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_f_p_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_f_p_p)8 int ostei_d_f_p_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_f_p_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__d_f_p_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__s_f_p_s = work + (SIMINT_NSHELL_SIMD * 0);
30 double * const INT__s_f_d_s = work + (SIMINT_NSHELL_SIMD * 30);
31 double * const INT__s_g_p_s = work + (SIMINT_NSHELL_SIMD * 90);
32 double * const INT__s_g_d_s = work + (SIMINT_NSHELL_SIMD * 135);
33 double * const INT__s_h_p_s = work + (SIMINT_NSHELL_SIMD * 225);
34 double * const INT__s_h_d_s = 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_p_s_s = primwork + 8;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_s_s = primwork + 29;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_p_s = primwork + 65;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_s_s = primwork + 101;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_p_s = primwork + 151;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_d_s = primwork + 211;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_s_s = primwork + 271;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_p_s = primwork + 331;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_d_s = primwork + 421;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__s_h_s_s = primwork + 511;
47 SIMINT_DBLTYPE * const restrict PRIM_INT__s_h_p_s = primwork + 574;
48 SIMINT_DBLTYPE * const restrict PRIM_INT__s_h_d_s = primwork + 700;
49 double * const hrrwork = (double *)(primwork + 826);
50 double * const HRR_INT__p_f_p_s = hrrwork + 0;
51 double * const HRR_INT__p_f_d_s = hrrwork + 90;
52 double * const HRR_INT__p_g_p_s = hrrwork + 270;
53 double * const HRR_INT__p_g_d_s = hrrwork + 405;
54 double * const HRR_INT__d_f_p_s = hrrwork + 675;
55 double * const HRR_INT__d_f_d_s = hrrwork + 855;
56
57
58 // Create constants
59 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
60 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
61 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
62 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
63 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
64 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
65
66
67 ////////////////////////////////////////
68 // Loop over shells and primitives
69 ////////////////////////////////////////
70
71 real_abcd = 0;
72 istart = 0;
73 for(ab = 0; ab < P.nshell12_clip; ++ab)
74 {
75 const int iend = istart + P.nprim12[ab];
76
77 cd = 0;
78 jstart = 0;
79
80 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
81 {
82 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
83 int jend = jstart;
84 for(i = 0; i < nshellbatch; i++)
85 jend += Q.nprim12[cd+i];
86
87 // Clear the beginning of the workspace (where we are accumulating integrals)
88 memset(work, 0, SIMINT_NSHELL_SIMD * 414 * sizeof(double));
89 abcd = 0;
90
91
92 for(i = istart; i < iend; ++i)
93 {
94 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
95
96 if(check_screen)
97 {
98 // Skip this whole thing if always insignificant
99 if((P.screen[i] * Q.screen_max) < screen_tol)
100 continue;
101 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
102 }
103
104 icd = 0;
105 iprimcd = 0;
106 nprim_icd = Q.nprim12[cd];
107 double * restrict PRIM_PTR_INT__s_f_p_s = INT__s_f_p_s + abcd * 30;
108 double * restrict PRIM_PTR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
109 double * restrict PRIM_PTR_INT__s_g_p_s = INT__s_g_p_s + abcd * 45;
110 double * restrict PRIM_PTR_INT__s_g_d_s = INT__s_g_d_s + abcd * 90;
111 double * restrict PRIM_PTR_INT__s_h_p_s = INT__s_h_p_s + abcd * 63;
112 double * restrict PRIM_PTR_INT__s_h_d_s = INT__s_h_d_s + abcd * 126;
113
114
115
116 // Load these one per loop over i
117 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
118 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
119 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
120
121 const SIMINT_DBLTYPE P_PB[3] = { SIMINT_DBLSET1(P.PB_x[i]), SIMINT_DBLSET1(P.PB_y[i]), SIMINT_DBLSET1(P.PB_z[i]) };
122
123 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
124 {
125 // calculate the shell offsets
126 // these are the offset from the shell pointed to by cd
127 // for each element
128 int shelloffsets[SIMINT_SIMD_LEN] = {0};
129 int lastoffset = 0;
130 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
131
132 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
133 {
134 // Handle if the first element of the vector is a new shell
135 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
136 {
137 nprim_icd += Q.nprim12[cd + (++icd)];
138 PRIM_PTR_INT__s_f_p_s += 30;
139 PRIM_PTR_INT__s_f_d_s += 60;
140 PRIM_PTR_INT__s_g_p_s += 45;
141 PRIM_PTR_INT__s_g_d_s += 90;
142 PRIM_PTR_INT__s_h_p_s += 63;
143 PRIM_PTR_INT__s_h_d_s += 126;
144 }
145 iprimcd++;
146 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
147 {
148 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
149 {
150 shelloffsets[n] = shelloffsets[n-1] + 1;
151 lastoffset++;
152 nprim_icd += Q.nprim12[cd + (++icd)];
153 }
154 else
155 shelloffsets[n] = shelloffsets[n-1];
156 iprimcd++;
157 }
158 }
159 else
160 iprimcd += SIMINT_SIMD_LEN;
161
162 // Do we have to compute this vector (or has it been screened out)?
