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_p_p_d_f(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__p_p_d_f)8 int ostei_p_p_d_f(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__p_p_d_f)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__p_p_d_f);
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__p_s_s_f = work + (SIMINT_NSHELL_SIMD * 0);
30 double * const INT__p_s_s_g = work + (SIMINT_NSHELL_SIMD * 30);
31 double * const INT__p_s_s_h = work + (SIMINT_NSHELL_SIMD * 75);
32 double * const INT__d_s_s_f = work + (SIMINT_NSHELL_SIMD * 138);
33 double * const INT__d_s_s_g = work + (SIMINT_NSHELL_SIMD * 198);
34 double * const INT__d_s_s_h = work + (SIMINT_NSHELL_SIMD * 288);
35 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*414);
36 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_p = primwork + 8;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_d = primwork + 29;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_f = primwork + 65;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_g = primwork + 115;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_h = primwork + 175;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_d = primwork + 238;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_f = primwork + 274;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_g = primwork + 334;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_h = primwork + 424;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_f = primwork + 550;
47 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_g = primwork + 610;
48 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_h = primwork + 700;
49 double * const hrrwork = (double *)(primwork + 826);
50 double * const HRR_INT__p_p_s_f = hrrwork + 0;
51 double * const HRR_INT__p_p_s_g = hrrwork + 90;
52 double * const HRR_INT__p_p_s_h = hrrwork + 225;
53 double * const HRR_INT__p_p_p_f = hrrwork + 414;
54 double * const HRR_INT__p_p_p_g = hrrwork + 684;
55
56
57 // Create constants
58 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
59 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
60 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
61 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
62 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
63 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
64
65
66 ////////////////////////////////////////
67 // Loop over shells and primitives
68 ////////////////////////////////////////
69
70 real_abcd = 0;
71 istart = 0;
72 for(ab = 0; ab < P.nshell12_clip; ++ab)
73 {
74 const int iend = istart + P.nprim12[ab];
75
76 cd = 0;
77 jstart = 0;
78
79 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
80 {
81 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
82 int jend = jstart;
83 for(i = 0; i < nshellbatch; i++)
84 jend += Q.nprim12[cd+i];
85
86 // Clear the beginning of the workspace (where we are accumulating integrals)
87 memset(work, 0, SIMINT_NSHELL_SIMD * 414 * sizeof(double));
88 abcd = 0;
89
90
91 for(i = istart; i < iend; ++i)
92 {
93 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
94
95 if(check_screen)
96 {
97 // Skip this whole thing if always insignificant
98 if((P.screen[i] * Q.screen_max) < screen_tol)
99 continue;
100 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
101 }
102
103 icd = 0;
104 iprimcd = 0;
105 nprim_icd = Q.nprim12[cd];
106 double * restrict PRIM_PTR_INT__p_s_s_f = INT__p_s_s_f + abcd * 30;
107 double * restrict PRIM_PTR_INT__p_s_s_g = INT__p_s_s_g + abcd * 45;
108 double * restrict PRIM_PTR_INT__p_s_s_h = INT__p_s_s_h + abcd * 63;
109 double * restrict PRIM_PTR_INT__d_s_s_f = INT__d_s_s_f + abcd * 60;
110 double * restrict PRIM_PTR_INT__d_s_s_g = INT__d_s_s_g + abcd * 90;
111 double * restrict PRIM_PTR_INT__d_s_s_h = INT__d_s_s_h + abcd * 126;
112
113
114
115 // Load these one per loop over i
116 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
117 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
118 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
119
120 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
121
122 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
123 {
124 // calculate the shell offsets
125 // these are the offset from the shell pointed to by cd
126 // for each element
127 int shelloffsets[SIMINT_SIMD_LEN] = {0};
128 int lastoffset = 0;
129 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
130
131 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
132 {
133 // Handle if the first element of the vector is a new shell
134 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
135 {
136 nprim_icd += Q.nprim12[cd + (++icd)];
137 PRIM_PTR_INT__p_s_s_f += 30;
138 PRIM_PTR_INT__p_s_s_g += 45;
139 PRIM_PTR_INT__p_s_s_h += 63;
140 PRIM_PTR_INT__d_s_s_f += 60;
141 PRIM_PTR_INT__d_s_s_g += 90;
142 PRIM_PTR_INT__d_s_s_h += 126;
143 }
144 iprimcd++;
145 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
146 {
147 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
148 {
149 shelloffsets[n] = shelloffsets[n-1] + 1;
150 lastoffset++;
151 nprim_icd += Q.nprim12[cd + (++icd)];
152 }
153 else
154 shelloffsets[n] = shelloffsets[n-1];
155 iprimcd++;
156 }
157 }
158 else
159 iprimcd += SIMINT_SIMD_LEN;
160
161 // Do we have to compute this vector (or has it been screened out)?
