root / TrouNoir / TrouNoir.py @ 235
Historique | Voir | Annoter | Télécharger (50,76 ko)
1 |
#!/usr/bin/env python
|
---|---|
2 |
#
|
3 |
# TrouNoir model using PyOpenCL
|
4 |
#
|
5 |
# CC BY-NC-SA 2019 : <emmanuel.quemener@ens-lyon.fr>
|
6 |
#
|
7 |
# Thanks to Andreas Klockner for PyOpenCL:
|
8 |
# http://mathema.tician.de/software/pyopencl
|
9 |
#
|
10 |
# Original code programmed in Fortran 77 in mars 1994
|
11 |
# for Practical Work of Numerical Simulation
|
12 |
# DEA (old Master2) in astrophysics and spatial techniques in Meudon
|
13 |
# by Herve Aussel & Emmanuel Quemener
|
14 |
#
|
15 |
# Conversion in C done by Emmanuel Quemener in august 1997
|
16 |
# GPUfication in OpenCL under Python in july 2019
|
17 |
# GPUfication in CUDA under Python in august 2019
|
18 |
#
|
19 |
# Thanks to :
|
20 |
#
|
21 |
# - Herve Aussel for his part of code of black body spectrum
|
22 |
# - Didier Pelat for his help to perform this work
|
23 |
# - Jean-Pierre Luminet for his article published in 1979
|
24 |
# - Numerical Recipies for Runge Kutta recipies
|
25 |
# - Luc Blanchet for his disponibility about my questions in General Relativity
|
26 |
# - Pierre Lena for his passion about science and vulgarisation
|
27 |
|
28 |
# If crash on OpenCL Intel implementation, add following options and force
|
29 |
#export PYOPENCL_COMPILER_OUTPUT=1
|
30 |
#export CL_CONFIG_USE_VECTORIZER=True
|
31 |
#export CL_CONFIG_CPU_VECTORIZER_MODE=16
|
32 |
|
33 |
import pyopencl as cl |
34 |
import numpy |
35 |
import time,string |
36 |
from numpy.random import randint as nprnd |
37 |
import sys |
38 |
import getopt |
39 |
import matplotlib.pyplot as plt |
40 |
from socket import gethostname |
41 |
|
42 |
def DictionariesAPI(): |
43 |
PhysicsList={'Einstein':0,'Newton':1} |
44 |
return(PhysicsList)
|
45 |
|
46 |
#
|
47 |
# Blank space below to simplify debugging on OpenCL code
|
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 |
BlobOpenCL= """
|
104 |
|
105 |
#define PI (float)3.14159265359e0f
|
106 |
#define nbr 256
|
107 |
|
108 |
#define EINSTEIN 0
|
109 |
#define NEWTON 1
|
110 |
|
111 |
#ifdef SETTRACKPOINTS
|
112 |
#define TRACKPOINTS SETTRACKPOINTS
|
113 |
#else
|
114 |
#define TRACKPOINTS 2048
|
115 |
#endif
|
116 |
|
117 |
float atanp(float x,float y)
|
118 |
{
|
119 |
float angle;
|
120 |
|
121 |
angle=atan2(y,x);
|
122 |
|
123 |
if (angle<0.e0f)
|
124 |
{
|
125 |
angle+=(float)2.e0f*PI;
|
126 |
}
|
127 |
|
128 |
return angle;
|
129 |
}
|
130 |
|
131 |
float f(float v)
|
132 |
{
|
133 |
return v;
|
134 |
}
|
135 |
|
136 |
#if PHYSICS == NEWTON
|
137 |
float g(float u,float m,float b)
|
138 |
{
|
139 |
return (-u);
|
140 |
}
|
141 |
#else
|
142 |
float g(float u,float m,float b)
|
143 |
{
|
144 |
return (3.e0f*m/b*pow(u,2)-u);
|
145 |
}
|
146 |
#endif
|
147 |
|
148 |
void calcul(float *us,float *vs,float up,float vp,
|
149 |
float h,float m,float b)
|
150 |
{
|
151 |
float c0,c1,c2,c3,d0,d1,d2,d3;
|
152 |
|
153 |
c0=h*f(vp);
|
154 |
c1=h*f(vp+c0/2.e0f);
|
155 |
c2=h*f(vp+c1/2.e0f);
|
156 |
c3=h*f(vp+c2);
|
157 |
d0=h*g(up,m,b);
|
158 |
d1=h*g(up+d0/2.e0f,m,b);
|
159 |
d2=h*g(up+d1/2.e0f,m,b);
|
160 |
d3=h*g(up+d2,m,b);
|
161 |
|
162 |
*us=up+(c0+2.e0f*c1+2.e0f*c2+c3)/6.e0f;
|
163 |
*vs=vp+(d0+2.e0f*d1+2.e0f*d2+d3)/6.e0f;
|
164 |
}
|
165 |
|
166 |
void rungekutta(float *ps,float *us,float *vs,
|
167 |
float pp,float up,float vp,
|
168 |
float h,float m,float b)
|
169 |
{
|
170 |
calcul(us,vs,up,vp,h,m,b);
|
171 |
*ps=pp+h;
|
172 |
}
|
173 |
|
174 |
float decalage_spectral(float r,float b,float phi,
|
175 |
float tho,float m)
|
176 |
{
|
177 |
return (sqrt(1-3*m/r)/(1+sqrt(m/pow(r,3))*b*sin(tho)*sin(phi)));
|
178 |
}
|
179 |
|
180 |
float spectre(float rf,int q,float b,float db,
|
181 |
float h,float r,float m,float bss)
|
182 |
{
|
183 |
float flx;
|
184 |
|
185 |
// flx=exp(q*log(r/m))*pow(rf,4)*b*db*h;
|
186 |
flx=exp(q*log(r/m)+4.e0f*log(rf))*b*db*h;
|
187 |
return(flx);
|
188 |
}
|
189 |
|
190 |
float spectre_cn(float rf32,float b32,float db32,
|
191 |
float h32,float r32,float m32,float bss32)
|
192 |
{
|
193 |
|
194 |
#define MYFLOAT float
|
195 |
|
196 |
MYFLOAT rf=(MYFLOAT)(rf32);
|
197 |
MYFLOAT b=(MYFLOAT)(b32);
|
198 |
MYFLOAT db=(MYFLOAT)(db32);
|
199 |
MYFLOAT h=(MYFLOAT)(h32);
|
200 |
MYFLOAT r=(MYFLOAT)(r32);
|
201 |
MYFLOAT m=(MYFLOAT)(m32);
|
202 |
MYFLOAT bss=(MYFLOAT)(bss32);
|
203 |
|
204 |
MYFLOAT flx;
|
205 |
MYFLOAT nu_rec,nu_em,qu,temp_em,flux_int;
|
206 |
int fi,posfreq;
|
207 |
|
208 |
#define planck 6.62e-34f
|
209 |
#define k 1.38e-23f
|
210 |
#define c2 9.e16f
|
211 |
#define temp 3.e7f
|
212 |
#define m_point 1.e0f
|
213 |
|
214 |
#define lplanck (log(6.62e0f)-34.e0f*log(10.e0f))
|
215 |
#define lk (log(1.38e0f)-23.e0f*log(10.e0f))
|
216 |
#define lc2 (log(9.e0f)+16.e0f*log(10.e0f))
|
217 |
|
218 |
MYFLOAT v=1.e0f-3.e0f/r;
|
219 |
|
220 |
qu=1.e0f/sqrt((1.e0f-3.e0f/r)*r)*(sqrt(r)-sqrt(6.e0f)+sqrt(3.e0f)/2.e0f*log((sqrt(r)+sqrt(3.e0f))/(sqrt(r)-sqrt(3.e0f))* 0.17157287525380988e0f ));
|
221 |
|
222 |
temp_em=temp*sqrt(m)*exp(0.25e0f*log(m_point)-0.75e0f*log(r)-0.125e0f*log(v)+0.25e0f*log(fabs(qu)));
|
223 |
|
224 |
flux_int=0.e0f;
|
225 |
flx=0.e0f;
|
226 |
|
227 |
for (fi=0;fi<nbr;fi++)
|
228 |
{
|
229 |
nu_em=bss*(MYFLOAT)fi/(MYFLOAT)nbr;
|
230 |
nu_rec=nu_em*rf;
|
231 |
posfreq=(int)(nu_rec*(MYFLOAT)nbr/bss);
|
232 |
if ((posfreq>0)&&(posfreq<nbr))
|
233 |
{
|
234 |
// Initial version
|
235 |
// flux_int=2.*planck/c2*pow(nu_em,3)/(exp(planck*nu_em/(k*temp_em))-1.);
|
236 |
// Version with log used
|
237 |
//flux_int=2.*exp(lplanck-lc2+3.*log(nu_em))/(exp(exp(lplanck-lk+log(nu_em/temp_em)))-1.);
|
238 |
// flux_int*=pow(rf,3)*b*db*h;
|
239 |
//flux_int*=exp(3.e0f*log(rf))*b*db*h;
|
240 |
flux_int=2.e0f*exp(lplanck-lc2+3.e0f*log(nu_em))/(exp(exp(lplanck-lk+log(nu_em/temp_em)))-1.e0f)*exp(3.e0f*log(rf))*b*db*h;
|
241 |
|
242 |
flx+=flux_int;
|
243 |
}
|
244 |
}
|
245 |
|
246 |
return((float)(flx));
|
247 |
}
|
248 |
|
249 |
void impact(float phi,float r,float b,float tho,float m,
|
250 |
float *zp,float *fp,
|
251 |
int q,float db,
|
252 |
float h,int raie)
|
253 |
{
|
254 |
float flx,rf,bss;
|
255 |
|
256 |
rf=decalage_spectral(r,b,phi,tho,m);
|
257 |
|
258 |
if (raie==0)
|
259 |
{
|
260 |
bss=1.e19f;
|
261 |
flx=spectre_cn(rf,b,db,h,r,m,bss);
|
262 |
}
|
263 |
else
|
264 |
{
|
265 |
bss=2.e0f;
|
266 |
flx=spectre(rf,q,b,db,h,r,m,bss);
|
267 |
}
|
268 |
|
269 |
*zp=1.e0f/rf;
|
270 |
*fp=flx;
|
271 |
|
272 |
}
|
273 |
|
274 |
__kernel void EachPixel(__global float *zImage,__global float *fImage,
|
275 |
float Mass,float InternalRadius,
|
276 |
float ExternalRadius,float Angle,
|
277 |
int Line)
|
278 |
{
|
279 |
uint xi=(uint)get_global_id(0);
|
280 |
uint yi=(uint)get_global_id(1);
|
281 |
uint sizex=(uint)get_global_size(0);
|
282 |
uint sizey=(uint)get_global_size(1);
|
283 |
|
284 |
// Perform trajectory for each pixel, exit on hit
|
285 |
|
286 |
float m,rs,ri,re,tho;
|
287 |
int q,raie;
|
288 |
|
289 |
m=Mass;
|
290 |
rs=2.e0f*m;
|
291 |
ri=InternalRadius;
|
292 |
re=ExternalRadius;
|
293 |
tho=Angle;
|
294 |
q=-2;
|
295 |
raie=Line;
|
296 |
|
297 |
float bmx,db,b,h;
|
298 |
float rp0,rps;
|
299 |
float phi,phd;
|
300 |
uint nh=0;
|
301 |
float zp=0.e0f,fp=0.e0f;
|
302 |
|
303 |
// Autosize for image
|
304 |
bmx=1.25e0f*re;
|
305 |
|
306 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
307 |
|
308 |
// set origin as center of image
|
309 |
float x=(float)xi-(float)(sizex/2)+(float)5.e-1f;
|
310 |
float y=(float)yi-(float)(sizey/2)+(float)5.e-1f;
|
311 |
// angle extracted from cylindric symmetry
|
312 |
phi=atanp(x,y);
|
313 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
314 |
|
315 |
|
316 |
float up,vp,pp,us,vs,ps;
|
317 |
|
318 |
// impact parameter
|
319 |
b=sqrt(x*x+y*y)*(float)2.e0f/(float)sizex*bmx;
