1 #include "simint/boys/boys.h"
2 #include "simint/ostei/gen/ostei_generated.h"
3 #include "simint/vectorization/vectorization.h"
4 #include <math.h>
5 #include <string.h>
6
7
ostei_d_d_i_p(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__d_d_i_p)8 int ostei_d_d_i_p(struct simint_multi_shellpair const P,
9 struct simint_multi_shellpair const Q,
10 double screen_tol,
11 double * const restrict work,
12 double * const restrict INT__d_d_i_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__d_d_i_p);
17 int ab, cd, abcd;
18 int istart, jstart;
19 int iprimcd, nprim_icd, icd;
20 const int check_screen = (screen_tol > 0.0);
21 int i, j;
22 int n;
23 int not_screened;
24 int real_abcd;
25 int iket;
26 int ibra;
27
28 // partition workspace
29 double * const INT__d_s_i_s = work + (SIMINT_NSHELL_SIMD * 0);
30 double * const INT__d_s_k_s = work + (SIMINT_NSHELL_SIMD * 168);
31 double * const INT__f_s_i_s = work + (SIMINT_NSHELL_SIMD * 384);
32 double * const INT__f_s_k_s = work + (SIMINT_NSHELL_SIMD * 664);
33 double * const INT__g_s_i_s = work + (SIMINT_NSHELL_SIMD * 1024);
34 double * const INT__g_s_k_s = work + (SIMINT_NSHELL_SIMD * 1444);
35 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*1984);
36 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_p_s = primwork + 12;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_d_s = primwork + 45;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_f_s = primwork + 105;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_g_s = primwork + 195;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_h_s = primwork + 315;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_i_s = primwork + 462;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_k_s = primwork + 630;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_f_s = primwork + 810;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_g_s = primwork + 930;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_h_s = primwork + 1110;
47 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_i_s = primwork + 1362;
48 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_k_s = primwork + 1698;
49 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_g_s = primwork + 2130;
50 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_h_s = primwork + 2400;
51 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_i_s = primwork + 2778;
52 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_k_s = primwork + 3282;
53 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_h_s = primwork + 3930;
54 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_i_s = primwork + 4350;
55 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_k_s = primwork + 4910;
56 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_i_s = primwork + 5630;
57 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_k_s = primwork + 6050;
58 double * const hrrwork = (double *)(primwork + 6590);
59 double * const HRR_INT__d_p_i_s = hrrwork + 0;
60 double * const HRR_INT__d_p_k_s = hrrwork + 504;
61 double * const HRR_INT__d_d_i_s = hrrwork + 1152;
62 double * const HRR_INT__d_d_k_s = hrrwork + 2160;
63 double * const HRR_INT__f_p_i_s = hrrwork + 3456;
64 double * const HRR_INT__f_p_k_s = hrrwork + 4296;
65
66
67 // Create constants
68 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
69 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
70 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
71 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
72 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
73 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
74 const SIMINT_DBLTYPE const_7 = SIMINT_DBLSET1(7);
75 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
76
77
78 ////////////////////////////////////////
79 // Loop over shells and primitives
80 ////////////////////////////////////////
81
82 real_abcd = 0;
83 istart = 0;
84 for(ab = 0; ab < P.nshell12_clip; ++ab)
85 {
86 const int iend = istart + P.nprim12[ab];
87
88 cd = 0;
89 jstart = 0;
90
91 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
92 {
93 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
94 int jend = jstart;
