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_d_f_s(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__s_d_f_s)8 int ostei_s_d_f_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__s_d_f_s)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__s_d_f_s);
17 memset(INT__s_d_f_s, 0, P.nshell12_clip * Q.nshell12_clip * 60 * sizeof(double));
18
19 int ab, cd, abcd;
20 int istart, jstart;
21 int iprimcd, nprim_icd, icd;
22 const int check_screen = (screen_tol > 0.0);
23 int i, j;
24 int n;
25 int not_screened;
26
27 // partition workspace
28 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*0);
29 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
30 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_p_s = primwork + 6;
31 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_d_s = primwork + 21;
32 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_f_s = primwork + 45;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__s_p_d_s = primwork + 75;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_p_f_s = primwork + 111;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_d_f_s = primwork + 171;
36 double * const hrrwork = (double *)(primwork + 231);
37
38
39 // Create constants
40 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
41 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
42 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
43 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
44
45
46 ////////////////////////////////////////
47 // Loop over shells and primitives
48 ////////////////////////////////////////
49
50 abcd = 0;
51 istart = 0;
52 for(ab = 0; ab < P.nshell12_clip; ++ab)
53 {
54 const int iend = istart + P.nprim12[ab];
55
56 cd = 0;
57 jstart = 0;
58
59 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
60 {
61 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
62 int jend = jstart;
63 for(i = 0; i < nshellbatch; i++)
64 jend += Q.nprim12[cd+i];
65
66
67 for(i = istart; i < iend; ++i)
68 {
69 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
70
71 if(check_screen)
72 {
73 // Skip this whole thing if always insignificant
74 if((P.screen[i] * Q.screen_max) < screen_tol)
75 continue;
76 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
77 }
78
79 icd = 0;
80 iprimcd = 0;
81 nprim_icd = Q.nprim12[cd];
82 double * restrict PRIM_PTR_INT__s_d_f_s = INT__s_d_f_s + abcd * 60;
83
84
85
86 // Load these one per loop over i
87 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
88 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
89 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
90
91 const SIMINT_DBLTYPE P_PB[3] = { SIMINT_DBLSET1(P.PB_x[i]), SIMINT_DBLSET1(P.PB_y[i]), SIMINT_DBLSET1(P.PB_z[i]) };
92
93 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
94 {
95 // calculate the shell offsets
96 // these are the offset from the shell pointed to by cd
97 // for each element
98 int shelloffsets[SIMINT_SIMD_LEN] = {0};
99 int lastoffset = 0;
100 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
101
102 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
103 {
104 // Handle if the first element of the vector is a new shell
105 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
106 {
107 nprim_icd += Q.nprim12[cd + (++icd)];
108 PRIM_PTR_INT__s_d_f_s += 60;
109 }
110 iprimcd++;
111 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
112 {
113 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
114 {
115 shelloffsets[n] = shelloffsets[n-1] + 1;
116 lastoffset++;
117 nprim_icd += Q.nprim12[cd + (++icd)];
118 }
119 else
120 shelloffsets[n] = shelloffsets[n-1];
121 iprimcd++;
122 }
123 }
124 else
125 iprimcd += SIMINT_SIMD_LEN;
126
127 // Do we have to compute this vector (or has it been screened out)?
128 // (not_screened != 0 means we have to do this vector)
129 if(check_screen)
130 {
131 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
132 if(vmax < screen_tol)
133 {
134 PRIM_PTR_INT__s_d_f_s += lastoffset*60;
135 continue;
136 }
137 }
138
139 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
140 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
141 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
142 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
143
144
145 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
146 SIMINT_DBLTYPE PQ[3];
147 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
148 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
149 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
150 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
151 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
152 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
153
154 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
155 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
156 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
157 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
158 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
159 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
160 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
161
162 // NOTE: Minus sign!
