real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m128i vfitab;
__m128i ifour = _mm_set1_epi32(4);
__m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
real *vftab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
__m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
real * vdwioffsetptr2;
__m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
real * vdwioffsetptr3;
__m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
__m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
__m256d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
BI_FORCE_INLINE inline avx_double operator/(const double& o1,
const avx_double& o2) {
avx_double res;
res.packed = _mm256_div_pd(_mm256_set1_pd(o1), o2.packed);
return res;
}
int *iinr,*jindex,*jjnr,*shiftidx,*gid;
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m128i ewitab;
__m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
__m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
real *ewtab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m128i vfitab;
__m128i ifour = _mm_set1_epi32(4);
__m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
real *vftab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
__m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
real * vdwioffsetptr2;
__m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
real * vdwioffsetptr3;
__m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
__m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
__m256d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
krf = _mm256_set1_pd(fr->ic->k_rf);
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
int *iinr,*jindex,*jjnr,*shiftidx,*gid;
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__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;
__m128i vfitab;
__m128i ifour = _mm_set1_epi32(4);
__m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
real *vftab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
real * vdwgridioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d c6grid_00;
real *vdwgridparam;
__m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
__m256d one_half = _mm256_set1_pd(0.5);
__m256d minus_one = _mm256_set1_pd(-1.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
real * vdwgridioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d c6grid_00;
real *vdwgridparam;
__m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
__m256d one_half = _mm256_set1_pd(0.5);
__m256d minus_one = _mm256_set1_pd(-1.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
real * vdwioffsetptr1;
__m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
real * vdwioffsetptr2;
__m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
__m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
krf = _mm256_set1_pd(fr->ic->k_rf);
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
krf = _mm256_set1_pd(fr->ic->k_rf);
real * vdwioffsetptr1;
__m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
real * vdwioffsetptr2;
__m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
__m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
void CalculateBasisComponents(const MDoubleArray& weights, const BaryCoords& coords,
const MIntArray& triangleVertices, const MPointArray& points,
const MFloatVectorArray& normals, const MIntArray& sampleIds,
double* alignedStorage,
MPoint& origin, MVector& up, MVector& normal) {
// Start with the recreated point and normal using the barycentric coordinates of the hit point.
unsigned int hitIndex = weights.length()-1;
#ifdef __AVX__
__m256d originV = Dot4<MPoint>(coords[0], coords[1], coords[2], 0.0,
points[triangleVertices[0]], points[triangleVertices[1]],
points[triangleVertices[2]], MPoint::origin);
__m256d hitNormalV = Dot4<MVector>(coords[0], coords[1], coords[2], 0.0,
normals[triangleVertices[0]], normals[triangleVertices[1]],
normals[triangleVertices[2]], MVector::zero);
__m256d hitWeightV = _mm256_set1_pd(weights[hitIndex]);
// Create the barycentric point and normal.
__m256d normalV = _mm256_mul_pd(hitNormalV, hitWeightV);
// Then use the weighted adjacent data.
for (unsigned int j = 0; j < hitIndex; j += 4) {
__m256d tempNormal = Dot4<MVector>(weights[j], weights[j+1], weights[j+2], weights[j+3],
normals[sampleIds[j]], normals[sampleIds[j+1]],
normals[sampleIds[j+2]], normals[sampleIds[j+3]]);
normalV = _mm256_add_pd(tempNormal, normalV);
}
_mm256_store_pd(alignedStorage, originV);
origin.x = alignedStorage[0];
origin.y = alignedStorage[1];
origin.z = alignedStorage[2];
_mm256_store_pd(alignedStorage, normalV);
normal.x = alignedStorage[0];
normal.y = alignedStorage[1];
normal.z = alignedStorage[2];
// Calculate the up vector
const MPoint& pt1 = points[triangleVertices[0]];
const MPoint& pt2 = points[triangleVertices[1]];
__m256d p1 = _mm256_set_pd(pt1.w, pt1.z, pt1.y, pt1.x);
__m256d p2 = _mm256_set_pd(pt2.w, pt2.z, pt2.y, pt2.x);
p1 = _mm256_add_pd(p1, p2);
__m256d half = _mm256_set_pd(0.5, 0.5, 0.5, 0.5);
p1 = _mm256_mul_pd(p1, half);
__m256d upV = _mm256_sub_pd(p1, originV);
_mm256_store_pd(alignedStorage, upV);
up.x = alignedStorage[0];
up.y = alignedStorage[1];
up.z = alignedStorage[2];
#else
MVector hitNormal;
// Create the barycentric point and normal.