95 for(i = 0; i < nshellbatch; i++)
96 jend += Q.nprim12[cd+i];
97
98 // Clear the beginning of the workspace (where we are accumulating integrals)
99 memset(work, 0, SIMINT_NSHELL_SIMD * 1984 * sizeof(double));
100 abcd = 0;
101
102
103 for(i = istart; i < iend; ++i)
104 {
105 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
106
107 if(check_screen)
108 {
109 // Skip this whole thing if always insignificant
110 if((P.screen[i] * Q.screen_max) < screen_tol)
111 continue;
112 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
113 }
114
115 icd = 0;
116 iprimcd = 0;
117 nprim_icd = Q.nprim12[cd];
118 double * restrict PRIM_PTR_INT__d_s_i_s = INT__d_s_i_s + abcd * 168;
119 double * restrict PRIM_PTR_INT__d_s_k_s = INT__d_s_k_s + abcd * 216;
120 double * restrict PRIM_PTR_INT__f_s_i_s = INT__f_s_i_s + abcd * 280;
121 double * restrict PRIM_PTR_INT__f_s_k_s = INT__f_s_k_s + abcd * 360;
122 double * restrict PRIM_PTR_INT__g_s_i_s = INT__g_s_i_s + abcd * 420;
123 double * restrict PRIM_PTR_INT__g_s_k_s = INT__g_s_k_s + abcd * 540;
124
125
126
127 // Load these one per loop over i
128 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
129 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
130 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
131
132 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
133
134 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
135 {
136 // calculate the shell offsets
137 // these are the offset from the shell pointed to by cd
138 // for each element
139 int shelloffsets[SIMINT_SIMD_LEN] = {0};
140 int lastoffset = 0;
141 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
142
143 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
144 {
145 // Handle if the first element of the vector is a new shell
146 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
147 {
148 nprim_icd += Q.nprim12[cd + (++icd)];
149 PRIM_PTR_INT__d_s_i_s += 168;
150 PRIM_PTR_INT__d_s_k_s += 216;
151 PRIM_PTR_INT__f_s_i_s += 280;
152 PRIM_PTR_INT__f_s_k_s += 360;
153 PRIM_PTR_INT__g_s_i_s += 420;
154 PRIM_PTR_INT__g_s_k_s += 540;
155 }
156 iprimcd++;
157 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
158 {
159 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
160 {
161 shelloffsets[n] = shelloffsets[n-1] + 1;
162 lastoffset++;
163 nprim_icd += Q.nprim12[cd + (++icd)];
164 }
165 else
166 shelloffsets[n] = shelloffsets[n-1];
167 iprimcd++;
168 }
169 }
170 else
171 iprimcd += SIMINT_SIMD_LEN;
172
173 // Do we have to compute this vector (or has it been screened out)?
174 // (not_screened != 0 means we have to do this vector)
175 if(check_screen)
176 {
177 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
178 if(vmax < screen_tol)
179 {
180 PRIM_PTR_INT__d_s_i_s += lastoffset*168;
181 PRIM_PTR_INT__d_s_k_s += lastoffset*216;
182 PRIM_PTR_INT__f_s_i_s += lastoffset*280;
183 PRIM_PTR_INT__f_s_k_s += lastoffset*360;
184 PRIM_PTR_INT__g_s_i_s += lastoffset*420;
185 PRIM_PTR_INT__g_s_k_s += lastoffset*540;
186 continue;
187 }
188 }
189
190 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
191 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
192 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
193 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
194
195
196 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
197 SIMINT_DBLTYPE PQ[3];
198 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
199 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
200 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
201 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
202 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
203 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
204
205 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
206 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
207 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
208 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
209 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
210 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
211 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
212
213 // NOTE: Minus sign!
