root / TrouNoir / TrouNoir.py @ 208
Historique | Voir | Annoter | Télécharger (23,65 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 |
#
|
18 |
# Thanks to :
|
19 |
#
|
20 |
# - Herve Aussel for his part of code of black body spectrum
|
21 |
# - Didier Pelat for his help to perform this work
|
22 |
# - Jean-Pierre Luminet for his article published in 1979
|
23 |
# - Numerical Recipies for Runge Kutta recipies
|
24 |
# - Luc Blanchet for his disponibility about my questions in General Relativity
|
25 |
# - Pierre Lena for his passion about science and vulgarisation
|
26 |
|
27 |
import pyopencl as cl |
28 |
import numpy |
29 |
from PIL import Image |
30 |
import time,string |
31 |
from numpy.random import randint as nprnd |
32 |
import sys |
33 |
import getopt |
34 |
import matplotlib.pyplot as plt |
35 |
|
36 |
|
37 |
|
38 |
|
39 |
|
40 |
|
41 |
|
42 |
|
43 |
KERNEL_CODE=string.Template("""
|
44 |
|
45 |
#define PI (float)3.14159265359
|
46 |
#define nbr 200
|
47 |
|
48 |
float atanp(float x,float y)
|
49 |
{
|
50 |
float angle;
|
51 |
|
52 |
angle=atan2(y,x);
|
53 |
|
54 |
if (angle<0.e0f)
|
55 |
{
|
56 |
angle+=(float)2.e0f*PI;
|
57 |
}
|
58 |
|
59 |
return angle;
|
60 |
}
|
61 |
|
62 |
float f(float v)
|
63 |
{
|
64 |
return v;
|
65 |
}
|
66 |
|
67 |
float g(float u,float m,float b)
|
68 |
{
|
69 |
return (3.e0f*m/b*pow(u,2)-u);
|
70 |
}
|
71 |
|
72 |
|
73 |
void calcul(float *us,float *vs,float up,float vp,
|
74 |
float h,float m,float b)
|
75 |
{
|
76 |
float c0,c1,c2,c3,d0,d1,d2,d3;
|
77 |
|
78 |
c0=h*f(vp);
|
79 |
c1=h*f(vp+c0/2.);
|
80 |
c2=h*f(vp+c1/2.);
|
81 |
c3=h*f(vp+c2);
|
82 |
d0=h*g(up,m,b);
|
83 |
d1=h*g(up+d0/2.,m,b);
|
84 |
d2=h*g(up+d1/2.,m,b);
|
85 |
d3=h*g(up+d2,m,b);
|
86 |
|
87 |
*us=up+(c0+2.*c1+2.*c2+c3)/6.;
|
88 |
*vs=vp+(d0+2.*d1+2.*d2+d3)/6.;
|
89 |
}
|
90 |
|
91 |
void rungekutta(float *ps,float *us,float *vs,
|
92 |
float pp,float up,float vp,
|
93 |
float h,float m,float b)
|
94 |
{
|
95 |
calcul(us,vs,up,vp,h,m,b);
|
96 |
*ps=pp+h;
|
97 |
}
|
98 |
|
99 |
float decalage_spectral(float r,float b,float phi,
|
100 |
float tho,float m)
|
101 |
{
|
102 |
return (sqrt(1-3*m/r)/(1+sqrt(m/pow(r,3))*b*sin(tho)*sin(phi)));
|
103 |
}
|
104 |
|
105 |
float spectre(float rf,int q,float b,float db,
|
106 |
float h,float r,float m,float bss)
|
107 |
{
|
108 |
float flx;
|
109 |
|
110 |
flx=pow(r/m,q)*pow(rf,4)*b;
|
111 |
return(flx);
|
112 |
}
|
113 |
|
114 |
float spectre_cn(float rf,float b,float db,float h,float r,float m,float bss)
|
115 |
{
|
116 |
double flx;
|
117 |
double nu_rec,nu_em,qu,v,temp,temp_em,flux_int,m_point,planck,c,k,psc2,psk;
|
118 |
int fi,posfreq;
|
119 |
|
120 |
planck=6.62e-34;
|
121 |
k=1.38e-23;
|
122 |
psk=5.38e-11;
|
123 |
psc2=7.35e-51;
|
124 |
temp=3.e7;
|
125 |
m_point=1.e14;
|
126 |
v=1.-3./r;
|
127 |
|
128 |
qu=1./sqrt(1.-3./r)/sqrt(r)*(sqrt(r)-sqrt(6.)+sqrt(3.)/2.*log((sqrt(r)+sqrt(3.))/(sqrt(r)-sqrt(3.))*(sqrt(6.)-sqrt(3.))/(sqrt(6.)+sqrt(3.))));
|
129 |
|
130 |
temp_em=temp*sqrt(m)*exp(0.25*log(m_point))*exp(-0.75*log(r))*exp(-0.125*log(v))*exp(0.25*log(qu));
|
131 |
|
132 |
flux_int=0;
|
133 |
flx=0;
|
134 |
|
135 |
for (fi=0;fi<nbr;fi++)
|
136 |
