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_g_g_d_s(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__g_g_d_s)8 int ostei_g_g_d_s(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__g_g_d_s)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__g_g_d_s);
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
27 // partition workspace
28 double * const INT__g_s_d_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__h_s_d_s = work + (SIMINT_NSHELL_SIMD * 90);
30 double * const INT__i_s_d_s = work + (SIMINT_NSHELL_SIMD * 216);
31 double * const INT__k_s_d_s = work + (SIMINT_NSHELL_SIMD * 384);
32 double * const INT__l_s_d_s = work + (SIMINT_NSHELL_SIMD * 600);
33 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*870);
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 11;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 41;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 95;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_p_s = primwork + 175;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 235;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_p_s = primwork + 340;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_d_s = primwork + 430;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 520;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_p_s = primwork + 646;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_d_s = primwork + 772;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_s_s = primwork + 898;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_p_s = primwork + 1038;
47 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_d_s = primwork + 1206;
48 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_s_s = primwork + 1374;
49 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_p_s = primwork + 1518;
50 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_d_s = primwork + 1734;
51 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_s_s = primwork + 1950;
52 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_p_s = primwork + 2085;
53 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_d_s = primwork + 2355;
54 double * const hrrwork = (double *)(primwork + 2625);
55 double * const HRR_INT__g_p_d_s = hrrwork + 0;
56 double * const HRR_INT__g_d_d_s = hrrwork + 270;
57 double * const HRR_INT__g_f_d_s = hrrwork + 810;
58 double * const HRR_INT__h_p_d_s = hrrwork + 1710;
59 double * const HRR_INT__h_d_d_s = hrrwork + 2088;
60 double * const HRR_INT__h_f_d_s = hrrwork + 2844;
61 double * const HRR_INT__i_p_d_s = hrrwork + 4104;
62 double * const HRR_INT__i_d_d_s = hrrwork + 4608;
63 double * const HRR_INT__k_p_d_s = hrrwork + 5616;
64
65
66 // Create constants
67 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
68 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
69 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
70 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
71 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
72 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
73 const SIMINT_DBLTYPE const_7 = SIMINT_DBLSET1(7);
74 const SIMINT_DBLTYPE const_8 = SIMINT_DBLSET1(8);
75 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
76
77
78 ////////////////////////////////////////
79 // Loop over shells and primitives
80 ////////////////////////////////////////
81
82 real_abcd = 0;
83 istart = 0;
84 for(ab = 0; ab < P.nshell12_clip; ++ab)
85 {
86 const int iend = istart + P.nprim12[ab];
87
88 cd = 0;
89 jstart = 0;
90
91 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
92 {
93 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
94 int jend = jstart;
95 for(i = 0; i < nshellbatch; i++)
96 jend += Q.nprim12[cd+i];
97
98 // Clear the beginning of the workspace (where we are accumulating integrals)
99 memset(work, 0, SIMINT_NSHELL_SIMD * 870 * sizeof(double));
100 abcd = 0;
101
102
103 for(i = istart; i < iend; ++i)
104 {
105 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
106
107 if(check_screen)
108 {
109 // Skip this whole thing if always insignificant
110 if((P.screen[i] * Q.screen_max) < screen_tol)
111 continue;
112 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
113 }
114
115 icd = 0;
116 iprimcd = 0;
117 nprim_icd = Q.nprim12[cd];
118 double * restrict PRIM_PTR_INT__g_s_d_s = INT__g_s_d_s + abcd * 90;
119 double * restrict PRIM_PTR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126;
120 double * restrict PRIM_PTR_INT__i_s_d_s = INT__i_s_d_s + abcd * 168;
121 double * restrict PRIM_PTR_INT__k_s_d_s = INT__k_s_d_s + abcd * 216;
122 double * restrict PRIM_PTR_INT__l_s_d_s = INT__l_s_d_s + abcd * 270;
123
124
125
126 // Load these one per loop over i
127 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
128 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
129 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
130
131 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
132
133 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
134 {
135 // calculate the shell offsets
136 // these are the offset from the shell pointed to by cd
137 // for each element
138 int shelloffsets[SIMINT_SIMD_LEN] = {0};
139 int lastoffset = 0;
140 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
141
142 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
143 {
144 // Handle if the first element of the vector is a new shell
145 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
146 {
147 nprim_icd += Q.nprim12[cd + (++icd)];
148 PRIM_PTR_INT__g_s_d_s += 90;
149 PRIM_PTR_INT__h_s_d_s += 126;
150 PRIM_PTR_INT__i_s_d_s += 168;
151 PRIM_PTR_INT__k_s_d_s += 216;
152 PRIM_PTR_INT__l_s_d_s += 270;
153 }
154 iprimcd++;
155 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
156 {
157 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
158 {
159 shelloffsets[n] = shelloffsets[n-1] + 1;
160 lastoffset++;
161 nprim_icd += Q.nprim12[cd + (++icd)];
162 }
163 else
164 shelloffsets[n] = shelloffsets[n-1];
165 iprimcd++;
166 }
167 }
168 else
169 iprimcd += SIMINT_SIMD_LEN;
170
171 // Do we have to compute this vector (or has it been screened out)?
