1 | |
---|
2 | /* |
---|
3 | ########################################################## |
---|
4 | # # |
---|
5 | # !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! # |
---|
6 | # !! !! # |
---|
7 | # !! KEEP THIS CODE CONSISTENT WITH KERNELPY.PY !! # |
---|
8 | # !! !! # |
---|
9 | # !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! # |
---|
10 | # # |
---|
11 | ########################################################## |
---|
12 | */ |
---|
13 | |
---|
14 | #ifndef _PAR_BLOCK_ // protected block so we can include this code twice. |
---|
15 | #define _PAR_BLOCK_ |
---|
16 | |
---|
17 | typedef struct { |
---|
18 | #if MAX_PD > 0 |
---|
19 | int32_t pd_par[MAX_PD]; // id of the nth polydispersity variable |
---|
20 | int32_t pd_length[MAX_PD]; // length of the nth polydispersity weight vector |
---|
21 | int32_t pd_offset[MAX_PD]; // offset of pd weights in the value & weight vector |
---|
22 | int32_t pd_stride[MAX_PD]; // stride to move to the next index at this level |
---|
23 | #endif // MAX_PD > 0 |
---|
24 | int32_t pd_prod; // total number of voxels in hypercube |
---|
25 | int32_t pd_sum; // total length of the weights vector |
---|
26 | int32_t num_active; // number of non-trivial pd loops |
---|
27 | int32_t theta_par; // id of spherical correction variable |
---|
28 | } ProblemDetails; |
---|
29 | |
---|
30 | typedef struct { |
---|
31 | PARAMETER_TABLE; |
---|
32 | } ParameterBlock; |
---|
33 | #endif |
---|
34 | |
---|
35 | |
---|
36 | kernel |
---|
37 | void KERNEL_NAME( |
---|
38 | int32_t nq, // number of q values |
---|
39 | const int32_t pd_start, // where we are in the polydispersity loop |
---|
40 | const int32_t pd_stop, // where we are stopping in the polydispersity loop |
---|
41 | global const ProblemDetails *details, |
---|
42 | global const double *values, |
---|
43 | global const double *q, // nq q values, with padding to boundary |
---|
44 | global double *result, // nq+3 return values, again with padding |
---|
45 | const double cutoff // cutoff in the polydispersity weight product |
---|
46 | ) |
---|
47 | { |
---|
48 | // Storage for the current parameter values. These will be updated as we |
---|
49 | // walk the polydispersity cube. |
---|
50 | ParameterBlock local_values; // current parameter values |
---|
51 | double *pvec = (double *)(&local_values); // Alias named parameters with a vector |
---|
52 | |
---|
53 | // Fill in the initial variables |
---|
54 | #ifdef USE_OPENMP |
---|
55 | #pragma omp parallel for |
---|
56 | #endif |
---|
57 | for (int k=0; k < NPARS; k++) { |
---|
58 | pvec[k] = values[k+2]; |
---|
59 | } |
---|
60 | |
---|
61 | // Monodisperse computation |
---|
62 | if (details->num_active == 0) { |
---|
63 | double norm, scale, background; |
---|
64 | #ifdef INVALID |
---|
65 | if (INVALID(local_values)) { return; } |
---|
66 | #endif |
---|
67 | |
---|
68 | norm = CALL_VOLUME(local_values); |
---|
69 | scale = values[0]; |
---|
70 | background = values[1]; |
---|
71 | |
---|
72 | #ifdef USE_OPENMP |
---|
73 | #pragma omp parallel for |
---|
74 | #endif |
---|
75 | for (int q_index=0; q_index < nq; q_index++) { |
---|
76 | double scattering = CALL_IQ(q, q_index, local_values); |
---|
77 | result[q_index] = (norm>0. ? scale*scattering/norm + background : background); |
---|
78 | } |
---|
79 | return; |
---|
80 | } |
---|
81 | |
---|
82 | #if MAX_PD > 0 |
---|
83 | |
---|
84 | const double *pd_value = values+2+NPARS; |
---|
85 | const double *pd_weight = pd_value+details->pd_sum; |
---|
86 | |
---|
87 | // need product of weights at every Iq calc, so keep product of |
---|
88 | // weights from the outer loops so that weight = partial_weight * fast_weight |
---|
89 | double pd_norm; |
---|
90 | double partial_weight; // product of weight w4*w3*w2 but not w1 |
---|
91 | double spherical_correction; // cosine correction for latitude variation |
---|
92 | double weight; // product of partial_weight*w1*spherical_correction |
---|
93 | |
---|