163 // (not_screened != 0 means we have to do this vector)
164 if(check_screen)
165 {
166 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
167 if(vmax < screen_tol)
168 {
169 PRIM_PTR_INT__s_f_p_s += lastoffset*30;
170 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
171 PRIM_PTR_INT__s_g_p_s += lastoffset*45;
172 PRIM_PTR_INT__s_g_d_s += lastoffset*90;
173 PRIM_PTR_INT__s_h_p_s += lastoffset*63;
174 PRIM_PTR_INT__s_h_d_s += lastoffset*126;
175 continue;
176 }
177 }
178
179 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
180 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
181 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
182 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
183
184
185 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
186 SIMINT_DBLTYPE PQ[3];
187 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
188 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
189 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
190 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
191 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
192 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
193
194 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
195 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
196 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
197 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
198 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
199 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
200 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
201
202 // NOTE: Minus sign!
203 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
204 SIMINT_DBLTYPE aop_PQ[3];
205 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
206 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
207 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
208
209 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
210 SIMINT_DBLTYPE aoq_PQ[3];
211 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
212 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
213 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
214 // Put a minus sign here so we don't have to in RR routines
215 a_over_q = SIMINT_NEG(a_over_q);
216
217
218 //////////////////////////////////////////////
219 // Fjt function section
220 // Maximum v value: 7
221 //////////////////////////////////////////////
222 // The parameter to the Fjt function
223 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
224
225
226 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
227
228
229 boys_F_split(PRIM_INT__s_s_s_s, F_x, 7);
230 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
231 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
232 for(n = 0; n <= 7; n++)
233 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
234
235 //////////////////////////////////////////////
236 // Primitive integrals: Vertical recurrance
237 //////////////////////////////////////////////
238
239 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
240 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
241 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
242 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
243 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
244 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
245 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
246 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
247 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
248 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
249
250
251
252 // Forming PRIM_INT__s_p_s_s[7 * 3];
253 for(n = 0; n < 7; ++n) // loop over orders of auxiliary function
254 {
255
256 PRIM_INT__s_p_s_s[n * 3 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
