/*
* This file is part of the GROMACS molecular simulation package.
*
- * Copyright (c) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
+ * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
* Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
* and including many others, as listed in the AUTHORS file in the
* top-level source directory and at http://www.gromacs.org.
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
/* #endif */
- /* #if 'GeneralizedBorn' in KERNEL_ELEC */
- __m128i gbitab;
- __m256d vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
- __m256d minushalf = _mm256_set1_pd(-0.5);
- real *invsqrta,*dvda,*gbtab;
- /* #endif */
/* #if KERNEL_VDW != 'None' */
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
/* #endif */
- /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
+ /* #if 'Table' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
__m128i vfitab;
__m128i ifour = _mm_set1_epi32(4);
__m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
/* #endif */
/* #endif */
- /* #if KERNEL_ELEC=='GeneralizedBorn' */
- invsqrta = fr->invsqrta;
- dvda = fr->dvda;
- gbtabscale = _mm256_set1_pd(fr->gbtab->scale);
- gbtab = fr->gbtab->data;
- gbinvepsdiff = _mm256_set1_pd((1.0/fr->ic->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
- /* #endif */
-
/* #if 'Water' in GEOMETRY_I */
/* Setup water-specific parameters */
inr = nlist->iinr[0];
/* #for I in PARTICLES_ELEC_I */
iq{I} = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+{I}]));
/* #define OUTERFLOPS OUTERFLOPS+1 */
- /* #if KERNEL_ELEC=='GeneralizedBorn' */
- isai{I} = _mm256_set1_pd(invsqrta[inr+{I}]);
- /* #endif */
/* #endfor */
/* #for I in PARTICLES_VDW_I */
vdwioffsetptr{I} = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
/* #if KERNEL_ELEC != 'None' */
velecsum = _mm256_setzero_pd();
/* #endif */
- /* #if 'GeneralizedBorn' in KERNEL_ELEC */
- vgbsum = _mm256_setzero_pd();
- /* #endif */
/* #if KERNEL_VDW != 'None' */
vvdwsum = _mm256_setzero_pd();
/* #endif */
/* #endif */
- /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
- dvdasum = _mm256_setzero_pd();
- /* #endif */
/* #for ROUND in ['Loop','Epilogue'] */
/* #for J in PARTICLES_ELEC_J */
jq{J} = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+{J},charge+jnrB+{J},
charge+jnrC+{J},charge+jnrD+{J});
- /* #if KERNEL_ELEC=='GeneralizedBorn' */
- isaj{J} = gmx_mm256_load_4real_swizzle_pd(invsqrta+jnrA+{J},invsqrta+jnrB+{J},
- invsqrta+jnrC+{J},invsqrta+jnrD+{J});
- /* #endif */
/* #endfor */
/* #for J in PARTICLES_VDW_J */
vdwjidx{J}A = 2*vdwtype[jnrA+{J}];
/* #define INNERFLOPS INNERFLOPS+3 */
/* #endif */
- /* #elif KERNEL_ELEC=='GeneralizedBorn' */
-
- /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
- isaprod = _mm256_mul_pd(isai{I},isaj{J});
- gbqqfactor = _mm256_xor_pd(signbit,_mm256_mul_pd(qq{I}{J},_mm256_mul_pd(isaprod,gbinvepsdiff)));
- gbscale = _mm256_mul_pd(isaprod,gbtabscale);
- /* #define INNERFLOPS INNERFLOPS+5 */
-
- /* Calculate generalized born table index - this is a separate table from the normal one,
- * but we use the same procedure by multiplying r with scale and truncating to integer.