{
|
137 |
nu_em=bss*(float)fi/(float)nbr;
|
138 |
nu_rec=nu_em*rf;
|
139 |
posfreq=(int)(1./bss*nu_rec*(float)nbr);
|
140 |
if ((posfreq>0)&&(posfreq<nbr))
|
141 |
{
|
142 |
//flux_int=2*planck/9.e16f*pow(nu_em,3)/(exp(planck*nu_em/k/temp_em)-1.)*pow(rf,3)*b*db*h;
|
143 |
//flux_int=pow(nu_em,3)/(exp(planck*nu_em/k/temp_em)-1.)*pow(rf,3)*b;
|
144 |
flux_int=2.*psc2*pow(nu_em,3)/(exp(psk*nu_em/temp_em)-1.)*pow(rf,3)*b;
|
145 |
flx+=flux_int;
|
146 |
}
|
147 |
}
|
148 |
//return(pow(rf,3)*b*temp_em);
|
149 |
return((float)flx);
|
150 |
return(flx);
|
151 |
}
|
152 |
|
153 |
void impact(float phi,float r,float b,float tho,float m,
|
154 |
float *zp,float *fp,
|
155 |
int q,float db,
|
156 |
float h,int raie)
|
157 |
{
|
158 |
float flx,rf,bss;
|
159 |
|
160 |
rf=decalage_spectral(r,b,phi,tho,m);
|
161 |
|
162 |
if (raie==0)
|
163 |
{
|
164 |
bss=2.;
|
165 |
flx=spectre(rf,q,b,db,h,r,m,bss);
|
166 |
}
|
167 |
else
|
168 |
{
|
169 |
bss=1.e6;
|
170 |
flx=spectre_cn(rf,b,db,h,r,m,bss);
|
171 |
}
|
172 |
|
173 |
*zp=1./rf;
|
174 |
*fp=flx;
|
175 |
|
176 |
}
|
177 |
|
178 |
__kernel void EachPixel(__global float *zImage,__global float *fImage,
|
179 |
float Mass,float InternalRadius,
|
180 |
float ExternalRadius,float Angle,
|
181 |
float ObserverDistance,
|
182 |
int BlackBody,int AngularCamera)
|
183 |
{
|
184 |
uint xi=(uint)get_global_id(0);
|
185 |
uint yi=(uint)get_global_id(1);
|
186 |
uint sizex=(uint)get_global_size(0);
|
187 |
uint sizey=(uint)get_global_size(1);
|
188 |
|
189 |
// Perform trajectory for each pixel, exit on hit
|
190 |
|
191 |
float m,rs,ri,re,tho,ro;
|
192 |
int q,dist,raie;
|
193 |
|
194 |
m=Mass;
|
195 |
rs=2.*m;
|
196 |
ri=InternalRadius;
|
197 |
re=ExternalRadius;
|
198 |
ro=ObserverDistance;
|
199 |
tho=Angle;
|
200 |
q=-2;
|
201 |
dist=AngularCamera;
|
202 |
raie=BlackBody;
|
203 |
|
204 |
float d,bmx,db,b,h;
|
205 |
float rp[2048];
|
206 |
float phi,thi,phd,php,nr,r;
|
207 |
int nh;
|
208 |
float zp,fp;
|
209 |
|
210 |
// Autosize for image
|
211 |
bmx=1.25*re;
|
212 |
b=0.;
|
213 |
|
214 |
h=4.e0f*PI/(float)2048;
|
215 |
|
216 |
// set origin as center of image
|
217 |
float x=(float)xi-(float)(sizex/2)+(float)5e-1f;
|
218 |
float y=(float)yi-(float)(sizey/2)+(float)5e-1f;
|
219 |
// angle extracted from cylindric symmetry
|
220 |
phi=atanp(x,y);
|
221 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
222 |
|
223 |
float up,vp,pp,us,vs,ps;
|
224 |
|
225 |
// impact parameter
|
226 |
b=sqrt(x*x+y*y)*(float)2.e0f/(float)sizex*bmx;
|
227 |
// step of impact parameter;
|
228 |
db=bmx/(float)(sizex/2);
|
229 |
|
230 |
if (dist==1)
|
231 |
{
|
232 |
up=0.;
|
233 |
vp=1.;
|
234 |
}
|
235 |
else
|
236 |
{
|
237 |
up=sin(thi);
|
238 |
vp=cos(thi);
|
239 |
}
|
240 |
|
241 |
pp=0.;
|
242 |
nh=0;
|
243 |
|
244 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
245 |
|
246 |
rp[nh]=fabs(b/us);
|
247 |
|
248 |
int ExitOnImpact=0;
|
249 |
|
250 |
do
|
251 |
{
|
252 |
nh++;
|
253 |
pp=ps;
|
254 |
up=us;
|
255 |
vp=vs;
|
256 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
257 |
rp[nh]=fabs(b/us);
|
258 |
ExitOnImpact = ((fmod(pp,PI)<fmod(phd,PI))&&(fmod(ps,PI)>fmod(phd,PI)))&&(rp[nh]>ri)&&(rp[nh]<re)?1:0;
