cudaKK.h
4.56 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
__device__ float g(float v0, float v1)
{
return (v0 + v1)*log(abs(v0+v1)) + (v0-v1)*log(abs(v0-v1));
}
__device__ float hfin(float v0, float v1, float dv)
{
float e = 0.001;
float t0 = g(v0+e, v1-dv)/dv;
float t1 = 2*g(v0+e, v1)/dv;
float t2 = g(v0+e, v1+dv)/dv;
return -1.0/PI * (t0 - t1 + t2);
}
__global__ void devKramersKronig(float* gpuN, float* gpuK, int numVals, float nuStart, float nuEnd, float nOffset)
{
int i = blockIdx.x * blockDim.x + threadIdx.x;
if(i >= numVals) return;
float nuDelta = (nuEnd - nuStart)/(numVals - 1);
float nu = nuStart + i*nuDelta;
float n = 0.0;
float jNu;
float jK;
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;
}
void cudaKramersKronig(float* cpuN, float* cpuK, int nVals, float nuStart, float nuEnd, float nOffset)
{
float* gpuK;
HANDLE_ERROR(cudaMalloc(&gpuK, sizeof(float)*nVals));
HANDLE_ERROR(cudaMemcpy(gpuK, cpuK, sizeof(float)*nVals, cudaMemcpyHostToDevice));
float* gpuN;
HANDLE_ERROR(cudaMalloc(&gpuN, sizeof(float)*nVals));
dim3 block(BLOCK_SIZE*BLOCK_SIZE);
dim3 grid(nVals/block.x + 1);
devKramersKronig<<<grid, block>>>(gpuN, gpuK, nVals, nuStart, nuEnd, nOffset);
HANDLE_ERROR(cudaMemcpy(cpuN, gpuN, sizeof(float)*nVals, cudaMemcpyDeviceToHost));
//free resources
HANDLE_ERROR(cudaFree(gpuK));
HANDLE_ERROR(cudaFree(gpuN));
}
__global__ void devComputeSpectrum(float* I, float2* B, float* alpha, int Nl,
int nSamples, float oThetaI, float oThetaO, float cThetaI, float cThetaO)
{
int i = blockIdx.x * blockDim.x + threadIdx.x;
//compute the delta-theta value
float dTheta = (oThetaO - oThetaI)/nSamples;
//allocate space for the Legendre polynomials
float Ptheta[2];
float cosTheta, theta;
cuComplex Us;
cuComplex UsSample;
cuComplex U;
cuComplex Ui;
Ui.x = 2*PI;
Ui.y = 0.0;
cuComplex numer;
numer.x = 0.0;
cuComplex exp_numer;
cuComplex iL;
cuComplex imag;
imag.x = 0.0; imag.y = 1.0;
float realFac;
cuComplex complexFac;
float PlTheta;
float Isum = 0.0;
float maxVal = 0;
float val;
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);
//val = cMag(Us);
//if(val > maxVal)
// maxVal = val;
//Us += B[l] * exp(numer) * Ptheta[l] * alpha * M * pow(complex<float>(0.0, 1.0), l);
}
//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;
}
void cudaComputeSpectrum(float* cpuI, float* cpuB, float* cpuAlpha,
int Nl, int nLambda, float oThetaI, float oThetaO, float cThetaI, float cThetaO, int nSamples)
{
//copy everything to the GPU
float2* gpuB;
HANDLE_ERROR(cudaMalloc(&gpuB, sizeof(float2) * nLambda * Nl));
HANDLE_ERROR(cudaMemcpy(gpuB, cpuB, sizeof(float2) * nLambda * Nl, cudaMemcpyHostToDevice));
float* gpuAlpha;
HANDLE_ERROR(cudaMalloc(&gpuAlpha, sizeof(float) * Nl));
HANDLE_ERROR(cudaMemcpy(gpuAlpha, cpuAlpha, sizeof(float) * Nl, cudaMemcpyHostToDevice));
float* gpuI;
HANDLE_ERROR(cudaMalloc(&gpuI, sizeof(float) * nLambda));
//call the kernel to compute the spectrum
dim3 block(BLOCK_SIZE*BLOCK_SIZE);
dim3 grid(nLambda/block.x + 1);
//devComputeSpectrum
devComputeSpectrum<<<grid, block>>>(gpuI, (float2*)gpuB, gpuAlpha, Nl,
nSamples, oThetaI, oThetaO, cThetaI, cThetaO);
HANDLE_ERROR(cudaMemcpy(cpuI, gpuI, sizeof(float) * nLambda, cudaMemcpyDeviceToHost));
HANDLE_ERROR(cudaFree(gpuB));
HANDLE_ERROR(cudaFree(gpuAlpha));
HANDLE_ERROR(cudaFree(gpuI));
}