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_s_d_p(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__p_s_d_p)8 int ostei_p_s_d_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__p_s_d_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__p_s_d_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 ibra;
26
27 // partition workspace
28 double * const INT__p_s_d_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__p_s_f_s = work + (SIMINT_NSHELL_SIMD * 18);
30 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*48);
31 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
32 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_p_s = primwork + 5;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_d_s = primwork + 17;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_f_s = primwork + 35;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_d_s = primwork + 55;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_f_s = primwork + 73;
37 double * const hrrwork = (double *)(primwork + 103);
38
39
40 // Create constants
41 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
42 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
43 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
44 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
45
46
47 ////////////////////////////////////////
48 // Loop over shells and primitives
49 ////////////////////////////////////////
50
51 real_abcd = 0;
52 istart = 0;
53 for(ab = 0; ab < P.nshell12_clip; ++ab)
54 {
55 const int iend = istart + P.nprim12[ab];
56
57 cd = 0;
58 jstart = 0;
59
60 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
61 {
62 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
63 int jend = jstart;
64 for(i = 0; i < nshellbatch; i++)
65 jend += Q.nprim12[cd+i];
66
67 // Clear the beginning of the workspace (where we are accumulating integrals)
68 memset(work, 0, SIMINT_NSHELL_SIMD * 48 * sizeof(double));
69 abcd = 0;
70
71
72 for(i = istart; i < iend; ++i)
73 {
74 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
75
76 if(check_screen)
77 {
78 // Skip this whole thing if always insignificant
79 if((P.screen[i] * Q.screen_max) < screen_tol)
80 continue;
81 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
82 }
83
84 icd = 0;
85 iprimcd = 0;
86 nprim_icd = Q.nprim12[cd];
87 double * restrict PRIM_PTR_INT__p_s_d_s = INT__p_s_d_s + abcd * 18;
88 double * restrict PRIM_PTR_INT__p_s_f_s = INT__p_s_f_s + abcd * 30;
89
90
91
92 // Load these one per loop over i
93 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
94 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
95 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
96
97 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
98
99 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
100 {
101 // calculate the shell offsets
102 // these are the offset from the shell pointed to by cd
103 // for each element
104 int shelloffsets[SIMINT_SIMD_LEN] = {0};
105 int lastoffset = 0;
106 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
107
108 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
109 {
110 // Handle if the first element of the vector is a new shell
111 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
112 {
113 nprim_icd += Q.nprim12[cd + (++icd)];
114 PRIM_PTR_INT__p_s_d_s += 18;
115 PRIM_PTR_INT__p_s_f_s += 30;
116 }
117 iprimcd++;
118 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
119 {
120 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
121 {
122 shelloffsets[n] = shelloffsets[n-1] + 1;
123 lastoffset++;
124 nprim_icd += Q.nprim12[cd + (++icd)];
125 }
126 else
127 shelloffsets[n] = shelloffsets[n-1];
128 iprimcd++;
129 }
130 }
131 else
132 iprimcd += SIMINT_SIMD_LEN;
133
134 // Do we have to compute this vector (or has it been screened out)?
135 // (not_screened != 0 means we have to do this vector)
136 if(check_screen)
137 {
138 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
139 if(vmax < screen_tol)
140 {
141 PRIM_PTR_INT__p_s_d_s += lastoffset*18;
142 PRIM_PTR_INT__p_s_f_s += lastoffset*30;
143 continue;
144 }
145 }
146
147 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
148 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
149 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
150 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
151
152
153 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
154 SIMINT_DBLTYPE PQ[3];
155 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
156 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
157 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
158 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
159 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
160 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
161
162 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
163 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
164 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
165 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
166 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
167 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
168 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
169
170 // NOTE: Minus sign!