|
259 |
|
260 |
} while ((rp[nh]>=rs)&&(rp[nh]<=rp[0])&&(ExitOnImpact==0));
|
261 |
|
262 |
if (ExitOnImpact==1) {
|
263 |
impact(phi,rp[nh-1],b,tho,m,&zp,&fp,q,db,h,raie);
|
264 |
}
|
265 |
else
|
266 |
{
|
267 |
zp=0.;
|
268 |
fp=0.;
|
269 |
}
|
270 |
|
271 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
272 |
|
273 |
zImage[yi+sizex*xi]=(float)zp;
|
274 |
fImage[yi+sizex*xi]=(float)fp;
|
275 |
}
|
276 |
|
277 |
__kernel void Pixel(__global float *zImage,__global float *fImage,
|
278 |
__global float *Trajectories,__global int *IdLast,
|
279 |
uint ImpactParameter,uint TrackPoints,
|
280 |
float Mass,float InternalRadius,
|
281 |
float ExternalRadius,float Angle,
|
282 |
float ObserverDistance,
|
283 |
int BlackBody,int AngularCamera)
|
284 |
{
|
285 |
uint xi=(uint)get_global_id(0);
|
286 |
uint yi=(uint)get_global_id(1);
|
287 |
uint sizex=(uint)get_global_size(0);
|
288 |
uint sizey=(uint)get_global_size(1);
|
289 |
|
290 |
// Perform trajectory for each pixel
|
291 |
|
292 |
float m,rs,ri,re,tho,ro;
|
293 |
int q,dist,raie;
|
294 |
|
295 |
m=Mass;
|
296 |
rs=2.*m;
|
297 |
ri=InternalRadius;
|
298 |
re=ExternalRadius;
|
299 |
ro=ObserverDistance;
|
300 |
tho=Angle;
|
301 |
q=-2;
|
302 |
dist=AngularCamera;
|
303 |
raie=BlackBody;
|
304 |
|
305 |
float d,bmx,db,b,h;
|
306 |
float phi,thi,phd,php,nr,r;
|
307 |
int nh;
|
308 |
float zp=0,fp=0;
|
309 |
|
310 |
// Autosize for image, 25% greater than externa radius
|
311 |
bmx=1.25*re;
|
312 |
|
313 |
// Angular step of integration
|
314 |
h=4.e0f*PI/(float)TrackPoints;
|
315 |
|
316 |
// Step of Impact Parameter
|
317 |
db=bmx/(float)ImpactParameter;
|
318 |
|
319 |
// set origin as center of image
|
320 |
float x=(float)xi-(float)(sizex/2)+(float)5e-1f;
|
321 |
float y=(float)yi-(float)(sizey/2)+(float)5e-1f;
|
322 |
|
323 |
// angle extracted from cylindric symmetry
|
324 |
phi=atanp(x,y);
|
325 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
326 |
|
327 |
// Real Impact Parameter
|
328 |
b=sqrt(x*x+y*y)*bmx/(float)ImpactParameter;
|
329 |
|
330 |
// Integer Impact Parameter
|
331 |
uint bi=(uint)sqrt(x*x+y*y);
|
332 |
|
333 |
int HalfLap=0,ExitOnImpact=0,ni;
|
334 |
|
335 |
if (bi<ImpactParameter)
|
336 |
{
|
337 |
do
|
338 |
{
|
339 |
php=phd+(float)HalfLap*PI;
|
340 |
nr=php/h;
|
341 |
ni=(int)nr;
|
342 |
|
343 |
if (ni<IdLast[bi])
|
344 |
{
|
345 |
r=(Trajectories[bi*TrackPoints+ni+1]-Trajectories[bi*TrackPoints+ni])*(nr-ni*1.)+Trajectories[bi*TrackPoints+ni];
|
346 |
}
|
347 |
else
|
348 |
{
|
349 |
r=Trajectories[bi*TrackPoints+ni];
|
350 |
}
|
351 |
|
352 |
if ((r<=re)&&(r>=ri))
|
353 |
{
|
354 |
ExitOnImpact=1;
|
355 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
356 |
}
|
357 |
|
358 |
HalfLap++;
|
359 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
360 |
|
361 |
}
|
362 |
|
363 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
364 |
|
365 |
zImage[yi+sizex*xi]=zp;
|
366 |
fImage[yi+sizex*xi]=fp;
|
367 |
}
|
368 |
|
369 |
__kernel void Circle(__global float *Trajectories,__global int *IdLast,
|
370 |
__global float *zImage,__global float *fImage,
|
371 |
int TrackPoints,
|
372 |