172 // (not_screened != 0 means we have to do this vector)
173 if(check_screen)
174 {
175 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
176 if(vmax < screen_tol)
177 {
178 PRIM_PTR_INT__g_s_d_s += lastoffset*90;
179 PRIM_PTR_INT__h_s_d_s += lastoffset*126;
180 PRIM_PTR_INT__i_s_d_s += lastoffset*168;
181 PRIM_PTR_INT__k_s_d_s += lastoffset*216;
182 PRIM_PTR_INT__l_s_d_s += lastoffset*270;
183 continue;
184 }
185 }
186
187 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
188 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
189 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
190 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
191
192
193 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
194 SIMINT_DBLTYPE PQ[3];
195 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
196 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
197 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
198 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
199 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
200 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
201
202 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
203 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
204 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
205 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
206 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
207 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
208 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
209
210 // NOTE: Minus sign!
211 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
212 SIMINT_DBLTYPE aop_PQ[3];
213 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
214 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
215 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
216
217 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
218 SIMINT_DBLTYPE aoq_PQ[3];
219 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
220 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
221 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
222 // Put a minus sign here so we don't have to in RR routines
223 a_over_q = SIMINT_NEG(a_over_q);
224
225
226 //////////////////////////////////////////////
227 // Fjt function section
228 // Maximum v value: 10
229 //////////////////////////////////////////////
230 // The parameter to the Fjt function
231 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
232
233
234 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
235
236
237 boys_F_split(PRIM_INT__s_s_s_s, F_x, 10);
238 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
239 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
240 for(n = 0; n <= 10; n++)
241 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
242
243 //////////////////////////////////////////////
244 // Primitive integrals: Vertical recurrance
245 //////////////////////////////////////////////
246
247 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
248 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
249 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
250 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
251 const SIMINT_DBLTYPE vrr_const_5_over_2p = SIMINT_MUL(const_5, one_over_2p);
252 const SIMINT_DBLTYPE vrr_const_6_over_2p = SIMINT_MUL(const_6, one_over_2p);
253 const SIMINT_DBLTYPE vrr_const_7_over_2p = SIMINT_MUL(const_7, one_over_2p);
254 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
255 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
256 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
257 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
258 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
259 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