257 PRIM_INT__s_p_s_s[n * 3 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_s_s[n * 3 + 0]);
258
259 PRIM_INT__s_p_s_s[n * 3 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
260 PRIM_INT__s_p_s_s[n * 3 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_s_s[n * 3 + 1]);
261
262 PRIM_INT__s_p_s_s[n * 3 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
263 PRIM_INT__s_p_s_s[n * 3 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_s_s[n * 3 + 2]);
264
265 }
266
267
268
269 // Forming PRIM_INT__s_d_s_s[6 * 6];
270 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
271 {
272
273 PRIM_INT__s_d_s_s[n * 6 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_p_s_s[n * 3 + 0]);
274 PRIM_INT__s_d_s_s[n * 6 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_s_s[n * 6 + 0]);
275 PRIM_INT__s_d_s_s[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_d_s_s[n * 6 + 0]);
276
277 PRIM_INT__s_d_s_s[n * 6 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 0]);
278 PRIM_INT__s_d_s_s[n * 6 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_s_s[n * 6 + 1]);
279
280 PRIM_INT__s_d_s_s[n * 6 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 0]);
281 PRIM_INT__s_d_s_s[n * 6 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_s_s[n * 6 + 2]);
282
283 PRIM_INT__s_d_s_s[n * 6 + 3] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 1]);
284 PRIM_INT__s_d_s_s[n * 6 + 3] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_d_s_s[n * 6 + 3]);
285 PRIM_INT__s_d_s_s[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_d_s_s[n * 6 + 3]);
286
287 PRIM_INT__s_d_s_s[n * 6 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 1]);
288 PRIM_INT__s_d_s_s[n * 6 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_d_s_s[n * 6 + 4]);
289
290 PRIM_INT__s_d_s_s[n * 6 + 5] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 2]);
291 PRIM_INT__s_d_s_s[n * 6 + 5] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_d_s_s[n * 6 + 5]);
292 PRIM_INT__s_d_s_s[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_d_s_s[n * 6 + 5]);
293
294 }
295
296
297
298 // Forming PRIM_INT__s_f_s_s[5 * 10];
299 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
300 {
301
302 PRIM_INT__s_f_s_s[n * 10 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
303 PRIM_INT__s_f_s_s[n * 10 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_f_s_s[n * 10 + 0]);
304 PRIM_INT__s_f_s_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_p_s_s[n * 3 + 0]), PRIM_INT__s_f_s_s[n * 10 + 0]);
305
306 PRIM_INT__s_f_s_s[n * 10 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
307 PRIM_INT__s_f_s_s[n * 10 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_f_s_s[n * 10 + 1]);
308
309 PRIM_INT__s_f_s_s[n * 10 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
310 PRIM_INT__s_f_s_s[n * 10 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_f_s_s[n * 10 + 2]);
311
312 PRIM_INT__s_f_s_s[n * 10 + 3] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
313 PRIM_INT__s_f_s_s[n * 10 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_f_s_s[n * 10 + 3]);
314
315 PRIM_INT__s_f_s_s[n * 10 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
316 PRIM_INT__s_f_s_s[n * 10 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 1], PRIM_INT__s_f_s_s[n * 10 + 4]);
317
318 PRIM_INT__s_f_s_s[n * 10 + 5] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
319 PRIM_INT__s_f_s_s[n * 10 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_f_s_s[n * 10 + 5]);
320
321 PRIM_INT__s_f_s_s[n * 10 + 6] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
322 PRIM_INT__s_f_s_s[n * 10 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_f_s_s[n * 10 + 6]);
323 PRIM_INT__s_f_s_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_p_s_s[n * 3 + 1]), PRIM_INT__s_f_s_s[n * 10 + 6]);
324