float Mass,float InternalRadius,
|
373 |
float ExternalRadius,float Angle,
|
374 |
float ObserverDistance,
|
375 |
int BlackBody,int AngularCamera)
|
376 |
{
|
377 |
// Integer Impact Parameter ID
|
378 |
int bi=get_global_id(0);
|
379 |
// Integer points on circle
|
380 |
int i=get_global_id(1);
|
381 |
// Integer Impact Parameter Size (half of image)
|
382 |
int bmaxi=get_global_size(0);
|
383 |
// Integer Points on circle
|
384 |
int imx=get_global_size(1);
|
385 |
|
386 |
// Perform trajectory for each pixel
|
387 |
|
388 |
float m,rs,ri,re,tho,ro;
|
389 |
int q,dist,raie;
|
390 |
|
391 |
m=Mass;
|
392 |
rs=2.*m;
|
393 |
ri=InternalRadius;
|
394 |
re=ExternalRadius;
|
395 |
ro=ObserverDistance;
|
396 |
tho=Angle;
|
397 |
q=-2;
|
398 |
dist=AngularCamera;
|
399 |
raie=BlackBody;
|
400 |
|
401 |
float bmx,db,b,h;
|
402 |
float phi,thi,phd;
|
403 |
int nh;
|
404 |
float zp=0,fp=0;
|
405 |
|
406 |
// Autosize for image
|
407 |
bmx=1.25*re;
|
408 |
|
409 |
// Angular step of integration
|
410 |
h=4.e0f*PI/(float)TrackPoints;
|
411 |
|
412 |
// impact parameter
|
413 |
b=(float)bi/(float)bmaxi*bmx;
|
414 |
db=bmx/(float)bmaxi;
|
415 |
|
416 |
phi=2.*PI/(float)imx*(float)i;
|
417 |
phd=atanp(cos(phi)*sin(tho),cos(tho));
|
418 |
int yi=(int)((float)bi*sin(phi))+bmaxi;
|
419 |
int xi=(int)((float)bi*cos(phi))+bmaxi;
|
420 |
|
421 |
int HalfLap=0,ExitOnImpact=0,ni;
|
422 |
float php,nr,r;
|
423 |
|
424 |
do
|
425 |
{
|
426 |
php=phd+(float)HalfLap*PI;
|
427 |
nr=php/h;
|
428 |
ni=(int)nr;
|
429 |
|
430 |
if (ni<IdLast[bi])
|
431 |
{
|
432 |
r=(Trajectories[bi*TrackPoints+ni+1]-Trajectories[bi*TrackPoints+ni])*(nr-ni*1.)+Trajectories[bi*TrackPoints+ni];
|
433 |
}
|
434 |
else
|
435 |
{
|
436 |
r=Trajectories[bi*TrackPoints+ni];
|
437 |
}
|
438 |
|
439 |
if ((r<=re)&&(r>=ri))
|
440 |
{
|
441 |
ExitOnImpact=1;
|
442 |
impact(phi,r,b,tho,m,&zp,&fp,q,db,h,raie);
|
443 |
}
|
444 |
|
445 |
HalfLap++;
|
446 |
} while ((HalfLap<=2)&&(ExitOnImpact==0));
|
447 |
|
448 |
zImage[yi+2*bmaxi*xi]=zp;
|
449 |
fImage[yi+2*bmaxi*xi]=fp;
|
450 |
|
451 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
452 |
|
453 |
}
|
454 |
|
455 |
__kernel void Trajectory(__global float *Trajectories,
|
456 |
__global int *IdLast,int TrackPoints,
|
457 |
float Mass,float InternalRadius,
|
458 |
float ExternalRadius,float Angle,
|
459 |
float ObserverDistance,
|
460 |
int BlackBody,int AngularCamera)
|
461 |
{
|
462 |
// Integer Impact Parameter ID
|
463 |
int bi=get_global_id(0);
|
464 |
// Integer Impact Parameter Size (half of image)
|
465 |
int bmaxi=get_global_size(0);
|
466 |
|
467 |
// Perform trajectory for each pixel
|
468 |
|
469 |
float m,rs,ri,re,tho,ro;
|
470 |
int dist,raie,q;
|
471 |
|
472 |
m=Mass;
|
473 |
rs=2.*m;
|
474 |
ri=InternalRadius;
|
475 |
re=ExternalRadius;
|
476 |
ro=ObserverDistance;
|
477 |
tho=Angle;
|
478 |
q=-2;
|
479 |
dist=AngularCamera;
|
480 |
raie=BlackBody;
|
481 |
|
482 |
float d,bmx,db,b,h;
|
483 |
float phi,thi,phd,php,nr,r;
|
484 |
int nh;
|
485 |
float zp,fp;
|
486 |
|
487 |
// Autosize for image
|
488 |
bmx=1.25*re;
|
489 |
|
490 |
// Angular step of integration
|
491 |
h=4.e0f*PI/(float)TrackPoints;
|
492 |
|
493 |