162 // (not_screened != 0 means we have to do this vector)
163 if(check_screen)
164 {
165 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
166 if(vmax < screen_tol)
167 {
168 PRIM_PTR_INT__p_s_s_f += lastoffset*30;
169 PRIM_PTR_INT__p_s_s_g += lastoffset*45;
170 PRIM_PTR_INT__p_s_s_h += lastoffset*63;
171 PRIM_PTR_INT__d_s_s_f += lastoffset*60;
172 PRIM_PTR_INT__d_s_s_g += lastoffset*90;
173 PRIM_PTR_INT__d_s_s_h += lastoffset*126;
174 continue;
175 }
176 }
177
178 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
179 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
180 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
181 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
182
183
184 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
185 SIMINT_DBLTYPE PQ[3];
186 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
187 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
188 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
189 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
190 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
191 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
192
193 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
194 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
195 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
196 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
197 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
198 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
199 const SIMINT_DBLTYPE Q_PB[3] = { SIMINT_DBLLOAD(Q.PB_x, j), SIMINT_DBLLOAD(Q.PB_y, j), SIMINT_DBLLOAD(Q.PB_z, j) };
200
201 // NOTE: Minus sign!
202 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
203 SIMINT_DBLTYPE aop_PQ[3];
204 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
205 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
206 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
207
208 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
209 SIMINT_DBLTYPE aoq_PQ[3];
210 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
211 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
212 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
213 // Put a minus sign here so we don't have to in RR routines
214 a_over_q = SIMINT_NEG(a_over_q);
215
216
217 //////////////////////////////////////////////
218 // Fjt function section
219 // Maximum v value: 7
220 //////////////////////////////////////////////
221 // The parameter to the Fjt function
222 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
223
224
225 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
226
227
228 boys_F_split(PRIM_INT__s_s_s_s, F_x, 7);
229 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
230 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
231 for(n = 0; n <= 7; n++)
232 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
233
234 //////////////////////////////////////////////
235 // Primitive integrals: Vertical recurrance
236 //////////////////////////////////////////////
237
238 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
239 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
240 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
241 const SIMINT_DBLTYPE vrr_const_3_over_2q = SIMINT_MUL(const_3, one_over_2q);
242 const SIMINT_DBLTYPE vrr_const_4_over_2q = SIMINT_MUL(const_4, one_over_2q);
243 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
244 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