|
320 |
// step of impact parameter;
|
321 |
db=bmx/(float)(sizex);
|
322 |
|
323 |
up=0.e0f;
|
324 |
vp=1.e0f;
|
325 |
pp=0.e0f;
|
326 |
|
327 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
328 |
|
329 |
rps=fabs(b/us);
|
330 |
rp0=rps;
|
331 |
|
332 |
int ExitOnImpact=0;
|
333 |
|
334 |
do
|
335 |
{
|
336 |
nh++;
|
337 |
pp=ps;
|
338 |
up=us;
|
339 |
vp=vs;
|
340 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
341 |
rps=fabs(b/us);
|
342 |
ExitOnImpact = ((fmod(pp,PI)<fmod(phd,PI))&&(fmod(ps,PI)>fmod(phd,PI)))&&(rps>=ri)&&(rps<=re)?1:0;
|
343 |
|
344 |
} while ((rps>=rs)&&(rps<=rp0)&&(ExitOnImpact==0)&&(nh<TRACKPOINTS));
|
345 |
|
346 |
|
347 |
if (ExitOnImpact==1) {
|
348 |
impact(phi,rps,b,tho,m,&zp,&fp,q,db,h,raie);
|
349 |
}
|
350 |
else
|
351 |
{
|
352 |
zp=0.e0f;
|
353 |
fp=0.e0f;
|
354 |
}
|
355 |
|
356 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
357 |
|
358 |
zImage[yi+sizex*xi]=(float)zp;
|
359 |
fImage[yi+sizex*xi]=(float)fp;
|
360 |
}
|
361 |
|
362 |
__kernel void Pixel(__global float *zImage,__global float *fImage,
|
363 |
__global float *Trajectories,__global int *IdLast,
|
364 |
uint ImpactParameter,
|
365 |
float Mass,float InternalRadius,
|
366 |
float ExternalRadius,float Angle,
|
367 |
int Line)
|
368 |
{
|
369 |
uint xi=(uint)get_global_id(0);
|
370 |
uint yi=(uint)get_global_id(1);
|
371 |
uint sizex=(uint)get_global_size(0);
|
372 |
uint sizey=(uint)get_global_size(1);
|
373 |
|
374 |
// Perform trajectory for each pixel
|
375 |
|
376 |
float m,ri,re,tho;
|
377 |
int q,raie;
|
378 |
|
379 |
m=Mass;
|
380 |
ri=InternalRadius;
|
381 |
re=ExternalRadius;
|
382 |
tho=Angle;
|
383 |
q=-2;
|
384 |
raie=Line;
|
385 |
|
386 |
float bmx,db,b,h;
|
387 |
float phi,phd,php,nr,r;
|
388 |
float zp=0.e0f,fp=0.e0f;
|
389 |
|
390 |
// Autosize for image, 25% greater than external radius
|
391 |
bmx=1.25e0f*re;
|
392 |
|
393 |
// Angular step of integration
|
394 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
395 |
|
396 |
// Step of Impact Parameter
|
397 |
db=bmx/(2.e0f*(float)ImpactParameter);
|
398 |
|
399 |
// set origin as center of image
|
400 |
float x=(float)xi-(float)(sizex/2)+(float)5.e-1f;
|
401 |
float y=(float)yi-(float)(sizey/2)+(float)5.e-1f;
|
402 |
|
403 |
// angle extracted from cylindric symmetry
|
404 |
phi=atanp(x,y);
|
405 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
406 |
|
407 |
// Real Impact Parameter
|
408 |
b=sqrt(x*x+y*y)*bmx/(float)ImpactParameter;
|
409 |
|
410 |
// Integer Impact Parameter
|
411 |
uint bi=(uint)sqrt(x*x+y*y);
|
412 |
|
413 |
int HalfLap=0,ExitOnImpact=0,ni;
|
414 |
|
415 |
if (bi<ImpactParameter)
|
416 |
{
|
417 |
do
|
418 |
{
|
419 |
php=phd+(float)HalfLap*PI;
|
420 |
nr=php/h;
|
421 |
ni=(int)nr;
|
422 |
|
423 |
if (ni<IdLast[bi])
|
424 |
{
|
425 |
r=(Trajectories[bi*TRACKPOINTS+ni+1]-Trajectories[bi*TRACKPOINTS+ni])*(nr-ni*1.e0f)+Trajectories[bi*TRACKPOINTS+ni];
|
426 |
}
|
427 |
else
|
428 |
{
|
429 |
r=Trajectories[bi*TRACKPOINTS+ni];
|
430 |
}
|
431 |
|
432 |
if ((r<=re)&&(r>=ri))
|
433 |
{
|
434 |
ExitOnImpact=1;
|
435 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
436 |
}
|
437 |
|
438 |
HalfLap++;
|
439 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
440 |
|
441 |
}
|
442 |
|
443 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
444 |
|
445 |
zImage[yi+sizex*xi]=zp;
|
446 |
fImage[yi+sizex*xi]=fp;
|
447 |
}
|
448 |
|
449 |
__kernel void Circle(__global float *Trajectories,__global int *IdLast,
|
450 |
__global float *zImage,__global float *fImage,
|
451 |
float Mass,float InternalRadius,
|
452 |
float ExternalRadius,float Angle,
|
453 |
int Line)
|
454 |
{
|
455 |
// Integer Impact Parameter ID
|
456 |
int bi=get_global_id(0);
|
457 |
// Integer points on circle
|
458 |
int i=get_global_id(1);
|
459 |
// Integer Impact Parameter Size (half of image)
|
460 |
int bmaxi=get_global_size(0);
|
461 |
// Integer Points on circle
|
462 |
int imx=get_global_size(1);
|
463 |
|
464 |
// Perform trajectory for each pixel
|
465 |
|
466 |
float m,ri,re,tho;
|
467 |
int q,raie;
|
468 |
|
469 |
m=Mass;
|
470 |
ri=InternalRadius;
|
471 |
re=ExternalRadius;
|
472 |
tho=Angle;
|
473 |
raie=Line;
|
474 |
|
475 |
float bmx,db,b,h;
|
476 |
float phi,phd;
|
477 |
float zp=0.e0f,fp=0.e0f;
|
478 |
|
479 |
// Autosize for image
|
480 |
bmx=1.25e0f*re;
|
481 |
|
482 |
// Angular step of integration
|
483 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
484 |
|
485 |
// impact parameter
|
486 |
b=(float)bi/(float)bmaxi*bmx;
|
487 |
db=bmx/(2.e0f*(float)bmaxi);
|
488 |
|
489 |
phi=2.e0f*PI/(float)imx*(float)i;
|
490 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
491 |
int yi=(int)((float)bi*sin(phi))+bmaxi;
|
492 |
int xi=(int)((float)bi*cos(phi))+bmaxi;
|
493 |
|
494 |
int HalfLap=0,ExitOnImpact=0,ni;
|
495 |
float php,nr,r;
|
496 |
|
497 |
do
|
498 |
{
|
499 |
php=phd+(float)HalfLap*PI;
|
500 |
nr=php/h;
|
501 |
ni=(int)nr;
|
502 |
|
503 |
if (ni<IdLast[bi])
|
504 |
{
|
505 |
r=(Trajectories[bi*TRACKPOINTS+ni+1]-Trajectories[bi*TRACKPOINTS+ni])*(nr-ni*1.e0f)+Trajectories[bi*TRACKPOINTS+ni];
|
506 |
}
|
507 |
else
|
508 |
{
|
509 |
r=Trajectories[bi*TRACKPOINTS+ni];
|
510 |
}
|
511 |
|
512 |
if ((r<=re)&&(r>=ri))
|
513 |
{
|
514 |
ExitOnImpact=1;
|
515 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
516 |
}
|
517 |
|
518 |
HalfLap++;
|
519 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
520 |
|
521 |
zImage[yi+2*bmaxi*xi]=zp;
|
522 |
fImage[yi+2*bmaxi*xi]=fp;
|
523 |
|
524 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
525 |
|
526 |
}
|
527 |
|
528 |
__kernel void Trajectory(__global float *Trajectories,__global int *IdLast,
|
529 |
float Mass,float InternalRadius,
|
530 |
float ExternalRadius,float Angle,
|
531 |
int Line)
|
532 |
{
|
533 |
// Integer Impact Parameter ID
|
534 |
int bi=get_global_id(0);
|
535 |
// Integer Impact Parameter Size (half of image)
|
536 |
int bmaxi=get_global_size(0);
|
537 |
|
538 |
// Perform trajectory for each pixel
|
539 |
|
540 |
float m,rs,re;
|
541 |
|
542 |
m=Mass;
|
543 |
rs=2.e0f*m;
|
544 |
re=ExternalRadius;
|
545 |
|
546 |
float bmx,b,h;
|
547 |
int nh;
|
548 |
|
549 |
// Autosize for image
|
550 |
bmx=1.25e0f*re;
|
551 |
|
552 |
// Angular step of integration
|
553 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
554 |
|
555 |
// impact parameter
|
556 |
b=(float)bi/(float)bmaxi*bmx;
|
557 |
|
558 |
float up,vp,pp,us,vs,ps;
|
559 |
|
560 |
up=0.e0f;
|
561 |
vp=1.e0f;
|
562 |
|
563 |
pp=0.e0f;
|
564 |
nh=0;
|
565 |
|
566 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
567 |
|
568 |
// b versus us
|
569 |
float bvus=fabs(b/us);
|
570 |
float bvus0=bvus;
|
571 |
Trajectories[bi*TRACKPOINTS+nh]=bvus;
|
572 |
|
573 |
do
|
574 |
{
|
575 |
nh++;
|
576 |
pp=ps;
|
577 |
up=us;
|
578 |
vp=vs;
|
579 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
580 |
bvus=fabs(b/us);
|
581 |
Trajectories[bi*TRACKPOINTS+nh]=bvus;
|
582 |
|
583 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
584 |
|
585 |
IdLast[bi]=nh;
|
586 |
|
587 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
588 |
|
589 |
}
|
590 |
|
591 |
__kernel void EachCircle(__global float *zImage,__global float *fImage,
|
592 |
float Mass,float InternalRadius,
|
593 |
float ExternalRadius,float Angle,
|
594 |
int Line)
|
595 |
{
|
596 |
// Integer Impact Parameter ID
|
597 |
uint bi=(uint)get_global_id(0);
|
598 |
// Integer Impact Parameter Size (half of image)
|
599 |
uint bmaxi=(uint)get_global_size(0);
|
600 |
|
601 |
private float Trajectory[TRACKPOINTS];
|
602 |
|
603 |
float m,rs,ri,re,tho;
|
604 |
int raie,q;
|
605 |
|
606 |
m=Mass;
|
607 |
rs=2.e0f*m;
|
608 |
ri=InternalRadius;
|
609 |
re=ExternalRadius;
|
610 |
tho=Angle;
|
611 |
q=-2;
|
612 |
raie=Line;
|
613 |
|
614 |
