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_k_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__k_f_p_p)8 int ostei_k_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__k_f_p_p)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__k_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__k_s_p_s = work + (SIMINT_NSHELL_SIMD * 0);
30 double * const INT__k_s_d_s = work + (SIMINT_NSHELL_SIMD * 108);
31 double * const INT__l_s_p_s = work + (SIMINT_NSHELL_SIMD * 324);
32 double * const INT__l_s_d_s = work + (SIMINT_NSHELL_SIMD * 459);
33 double * const INT__m_s_p_s = work + (SIMINT_NSHELL_SIMD * 729);
34 double * const INT__m_s_d_s = work + (SIMINT_NSHELL_SIMD * 894);
35 double * const INT__n_s_p_s = work + (SIMINT_NSHELL_SIMD * 1224);
36 double * const INT__n_s_d_s = work + (SIMINT_NSHELL_SIMD * 1422);
37 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*1818);
38 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 13;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 49;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 115;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 215;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 350;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_s_s = primwork + 518;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_p_s = primwork + 714;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_s_s = primwork + 882;
47 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_p_s = primwork + 1098;
48 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_d_s = primwork + 1314;
49 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_s_s = primwork + 1530;
50 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_p_s = primwork + 1755;
51 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_d_s = primwork + 2025;
52 SIMINT_DBLTYPE * const restrict PRIM_INT__m_s_s_s = primwork + 2295;
53 SIMINT_DBLTYPE * const restrict PRIM_INT__m_s_p_s = primwork + 2515;
54 SIMINT_DBLTYPE * const restrict PRIM_INT__m_s_d_s = primwork + 2845;
55 SIMINT_DBLTYPE * const restrict PRIM_INT__n_s_s_s = primwork + 3175;
56 SIMINT_DBLTYPE * const restrict PRIM_INT__n_s_p_s = primwork + 3373;
57 SIMINT_DBLTYPE * const restrict PRIM_INT__n_s_d_s = primwork + 3769;
58 double * const hrrwork = (double *)(primwork + 4165);
59 double * const HRR_INT__k_p_p_s = hrrwork + 0;
60 double * const HRR_INT__k_p_d_s = hrrwork + 324;
61 double * const HRR_INT__k_d_p_s = hrrwork + 972;
62 double * const HRR_INT__k_d_d_s = hrrwork + 1620;
63 double * const HRR_INT__k_f_p_s = hrrwork + 2916;
64 double * const HRR_INT__k_f_d_s = hrrwork + 3996;
65 double * const HRR_INT__l_p_p_s = hrrwork + 6156;
66 double * const HRR_INT__l_p_d_s = hrrwork + 6561;
67 double * const HRR_INT__l_d_p_s = hrrwork + 7371;
68 double * const HRR_INT__l_d_d_s = hrrwork + 8181;
69 double * const HRR_INT__m_p_p_s = hrrwork + 9801;
70 double * const HRR_INT__m_p_d_s = hrrwork + 10296;
71
72
73 // Create constants
74 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
75 const SIMINT_DBLTYPE const_10 = SIMINT_DBLSET1(10);
76 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
77 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
78 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
79 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
80 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
81 const SIMINT_DBLTYPE const_7 = SIMINT_DBLSET1(7);
82 const SIMINT_DBLTYPE const_8 = SIMINT_DBLSET1(8);
83 const SIMINT_DBLTYPE const_9 = SIMINT_DBLSET1(9);
84 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
85
86
87 ////////////////////////////////////////
88 // Loop over shells and primitives
89 ////////////////////////////////////////
90
91 real_abcd = 0;
92 istart = 0;
93 for(ab = 0; ab < P.nshell12_clip; ++ab)
94 {
95 const int iend = istart + P.nprim12[ab];
96
97 cd = 0;
98 jstart = 0;
99
100 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
101 {
102 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
103 int jend = jstart;
104 for(i = 0; i < nshellbatch; i++)
105 jend += Q.nprim12[cd+i];
106
107 // Clear the beginning of the workspace (where we are accumulating integrals)
108 memset(work, 0, SIMINT_NSHELL_SIMD * 1818 * sizeof(double));