260 const SIMINT_DBLTYPE vrr_const_6_over_2pq = SIMINT_MUL(const_6, one_over_2pq);
261 const SIMINT_DBLTYPE vrr_const_7_over_2pq = SIMINT_MUL(const_7, one_over_2pq);
262 const SIMINT_DBLTYPE vrr_const_8_over_2pq = SIMINT_MUL(const_8, one_over_2pq);
263
264
265
266 // Forming PRIM_INT__p_s_s_s[10 * 3];
267 for(n = 0; n < 10; ++n) // loop over orders of auxiliary function
268 {
269
270 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
271 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]);
272
273 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
274 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 1]);
275
276 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
277 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 2]);
278
279 }
280
281
282
283 // Forming PRIM_INT__d_s_s_s[9 * 6];
284 for(n = 0; n < 9; ++n) // loop over orders of auxiliary function
285 {
286
287 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
288 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 0]);
289 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 0]);
290
291 PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 0]);
292 PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 1]);
293
294 PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 0]);
295 PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 2]);
296
297 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
298 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 3]);
299 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 3]);
300
301 PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 1]);
302 PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 4]);
303
304 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
305 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__d_s_s_s[n * 6 + 5]);
306 PRIM_INT__d_s_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__d_s_s_s[n * 6 + 5]);
307
308 }
309
310
311
312 // Forming PRIM_INT__f_s_s_s[8 * 10];
313 for(n = 0; n < 8; ++n) // loop over orders of auxiliary function
314 {
315
316 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
317 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 0]);
318 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]), PRIM_INT__f_s_s_s[n * 10 + 0]);
319
320 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
321 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 1]);
322
323 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]);
324 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 2]);
325
326 PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 3]);
327 PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 3]);
328
329 PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 1]);
330 PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 1], PRIM_INT__f_s_s_s[n * 10 + 4]);
331
332 PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 5]);
333 PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 5]);
334
335 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
336 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 6]);
337 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__p_s_s_s[n * 3 + 1]), PRIM_INT__f_s_s_s[n * 10 + 6]);
338
339 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]);
340 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 7]);
341
342 PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 5]);
343 PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 8]);
344
345 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
346 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 9]);
347 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__p_s_s_s[n * 3 + 2]), PRIM_INT__f_s_s_s[n * 10 + 9]);
348
349 }
350
351
352 VRR_I_g_s_s_s(
