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_s_f_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__s_f_d_p)8 int ostei_s_f_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__s_f_d_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__s_f_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__s_f_d_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__s_f_f_s = work + (SIMINT_NSHELL_SIMD * 60);
30 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*160);
31 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
32 SIMINT_DBLTYPE * const restrict PRIM_INT__s_p_s_s = primwork + 7;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__s_p_p_s = primwork + 25;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_s_s = primwork + 52;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_p_s = primwork + 82;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_d_s = primwork + 136;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_s_s = primwork + 208;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_p_s = primwork + 248;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_d_s = primwork + 338;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_f_s = primwork + 458;
41 double * const hrrwork = (double *)(primwork + 558);
42
43
44 // Create constants
45 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
46 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
47 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
48 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
49
50
51 ////////////////////////////////////////
52 // Loop over shells and primitives
53 ////////////////////////////////////////
54
55 real_abcd = 0;
56 istart = 0;
57 for(ab = 0; ab < P.nshell12_clip; ++ab)
58 {
59 const int iend = istart + P.nprim12[ab];
60
61 cd = 0;
62 jstart = 0;
63
64 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
65 {
66 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
67 int jend = jstart;
68 for(i = 0; i < nshellbatch; i++)
69 jend += Q.nprim12[cd+i];
70
71 // Clear the beginning of the workspace (where we are accumulating integrals)
72 memset(work, 0, SIMINT_NSHELL_SIMD * 160 * sizeof(double));
73 abcd = 0;
74
75
76 for(i = istart; i < iend; ++i)
77 {
78 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
79
80 if(check_screen)
81 {
82 // Skip this whole thing if always insignificant
83 if((P.screen[i] * Q.screen_max) < screen_tol)
84 continue;
85 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
86 }
87
88 icd = 0;
89 iprimcd = 0;
90 nprim_icd = Q.nprim12[cd];
91 double * restrict PRIM_PTR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
92 double * restrict PRIM_PTR_INT__s_f_f_s = INT__s_f_f_s + abcd * 100;
93
94
95
96 // Load these one per loop over i
97 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
98 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
99 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
100
101 const SIMINT_DBLTYPE P_PB[3] = { SIMINT_DBLSET1(P.PB_x[i]), SIMINT_DBLSET1(P.PB_y[i]), SIMINT_DBLSET1(P.PB_z[i]) };
102
103 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
104 {
105 // calculate the shell offsets
106 // these are the offset from the shell pointed to by cd
107 // for each element
108 int shelloffsets[SIMINT_SIMD_LEN] = {0};
109 int lastoffset = 0;
110 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
111
112 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
113 {
114 // Handle if the first element of the vector is a new shell
115 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
116 {
117 nprim_icd += Q.nprim12[cd + (++icd)];
118 PRIM_PTR_INT__s_f_d_s += 60;
119 PRIM_PTR_INT__s_f_f_s += 100;
120 }
121 iprimcd++;
122 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
123 {
124 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
125 {
126 shelloffsets[n] = shelloffsets[n-1] + 1;
127 lastoffset++;
128 nprim_icd += Q.nprim12[cd + (++icd)];
129 }
130 else
131 shelloffsets[n] = shelloffsets[n-1];
132 iprimcd++;
133 }
134 }
135 else
136 iprimcd += SIMINT_SIMD_LEN;
137
138 // Do we have to compute this vector (or has it been screened out)?
139 // (not_screened != 0 means we have to do this vector)
140 if(check_screen)
141 {
142 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
143 if(vmax < screen_tol)
144 {
145 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
146 PRIM_PTR_INT__s_f_f_s += lastoffset*100;
147 continue;
148 }
149 }
150
151 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
152 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
153 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
154 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
155
156
157 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
158 SIMINT_DBLTYPE PQ[3];
159 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
160 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
161 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
162 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
163 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
164 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
165
166 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
167 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
168 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
169 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
170 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
171 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
172 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
173
174 // NOTE: Minus sign!