- */
- rt = _mm256_mul_pd(r{I}{J},gbscale);
- gbitab = _mm256_cvttpd_epi32(rt);
- gbeps = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
- gbitab = _mm_slli_epi32(gbitab,2);
- Y = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
- F = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,1) );
- G = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,2) );
- H = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,3) );
- GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
- Heps = _mm256_mul_pd(gbeps,H);
- Fp = _mm256_add_pd(F,_mm256_mul_pd(gbeps,_mm256_add_pd(G,Heps)));
- VV = _mm256_add_pd(Y,_mm256_mul_pd(gbeps,Fp));
- vgb = _mm256_mul_pd(gbqqfactor,VV);
- /* #define INNERFLOPS INNERFLOPS+10 */
-
- /* #if 'Force' in KERNEL_VF */
- FF = _mm256_add_pd(Fp,_mm256_mul_pd(gbeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
- fgb = _mm256_mul_pd(gbqqfactor,_mm256_mul_pd(FF,gbscale));
- dvdatmp = _mm256_mul_pd(minushalf,_mm256_add_pd(vgb,_mm256_mul_pd(fgb,r{I}{J})));
- /* #if ROUND == 'Epilogue' */
- dvdatmp = _mm256_andnot_pd(dummy_mask,dvdatmp);
- /* #endif */
- dvdasum = _mm256_add_pd(dvdasum,dvdatmp);
- /* #if ROUND == 'Loop' */
- fjptrA = dvda+jnrA;
- fjptrB = dvda+jnrB;
- fjptrC = dvda+jnrC;
- fjptrD = dvda+jnrD;
- /* #else */
- /* The pointers to scratch make sure that this code with compilers that take gmx_restrict seriously (e.g. icc 13) really can't screw things up. */
- fjptrA = (jnrlistA>=0) ? dvda+jnrA : scratch;
- fjptrB = (jnrlistB>=0) ? dvda+jnrB : scratch;
- fjptrC = (jnrlistC>=0) ? dvda+jnrC : scratch;
- fjptrD = (jnrlistD>=0) ? dvda+jnrD : scratch;
- /* #endif */
- gmx_mm256_increment_4real_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
- _mm256_mul_pd(dvdatmp,_mm256_mul_pd(isaj{J},isaj{J})));
- /* #define INNERFLOPS INNERFLOPS+12 */
- /* #endif */
- velec = _mm256_mul_pd(qq{I}{J},rinv{I}{J});
- /* #define INNERFLOPS INNERFLOPS+1 */
- /* #if 'Force' in KERNEL_VF */
- felec = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(velec,rinv{I}{J}),fgb),rinv{I}{J});
- /* #define INNERFLOPS INNERFLOPS+3 */
- /* #endif */
-
/* #elif KERNEL_ELEC=='Ewald' */
/* EWALD ELECTROSTATICS */
/* #endif */
velecsum = _mm256_add_pd(velecsum,velec);
/* #define INNERFLOPS INNERFLOPS+1 */
- /* #if KERNEL_ELEC=='GeneralizedBorn' */
- /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
- vgb = _mm256_and_pd(vgb,cutoff_mask);
- /* #define INNERFLOPS INNERFLOPS+1 */
- /* #endif */
- /* #if ROUND == 'Epilogue' */
- vgb = _mm256_andnot_pd(dummy_mask,vgb);
- /* #endif */
- vgbsum = _mm256_add_pd(vgbsum,vgb);
- /* #define INNERFLOPS INNERFLOPS+1 */
- /* #endif */
/* #endif */
/* #if 'vdw' in INTERACTION_FLAGS[I][J] */
/* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
/* #define OUTERFLOPS OUTERFLOPS+1 */
/* #endif */
- /* #if 'GeneralizedBorn' in KERNEL_ELEC */
- gmx_mm256_update_1pot_pd(vgbsum,kernel_data->energygrp_polarization+ggid);
- /* #define OUTERFLOPS OUTERFLOPS+1 */
- /* #endif */
/* #if KERNEL_VDW != 'None' */
gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
/* #define OUTERFLOPS OUTERFLOPS+1 */
/* #endif */
/* #endif */
- /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
- dvdasum = _mm256_mul_pd(dvdasum, _mm256_mul_pd(isai{I},isai{I}));
- gmx_mm256_update_1pot_pd(dvdasum,dvda+inr);
- /* #endif */
/* Increment number of inner iterations */
inneriter += j_index_end - j_index_start;
biod.pnpi.spb.ru / alexxy / gromacs.git / blobdiff