163 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
164 SIMINT_DBLTYPE aop_PQ[3];
165 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
166 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
167 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
168
169 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
170 SIMINT_DBLTYPE aoq_PQ[3];
171 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
172 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
173 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
174 // Put a minus sign here so we don't have to in RR routines
175 a_over_q = SIMINT_NEG(a_over_q);
176
177
178 //////////////////////////////////////////////
179 // Fjt function section
180 // Maximum v value: 5
181 //////////////////////////////////////////////
182 // The parameter to the Fjt function
183 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
184
185
186 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
187
188
189 boys_F_split(PRIM_INT__s_s_s_s, F_x, 5);
190 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
191 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
192 for(n = 0; n <= 5; n++)
193 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
194
195 //////////////////////////////////////////////
196 // Primitive integrals: Vertical recurrance
197 //////////////////////////////////////////////
198
199 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
200 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
201 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
202 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
203 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
204 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
205
206
207
208 // Forming PRIM_INT__s_s_p_s[5 * 3];
209 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
210 {
211
212 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
213 PRIM_INT__s_s_p_s[n * 3 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]);
214
215 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
216 PRIM_INT__s_s_p_s[n * 3 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 1]);
217
218 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
219 PRIM_INT__s_s_p_s[n * 3 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_p_s[n * 3 + 2]);
220
221 }
222
223
224
225 // Forming PRIM_INT__s_s_d_s[4 * 6];
226 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
227 {
228
229 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_p_s[n * 3 + 0]);
230 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 0]);
231 PRIM_INT__s_s_d_s[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 0]);
232
233 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 0]);
234 PRIM_INT__s_s_d_s[n * 6 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 1]);
235
236 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 0]);
237 PRIM_INT__s_s_d_s[n * 6 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_d_s[n * 6 + 2]);
238
239 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_p_s[n * 3 + 1]);
240 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 3]);
241 PRIM_INT__s_s_d_s[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 3]);
242
243 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 1]);
244 PRIM_INT__s_s_d_s[n * 6 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_d_s[n * 6 + 4]);
245
246 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_p_s[n * 3 + 2]);
247 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_d_s[n * 6 + 5]);
248 PRIM_INT__s_s_d_s[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_d_s[n * 6 + 5]);
249
250 }
251
252
253
254 // Forming PRIM_INT__s_s_f_s[3 * 10];
255 for(n = 0; n < 3; ++n) // loop over orders of auxiliary function
256 {
257
258 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
259 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 0]);
260 PRIM_INT__s_s_f_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_s_p_s[n * 3 + 0]), PRIM_INT__s_s_f_s[n * 10 + 0]);
261
262 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
263 PRIM_INT__s_s_f_s[n * 10 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 1]);
264
265 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
266 PRIM_INT__s_s_f_s[n * 10 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_s_f_s[n * 10 + 2]);
267
268 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
269 PRIM_INT__s_s_f_s[n * 10 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 3]);
270
271 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
272 PRIM_INT__s_s_f_s[n * 10 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_s_f_s[n * 10 + 4]);
273
274 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_MUL(Q_PA[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
275 PRIM_INT__s_s_f_s[n * 10 + 5] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 5]);
276
277 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
278 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 6]);
279 PRIM_INT__s_s_f_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_s_p_s[n * 3 + 1]), PRIM_INT__s_s_f_s[n * 10 + 6]);
280
281 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
282 PRIM_INT__s_s_f_s[n * 10 + 7] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_s_f_s[n * 10 + 7]);
283
284 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_MUL(Q_PA[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
285 PRIM_INT__s_s_f_s[n * 10 + 8] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 8]);
286
287 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_MUL(Q_PA[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
288 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_s_f_s[n * 10 + 9]);
289 PRIM_INT__s_s_f_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_s_p_s[n * 3 + 2]), PRIM_INT__s_s_f_s[n * 10 + 9]);
290
291 }
292
293
294 VRR_J_s_p_f_s(
295 PRIM_INT__s_p_f_s,
296 PRIM_INT__s_s_f_s,
297 PRIM_INT__s_s_d_s,
