[5378e40] | 1 | #!/usr/bin/env python |
---|
| 2 | # -*- coding: utf-8 -*- |
---|
| 3 | |
---|
| 4 | import numpy as np |
---|
| 5 | import math |
---|
| 6 | import pyopencl as cl |
---|
| 7 | from weights import GaussianDispersion |
---|
[8a20be5] | 8 | from sasmodel import card |
---|
[5378e40] | 9 | |
---|
[8a20be5] | 10 | def set_precision(src, qx, qy, dtype): |
---|
| 11 | qx = np.ascontiguousarray(qx, dtype=dtype) |
---|
| 12 | qy = np.ascontiguousarray(qy, dtype=dtype) |
---|
| 13 | if np.dtype(dtype) == np.dtype('float32'): |
---|
| 14 | header = """\ |
---|
| 15 | #define real float |
---|
| 16 | """ |
---|
| 17 | else: |
---|
| 18 | header = """\ |
---|
| 19 | #pragma OPENCL EXTENSION cl_khr_fp64: enable |
---|
| 20 | #define real double |
---|
| 21 | """ |
---|
| 22 | return header+src, qx, qy |
---|
[5378e40] | 23 | |
---|
| 24 | class GpuCylinder(object): |
---|
| 25 | PARS = { |
---|
| 26 | 'scale':1,'radius':1,'length':1,'sldCyl':1e-6,'sldSolv':0,'background':0, |
---|
| 27 | 'cyl_theta':0,'cyl_phi':0, |
---|
| 28 | } |
---|
| 29 | PD_PARS = ['radius', 'length', 'cyl_theta', 'cyl_phi'] |
---|
| 30 | |
---|
[8a20be5] | 31 | def __init__(self, qx, qy, dtype='float32'): |
---|
[5378e40] | 32 | |
---|
| 33 | #create context, queue, and build program |
---|
[8a20be5] | 34 | ctx,_queue = card() |
---|
| 35 | src, qx, qy = set_precision(open('NR_BessJ1.cpp').read()+"\n"+open('Kernel-Cylinder.cpp').read(), qx, qy, dtype=dtype) |
---|
| 36 | self.prg = cl.Program(ctx, src).build() |
---|
| 37 | self.qx, self.qy = qx, qy |
---|
[5378e40] | 38 | |
---|
| 39 | #buffers |
---|
| 40 | mf = cl.mem_flags |
---|
[8a20be5] | 41 | self.qx_b = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=self.qx) |
---|
| 42 | self.qy_b = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=self.qy) |
---|
| 43 | self.res_b = cl.Buffer(ctx, mf.WRITE_ONLY, qx.nbytes) |
---|
[5378e40] | 44 | self.res = np.empty_like(self.qx) |
---|
| 45 | |
---|
| 46 | def eval(self, pars): |
---|
| 47 | |
---|
[8a20be5] | 48 | _ctx,queue = card() |
---|
| 49 | radius, length, cyl_theta, cyl_phi = \ |
---|
[5378e40] | 50 | [GaussianDispersion(int(pars[base+'_pd_n']), pars[base+'_pd'], pars[base+'_pd_nsigma']) |
---|
| 51 | for base in GpuCylinder.PD_PARS] |
---|
| 52 | |
---|
| 53 | #Get the weights for each |
---|
| 54 | radius.value, radius.weight = radius.get_weights(pars['radius'], 0, 1000, True) |
---|
| 55 | length.value, length.weight = length.get_weights(pars['length'], 0, 1000, True) |
---|
| 56 | cyl_theta.value, cyl_theta.weight = cyl_theta.get_weights(pars['cyl_theta'], -90, 180, False) |
---|
| 57 | cyl_phi.value, cyl_phi.weight = cyl_phi.get_weights(pars['cyl_phi'], -90, 180, False) |
---|
| 58 | |
---|
| 59 | #Perform the computation, with all weight points |
---|
| 60 | sum, norm, norm_vol, vol = 0.0, 0.0, 0.0, 0.0 |