94 | // Number of elements in the longest polydispersity loop |
---|
95 | const int p0_par = details->pd_par[0]; |
---|
96 | const int p0_length = details->pd_length[0]; |
---|
97 | const int p0_offset = details->pd_offset[0]; |
---|
98 | const int p0_is_theta = (p0_par == details->theta_par); |
---|
99 | int p0_index; |
---|
100 | |
---|
101 | // Trigger the reset behaviour that happens at the end the fast loop |
---|
102 | // by setting the initial index >= weight vector length. |
---|
103 | p0_index = p0_length; |
---|
104 | |
---|
105 | // Default the spherical correction to 1.0 in case it is not otherwise set |
---|
106 | spherical_correction = 1.0; |
---|
107 | |
---|
108 | // Since we are no longer looping over the entire polydispersity hypercube |
---|
109 | // for each q, we need to track the result and normalization values between |
---|
110 | // calls. This means initializing them to 0 at the start and accumulating |
---|
111 | // them between calls. |
---|
112 | pd_norm = (pd_start == 0 ? 0.0 : result[nq]); |
---|
113 | |
---|
114 | if (pd_start == 0) { |
---|
115 | #ifdef USE_OPENMP |
---|
116 | #pragma omp parallel for |
---|
117 | #endif |
---|
118 | for (int q_index=0; q_index < nq; q_index++) { |
---|
119 | result[q_index] = 0.0; |
---|
120 | } |
---|
121 | } |
---|
122 | |
---|
123 | // Loop over the weights then loop over q, accumulating values |
---|
124 | for (int loop_index=pd_start; loop_index < pd_stop; loop_index++) { |
---|
125 | // check if fast loop needs to be reset |
---|
126 | if (p0_index == p0_length) { |
---|
127 | |
---|
128 | // Compute position in polydispersity hypercube and partial weight |
---|
129 | partial_weight = 1.0; |
---|
130 | for (int k=1; k < details->num_active; k++) { |
---|
131 | int pk = details->pd_par[k]; |
---|
132 | int index = details->pd_offset[k] + (loop_index/details->pd_stride[k])%details->pd_length[k]; |
---|
133 | pvec[pk] = pd_value[index]; |
---|
134 | partial_weight *= pd_weight[index]; |
---|
135 | if (pk == details->theta_par) { |
---|
136 | spherical_correction = fmax(fabs(cos(M_PI_180*pvec[pk])), 1.e-6); |
---|
137 | } |
---|
138 | } |
---|
139 | p0_index = loop_index%p0_length; |
---|
140 | } |
---|
141 | |
---|
142 | // Update parameter p0 |
---|
143 | weight = partial_weight*pd_weight[p0_offset + p0_index]; |
---|
144 | pvec[p0_par] = pd_value[p0_offset + p0_index]; |
---|
145 | if (p0_is_theta) { |
---|
146 | spherical_correction = fmax(fabs(cos(M_PI_180*pvec[p0_par])), 1.e-6); |
---|
147 | } |
---|
148 | p0_index++; |
---|
149 | |
---|
150 | #ifdef INVALID |
---|
151 | if (INVALID(local_values)) continue; |
---|
152 | #endif |
---|
153 | |
---|
154 | // Accumulate I(q) |
---|
155 | // Note: weight==0 must always be excluded |
---|
156 | if (weight > cutoff) { |
---|
157 | // spherical correction has some nasty effects when theta is +90 or -90 |
---|
158 | // where it becomes zero. If the entirety of the correction |
---|
159 | weight *= spherical_correction; |
---|
160 | pd_norm += weight * CALL_VOLUME(local_values); |
---|
161 | |
---|
162 | #ifdef USE_OPENMP |
---|
163 | #pragma omp parallel for |
---|
164 | #endif |
---|
165 | for (int q_index=0; q_index < nq; q_index++) { |
---|
166 | const double scattering = CALL_IQ(q, q_index, local_values); |
---|
167 | result[q_index] += weight*scattering; |
---|
168 | } |
---|
169 | } |
---|
170 | } |
---|
171 | |
---|
172 | if (pd_stop >= details->pd_prod) { |
---|
173 | // End of the PD loop we can normalize |
---|
174 | double scale, background; |
---|
175 | scale = values[0]; |
---|
176 | background = values[1]; |
---|
177 | #ifdef USE_OPENMP |
---|
178 | #pragma omp parallel for |
---|
179 | #endif |
---|
180 | for (int q_index=0; q_index < nq; q_index++) { |
---|
181 | result[q_index] = (pd_norm>0. ? scale*result[q_index]/pd_norm + background : background); |
---|
182 | } |
---|
183 | } |
---|
184 | |
---|
185 | // Remember the updated norm. |
---|
186 | result[nq] = pd_norm; |
---|
187 | #endif // MAX_PD > 0 |
---|
188 | } |
---|