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_f_p_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_f_p_p)8 int ostei_p_f_p_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_f_p_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__p_f_p_p);
17 int ab, cd, abcd;
18 int istart, jstart;
19 int iprimcd, nprim_icd, icd;
20 const int check_screen = (screen_tol > 0.0);
21 int i, j;
22 int n;
23 int not_screened;
24 int real_abcd;
25 int iket;
26 int ibra;
27
28 // partition workspace
29 double * const INT__s_f_p_s = work + (SIMINT_NSHELL_SIMD * 0);
30 double * const INT__s_f_d_s = work + (SIMINT_NSHELL_SIMD * 30);
31 double * const INT__s_g_p_s = work + (SIMINT_NSHELL_SIMD * 90);
32 double * const INT__s_g_d_s = work + (SIMINT_NSHELL_SIMD * 135);
33 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*225);
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_p_s_s = primwork + 7;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_s_s = primwork + 25;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_p_s = primwork + 55;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_s_s = primwork + 91;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_p_s = primwork + 131;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__s_f_d_s = primwork + 191;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_s_s = primwork + 251;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_p_s = primwork + 296;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__s_g_d_s = primwork + 386;
44 double * const hrrwork = (double *)(primwork + 476);
45 double * const HRR_INT__p_f_p_s = hrrwork + 0;
46 double * const HRR_INT__p_f_d_s = hrrwork + 90;
47
48
49 // Create constants
50 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
51 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
52 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
53 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
54 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
55
56
57 ////////////////////////////////////////
58 // Loop over shells and primitives
59 ////////////////////////////////////////
60
61 real_abcd = 0;
62 istart = 0;
63 for(ab = 0; ab < P.nshell12_clip; ++ab)
64 {
65 const int iend = istart + P.nprim12[ab];
66
67 cd = 0;
68 jstart = 0;
69
70 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
71 {
72 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
73 int jend = jstart;
74 for(i = 0; i < nshellbatch; i++)
75 jend += Q.nprim12[cd+i];
76
77 // Clear the beginning of the workspace (where we are accumulating integrals)
78 memset(work, 0, SIMINT_NSHELL_SIMD * 225 * sizeof(double));
79 abcd = 0;
80
81
82 for(i = istart; i < iend; ++i)
83 {
84 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
85
86 if(check_screen)
87 {
88 // Skip this whole thing if always insignificant
89 if((P.screen[i] * Q.screen_max) < screen_tol)
90 continue;
91 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
92 }
93
94 icd = 0;
95 iprimcd = 0;
96 nprim_icd = Q.nprim12[cd];
97 double * restrict PRIM_PTR_INT__s_f_p_s = INT__s_f_p_s + abcd * 30;
98 double * restrict PRIM_PTR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
99 double * restrict PRIM_PTR_INT__s_g_p_s = INT__s_g_p_s + abcd * 45;
100 double * restrict PRIM_PTR_INT__s_g_d_s = INT__s_g_d_s + abcd * 90;
101
102
103
104 // Load these one per loop over i
105 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
106 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
107 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
108
109 const SIMINT_DBLTYPE P_PB[3] = { SIMINT_DBLSET1(P.PB_x[i]), SIMINT_DBLSET1(P.PB_y[i]), SIMINT_DBLSET1(P.PB_z[i]) };
110
111 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
112 {
113 // calculate the shell offsets
114 // these are the offset from the shell pointed to by cd
115 // for each element
116 int shelloffsets[SIMINT_SIMD_LEN] = {0};
117 int lastoffset = 0;
118 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
119
120 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
121 {
122 // Handle if the first element of the vector is a new shell
123 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
124 {
125 nprim_icd += Q.nprim12[cd + (++icd)];
126 PRIM_PTR_INT__s_f_p_s += 30;
127 PRIM_PTR_INT__s_f_d_s += 60;
128 PRIM_PTR_INT__s_g_p_s += 45;
129 PRIM_PTR_INT__s_g_d_s += 90;
130 }
131 iprimcd++;
132 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
133 {
134 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
135 {
136 shelloffsets[n] = shelloffsets[n-1] + 1;
137 lastoffset++;
138 nprim_icd += Q.nprim12[cd + (++icd)];
139 }
140 else
141 shelloffsets[n] = shelloffsets[n-1];
142 iprimcd++;
143 }
144 }
145 else
146 iprimcd += SIMINT_SIMD_LEN;
147
148 // Do we have to compute this vector (or has it been screened out)?