175 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
176 SIMINT_DBLTYPE aop_PQ[3];
177 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
178 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
179 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
180
181 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
182 SIMINT_DBLTYPE aoq_PQ[3];
183 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
184 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
185 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
186 // Put a minus sign here so we don't have to in RR routines
187 a_over_q = SIMINT_NEG(a_over_q);
188
189
190 //////////////////////////////////////////////
191 // Fjt function section
192 // Maximum v value: 6
193 //////////////////////////////////////////////
194 // The parameter to the Fjt function
195 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
196
197
198 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
199
200
201 boys_F_split(PRIM_INT__s_s_s_s, F_x, 6);
202 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
203 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
204 for(n = 0; n <= 6; n++)
205 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
206
207 //////////////////////////////////////////////
208 // Primitive integrals: Vertical recurrance
209 //////////////////////////////////////////////
210
211 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
212 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
213 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
214 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
215 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
216 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
217 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
218
219
220
221 // Forming PRIM_INT__s_p_s_s[6 * 3];
222 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
223 {
224
225 PRIM_INT__s_p_s_s[n * 3 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
226 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]);
227
228 PRIM_INT__s_p_s_s[n * 3 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
229 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]);
230
231 PRIM_INT__s_p_s_s[n * 3 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
232 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]);
233
234 }
235
236
237
238 // Forming PRIM_INT__s_d_s_s[5 * 6];
239 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
240 {
241
242 PRIM_INT__s_d_s_s[n * 6 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_p_s_s[n * 3 + 0]);
243 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]);
244 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]);
245
246 PRIM_INT__s_d_s_s[n * 6 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 0]);
247 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]);
248
249 PRIM_INT__s_d_s_s[n * 6 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 0]);
250 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]);
251
252 PRIM_INT__s_d_s_s[n * 6 + 3] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 1]);
253 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]);
254 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]);
255
256 PRIM_INT__s_d_s_s[n * 6 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 1]);
257 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]);
258
259 PRIM_INT__s_d_s_s[n * 6 + 5] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 2]);
260 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]);
261 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]);
262
263 }
264
265
266
267 // Forming PRIM_INT__s_f_s_s[4 * 10];
268 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
269 {
270
271 PRIM_INT__s_f_s_s[n * 10 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
272 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]);
273 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]);
274
275 PRIM_INT__s_f_s_s[n * 10 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
276 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]);
277
278 PRIM_INT__s_f_s_s[n * 10 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
279 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]);
280
281 PRIM_INT__s_f_s_s[n * 10 + 3] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
282 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]);
283
284 PRIM_INT__s_f_s_s[n * 10 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
285 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]);
286
287 PRIM_INT__s_f_s_s[n * 10 + 5] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
288 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]);
289
290 PRIM_INT__s_f_s_s[n * 10 + 6] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
291 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]);
292 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]);
293
294 PRIM_INT__s_f_s_s[n * 10 + 7] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
295 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]);
296
297 PRIM_INT__s_f_s_s[n * 10 + 8] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
298 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]);
299
300 PRIM_INT__s_f_s_s[n * 10 + 9] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