float bmx,db,b,h;
|
615 |
uint nh;
|
616 |
|
617 |
|
618 |
// Autosize for image
|
619 |
bmx=1.25e0f*re;
|
620 |
|
621 |
// Angular step of integration
|
622 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
623 |
|
624 |
// impact parameter
|
625 |
b=(float)bi/(float)bmaxi*bmx;
|
626 |
db=bmx/(2.e0f*(float)bmaxi);
|
627 |
|
628 |
float up,vp,pp,us,vs,ps;
|
629 |
|
630 |
up=0.e0f;
|
631 |
vp=1.e0f;
|
632 |
|
633 |
pp=0.e0f;
|
634 |
nh=0;
|
635 |
|
636 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
637 |
|
638 |
// b versus us
|
639 |
float bvus=fabs(b/us);
|
640 |
float bvus0=bvus;
|
641 |
Trajectory[nh]=bvus;
|
642 |
|
643 |
do
|
644 |
{
|
645 |
nh++;
|
646 |
pp=ps;
|
647 |
up=us;
|
648 |
vp=vs;
|
649 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
650 |
bvus=(float)fabs(b/us);
|
651 |
Trajectory[nh]=bvus;
|
652 |
|
653 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
654 |
|
655 |
|
656 |
for (uint i=(uint)nh+1;i<TRACKPOINTS;i++) {
|
657 |
Trajectory[i]=0.e0f;
|
658 |
}
|
659 |
|
660 |
|
661 |
uint imx=(uint)(16*bi);
|
662 |
|
663 |
for (uint i=0;i<imx;i++)
|
664 |
{
|
665 |
float zp=0.e0f,fp=0.e0f;
|
666 |
float phi=2.e0f*PI/(float)imx*(float)i;
|
667 |
float phd=atanp(cos(phi)*sin(tho),cos(tho));
|
668 |
uint yi=(uint)((float)bi*sin(phi)+bmaxi);
|
669 |
uint xi=(uint)((float)bi*cos(phi)+bmaxi);
|
670 |
|
671 |
uint HalfLap=0,ExitOnImpact=0,ni;
|
672 |
float php,nr,r;
|
673 |
|
674 |
do
|
675 |
{
|
676 |
php=phd+(float)HalfLap*PI;
|
677 |
nr=php/h;
|
678 |
ni=(int)nr;
|
679 |
|
680 |
if (ni<nh)
|
681 |
{
|
682 |
r=(Trajectory[ni+1]-Trajectory[ni])*(nr-ni*1.e0f)+Trajectory[ni];
|
683 |
}
|
684 |
else
|
685 |
{
|
686 |
r=Trajectory[ni];
|
687 |
}
|
688 |
|
689 |
if ((r<=re)&&(r>=ri))
|
690 |
{
|
691 |
ExitOnImpact=1;
|
692 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
693 |
}
|
694 |
|
695 |
HalfLap++;
|
696 |
|
697 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
698 |
|
699 |
zImage[yi+2*bmaxi*xi]=zp;
|
700 |
fImage[yi+2*bmaxi*xi]=fp;
|
701 |
|
702 |
}
|
703 |
|
704 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
705 |
|
706 |
}
|
707 |
|
708 |
__kernel void Original(__global float *zImage,__global float *fImage,
|
709 |
uint Size,float Mass,float InternalRadius,
|
710 |
float ExternalRadius,float Angle,
|
711 |
int Line)
|
712 |
{
|
713 |
// Integer Impact Parameter Size (half of image)
|
714 |
uint bmaxi=(uint)Size;
|
715 |
|
716 |
float Trajectory[TRACKPOINTS];
|
717 |
|
718 |
// Perform trajectory for each pixel
|
719 |
|
720 |
float m,rs,ri,re,tho;
|
721 |
int raie,q;
|
722 |
|
723 |
m=Mass;
|
724 |
rs=2.e0f*m;
|
725 |
ri=InternalRadius;
|
726 |
re=ExternalRadius;
|
727 |
tho=Angle;
|
728 |
q=-2;
|
729 |
raie=Line;
|
730 |
|
731 |
float bmx,db,b,h;
|
732 |
uint nh;
|
733 |
|
734 |
// Autosize for image
|
735 |
bmx=1.25e0f*re;
|
736 |
|
737 |
// Angular step of integration
|
738 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
739 |
|
740 |
// Integer Impact Parameter ID
|
741 |
for (int bi=0;bi<bmaxi;bi++)
|
742 |
{
|
743 |
// impact parameter
|
744 |
b=(float)bi/(float)bmaxi*bmx;
|
745 |
db=bmx/(2.e0f*(float)bmaxi);
|
746 |
|
747 |
float up,vp,pp,us,vs,ps;
|
748 |
|
749 |
up=0.e0f;
|
750 |
vp=1.e0f;
|
751 |
|
752 |
pp=0.e0f;
|
753 |
nh=0;
|
754 |
|
755 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
756 |
|
757 |
// b versus us
|
758 |
float bvus=fabs(b/us);
|
759 |
float bvus0=bvus;
|
760 |
Trajectory[nh]=bvus;
|
761 |
|
762 |
do
|
763 |
{
|
764 |
nh++;
|
765 |
pp=ps;
|
766 |
up=us;
|
767 |
vp=vs;
|
768 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
769 |
bvus=fabs(b/us);
|
770 |
Trajectory[nh]=bvus;
|
771 |
|
772 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
773 |
|
774 |
for (uint i=(uint)nh+1;i<TRACKPOINTS;i++) {
|
775 |
Trajectory[i]=0.e0f;
|
776 |
}
|
777 |
|
778 |
int imx=(int)(16*bi);
|
779 |
|
780 |
for (int i=0;i<imx;i++)
|
781 |
{
|
782 |
float zp=0.e0f,fp=0.e0f;
|
783 |
float phi=2.e0f*PI/(float)imx*(float)i;
|
784 |
float phd=atanp(cos(phi)*sin(tho),cos(tho));
|
785 |
uint yi=(uint)((float)bi*sin(phi)+bmaxi);
|
786 |
uint xi=(uint)((float)bi*cos(phi)+bmaxi);
|
787 |
|
788 |
uint HalfLap=0,ExitOnImpact=0,ni;
|
789 |
float php,nr,r;
|
790 |
|
791 |
do
|
792 |
{
|
793 |
php=phd+(float)HalfLap*PI;
|
794 |
nr=php/h;
|
795 |
ni=(int)nr;
|
796 |
|
797 |
if (ni<nh)
|
798 |
{
|
799 |
r=(Trajectory[ni+1]-Trajectory[ni])*(nr-ni*1.e0f)+Trajectory[ni];
|
800 |
}
|
801 |
else
|
802 |
{
|
803 |
r=Trajectory[ni];
|
804 |
}
|
805 |
|
806 |
if ((r<=re)&&(r>=ri))
|
807 |
{
|
808 |
ExitOnImpact=1;
|
809 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
810 |
}
|
811 |
|
812 |
HalfLap++;
|
813 |
|
814 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
815 |
|
816 |
zImage[yi+2*bmaxi*xi]=zp;
|
817 |
fImage[yi+2*bmaxi*xi]=fp;
|
818 |
|
819 |
}
|
820 |
|
821 |
}
|
822 |
|
823 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
824 |
|
825 |
}
|
826 |
"""
|
827 |
|
828 |
def KernelCodeCuda(): |
829 |
BlobCUDA= """
|
830 |
|
831 |
#define PI (float)3.14159265359
|
832 |
#define nbr 256
|
833 |
|
834 |
#define EINSTEIN 0
|
835 |
#define NEWTON 1
|
836 |
|
837 |
#ifdef SETTRACKPOINTS
|
838 |
#define TRACKPOINTS SETTRACKPOINTS
|
839 |
#else
|
840 |
#define TRACKPOINTS
|
841 |
#endif
|
842 |
|
843 |
__device__ float nothing(float x)
|
844 |
{
|
845 |
return(x);
|
846 |
}
|
847 |
|
848 |
__device__ float atanp(float x,float y)
|
849 |
{
|
850 |
float angle;
|
851 |
|
852 |
angle=atan2(y,x);
|
853 |
|
854 |
if (angle<0.e0f)
|
855 |
{
|
856 |
angle+=(float)2.e0f*PI;
|
857 |
}
|
858 |
|
859 |
return(angle);
|
860 |
}
|
861 |
|
862 |
__device__ float f(float v)
|
863 |
{
|
864 |
return(v);
|
865 |
}
|
866 |
|
867 |
#if PHYSICS == NEWTON
|
868 |
__device__ float g(float u,float m,float b)
|
869 |
{
|
870 |
return (-u);
|
871 |
}
|
872 |
#else
|
873 |
__device__ float g(float u,float m,float b)
|
874 |
{
|
875 |
return (3.e0f*m/b*pow(u,2)-u);
|
876 |
}
|
877 |
#endif
|
878 |
|
879 |
__device__ void calcul(float *us,float *vs,float up,float vp,
|
880 |
float h,float m,float b)
|
881 |
{
|
882 |
float c0,c1,c2,c3,d0,d1,d2,d3;
|
883 |
|
884 |
c0=h*f(vp);
|
885 |
c1=h*f(vp+c0/2.);
|
886 |
c2=h*f(vp+c1/2.);
|
887 |
c3=h*f(vp+c2);
|
888 |
d0=h*g(up,m,b);
|
889 |
d1=h*g(up+d0/2.,m,b);
|
890 |
d2=h*g(up+d1/2.,m,b);
|
891 |
d3=h*g(up+d2,m,b);
|
892 |
|
893 |
*us=up+(c0+2.*c1+2.*c2+c3)/6.;
|
894 |
*vs=vp+(d0+2.*d1+2.*d2+d3)/6.;
|
895 |
}
|
896 |
|
897 |
__device__ void rungekutta(float *ps,float *us,float *vs,
|
898 |
float pp,float up,float vp,
|
899 |
float h,float m,float b)
|
900 |
{
|
901 |
calcul(us,vs,up,vp,h,m,b);
|
902 |
*ps=pp+h;
|
903 |
}
|
904 |
|
905 |
__device__ float decalage_spectral(float r,float b,float phi,
|
906 |
float tho,float m)
|
907 |
{
|
908 |
return (sqrt(1-3*m/r)/(1+sqrt(m/pow(r,3))*b*sin(tho)*sin(phi)));
|
909 |
}
|
910 |
|
911 |
__device__ float spectre(float rf,int q,float b,float db,
|
912 |
float h,float r,float m,float bss)
|
913 |
{
|
914 |
float flx;
|
915 |
|
916 |
// flx=exp(q*log(r/m))*pow(rf,4)*b*db*h;
|
917 |
flx=exp(q*log(r/m)+4.*log(rf))*b*db*h;
|
918 |
return(flx);
|
919 |
}
|
920 |
|
921 |
__device__ float spectre_cn(float rf32,float b32,float db32,
|
922 |
float h32,float r32,float m32,float bss32)
|
923 |
{
|
924 |
|
925 |
#define MYFLOAT float
|
926 |
|
927 |
MYFLOAT rf=(MYFLOAT)(rf32);
|
928 |
MYFLOAT b=(MYFLOAT)(b32);
|
929 |
MYFLOAT db=(MYFLOAT)(db32);
|
930 |
MYFLOAT h=(MYFLOAT)(h32);
|
931 |
MYFLOAT r=(MYFLOAT)(r32);
|
932 |
MYFLOAT m=(MYFLOAT)(m32);
|
933 |
MYFLOAT bss=(MYFLOAT)(bss32);
|
934 |
|
935 |
MYFLOAT flx;
|
936 |
MYFLOAT nu_rec,nu_em,qu,temp_em,flux_int;
|
937 |
int fi,posfreq;
|
938 |
|
939 |
#define planck 6.62e-34
|
940 |
#define k 1.38e-23
|
941 |
#define c2 9.e16
|
942 |
#define temp 3.e7
|
943 |
#define m_point 1.