149 // (not_screened != 0 means we have to do this vector)
150 if(check_screen)
151 {
152 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
153 if(vmax < screen_tol)
154 {
155 PRIM_PTR_INT__s_f_p_s += lastoffset*30;
156 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
157 PRIM_PTR_INT__s_g_p_s += lastoffset*45;
158 PRIM_PTR_INT__s_g_d_s += lastoffset*90;
159 continue;
160 }
161 }
162
163 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
164 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
165 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
166 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
167
168
169 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
170 SIMINT_DBLTYPE PQ[3];
171 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
172 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
173 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
174 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
175 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
176 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
177
178 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
179 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
180 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
181 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
182 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
183 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
184 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
185
186 // NOTE: Minus sign!
187 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
188 SIMINT_DBLTYPE aop_PQ[3];
189 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
190 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
191 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
192
193 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
194 SIMINT_DBLTYPE aoq_PQ[3];
195 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
196 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
197 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
198 // Put a minus sign here so we don't have to in RR routines
199 a_over_q = SIMINT_NEG(a_over_q);
200
201
202 //////////////////////////////////////////////
203 // Fjt function section
204 // Maximum v value: 6
205 //////////////////////////////////////////////
206 // The parameter to the Fjt function
207 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
208
209
210 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
211
212
213 boys_F_split(PRIM_INT__s_s_s_s, F_x, 6);
214 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
215 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
216 for(n = 0; n <= 6; n++)
217 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
218
219 //////////////////////////////////////////////
220 // Primitive integrals: Vertical recurrance
221 //////////////////////////////////////////////
222
223 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
224 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
225 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
226 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
227 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
228 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
229 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
230 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
231
232
233
234 // Forming PRIM_INT__s_p_s_s[6 * 3];
235 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
236 {
237
238 PRIM_INT__s_p_s_s[n * 3 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
239 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]);
240
241 PRIM_INT__s_p_s_s[n * 3 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
242 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]);
243
244 PRIM_INT__s_p_s_s[n * 3 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
245 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]);
246
247 }
248
249
250
251 // Forming PRIM_INT__s_d_s_s[5 * 6];
252 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
253 {
254
255 PRIM_INT__s_d_s_s[n * 6 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_p_s_s[n * 3 + 0]);
256 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]);
257 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]);
258
259 PRIM_INT__s_d_s_s[n * 6 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 0]);
260 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]);
261
262 PRIM_INT__s_d_s_s[n * 6 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 0]);
263 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]);
264
265 PRIM_INT__s_d_s_s[n * 6 + 3] = SIMINT_MUL(P_PB[1], PRIM_INT__s_p_s_s[n * 3 + 1]);
266 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]);
267 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]);
268
269 PRIM_INT__s_d_s_s[n * 6 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 1]);
270 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]);
271