325 PRIM_INT__s_f_s_s[n * 10 + 7] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
326 PRIM_INT__s_f_s_s[n * 10 + 7] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_f_s_s[n * 10 + 7]);
327
328 PRIM_INT__s_f_s_s[n * 10 + 8] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
329 PRIM_INT__s_f_s_s[n * 10 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_f_s_s[n * 10 + 8]);
330
331 PRIM_INT__s_f_s_s[n * 10 + 9] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
332 PRIM_INT__s_f_s_s[n * 10 + 9] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_f_s_s[n * 10 + 9]);
333 PRIM_INT__s_f_s_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_p_s_s[n * 3 + 2]), PRIM_INT__s_f_s_s[n * 10 + 9]);
334
335 }
336
337
338 VRR_K_s_f_p_s(
339 PRIM_INT__s_f_p_s,
340 PRIM_INT__s_f_s_s,
341 PRIM_INT__s_d_s_s,
342 Q_PA,
343 aoq_PQ,
344 one_over_2pq,
345 2);
346
347
348
349 // Forming PRIM_INT__s_d_p_s[2 * 18];
350 for(n = 0; n < 2; ++n) // loop over orders of auxiliary function
351 {
352
353 PRIM_INT__s_d_p_s[n * 18 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
354 PRIM_INT__s_d_p_s[n * 18 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_d_p_s[n * 18 + 0]);
355 PRIM_INT__s_d_p_s[n * 18 + 0] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_p_s[n * 18 + 0]);
356
357 PRIM_INT__s_d_p_s[n * 18 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
358 PRIM_INT__s_d_p_s[n * 18 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_d_p_s[n * 18 + 1]);
359
360 PRIM_INT__s_d_p_s[n * 18 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
361 PRIM_INT__s_d_p_s[n * 18 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 0], PRIM_INT__s_d_p_s[n * 18 + 2]);
362
363 PRIM_INT__s_d_p_s[n * 18 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 1]);
364 PRIM_INT__s_d_p_s[n * 18 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 1], PRIM_INT__s_d_p_s[n * 18 + 3]);
365 PRIM_INT__s_d_p_s[n * 18 + 3] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_d_p_s[n * 18 + 3]);
366
367 PRIM_INT__s_d_p_s[n * 18 + 4] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 1]);
368 PRIM_INT__s_d_p_s[n * 18 + 4] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 1], PRIM_INT__s_d_p_s[n * 18 + 4]);
369 PRIM_INT__s_d_p_s[n * 18 + 4] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_p_s[n * 18 + 4]);
370
371 PRIM_INT__s_d_p_s[n * 18 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
372 PRIM_INT__s_d_p_s[n * 18 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 1], PRIM_INT__s_d_p_s[n * 18 + 5]);
373
374 PRIM_INT__s_d_p_s[n * 18 + 6] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 2]);
375 PRIM_INT__s_d_p_s[n * 18 + 6] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 2], PRIM_INT__s_d_p_s[n * 18 + 6]);
376 PRIM_INT__s_d_p_s[n * 18 + 6] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_d_p_s[n * 18 + 6]);
377
378 PRIM_INT__s_d_p_s[n * 18 + 7] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 2]);
379 PRIM_INT__s_d_p_s[n * 18 + 7] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 2], PRIM_INT__s_d_p_s[n * 18 + 7]);
380
381 PRIM_INT__s_d_p_s[n * 18 + 8] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 2]);
382 PRIM_INT__s_d_p_s[n * 18 + 8] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 2], PRIM_INT__s_d_p_s[n * 18 + 8]);
383 PRIM_INT__s_d_p_s[n * 18 + 8] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_d_p_s[n * 18 + 8]);
384
385 PRIM_INT__s_d_p_s[n * 18 + 9] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
386 PRIM_INT__s_d_p_s[n * 18 + 9] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_d_p_s[n * 18 + 9]);
387
388 PRIM_INT__s_d_p_s[n * 18 + 10] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
389 PRIM_INT__s_d_p_s[n * 18 + 10] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_d_p_s[n * 18 + 10]);
