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_s_f_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_s_f_p)8 int ostei_f_s_f_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_s_f_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__f_s_f_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__f_s_f_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__f_s_g_s = work + (SIMINT_NSHELL_SIMD * 100);
30 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*250);
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 + 8;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_d_s = primwork + 29;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_f_s = primwork + 65;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_g_s = primwork + 115;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_p_s = primwork + 175;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_d_s = primwork + 202;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_f_s = primwork + 256;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_g_s = primwork + 346;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_d_s = primwork + 481;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_f_s = primwork + 553;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_g_s = primwork + 673;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_f_s = primwork + 853;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_g_s = primwork + 953;
45 double * const hrrwork = (double *)(primwork + 1103);
46
47
48 // Create constants
49 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
50 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
51 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
52 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
53 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
54
55
56 ////////////////////////////////////////
57 // Loop over shells and primitives
58 ////////////////////////////////////////
59
60 real_abcd = 0;
61 istart = 0;
62 for(ab = 0; ab < P.nshell12_clip; ++ab)
63 {
64 const int iend = istart + P.nprim12[ab];
65
66 cd = 0;
67 jstart = 0;
68
69 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
70 {
71 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
72 int jend = jstart;
73 for(i = 0; i < nshellbatch; i++)
74 jend += Q.nprim12[cd+i];
75
76 // Clear the beginning of the workspace (where we are accumulating integrals)
77 memset(work, 0, SIMINT_NSHELL_SIMD * 250 * sizeof(double));
78 abcd = 0;
79
80
81 for(i = istart; i < iend; ++i)
82 {
83 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
84
85 if(check_screen)
86 {
87 // Skip this whole thing if always insignificant
88 if((P.screen[i] * Q.screen_max) < screen_tol)
89 continue;
90 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
91 }
92
93 icd = 0;
94 iprimcd = 0;
95 nprim_icd = Q.nprim12[cd];
96 double * restrict PRIM_PTR_INT__f_s_f_s = INT__f_s_f_s + abcd * 100;
97 double * restrict PRIM_PTR_INT__f_s_g_s = INT__f_s_g_s + abcd * 150;
98
99
100
101 // Load these one per loop over i
102 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
103 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
104 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
105
106 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
107
108 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
109 {
110 // calculate the shell offsets
111 // these are the offset from the shell pointed to by cd
112 // for each element
113 int shelloffsets[SIMINT_SIMD_LEN] = {0};
114 int lastoffset = 0;
115 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
116
117 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
118 {
119 // Handle if the first element of the vector is a new shell
120 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
121 {
122 nprim_icd += Q.nprim12[cd + (++icd)];
123 PRIM_PTR_INT__f_s_f_s += 100;
124 PRIM_PTR_INT__f_s_g_s += 150;
125 }
126 iprimcd++;
127 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
128 {
129 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
130 {
131 shelloffsets[n] = shelloffsets[n-1] + 1;
132 lastoffset++;
133 nprim_icd += Q.nprim12[cd + (++icd)];
134 }
135 else
136 shelloffsets[n] = shelloffsets[n-1];
137 iprimcd++;
138 }
139 }
140 else
141 iprimcd += SIMINT_SIMD_LEN;
142
143 // Do we have to compute this vector (or has it been screened out)?
144 // (not_screened != 0 means we have to do this vector)
145 if(check_screen)
146 {
147 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
148 if(vmax < screen_tol)
149 {
150 PRIM_PTR_INT__f_s_f_s += lastoffset*100;
151 PRIM_PTR_INT__f_s_g_s += lastoffset*150;
152 continue;
153 }
154 }
155
156 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
157 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
158 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
159 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
160
161
162 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
163 SIMINT_DBLTYPE PQ[3];
164 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
165 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
166 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
167 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
168 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
169 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
170
171 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
172 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
173 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
174 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
175 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
176 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
177 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
178
179 // NOTE: Minus sign!
