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