245 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
246 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
247 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
248
249
250
251 // Forming PRIM_INT__s_s_s_p[7 * 3];
252 for(n = 0; n < 7; ++n) // loop over orders of auxiliary function
253 {
254
255 PRIM_INT__s_s_s_p[n * 3 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
256 PRIM_INT__s_s_s_p[n * 3 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 0]);
257
258 PRIM_INT__s_s_s_p[n * 3 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
259 PRIM_INT__s_s_s_p[n * 3 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 1]);
260
261 PRIM_INT__s_s_s_p[n * 3 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
262 PRIM_INT__s_s_s_p[n * 3 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 2]);
263
264 }
265
266
267
268 // Forming PRIM_INT__s_s_s_d[6 * 6];
269 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
270 {
271
272 PRIM_INT__s_s_s_d[n * 6 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_p[n * 3 + 0]);
273 PRIM_INT__s_s_s_d[n * 6 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 0]);
274 PRIM_INT__s_s_s_d[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_s_d[n * 6 + 0]);
275
276 PRIM_INT__s_s_s_d[n * 6 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_p[n * 3 + 0]);
277 PRIM_INT__s_s_s_d[n * 6 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 1]);
278
279 PRIM_INT__s_s_s_d[n * 6 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 0]);
280 PRIM_INT__s_s_s_d[n * 6 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 2]);
281
282 PRIM_INT__s_s_s_d[n * 6 + 3] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_p[n * 3 + 1]);
283 PRIM_INT__s_s_s_d[n * 6 + 3] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_d[n * 6 + 3]);
284 PRIM_INT__s_s_s_d[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_s_d[n * 6 + 3]);
285
286 PRIM_INT__s_s_s_d[n * 6 + 4] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 1]);
287 PRIM_INT__s_s_s_d[n * 6 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_d[n * 6 + 4]);
288
289 PRIM_INT__s_s_s_d[n * 6 + 5] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 2]);
290 PRIM_INT__s_s_s_d[n * 6 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__s_s_s_d[n * 6 + 5]);
291 PRIM_INT__s_s_s_d[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_s_d[n * 6 + 5]);
292
293 }
294
295
296
297 // Forming PRIM_INT__s_s_s_f[5 * 10];
298 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
299 {
300
301 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 0]);
302 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 0]);
303 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_p[n * 3 + 0]), PRIM_INT__s_s_s_f[n * 10 + 0]);
304
305 PRIM_INT__s_s_s_f[n * 10 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 0]);
306 PRIM_INT__s_s_s_f[n * 10 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 1]);
307
308 PRIM_INT__s_s_s_f[n * 10 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 0]);
309 PRIM_INT__s_s_s_f[n * 10 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 2]);
310
311 PRIM_INT__s_s_s_f[n * 10 + 3] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 3]);
312 PRIM_INT__s_s_s_f[n * 10 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 3]);
313
314 PRIM_INT__s_s_s_f[n * 10 + 4] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 1]);
315 PRIM_INT__s_s_s_f[n * 10 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 1], PRIM_INT__s_s_s_f[n * 10 + 4]);