298 P_PB,
299 aop_PQ,
300 one_over_2pq,
301 2);
302
303
304
305 // Forming PRIM_INT__s_p_d_s[2 * 18];
306 for(n = 0; n < 2; ++n) // loop over orders of auxiliary function
307 {
308
309 PRIM_INT__s_p_d_s[n * 18 + 0] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 0]);
310 PRIM_INT__s_p_d_s[n * 18 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_p_d_s[n * 18 + 0]);
311 PRIM_INT__s_p_d_s[n * 18 + 0] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_p_d_s[n * 18 + 0]);
312
313 PRIM_INT__s_p_d_s[n * 18 + 1] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 1]);
314 PRIM_INT__s_p_d_s[n * 18 + 1] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_p_d_s[n * 18 + 1]);
315 PRIM_INT__s_p_d_s[n * 18 + 1] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_p_d_s[n * 18 + 1]);
316
317 PRIM_INT__s_p_d_s[n * 18 + 2] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 2]);
318 PRIM_INT__s_p_d_s[n * 18 + 2] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__s_p_d_s[n * 18 + 2]);
319 PRIM_INT__s_p_d_s[n * 18 + 2] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_p_d_s[n * 18 + 2]);
320
321 PRIM_INT__s_p_d_s[n * 18 + 3] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 3]);
322 PRIM_INT__s_p_d_s[n * 18 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_p_d_s[n * 18 + 3]);
323
324 PRIM_INT__s_p_d_s[n * 18 + 4] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 4]);
325 PRIM_INT__s_p_d_s[n * 18 + 4] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__s_p_d_s[n * 18 + 4]);
326
327 PRIM_INT__s_p_d_s[n * 18 + 5] = SIMINT_MUL(P_PB[0], PRIM_INT__s_s_d_s[n * 6 + 5]);
328 PRIM_INT__s_p_d_s[n * 18 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_p_d_s[n * 18 + 5]);
329
330 PRIM_INT__s_p_d_s[n * 18 + 6] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 0]);
331 PRIM_INT__s_p_d_s[n * 18 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_p_d_s[n * 18 + 6]);
332
333 PRIM_INT__s_p_d_s[n * 18 + 7] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 1]);
334 PRIM_INT__s_p_d_s[n * 18 + 7] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_p_d_s[n * 18 + 7]);
335 PRIM_INT__s_p_d_s[n * 18 + 7] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_p_d_s[n * 18 + 7]);
336
337 PRIM_INT__s_p_d_s[n * 18 + 8] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 2]);
338 PRIM_INT__s_p_d_s[n * 18 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__s_p_d_s[n * 18 + 8]);
339
340 PRIM_INT__s_p_d_s[n * 18 + 9] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 3]);
341 PRIM_INT__s_p_d_s[n * 18 + 9] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_p_d_s[n * 18 + 9]);
342 PRIM_INT__s_p_d_s[n * 18 + 9] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_p_d_s[n * 18 + 9]);
343
344 PRIM_INT__s_p_d_s[n * 18 + 10] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 4]);
345 PRIM_INT__s_p_d_s[n * 18 + 10] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__s_p_d_s[n * 18 + 10]);
346 PRIM_INT__s_p_d_s[n * 18 + 10] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_p_d_s[n * 18 + 10]);
347
348 PRIM_INT__s_p_d_s[n * 18 + 11] = SIMINT_MUL(P_PB[1], PRIM_INT__s_s_d_s[n * 6 + 5]);
349 PRIM_INT__s_p_d_s[n * 18 + 11] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_p_d_s[n * 18 + 11]);
350
351 PRIM_INT__s_p_d_s[n * 18 + 12] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 0]);
352 PRIM_INT__s_p_d_s[n * 18 + 12] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 0], PRIM_INT__s_p_d_s[n * 18 + 12]);
353
354 PRIM_INT__s_p_d_s[n * 18 + 13] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 1]);
355 PRIM_INT__s_p_d_s[n * 18 + 13] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 1], PRIM_INT__s_p_d_s[n * 18 + 13]);
356
357 PRIM_INT__s_p_d_s[n * 18 + 14] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 2]);
358 PRIM_INT__s_p_d_s[n * 18 + 14] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 2], PRIM_INT__s_p_d_s[n * 18 + 14]);
359 PRIM_INT__s_p_d_s[n * 18 + 14] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 0], PRIM_INT__s_p_d_s[n * 18 + 14]);
360
361 PRIM_INT__s_p_d_s[n * 18 + 15] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 3]);
362 PRIM_INT__s_p_d_s[n * 18 + 15] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 3], PRIM_INT__s_p_d_s[n * 18 + 15]);
363
364 PRIM_INT__s_p_d_s[n * 18 + 16] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 4]);
365 PRIM_INT__s_p_d_s[n * 18 + 16] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 4], PRIM_INT__s_p_d_s[n * 18 + 16]);
366 PRIM_INT__s_p_d_s[n * 18 + 16] = SIMINT_FMADD( vrr_const_1_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 1], PRIM_INT__s_p_d_s[n * 18 + 16]);
367
368 PRIM_INT__s_p_d_s[n * 18 + 17] = SIMINT_MUL(P_PB[2], PRIM_INT__s_s_d_s[n * 6 + 5]);
369 PRIM_INT__s_p_d_s[n * 18 + 17] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_d_s[(n+1) * 6 + 5], PRIM_INT__s_p_d_s[n * 18 + 17]);
370 PRIM_INT__s_p_d_s[n * 18 + 17] = SIMINT_FMADD( vrr_const_2_over_2pq, PRIM_INT__s_s_p_s[(n+1) * 3 + 2], PRIM_INT__s_p_d_s[n * 18 + 17]);
371
372 }
373
374
375 VRR_J_s_d_f_s(
376 PRIM_INT__s_d_f_s,
377 PRIM_INT__s_p_f_s,
378 PRIM_INT__s_s_f_s,
379 PRIM_INT__s_p_d_s,
380 P_PB,
381 a_over_p,
382 aop_PQ,
383 one_over_2p,
384 one_over_2pq,
385 1);
386
387
388
389
390 ////////////////////////////////////
391 // Accumulate contracted integrals
392 ////////////////////////////////////
393 if(lastoffset == 0)
394 {
395 contract_all(60, PRIM_INT__s_d_f_s, PRIM_PTR_INT__s_d_f_s);
396 }
397 else
398 {
399 contract(60, shelloffsets, PRIM_INT__s_d_f_s, PRIM_PTR_INT__s_d_f_s);
400 PRIM_PTR_INT__s_d_f_s += lastoffset*60;
401 }
402
403 } // close loop over j
404 } // close loop over i
405
406 //Advance to the next batch
407 jstart = SIMINT_SIMD_ROUND(jend);
408 abcd += nshellbatch;
409
410 } // close loop cdbatch
411
412 istart = iend;
413 } // close loop over ab
414
415 return P.nshell12_clip * Q.nshell12_clip;
416 }
417
418