52a5fe9d
dmayerich
Added double supp...
|
1
|
__device__ double g(double v0, double v1)
|
da3d4e0e
dmayerich
Initial commit.
|
2
3
4
5
|
{
return (v0 + v1)*log(abs(v0+v1)) + (v0-v1)*log(abs(v0-v1));
}
|
52a5fe9d
dmayerich
Added double supp...
|
6
|
__device__ double hfin(double v0, double v1, double dv)
|
da3d4e0e
dmayerich
Initial commit.
|
7
|
{
|
52a5fe9d
dmayerich
Added double supp...
|
8
9
10
11
|
double e = 0.001;
double t0 = g(v0+e, v1-dv)/dv;
double t1 = 2*g(v0+e, v1)/dv;
double t2 = g(v0+e, v1+dv)/dv;
|
da3d4e0e
dmayerich
Initial commit.
|
12
13
14
15
|
return -1.0/PI * (t0 - t1 + t2);
}
|
52a5fe9d
dmayerich
Added double supp...
|
16
|
__global__ void devKramersKronig(double* gpuN, double* gpuK, int numVals, double nuStart, double nuEnd, double nOffset)
|
da3d4e0e
dmayerich
Initial commit.
|
17
18
19
20
|
{
int i = blockIdx.x * blockDim.x + threadIdx.x;
if(i >= numVals) return;
|
52a5fe9d
dmayerich
Added double supp...
|
21
|
double nuDelta = (nuEnd - nuStart)/(numVals - 1);
|
da3d4e0e
dmayerich
Initial commit.
|
22
|
|
52a5fe9d
dmayerich
Added double supp...
|
23
24
25
26
|
double nu = nuStart + i*nuDelta;
double n = 0.0;
double jNu;
double jK;
|
da3d4e0e
dmayerich
Initial commit.
|
27
28
29
30
31
32
33
34
35
36
37
|
for(int j=1; j<numVals-1; j++)
{
jNu = nuStart + j*nuDelta;
jK = gpuK[j];
n += hfin(nu, jNu, nuDelta) * jK;
}
gpuN[i] = n + nOffset;
}
|
52a5fe9d
dmayerich
Added double supp...
|
38
|
void cudaKramersKronig(double* cpuN, double* cpuK, int nVals, double nuStart, double nuEnd, double nOffset)
|
da3d4e0e
dmayerich
Initial commit.
|
39
|
{
|
52a5fe9d
dmayerich
Added double supp...
|
40
41
42
43
44
|
double* gpuK;
HANDLE_ERROR(cudaMalloc(&gpuK, sizeof(double)*nVals));
HANDLE_ERROR(cudaMemcpy(gpuK, cpuK, sizeof(double)*nVals, cudaMemcpyHostToDevice));
double* gpuN;
HANDLE_ERROR(cudaMalloc(&gpuN, sizeof(double)*nVals));
|
da3d4e0e
dmayerich
Initial commit.
|
45
46
47
48
49
|
dim3 block(BLOCK_SIZE*BLOCK_SIZE);
dim3 grid(nVals/block.x + 1);
devKramersKronig<<<grid, block>>>(gpuN, gpuK, nVals, nuStart, nuEnd, nOffset);
|
52a5fe9d
dmayerich
Added double supp...
|
50
|
HANDLE_ERROR(cudaMemcpy(cpuN, gpuN, sizeof(double)*nVals, cudaMemcpyDeviceToHost));
|
da3d4e0e
dmayerich
Initial commit.
|
51
52
53
54
55
56
|
//free resources
HANDLE_ERROR(cudaFree(gpuK));
HANDLE_ERROR(cudaFree(gpuN));
}
|
52a5fe9d
dmayerich
Added double supp...
|
57
58
|
__global__ void devComputeSpectrum(double* I, double2* B, double* alpha, int Nl,
int nSamples, double oThetaI, double oThetaO, double cThetaI, double cThetaO)
|
da3d4e0e
dmayerich
Initial commit.
|
59
60
61
62
|
{
int i = blockIdx.x * blockDim.x + threadIdx.x;
//compute the delta-theta value
|
52a5fe9d
dmayerich
Added double supp...
|
63
|
double dTheta = (oThetaO - oThetaI)/nSamples;
|
da3d4e0e
dmayerich
Initial commit.
|
64
65
|
//allocate space for the Legendre polynomials
|
52a5fe9d
dmayerich
Added double supp...
|
66
67
68
69
70
71
72
73
74
75
|
double Ptheta[2];
double cosTheta, theta;
cuDoubleComplex Us;
cuDoubleComplex UsSample;
cuDoubleComplex U;
//cuComplex Ui;
//Ui.x = 2*PI;
//Ui.y = 0.0;
cuDoubleComplex numer;
|
da3d4e0e
dmayerich
Initial commit.
|
76
|
numer.x = 0.0;
|
52a5fe9d
dmayerich
Added double supp...
|
77
78
79
|
cuDoubleComplex exp_numer;
cuDoubleComplex iL;
cuDoubleComplex imag;
|
da3d4e0e
dmayerich
Initial commit.