316
317 PRIM_INT__s_s_s_f[n * 10 + 5] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 5]);
318 PRIM_INT__s_s_s_f[n * 10 + 5] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 5]);
319
320 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 3]);
321 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 6]);
322 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_p[n * 3 + 1]), PRIM_INT__s_s_s_f[n * 10 + 6]);
323
324 PRIM_INT__s_s_s_f[n * 10 + 7] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 3]);
325 PRIM_INT__s_s_s_f[n * 10 + 7] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 7]);
326
327 PRIM_INT__s_s_s_f[n * 10 + 8] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 5]);
328 PRIM_INT__s_s_s_f[n * 10 + 8] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 8]);
329
330 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 5]);
331 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 9]);
332 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__s_s_s_p[n * 3 + 2]), PRIM_INT__s_s_s_f[n * 10 + 9]);
333
334 }
335
336
337 VRR_I_p_s_s_f(
338 PRIM_INT__p_s_s_f,
339 PRIM_INT__s_s_s_f,
340 PRIM_INT__s_s_s_d,
341 P_PA,
342 aop_PQ,
343 one_over_2pq,
344 2);
345
346
347 VRR_L_s_s_s_g(
348 PRIM_INT__s_s_s_g,
349 PRIM_INT__s_s_s_f,
350 PRIM_INT__s_s_s_d,
351 Q_PB,
352 a_over_q,
353 aoq_PQ,
354 one_over_2q,
355 4);
356
357
358 VRR_I_p_s_s_g(
359 PRIM_INT__p_s_s_g,
360 PRIM_INT__s_s_s_g,
361 PRIM_INT__s_s_s_f,
362 P_PA,
363 aop_PQ,
364 one_over_2pq,
365 2);
366
367
368
369 // Forming PRIM_INT__p_s_s_d[2 * 18];
370 for(n = 0; n < 2; ++n) // loop over orders of auxiliary function
371 {
372
373 PRIM_INT__p_s_s_d[n * 18 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 0]);
374 PRIM_INT__p_s_s_d[n * 18 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__p_s_s_d[n * 18 + 0]);
375 PRIM_INT__p_s_s_d[n * 18 + 0] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__p_s_s_d[n * 18 + 0]);
376
377 PRIM_INT__p_s_s_d[n * 18 + 1] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 1]);
378 PRIM_INT__p_s_s_d[n * 18 + 1] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 1], PRIM_INT__p_s_s_d[n * 18 + 1]);
379 PRIM_INT__p_s_s_d[n * 18 + 1] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__p_s_s_d[n * 18 + 1]);
380
381 PRIM_INT__p_s_s_d[n * 18 + 2] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 2]);
382 PRIM_INT__p_s_s_d[n * 18 + 2] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 2], PRIM_INT__p_s_s_d[n * 18 + 2]);
383 PRIM_INT__p_s_s_d[n * 18 + 2] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__p_s_s_d[n * 18 + 2]);
384
385 PRIM_INT__p_s_s_d[n * 18 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 3]);
386 PRIM_INT__p_s_s_d[n * 18 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__p_s_s_d[n * 18 + 3]);
387
388 PRIM_INT__p_s_s_d[n * 18 + 4] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 4]);
389 PRIM_INT__p_s_s_d[n * 18 + 4] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 4], PRIM_INT__p_s_s_d[n * 18 + 4]);
390
391 PRIM_INT__p_s_s_d[n * 18 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_d[n * 6 + 5]);
392 PRIM_INT__p_s_s_d[n * 18 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__p_s_s_d[n * 18 + 5]);
393
394 PRIM_INT__p_s_s_d[n * 18 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 0]);
395 PRIM_INT__p_s_s_d[n * 18 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__p_s_s_d[n * 18 + 6]);
396