214 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
215 SIMINT_DBLTYPE aop_PQ[3];
216 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
217 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
218 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
219
220 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
221 SIMINT_DBLTYPE aoq_PQ[3];
222 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
223 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
224 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
225 // Put a minus sign here so we don't have to in RR routines
226 a_over_q = SIMINT_NEG(a_over_q);
227
228
229 //////////////////////////////////////////////
230 // Fjt function section
231 // Maximum v value: 11
232 //////////////////////////////////////////////
233 // The parameter to the Fjt function
234 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
235
236
237 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
238
239
240 boys_F_split(PRIM_INT__s_s_s_s, F_x, 11);
241 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
242 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
243 for(n = 0; n <= 11; n++)
244 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
245
246 //////////////////////////////////////////////
247 // Primitive integrals: Vertical recurrance
248 //////////////////////////////////////////////
249
250 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
251 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
252 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
253 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
254 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
255 const SIMINT_DBLTYPE vrr_const_3_over_2q = SIMINT_MUL(const_3, one_over_2q);
256 const SIMINT_DBLTYPE vrr_const_4_over_2q = SIMINT_MUL(const_4, one_over_2q);
257 const SIMINT_DBLTYPE vrr_const_5_over_2q = SIMINT_MUL(const_5, one_over_2q);
258 const SIMINT_DBLTYPE vrr_const_6_over_2q = SIMINT_MUL(const_6, one_over_2q);
259 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
260 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
261 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
262 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
263 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
264 const SIMINT_DBLTYPE vrr_const_6_over_2pq = SIMINT_MUL(const_6, one_over_2pq);
265 const SIMINT_DBLTYPE vrr_const_7_over_2pq = SIMINT_MUL(const_7, one_over_2pq);
266
267
268
269 // Forming PRIM_INT__s_s_p_s[11 * 3];
270 for(n = 0; n < 11; ++n) // loop over orders of auxiliary function
271 {
272
273 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
274 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]);
275
276 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
277 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 1]);
278
279 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
280 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 2]);
281
282 }
283
284
285
286 // Forming PRIM_INT__s_s_d_s[10 * 6];
287 for(n = 0; n < 10; ++n) // loop over orders of auxiliary function
288 {
289
290 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_p_s[n * 3 + 0]);
291 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 0]);
292 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 0]);
293
294 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 0]);
295 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 1]);
296
297 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 0]);
298 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 2]);
299
300 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 1]);
301 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 3]);
