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