397 PRIM_INT__p_s_s_d[n * 18 + 7] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 1]);
398 PRIM_INT__p_s_s_d[n * 18 + 7] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 1], PRIM_INT__p_s_s_d[n * 18 + 7]);
399 PRIM_INT__p_s_s_d[n * 18 + 7] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__p_s_s_d[n * 18 + 7]);
400
401 PRIM_INT__p_s_s_d[n * 18 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 2]);
402 PRIM_INT__p_s_s_d[n * 18 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 2], PRIM_INT__p_s_s_d[n * 18 + 8]);
403
404 PRIM_INT__p_s_s_d[n * 18 + 9] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 3]);
405 PRIM_INT__p_s_s_d[n * 18 + 9] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__p_s_s_d[n * 18 + 9]);
406 PRIM_INT__p_s_s_d[n * 18 + 9] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__p_s_s_d[n * 18 + 9]);
407
408 PRIM_INT__p_s_s_d[n * 18 + 10] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 4]);
409 PRIM_INT__p_s_s_d[n * 18 + 10] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 4], PRIM_INT__p_s_s_d[n * 18 + 10]);
410 PRIM_INT__p_s_s_d[n * 18 + 10] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__p_s_s_d[n * 18 + 10]);
411
412 PRIM_INT__p_s_s_d[n * 18 + 11] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_d[n * 6 + 5]);
413 PRIM_INT__p_s_s_d[n * 18 + 11] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__p_s_s_d[n * 18 + 11]);
414
415 PRIM_INT__p_s_s_d[n * 18 + 12] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 0]);
416 PRIM_INT__p_s_s_d[n * 18 + 12] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__p_s_s_d[n * 18 + 12]);
417
418 PRIM_INT__p_s_s_d[n * 18 + 13] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 1]);
419 PRIM_INT__p_s_s_d[n * 18 + 13] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 1], PRIM_INT__p_s_s_d[n * 18 + 13]);
420
421 PRIM_INT__p_s_s_d[n * 18 + 14] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 2]);
422 PRIM_INT__p_s_s_d[n * 18 + 14] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 2], PRIM_INT__p_s_s_d[n * 18 + 14]);
423 PRIM_INT__p_s_s_d[n * 18 + 14] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__p_s_s_d[n * 18 + 14]);
424
425 PRIM_INT__p_s_s_d[n * 18 + 15] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 3]);
426 PRIM_INT__p_s_s_d[n * 18 + 15] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__p_s_s_d[n * 18 + 15]);
427
428 PRIM_INT__p_s_s_d[n * 18 + 16] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 4]);
429 PRIM_INT__p_s_s_d[n * 18 + 16] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 4], PRIM_INT__p_s_s_d[n * 18 + 16]);
430 PRIM_INT__p_s_s_d[n * 18 + 16] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__p_s_s_d[n * 18 + 16]);
431
432 PRIM_INT__p_s_s_d[n * 18 + 17] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_d[n * 6 + 5]);
433 PRIM_INT__p_s_s_d[n * 18 + 17] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__p_s_s_d[n * 18 + 17]);
434 PRIM_INT__p_s_s_d[n * 18 + 17] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__p_s_s_d[n * 18 + 17]);
435
436 }
437
438
439 VRR_I_d_s_s_f(
440 PRIM_INT__d_s_s_f,
441 PRIM_INT__p_s_s_f,
442 PRIM_INT__s_s_s_f,
443 PRIM_INT__p_s_s_d,
444 P_PA,
445 a_over_p,
446 aop_PQ,
447 one_over_2p,
448 one_over_2pq,
449 1);
450
451
452 VRR_L_s_s_s_h(
453 PRIM_INT__s_s_s_h,
454 PRIM_INT__s_s_s_g,
455 PRIM_INT__s_s_s_f,
456 Q_PB,
457 a_over_q,
458 aoq_PQ,
459 one_over_2q,
460 3);
461
462
463 ostei_general_vrr_I(1, 0, 0, 5, 2,