302 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 3]);
303
304 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 1]);
305 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 4]);
306
307 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 2]);
308 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_d_s[n * 6 + 5]);
309 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 5]);
310
311 }
312
313
314
315 // Forming PRIM_INT__s_s_f_s[9 * 10];
316 for(n = 0; n < 9; ++n) // loop over orders of auxiliary function
317 {
318
319 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
320 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 0]);
321 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]), PRIM_INT__s_s_f_s[n * 10 + 0]);
322
323 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
324 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 1]);
325
326 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
327 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 2]);
328
329 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
330 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 3]);
331
332 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
333 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_s_f_s[n * 10 + 4]);
334
335 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
336 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 5]);
337
338 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
339 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 6]);
340 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_p_s[n * 3 + 1]), PRIM_INT__s_s_f_s[n * 10 + 6]);
341
342 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
343 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 7]);
344
345 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
346 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 8]);
347
348 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
349 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 9]);
350 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_p_s[n * 3 + 2]), PRIM_INT__s_s_f_s[n * 10 + 9]);
351
352 }
353
354
355 VRR_K_s_s_g_s(
356 PRIM_INT__s_s_g_s,
357 PRIM_INT__s_s_f_s,
358 PRIM_INT__s_s_d_s,
359 Q_PA,
360 a_over_q,
361 aoq_PQ,
362 one_over_2q,
363 8);
364
365
366 VRR_K_s_s_h_s(
367 PRIM_INT__s_s_h_s,
368 PRIM_INT__s_s_g_s,
369 PRIM_INT__s_s_f_s,
370 Q_PA,
371 a_over_q,
372 aoq_PQ,
373 one_over_2q,
374 7);
375
376
377 ostei_general_vrr1_K(6, 6,
378 one_over_2q, a_over_q, aoq_PQ, Q_PA,
379 PRIM_INT__s_s_h_s, PRIM_INT__s_s_g_s, PRIM_INT__s_s_i_s);
380
381
382 ostei_general_vrr_I(1, 0, 6, 0, 4,
383 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
384 PRIM_INT__s_s_i_s, NULL, NULL, PRIM_INT__s_s_h_s, NULL, PRIM_INT__p_s_i_s);
385
386
387 ostei_general_vrr_I(1, 0, 5, 0, 4,
388 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
389 PRIM_INT__s_s_h_s, NULL, NULL, PRIM_INT__s_s_g_s, NULL, PRIM_INT__p_s_h_s);
390
391
392 ostei_general_vrr_I(2, 0, 6, 0, 3,
393 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
394 PRIM_INT__p_s_i_s, PRIM_INT__s_s_i_s, NULL, PRIM_INT__p_s_h_s, NULL, PRIM_INT__d_s_i_s);
395
396
397 ostei_general_vrr1_K(7, 5,
398 one_over_2q, a_over_q, aoq_PQ, Q_PA,
399 PRIM_INT__s_s_i_s, PRIM_INT__s_s_h_s, PRIM_INT__s_s_k_s);
400
401
402 ostei_general_vrr_I(1, 0, 7, 0, 4,
403 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
404 PRIM_INT__s_s_k_s, NULL, NULL, PRIM_INT__s_s_i_s, NULL, PRIM_INT__p_s_k_s);
405
406
407 ostei_general_vrr_I(2, 0, 7, 0, 3,
408 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
409 PRIM_INT__p_s_k_s, PRIM_INT__s_s_k_s, NULL, PRIM_INT__p_s_i_s, NULL, PRIM_INT__d_s_k_s);