---|
| 61 | size = len(cyl_theta.weight) |
---|
| 62 | sub = pars['sldCyl'] - pars['sldSolv'] |
---|
| 63 | |
---|
[8a20be5] | 64 | real = np.float32 if self.qx.dtype == np.dtype('float32') else np.float64 |
---|
[5378e40] | 65 | #Loop over radius, length, theta, phi weight points |
---|
| 66 | for i in xrange(len(radius.weight)): |
---|
| 67 | for j in xrange(len(length.weight)): |
---|
| 68 | for k in xrange(len(cyl_theta.weight)): |
---|
| 69 | for l in xrange(len(cyl_phi.weight)): |
---|
| 70 | |
---|
[8a20be5] | 71 | |
---|
| 72 | self.prg.CylinderKernel(queue, self.qx.shape, None, self.qx_b, self.qy_b, self.res_b, real(sub), |
---|
| 73 | real(radius.value[i]), real(length.value[j]), real(pars['scale']), |
---|
| 74 | real(radius.weight[i]), real(length.weight[j]), real(cyl_theta.weight[k]), |
---|
| 75 | real(cyl_phi.weight[l]), real(cyl_theta.value[k]), real(cyl_phi.value[l]), |
---|
[5378e40] | 76 | np.uint32(self.qx.size), np.uint32(size)) |
---|
[8a20be5] | 77 | cl.enqueue_copy(queue, self.res, self.res_b) |
---|
[5378e40] | 78 | sum += self.res |
---|
| 79 | vol += radius.weight[i]*length.weight[j]*pow(radius.value[i], 2)*length.value[j] |
---|
| 80 | norm_vol += radius.weight[i]*length.weight[j] |
---|
| 81 | norm += radius.weight[i]*length.weight[j]*cyl_theta.weight[k]*cyl_phi.weight[l] |
---|
| 82 | |
---|
| 83 | if size > 1: |
---|
| 84 | norm /= math.asin(1.0) |
---|
| 85 | if vol != 0.0 and norm_vol != 0.0: |
---|
| 86 | sum *= norm_vol/vol |
---|
| 87 | |
---|
| 88 | return sum/norm+pars['background'] |
---|
| 89 | |
---|
| 90 | def demo(): |
---|
| 91 | from time import time |
---|
| 92 | import matplotlib.pyplot as plt |
---|
| 93 | |
---|
| 94 | #create qx and qy evenly spaces |
---|
| 95 | qx = np.linspace(-.02, .02, 128) |
---|
| 96 | qy = np.linspace(-.02, .02, 128) |
---|
| 97 | qx, qy = np.meshgrid(qx, qy) |
---|
| 98 | |
---|
| 99 | #saved shape of qx |
---|
| 100 | r_shape = qx.shape |
---|
| 101 | |
---|
| 102 | #reshape for calculation; resize as float32 |
---|
| 103 | qx = qx.flatten() |
---|
| 104 | qy = qy.flatten() |
---|
| 105 | |
---|
| 106 | pars = CylinderParameters(scale=1, radius=64.1, length=266.96, sldCyl=.291e-6, sldSolv=5.77e-6, background=0, |
---|
| 107 | cyl_theta=0, cyl_phi=0) |
---|
| 108 | t = time() |
---|
| 109 | result = GpuCylinder(qx, qy) |
---|
| 110 | result.x = result.cylinder_fit(pars, r_n=10, t_n=10, l_n=10, p_n=10, r_w=.1, t_w=.1, l_w=.1, p_w=.1, sigma=3) |
---|
| 111 | result.x = np.reshape(result.x, r_shape) |
---|
| 112 | tt = time() |
---|
| 113 | |
---|
| 114 | print("Time taken: %f" % (tt - t)) |
---|
| 115 | |
---|
| 116 | plt.pcolormesh(result.x) |
---|
| 117 | plt.show() |
---|
| 118 | |
---|
| 119 | |
---|
| 120 | if __name__=="__main__": |
---|
| 121 | demo() |
---|
| 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 | |
---|