// impact parameter
|
494 |
b=(float)bi/(float)bmaxi*bmx;
|
495 |
|
496 |
float up,vp,pp,us,vs,ps;
|
497 |
|
498 |
if (dist==1)
|
499 |
{
|
500 |
up=0.;
|
501 |
vp=1.;
|
502 |
}
|
503 |
else
|
504 |
{
|
505 |
thi=asin(b/ro);
|
506 |
up=sin(thi);
|
507 |
vp=cos(thi);
|
508 |
}
|
509 |
|
510 |
pp=0.;
|
511 |
nh=0;
|
512 |
|
513 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
514 |
|
515 |
// b versus us
|
516 |
float bvus=fabs(b/us);
|
517 |
float bvus0=bvus;
|
518 |
Trajectories[bi*TrackPoints+nh]=bvus;
|
519 |
|
520 |
do
|
521 |
{
|
522 |
nh++;
|
523 |
pp=ps;
|
524 |
up=us;
|
525 |
vp=vs;
|
526 |
rungekutta(&ps,&us,&vs,pp,up,vp,h,m,b);
|
527 |
bvus=fabs(b/us);
|
528 |
Trajectories[bi*TrackPoints+nh]=bvus;
|
529 |
|
530 |
} while ((bvus>=rs)&&(bvus<=bvus0));
|
531 |
|
532 |
IdLast[bi]=nh;
|
533 |
|
534 |
barrier(CLK_GLOBAL_MEM_FENCE);
|
535 |
|
536 |
}
|
537 |
""")
|
538 |
|
539 |
def ImageOutput(sigma,prefix): |
540 |
Max=sigma.max() |
541 |
Min=sigma.min() |
542 |
|
543 |
# Normalize value as 8bits Integer
|
544 |
SigmaInt=(255*(sigma-Min)/(Max-Min)).astype('uint8') |
545 |
image = Image.fromarray(SigmaInt) |
546 |
image.save("%s.jpg" % prefix)
|
547 |
|
548 |
def BlackHoleCL(zImage,fImage,InputCL): |
549 |
|
550 |
kernel_params = {} |
551 |
|
552 |
print(InputCL) |
553 |
|
554 |
Device=InputCL['Device']
|
555 |
GpuStyle=InputCL['GpuStyle']
|
556 |
VariableType=InputCL['VariableType']
|
557 |
Size=InputCL['Size']
|
558 |
Mass=InputCL['Mass']
|
559 |
InternalRadius=InputCL['InternalRadius']
|
560 |
ExternalRadius=InputCL['ExternalRadius']
|
561 |
Angle=InputCL['Angle']
|
562 |
ObserverDistance=InputCL['ObserverDistance']
|
563 |
Method=InputCL['Method']
|
564 |
TrackPoints=InputCL['TrackPoints']
|
565 |
|
566 |
if InputCL['BlackBody']: |
567 |
BlackBody=0
|
568 |
else:
|
569 |
BlackBody=1
|
570 |
|
571 |
if InputCL['AngularCamera']: |
572 |
AngularCamera=1
|
573 |
else:
|
574 |
AngularCamera=0
|
575 |
|
576 |
Trajectories=numpy.zeros((InputCL['Size']/2,InputCL['TrackPoints']), |
577 |
dtype=numpy.float32) |
578 |
IdLast=numpy.zeros(InputCL['Size']/2,dtype=numpy.int32) |
579 |
|
580 |
# Je detecte un peripherique GPU dans la liste des peripheriques
|
581 |
Id=0
|
582 |
HasXPU=False
|
583 |
for platform in cl.get_platforms(): |
584 |
for device in platform.get_devices(): |
585 |
if Id==Device:
|
586 |
XPU=device |
587 |
print "CPU/GPU selected: ",device.name.lstrip() |
588 |
HasXPU=True
|
589 |
Id+=1
|
590 |
|
591 |
if HasXPU==False: |
592 |
print "No XPU #%i found in all of %i devices, sorry..." % (Device,Id-1) |
593 |
sys.exit() |
594 |
|
595 |
ctx = cl.Context([XPU]) |
596 |
queue = cl.CommandQueue(ctx, |
597 |
properties=cl.command_queue_properties.PROFILING_ENABLE) |
598 |
|
599 |
BlackHoleCL = cl.Program(ctx,KERNEL_CODE.substitute(kernel_params)).build() |
600 |
|
601 |
# Je recupere les flag possibles pour les buffers
|
602 |
mf = cl.mem_flags |
603 |
|
604 |
print(zImage) |
605 |
|
606 |
TrajectoriesCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=Trajectories) |
607 |
zImageCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=zImage) |
608 |
fImageCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=fImage) |