|
944 |
|
945 |
#define lplanck (log(6.62)-34.*log(10.))
|
946 |
#define lk (log(1.38)-23.*log(10.))
|
947 |
#define lc2 (log(9.)+16.*log(10.))
|
948 |
|
949 |
MYFLOAT v=1.-3./r;
|
950 |
|
951 |
qu=1./sqrt((1.-3./r)*r)*(sqrt(r)-sqrt(6.)+sqrt(3.)/2.*log((sqrt(r)+sqrt(3.))/(sqrt(r)-sqrt(3.))* 0.17157287525380988 ));
|
952 |
|
953 |
temp_em=temp*sqrt(m)*exp(0.25*log(m_point)-0.75*log(r)-0.125*log(v)+0.25*log(fabs(qu)));
|
954 |
|
955 |
flux_int=0.;
|
956 |
flx=0.;
|
957 |
|
958 |
for (fi=0;fi<nbr;fi++)
|
959 |
{
|
960 |
nu_em=bss*(MYFLOAT)fi/(MYFLOAT)nbr;
|
961 |
nu_rec=nu_em*rf;
|
962 |
posfreq=(int)(nu_rec*(MYFLOAT)nbr/bss);
|
963 |
if ((posfreq>0)&&(posfreq<nbr))
|
964 |
{
|
965 |
// Initial version
|
966 |
// flux_int=2.*planck/c2*pow(nu_em,3)/(exp(planck*nu_em/(k*temp_em))-1.);
|
967 |
// Version with log used
|
968 |
//flux_int=2.*exp(lplanck-lc2+3.*log(nu_em))/(exp(exp(lplanck-lk+log(nu_em/temp_em)))-1.);
|
969 |
// flux_int*=pow(rf,3)*b*db*h;
|
970 |
//flux_int*=exp(3.*log(rf))*b*db*h;
|
971 |
flux_int=2.*exp(lplanck-lc2+3.*log(nu_em))/(exp(exp(lplanck-lk+log(nu_em/temp_em)))-1.)*exp(3.*log(rf))*b*db*h;
|
972 |
|
973 |
flx+=flux_int;
|
974 |
}
|
975 |
}
|
976 |
|
977 |
return((float)(flx));
|
978 |
}
|
979 |
|
980 |
__device__ void impact(float phi,float r,float b,float tho,float m,
|
981 |
float *zp,float *fp,
|
982 |
int q,float db,
|
983 |
float h,int raie)
|
984 |
{
|
985 |
float flx,rf,bss;
|
986 |
|
987 |
rf=decalage_spectral(r,b,phi,tho,m);
|
988 |
|
989 |
if (raie==0)
|
990 |
{
|
991 |
bss=1.e19;
|
992 |
flx=spectre_cn(rf,b,db,h,r,m,bss);
|
993 |
}
|
994 |
else
|
995 |
{
|
996 |
bss=2.;
|
997 |
flx=spectre(rf,q,b,db,h,r,m,bss);
|
998 |
}
|
999 |
|
1000 |
*zp=1./rf;
|
1001 |
*fp=flx;
|
1002 |
|
1003 |
}
|
1004 |
|
1005 |
__global__ void EachPixel(float *zImage,float *fImage,
|
1006 |
float Mass,float InternalRadius,
|
1007 |
float ExternalRadius,float Angle,
|
1008 |
int Line)
|
1009 |
{
|
1010 |
uint xi=(uint)(blockIdx.x*blockDim.x+threadIdx.x);
|
1011 |
uint yi=(uint)(blockIdx.y*blockDim.y+threadIdx.y);
|
1012 |
uint sizex=(uint)gridDim.x*blockDim.x;
|
1013 |
uint sizey=(uint)gridDim.y*blockDim.y;
|
1014 |
|
1015 |
// Perform trajectory for each pixel, exit on hit
|
1016 |
|
1017 |
float m,rs,ri,re,tho;
|
1018 |
int q,raie;
|
1019 |
|
1020 |
m=Mass;
|
1021 |
rs=2.*m;
|
1022 |
ri=InternalRadius;
|
1023 |
re=ExternalRadius;
|
1024 |
tho=Angle;
|
1025 |
q=-2;
|
1026 |
raie=Line;
|
1027 |
|
1028 |
float d,bmx,db,b,h;
|
1029 |
float rp0,rpp,rps;
|
1030 |
float phi,thi,phd,php,nr,r;
|
1031 |
int nh;
|
1032 |
float zp,fp;
|
1033 |
|
1034 |
// Autosize for image
|
1035 |
bmx=1.25*re;
|
1036 |
b=0.;
|
1037 |
|
1038 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1039 |
|
1040 |
// set origin as center of image
|
1041 |
float x=(float)xi-(float)(sizex/2)+(float)5e-1f;
|
1042 |
float y=(float)yi-(float)(sizey/2)+(float)5e-1f;
|
1043 |
// angle extracted from cylindric symmetry
|
1044 |
phi=atanp(x,y);
|
1045 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
1046 |
|
1047 |
float up,vp,pp,us,vs,ps;
|
1048 |
|
1049 |
// impact parameter
|
1050 |
b=sqrt(x*x+y*y)*(float)2.e0f/(float)sizex*bmx;
|
1051 |
// step of impact parameter;
|
1052 |
// db=bmx/(float)(sizex/2);
|
1053 |
db=bmx/(float)(sizex);
|
1054 |
|
1055 |
up=0.;
|
1056 |
vp=1.;
|
1057 |
pp=0.;
|
1058 |
nh=0;
|
1059 |
|
1060 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1061 |
|
1062 |
rps=fabs(b/us);
|
1063 |
rp0=rps;
|
1064 |
|
1065 |
int ExitOnImpact=0;
|
1066 |
|
1067 |
do
|
1068 |
{
|
1069 |
nh++;
|
1070 |
pp=ps;
|
1071 |
up=us;
|
1072 |
vp=vs;
|
1073 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1074 |
rpp=rps;
|
1075 |
rps=fabs(b/us);
|
1076 |
ExitOnImpact = ((fmod(pp,PI)<fmod(phd,PI))&&(fmod(ps,PI)>fmod(phd,PI)))&&(rps>ri)&&(rps<re)?1:0;
|
1077 |
|
1078 |
} while ((rps>=rs)&&(rps<=rp0)&&(ExitOnImpact==0));
|
1079 |
|
1080 |
if (ExitOnImpact==1) {
|
1081 |
impact(phi,rpp,b,tho,m,&zp,&fp,q,db,h,raie);
|
1082 |
}
|
1083 |
else
|
1084 |
{
|
1085 |
zp=0.e0f;
|
1086 |
fp=0.e0f;
|
1087 |
}
|
1088 |
|
1089 |
__syncthreads();
|
1090 |
|
1091 |
zImage[yi+sizex*xi]=(float)zp;
|
1092 |
fImage[yi+sizex*xi]=(float)fp;
|
1093 |
}
|
1094 |
|
1095 |
__global__ void Pixel(float *zImage,float *fImage,
|
1096 |
float *Trajectories,int *IdLast,
|
1097 |
uint ImpactParameter,
|
1098 |
float Mass,float InternalRadius,
|
1099 |
float ExternalRadius,float Angle,
|
1100 |
int Line)
|
1101 |
{
|
1102 |
uint xi=(uint)(blockIdx.x*blockDim.x+threadIdx.x);
|
1103 |
uint yi=(uint)(blockIdx.y*blockDim.y+threadIdx.y);
|
1104 |
uint sizex=(uint)gridDim.x*blockDim.x;
|
1105 |
uint sizey=(uint)gridDim.y*blockDim.y;
|
1106 |
|
1107 |
// Perform trajectory for each pixel
|
1108 |
|
1109 |
float m,rs,ri,re,tho;
|
1110 |
int q,raie;
|
1111 |
|
1112 |
m=Mass;
|
1113 |
rs=2.e0f*m;
|
1114 |
ri=InternalRadius;
|
1115 |
re=ExternalRadius;
|
1116 |
tho=Angle;
|
1117 |
q=-2;
|
1118 |
raie=Line;
|
1119 |
|
1120 |
float d,bmx,db,b,h;
|
1121 |
float phi,thi,phd,php,nr,r;
|
1122 |
int nh;
|
1123 |
float zp=0,fp=0;
|
1124 |
// Autosize for image, 25% greater than external radius
|
1125 |
bmx=1.25e0f*re;
|
1126 |
|
1127 |
// Angular step of integration
|
1128 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1129 |
|
1130 |
// Step of Impact Parameter
|
1131 |
db=bmx/(2.e0f*(float)ImpactParameter);
|
1132 |
|
1133 |
// set origin as center of image
|
1134 |
float x=(float)xi-(float)(sizex/2)+(float)5e-1f;
|
1135 |
float y=(float)yi-(float)(sizey/2)+(float)5e-1f;
|
1136 |
// angle extracted from cylindric symmetry
|
1137 |
phi=atanp(x,y);
|
1138 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
1139 |
|
1140 |
// Real Impact Parameter
|
1141 |
b=sqrt(x*x+y*y)*bmx/(float)ImpactParameter;
|
1142 |
|
1143 |
// Integer Impact Parameter
|
1144 |
uint bi=(uint)sqrt(x*x+y*y);
|
1145 |
|
1146 |
int HalfLap=0,ExitOnImpact=0,ni;
|
1147 |
|
1148 |
if (bi<ImpactParameter)
|
1149 |
{
|
1150 |
do
|
1151 |
{
|
1152 |
php=phd+(float)HalfLap*PI;
|
1153 |
nr=php/h;
|
1154 |
ni=(int)nr;
|
1155 |
|
1156 |
if (ni<IdLast[bi])
|
1157 |
{
|
1158 |
r=(Trajectories[bi*TRACKPOINTS+ni+1]-Trajectories[bi*TRACKPOINTS+ni])*(nr-ni*1.e0f)+Trajectories[bi*TRACKPOINTS+ni];
|
1159 |
}
|
1160 |
else
|
1161 |
{
|
1162 |
r=Trajectories[bi*TRACKPOINTS+ni];
|
1163 |
}
|
1164 |
|
1165 |