410
411
412 VRR_I_p_s_g_s(
413 PRIM_INT__p_s_g_s,
414 PRIM_INT__s_s_g_s,
415 PRIM_INT__s_s_f_s,
416 P_PA,
417 aop_PQ,
418 one_over_2pq,
419 4);
420
421
422 ostei_general_vrr_I(2, 0, 5, 0, 3,
423 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
424 PRIM_INT__p_s_h_s, PRIM_INT__s_s_h_s, NULL, PRIM_INT__p_s_g_s, NULL, PRIM_INT__d_s_h_s);
425
426
427 ostei_general_vrr_I(3, 0, 6, 0, 2,
428 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
429 PRIM_INT__d_s_i_s, PRIM_INT__p_s_i_s, NULL, PRIM_INT__d_s_h_s, NULL, PRIM_INT__f_s_i_s);
430
431
432 ostei_general_vrr_I(3, 0, 7, 0, 2,
433 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
434 PRIM_INT__d_s_k_s, PRIM_INT__p_s_k_s, NULL, PRIM_INT__d_s_i_s, NULL, PRIM_INT__f_s_k_s);
435
436
437 VRR_I_p_s_f_s(
438 PRIM_INT__p_s_f_s,
439 PRIM_INT__s_s_f_s,
440 PRIM_INT__s_s_d_s,
441 P_PA,
442 aop_PQ,
443 one_over_2pq,
444 4);
445
446
447 ostei_general_vrr_I(2, 0, 4, 0, 3,
448 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
449 PRIM_INT__p_s_g_s, PRIM_INT__s_s_g_s, NULL, PRIM_INT__p_s_f_s, NULL, PRIM_INT__d_s_g_s);
450
451
452 ostei_general_vrr_I(3, 0, 5, 0, 2,
453 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
454 PRIM_INT__d_s_h_s, PRIM_INT__p_s_h_s, NULL, PRIM_INT__d_s_g_s, NULL, PRIM_INT__f_s_h_s);
455
456
457 ostei_general_vrr_I(4, 0, 6, 0, 1,
458 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
459 PRIM_INT__f_s_i_s, PRIM_INT__d_s_i_s, NULL, PRIM_INT__f_s_h_s, NULL, PRIM_INT__g_s_i_s);
460
461
462 ostei_general_vrr_I(4, 0, 7, 0, 1,
463 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
464 PRIM_INT__f_s_k_s, PRIM_INT__d_s_k_s, NULL, PRIM_INT__f_s_i_s, NULL, PRIM_INT__g_s_k_s);
465
466
467
468
469 ////////////////////////////////////
470 // Accumulate contracted integrals
471 ////////////////////////////////////
472 if(lastoffset == 0)
473 {
474 contract_all(168, PRIM_INT__d_s_i_s, PRIM_PTR_INT__d_s_i_s);
475 contract_all(216, PRIM_INT__d_s_k_s, PRIM_PTR_INT__d_s_k_s);
476 contract_all(280, PRIM_INT__f_s_i_s, PRIM_PTR_INT__f_s_i_s);
477 contract_all(360, PRIM_INT__f_s_k_s, PRIM_PTR_INT__f_s_k_s);
478 contract_all(420, PRIM_INT__g_s_i_s, PRIM_PTR_INT__g_s_i_s);
479 contract_all(540, PRIM_INT__g_s_k_s, PRIM_PTR_INT__g_s_k_s);
480 }
481 else
482 {
483 contract(168, shelloffsets, PRIM_INT__d_s_i_s, PRIM_PTR_INT__d_s_i_s);
484 contract(216, shelloffsets, PRIM_INT__d_s_k_s, PRIM_PTR_INT__d_s_k_s);
485 contract(280, shelloffsets, PRIM_INT__f_s_i_s, PRIM_PTR_INT__f_s_i_s);
486 contract(360, shelloffsets, PRIM_INT__f_s_k_s, PRIM_PTR_INT__f_s_k_s);
487 contract(420, shelloffsets, PRIM_INT__g_s_i_s, PRIM_PTR_INT__g_s_i_s);
488 contract(540, shelloffsets, PRIM_INT__g_s_k_s, PRIM_PTR_INT__g_s_k_s);
489 PRIM_PTR_INT__d_s_i_s += lastoffset*168;
490 PRIM_PTR_INT__d_s_k_s += lastoffset*216;
491 PRIM_PTR_INT__f_s_i_s += lastoffset*280;
492 PRIM_PTR_INT__f_s_k_s += lastoffset*360;
493 PRIM_PTR_INT__g_s_i_s += lastoffset*420;
494 PRIM_PTR_INT__g_s_k_s += lastoffset*540;
495 }
496
497 } // close loop over j
498 } // close loop over i
499
500 //Advance to the next batch
501 jstart = SIMINT_SIMD_ROUND(jend);
502
503 //////////////////////////////////////////////
504 // Contracted integrals: Horizontal recurrance
505 //////////////////////////////////////////////
506
507
508 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
509
510
511 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
512 {
513 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
514
515 // set up HRR pointers
516 double const * restrict HRR_INT__d_s_i_s = INT__d_s_i_s + abcd * 168;
517 double const * restrict HRR_INT__d_s_k_s = INT__d_s_k_s + abcd * 216;