609 |
fImageCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=fImage) |
610 |
IdLastCL = cl.Buffer(ctx, mf.WRITE_ONLY | mf.COPY_HOST_PTR, hostbuf=IdLast) |
611 |
|
612 |
start_time=time.time() |
613 |
|
614 |
print(Trajectories.shape) |
615 |
if Method=='EachPixel': |
616 |
CLLaunch=BlackHoleCL.EachPixel(queue,(zImage.shape[0],zImage.shape[1]), |
617 |
None,zImageCL,fImageCL,
|
618 |
numpy.float32(Mass), |
619 |
numpy.float32(InternalRadius), |
620 |
numpy.float32(ExternalRadius), |
621 |
numpy.float32(Angle), |
622 |
numpy.float32(ObserverDistance), |
623 |
numpy.int32(BlackBody), |
624 |
numpy.int32(AngularCamera)) |
625 |
CLLaunch.wait() |
626 |
elif Method=='TrajectoCircle': |
627 |
CLLaunch=BlackHoleCL.Trajectory(queue,(Trajectories.shape[0],),
|
628 |
None,TrajectoriesCL,IdLastCL,
|
629 |
numpy.uint32(Trajectories.shape[1]),
|
630 |
numpy.float32(Mass), |
631 |
numpy.float32(InternalRadius), |
632 |
numpy.float32(ExternalRadius), |
633 |
numpy.float32(Angle), |
634 |
numpy.float32(ObserverDistance), |
635 |
numpy.int32(BlackBody), |
636 |
numpy.int32(AngularCamera)) |
637 |
|
638 |
CLLaunch=BlackHoleCL.Circle(queue,(Trajectories.shape[0],
|
639 |
zImage.shape[0]*4),None, |
640 |
TrajectoriesCL,IdLastCL, |
641 |
zImageCL,fImageCL, |
642 |
numpy.uint32(Trajectories.shape[1]),
|
643 |
numpy.float32(Mass), |
644 |
numpy.float32(InternalRadius), |
645 |
numpy.float32(ExternalRadius), |
646 |
numpy.float32(Angle), |
647 |
numpy.float32(ObserverDistance), |
648 |
numpy.int32(BlackBody), |
649 |
numpy.int32(AngularCamera)) |
650 |
CLLaunch.wait() |
651 |
else:
|
652 |
CLLaunch=BlackHoleCL.Trajectory(queue,(Trajectories.shape[0],),
|
653 |
None,TrajectoriesCL,IdLastCL,
|
654 |
numpy.uint32(Trajectories.shape[1]),
|
655 |
numpy.float32(Mass), |
656 |
numpy.float32(InternalRadius), |
657 |
numpy.float32(ExternalRadius), |
658 |
numpy.float32(Angle), |
659 |
numpy.float32(ObserverDistance), |
660 |
numpy.int32(BlackBody), |
661 |
numpy.int32(AngularCamera)) |
662 |
|
663 |
CLLaunch=BlackHoleCL.Pixel(queue,(zImage.shape[0],
|
664 |
zImage.shape[1]),None, |
665 |
zImageCL,fImageCL,TrajectoriesCL,IdLastCL, |
666 |
numpy.uint32(Trajectories.shape[0]),
|
667 |
numpy.uint32(Trajectories.shape[1]),
|
668 |
numpy.float32(Mass), |
669 |
numpy.float32(InternalRadius), |
670 |
numpy.float32(ExternalRadius), |
671 |
numpy.float32(Angle), |
672 |
numpy.float32(ObserverDistance), |
673 |
numpy.int32(BlackBody), |
674 |
numpy.int32(AngularCamera)) |
675 |
CLLaunch.wait() |
676 |
|
677 |
elapsed = time.time()-start_time |
678 |
print("Elapsed %f: " % elapsed)
|
679 |
|
680 |
cl.enqueue_copy(queue,zImage,zImageCL).wait() |
681 |
cl.enqueue_copy(queue,fImage,fImageCL).wait() |
682 |
cl.enqueue_copy(queue,Trajectories,TrajectoriesCL).wait() |
683 |
cl.enqueue_copy(queue,IdLast,IdLastCL).wait() |
684 |
|
685 |
print(zImage.max()) |
686 |
print(fImage.max()) |
687 |
zImageCL.release() |
688 |
fImageCL.release() |
689 |
|
690 |
TrajectoriesCL.release() |
691 |
IdLastCL.release() |
692 |
|
693 |
return(elapsed)
|
694 |
|
695 |
if __name__=='__main__': |
696 |
|
697 |
GpuStyle = 'OpenCL'
|
698 |
Mass = 1.