if ((r<=re)&&(r>=ri))
|
1166 |
{
|
1167 |
ExitOnImpact=1;
|
1168 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
1169 |
}
|
1170 |
|
1171 |
HalfLap++;
|
1172 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
1173 |
|
1174 |
}
|
1175 |
|
1176 |
zImage[yi+sizex*xi]=zp;
|
1177 |
fImage[yi+sizex*xi]=fp;
|
1178 |
}
|
1179 |
|
1180 |
__global__ void Circle(float *Trajectories,int *IdLast,
|
1181 |
float *zImage,float *fImage,
|
1182 |
float Mass,float InternalRadius,
|
1183 |
float ExternalRadius,float Angle,
|
1184 |
int Line)
|
1185 |
{
|
1186 |
// Integer Impact Parameter ID
|
1187 |
int bi=blockIdx.x*blockDim.x+threadIdx.x;
|
1188 |
// Integer points on circle
|
1189 |
int i=blockIdx.y*blockDim.y+threadIdx.y;
|
1190 |
// Integer Impact Parameter Size (half of image)
|
1191 |
int bmaxi=gridDim.x*blockDim.x;
|
1192 |
// Integer Points on circle
|
1193 |
int imx=gridDim.y*blockDim.y;
|
1194 |
|
1195 |
// Perform trajectory for each pixel
|
1196 |
|
1197 |
float m,rs,ri,re,tho;
|
1198 |
int q,raie;
|
1199 |
|
1200 |
m=Mass;
|
1201 |
rs=2.e0f*m;
|
1202 |
ri=InternalRadius;
|
1203 |
re=ExternalRadius;
|
1204 |
tho=Angle;
|
1205 |
raie=Line;
|
1206 |
|
1207 |
float bmx,db,b,h;
|
1208 |
float phi,thi,phd;
|
1209 |
int nh;
|
1210 |
float zp=0,fp=0;
|
1211 |
|
1212 |
// Autosize for image
|
1213 |
bmx=1.25e0f*re;
|
1214 |
|
1215 |
// Angular step of integration
|
1216 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1217 |
|
1218 |
// impact parameter
|
1219 |
b=(float)bi/(float)bmaxi*bmx;
|
1220 |
db=bmx/(2.e0f*(float)bmaxi);
|
1221 |
|
1222 |
phi=2.e0f*PI/(float)imx*(float)i;
|
1223 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
1224 |
int yi=(int)((float)bi*sin(phi))+bmaxi;
|
1225 |
int xi=(int)((float)bi*cos(phi))+bmaxi;
|
1226 |
|
1227 |
int HalfLap=0,ExitOnImpact=0,ni;
|
1228 |
float php,nr,r;
|
1229 |
|
1230 |
do
|
1231 |
{
|
1232 |
php=phd+(float)HalfLap*PI;
|
1233 |
nr=php/h;
|
1234 |
ni=(int)nr;
|
1235 |
|
1236 |
if (ni<IdLast[bi])
|
1237 |
{
|
1238 |
r=(Trajectories[bi*TRACKPOINTS+ni+1]-Trajectories[bi*TRACKPOINTS+ni])*(nr-ni*1.e0f)+Trajectories[bi*TRACKPOINTS+ni];
|
1239 |
}
|
1240 |
else
|
1241 |
{
|
1242 |
r=Trajectories[bi*TRACKPOINTS+ni];
|
1243 |
}
|
1244 |
|
1245 |
if ((r<=re)&&(r>=ri))
|
1246 |
{
|
1247 |
ExitOnImpact=1;
|
1248 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
1249 |
}
|
1250 |
|
1251 |
HalfLap++;
|
1252 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
1253 |
|
1254 |
zImage[yi+2*bmaxi*xi]=zp;
|
1255 |
fImage[yi+2*bmaxi*xi]=fp;
|
1256 |
|
1257 |
}
|
1258 |
|
1259 |
__global__ void Trajectory(float *Trajectories,int *IdLast,
|
1260 |
float Mass,float InternalRadius,
|
1261 |
float ExternalRadius,float Angle,
|
1262 |
int Line)
|
1263 |
{
|
1264 |
// Integer Impact Parameter ID
|
1265 |
int bi=blockIdx.x*blockDim.x+threadIdx.x;
|
1266 |
// Integer Impact Parameter Size (half of image)
|
1267 |
int bmaxi=gridDim.x*blockDim.x;
|
1268 |
|
1269 |
// Perform trajectory for each pixel
|
1270 |
|
1271 |
float m,rs,ri,re,tho;
|
1272 |
int raie,q;
|
1273 |
|
1274 |
m=Mass;
|
1275 |
rs=2.e0f*m;
|
1276 |
ri=InternalRadius;
|
1277 |
re=ExternalRadius;
|
1278 |
tho=Angle;
|
1279 |
q=-2;
|
1280 |
raie=Line;
|
1281 |
|
1282 |
float d,bmx,db,b,h;
|
1283 |
float phi,thi,phd,php,nr,r;
|
1284 |
int nh;
|
1285 |
float zp,fp;
|
1286 |
|
1287 |
// Autosize for image
|
1288 |
bmx=1.25e0f*re;
|
1289 |
|
1290 |
// Angular step of integration
|
1291 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1292 |
|
1293 |
// impact parameter
|
1294 |
b=(float)bi/(float)bmaxi*bmx;
|
1295 |
|
1296 |
float up,vp,pp,us,vs,ps;
|
1297 |
|
1298 |
up=0.e0f;
|
1299 |
vp=1.e0f;
|
1300 |
pp=0.e0f;
|
1301 |
nh=0;
|
1302 |
|
1303 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1304 |
|
1305 |
// b versus us
|
1306 |
float bvus=fabs(b/us);
|
1307 |
float bvus0=bvus;
|
1308 |
Trajectories[bi*TRACKPOINTS+nh]=bvus;
|
1309 |
|
1310 |
do
|
1311 |
{
|
1312 |
nh++;
|
1313 |
pp=ps;
|
1314 |
up=us;
|
1315 |
vp=vs;
|
1316 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1317 |
bvus=fabs(b/us);
|
1318 |
Trajectories[bi*TRACKPOINTS+nh]=bvus;
|
1319 |
|
1320 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
1321 |
|
1322 |
IdLast[bi]=nh;
|
1323 |
|
1324 |
}
|
1325 |
|
1326 |
__global__ void EachCircle(float *zImage,float *fImage,
|
1327 |
float Mass,float InternalRadius,
|
1328 |
float ExternalRadius,float Angle,
|
1329 |
int Line)
|
1330 |
{
|
1331 |
// Integer Impact Parameter ID
|
1332 |
int bi=blockIdx.x*blockDim.x+threadIdx.x;
|
1333 |
|
1334 |
// Integer Impact Parameter Size (half of image)
|
1335 |
int bmaxi=gridDim.x*blockDim.x;
|
1336 |
|
1337 |
float Trajectory[2048];
|
1338 |
|
1339 |
// Perform trajectory for each pixel
|
1340 |
|
1341 |
float m,rs,ri,re,tho;
|
1342 |
int raie,q;
|
1343 |
|
1344 |
m=Mass;
|
1345 |
rs=2.*m;
|
1346 |
ri=InternalRadius;
|
1347 |
re=ExternalRadius;
|
1348 |
tho=Angle;
|
1349 |
q=-2;
|
1350 |
raie=Line;
|
1351 |
|
1352 |
float bmx,db,b,h;
|
1353 |
int nh;
|
1354 |
|
1355 |
// Autosize for image
|
1356 |
bmx=1.25e0f*re;
|
1357 |
|
1358 |
// Angular step of integration
|
1359 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1360 |
|
1361 |
// impact parameter
|
1362 |
b=(float)bi/(float)bmaxi*bmx;
|
1363 |
db=bmx/(2.e0f*(float)bmaxi);
|
1364 |
|
1365 |
float up,vp,pp,us,vs,ps;
|
1366 |
|
1367 |
up=0.;
|
1368 |
vp=1.;
|
1369 |
pp=0.;
|
1370 |
nh=0;
|
1371 |
|
1372 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1373 |
|
1374 |
// b versus us
|
1375 |
float bvus=fabs(b/us);
|
1376 |
float bvus0=bvus;
|
1377 |
Trajectory[nh]=bvus;
|
1378 |
|
1379 |
do
|
1380 |
{
|
1381 |
nh++;
|
1382 |
pp=ps;
|
1383 |
up=us;
|
1384 |
vp=vs;
|
1385 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1386 |
bvus=fabs(b/us);
|
1387 |
Trajectory[nh]=bvus;
|
1388 |
|
1389 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
1390 |
|
1391 |
int imx=(int)(16*bi);
|
1392 |
|
1393 |
for (int i=0;i<imx;i++)
|
1394 |
{
|
1395 |
float zp=0,fp=0;
|
1396 |
float phi=2.*PI/(float)imx*(float)i;
|
1397 |
float phd=atanp(cos(phi)*sin(tho),cos(tho));
|
1398 |
uint yi=(uint)((float)bi*sin(phi)+bmaxi);
|
1399 |
uint xi=(uint)((float)bi*cos(phi)+bmaxi);
|
1400 |
|
1401 |
int HalfLap=0,ExitOnImpact=0,ni;
|
1402 |
float php,nr,r;
|
1403 |
|
1404 |
do
|
1405 |
{
|
1406 |