301 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]);
302 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]);
303
304 }
305
306
307 VRR_K_s_f_p_s(
308 PRIM_INT__s_f_p_s,
309 PRIM_INT__s_f_s_s,
310 PRIM_INT__s_d_s_s,
311 Q_PA,
312 aoq_PQ,
313 one_over_2pq,
314 3);
315
316
317
318 // Forming PRIM_INT__s_d_p_s[3 * 18];
319 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
320 {
321
322 PRIM_INT__s_d_p_s[n * 18 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
323 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]);
324 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]);
325
326 PRIM_INT__s_d_p_s[n * 18 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
327 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]);
328
329 PRIM_INT__s_d_p_s[n * 18 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
330 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]);
331
332 PRIM_INT__s_d_p_s[n * 18 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 1]);
333 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]);
334 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]);
335
336 PRIM_INT__s_d_p_s[n * 18 + 4] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 1]);
337 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]);
338 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]);
339
340 PRIM_INT__s_d_p_s[n * 18 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
341 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]);
342
343 PRIM_INT__s_d_p_s[n * 18 + 6] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 2]);
344 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]);
345 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]);
346
347 PRIM_INT__s_d_p_s[n * 18 + 7] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 2]);
348 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]);
349
350 PRIM_INT__s_d_p_s[n * 18 + 8] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 2]);
351 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]);
352 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]);
353
354 PRIM_INT__s_d_p_s[n * 18 + 9] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
355 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]);
356
357 PRIM_INT__s_d_p_s[n * 18 + 10] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
358 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]);
359 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]);
360
361 PRIM_INT__s_d_p_s[n * 18 + 11] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
362 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]);
363
364 PRIM_INT__s_d_p_s[n * 18 + 12] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 4]);
365 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]);
366
367 PRIM_INT__s_d_p_s[n * 18 + 13] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 4]);
368 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]);
369 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]);
370
371 PRIM_INT__s_d_p_s[n * 18 + 14] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 4]);
372 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]);
373 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]);
374
375 PRIM_INT__s_d_p_s[n * 18 + 15] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
376 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]);
377
378 PRIM_INT__s_d_p_s[n * 18 + 16] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
379 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]);
380
381 PRIM_INT__s_d_p_s[n * 18 + 17] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
382 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]);
383 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]);
384
385 }
386
387
388 VRR_K_s_f_d_s(
389 PRIM_INT__s_f_d_s,
390 PRIM_INT__s_f_p_s,
391 PRIM_INT__s_f_s_s,
392 PRIM_INT__s_d_p_s,
393 Q_PA,
394 a_over_q,
395 aoq_PQ,
396 one_over_2pq,
397 one_over_2q,
398 2);
399
400
401
402 // Forming PRIM_INT__s_p_p_s[3 * 9];
403 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
404 {
405
406 PRIM_INT__s_p_p_s[n * 9 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_p_s_s[n * 3 + 0]);
407 PRIM_INT__s_p_p_s[n * 9 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_p_p_s[n * 9 + 0]);
408 PRIM_INT__s_p_p_s[n * 9 + 0] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_p_s[n * 9 + 0]);
409
410 PRIM_INT__s_p_p_s[n * 9 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_p_s_s[n * 3 + 0]);
411 PRIM_INT__s_p_p_s[n * 9 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_p_p_s[n * 9 + 1]);
412
413 PRIM_INT__s_p_p_s[n * 9 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_p_s_s[n * 3 + 0]);
414 PRIM_INT__s_p_p_s[n * 9 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 0], PRIM_INT__s_p_p_s[n * 9 + 2]);
415
416 PRIM_INT__s_p_p_s[n * 9 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_p_s_s[n * 3 + 1]);
417 PRIM_INT__s_p_p_s[n * 9 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_p_p_s[n * 9 + 3]);
418
419 PRIM_INT__s_p_p_s[n * 9 + 4] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_p_s_s[n * 3 + 1]);
420 PRIM_INT__s_p_p_s[n * 9 + 4] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_p_p_s[n * 9 + 4]);
421 PRIM_INT__s_p_p_s[n * 9 + 4] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_p_s[n * 9 + 4]);
422