390 PRIM_INT__s_d_p_s[n * 18 + 10] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_d_p_s[n * 18 + 10]);
391
392 PRIM_INT__s_d_p_s[n * 18 + 11] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
393 PRIM_INT__s_d_p_s[n * 18 + 11] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 3], PRIM_INT__s_d_p_s[n * 18 + 11]);
394
395 PRIM_INT__s_d_p_s[n * 18 + 12] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 4]);
396 PRIM_INT__s_d_p_s[n * 18 + 12] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 4], PRIM_INT__s_d_p_s[n * 18 + 12]);
397
398 PRIM_INT__s_d_p_s[n * 18 + 13] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 4]);
399 PRIM_INT__s_d_p_s[n * 18 + 13] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 4], PRIM_INT__s_d_p_s[n * 18 + 13]);
400 PRIM_INT__s_d_p_s[n * 18 + 13] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_d_p_s[n * 18 + 13]);
401
402 PRIM_INT__s_d_p_s[n * 18 + 14] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 4]);
403 PRIM_INT__s_d_p_s[n * 18 + 14] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 4], PRIM_INT__s_d_p_s[n * 18 + 14]);
404 PRIM_INT__s_d_p_s[n * 18 + 14] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_d_p_s[n * 18 + 14]);
405
406 PRIM_INT__s_d_p_s[n * 18 + 15] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
407 PRIM_INT__s_d_p_s[n * 18 + 15] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_d_p_s[n * 18 + 15]);
408
409 PRIM_INT__s_d_p_s[n * 18 + 16] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
410 PRIM_INT__s_d_p_s[n * 18 + 16] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_d_p_s[n * 18 + 16]);
411
412 PRIM_INT__s_d_p_s[n * 18 + 17] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
413 PRIM_INT__s_d_p_s[n * 18 + 17] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_d_s_s[(n+1) * 6 + 5], PRIM_INT__s_d_p_s[n * 18 + 17]);
414 PRIM_INT__s_d_p_s[n * 18 + 17] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_d_p_s[n * 18 + 17]);
415
416 }
417
418
419 VRR_K_s_f_d_s(
420 PRIM_INT__s_f_d_s,
421 PRIM_INT__s_f_p_s,
422 PRIM_INT__s_f_s_s,
423 PRIM_INT__s_d_p_s,
424 Q_PA,
425 a_over_q,
426 aoq_PQ,
427 one_over_2pq,
428 one_over_2q,
429 1);
430
431
432 VRR_J_s_g_s_s(
433 PRIM_INT__s_g_s_s,
434 PRIM_INT__s_f_s_s,
435 PRIM_INT__s_d_s_s,
436 P_PB,
437 a_over_p,
438 aop_PQ,
439 one_over_2p,
440 4);
441
442
443 VRR_K_s_g_p_s(
444 PRIM_INT__s_g_p_s,
445 PRIM_INT__s_g_s_s,
446 PRIM_INT__s_f_s_s,
447 Q_PA,
448 aoq_PQ,
449 one_over_2pq,
450 2);
451
452
453 ostei_general_vrr_K(0, 4, 2, 0, 1,
454 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
455 PRIM_INT__s_g_p_s, PRIM_INT__s_g_s_s, NULL, NULL, PRIM_INT__s_f_p_s, PRIM_INT__s_g_d_s);
456
457
458 VRR_J_s_h_s_s(
459 PRIM_INT__s_h_s_s,
460 PRIM_INT__s_g_s_s,
461 PRIM_INT__s_f_s_s,
462 P_PB,
463 a_over_p,
464 aop_PQ,
465 one_over_2p,
466 3);
467
468
469 ostei_general_vrr_K(0, 5, 1, 0, 2,
470 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
471 PRIM_INT__s_h_s_s, NULL, NULL, NULL, PRIM_INT__s_g_s_s, PRIM_INT__s_h_p_s);
472
473
474 ostei_general_vrr_K(0, 5, 2, 0, 1,
475 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
476 PRIM_INT__s_h_p_s, PRIM_INT__s_h_s_s, NULL, NULL, PRIM_INT__s_g_p_s, PRIM_INT__s_h_d_s);
477
478
479
480
481 ////////////////////////////////////
482 // Accumulate contracted integrals
483 ////////////////////////////////////
484 if(lastoffset == 0)
485 {
486 contract_all(30, PRIM_INT__s_f_p_s, PRIM_PTR_INT__s_f_p_s);
487 contract_all(60, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
488 contract_all(45, PRIM_INT__s_g_p_s, PRIM_PTR_INT__s_g_p_s);
489 contract_all(90, PRIM_INT__s_g_d_s, PRIM_PTR_INT__s_g_d_s);
490 contract_all(63, PRIM_INT__s_h_p_s, PRIM_PTR_INT__s_h_p_s);
491 contract_all(126, PRIM_INT__s_h_d_s, PRIM_PTR_INT__s_h_d_s);
492 }
493 else
494 {
495 contract(30, shelloffsets, PRIM_INT__s_f_p_s, PRIM_PTR_INT__s_f_p_s);