php=phd+(float)HalfLap*PI;
|
1407 |
nr=php/h;
|
1408 |
ni=(int)nr;
|
1409 |
|
1410 |
if (ni<nh)
|
1411 |
{
|
1412 |
r=(Trajectory[ni+1]-Trajectory[ni])*(nr-ni*1.)+Trajectory[ni];
|
1413 |
}
|
1414 |
else
|
1415 |
{
|
1416 |
r=Trajectory[ni];
|
1417 |
}
|
1418 |
|
1419 |
if ((r<=re)&&(r>=ri))
|
1420 |
{
|
1421 |
ExitOnImpact=1;
|
1422 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
1423 |
}
|
1424 |
|
1425 |
HalfLap++;
|
1426 |
|
1427 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
1428 |
|
1429 |
__syncthreads();
|
1430 |
|
1431 |
zImage[yi+2*bmaxi*xi]=zp;
|
1432 |
fImage[yi+2*bmaxi*xi]=fp;
|
1433 |
|
1434 |
}
|
1435 |
|
1436 |
}
|
1437 |
|
1438 |
__global__ void Original(float *zImage,float *fImage,
|
1439 |
uint Size,float Mass,float InternalRadius,
|
1440 |
float ExternalRadius,float Angle,
|
1441 |
int Line)
|
1442 |
{
|
1443 |
// Integer Impact Parameter Size (half of image)
|
1444 |
uint bmaxi=(uint)Size;
|
1445 |
|
1446 |
float Trajectory[TRACKPOINTS];
|
1447 |
|
1448 |
// Perform trajectory for each pixel
|
1449 |
|
1450 |
float m,rs,ri,re,tho;
|
1451 |
int raie,q;
|
1452 |
|
1453 |
m=Mass;
|
1454 |
rs=2.e0f*m;
|
1455 |
ri=InternalRadius;
|
1456 |
re=ExternalRadius;
|
1457 |
tho=Angle;
|
1458 |
q=-2;
|
1459 |
raie=Line;
|
1460 |
|
1461 |
float bmx,db,b,h;
|
1462 |
int nh;
|
1463 |
|
1464 |
// Autosize for image
|
1465 |
bmx=1.25e0f*re;
|
1466 |
|
1467 |
// Angular step of integration
|
1468 |
h=4.e0f*PI/(float)TRACKPOINTS;
|
1469 |
|
1470 |
// Integer Impact Parameter ID
|
1471 |
for (int bi=0;bi<bmaxi;bi++)
|
1472 |
{
|
1473 |
// impact parameter
|
1474 |
b=(float)bi/(float)bmaxi*bmx;
|
1475 |
db=bmx/(2.e0f*(float)bmaxi);
|
1476 |
|
1477 |
float up,vp,pp,us,vs,ps;
|
1478 |
|
1479 |
up=0.;
|
1480 |
vp=1.;
|
1481 |
pp=0.;
|
1482 |
nh=0;
|
1483 |
|
1484 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1485 |
|
1486 |
// b versus us
|
1487 |
float bvus=fabs(b/us);
|
1488 |
float bvus0=bvus;
|
1489 |
Trajectory[nh]=bvus;
|
1490 |
|
1491 |
do
|
1492 |
{
|
1493 |
nh++;
|
1494 |
pp=ps;
|
1495 |
up=us;
|
1496 |
vp=vs;
|
1497 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
1498 |
bvus=fabs(b/us);
|
1499 |
Trajectory[nh]=bvus;
|
1500 |
|
1501 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
1502 |
|
1503 |
for (uint i=(uint)nh+1;i<TRACKPOINTS;i++) {
|
1504 |
Trajectory[i]=0.e0f;
|
1505 |
}
|
1506 |
|
1507 |
int imx=(int)(16*bi);
|
1508 |
|
1509 |
for (int i=0;i<imx;i++)
|
1510 |
{
|
1511 |
float zp=0,fp=0;
|
1512 |
float phi=2.e0f*PI/(float)imx*(float)i;
|
1513 |
float phd=atanp(cos(phi)*sin(tho),cos(tho));
|
1514 |
uint yi=(uint)((float)bi*sin(phi)+bmaxi);
|
1515 |
uint xi=(uint)((float)bi*cos(phi)+bmaxi);
|
1516 |
|
1517 |
int HalfLap=0,ExitOnImpact=0,ni;
|
1518 |
float php,nr,r;
|
1519 |
|
1520 |
do
|
1521 |
{
|
1522 |
php=phd+(float)HalfLap*PI;
|
1523 |
nr=php/h;
|
1524 |
ni=(int)nr;
|
1525 |
|
1526 |
if (ni<nh)
|
1527 |
{
|
1528 |
r=(Trajectory[ni+1]-Trajectory[ni])*(nr-ni*1.)+Trajectory[ni];
|
1529 |
}
|
1530 |
else
|
1531 |
{
|
1532 |
r=Trajectory[ni];
|
1533 |
}
|
1534 |
|
1535 |
if ((r<=re)&&(r>=ri))
|
1536 |
{
|
1537 |
ExitOnImpact=1;
|
1538 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
1539 |
}
|
1540 |
|
1541 |
HalfLap++;
|
1542 |
|
1543 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
1544 |
|
1545 |
zImage[yi+2*bmaxi*xi]=zp;
|
1546 |
fImage[yi+2*bmaxi*xi]=fp;
|
1547 |
|
1548 |
}
|
1549 |
|
1550 |
}
|
1551 |
|
1552 |
}
|
1553 |
"""
|
1554 |
return(BlobCUDA)
|
1555 |
|
1556 |
# def ImageOutput(sigma,prefix):
|
1557 |
# from PIL import Image
|
1558 |
# Max=sigma.max()
|
1559 |
# Min=sigma.min()
|
1560 |
# # Normalize value as 8bits Integer
|
1561 |
# SigmaInt=(255*(sigma-Min)/(Max-Min)).astype('uint8')
|
1562 |
# image = Image.fromarray(SigmaInt)
|
1563 |
# image.save("%s.jpg" % prefix)
|
1564 |
|
1565 |
def ImageOutput(sigma,prefix,Colors): |
1566 |
import matplotlib.pyplot as plt |
1567 |
start_time=time.time() |
1568 |
if Colors == 'Red2Yellow': |
1569 |
plt.imsave("%s.png" % prefix, sigma, cmap='afmhot') |
1570 |
else:
|
1571 |
plt.imsave("%s.png" % prefix, sigma, cmap='Greys_r') |
1572 |
save_time = time.time()-start_time |
1573 |
print("Save image as %s.png file" % prefix)
|
1574 |
print("Save Time : %f" % save_time)
|
1575 |
|
1576 |
def BlackHoleCL(zImage,fImage,InputCL): |
1577 |
|
1578 |
kernel_params = {} |
1579 |
|
1580 |
print(InputCL) |
1581 |
|
1582 |
Device=InputCL['Device']
|
1583 |
GpuStyle=InputCL['GpuStyle']
|
1584 |
VariableType=InputCL['VariableType']
|
1585 |
Size=InputCL['Size']
|
1586 |
Mass=InputCL['Mass']
|
1587 |
InternalRadius=InputCL['InternalRadius']
|
1588 |
ExternalRadius=InputCL['ExternalRadius']
|
1589 |
Angle=InputCL['Angle']
|
1590 |
Method=InputCL['Method']
|
1591 |
TrackPoints=InputCL['TrackPoints']
|
1592 |
Physics=InputCL['Physics']
|
1593 |
|
1594 |
PhysicsList=DictionariesAPI() |
1595 |
|
1596 |
if InputCL['BlackBody']: |
1597 |
# Spectrum is Black Body one
|
1598 |
Line=0
|
1599 |
else:
|
1600 |
# Spectrum is Monochromatic Line one
|
1601 |
Line=1
|
1602 |
|
1603 |
Trajectories=numpy.zeros((int(InputCL['Size']/2),InputCL['TrackPoints']),dtype=numpy.float32) |
1604 |
IdLast=numpy.zeros(int(InputCL['Size']/2),dtype=numpy.int32) |
1605 |
|
1606 |
# Je detecte un peripherique GPU dans la liste des peripheriques
|
1607 |
Id=0
|
1608 |
HasXPU=False
|
1609 |
for platform in cl.get_platforms(): |
1610 |
for device in platform.get_devices(): |
1611 |
if Id==Device:
|
1612 |
PF4XPU=platform.name |
1613 |
XPU=device |
1614 |
print("CPU/GPU selected: ",device.name.lstrip())
|
1615 |
HasXPU=True
|
1616 |
Id+=1
|
1617 |
|
1618 |
if HasXPU==False: |
1619 |
print("No XPU #%i found in all of %i devices, sorry..." % (Device,Id-1)) |
1620 |
sys.exit() |
1621 |
|
1622 |
ctx = cl.Context([XPU]) |
1623 |
queue = cl.CommandQueue(ctx, |
1624 |
properties=cl.command_queue_properties.PROFILING_ENABLE) |
1625 |
|
1626 |
|
1627 |
# BlackHoleCL = cl.Program(ctx,KERNEL_CODE.substitute(kernel_params)).build()
|
1628 |
|
1629 |
BuildOptions="-DPHYSICS=%i -DSETTRACKPOINTS=%i " % (PhysicsList[Physics],InputCL['TrackPoints']) |
1630 |
|
1631 |
print('My Platform is ',PF4XPU)
|
1632 |
|
1633 |
if 'Intel' in PF4XPU or 'Experimental' in PF4XPU or 'Clover' in PF4XPU or 'Portable' in PF4XPU : |
1634 |
print('No extra options for Intel and Clover!')