|
699 |
# Internal Radius 3 times de Schwarzschild Radius
|
700 |
InternalRadius=6.*Mass
|
701 |
#
|
702 |
ExternalRadius=12.
|
703 |
#
|
704 |
ObserverDistance=100.
|
705 |
# Angle with normal to disc 10 degrees
|
706 |
Angle = numpy.pi/180.*(90.-10.) |
707 |
# Radiation of disc : BlackBody or Monochromatic
|
708 |
BlackBody = True
|
709 |
# Camera : Angular Camera or plate with the size of camera
|
710 |
AngularCamera = True
|
711 |
# Size of image
|
712 |
Size=256
|
713 |
# Variable Type
|
714 |
VariableType='FP32'
|
715 |
# ?
|
716 |
q=-2
|
717 |
# Method of resolution
|
718 |
Method='Trajecto'
|
719 |
|
720 |
HowToUse='%s -h [Help] -b [BlackBodyEmission] -c [AngularCamera] -s <SizeInPixels> -m <Mass> -i <DiscInternalRadius> -x <DiscExternalRadius> -o <ObservatorDistance> -a <AngleAboveDisc> -d <DeviceId> -g <CUDA/OpenCL> -t <EachPixel/TrajectoCircle/TrajectoPixel> -v <FP32/FP64>'
|
721 |
|
722 |
try:
|
723 |
opts, args = getopt.getopt(sys.argv[1:],"hbcs:m:i:x:o:a:d:g:v:t:",["blackbody","camera","size=","mass=","internal=","external=","observer=","angle=","device=","gpustyle=","variabletype=","method="]) |
724 |
except getopt.GetoptError:
|
725 |
print(HowToUse % sys.argv[0])
|
726 |
sys.exit(2)
|
727 |
|
728 |
# List of Devices
|
729 |
Devices=[] |
730 |
Alu={} |
731 |
|
732 |
for opt, arg in opts: |
733 |
if opt == '-h': |
734 |
print(HowToUse % sys.argv[0])
|
735 |
|
736 |
print("\nInformations about devices detected under OpenCL API:")
|
737 |
# For PyOpenCL import
|
738 |
try:
|
739 |
import pyopencl as cl |
740 |
Id=0
|
741 |
for platform in cl.get_platforms(): |
742 |
for device in platform.get_devices(): |
743 |
#deviceType=cl.device_type.to_string(device.type)
|
744 |
deviceType="xPU"
|
745 |
print("Device #%i from %s of type %s : %s" % (Id,platform.vendor.lstrip(),deviceType,device.name.lstrip()))
|
746 |
Id=Id+1
|
747 |
|
748 |
except:
|
749 |
print("Your platform does not seem to support OpenCL")
|
750 |
|
751 |
print("\nInformations about devices detected under CUDA API:")
|
752 |
# For PyCUDA import
|
753 |
try:
|
754 |
import pycuda.driver as cuda |
755 |
cuda.init() |
756 |
for Id in range(cuda.Device.count()): |
757 |
device=cuda.Device(Id) |
758 |
print("Device #%i of type GPU : %s" % (Id,device.name()))
|
759 |
print
|
760 |
except:
|
761 |
print("Your platform does not seem to support CUDA")
|
762 |
|
763 |
sys.exit() |
764 |
|
765 |
elif opt in ("-d", "--device"): |
766 |
# Devices.append(int(arg))
|
767 |
Device=int(arg)
|
768 |
elif opt in ("-g", "--gpustyle"): |
769 |
GpuStyle = arg |
770 |
elif opt in ("-v", "--variabletype"): |
771 |
VariableType = arg |
772 |
elif opt in ("-s", "--size"): |
773 |
Size = int(arg)
|
774 |
elif opt in ("-m", "--mass"): |
775 |
Mass = float(arg)
|
776 |
elif opt in ("-i", "--internal"): |
777 |
InternalRadius = float(arg)
|
778 |
elif opt in ("-e", "--external"): |
779 |
ExternalRadius = float(arg)
|
780 |
elif opt in ("-o", "--observer"): |
781 |
ObserverDistance = float(arg)
|
782 |
elif opt in ("-a", "--angle"): |
783 |