496 contract(60, shelloffsets, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
497 contract(45, shelloffsets, PRIM_INT__s_g_p_s, PRIM_PTR_INT__s_g_p_s);
498 contract(90, shelloffsets, PRIM_INT__s_g_d_s, PRIM_PTR_INT__s_g_d_s);
499 contract(63, shelloffsets, PRIM_INT__s_h_p_s, PRIM_PTR_INT__s_h_p_s);
500 contract(126, shelloffsets, PRIM_INT__s_h_d_s, PRIM_PTR_INT__s_h_d_s);
501 PRIM_PTR_INT__s_f_p_s += lastoffset*30;
502 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
503 PRIM_PTR_INT__s_g_p_s += lastoffset*45;
504 PRIM_PTR_INT__s_g_d_s += lastoffset*90;
505 PRIM_PTR_INT__s_h_p_s += lastoffset*63;
506 PRIM_PTR_INT__s_h_d_s += lastoffset*126;
507 }
508
509 } // close loop over j
510 } // close loop over i
511
512 //Advance to the next batch
513 jstart = SIMINT_SIMD_ROUND(jend);
514
515 //////////////////////////////////////////////
516 // Contracted integrals: Horizontal recurrance
517 //////////////////////////////////////////////
518
519
520 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
521
522
523 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
524 {
525 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
526
527 // set up HRR pointers
528 double const * restrict HRR_INT__s_f_p_s = INT__s_f_p_s + abcd * 30;
529 double const * restrict HRR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
530 double const * restrict HRR_INT__s_g_p_s = INT__s_g_p_s + abcd * 45;
531 double const * restrict HRR_INT__s_g_d_s = INT__s_g_d_s + abcd * 90;
532 double const * restrict HRR_INT__s_h_p_s = INT__s_h_p_s + abcd * 63;
533 double const * restrict HRR_INT__s_h_d_s = INT__s_h_d_s + abcd * 126;
534 double * restrict HRR_INT__d_f_p_p = INT__d_f_p_p + real_abcd * 540;
535
536 // form INT__p_f_p_s
537 HRR_I_p_f(
538 HRR_INT__p_f_p_s,
539 HRR_INT__s_f_p_s,
540 HRR_INT__s_g_p_s,
541 hAB, 3);
542
543 // form INT__p_f_d_s
544 HRR_I_p_f(
545 HRR_INT__p_f_d_s,
546 HRR_INT__s_f_d_s,
547 HRR_INT__s_g_d_s,
548 hAB, 6);
549
550 // form INT__p_g_p_s
551 HRR_I_p_g(
552 HRR_INT__p_g_p_s,
553 HRR_INT__s_g_p_s,
554 HRR_INT__s_h_p_s,
555 hAB, 3);
556
557 // form INT__p_g_d_s
558 HRR_I_p_g(
559 HRR_INT__p_g_d_s,
560 HRR_INT__s_g_d_s,
561 HRR_INT__s_h_d_s,
562 hAB, 6);
563
564 // form INT__d_f_p_s
565 HRR_I_d_f(
566 HRR_INT__d_f_p_s,
567 HRR_INT__p_f_p_s,
568 HRR_INT__p_g_p_s,
569 hAB, 3);
570
571 // form INT__d_f_d_s
572 HRR_I_d_f(
573 HRR_INT__d_f_d_s,
574 HRR_INT__p_f_d_s,
575 HRR_INT__p_g_d_s,
576 hAB, 6);
577
578 // form INT__d_f_p_p
579 for(ibra = 0; ibra < 60; ++ibra)
580 {
581 HRR_INT__d_f_p_p[ibra * 9 + 0] = HRR_INT__d_f_d_s[ibra * 6 + 0] + ( hCD[0] * HRR_INT__d_f_p_s[ibra * 3 + 0] );
582
583 HRR_INT__d_f_p_p[ibra * 9 + 1] = HRR_INT__d_f_d_s[ibra * 6 + 1] + ( hCD[1] * HRR_INT__d_f_p_s[ibra * 3 + 0] );
584
585 HRR_INT__d_f_p_p[ibra * 9 + 2] = HRR_INT__d_f_d_s[ibra * 6 + 2] + ( hCD[2] * HRR_INT__d_f_p_s[ibra * 3 + 0] );
586
587 HRR_INT__d_f_p_p[ibra * 9 + 3] = HRR_INT__d_f_d_s[ibra * 6 + 1] + ( hCD[0] * HRR_INT__d_f_p_s[ibra * 3 + 1] );
588
589 HRR_INT__d_f_p_p[ibra * 9 + 4] = HRR_INT__d_f_d_s[ibra * 6 + 3] + ( hCD[1] * HRR_INT__d_f_p_s[ibra * 3 + 1] );
590
591 HRR_INT__d_f_p_p[ibra * 9 + 5] = HRR_INT__d_f_d_s[ibra * 6 + 4] + ( hCD[2] * HRR_INT__d_f_p_s[ibra * 3 + 1] );
592
593 HRR_INT__d_f_p_p[ibra * 9 + 6] = HRR_INT__d_f_d_s[ibra * 6 + 2] + ( hCD[0] * HRR_INT__d_f_p_s[ibra * 3 + 2] );
594
595 HRR_INT__d_f_p_p[ibra * 9 + 7] = HRR_INT__d_f_d_s[ibra * 6 + 4] + ( hCD[1] * HRR_INT__d_f_p_s[ibra * 3 + 2] );
596
597 HRR_INT__d_f_p_p[ibra * 9 + 8] = HRR_INT__d_f_d_s[ibra * 6 + 5] + ( hCD[2] * HRR_INT__d_f_p_s[ibra * 3 + 2] );
598
599 }
600
601
602 } // close HRR loop
603
604
605 } // close loop cdbatch
606
607 istart = iend;
608 } // close loop over ab
609
610 return P.nshell12_clip * Q.nshell12_clip;
611 }
612
613