|
1635 |
else:
|
1636 |
BuildOptions = BuildOptions+" -cl-mad-enable"
|
1637 |
|
1638 |
print(BuildOptions) |
1639 |
|
1640 |
BlackHoleCL = cl.Program(ctx,BlobOpenCL).build(options = BuildOptions) |
1641 |
|
1642 |
# Je recupere les flag possibles pour les buffers
|
1643 |
mf = cl.mem_flags |
1644 |
|
1645 |
if Method=='TrajectoPixel' or Method=='TrajectoCircle': |
1646 |
TrajectoriesCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=Trajectories) |
1647 |
IdLastCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=IdLast) |
1648 |
|
1649 |
zImageCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=zImage) |
1650 |
fImageCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=fImage) |
1651 |
|
1652 |
start_time=time.time() |
1653 |
|
1654 |
if Method=='EachPixel': |
1655 |
CLLaunch=BlackHoleCL.EachPixel(queue,(zImage.shape[0],zImage.shape[1]), |
1656 |
None,zImageCL,fImageCL,
|
1657 |
numpy.float32(Mass), |
1658 |
numpy.float32(InternalRadius), |
1659 |
numpy.float32(ExternalRadius), |
1660 |
numpy.float32(Angle), |
1661 |
numpy.int32(Line)) |
1662 |
CLLaunch.wait() |
1663 |
elif Method=='Original': |
1664 |
CLLaunch=BlackHoleCL.Original(queue,(1,),
|
1665 |
None,zImageCL,fImageCL,
|
1666 |
numpy.uint32(zImage.shape[0]/2), |
1667 |
numpy.float32(Mass), |
1668 |
numpy.float32(InternalRadius), |
1669 |
numpy.float32(ExternalRadius), |
1670 |
numpy.float32(Angle), |
1671 |
numpy.int32(Line)) |
1672 |
CLLaunch.wait() |
1673 |
elif Method=='EachCircle': |
1674 |
CLLaunch=BlackHoleCL.EachCircle(queue,(zImage.shape[0]/2,), |
1675 |
None,zImageCL,fImageCL,
|
1676 |
numpy.float32(Mass), |
1677 |
numpy.float32(InternalRadius), |
1678 |
numpy.float32(ExternalRadius), |
1679 |
numpy.float32(Angle), |
1680 |
numpy.int32(Line)) |
1681 |
CLLaunch.wait() |
1682 |
elif Method=='TrajectoCircle': |
1683 |
CLLaunch=BlackHoleCL.Trajectory(queue,(Trajectories.shape[0],),
|
1684 |
None,TrajectoriesCL,IdLastCL,
|
1685 |
numpy.float32(Mass), |
1686 |
numpy.float32(InternalRadius), |
1687 |
numpy.float32(ExternalRadius), |
1688 |
numpy.float32(Angle), |
1689 |
numpy.int32(Line)) |
1690 |
|
1691 |
CLLaunch=BlackHoleCL.Circle(queue,(Trajectories.shape[0],
|
1692 |
zImage.shape[0]*4),None, |
1693 |
TrajectoriesCL,IdLastCL, |
1694 |
zImageCL,fImageCL, |
1695 |
numpy.float32(Mass), |
1696 |
numpy.float32(InternalRadius), |
1697 |
numpy.float32(ExternalRadius), |
1698 |
numpy.float32(Angle), |
1699 |
numpy.int32(Line)) |
1700 |
CLLaunch.wait() |
1701 |
else:
|
1702 |
CLLaunch=BlackHoleCL.Trajectory(queue,(Trajectories.shape[0],),
|
1703 |
None,TrajectoriesCL,IdLastCL,
|
1704 |
numpy.float32(Mass), |
1705 |
numpy.float32(InternalRadius), |
1706 |
numpy.float32(ExternalRadius), |
1707 |
numpy.float32(Angle), |
1708 |
numpy.int32(Line)) |
1709 |
|
1710 |
CLLaunch=BlackHoleCL.Pixel(queue,(zImage.shape[0],zImage.shape[1]),None, |
1711 |
zImageCL,fImageCL,TrajectoriesCL,IdLastCL, |
1712 |
numpy.uint32(Trajectories.shape[0]),
|
1713 |
numpy.float32(Mass), |
1714 |
numpy.float32(InternalRadius), |
1715 |
numpy.float32(ExternalRadius), |
1716 |
numpy.float32(Angle), |
1717 |
numpy.int32(Line)) |
1718 |
CLLaunch.wait() |
1719 |
|
1720 |
compute = time.time()-start_time |
1721 |
|
1722 |
cl.enqueue_copy(queue,zImage,zImageCL).wait() |
1723 |
cl.enqueue_copy(queue,fImage,fImageCL).wait() |
1724 |
if Method=='TrajectoPixel' or Method=='TrajectoCircle': |
1725 |
cl.enqueue_copy(queue,Trajectories,TrajectoriesCL).wait() |
1726 |
cl.enqueue_copy(queue,IdLast,IdLastCL).wait() |
1727 |
elapsed = time.time()-start_time |
1728 |
print("\nCompute Time : %f" % compute)
|
1729 |
print("Elapsed Time : %f\n" % elapsed)
|
1730 |
|
1731 |
zMaxPosition=numpy.where(zImage[:,:]==zImage.max()) |
1732 |
fMaxPosition=numpy.where(fImage[:,:]==fImage.max()) |
1733 |
print("Z max @(%i,%i) : %f" % (zMaxPosition[1][0],zMaxPosition[0][0],zImage.max())) |
1734 |
print("Flux max @(%i,%i) : %f\n" % (fMaxPosition[1][0],fMaxPosition[0][0],fImage.max())) |
1735 |
zImageCL.release() |
1736 |
fImageCL.release() |
1737 |
|
1738 |
if Method=='TrajectoPixel' or Method=='TrajectoCircle': |
1739 |
TrajectoriesCL.release() |
1740 |
IdLastCL.release() |
1741 |
|
1742 |
return(elapsed)
|
1743 |
|
1744 |
def BlackHoleCUDA(zImage,fImage,InputCL): |
1745 |
|
1746 |
kernel_params = {} |
1747 |
|
1748 |
print(InputCL) |
1749 |
|
1750 |
Device=InputCL['Device']
|
1751 |
GpuStyle=InputCL['GpuStyle']
|
1752 |
VariableType=InputCL['VariableType']
|
1753 |
Size=InputCL['Size']
|
1754 |
Mass=InputCL['Mass']
|
1755 |
InternalRadius=InputCL['InternalRadius']
|
1756 |
ExternalRadius=InputCL['ExternalRadius']
|
1757 |
Angle=InputCL['Angle']
|
1758 |
Method=InputCL['Method']
|
1759 |
TrackPoints=InputCL['TrackPoints']
|
1760 |
Physics=InputCL['Physics']
|
1761 |
Threads=InputCL['Threads']
|
1762 |
|
1763 |
PhysicsList=DictionariesAPI() |
1764 |
|
1765 |
if InputCL['BlackBody']: |
1766 |
# Spectrum is Black Body one
|
1767 |
Line=0
|
1768 |
else:
|
1769 |
# Spectrum is Monochromatic Line one
|
1770 |
Line=1
|
1771 |
|
1772 |
Trajectories=numpy.zeros((int(InputCL['Size']/2),InputCL['TrackPoints']),dtype=numpy.float32) |
1773 |
IdLast=numpy.zeros(int(InputCL['Size']/2),dtype=numpy.int32) |
1774 |
|
1775 |
try:
|
1776 |
# For PyCUDA import
|
1777 |
import pycuda.driver as cuda |
1778 |
from pycuda.compiler import SourceModule |
1779 |
|
1780 |
cuda.init() |
1781 |
for Id in range(cuda.Device.count()): |
1782 |
if Id==Device:
|
1783 |
XPU=cuda.Device(Id) |
1784 |
print("GPU selected %s" % XPU.name())
|
1785 |
print
|
1786 |
|
1787 |
except ImportError: |
1788 |
print("Platform does not seem to support CUDA")
|
1789 |
|
1790 |
Context=XPU.make_context() |
1791 |
|
1792 |
try:
|
1793 |
mod = SourceModule(KernelCodeCuda(),options=['--compiler-options','-DPHYSICS=%i -DSETTRACKPOINTS=%i' % (PhysicsList[Physics],TrackPoints)]) |
1794 |
print("Compilation seems to be OK")
|
1795 |
except:
|
1796 |
print("Compilation seems to break")
|
1797 |
|
1798 |
EachPixelCU=mod.get_function("EachPixel")
|
1799 |
OriginalCU=mod.get_function("Original")
|
1800 |
EachCircleCU=mod.get_function("EachCircle")
|
1801 |
TrajectoryCU=mod.get_function("Trajectory")
|
1802 |
PixelCU=mod.get_function("Pixel")
|
1803 |
CircleCU=mod.get_function("Circle")
|
1804 |
|
1805 |
TrajectoriesCU = cuda.mem_alloc(Trajectories.size*Trajectories.dtype.itemsize) |
1806 |
cuda.memcpy_htod(TrajectoriesCU, Trajectories) |
1807 |
zImageCU = cuda.mem_alloc(zImage.size*zImage.dtype.itemsize) |
1808 |
cuda.memcpy_htod(zImageCU, zImage) |
1809 |
fImageCU = cuda.mem_alloc(fImage.size*fImage.dtype.itemsize) |
1810 |
cuda.memcpy_htod(zImageCU, fImage) |
1811 |
IdLastCU = cuda.mem_alloc(IdLast.size*IdLast.dtype.itemsize) |
1812 |
cuda.memcpy_htod(IdLastCU, IdLast) |
1813 |
|
1814 |
start_time=time.time() |
1815 |
|
1816 |
if Method=='EachPixel': |
1817 |
EachPixelCU(zImageCU,fImageCU, |
1818 |
numpy.float32(Mass), |
1819 |
numpy.float32(InternalRadius), |
1820 |
numpy.float32(ExternalRadius), |
1821 |
numpy.float32(Angle), |
1822 |
numpy.int32(Line), |
1823 |
grid=(zImage.shape[0]/32,zImage.shape[1]/Threads), |
1824 |
block=(Threads,Threads,1))
|
1825 |
elif Method=='EachCircle': |
1826 |
EachCircleCU(zImageCU,fImageCU, |
1827 |
numpy.float32(Mass), |
1828 |
numpy.float32(InternalRadius), |
1829 |
numpy.float32(ExternalRadius), |
1830 |
numpy.float32(Angle), |
1831 |
numpy.int32(Line), |
1832 |
grid=(zImage.shape[0]/Threads/2,1), |
1833 |
block=(Threads,1,1)) |
1834 |
elif Method=='Original': |
1835 |
OriginalCU(zImageCU,fImageCU, |
1836 |
numpy.uint32(zImage.shape[0]/2), |
1837 |
numpy.float32(Mass), |
1838 |
numpy.float32(InternalRadius), |
1839 |
numpy.float32(ExternalRadius), |
1840 |
numpy.float32(Angle), |
1841 |
numpy.int32(Line), |
1842 |
grid=(1,1), |
1843 |
block=(1,1,1)) |
1844 |
elif Method=='TrajectoCircle': |
1845 |
TrajectoryCU(TrajectoriesCU,IdLastCU, |
1846 |
numpy.float32(Mass), |
1847 |
numpy.float32(InternalRadius), |
1848 |
numpy.float32(ExternalRadius), |
1849 |
numpy.float32(Angle), |
1850 |
numpy.int32(Line), |
1851 |
grid=(Trajectories.shape[0]/Threads,1), |
1852 |
block=(Threads,1,1)) |
1853 |
|
1854 |
CircleCU(TrajectoriesCU,IdLastCU,zImageCU,fImageCU, |
1855 |
numpy.float32(Mass), |
1856 |
numpy.float32(InternalRadius), |
1857 |
numpy.float32(ExternalRadius), |
1858 |
numpy.float32(Angle), |
1859 |
numpy.int32(Line), |
1860 |
grid=(Trajectories.shape[0]/Threads,zImage.shape[0]*4/Threads), |
1861 |
block=(Threads,Threads,1))
|
1862 |
else:
|
1863 |
TrajectoryCU(TrajectoriesCU,IdLastCU, |
1864 |
numpy.float32(Mass), |
1865 |
numpy.float32(InternalRadius), |
1866 |
numpy.float32(ExternalRadius), |
1867 |
numpy.float32(Angle), |
1868 |
numpy.int32(Line), |
1869 |
grid=(Trajectories.shape[0]/Threads,1), |
1870 |
block=(Threads,1,1)) |
1871 |
|
1872 |
PixelCU(zImageCU,fImageCU,TrajectoriesCU,IdLastCU, |
1873 |
numpy.uint32(Trajectories.shape[0]),
|
1874 |
numpy.float32(Mass), |
1875 |
numpy.float32(InternalRadius), |
1876 |
numpy.float32(ExternalRadius), |
1877 |
numpy.float32(Angle), |
1878 |
numpy.int32(Line), |
1879 |
grid=(zImage.shape[0]/Threads,zImage.shape[1]/Threads,1), |
1880 |
block=(Threads,Threads,1))
|
1881 |
|
1882 |
Context.synchronize() |
1883 |
|
1884 |
compute = time.time()-start_time |
1885 |
|
1886 |
cuda.memcpy_dtoh(zImage,zImageCU) |
1887 |
cuda.memcpy_dtoh(fImage,fImageCU) |
1888 |
elapsed = time.time()-start_time |
1889 |
print("\nCompute Time : %f" % compute)
|
1890 |
print("Elapsed Time : %f\n" % elapsed)
|
1891 |
|
1892 |
zMaxPosition=numpy.where(zImage[:,:]==zImage.max()) |
1893 |
fMaxPosition=numpy.where(fImage[:,:]==fImage.max()) |
1894 |
print("Z max @(%i,%i) : %f" % (zMaxPosition[1][0],zMaxPosition[0][0],zImage.max())) |
1895 |
print("Flux max @(%i,%i) : %f\n" % (fMaxPosition[1][0],fMaxPosition[0][0],fImage.max())) |
1896 |
|
1897 |
|
1898 |
Context.pop() |
1899 |
|
1900 |
Context.detach() |
1901 |
|
1902 |
return(elapsed)
|
1903 |
|
1904 |
if __name__=='__main__': |
1905 |
|
1906 |
GpuStyle = 'OpenCL'
|
1907 |
Mass = 1.