180 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
181 SIMINT_DBLTYPE aop_PQ[3];
182 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
183 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
184 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
185
186 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
187 SIMINT_DBLTYPE aoq_PQ[3];
188 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
189 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
190 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
191 // Put a minus sign here so we don't have to in RR routines
192 a_over_q = SIMINT_NEG(a_over_q);
193
194
195 //////////////////////////////////////////////
196 // Fjt function section
197 // Maximum v value: 7
198 //////////////////////////////////////////////
199 // The parameter to the Fjt function
200 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
201
202
203 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
204
205
206 boys_F_split(PRIM_INT__s_s_s_s, F_x, 7);
207 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
208 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
209 for(n = 0; n <= 7; n++)
210 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
211
212 //////////////////////////////////////////////
213 // Primitive integrals: Vertical recurrance
214 //////////////////////////////////////////////
215
216 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
217 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
218 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
219 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
220 const SIMINT_DBLTYPE vrr_const_3_over_2q = SIMINT_MUL(const_3, one_over_2q);
221 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
222 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
223 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
224 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
225
226
227
228 // Forming PRIM_INT__s_s_p_s[7 * 3];
229 for(n = 0; n < 7; ++n) // loop over orders of auxiliary function
230 {
231
232 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
233 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]);
234
235 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
236 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]);
237
238 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
239 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]);
240
241 }
242
243
244
245 // Forming PRIM_INT__s_s_d_s[6 * 6];
246 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
247 {
248
249 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_p_s[n * 3 + 0]);
250 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]);
251 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]);
252
253 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 0]);
254 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]);
255
256 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 0]);
257 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]);
258
259 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 1]);
260 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]);
261 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]);
262
263 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 1]);
264 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]);
265
266 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 2]);
267 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]);
268 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]);
269
270 }
271
272
273
274 // Forming PRIM_INT__s_s_f_s[5 * 10];
275 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
276 {
277
278 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
279 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]);
280 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]);
281
282 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
283 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]);
284
285 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
286 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]);
287
288 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
289 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]);
290
291 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
292 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]);
293
294 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
295 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]);
296
297 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
298 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]);
299 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]);
300
301 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
302 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]);
303
304 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
305 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]);
306
307 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
308 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]);
309 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]);
310
311 }
312
313
314 VRR_I_p_s_f_s(
315 PRIM_INT__p_s_f_s,
316 PRIM_INT__s_s_f_s,
317 PRIM_INT__s_s_d_s,
318 P_PA,
319 aop_PQ,
320 one_over_2pq,
321 3);
322
323
324
325 // Forming PRIM_INT__p_s_d_s[3 * 18];
326 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
327 {
328
329 PRIM_INT__p_s_d_s[n * 18 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
330 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]);
331 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]);
332
333 PRIM_INT__p_s_d_s[n * 18 + 1] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 1]);
334 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]);
335 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]);
336
337 PRIM_INT__p_s_d_s[n * 18 + 2] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 2]);
338 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]);
339 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]);
340
341 PRIM_INT__p_s_d_s[n * 18 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
342 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]);
343
344 PRIM_INT__p_s_d_s[n * 18 + 4] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 4]);
345 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]);
346
347 PRIM_INT__p_s_d_s[n * 18 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
348 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]);
349
350 PRIM_INT__p_s_d_s[n * 18 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
351 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]);
352
353 PRIM_INT__p_s_d_s[n * 18 + 7] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 1]);
354 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]);
355 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]);
356
357 PRIM_INT__p_s_d_s[n * 18 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 2]);