464 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
465 PRIM_INT__s_s_s_h, NULL, NULL, NULL, PRIM_INT__s_s_s_g, PRIM_INT__p_s_s_h);
466
467
468 ostei_general_vrr_I(2, 0, 0, 4, 1,
469 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
470 PRIM_INT__p_s_s_g, PRIM_INT__s_s_s_g, NULL, NULL, PRIM_INT__p_s_s_f, PRIM_INT__d_s_s_g);
471
472
473 ostei_general_vrr_I(2, 0, 0, 5, 1,
474 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
475 PRIM_INT__p_s_s_h, PRIM_INT__s_s_s_h, NULL, NULL, PRIM_INT__p_s_s_g, PRIM_INT__d_s_s_h);
476
477
478
479
480 ////////////////////////////////////
481 // Accumulate contracted integrals
482 ////////////////////////////////////
483 if(lastoffset == 0)
484 {
485 contract_all(30, PRIM_INT__p_s_s_f, PRIM_PTR_INT__p_s_s_f);
486 contract_all(45, PRIM_INT__p_s_s_g, PRIM_PTR_INT__p_s_s_g);
487 contract_all(63, PRIM_INT__p_s_s_h, PRIM_PTR_INT__p_s_s_h);
488 contract_all(60, PRIM_INT__d_s_s_f, PRIM_PTR_INT__d_s_s_f);
489 contract_all(90, PRIM_INT__d_s_s_g, PRIM_PTR_INT__d_s_s_g);
490 contract_all(126, PRIM_INT__d_s_s_h, PRIM_PTR_INT__d_s_s_h);
491 }
492 else
493 {
494 contract(30, shelloffsets, PRIM_INT__p_s_s_f, PRIM_PTR_INT__p_s_s_f);
495 contract(45, shelloffsets, PRIM_INT__p_s_s_g, PRIM_PTR_INT__p_s_s_g);
496 contract(63, shelloffsets, PRIM_INT__p_s_s_h, PRIM_PTR_INT__p_s_s_h);
497 contract(60, shelloffsets, PRIM_INT__d_s_s_f, PRIM_PTR_INT__d_s_s_f);
498 contract(90, shelloffsets, PRIM_INT__d_s_s_g, PRIM_PTR_INT__d_s_s_g);
499 contract(126, shelloffsets, PRIM_INT__d_s_s_h, PRIM_PTR_INT__d_s_s_h);
500 PRIM_PTR_INT__p_s_s_f += lastoffset*30;
501 PRIM_PTR_INT__p_s_s_g += lastoffset*45;
502 PRIM_PTR_INT__p_s_s_h += lastoffset*63;
503 PRIM_PTR_INT__d_s_s_f += lastoffset*60;
504 PRIM_PTR_INT__d_s_s_g += lastoffset*90;
505 PRIM_PTR_INT__d_s_s_h += lastoffset*126;
506 }
507
508 } // close loop over j
509 } // close loop over i
510
511 //Advance to the next batch
512 jstart = SIMINT_SIMD_ROUND(jend);
513
514 //////////////////////////////////////////////
515 // Contracted integrals: Horizontal recurrance
516 //////////////////////////////////////////////
517
518
519 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
520
521
522 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
523 {
524 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
525
526 // set up HRR pointers
527 double const * restrict HRR_INT__p_s_s_f = INT__p_s_s_f + abcd * 30;
528 double const * restrict HRR_INT__p_s_s_g = INT__p_s_s_g + abcd * 45;
529 double const * restrict HRR_INT__p_s_s_h = INT__p_s_s_h + abcd * 63;
530 double const * restrict HRR_INT__d_s_s_f = INT__d_s_s_f + abcd * 60;
531 double const * restrict HRR_INT__d_s_s_g = INT__d_s_s_g + abcd * 90;
532 double const * restrict HRR_INT__d_s_s_h = INT__d_s_s_h + abcd * 126;
533 double * restrict HRR_INT__p_p_d_f = INT__p_p_d_f + real_abcd * 540;
534
535 // form INT__p_p_s_f
536 for(iket = 0; iket < 10; ++iket)
537 {
538 HRR_INT__p_p_s_f[0 * 10 + iket] = HRR_INT__d_s_s_f[0 * 10 + iket] + ( hAB[0] * HRR_INT__p_s_s_f[0 * 10 + iket] );
539
540 HRR_INT__p_p_s_f[1 * 10 + iket] = HRR_INT__d_s_s_f[1 * 10 + iket] + ( hAB[1] * HRR_INT__p_s_s_f[0 * 10 + iket] );
541
542 HRR_INT__p_p_s_f[2 * 10 + iket] = HRR_INT__d_s_s_f[2 * 10 + iket] + ( hAB[2] * HRR_INT__p_s_s_f[0 * 10 + iket] );
543
544 HRR_INT__p_p_s_f[3 * 10 + iket] = HRR_INT__d_s_s_f[1 * 10 + iket] + ( hAB[0] * HRR_INT__p_s_s_f[1 * 10 + iket] );
545
546 HRR_INT__p_p_s_f[4 * 10 + iket] = HRR_INT__d_s_s_f[3 * 10 + iket] + ( hAB[1] * HRR_INT__p_s_s_f[1 * 10 + iket] );