|
1908 |
# Internal Radius 3 times de Schwarzschild Radius
|
1909 |
InternalRadius=6.*Mass
|
1910 |
#
|
1911 |
ExternalRadius=12.
|
1912 |
#
|
1913 |
# Angle with normal to disc 10 degrees
|
1914 |
Angle = numpy.pi/180.*(90.-10.) |
1915 |
# Radiation of disc : BlackBody or Monochromatic
|
1916 |
BlackBody = False
|
1917 |
# Size of image
|
1918 |
Size=256
|
1919 |
# Variable Type
|
1920 |
VariableType='FP32'
|
1921 |
# ?
|
1922 |
q=-2
|
1923 |
# Method of resolution
|
1924 |
Method='TrajectoPixel'
|
1925 |
# Colors for output image
|
1926 |
Colors='Greyscale'
|
1927 |
# Physics
|
1928 |
Physics='Einstein'
|
1929 |
# No output as image
|
1930 |
NoImage = False
|
1931 |
# Threads in CUDA
|
1932 |
Threads = 32
|
1933 |
|
1934 |
HowToUse='%s -h [Help] -b [BlackBodyEmission] -n [NoImage] -p <Einstein/Newton> -s <SizeInPixels> -m <Mass> -i <DiscInternalRadius> -x <DiscExternalRadius> -a <AngleAboveDisc> -d <DeviceId> -c <Greyscale/Red2Yellow> -g <CUDA/OpenCL> -o <EachPixel/TrajectoCircle/TrajectoPixel/EachCircle/Original> -t <ThreadsInCuda> -v <FP32/FP64>'
|
1935 |
|
1936 |
try:
|
1937 |
opts, args = getopt.getopt(sys.argv[1:],"hbns:m:i:x:a:d:g:v:o:t:c:p:",["blackbody","noimage","camera","size=","mass=","internal=","external=","angle=","device=","gpustyle=","variabletype=","method=","threads=","colors=","physics="]) |
1938 |
except getopt.GetoptError:
|
1939 |
print(HowToUse % sys.argv[0])
|
1940 |
sys.exit(2)
|
1941 |
|
1942 |
# List of Devices
|
1943 |
Devices=[] |
1944 |
Alu={} |
1945 |
|
1946 |
for opt, arg in opts: |
1947 |
if opt == '-h': |
1948 |
print(HowToUse % sys.argv[0])
|
1949 |
|
1950 |
print("\nInformations about devices detected under OpenCL API:")
|
1951 |
# For PyOpenCL import
|
1952 |
try:
|
1953 |
import pyopencl as cl |
1954 |
Id=0
|
1955 |
for platform in cl.get_platforms(): |
1956 |
for device in platform.get_devices(): |
1957 |
#deviceType=cl.device_type.to_string(device.type)
|
1958 |
deviceType="xPU"
|
1959 |
print("Device #%i from %s of type %s : %s" % (Id,platform.vendor.lstrip(),deviceType,device.name.lstrip()))
|
1960 |
Id=Id+1
|
1961 |
|
1962 |
except:
|
1963 |
print("Your platform does not seem to support OpenCL")
|
1964 |
|
1965 |
print("\nInformations about devices detected under CUDA API:")
|
1966 |
# For PyCUDA import
|
1967 |
try:
|
1968 |
import pycuda.driver as cuda |
1969 |
cuda.init() |
1970 |
for Id in range(cuda.Device.count()): |
1971 |
device=cuda.Device(Id) |
1972 |
print("Device #%i of type GPU : %s" % (Id,device.name()))
|
1973 |
print
|
1974 |
except:
|
1975 |
print("Your platform does not seem to support CUDA")
|
1976 |
|
1977 |
sys.exit() |
1978 |
|
1979 |
elif opt in ("-d", "--device"): |
1980 |
# Devices.append(int(arg))
|
1981 |
Device=int(arg)
|
1982 |
elif opt in ("-g", "--gpustyle"): |
1983 |
GpuStyle = arg |
1984 |
elif opt in ("-v", "--variabletype"): |
1985 |
VariableType = arg |
1986 |
elif opt in ("-s", "--size"): |
1987 |
Size = int(arg)
|
1988 |
elif opt in ("-m", "--mass"): |
1989 |
Mass = float(arg)
|
1990 |
elif opt in ("-i", "--internal"): |
1991 |
InternalRadius = float(arg)
|
1992 |
elif opt in ("-e", "--external"): |
1993 |
ExternalRadius = float(arg)
|
1994 |
elif opt in ("-a", "--angle"): |
1995 |
Angle = numpy.pi/180.*(90.-float(arg)) |
1996 |
elif opt in ("-b", "--blackbody"): |
1997 |
BlackBody = True
|
1998 |
elif opt in ("-n", "--noimage"): |
1999 |
NoImage = True
|
2000 |
elif opt in ("-o", "--method"): |
2001 |
Method = arg |
2002 |
elif opt in ("-t", "--threads"): |
2003 |
Threads = int(arg)
|
2004 |
elif opt in ("-c", "--colors"): |
2005 |
Colors = arg |
2006 |
elif opt in ("-p", "--physics"): |
2007 |
Physics = arg |
2008 |
|
2009 |
print("Device Identification selected : %s" % Device)
|
2010 |
print("GpuStyle used : %s" % GpuStyle)
|
2011 |
print("VariableType : %s" % VariableType)
|
2012 |
print("Size : %i" % Size)
|
2013 |
print("Mass : %f" % Mass)
|
2014 |
print("Internal Radius : %f" % InternalRadius)
|
2015 |
print("External Radius : %f" % ExternalRadius)
|
2016 |
print("Angle with normal of (in radians) : %f" % Angle)
|
2017 |
print("Black Body Disc Emission (monochromatic instead) : %s" % BlackBody)
|
2018 |
print("Method of resolution : %s" % Method)
|
2019 |
print("Colors for output images : %s" % Colors)
|
2020 |
print("Physics used for Trajectories : %s" % Physics)
|
2021 |
|
2022 |
if GpuStyle=='CUDA': |
2023 |
print("\nSelection of CUDA device")
|
2024 |
try:
|
2025 |
# For PyCUDA import
|
2026 |
import pycuda.driver as cuda |
2027 |
|
2028 |
cuda.init() |
2029 |
for Id in range(cuda.Device.count()): |
2030 |
device=cuda.Device(Id) |
2031 |
print("Device #%i of type GPU : %s" % (Id,device.name()))
|
2032 |
if Id in Devices: |
2033 |
Alu[Id]='GPU'
|
2034 |
|
2035 |
except ImportError: |
2036 |
print("Platform does not seem to support CUDA")
|
2037 |
|
2038 |
if GpuStyle=='OpenCL': |
2039 |
print("\nSelection of OpenCL device")
|
2040 |
try:
|
2041 |
# For PyOpenCL import
|
2042 |
import pyopencl as cl |
2043 |
Id=0
|
2044 |
for platform in cl.get_platforms(): |
2045 |
for device in platform.get_devices(): |
2046 |
#deviceType=cl.device_type.to_string(device.type)
|
2047 |
deviceType="xPU"
|
2048 |
print("Device #%i from %s of type %s : %s" % (Id,platform.vendor.lstrip().rstrip(),deviceType,device.name.lstrip().rstrip()))
|
2049 |
|
2050 |
if Id in Devices: |
2051 |
# Set the Alu as detected Device Type
|
2052 |
Alu[Id]=deviceType |
2053 |
Id=Id+1
|
2054 |
except ImportError: |
2055 |
print("Platform does not seem to support OpenCL")
|
2056 |
|
2057 |
# print(Devices,Alu)
|
2058 |
|
2059 |
# MyImage=numpy.where(numpy.random.zeros(Size,Size)>0,1,-1).astype(numpy.float32)
|
2060 |
TrackPoints=2048
|
2061 |
zImage=numpy.zeros((Size,Size),dtype=numpy.float32) |
2062 |
fImage=numpy.zeros((Size,Size),dtype=numpy.float32) |
2063 |
|
2064 |
InputCL={} |
2065 |
InputCL['Device']=Device
|
2066 |
InputCL['GpuStyle']=GpuStyle
|
2067 |
InputCL['VariableType']=VariableType
|
2068 |
InputCL['Size']=Size
|
2069 |
InputCL['Mass']=Mass
|
2070 |
InputCL['InternalRadius']=InternalRadius
|
2071 |
InputCL['ExternalRadius']=ExternalRadius
|
2072 |
InputCL['Angle']=Angle
|
2073 |
InputCL['BlackBody']=BlackBody
|
2074 |
InputCL['Method']=Method
|
2075 |
InputCL['TrackPoints']=TrackPoints
|
2076 |
InputCL['Physics']=Physics
|
2077 |
InputCL['Threads']=Threads
|
2078 |
|
2079 |
if GpuStyle=='OpenCL': |
2080 |
duration=BlackHoleCL(zImage,fImage,InputCL) |
2081 |
else:
|
2082 |
duration=BlackHoleCUDA(zImage,fImage,InputCL) |
2083 |
|
2084 |
Hostname=gethostname() |
2085 |
Date=time.strftime("%Y%m%d_%H%M%S")
|
2086 |
ImageInfo="%s_Device%i_%s_%s" % (Method,Device,Hostname,Date)
|
2087 |
|
2088 |
|
2089 |
if not NoImage: |
2090 |
ImageOutput(zImage,"TrouNoirZ_%s" % ImageInfo,Colors)
|
2091 |
ImageOutput(fImage,"TrouNoirF_%s" % ImageInfo,Colors)
|