first float
permfill.c[4] = 16; permfill.c[5] = 17; permfill.c[6] = 18; permfill.c[7] = 19; //second float
permfill.c[8] = 4; permfill.c[9] = 5; permfill.c[10] = 6; permfill.c[11] = 7; //third float
permfill.c[12] = 20; permfill.c[13] = 21; permfill.c[14] = 22; permfill.c[15] = 23; //fourth float
perm_0123 = permfill.v;
// perm_4567 from [0,2,4,6] and [1,3,5,7]
// 8.9.10.11 24,25,26,27 12,13,14,15 28,29,30,31
permfill.c[0] = 8; permfill.c[1] = 9; permfill.c[2] = 10; permfill.c[3] = 11; //first float
permfill.c[4] = 24; permfill.c[5] = 25; permfill.c[6] = 26; permfill.c[7] = 27; //second float
permfill.c[8] = 12; permfill.c[9] = 13; permfill.c[10] = 14; permfill.c[11] = 15; //third float
permfill.c[12] = 28; permfill.c[13] = 29; permfill.c[14] = 30; permfill.c[15] = 31; //fourth float
// perm_4567 from [0,2,4,6] and [1,3,5,7]
// 8.9.10.11 24,25,26,27 12,13,14,15 28,29,30,31
permfill.c[0] = 8; permfill.c[1] = 9; permfill.c[2] = 10; permfill.c[3] = 11; //first float
permfill.c[4] = 24; permfill.c[5] = 25; permfill.c[6] = 26; permfill.c[7] = 27; //second float
permfill.c[8] = 12; permfill.c[9] = 13; permfill.c[10] = 14; permfill.c[11] = 15; //third float
permfill.c[12] = 28; permfill.c[13] = 29; permfill.c[14] = 30; permfill.c[15] = 31; //fourth float
perm_4567 = permfill.v;
memcpy(&(x->inbuf[x->inbufpos]), in, n*sizeof(float)); // gather a block of input into input buffer
x->inbufpos += n;
if (x->inbufpos >= x->partsize) {
// input buffer is full, so we begin a new cycle
if (x->pd_blocksize != n) {
// the patch's blocksize has change since we last dealt the work
x->pd_blocksize = n;
partconv_deal_work(x);
}
x->inbufpos = 0;
x->curcall = 0;
x->curpart = 0;
memcpy(x->input_td, x->inbuf, x->partsize * sizeof(float)); // copy 'gathering' input buffer into 'transform' buffer
memset(&(x->input_td[x->partsize]), 0, (x->paddedsize - x->partsize) * sizeof(float)); // pad
fftwf_execute(x->input_plan); // transform the input
// everything has been read out of prev sumbuf, so clear it
memset(x->sumbuf->prev->td, 0, x->paddedsize * sizeof(float));
// advance sumbuf pointers
x->sumbuf = x->sumbuf->next;
x->sumbuf->readpos = 0;
x->sumbuf->prev->readpos = x->partsize;
}
// convolve this call's portion of partitions
endpart = x->curpart + x->parts_per_call[x->curcall];
if (endpart > x->nparts) // FIXME does this ever happen?