109 abcd = 0;
110
111
112 for(i = istart; i < iend; ++i)
113 {
114 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
115
116 if(check_screen)
117 {
118 // Skip this whole thing if always insignificant
119 if((P.screen[i] * Q.screen_max) < screen_tol)
120 continue;
121 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
122 }
123
124 icd = 0;
125 iprimcd = 0;
126 nprim_icd = Q.nprim12[cd];
127 double * restrict PRIM_PTR_INT__k_s_p_s = INT__k_s_p_s + abcd * 108;
128 double * restrict PRIM_PTR_INT__k_s_d_s = INT__k_s_d_s + abcd * 216;
129 double * restrict PRIM_PTR_INT__l_s_p_s = INT__l_s_p_s + abcd * 135;
130 double * restrict PRIM_PTR_INT__l_s_d_s = INT__l_s_d_s + abcd * 270;
131 double * restrict PRIM_PTR_INT__m_s_p_s = INT__m_s_p_s + abcd * 165;
132 double * restrict PRIM_PTR_INT__m_s_d_s = INT__m_s_d_s + abcd * 330;
133 double * restrict PRIM_PTR_INT__n_s_p_s = INT__n_s_p_s + abcd * 198;
134 double * restrict PRIM_PTR_INT__n_s_d_s = INT__n_s_d_s + abcd * 396;
135
136
137
138 // Load these one per loop over i
139 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
140 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
141 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
142
143 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
144
145 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
146 {
147 // calculate the shell offsets
148 // these are the offset from the shell pointed to by cd
149 // for each element
150 int shelloffsets[SIMINT_SIMD_LEN] = {0};
151 int lastoffset = 0;
152 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
153
154 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
155 {
156 // Handle if the first element of the vector is a new shell
157 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
158 {
159 nprim_icd += Q.nprim12[cd + (++icd)];
160 PRIM_PTR_INT__k_s_p_s += 108;
161 PRIM_PTR_INT__k_s_d_s += 216;
162 PRIM_PTR_INT__l_s_p_s += 135;
163 PRIM_PTR_INT__l_s_d_s += 270;
164 PRIM_PTR_INT__m_s_p_s += 165;
165 PRIM_PTR_INT__m_s_d_s += 330;
166 PRIM_PTR_INT__n_s_p_s += 198;
167 PRIM_PTR_INT__n_s_d_s += 396;
168 }
169 iprimcd++;
170 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
171 {
172 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
173 {
174 shelloffsets[n] = shelloffsets[n-1] + 1;
175 lastoffset++;
176 nprim_icd += Q.nprim12[cd + (++icd)];
177 }
178 else
179 shelloffsets[n] = shelloffsets[n-1];
180 iprimcd++;
181 }
182 }
183 else
184 iprimcd += SIMINT_SIMD_LEN;
185
186 // Do we have to compute this vector (or has it been screened out)?
187 // (not_screened != 0 means we have to do this vector)
188 if(check_screen)
189 {
190 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
191 if(vmax < screen_tol)
192 {
193 PRIM_PTR_INT__k_s_p_s += lastoffset*108;
194 PRIM_PTR_INT__k_s_d_s += lastoffset*216;
195 PRIM_PTR_INT__l_s_p_s += lastoffset*135;
196 PRIM_PTR_INT__l_s_d_s += lastoffset*270;
197 PRIM_PTR_INT__m_s_p_s += lastoffset*165;
198 PRIM_PTR_INT__m_s_d_s += lastoffset*330;
199 PRIM_PTR_INT__n_s_p_s += lastoffset*198;
200 PRIM_PTR_INT__n_s_d_s += lastoffset*396;
201 continue;
202 }
203 }
204
205 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
206 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
207 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
208 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
209
210
211 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
212 SIMINT_DBLTYPE PQ[3];
213 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
214 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
215 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
216 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
217 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
218 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
219
220 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
221 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
222 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
223 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
224 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
225 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
226 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
227
228 // NOTE: Minus sign!