272 PRIM_INT__s_d_s_s[n * 6 + 5] = SIMINT_MUL(P_PB[2], PRIM_INT__s_p_s_s[n * 3 + 2]);
273 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]);
274 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]);
275
276 }
277
278
279
280 // Forming PRIM_INT__s_f_s_s[4 * 10];
281 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
282 {
283
284 PRIM_INT__s_f_s_s[n * 10 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
285 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]);
286 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]);
287
288 PRIM_INT__s_f_s_s[n * 10 + 1] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
289 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]);
290
291 PRIM_INT__s_f_s_s[n * 10 + 2] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
292 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]);
293
294 PRIM_INT__s_f_s_s[n * 10 + 3] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
295 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]);
296
297 PRIM_INT__s_f_s_s[n * 10 + 4] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
298 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]);
299
300 PRIM_INT__s_f_s_s[n * 10 + 5] = SIMINT_MUL(P_PB[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
301 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]);
302
303 PRIM_INT__s_f_s_s[n * 10 + 6] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
304 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]);
305 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]);
306
307 PRIM_INT__s_f_s_s[n * 10 + 7] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
308 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]);
309
310 PRIM_INT__s_f_s_s[n * 10 + 8] = SIMINT_MUL(P_PB[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
311 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]);
312
313 PRIM_INT__s_f_s_s[n * 10 + 9] = SIMINT_MUL(P_PB[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
314 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]);
315 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]);
316
317 }
318
319
320 VRR_K_s_f_p_s(
321 PRIM_INT__s_f_p_s,
322 PRIM_INT__s_f_s_s,
323 PRIM_INT__s_d_s_s,
324 Q_PA,
325 aoq_PQ,
326 one_over_2pq,
327 2);
328
329
330
331 // Forming PRIM_INT__s_d_p_s[2 * 18];
332 for(n = 0; n < 2; ++n) // loop over orders of auxiliary function
333 {
334
335 PRIM_INT__s_d_p_s[n * 18 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 0]);
336 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]);
337 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]);
338
339 PRIM_INT__s_d_p_s[n * 18 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 0]);
340 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]);
341
342 PRIM_INT__s_d_p_s[n * 18 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 0]);
343 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]);
344
345 PRIM_INT__s_d_p_s[n * 18 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 1]);
346 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]);
347 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]);
348
349 PRIM_INT__s_d_p_s[n * 18 + 4] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 1]);
350 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]);
351 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]);
352
353 PRIM_INT__s_d_p_s[n * 18 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 1]);
354 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]);
355
356 PRIM_INT__s_d_p_s[n * 18 + 6] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 2]);
357 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]);
358 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]);
359
360 PRIM_INT__s_d_p_s[n * 18 + 7] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 2]);
361 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]);
362
363 PRIM_INT__s_d_p_s[n * 18 + 8] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 2]);
364 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]);
365 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]);
366
367 PRIM_INT__s_d_p_s[n * 18 + 9] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 3]);
368 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]);
369
370 PRIM_INT__s_d_p_s[n * 18 + 10] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 3]);
371 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]);
372 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]);
373
374 PRIM_INT__s_d_p_s[n * 18 + 11] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 3]);
375 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]);
376
377 PRIM_INT__s_d_p_s[n * 18 + 12] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 4]);
378 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]);
379
380 PRIM_INT__s_d_p_s[n * 18 + 13] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 4]);
381 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]);