518 double const * restrict HRR_INT__f_s_i_s = INT__f_s_i_s + abcd * 280;
519 double const * restrict HRR_INT__f_s_k_s = INT__f_s_k_s + abcd * 360;
520 double const * restrict HRR_INT__g_s_i_s = INT__g_s_i_s + abcd * 420;
521 double const * restrict HRR_INT__g_s_k_s = INT__g_s_k_s + abcd * 540;
522 double * restrict HRR_INT__d_d_i_p = INT__d_d_i_p + real_abcd * 3024;
523
524 // form INT__d_p_i_s
525 for(iket = 0; iket < 28; ++iket)
526 {
527 HRR_INT__d_p_i_s[0 * 28 + iket] = HRR_INT__f_s_i_s[0 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[0 * 28 + iket] );
528
529 HRR_INT__d_p_i_s[1 * 28 + iket] = HRR_INT__f_s_i_s[1 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[0 * 28 + iket] );
530
531 HRR_INT__d_p_i_s[2 * 28 + iket] = HRR_INT__f_s_i_s[2 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[0 * 28 + iket] );
532
533 HRR_INT__d_p_i_s[3 * 28 + iket] = HRR_INT__f_s_i_s[1 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[1 * 28 + iket] );
534
535 HRR_INT__d_p_i_s[4 * 28 + iket] = HRR_INT__f_s_i_s[3 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[1 * 28 + iket] );
536
537 HRR_INT__d_p_i_s[5 * 28 + iket] = HRR_INT__f_s_i_s[4 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[1 * 28 + iket] );
538
539 HRR_INT__d_p_i_s[6 * 28 + iket] = HRR_INT__f_s_i_s[2 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[2 * 28 + iket] );
540
541 HRR_INT__d_p_i_s[7 * 28 + iket] = HRR_INT__f_s_i_s[4 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[2 * 28 + iket] );
542
543 HRR_INT__d_p_i_s[8 * 28 + iket] = HRR_INT__f_s_i_s[5 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[2 * 28 + iket] );
544
545 HRR_INT__d_p_i_s[9 * 28 + iket] = HRR_INT__f_s_i_s[3 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[3 * 28 + iket] );
546
547 HRR_INT__d_p_i_s[10 * 28 + iket] = HRR_INT__f_s_i_s[6 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[3 * 28 + iket] );
548
549 HRR_INT__d_p_i_s[11 * 28 + iket] = HRR_INT__f_s_i_s[7 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[3 * 28 + iket] );
550
551 HRR_INT__d_p_i_s[12 * 28 + iket] = HRR_INT__f_s_i_s[4 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[4 * 28 + iket] );
552
553 HRR_INT__d_p_i_s[13 * 28 + iket] = HRR_INT__f_s_i_s[7 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[4 * 28 + iket] );
554
555 HRR_INT__d_p_i_s[14 * 28 + iket] = HRR_INT__f_s_i_s[8 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[4 * 28 + iket] );
556
557 HRR_INT__d_p_i_s[15 * 28 + iket] = HRR_INT__f_s_i_s[5 * 28 + iket] + ( hAB[0] * HRR_INT__d_s_i_s[5 * 28 + iket] );
558
559 HRR_INT__d_p_i_s[16 * 28 + iket] = HRR_INT__f_s_i_s[8 * 28 + iket] + ( hAB[1] * HRR_INT__d_s_i_s[5 * 28 + iket] );
560
561 HRR_INT__d_p_i_s[17 * 28 + iket] = HRR_INT__f_s_i_s[9 * 28 + iket] + ( hAB[2] * HRR_INT__d_s_i_s[5 * 28 + iket] );
562
563 }
564
565
566 // form INT__d_p_k_s
567 for(iket = 0; iket < 36; ++iket)
568 {
569 HRR_INT__d_p_k_s[0 * 36 + iket] = HRR_INT__f_s_k_s[0 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[0 * 36 + iket] );
570
571 HRR_INT__d_p_k_s[1 * 36 + iket] = HRR_INT__f_s_k_s[1 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[0 * 36 + iket] );
572
573 HRR_INT__d_p_k_s[2 * 36 + iket] = HRR_INT__f_s_k_s[2 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[0 * 36 + iket] );
574
575 HRR_INT__d_p_k_s[3 * 36 + iket] = HRR_INT__f_s_k_s[1 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[1 * 36 + iket] );
576
577 HRR_INT__d_p_k_s[4 * 36 + iket] = HRR_INT__f_s_k_s[3 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[1 * 36 + iket] );
578