229 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
230 SIMINT_DBLTYPE aop_PQ[3];
231 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
232 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
233 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
234
235 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
236 SIMINT_DBLTYPE aoq_PQ[3];
237 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
238 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
239 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
240 // Put a minus sign here so we don't have to in RR routines
241 a_over_q = SIMINT_NEG(a_over_q);
242
243
244 //////////////////////////////////////////////
245 // Fjt function section
246 // Maximum v value: 12
247 //////////////////////////////////////////////
248 // The parameter to the Fjt function
249 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
250
251
252 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
253
254
255 boys_F_split(PRIM_INT__s_s_s_s, F_x, 12);
256 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
257 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
258 for(n = 0; n <= 12; n++)
259 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
260
261 //////////////////////////////////////////////
262 // Primitive integrals: Vertical recurrance
263 //////////////////////////////////////////////
264
265 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
266 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
267 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
268 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
269 const SIMINT_DBLTYPE vrr_const_5_over_2p = SIMINT_MUL(const_5, one_over_2p);
270 const SIMINT_DBLTYPE vrr_const_6_over_2p = SIMINT_MUL(const_6, one_over_2p);
271 const SIMINT_DBLTYPE vrr_const_7_over_2p = SIMINT_MUL(const_7, one_over_2p);
272 const SIMINT_DBLTYPE vrr_const_8_over_2p = SIMINT_MUL(const_8, one_over_2p);
273 const SIMINT_DBLTYPE vrr_const_9_over_2p = SIMINT_MUL(const_9, one_over_2p);
274 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
275 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
276 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
277 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
278 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
279 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
280 const SIMINT_DBLTYPE vrr_const_6_over_2pq = SIMINT_MUL(const_6, one_over_2pq);
281 const SIMINT_DBLTYPE vrr_const_7_over_2pq = SIMINT_MUL(const_7, one_over_2pq);
282 const SIMINT_DBLTYPE vrr_const_8_over_2pq = SIMINT_MUL(const_8, one_over_2pq);
283 const SIMINT_DBLTYPE vrr_const_9_over_2pq = SIMINT_MUL(const_9, one_over_2pq);
284 const SIMINT_DBLTYPE vrr_const_10_over_2pq = SIMINT_MUL(const_10, one_over_2pq);
285
286
287
288 // Forming PRIM_INT__p_s_s_s[12 * 3];
289 for(n = 0; n < 12; ++n) // loop over orders of auxiliary function
290 {
291
292 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
293 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]);
294
295 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
296 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]);
297
298 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
299 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]);
300
301 }
302
303
304
305 // Forming PRIM_INT__d_s_s_s[11 * 6];
306 for(n = 0; n < 11; ++n) // loop over orders of auxiliary function
307 {
308
309 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
310 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]);