171 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
172 SIMINT_DBLTYPE aop_PQ[3];
173 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
174 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
175 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
176
177 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
178 SIMINT_DBLTYPE aoq_PQ[3];
179 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
180 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
181 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
182 // Put a minus sign here so we don't have to in RR routines
183 a_over_q = SIMINT_NEG(a_over_q);
184
185
186 //////////////////////////////////////////////
187 // Fjt function section
188 // Maximum v value: 4
189 //////////////////////////////////////////////
190 // The parameter to the Fjt function
191 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
192
193
194 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
195
196
197 boys_F_split(PRIM_INT__s_s_s_s, F_x, 4);
198 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
199 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
200 for(n = 0; n <= 4; n++)
201 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
202
203 //////////////////////////////////////////////
204 // Primitive integrals: Vertical recurrance
205 //////////////////////////////////////////////
206
207 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
208 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
209 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
210 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
211 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
212
213
214
215 // Forming PRIM_INT__s_s_p_s[4 * 3];
216 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
217 {
218
219 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
220 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]);
221
222 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
223 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 1]);
224
225 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
226 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 2]);
227
228 }
229
230
231
232 // Forming PRIM_INT__s_s_d_s[3 * 6];
233 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
234 {
235
236 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_p_s[n * 3 + 0]);
237 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 0]);
238 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 0]);
239
240 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 0]);
241 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 1]);
242
243 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 0]);
244 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 2]);
245
246 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 1]);
247 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 3]);
248 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 3]);
249
250 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 1]);
251 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 4]);
252
253 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 2]);
254 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_d_s[n * 6 + 5]);
255 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 5]);
256
257 }
258
259
260
261 // Forming PRIM_INT__p_s_d_s[1 * 18];
262 for(n = 0; n < 1; ++n) // loop over orders of auxiliary function
263 {
264
265 PRIM_INT__p_s_d_s[n * 18 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
266 PRIM_INT__p_s_d_s[n * 18 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__p_s_d_s[n * 18 + 0]);
267 PRIM_INT__p_s_d_s[n * 18 + 0] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_d_s[n * 18 + 0]);
268
269 PRIM_INT__p_s_d_s[n * 18 + 1] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 1]);
270 PRIM_INT__p_s_d_s[n * 18 + 1] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__p_s_d_s[n * 18 + 1]);
271 PRIM_INT__p_s_d_s[n * 18 + 1] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_d_s[n * 18 + 1]);
272
273 PRIM_INT__p_s_d_s[n * 18 + 2] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 2]);
274 PRIM_INT__p_s_d_s[n * 18 + 2] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__p_s_d_s[n * 18 + 2]);
275 PRIM_INT__p_s_d_s[n * 18 + 2] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_d_s[n * 18 + 2]);
276
277 PRIM_INT__p_s_d_s[n * 18 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
278 PRIM_INT__p_s_d_s[n * 18 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__p_s_d_s[n * 18 + 3]);
279
280 PRIM_INT__p_s_d_s[n * 18 + 4] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 4]);