423 PRIM_INT__s_p_p_s[n * 9 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_p_s_s[n * 3 + 1]);
424 PRIM_INT__s_p_p_s[n * 9 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 1], PRIM_INT__s_p_p_s[n * 9 + 5]);
425
426 PRIM_INT__s_p_p_s[n * 9 + 6] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_p_s_s[n * 3 + 2]);
427 PRIM_INT__s_p_p_s[n * 9 + 6] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_p_p_s[n * 9 + 6]);
428
429 PRIM_INT__s_p_p_s[n * 9 + 7] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_p_s_s[n * 3 + 2]);
430 PRIM_INT__s_p_p_s[n * 9 + 7] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_p_p_s[n * 9 + 7]);
431
432 PRIM_INT__s_p_p_s[n * 9 + 8] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_p_s_s[n * 3 + 2]);
433 PRIM_INT__s_p_p_s[n * 9 + 8] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_p_s_s[(n+1) * 3 + 2], PRIM_INT__s_p_p_s[n * 9 + 8]);
434 PRIM_INT__s_p_p_s[n * 9 + 8] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_p_p_s[n * 9 + 8]);
435
436 }
437
438
439 VRR_K_s_d_d_s(
440 PRIM_INT__s_d_d_s,
441 PRIM_INT__s_d_p_s,
442 PRIM_INT__s_d_s_s,
443 PRIM_INT__s_p_p_s,
444 Q_PA,
445 a_over_q,
446 aoq_PQ,
447 one_over_2pq,
448 one_over_2q,
449 2);
450
451
452 ostei_general_vrr_K(0, 3, 3, 0, 1,
453 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
454 PRIM_INT__s_f_d_s, PRIM_INT__s_f_p_s, NULL, NULL, PRIM_INT__s_d_d_s, PRIM_INT__s_f_f_s);
455
456
457
458
459 ////////////////////////////////////
460 // Accumulate contracted integrals
461 ////////////////////////////////////
462 if(lastoffset == 0)
463 {
464 contract_all(60, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
465 contract_all(100, PRIM_INT__s_f_f_s, PRIM_PTR_INT__s_f_f_s);
466 }
467 else
468 {
469 contract(60, shelloffsets, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
470 contract(100, shelloffsets, PRIM_INT__s_f_f_s, PRIM_PTR_INT__s_f_f_s);
471 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
472 PRIM_PTR_INT__s_f_f_s += lastoffset*100;
473 }
474
475 } // close loop over j
476 } // close loop over i
477
478 //Advance to the next batch
479 jstart = SIMINT_SIMD_ROUND(jend);
480
481 //////////////////////////////////////////////
482 // Contracted integrals: Horizontal recurrance
483 //////////////////////////////////////////////
484
485
486
487
488 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
489 {
490 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
491
492 // set up HRR pointers
493 double const * restrict HRR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
494 double const * restrict HRR_INT__s_f_f_s = INT__s_f_f_s + abcd * 100;
495 double * restrict HRR_INT__s_f_d_p = INT__s_f_d_p + real_abcd * 180;
496
497 // form INT__s_f_d_p
498 for(ibra = 0; ibra < 10; ++ibra)
499 {
500 HRR_INT__s_f_d_p[ibra * 18 + 0] = HRR_INT__s_f_f_s[ibra * 10 + 0] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 0] );
501
502 HRR_INT__s_f_d_p[ibra * 18 + 1] = HRR_INT__s_f_f_s[ibra * 10 + 1] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 0] );
503
504 HRR_INT__s_f_d_p[ibra * 18 + 2] = HRR_INT__s_f_f_s[ibra * 10 + 2] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 0] );
505
506 HRR_INT__s_f_d_p[ibra * 18 + 3] = HRR_INT__s_f_f_s[ibra * 10 + 1] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 1] );
507
508 HRR_INT__s_f_d_p[ibra * 18 + 4] = HRR_INT__s_f_f_s[ibra * 10 + 3] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 1] );
509
510 HRR_INT__s_f_d_p[ibra * 18 + 5] = HRR_INT__s_f_f_s[ibra * 10 + 4] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 1] );
511
512 HRR_INT__s_f_d_p[ibra * 18 + 6] = HRR_INT__s_f_f_s[ibra * 10 + 2] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 2] );
513
514 HRR_INT__s_f_d_p[ibra * 18 + 7] = HRR_INT__s_f_f_s[ibra * 10 + 4] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 2] );
515
516 HRR_INT__s_f_d_p[ibra * 18 + 8] = HRR_INT__s_f_f_s[ibra * 10 + 5] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 2] );
517
518 HRR_INT__s_f_d_p[ibra * 18 + 9] = HRR_INT__s_f_f_s[ibra * 10 + 3] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 3] );
519
520 HRR_INT__s_f_d_p[ibra * 18 + 10] = HRR_INT__s_f_f_s[ibra * 10 + 6] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 3] );
521
522 HRR_INT__s_f_d_p[ibra * 18 + 11] = HRR_INT__s_f_f_s[ibra * 10 + 7] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 3] );
523
524 HRR_INT__s_f_d_p[ibra * 18 + 12] = HRR_INT__s_f_f_s[ibra * 10 + 4] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 4] );
525
526 HRR_INT__s_f_d_p[ibra * 18 + 13] = HRR_INT__s_f_f_s[ibra * 10 + 7] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 4] );
527
528 HRR_INT__s_f_d_p[ibra * 18 + 14] = HRR_INT__s_f_f_s[ibra * 10 + 8] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 4] );
529
530 HRR_INT__s_f_d_p[ibra * 18 + 15] = HRR_INT__s_f_f_s[ibra * 10 + 5] + ( hCD[0] * HRR_INT__s_f_d_s[ibra * 6 + 5] );
531
532 HRR_INT__s_f_d_p[ibra * 18 + 16] = HRR_INT__s_f_f_s[ibra * 10 + 8] + ( hCD[1] * HRR_INT__s_f_d_s[ibra * 6 + 5] );
533
534 HRR_INT__s_f_d_p[ibra * 18 + 17] = HRR_INT__s_f_f_s[ibra * 10 + 9] + ( hCD[2] * HRR_INT__s_f_d_s[ibra * 6 + 5] );
535
536 }
537
538
539 } // close HRR loop
540
541
542 } // close loop cdbatch
543
544 istart = iend;
545 } // close loop over ab
546
547 return P.nshell12_clip * Q.nshell12_clip;
548 }
549
550