353 PRIM_INT__g_s_s_s,
354 PRIM_INT__f_s_s_s,
355 PRIM_INT__d_s_s_s,
356 P_PA,
357 a_over_p,
358 aop_PQ,
359 one_over_2p,
360 7);
361
362
363 VRR_K_g_s_p_s(
364 PRIM_INT__g_s_p_s,
365 PRIM_INT__g_s_s_s,
366 PRIM_INT__f_s_s_s,
367 Q_PA,
368 aoq_PQ,
369 one_over_2pq,
370 2);
371
372
373 VRR_K_f_s_p_s(
374 PRIM_INT__f_s_p_s,
375 PRIM_INT__f_s_s_s,
376 PRIM_INT__d_s_s_s,
377 Q_PA,
378 aoq_PQ,
379 one_over_2pq,
380 2);
381
382
383 ostei_general_vrr_K(4, 0, 2, 0, 1,
384 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
385 PRIM_INT__g_s_p_s, PRIM_INT__g_s_s_s, NULL, PRIM_INT__f_s_p_s, NULL, PRIM_INT__g_s_d_s);
386
387
388 VRR_I_h_s_s_s(
389 PRIM_INT__h_s_s_s,
390 PRIM_INT__g_s_s_s,
391 PRIM_INT__f_s_s_s,
392 P_PA,
393 a_over_p,
394 aop_PQ,
395 one_over_2p,
396 6);
397
398
399 ostei_general_vrr_K(5, 0, 1, 0, 2,
400 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
401 PRIM_INT__h_s_s_s, NULL, NULL, PRIM_INT__g_s_s_s, NULL, PRIM_INT__h_s_p_s);
402
403
404 ostei_general_vrr_K(5, 0, 2, 0, 1,
405 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
406 PRIM_INT__h_s_p_s, PRIM_INT__h_s_s_s, NULL, PRIM_INT__g_s_p_s, NULL, PRIM_INT__h_s_d_s);
407
408
409 ostei_general_vrr1_I(6, 5,
410 one_over_2p, a_over_p, aop_PQ, P_PA,
411 PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__i_s_s_s);
412
413
414 ostei_general_vrr_K(6, 0, 1, 0, 2,
415 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
416 PRIM_INT__i_s_s_s, NULL, NULL, PRIM_INT__h_s_s_s, NULL, PRIM_INT__i_s_p_s);
417
418
419 ostei_general_vrr_K(6, 0, 2, 0, 1,
420 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
421 PRIM_INT__i_s_p_s, PRIM_INT__i_s_s_s, NULL, PRIM_INT__h_s_p_s, NULL, PRIM_INT__i_s_d_s);
422
423
424 ostei_general_vrr1_I(7, 4,
425 one_over_2p, a_over_p, aop_PQ, P_PA,
426 PRIM_INT__i_s_s_s, PRIM_INT__h_s_s_s, PRIM_INT__k_s_s_s);
427
428
429 ostei_general_vrr_K(7, 0, 1, 0, 2,
430 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
431 PRIM_INT__k_s_s_s, NULL, NULL, PRIM_INT__i_s_s_s, NULL, PRIM_INT__k_s_p_s);
432
433
434 ostei_general_vrr_K(7, 0, 2, 0, 1,
435 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
436 PRIM_INT__k_s_p_s, PRIM_INT__k_s_s_s, NULL, PRIM_INT__i_s_p_s, NULL, PRIM_INT__k_s_d_s);
437
438
439 ostei_general_vrr1_I(8, 3,
440 one_over_2p, a_over_p, aop_PQ, P_PA,
441 PRIM_INT__k_s_s_s, PRIM_INT__i_s_s_s, PRIM_INT__l_s_s_s);
442
443
444 ostei_general_vrr_K(8, 0, 1, 0, 2,
445 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
446 PRIM_INT__l_s_s_s, NULL, NULL, PRIM_INT__k_s_s_s, NULL, PRIM_INT__l_s_p_s);
447
448
449 ostei_general_vrr_K(8, 0, 2, 0, 1,
450 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
451 PRIM_INT__l_s_p_s, PRIM_INT__l_s_s_s, NULL, PRIM_INT__k_s_p_s, NULL, PRIM_INT__l_s_d_s);
452
453
454
455
456 ////////////////////////////////////
457 // Accumulate contracted integrals
458 ////////////////////////////////////
459 if(lastoffset == 0)
460 {
461 contract_all(90, PRIM_INT__g_s_d_s, PRIM_PTR_INT__g_s_d_s);
462 contract_all(126, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s);
463 contract_all(168, PRIM_INT__i_s_d_s, PRIM_PTR_INT__i_s_d_s);
464 contract_all(216, PRIM_INT__k_s_d_s, PRIM_PTR_INT__k_s_d_s);
465 contract_all(270, PRIM_INT__l_s_d_s, PRIM_PTR_INT__l_s_d_s);
466 }
467 else
468 {
469 contract(90, shelloffsets, PRIM_INT__g_s_d_s, PRIM_PTR_INT__g_s_d_s);
470 contract(126, shelloffsets, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s);
471 contract(168, shelloffsets, PRIM_INT__i_s_d_s, PRIM_PTR_INT__i_s_d_s);
472 contract(216, shelloffsets, PRIM_INT__k_s_d_s, PRIM_PTR_INT__k_s_d_s);
473 contract(270, shelloffsets, PRIM_INT__l_s_d_s, PRIM_PTR_INT__l_s_d_s);
474 PRIM_PTR_INT__g_s_d_s += lastoffset*90;
