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_f_p_s_p(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__f_p_s_p)8 int ostei_f_p_s_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__f_p_s_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__f_p_s_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
27 // partition workspace
28 double * const INT__f_s_s_p = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__g_s_s_p = work + (SIMINT_NSHELL_SIMD * 30);
30 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*75);
31 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
32 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 6;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 21;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 45;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_p = primwork + 75;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 105;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_p = primwork + 135;
38 double * const hrrwork = (double *)(primwork + 180);
39
40
41 // Create constants
42 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
43 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
44 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
45 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
46 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
47
48
49 ////////////////////////////////////////
50 // Loop over shells and primitives
51 ////////////////////////////////////////
52
53 real_abcd = 0;
54 istart = 0;
55 for(ab = 0; ab < P.nshell12_clip; ++ab)
56 {
57 const int iend = istart + P.nprim12[ab];
58
59 cd = 0;
60 jstart = 0;
61
62 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
63 {
64 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
65 int jend = jstart;
66 for(i = 0; i < nshellbatch; i++)
67 jend += Q.nprim12[cd+i];
68
69 // Clear the beginning of the workspace (where we are accumulating integrals)
70 memset(work, 0, SIMINT_NSHELL_SIMD * 75 * sizeof(double));
71 abcd = 0;
72
73
74 for(i = istart; i < iend; ++i)
75 {
76 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
77
78 if(check_screen)
79 {
80 // Skip this whole thing if always insignificant
81 if((P.screen[i] * Q.screen_max) < screen_tol)
82 continue;
83 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
84 }
85
86 icd = 0;
87 iprimcd = 0;
88 nprim_icd = Q.nprim12[cd];
89 double * restrict PRIM_PTR_INT__f_s_s_p = INT__f_s_s_p + abcd * 30;
90 double * restrict PRIM_PTR_INT__g_s_s_p = INT__g_s_s_p + abcd * 45;
91
92
93
94 // Load these one per loop over i
95 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
96 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
97 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
98
99 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
100
101 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
102 {
103 // calculate the shell offsets
104 // these are the offset from the shell pointed to by cd
105 // for each element
106 int shelloffsets[SIMINT_SIMD_LEN] = {0};
107 int lastoffset = 0;
108 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
109
110 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
111 {
112 // Handle if the first element of the vector is a new shell
113 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
114 {
115 nprim_icd += Q.nprim12[cd + (++icd)];
116 PRIM_PTR_INT__f_s_s_p += 30;
117 PRIM_PTR_INT__g_s_s_p += 45;
118 }
119 iprimcd++;
120 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
121 {
122 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
123 {
124 shelloffsets[n] = shelloffsets[n-1] + 1;
125 lastoffset++;
126 nprim_icd += Q.nprim12[cd + (++icd)];
127 }
128 else
129 shelloffsets[n] = shelloffsets[n-1];
130 iprimcd++;
131 }
132 }
133 else
134 iprimcd += SIMINT_SIMD_LEN;
135
136 // Do we have to compute this vector (or has it been screened out)?
137 // (not_screened != 0 means we have to do this vector)
138 if(check_screen)
139 {
140 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
141 if(vmax < screen_tol)
142 {
143 PRIM_PTR_INT__f_s_s_p += lastoffset*30;
144 PRIM_PTR_INT__g_s_s_p += lastoffset*45;
145 continue;
146 }
147 }
148
149 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
150 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
151 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
152 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
153
154
155 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
156 SIMINT_DBLTYPE PQ[3];
157 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
158 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
159 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
160 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
161 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
162 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
163
164 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
165 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
166 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
167 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
168 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
169 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
170 const SIMINT_DBLTYPE Q_PB[3] = { SIMINT_DBLLOAD(Q.PB_x, j), SIMINT_DBLLOAD(Q.PB_y, j), SIMINT_DBLLOAD(Q.PB_z, j) };
171
172 // NOTE: Minus sign!