311 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]);
312
313 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
314 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]);
315 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]);
316
317 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
318 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]);
319 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]);
320
321 }
322
323
324
325 // Forming PRIM_INT__f_s_s_s[10 * 10];
326 for(n = 0; n < 10; ++n) // loop over orders of auxiliary function
327 {
328
329 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
330 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]);
331 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]);
332
333 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
334 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]);
335
336 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]);
337 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]);
338
339 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
340 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]);
341 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]);
342
343 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]);
344 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]);
345
346 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
347 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]);
348 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]);
349
350 }
351
352
353 VRR_I_g_s_s_s(
354 PRIM_INT__g_s_s_s,
355 PRIM_INT__f_s_s_s,
356 PRIM_INT__d_s_s_s,
357 P_PA,
358 a_over_p,
359 aop_PQ,
360 one_over_2p,
361 9);
362
363
364 VRR_I_h_s_s_s(
365 PRIM_INT__h_s_s_s,
366 PRIM_INT__g_s_s_s,
367 PRIM_INT__f_s_s_s,
368 P_PA,
369 a_over_p,
370 aop_PQ,
371 one_over_2p,
372 8);
373
374
375 ostei_general_vrr1_I(6, 7,
376 one_over_2p, a_over_p, aop_PQ, P_PA,
377 PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__i_s_s_s);
378
379
380 ostei_general_vrr1_I(7, 6,
381 one_over_2p, a_over_p, aop_PQ, P_PA,
382 PRIM_INT__i_s_s_s, PRIM_INT__h_s_s_s, PRIM_INT__k_s_s_s);
383
384
385 ostei_general_vrr_K(7, 0, 1, 0, 2,
386 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
387 PRIM_INT__k_s_s_s, NULL, NULL, PRIM_INT__i_s_s_s, NULL, PRIM_INT__k_s_p_s);
388
389
390 ostei_general_vrr_K(6, 0, 1, 0, 2,
391 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
392 PRIM_INT__i_s_s_s, NULL, NULL, PRIM_INT__h_s_s_s, NULL, PRIM_INT__i_s_p_s);
393
394
395 ostei_general_vrr_K(7, 0, 2, 0, 1,
396 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
397 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);
398
399
400 ostei_general_vrr1_I(8, 5,
401 one_over_2p, a_over_p, aop_PQ, P_PA,
402 PRIM_INT__k_s_s_s, PRIM_INT__i_s_s_s, PRIM_INT__l_s_s_s);
403
404
405 ostei_general_vrr_K(8, 0, 1, 0, 2,
406 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
407 PRIM_INT__l_s_s_s, NULL, NULL, PRIM_INT__k_s_s_s, NULL, PRIM_INT__l_s_p_s);
408
409
410 ostei_general_vrr_K(8, 0, 2, 0, 1,
411 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
412 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);
413
414
415 ostei_general_vrr1_I(9, 4,
416 one_over_2p, a_over_p, aop_PQ, P_PA,
417 PRIM_INT__l_s_s_s, PRIM_INT__k_s_s_s, PRIM_INT__m_s_s_s);
418
419
420 ostei_general_vrr_K(9, 0, 1, 0, 2,
421 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
422 PRIM_INT__m_s_s_s, NULL, NULL, PRIM_INT__l_s_s_s, NULL, PRIM_INT__m_s_p_s);
423
424
425 ostei_general_vrr_K(9, 0, 2, 0, 1,
426 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