475 PRIM_PTR_INT__h_s_d_s += lastoffset*126;
476 PRIM_PTR_INT__i_s_d_s += lastoffset*168;
477 PRIM_PTR_INT__k_s_d_s += lastoffset*216;
478 PRIM_PTR_INT__l_s_d_s += lastoffset*270;
479 }
480
481 } // close loop over j
482 } // close loop over i
483
484 //Advance to the next batch
485 jstart = SIMINT_SIMD_ROUND(jend);
486
487 //////////////////////////////////////////////
488 // Contracted integrals: Horizontal recurrance
489 //////////////////////////////////////////////
490
491
492 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
493
494
495 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
496 {
497
498 // set up HRR pointers
499 double const * restrict HRR_INT__g_s_d_s = INT__g_s_d_s + abcd * 90;
500 double const * restrict HRR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126;
501 double const * restrict HRR_INT__i_s_d_s = INT__i_s_d_s + abcd * 168;
502 double const * restrict HRR_INT__k_s_d_s = INT__k_s_d_s + abcd * 216;
503 double const * restrict HRR_INT__l_s_d_s = INT__l_s_d_s + abcd * 270;
504 double * restrict HRR_INT__g_g_d_s = INT__g_g_d_s + real_abcd * 1350;
505
506 // form INT__g_p_d_s
507 HRR_J_g_p(
508 HRR_INT__g_p_d_s,
509 HRR_INT__g_s_d_s,
510 HRR_INT__h_s_d_s,
511 hAB, 6);
512
513 // form INT__h_p_d_s
514 ostei_general_hrr_J(5, 1, 2, 0, hAB, HRR_INT__i_s_d_s, HRR_INT__h_s_d_s, HRR_INT__h_p_d_s);
515
516 // form INT__i_p_d_s
517 ostei_general_hrr_J(6, 1, 2, 0, hAB, HRR_INT__k_s_d_s, HRR_INT__i_s_d_s, HRR_INT__i_p_d_s);
518
519 // form INT__k_p_d_s
520 ostei_general_hrr_J(7, 1, 2, 0, hAB, HRR_INT__l_s_d_s, HRR_INT__k_s_d_s, HRR_INT__k_p_d_s);
521
522 // form INT__g_d_d_s
523 ostei_general_hrr_J(4, 2, 2, 0, hAB, HRR_INT__h_p_d_s, HRR_INT__g_p_d_s, HRR_INT__g_d_d_s);
524
525 // form INT__h_d_d_s
526 ostei_general_hrr_J(5, 2, 2, 0, hAB, HRR_INT__i_p_d_s, HRR_INT__h_p_d_s, HRR_INT__h_d_d_s);
527
528 // form INT__i_d_d_s
529 ostei_general_hrr_J(6, 2, 2, 0, hAB, HRR_INT__k_p_d_s, HRR_INT__i_p_d_s, HRR_INT__i_d_d_s);
530
531 // form INT__g_f_d_s
532 ostei_general_hrr_J(4, 3, 2, 0, hAB, HRR_INT__h_d_d_s, HRR_INT__g_d_d_s, HRR_INT__g_f_d_s);
533
534 // form INT__h_f_d_s
535 ostei_general_hrr_J(5, 3, 2, 0, hAB, HRR_INT__i_d_d_s, HRR_INT__h_d_d_s, HRR_INT__h_f_d_s);
536
537 // form INT__g_g_d_s
538 ostei_general_hrr_J(4, 4, 2, 0, hAB, HRR_INT__h_f_d_s, HRR_INT__g_f_d_s, HRR_INT__g_g_d_s);
539
540
541 } // close HRR loop
542
543
544 } // close loop cdbatch
545
546 istart = iend;
547 } // close loop over ab
548
549 return P.nshell12_clip * Q.nshell12_clip;
550 }
551
ostei_g_g_s_d(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__g_g_s_d)552 int ostei_g_g_s_d(struct simint_multi_shellpair const P,
553 struct simint_multi_shellpair const Q,
554 double screen_tol,
555 double * const restrict work,
556 double * const restrict INT__g_g_s_d)
557 {
558 double Q_AB[3*Q.nshell12];
559 struct simint_multi_shellpair Q_tmp = Q;
560 Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z;
561 Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z;
562 Q_tmp.AB_x = Q_AB;
563 Q_tmp.AB_y = Q_AB + Q.nshell12;
564 Q_tmp.AB_z = Q_AB + 2*Q.nshell12;
565
566 for(int i = 0; i < Q.nshell12; i++)
567 {
568 Q_tmp.AB_x[i] = -Q.AB_x[i];
569 Q_tmp.AB_y[i] = -Q.AB_y[i];
570 Q_tmp.AB_z[i] = -Q.AB_z[i];
571 }
572
573 int ret = ostei_g_g_d_s(P, Q_tmp, screen_tol, work, INT__g_g_s_d);
574 double buffer[1350] SIMINT_ALIGN_ARRAY_DBL;
575
576 for(int q = 0; q < ret; q++)
577 {
578 int idx = 0;
579 for(int a = 0; a < 15; ++a)
580 for(int b = 0; b < 15; ++b)
581 for(int c = 0; c < 1; ++c)
582 for(int d = 0; d < 6; ++d)
583 buffer[idx++] = INT__g_g_s_d[q*1350+a*90+b*6+d*1+c];
584
585 memcpy(INT__g_g_s_d+q*1350, buffer, 1350*sizeof(double));
586 }
587
588 return ret;
589 }
590
591