Angle = numpy.pi/180.*(90.-float(arg)) |
784 |
elif opt in ("-b", "--blackbody"): |
785 |
BlackBody = True
|
786 |
elif opt in ("-c", "--camera"): |
787 |
AngularCamera = True
|
788 |
elif opt in ("-t", "--method"): |
789 |
Method = arg |
790 |
|
791 |
print("Device Identification selected : %s" % Device)
|
792 |
print("GpuStyle used : %s" % GpuStyle)
|
793 |
print("VariableType : %s" % VariableType)
|
794 |
print("Size : %i" % Size)
|
795 |
print("Mass : %f" % Mass)
|
796 |
print("Internal Radius : %f" % InternalRadius)
|
797 |
print("External Radius : %f" % ExternalRadius)
|
798 |
print("Observer Distance : %f" % ObserverDistance)
|
799 |
print("Angle with normal of (in radians) : %f" % Angle)
|
800 |
print("Black Body Disc Emission (monochromatic instead) : %s" % BlackBody)
|
801 |
print("Angular Camera (dimension of object instead) : %s" % AngularCamera)
|
802 |
print("Method of resolution : %s" % Method)
|
803 |
|
804 |
if GpuStyle=='CUDA': |
805 |
print("\nSelection of CUDA device")
|
806 |
try:
|
807 |
# For PyCUDA import
|
808 |
import pycuda.driver as cuda |
809 |
|
810 |
cuda.init() |
811 |
for Id in range(cuda.Device.count()): |
812 |
device=cuda.Device(Id) |
813 |
print("Device #%i of type GPU : %s" % (Id,device.name()))
|
814 |
if Id in Devices: |
815 |
Alu[Id]='GPU'
|
816 |
|
817 |
except ImportError: |
818 |
print("Platform does not seem to support CUDA")
|
819 |
|
820 |
if GpuStyle=='OpenCL': |
821 |
print("\nSelection of OpenCL device")
|
822 |
try:
|
823 |
# For PyOpenCL import
|
824 |
import pyopencl as cl |
825 |
Id=0
|
826 |
for platform in cl.get_platforms(): |
827 |
for device in platform.get_devices(): |
828 |
#deviceType=cl.device_type.to_string(device.type)
|
829 |
deviceType="xPU"
|
830 |
print("Device #%i from %s of type %s : %s" % (Id,platform.vendor.lstrip().rstrip(),deviceType,device.name.lstrip().rstrip()))
|
831 |
|
832 |
if Id in Devices: |
833 |
# Set the Alu as detected Device Type
|
834 |
Alu[Id]=deviceType |
835 |
Id=Id+1
|
836 |
except ImportError: |
837 |
print("Platform does not seem to support OpenCL")
|
838 |
|
839 |
# print(Devices,Alu)
|
840 |
|
841 |
# MyImage=numpy.where(numpy.random.zeros(Size,Size)>0,1,-1).astype(numpy.float32)
|
842 |
TrackPoints=2048
|
843 |
zImage=numpy.zeros((Size,Size),dtype=numpy.float32) |
844 |
fImage=numpy.zeros((Size,Size),dtype=numpy.float32) |
845 |
|
846 |
InputCL={} |
847 |
InputCL['Device']=Device
|
848 |
InputCL['GpuStyle']=GpuStyle
|
849 |
InputCL['VariableType']=VariableType
|
850 |
InputCL['Size']=Size
|
851 |
InputCL['Mass']=Mass
|
852 |
InputCL['InternalRadius']=InternalRadius
|
853 |
InputCL['ExternalRadius']=ExternalRadius
|
854 |
InputCL['ObserverDistance']=ObserverDistance
|
855 |
InputCL['Angle']=Angle
|
856 |
InputCL['BlackBody']=BlackBody
|
857 |
InputCL['AngularCamera']=AngularCamera
|
858 |
InputCL['Method']=Method
|
859 |
InputCL['TrackPoints']=TrackPoints
|
860 |
|
861 |
duration=BlackHoleCL(zImage,fImage,InputCL) |
862 |
|
863 |
ImageOutput(zImage,"TrouNoirZ_%s" % Method)
|
864 |
ImageOutput(fImage,"TrouNoirF_%s" % Method)
|