427 PRIM_INT__m_s_p_s, PRIM_INT__m_s_s_s, NULL, PRIM_INT__l_s_p_s, NULL, PRIM_INT__m_s_d_s);
428
429
430 ostei_general_vrr1_I(10, 3,
431 one_over_2p, a_over_p, aop_PQ, P_PA,
432 PRIM_INT__m_s_s_s, PRIM_INT__l_s_s_s, PRIM_INT__n_s_s_s);
433
434
435 ostei_general_vrr_K(10, 0, 1, 0, 2,
436 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
437 PRIM_INT__n_s_s_s, NULL, NULL, PRIM_INT__m_s_s_s, NULL, PRIM_INT__n_s_p_s);
438
439
440 ostei_general_vrr_K(10, 0, 2, 0, 1,
441 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
442 PRIM_INT__n_s_p_s, PRIM_INT__n_s_s_s, NULL, PRIM_INT__m_s_p_s, NULL, PRIM_INT__n_s_d_s);
443
444
445
446
447 ////////////////////////////////////
448 // Accumulate contracted integrals
449 ////////////////////////////////////
450 if(lastoffset == 0)
451 {
452 contract_all(108, PRIM_INT__k_s_p_s, PRIM_PTR_INT__k_s_p_s);
453 contract_all(216, PRIM_INT__k_s_d_s, PRIM_PTR_INT__k_s_d_s);
454 contract_all(135, PRIM_INT__l_s_p_s, PRIM_PTR_INT__l_s_p_s);
455 contract_all(270, PRIM_INT__l_s_d_s, PRIM_PTR_INT__l_s_d_s);
456 contract_all(165, PRIM_INT__m_s_p_s, PRIM_PTR_INT__m_s_p_s);
457 contract_all(330, PRIM_INT__m_s_d_s, PRIM_PTR_INT__m_s_d_s);
458 contract_all(198, PRIM_INT__n_s_p_s, PRIM_PTR_INT__n_s_p_s);
459 contract_all(396, PRIM_INT__n_s_d_s, PRIM_PTR_INT__n_s_d_s);
460 }
461 else
462 {
463 contract(108, shelloffsets, PRIM_INT__k_s_p_s, PRIM_PTR_INT__k_s_p_s);
464 contract(216, shelloffsets, PRIM_INT__k_s_d_s, PRIM_PTR_INT__k_s_d_s);
465 contract(135, shelloffsets, PRIM_INT__l_s_p_s, PRIM_PTR_INT__l_s_p_s);
466 contract(270, shelloffsets, PRIM_INT__l_s_d_s, PRIM_PTR_INT__l_s_d_s);
467 contract(165, shelloffsets, PRIM_INT__m_s_p_s, PRIM_PTR_INT__m_s_p_s);
468 contract(330, shelloffsets, PRIM_INT__m_s_d_s, PRIM_PTR_INT__m_s_d_s);
469 contract(198, shelloffsets, PRIM_INT__n_s_p_s, PRIM_PTR_INT__n_s_p_s);
470 contract(396, shelloffsets, PRIM_INT__n_s_d_s, PRIM_PTR_INT__n_s_d_s);
471 PRIM_PTR_INT__k_s_p_s += lastoffset*108;
472 PRIM_PTR_INT__k_s_d_s += lastoffset*216;
473 PRIM_PTR_INT__l_s_p_s += lastoffset*135;
474 PRIM_PTR_INT__l_s_d_s += lastoffset*270;
475 PRIM_PTR_INT__m_s_p_s += lastoffset*165;
476 PRIM_PTR_INT__m_s_d_s += lastoffset*330;
477 PRIM_PTR_INT__n_s_p_s += lastoffset*198;
478 PRIM_PTR_INT__n_s_d_s += lastoffset*396;
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 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
498
499 // set up HRR pointers
500 double const * restrict HRR_INT__k_s_p_s = INT__k_s_p_s + abcd * 108;
501 double const * restrict HRR_INT__k_s_d_s = INT__k_s_d_s + abcd * 216;
502 double const * restrict HRR_INT__l_s_p_s = INT__l_s_p_s + abcd * 135;
503 double const * restrict HRR_INT__l_s_d_s = INT__l_s_d_s + abcd * 270;
504 double const * restrict HRR_INT__m_s_p_s = INT__m_s_p_s + abcd * 165;
505 double const * restrict HRR_INT__m_s_d_s = INT__m_s_d_s + abcd * 330;
506 double const * restrict HRR_INT__n_s_p_s = INT__n_s_p_s + abcd * 198;
507 double const * restrict HRR_INT__n_s_d_s = INT__n_s_d_s + abcd * 396;
508 double * restrict HRR_INT__k_f_p_p = INT__k_f_p_p + real_abcd * 3240;
509
510 // form INT__k_p_p_s
511 ostei_general_hrr_J(7, 1, 1, 0, hAB, HRR_INT__l_s_p_s, HRR_INT__k_s_p_s, HRR_INT__k_p_p_s);
512
513 // form INT__k_p_d_s
514 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);
515
516 // form INT__l_p_p_s
517 ostei_general_hrr_J(8, 1, 1, 0, hAB, HRR_INT__m_s_p_s, HRR_INT__l_s_p_s, HRR_INT__l_p_p_s);
518
519 // form INT__l_p_d_s
520 ostei_general_hrr_J(8, 1, 2, 0, hAB, HRR_INT__m_s_d_s, HRR_INT__l_s_d_s, HRR_INT__l_p_d_s);