579 HRR_INT__d_p_k_s[5 * 36 + iket] = HRR_INT__f_s_k_s[4 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[1 * 36 + iket] );
580
581 HRR_INT__d_p_k_s[6 * 36 + iket] = HRR_INT__f_s_k_s[2 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[2 * 36 + iket] );
582
583 HRR_INT__d_p_k_s[7 * 36 + iket] = HRR_INT__f_s_k_s[4 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[2 * 36 + iket] );
584
585 HRR_INT__d_p_k_s[8 * 36 + iket] = HRR_INT__f_s_k_s[5 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[2 * 36 + iket] );
586
587 HRR_INT__d_p_k_s[9 * 36 + iket] = HRR_INT__f_s_k_s[3 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[3 * 36 + iket] );
588
589 HRR_INT__d_p_k_s[10 * 36 + iket] = HRR_INT__f_s_k_s[6 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[3 * 36 + iket] );
590
591 HRR_INT__d_p_k_s[11 * 36 + iket] = HRR_INT__f_s_k_s[7 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[3 * 36 + iket] );
592
593 HRR_INT__d_p_k_s[12 * 36 + iket] = HRR_INT__f_s_k_s[4 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[4 * 36 + iket] );
594
595 HRR_INT__d_p_k_s[13 * 36 + iket] = HRR_INT__f_s_k_s[7 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[4 * 36 + iket] );
596
597 HRR_INT__d_p_k_s[14 * 36 + iket] = HRR_INT__f_s_k_s[8 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[4 * 36 + iket] );
598
599 HRR_INT__d_p_k_s[15 * 36 + iket] = HRR_INT__f_s_k_s[5 * 36 + iket] + ( hAB[0] * HRR_INT__d_s_k_s[5 * 36 + iket] );
600
601 HRR_INT__d_p_k_s[16 * 36 + iket] = HRR_INT__f_s_k_s[8 * 36 + iket] + ( hAB[1] * HRR_INT__d_s_k_s[5 * 36 + iket] );
602
603 HRR_INT__d_p_k_s[17 * 36 + iket] = HRR_INT__f_s_k_s[9 * 36 + iket] + ( hAB[2] * HRR_INT__d_s_k_s[5 * 36 + iket] );
604
605 }
606
607
608 // form INT__f_p_i_s
609 HRR_J_f_p(
610 HRR_INT__f_p_i_s,
611 HRR_INT__f_s_i_s,
612 HRR_INT__g_s_i_s,
613 hAB, 28);
614
615 // form INT__f_p_k_s
616 HRR_J_f_p(
617 HRR_INT__f_p_k_s,
618 HRR_INT__f_s_k_s,
619 HRR_INT__g_s_k_s,
620 hAB, 36);
621
622 // form INT__d_d_i_s
623 HRR_J_d_d(
624 HRR_INT__d_d_i_s,
625 HRR_INT__d_p_i_s,
626 HRR_INT__f_p_i_s,
627 hAB, 28);
628
629 // form INT__d_d_k_s
630 HRR_J_d_d(
631 HRR_INT__d_d_k_s,
632 HRR_INT__d_p_k_s,
633 HRR_INT__f_p_k_s,
634 hAB, 36);
635
636 // form INT__d_d_i_p
637 ostei_general_hrr_L(2, 2, 6, 1, hCD, HRR_INT__d_d_k_s, HRR_INT__d_d_i_s, HRR_INT__d_d_i_p);
638
639
640 } // close HRR loop
641
642
643 } // close loop cdbatch
644
645 istart = iend;
646 } // close loop over ab
647
648 return P.nshell12_clip * Q.nshell12_clip;
649 }
650
ostei_d_d_p_i(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__d_d_p_i)651 int ostei_d_d_p_i(struct simint_multi_shellpair const P,
652 struct simint_multi_shellpair const Q,
653 double screen_tol,
654 double * const restrict work,
655 double * const restrict INT__d_d_p_i)
656 {
657 double Q_AB[3*Q.nshell12];
658 struct simint_multi_shellpair Q_tmp = Q;
659 Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z;
660 Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z;
661 Q_tmp.AB_x = Q_AB;
662 Q_tmp.AB_y = Q_AB + Q.nshell12;
663 Q_tmp.AB_z = Q_AB + 2*Q.nshell12;
664
665 for(int i = 0; i < Q.nshell12; i++)
666 {
667 Q_tmp.AB_x[i] = -Q.AB_x[i];
668 Q_tmp.AB_y[i] = -Q.AB_y[i];
669 Q_tmp.AB_z[i] = -Q.AB_z[i];
670 }
671
672 int ret = ostei_d_d_i_p(P, Q_tmp, screen_tol, work, INT__d_d_p_i);
673 double buffer[3024] SIMINT_ALIGN_ARRAY_DBL;
674
675 for(int q = 0; q < ret; q++)
676 {
677 int idx = 0;
678 for(int a = 0; a < 6; ++a)
679 for(int b = 0; b < 6; ++b)
680 for(int c = 0; c < 3; ++c)
681 for(int d = 0; d < 28; ++d)
682 buffer[idx++] = INT__d_d_p_i[q*3024+a*504+b*84+d*3+c];
683
684 memcpy(INT__d_d_p_i+q*3024, buffer, 3024*sizeof(double));
685 }
686
687 return ret;
688 }
689
690