|
80
|
imag.x = 0.0; imag.y = 1.0;
|
52a5fe9d
dmayerich
Added double supp...
|
81
82
83
84
85
86
|
double realFac;
cuDoubleComplex complexFac;
double PlTheta;
double Isum = 0.0;
//float maxVal = 0;
//float val;
|
da3d4e0e
dmayerich
Initial commit.
|
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
|
for(int iTheta = 0; iTheta < nSamples; iTheta++)
{
//calculate theta
theta = iTheta * dTheta + oThetaI;
cosTheta = cos(theta);
//initialize the theta Legendre polynomial
Ptheta[0] = 1.0;
Ptheta[1] = cosTheta;
//initialize the scattered field
Us.x = Us.y = 0.0;
iL.x = 1.0;
iL.y = 0.0;
for(int l = 0; l<Nl; l++)
{
//compute the theta legendre polynomial
if(l == 0)
PlTheta = Ptheta[0];
else if(l == 1)
PlTheta = Ptheta[1];
else
{
PlTheta = ((2*l - 1)*cosTheta*Ptheta[1] - (l - 1)*Ptheta[0])/l;
Ptheta[0] = Ptheta[1];
Ptheta[1] = PlTheta;
}
//compute the real components of the scattered field
realFac = alpha[l] * PlTheta;
//compute the complex components of the scattered field
numer.x = 0.0;
numer.y = -(l*PI)/2.0;
exp_numer = cExp(numer);
complexFac = cMult(B[Nl * i + l], exp_numer);
complexFac = cMult(complexFac, iL);
//combine the real and complex components
UsSample = cMult(complexFac, realFac);
Us = cAdd(Us, UsSample);
//increment the imaginary exponent i^l
iL = cMult(iL, imag);
|
52a5fe9d
dmayerich
Added double supp...
|
134
|
|
da3d4e0e
dmayerich
Initial commit.
|
135
136
137
138
139
140
141
142
143
144
145
146
147
|
}
//sum the scattered and incident fields
if(theta >= cThetaI && theta <= cThetaO)
U = cAdd(Us, 2*PI);
else
U = Us;
Isum += (U.x*U.x + U.y*U.y) * sin(theta) * 2 * PI * dTheta;
}
I[i] = Isum;
}
|
52a5fe9d
dmayerich
Added double supp...
|
148
149
|
void cudaComputeSpectrum(double* cpuI, double* cpuB, double* cpuAlpha,
int Nl, int nLambda, double oThetaI, double oThetaO, double cThetaI, double cThetaO, int nSamples)
|
da3d4e0e
dmayerich
Initial commit.
|
150
151
|
{
//copy everything to the GPU
|
52a5fe9d
dmayerich
Added double supp...
|
152
153
154
|
double2* gpuB;
HANDLE_ERROR(cudaMalloc(&gpuB, sizeof(double2) * nLambda * Nl));
HANDLE_ERROR(cudaMemcpy(gpuB, cpuB, sizeof(double2) * nLambda * Nl, cudaMemcpyHostToDevice));
|
da3d4e0e
dmayerich
Initial commit.
|
155
|
|
52a5fe9d
dmayerich
Added double supp...
|
156
157
158
|
double* gpuAlpha;
HANDLE_ERROR(cudaMalloc(&gpuAlpha, sizeof(double) * Nl));
HANDLE_ERROR(cudaMemcpy(gpuAlpha, cpuAlpha, sizeof(double) * Nl, cudaMemcpyHostToDevice));
|
da3d4e0e
dmayerich
Initial commit.
|
159
|
|
52a5fe9d
dmayerich
Added double supp...
|
160
161
|
double* gpuI;
HANDLE_ERROR(cudaMalloc(&gpuI, sizeof(double) * nLambda));
|
da3d4e0e
dmayerich
Initial commit.
|
162
163
164
165
166
167
168
|
//call the kernel to compute the spectrum
dim3 block(BLOCK_SIZE*BLOCK_SIZE);
dim3 grid(nLambda/block.x + 1);
//devComputeSpectrum
|
52a5fe9d
dmayerich
Added double supp...
|
169
|
devComputeSpectrum<<<grid, block>>>(gpuI, (double2*)gpuB, gpuAlpha, Nl,
|
da3d4e0e
dmayerich
Initial commit.
|
170
171
|
nSamples, oThetaI, oThetaO, cThetaI, cThetaO);
|
52a5fe9d
dmayerich
Added double supp...
|
172
|
HANDLE_ERROR(cudaMemcpy(cpuI, gpuI, sizeof(double) * nLambda, cudaMemcpyDeviceToHost));
|
da3d4e0e
dmayerich
Initial commit.
|
173
174
175
176
177
178
179
180
181
|
HANDLE_ERROR(cudaFree(gpuB));
HANDLE_ERROR(cudaFree(gpuAlpha));
HANDLE_ERROR(cudaFree(gpuI));
}
|