173 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
174 SIMINT_DBLTYPE aop_PQ[3];
175 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
176 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
177 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
178
179 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
180 SIMINT_DBLTYPE aoq_PQ[3];
181 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
182 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
183 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
184 // Put a minus sign here so we don't have to in RR routines
185 a_over_q = SIMINT_NEG(a_over_q);
186
187
188 //////////////////////////////////////////////
189 // Fjt function section
190 // Maximum v value: 5
191 //////////////////////////////////////////////
192 // The parameter to the Fjt function
193 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
194
195
196 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
197
198
199 boys_F_split(PRIM_INT__s_s_s_s, F_x, 5);
200 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
201 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
202 for(n = 0; n <= 5; n++)
203 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
204
205 //////////////////////////////////////////////
206 // Primitive integrals: Vertical recurrance
207 //////////////////////////////////////////////
208
209 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
210 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
211 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
212 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
213 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
214 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
215 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
216
217
218
219 // Forming PRIM_INT__p_s_s_s[5 * 3];
220 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
221 {
222
223 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
224 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]);
225
226 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
227 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 1]);
228
229 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
230 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 2]);
231
232 }
233
234
235
236 // Forming PRIM_INT__d_s_s_s[4 * 6];
237 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
238 {
239
240 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
241 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 0]);
242 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 0]);
243
244 PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 0]);
245 PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 1]);
246
247 PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 0]);
248 PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 2]);
249
250 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
251 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 3]);
252 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 3]);
253
254 PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 1]);
255 PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 4]);
256
257 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
258 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__d_s_s_s[n * 6 + 5]);
259 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 5]);
260
261 }
262
263
264
265 // Forming PRIM_INT__f_s_s_s[3 * 10];
266 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
267 {
268
269 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
270 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 0]);
271 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]), PRIM_INT__f_s_s_s[n * 10 + 0]);
272
273 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
274 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 1]);
275
276 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]);
277 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 2]);
278
279 PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 3]);
280 PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 3]);
281
282 PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 1]);
283 PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 1], PRIM_INT__f_s_s_s[n * 10 + 4]);
284
285 PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 5]);
286 PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 5]);
287
288 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
289 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 6]);
290 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__p_s_s_s[n * 3 + 1]), PRIM_INT__f_s_s_s[n * 10 + 6]);
291
292 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]);
293 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 7]);
294
295 PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 5]);
296 PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 8]);
297
298 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
299 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 9]);
300 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__p_s_s_s[n * 3 + 2]), PRIM_INT__f_s_s_s[n * 10 + 9]);
301
302 }
303
304
305 VRR_L_f_s_s_p(
306 PRIM_INT__f_s_s_p,
307 PRIM_INT__f_s_s_s,
308 PRIM_INT__d_s_s_s,
309 Q_PB,
310 aoq_PQ,
311 one_over_2pq,
312 1);
313
314
315 VRR_I_g_s_s_s(
316 PRIM_INT__g_s_s_s,
317 PRIM_INT__f_s_s_s,
318 PRIM_INT__d_s_s_s,
319 P_PA,
320 a_over_p,
321 aop_PQ,
322 one_over_2p,
323 2);
324
325
326 VRR_L_g_s_s_p(
327 PRIM_INT__g_s_s_p,
328 PRIM_INT__g_s_s_s,
329 PRIM_INT__f_s_s_s,
330 Q_PB,
331 aoq_PQ,
332 one_over_2pq,
333 1);
334
335
336
337
338 ////////////////////////////////////
339 // Accumulate contracted integrals
340 ////////////////////////////////////
341 if(lastoffset == 0)
342 {
343 contract_all(30, PRIM_INT__f_s_s_p, PRIM_PTR_INT__f_s_s_p);
344 contract_all(45, PRIM_INT__g_s_s_p, PRIM_PTR_INT__g_s_s_p);
345 }
346 else
347 {
348 contract(30, shelloffsets, PRIM_INT__f_s_s_p, PRIM_PTR_INT__f_s_s_p);
349 contract(45, shelloffsets, PRIM_INT__g_s_s_p, PRIM_PTR_INT__g_s_s_p);
350 PRIM_PTR_INT__f_s_s_p += lastoffset*30;
351 PRIM_PTR_INT__g_s_s_p += lastoffset*45;
352 }
353
354 } // close loop over j
355 } // close loop over i
356
357 //Advance to the next batch
358 jstart = SIMINT_SIMD_ROUND(jend);
359
360 //////////////////////////////////////////////
361 // Contracted integrals: Horizontal recurrance
362 //////////////////////////////////////////////
363
364
365 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
366
367
368 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
369 {
370
371 // set up HRR pointers
372 double const * restrict HRR_INT__f_s_s_p = INT__f_s_s_p + abcd * 30;
373 double const * restrict HRR_INT__g_s_s_p = INT__g_s_s_p + abcd * 45;
374 double * restrict HRR_INT__f_p_s_p = INT__f_p_s_p + real_abcd * 90;
375
376 // form INT__f_p_s_p
377 HRR_J_f_p(
378 HRR_INT__f_p_s_p,
379 HRR_INT__f_s_s_p,
380 HRR_INT__g_s_s_p,
381 hAB, 3);
382
383
384 } // close HRR loop
385
386
387 } // close loop cdbatch
388
389 istart = iend;
390 } // close loop over ab
391
392 return P.nshell12_clip * Q.nshell12_clip;
393 }
394
395