281 PRIM_INT__p_s_d_s[n * 18 + 4] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__p_s_d_s[n * 18 + 4]);
282
283 PRIM_INT__p_s_d_s[n * 18 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
284 PRIM_INT__p_s_d_s[n * 18 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__p_s_d_s[n * 18 + 5]);
285
286 PRIM_INT__p_s_d_s[n * 18 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
287 PRIM_INT__p_s_d_s[n * 18 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__p_s_d_s[n * 18 + 6]);
288
289 PRIM_INT__p_s_d_s[n * 18 + 7] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 1]);
290 PRIM_INT__p_s_d_s[n * 18 + 7] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__p_s_d_s[n * 18 + 7]);
291 PRIM_INT__p_s_d_s[n * 18 + 7] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_d_s[n * 18 + 7]);
292
293 PRIM_INT__p_s_d_s[n * 18 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 2]);
294 PRIM_INT__p_s_d_s[n * 18 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__p_s_d_s[n * 18 + 8]);
295
296 PRIM_INT__p_s_d_s[n * 18 + 9] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
297 PRIM_INT__p_s_d_s[n * 18 + 9] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__p_s_d_s[n * 18 + 9]);
298 PRIM_INT__p_s_d_s[n * 18 + 9] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_d_s[n * 18 + 9]);
299
300 PRIM_INT__p_s_d_s[n * 18 + 10] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 4]);
301 PRIM_INT__p_s_d_s[n * 18 + 10] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__p_s_d_s[n * 18 + 10]);
302 PRIM_INT__p_s_d_s[n * 18 + 10] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_d_s[n * 18 + 10]);
303
304 PRIM_INT__p_s_d_s[n * 18 + 11] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
305 PRIM_INT__p_s_d_s[n * 18 + 11] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__p_s_d_s[n * 18 + 11]);
306
307 PRIM_INT__p_s_d_s[n * 18 + 12] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
308 PRIM_INT__p_s_d_s[n * 18 + 12] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__p_s_d_s[n * 18 + 12]);
309
310 PRIM_INT__p_s_d_s[n * 18 + 13] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
311 PRIM_INT__p_s_d_s[n * 18 + 13] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__p_s_d_s[n * 18 + 13]);
312
313 PRIM_INT__p_s_d_s[n * 18 + 14] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 2]);
314 PRIM_INT__p_s_d_s[n * 18 + 14] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__p_s_d_s[n * 18 + 14]);
315 PRIM_INT__p_s_d_s[n * 18 + 14] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_d_s[n * 18 + 14]);
316
317 PRIM_INT__p_s_d_s[n * 18 + 15] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
318 PRIM_INT__p_s_d_s[n * 18 + 15] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__p_s_d_s[n * 18 + 15]);
319
320 PRIM_INT__p_s_d_s[n * 18 + 16] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 4]);
321 PRIM_INT__p_s_d_s[n * 18 + 16] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__p_s_d_s[n * 18 + 16]);
322 PRIM_INT__p_s_d_s[n * 18 + 16] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_d_s[n * 18 + 16]);
323
324 PRIM_INT__p_s_d_s[n * 18 + 17] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
325 PRIM_INT__p_s_d_s[n * 18 + 17] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__p_s_d_s[n * 18 + 17]);
326 PRIM_INT__p_s_d_s[n * 18 + 17] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_d_s[n * 18 + 17]);
327
328 }
329
330
331
332 // Forming PRIM_INT__s_s_f_s[2 * 10];
333 for(n = 0; n < 2; ++n) // loop over orders of auxiliary function
334 {
335
336 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
337 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 0]);
338 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]), PRIM_INT__s_s_f_s[n * 10 + 0]);
339
340 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
341 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 1]);
342
343 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
344 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 2]);
345
346 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
347 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 3]);
348
349 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
350 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_s_f_s[n * 10 + 4]);
351
352 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
353 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 5]);
354
355 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
356 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 6]);