358 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]);
359
360 PRIM_INT__p_s_d_s[n * 18 + 9] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
361 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]);
362 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]);
363
364 PRIM_INT__p_s_d_s[n * 18 + 10] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 4]);
365 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]);
366 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]);
367
368 PRIM_INT__p_s_d_s[n * 18 + 11] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
369 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]);
370
371 PRIM_INT__p_s_d_s[n * 18 + 12] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
372 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]);
373
374 PRIM_INT__p_s_d_s[n * 18 + 13] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
375 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]);
376
377 PRIM_INT__p_s_d_s[n * 18 + 14] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 2]);
378 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]);
379 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]);
380
381 PRIM_INT__p_s_d_s[n * 18 + 15] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
382 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]);
383
384 PRIM_INT__p_s_d_s[n * 18 + 16] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 4]);
385 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]);
386 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]);
387
388 PRIM_INT__p_s_d_s[n * 18 + 17] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
389 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]);
390 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]);
391
392 }
393
394
395 VRR_I_d_s_f_s(
396 PRIM_INT__d_s_f_s,
397 PRIM_INT__p_s_f_s,
398 PRIM_INT__s_s_f_s,
399 PRIM_INT__p_s_d_s,
400 P_PA,
401 a_over_p,
402 aop_PQ,
403 one_over_2p,
404 one_over_2pq,
405 2);
406
407
408
409 // Forming PRIM_INT__p_s_p_s[3 * 9];
410 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
411 {
412
413 PRIM_INT__p_s_p_s[n * 9 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_p_s[n * 3 + 0]);
414 PRIM_INT__p_s_p_s[n * 9 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_p_s[n * 9 + 0]);
415 PRIM_INT__p_s_p_s[n * 9 + 0] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_p_s[n * 9 + 0]);
416
417 PRIM_INT__p_s_p_s[n * 9 + 1] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_p_s[n * 3 + 1]);
418 PRIM_INT__p_s_p_s[n * 9 + 1] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_p_s[n * 9 + 1]);
419
420 PRIM_INT__p_s_p_s[n * 9 + 2] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_p_s[n * 3 + 2]);
421 PRIM_INT__p_s_p_s[n * 9 + 2] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_p_s[n * 9 + 2]);
422
423 PRIM_INT__p_s_p_s[n * 9 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_p_s[n * 3 + 0]);
424 PRIM_INT__p_s_p_s[n * 9 + 3] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_p_s[n * 9 + 3]);
425
426 PRIM_INT__p_s_p_s[n * 9 + 4] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_p_s[n * 3 + 1]);
427 PRIM_INT__p_s_p_s[n * 9 + 4] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_p_s[n * 9 + 4]);
428 PRIM_INT__p_s_p_s[n * 9 + 4] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_p_s[n * 9 + 4]);
429
430 PRIM_INT__p_s_p_s[n * 9 + 5] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_p_s[n * 3 + 2]);
431 PRIM_INT__p_s_p_s[n * 9 + 5] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_p_s[n * 9 + 5]);
432
433 PRIM_INT__p_s_p_s[n * 9 + 6] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_p_s[n * 3 + 0]);
434 PRIM_INT__p_s_p_s[n * 9 + 6] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__p_s_p_s[n * 9 + 6]);
435
436 PRIM_INT__p_s_p_s[n * 9 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_p_s[n * 3 + 1]);
437 PRIM_INT__p_s_p_s[n * 9 + 7] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__p_s_p_s[n * 9 + 7]);
438
439 PRIM_INT__p_s_p_s[n * 9 + 8] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_p_s[n * 3 + 2]);
440 PRIM_INT__p_s_p_s[n * 9 + 8] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__p_s_p_s[n * 9 + 8]);
441 PRIM_INT__p_s_p_s[n * 9 + 8] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_p_s[n * 9 + 8]);
442
443 }
444
445
446 VRR_I_d_s_d_s(
447 PRIM_INT__d_s_d_s,
448 PRIM_INT__p_s_d_s,
449 PRIM_INT__s_s_d_s,
450 PRIM_INT__p_s_p_s,
451 P_PA,
452 a_over_p,
453 aop_PQ,
454 one_over_2p,
455 one_over_2pq,
456 2);
457
458
459 ostei_general_vrr_I(3, 0, 3, 0, 1,
460 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
461 PRIM_INT__d_s_f_s, PRIM_INT__p_s_f_s, NULL, PRIM_INT__d_s_d_s, NULL, PRIM_INT__f_s_f_s);
462
463
464 VRR_K_s_s_g_s(
465 PRIM_INT__s_s_g_s,
466 PRIM_INT__s_s_f_s,
467 PRIM_INT__s_s_d_s,
468 Q_PA,
469 a_over_q,
470 aoq_PQ,
471 one_over_2q,
472 4);
473
474
475 VRR_I_p_s_g_s(
476 PRIM_INT__p_s_g_s,
477 PRIM_INT__s_s_g_s,
478 PRIM_INT__s_s_f_s,
479 P_PA,
480 aop_PQ,
481 one_over_2pq,
482 3);
483
484
485 ostei_general_vrr_I(2, 0, 4, 0, 2,
486 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
487 PRIM_INT__p_s_g_s, PRIM_INT__s_s_g_s, NULL, PRIM_INT__p_s_f_s, NULL, PRIM_INT__d_s_g_s);
488
489
490 ostei_general_vrr_I(3, 0, 4, 0, 1,
491 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
492 PRIM_INT__d_s_g_s, PRIM_INT__p_s_g_s, NULL, PRIM_INT__d_s_f_s, NULL, PRIM_INT__f_s_g_s);
493
494
495
496
497 ////////////////////////////////////
498 // Accumulate contracted integrals
499 ////////////////////////////////////
500 if(lastoffset == 0)
501 {
502 contract_all(100, PRIM_INT__f_s_f_s, PRIM_PTR_INT__f_s_f_s);
503 contract_all(150, PRIM_INT__f_s_g_s, PRIM_PTR_INT__f_s_g_s);
504 }
505 else
506 {
507 contract(100, shelloffsets, PRIM_INT__f_s_f_s, PRIM_PTR_INT__f_s_f_s);
508 contract(150, shelloffsets, PRIM_INT__f_s_g_s, PRIM_PTR_INT__f_s_g_s);
509 PRIM_PTR_INT__f_s_f_s += lastoffset*100;
510 PRIM_PTR_INT__f_s_g_s += lastoffset*150;
511 }
512
513 } // close loop over j
514 } // close loop over i
515
516 //Advance to the next batch
517 jstart = SIMINT_SIMD_ROUND(jend);
518
519 //////////////////////////////////////////////
520 // Contracted integrals: Horizontal recurrance
521 //////////////////////////////////////////////
522
523
524
525
526 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
527 {
528 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
529
530 // set up HRR pointers
531 double const * restrict HRR_INT__f_s_f_s = INT__f_s_f_s + abcd * 100;
532 double const * restrict HRR_INT__f_s_g_s = INT__f_s_g_s + abcd * 150;
533 double * restrict HRR_INT__f_s_f_p = INT__f_s_f_p + real_abcd * 300;
534
535 // form INT__f_s_f_p
536 HRR_L_f_p(
537 HRR_INT__f_s_f_p,
538 HRR_INT__f_s_f_s,
539 HRR_INT__f_s_g_s,
540 hCD, 10);
541
542
543 } // close HRR loop
544
545
546 } // close loop cdbatch
547
548 istart = iend;
549 } // close loop over ab
550
551 return P.nshell12_clip * Q.nshell12_clip;
552 }
553
554