547
548 HRR_INT__p_p_s_f[5 * 10 + iket] = HRR_INT__d_s_s_f[4 * 10 + iket] + ( hAB[2] * HRR_INT__p_s_s_f[1 * 10 + iket] );
549
550 HRR_INT__p_p_s_f[6 * 10 + iket] = HRR_INT__d_s_s_f[2 * 10 + iket] + ( hAB[0] * HRR_INT__p_s_s_f[2 * 10 + iket] );
551
552 HRR_INT__p_p_s_f[7 * 10 + iket] = HRR_INT__d_s_s_f[4 * 10 + iket] + ( hAB[1] * HRR_INT__p_s_s_f[2 * 10 + iket] );
553
554 HRR_INT__p_p_s_f[8 * 10 + iket] = HRR_INT__d_s_s_f[5 * 10 + iket] + ( hAB[2] * HRR_INT__p_s_s_f[2 * 10 + iket] );
555
556 }
557
558
559 // form INT__p_p_s_g
560 for(iket = 0; iket < 15; ++iket)
561 {
562 HRR_INT__p_p_s_g[0 * 15 + iket] = HRR_INT__d_s_s_g[0 * 15 + iket] + ( hAB[0] * HRR_INT__p_s_s_g[0 * 15 + iket] );
563
564 HRR_INT__p_p_s_g[1 * 15 + iket] = HRR_INT__d_s_s_g[1 * 15 + iket] + ( hAB[1] * HRR_INT__p_s_s_g[0 * 15 + iket] );
565
566 HRR_INT__p_p_s_g[2 * 15 + iket] = HRR_INT__d_s_s_g[2 * 15 + iket] + ( hAB[2] * HRR_INT__p_s_s_g[0 * 15 + iket] );
567
568 HRR_INT__p_p_s_g[3 * 15 + iket] = HRR_INT__d_s_s_g[1 * 15 + iket] + ( hAB[0] * HRR_INT__p_s_s_g[1 * 15 + iket] );
569
570 HRR_INT__p_p_s_g[4 * 15 + iket] = HRR_INT__d_s_s_g[3 * 15 + iket] + ( hAB[1] * HRR_INT__p_s_s_g[1 * 15 + iket] );
571
572 HRR_INT__p_p_s_g[5 * 15 + iket] = HRR_INT__d_s_s_g[4 * 15 + iket] + ( hAB[2] * HRR_INT__p_s_s_g[1 * 15 + iket] );
573
574 HRR_INT__p_p_s_g[6 * 15 + iket] = HRR_INT__d_s_s_g[2 * 15 + iket] + ( hAB[0] * HRR_INT__p_s_s_g[2 * 15 + iket] );
575
576 HRR_INT__p_p_s_g[7 * 15 + iket] = HRR_INT__d_s_s_g[4 * 15 + iket] + ( hAB[1] * HRR_INT__p_s_s_g[2 * 15 + iket] );
577
578 HRR_INT__p_p_s_g[8 * 15 + iket] = HRR_INT__d_s_s_g[5 * 15 + iket] + ( hAB[2] * HRR_INT__p_s_s_g[2 * 15 + iket] );
579
580 }
581
582
583 // form INT__p_p_s_h
584 for(iket = 0; iket < 21; ++iket)
585 {
586 HRR_INT__p_p_s_h[0 * 21 + iket] = HRR_INT__d_s_s_h[0 * 21 + iket] + ( hAB[0] * HRR_INT__p_s_s_h[0 * 21 + iket] );
587
588 HRR_INT__p_p_s_h[1 * 21 + iket] = HRR_INT__d_s_s_h[1 * 21 + iket] + ( hAB[1] * HRR_INT__p_s_s_h[0 * 21 + iket] );
589
590 HRR_INT__p_p_s_h[2 * 21 + iket] = HRR_INT__d_s_s_h[2 * 21 + iket] + ( hAB[2] * HRR_INT__p_s_s_h[0 * 21 + iket] );
591
592 HRR_INT__p_p_s_h[3 * 21 + iket] = HRR_INT__d_s_s_h[1 * 21 + iket] + ( hAB[0] * HRR_INT__p_s_s_h[1 * 21 + iket] );
593
594 HRR_INT__p_p_s_h[4 * 21 + iket] = HRR_INT__d_s_s_h[3 * 21 + iket] + ( hAB[1] * HRR_INT__p_s_s_h[1 * 21 + iket] );
595
596 HRR_INT__p_p_s_h[5 * 21 + iket] = HRR_INT__d_s_s_h[4 * 21 + iket] + ( hAB[2] * HRR_INT__p_s_s_h[1 * 21 + iket] );
597
598 HRR_INT__p_p_s_h[6 * 21 + iket] = HRR_INT__d_s_s_h[2 * 21 + iket] + ( hAB[0] * HRR_INT__p_s_s_h[2 * 21 + iket] );
599
600 HRR_INT__p_p_s_h[7 * 21 + iket] = HRR_INT__d_s_s_h[4 * 21 + iket] + ( hAB[1] * HRR_INT__p_s_s_h[2 * 21 + iket] );
601
602 HRR_INT__p_p_s_h[8 * 21 + iket] = HRR_INT__d_s_s_h[5 * 21 + iket] + ( hAB[2] * HRR_INT__p_s_s_h[2 * 21 + iket] );
603
604 }
605
606
607 // form INT__p_p_p_f
608 HRR_K_p_f(
609 HRR_INT__p_p_p_f,
610 HRR_INT__p_p_s_f,
611 HRR_INT__p_p_s_g,
612 hCD, 9);
613
614 // form INT__p_p_p_g
615 HRR_K_p_g(
616 HRR_INT__p_p_p_g,
617 HRR_INT__p_p_s_g,
618 HRR_INT__p_p_s_h,
619 hCD, 9);
620
621 // form INT__p_p_d_f
622 HRR_K_d_f(
623 HRR_INT__p_p_d_f,
624 HRR_INT__p_p_p_f,
625 HRR_INT__p_p_p_g,
626 hCD, 9);
627
628
629 } // close HRR loop
630
631
632 } // close loop cdbatch
633
634 istart = iend;
635 } // close loop over ab
636
637 return P.nshell12_clip * Q.nshell12_clip;
638 }
639
640