endpart = x->nparts;
for (p = x->curpart; p < endpart; p++) {
//printf("convolving with partition %d\n", p);
//
// multiply the input block by the partition, accumulating the result in the appropriate sumbuf
//
// FIXME do this in a circular list-type fashion so we don't need "index"
cursumbuf_fd = x->sumbufs[(x->sumbuf->index + p) % x->nsumbufs].fd;
for (k = 0; k < x->nbins; k+=4) {
load_input = (vector float *)&x->input_fd[k][0];
vinput_fd0 = vec_ld(0, (vector float *) load_input);
vtemp1 = vec_perm(load_input[0],vzero,input_0022);
load_input = (vector float *)&x->input_fd[k][4];
//load input_fd[k][4]
//vector will have input_fd[4,5,6,7]
vinput_fd4 = vec_ld(0, (vector float *) &x->input_fd[k][4]);
vtemp3 = vec_perm(load_input[0],vzero,input_0022);
//vec_ld irpart[p][k][0]
//vector will have irpart_fd[0,1,2,3]
load_irpart = (vector float *) &x->irpart_fd[p][k][0];
virpart_fd0 = vec_ld(0,&x->irpart_fd[p][k][0]);
vtemp1 = vec_madd(vtemp1,load_irpart[0],vzero);
load_irpart = (vector float *) &x->irpart_fd[p][k][4];
virpart_fd4 = vec_ld(0,&x->irpart_fd[p][k][4]);
vtemp3 = vec_madd(vtemp3,load_irpart[0],vzero);
store_multbuf1 = vec_ld(0,&cursumbuf_fd[k][0]);
store_multbuf2 = vec_ld(0,&cursumbuf_fd[k][4]);
//vec_perm to line up the elements
// irpart is fine
// make vector of input_fd[0] [2] and [4] [6]
//make vector of input_fd[1] [3] and [5] [7]
//
// permute only works on bytes so the first float is bytes 0,1,2,3 the second is 4,5,6,7 etc
//
// 0,1,2,3, 8,9,10,11, 16,17,18,19, 24,25,26,27
//
// 4,5,6,7, 12,13,14,15, 20,21,22,23, 28,29,30,31
//vec_perm temp1 and temp3 into [0,2,4,6]
permtemp0246 = vec_perm(vtemp1,vtemp3,perm_0246);
//and [1,3,5,7]
permtemp1357 = vec_perm(vtemp1,vtemp3,perm_1357);
//vinput_fd[1,3,5,7]
vtemp2 = vec_perm(vinput_fd0,vinput_fd4,perm_1357);
//irpart[1,3,5,7]
vtemp4 = vec_perm(virpart_fd0,virpart_fd4,perm_1357);
//irpart[0,2,4,6]
vtemp5 = vec_perm(virpart_fd0,virpart_fd4,perm_0246);
//vec_nmsub input_fd[1,3,5,7] irpart[1,3,5,7] temp[0,2,4,6]
vtemp6 = vec_nmsub(vtemp2,vtemp4,permtemp0246);
//vec_madd input_fd[1,3,5,7] irpart[0,2,4,6] temp[1,3,5,7]
vtemp7 = vec_madd(vtemp2,vtemp5,permtemp1357);
//vec_madd all by scale - this is now done after the loop
// vtemp6 = vec_madd(vtemp6,vscale,vzero);
// vtemp7 = vec_madd(vtemp7,vscale,vzero);
//vec_perm data back into place - tricky!
//vec_perm nmsub_result[0,2,4,6] madd_result [1,3,5,7]
// results will be [0,1,2,3] [4,5,6,7]
vtemp1 = vec_perm(vtemp6,vtemp7,perm_0123);
vtemp2 = vec_perm(vtemp6,vtemp7,perm_4567);
//vec_st
store_multbuf1 = vec_add(store_multbuf1,vtemp1);
store_multbuf2 = vec_add(store_multbuf2,vtemp2);
vec_st(store_multbuf1,0,&cursumbuf_fd[k][0]);
vec_st(store_multbuf2,0,&cursumbuf_fd[k][4]);
/*
cursumbuf_fd[k][0]
+=
( x->input_fd[k][0] * x->irpart_fd[p][k][0]
- x->input_fd[k][1] * x->irpart_fd[p][k][1]);
cursumbuf_fd[k][1]
+=
( x->input_fd[k][0] * x->irpart_fd[p][k][1]
+ x->input_fd[k][1] * x->irpart_fd[p][k][0]);*/
}
}
x->curpart = p;
// The convolution of the fresh block of input with the first partition of the IR
// is the last thing that gets summed into the current sumbuf before it gets IFFTed and starts being output.
// This happens during the first call of every cycle.
if (x->curcall == 0) {
// current sumbuf has been filled, so transform it (TD to FD).
// Output loop will begin to read it and sum it with the last one
fftwf_execute(x->sumbuf->plan);
}
// we're summing and outputting the first half of the most recently IFFTed sumbuf
// and the second half of the previous one
sumbuf1ptr = &(x->sumbuf->td[x->sumbuf->readpos]);
sumbuf2ptr = &(x->sumbuf->prev->td[x->sumbuf->prev->readpos]);
for (i = 0; i < n; i++) {
*(out++) = (*(sumbuf1ptr++) + *(sumbuf2ptr++)) * x->scale;
}
x->sumbuf->readpos += n;
x->sumbuf->prev->readpos += n;
x->curcall++;
return (w+5);
}
|