382 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]);
383
384 PRIM_INT__s_d_p_s[n * 18 + 14] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 4]);
385 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]);
386 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]);
387
388 PRIM_INT__s_d_p_s[n * 18 + 15] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_d_s_s[n * 6 + 5]);
389 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]);
390
391 PRIM_INT__s_d_p_s[n * 18 + 16] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_d_s_s[n * 6 + 5]);
392 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]);
393
394 PRIM_INT__s_d_p_s[n * 18 + 17] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_d_s_s[n * 6 + 5]);
395 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]);
396 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]);
397
398 }
399
400
401 VRR_K_s_f_d_s(
402 PRIM_INT__s_f_d_s,
403 PRIM_INT__s_f_p_s,
404 PRIM_INT__s_f_s_s,
405 PRIM_INT__s_d_p_s,
406 Q_PA,
407 a_over_q,
408 aoq_PQ,
409 one_over_2pq,
410 one_over_2q,
411 1);
412
413
414 VRR_J_s_g_s_s(
415 PRIM_INT__s_g_s_s,
416 PRIM_INT__s_f_s_s,
417 PRIM_INT__s_d_s_s,
418 P_PB,
419 a_over_p,
420 aop_PQ,
421 one_over_2p,
422 3);
423
424
425 VRR_K_s_g_p_s(
426 PRIM_INT__s_g_p_s,
427 PRIM_INT__s_g_s_s,
428 PRIM_INT__s_f_s_s,
429 Q_PA,
430 aoq_PQ,
431 one_over_2pq,
432 2);
433
434
435 ostei_general_vrr_K(0, 4, 2, 0, 1,
436 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
437 PRIM_INT__s_g_p_s, PRIM_INT__s_g_s_s, NULL, NULL, PRIM_INT__s_f_p_s, PRIM_INT__s_g_d_s);
438
439
440
441
442 ////////////////////////////////////
443 // Accumulate contracted integrals
444 ////////////////////////////////////
445 if(lastoffset == 0)
446 {
447 contract_all(30, PRIM_INT__s_f_p_s, PRIM_PTR_INT__s_f_p_s);
448 contract_all(60, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
449 contract_all(45, PRIM_INT__s_g_p_s, PRIM_PTR_INT__s_g_p_s);
450 contract_all(90, PRIM_INT__s_g_d_s, PRIM_PTR_INT__s_g_d_s);
451 }
452 else
453 {
454 contract(30, shelloffsets, PRIM_INT__s_f_p_s, PRIM_PTR_INT__s_f_p_s);
455 contract(60, shelloffsets, PRIM_INT__s_f_d_s, PRIM_PTR_INT__s_f_d_s);
456 contract(45, shelloffsets, PRIM_INT__s_g_p_s, PRIM_PTR_INT__s_g_p_s);
457 contract(90, shelloffsets, PRIM_INT__s_g_d_s, PRIM_PTR_INT__s_g_d_s);
458 PRIM_PTR_INT__s_f_p_s += lastoffset*30;
459 PRIM_PTR_INT__s_f_d_s += lastoffset*60;
460 PRIM_PTR_INT__s_g_p_s += lastoffset*45;
461 PRIM_PTR_INT__s_g_d_s += lastoffset*90;
462 }
463
464 } // close loop over j
465 } // close loop over i
466
467 //Advance to the next batch
468 jstart = SIMINT_SIMD_ROUND(jend);
469
470 //////////////////////////////////////////////
471 // Contracted integrals: Horizontal recurrance
472 //////////////////////////////////////////////
473
474
475 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
476
477
478 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
479 {
480 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
481
482 // set up HRR pointers
483 double const * restrict HRR_INT__s_f_p_s = INT__s_f_p_s + abcd * 30;
484 double const * restrict HRR_INT__s_f_d_s = INT__s_f_d_s + abcd * 60;
485 double const * restrict HRR_INT__s_g_p_s = INT__s_g_p_s + abcd * 45;
486 double const * restrict HRR_INT__s_g_d_s = INT__s_g_d_s + abcd * 90;
487 double * restrict HRR_INT__p_f_p_p = INT__p_f_p_p + real_abcd * 270;
488
489 // form INT__p_f_p_s
490 HRR_I_p_f(
491 HRR_INT__p_f_p_s,
492 HRR_INT__s_f_p_s,
493 HRR_INT__s_g_p_s,
494 hAB, 3);
495
496 // form INT__p_f_d_s
497 HRR_I_p_f(
498 HRR_INT__p_f_d_s,
499 HRR_INT__s_f_d_s,
500 HRR_INT__s_g_d_s,
501 hAB, 6);
502
503 // form INT__p_f_p_p
504 for(ibra = 0; ibra < 30; ++ibra)
505 {
506 HRR_INT__p_f_p_p[ibra * 9 + 0] = HRR_INT__p_f_d_s[ibra * 6 + 0] + ( hCD[0] * HRR_INT__p_f_p_s[ibra * 3 + 0] );
507
508 HRR_INT__p_f_p_p[ibra * 9 + 1] = HRR_INT__p_f_d_s[ibra * 6 + 1] + ( hCD[1] * HRR_INT__p_f_p_s[ibra * 3 + 0] );
509
510 HRR_INT__p_f_p_p[ibra * 9 + 2] = HRR_INT__p_f_d_s[ibra * 6 + 2] + ( hCD[2] * HRR_INT__p_f_p_s[ibra * 3 + 0] );
511
512 HRR_INT__p_f_p_p[ibra * 9 + 3] = HRR_INT__p_f_d_s[ibra * 6 + 1] + ( hCD[0] * HRR_INT__p_f_p_s[ibra * 3 + 1] );
513
514 HRR_INT__p_f_p_p[ibra * 9 + 4] = HRR_INT__p_f_d_s[ibra * 6 + 3] + ( hCD[1] * HRR_INT__p_f_p_s[ibra * 3 + 1] );
515
516 HRR_INT__p_f_p_p[ibra * 9 + 5] = HRR_INT__p_f_d_s[ibra * 6 + 4] + ( hCD[2] * HRR_INT__p_f_p_s[ibra * 3 + 1] );
517
518 HRR_INT__p_f_p_p[ibra * 9 + 6] = HRR_INT__p_f_d_s[ibra * 6 + 2] + ( hCD[0] * HRR_INT__p_f_p_s[ibra * 3 + 2] );
519
520 HRR_INT__p_f_p_p[ibra * 9 + 7] = HRR_INT__p_f_d_s[ibra * 6 + 4] + ( hCD[1] * HRR_INT__p_f_p_s[ibra * 3 + 2] );
521
522 HRR_INT__p_f_p_p[ibra * 9 + 8] = HRR_INT__p_f_d_s[ibra * 6 + 5] + ( hCD[2] * HRR_INT__p_f_p_s[ibra * 3 + 2] );
523
524 }
525
526
527 } // close HRR loop
528
529
530 } // close loop cdbatch
531
532 istart = iend;
533 } // close loop over ab
534
535 return P.nshell12_clip * Q.nshell12_clip;
536 }
537
538