521
522 // form INT__m_p_p_s
523 ostei_general_hrr_J(9, 1, 1, 0, hAB, HRR_INT__n_s_p_s, HRR_INT__m_s_p_s, HRR_INT__m_p_p_s);
524
525 // form INT__m_p_d_s
526 ostei_general_hrr_J(9, 1, 2, 0, hAB, HRR_INT__n_s_d_s, HRR_INT__m_s_d_s, HRR_INT__m_p_d_s);
527
528 // form INT__k_d_p_s
529 ostei_general_hrr_J(7, 2, 1, 0, hAB, HRR_INT__l_p_p_s, HRR_INT__k_p_p_s, HRR_INT__k_d_p_s);
530
531 // form INT__k_d_d_s
532 ostei_general_hrr_J(7, 2, 2, 0, hAB, HRR_INT__l_p_d_s, HRR_INT__k_p_d_s, HRR_INT__k_d_d_s);
533
534 // form INT__l_d_p_s
535 ostei_general_hrr_J(8, 2, 1, 0, hAB, HRR_INT__m_p_p_s, HRR_INT__l_p_p_s, HRR_INT__l_d_p_s);
536
537 // form INT__l_d_d_s
538 ostei_general_hrr_J(8, 2, 2, 0, hAB, HRR_INT__m_p_d_s, HRR_INT__l_p_d_s, HRR_INT__l_d_d_s);
539
540 // form INT__k_f_p_s
541 ostei_general_hrr_J(7, 3, 1, 0, hAB, HRR_INT__l_d_p_s, HRR_INT__k_d_p_s, HRR_INT__k_f_p_s);
542
543 // form INT__k_f_d_s
544 ostei_general_hrr_J(7, 3, 2, 0, hAB, HRR_INT__l_d_d_s, HRR_INT__k_d_d_s, HRR_INT__k_f_d_s);
545
546 // form INT__k_f_p_p
547 for(ibra = 0; ibra < 360; ++ibra)
548 {
549 HRR_INT__k_f_p_p[ibra * 9 + 0] = HRR_INT__k_f_d_s[ibra * 6 + 0] + ( hCD[0] * HRR_INT__k_f_p_s[ibra * 3 + 0] );
550
551 HRR_INT__k_f_p_p[ibra * 9 + 1] = HRR_INT__k_f_d_s[ibra * 6 + 1] + ( hCD[1] * HRR_INT__k_f_p_s[ibra * 3 + 0] );
552
553 HRR_INT__k_f_p_p[ibra * 9 + 2] = HRR_INT__k_f_d_s[ibra * 6 + 2] + ( hCD[2] * HRR_INT__k_f_p_s[ibra * 3 + 0] );
554
555 HRR_INT__k_f_p_p[ibra * 9 + 3] = HRR_INT__k_f_d_s[ibra * 6 + 1] + ( hCD[0] * HRR_INT__k_f_p_s[ibra * 3 + 1] );
556
557 HRR_INT__k_f_p_p[ibra * 9 + 4] = HRR_INT__k_f_d_s[ibra * 6 + 3] + ( hCD[1] * HRR_INT__k_f_p_s[ibra * 3 + 1] );
558
559 HRR_INT__k_f_p_p[ibra * 9 + 5] = HRR_INT__k_f_d_s[ibra * 6 + 4] + ( hCD[2] * HRR_INT__k_f_p_s[ibra * 3 + 1] );
560
561 HRR_INT__k_f_p_p[ibra * 9 + 6] = HRR_INT__k_f_d_s[ibra * 6 + 2] + ( hCD[0] * HRR_INT__k_f_p_s[ibra * 3 + 2] );
562
563 HRR_INT__k_f_p_p[ibra * 9 + 7] = HRR_INT__k_f_d_s[ibra * 6 + 4] + ( hCD[1] * HRR_INT__k_f_p_s[ibra * 3 + 2] );
564
565 HRR_INT__k_f_p_p[ibra * 9 + 8] = HRR_INT__k_f_d_s[ibra * 6 + 5] + ( hCD[2] * HRR_INT__k_f_p_s[ibra * 3 + 2] );
566
567 }
568
569
570 } // close HRR loop
571
572
573 } // close loop cdbatch
574
575 istart = iend;
576 } // close loop over ab
577
578 return P.nshell12_clip * Q.nshell12_clip;
579 }
580
ostei_f_k_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__f_k_p_p)581 int ostei_f_k_p_p(struct simint_multi_shellpair const P,
582 struct simint_multi_shellpair const Q,
583 double screen_tol,
584 double * const restrict work,
585 double * const restrict INT__f_k_p_p)
586 {
587 double P_AB[3*P.nshell12];
588 struct simint_multi_shellpair P_tmp = P;
589 P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z;
590 P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z;
591 P_tmp.AB_x = P_AB;
592 P_tmp.AB_y = P_AB + P.nshell12;
593 P_tmp.AB_z = P_AB + 2*P.nshell12;
594
595 for(int i = 0; i < P.nshell12; i++)
596 {
597 P_tmp.AB_x[i] = -P.AB_x[i];
598 P_tmp.AB_y[i] = -P.AB_y[i];
599 P_tmp.AB_z[i] = -P.AB_z[i];
600 }
601
602 int ret = ostei_k_f_p_p(P_tmp, Q, screen_tol, work, INT__f_k_p_p);
603 double buffer[3240] SIMINT_ALIGN_ARRAY_DBL;
604
605 for(int q = 0; q < ret; q++)
606 {
607 int idx = 0;
608 for(int a = 0; a < 10; ++a)
609 for(int b = 0; b < 36; ++b)
610 for(int c = 0; c < 3; ++c)
611 for(int d = 0; d < 3; ++d)
612 buffer[idx++] = INT__f_k_p_p[q*3240+b*90+a*9+c*3+d];
613
614 memcpy(INT__f_k_p_p+q*3240, buffer, 3240*sizeof(double));
615 }
616
617 return ret;
618 }
619
620