357 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_p_s[n * 3 + 1]), PRIM_INT__s_s_f_s[n * 10 + 6]);
358
359 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
360 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 7]);
361
362 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
363 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 8]);
364
365 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
366 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 9]);
367 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_p_s[n * 3 + 2]), PRIM_INT__s_s_f_s[n * 10 + 9]);
368
369 }
370
371
372 VRR_I_p_s_f_s(
373 PRIM_INT__p_s_f_s,
374 PRIM_INT__s_s_f_s,
375 PRIM_INT__s_s_d_s,
376 P_PA,
377 aop_PQ,
378 one_over_2pq,
379 1);
380
381
382
383
384 ////////////////////////////////////
385 // Accumulate contracted integrals
386 ////////////////////////////////////
387 if(lastoffset == 0)
388 {
389 contract_all(18, PRIM_INT__p_s_d_s, PRIM_PTR_INT__p_s_d_s);
390 contract_all(30, PRIM_INT__p_s_f_s, PRIM_PTR_INT__p_s_f_s);
391 }
392 else
393 {
394 contract(18, shelloffsets, PRIM_INT__p_s_d_s, PRIM_PTR_INT__p_s_d_s);
395 contract(30, shelloffsets, PRIM_INT__p_s_f_s, PRIM_PTR_INT__p_s_f_s);
396 PRIM_PTR_INT__p_s_d_s += lastoffset*18;
397 PRIM_PTR_INT__p_s_f_s += lastoffset*30;
398 }
399
400 } // close loop over j
401 } // close loop over i
402
403 //Advance to the next batch
404 jstart = SIMINT_SIMD_ROUND(jend);
405
406 //////////////////////////////////////////////
407 // Contracted integrals: Horizontal recurrance
408 //////////////////////////////////////////////
409
410
411
412
413 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
414 {
415 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
416
417 // set up HRR pointers
418 double const * restrict HRR_INT__p_s_d_s = INT__p_s_d_s + abcd * 18;
419 double const * restrict HRR_INT__p_s_f_s = INT__p_s_f_s + abcd * 30;
420 double * restrict HRR_INT__p_s_d_p = INT__p_s_d_p + real_abcd * 54;
421
422 // form INT__p_s_d_p
423 for(ibra = 0; ibra < 3; ++ibra)
424 {
425 HRR_INT__p_s_d_p[ibra * 18 + 0] = HRR_INT__p_s_f_s[ibra * 10 + 0] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 0] );
426
427 HRR_INT__p_s_d_p[ibra * 18 + 1] = HRR_INT__p_s_f_s[ibra * 10 + 1] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 0] );
428
429 HRR_INT__p_s_d_p[ibra * 18 + 2] = HRR_INT__p_s_f_s[ibra * 10 + 2] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 0] );
430
431 HRR_INT__p_s_d_p[ibra * 18 + 3] = HRR_INT__p_s_f_s[ibra * 10 + 1] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 1] );
432
433 HRR_INT__p_s_d_p[ibra * 18 + 4] = HRR_INT__p_s_f_s[ibra * 10 + 3] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 1] );
434
435 HRR_INT__p_s_d_p[ibra * 18 + 5] = HRR_INT__p_s_f_s[ibra * 10 + 4] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 1] );
436
437 HRR_INT__p_s_d_p[ibra * 18 + 6] = HRR_INT__p_s_f_s[ibra * 10 + 2] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 2] );
438
439 HRR_INT__p_s_d_p[ibra * 18 + 7] = HRR_INT__p_s_f_s[ibra * 10 + 4] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 2] );
440
441 HRR_INT__p_s_d_p[ibra * 18 + 8] = HRR_INT__p_s_f_s[ibra * 10 + 5] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 2] );
442
443 HRR_INT__p_s_d_p[ibra * 18 + 9] = HRR_INT__p_s_f_s[ibra * 10 + 3] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 3] );
444
445 HRR_INT__p_s_d_p[ibra * 18 + 10] = HRR_INT__p_s_f_s[ibra * 10 + 6] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 3] );
446
447 HRR_INT__p_s_d_p[ibra * 18 + 11] = HRR_INT__p_s_f_s[ibra * 10 + 7] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 3] );
448
449 HRR_INT__p_s_d_p[ibra * 18 + 12] = HRR_INT__p_s_f_s[ibra * 10 + 4] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 4] );
450
451 HRR_INT__p_s_d_p[ibra * 18 + 13] = HRR_INT__p_s_f_s[ibra * 10 + 7] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 4] );
452
453 HRR_INT__p_s_d_p[ibra * 18 + 14] = HRR_INT__p_s_f_s[ibra * 10 + 8] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 4] );
454
455 HRR_INT__p_s_d_p[ibra * 18 + 15] = HRR_INT__p_s_f_s[ibra * 10 + 5] + ( hCD[0] * HRR_INT__p_s_d_s[ibra * 6 + 5] );
456
457 HRR_INT__p_s_d_p[ibra * 18 + 16] = HRR_INT__p_s_f_s[ibra * 10 + 8] + ( hCD[1] * HRR_INT__p_s_d_s[ibra * 6 + 5] );
458
459 HRR_INT__p_s_d_p[ibra * 18 + 17] = HRR_INT__p_s_f_s[ibra * 10 + 9] + ( hCD[2] * HRR_INT__p_s_d_s[ibra * 6 + 5] );
460
461 }
462
463
464 } // close HRR loop
465
466
467 } // close loop cdbatch
468
469 istart = iend;
470 } // close loop over ab
471
472 return P.nshell12_clip * Q.nshell12_clip;
473 }
474
475