for (int i = 0; i < 3; ++i) {
origin += points[triangleVertices[i]] * coords[i];
hitNormal += MVector(normals[triangleVertices[i]]) * coords[i];
}
// Use crawl data to calculate normal
normal = hitNormal * weights[hitIndex];
for (unsigned int j = 0; j < hitIndex; j++) {
normal += MVector(normals[sampleIds[j]]) * weights[j];
}
// Calculate the up vector
// The triangle vertices are sorted by decreasing barycentric coordinates so the first two are
// the two closest vertices in the triangle.
up = ((points[triangleVertices[0]] + points[triangleVertices[1]]) * 0.5) - origin;
#endif
normal.normalize();
GetValidUp(weights, points, sampleIds, origin, normal, up);
}
int *iinr,*jindex,*jjnr,*shiftidx,*gid; real rcutoff_scalar; real *shiftvec,*fshift,*x,*f; real *fjptrA,*fjptrB,*fjptrC,*fjptrD; real scratch[4*DIM]; __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; real * vdwioffsetptr0; __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; __m256d velec,felec,velecsum,facel,crf,krf,krf2; real *charge; __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; int nvdwtype; __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; int *vdwtype; real *vdwparam; __m256d one_sixth = _mm256_set1_pd(1.0/6.0); __m256d one_twelfth = _mm256_set1_pd(1.0/12.0); __m128i vfitab; __m128i ifour = _mm_set1_epi32(4); __m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF; real *vftab; __m256d dummy_mask,cutoff_mask; __m128 tmpmask0,tmpmask1; __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) ); __m256d one = _mm256_set1_pd(1.0);
static real probabilities_avx_hernquist(const AstronomyParameters* ap,
const StreamConstants* sc,
const real* RESTRICT sg_dx,
const real* RESTRICT r_point,
const real* RESTRICT qw_r3_N,
LBTrig lbt,
real gPrime,
real reff_xr_rp3,
real* RESTRICT streamTmps)
{
double bg_prob, dotted, xyz_norm;
int i, j, k, convolve, nStreams;
MW_ALIGN_V(64) double psgt[256], psgf[256], xyzstr[256];
MW_ALIGN_V(64) double xs[256], ys[256], zs[256];
const __m256d REF_XR = _mm256_set1_pd(reff_xr_rp3);
const __m256d COSBL = _mm256_set1_pd(lbt.lCosBCos);
const __m256d SINB = _mm256_set1_pd(lbt.bSin);
const __m256d SINCOSBL = _mm256_set1_pd(lbt.lSinBCos);
const __m256d SUNR0 = _mm256_set1_pd(ap->sun_r0);
const __m256d R0 = _mm256_set1_pd(ap->r0);
const __m256d QV_RECIP = _mm256_set1_pd(ap->q_inv);
__m256d RI, QI;
ssp_m256 xyz0, xyz1, xyz2, tmp0, tmp1, tmp2, PROD, PBXV, BGP;
//xyz0, 1, 2 = x, y, z
BGP.d = _mm256_setzero_pd();
convolve = ap->convolve;
nStreams = ap->number_streams;
int *iinr,*jindex,*jjnr,*shiftidx,*gid;
real rcutoff_scalar;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
real rswitch_scalar,d_scalar;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
static inline __m256d gmx_mm256_exp2_pd(__m256d x)
{
/* Lower bound: We do not allow numbers that would lead to an IEEE fp representation exponent smaller than -126. */
const __m256d arglimit = _mm256_set1_pd(1022.0);
const __m128i expbase = _mm_set1_epi32(1023);
const __m256d P2 = _mm256_set1_pd(2.30933477057345225087e-2);
const __m256d P1 = _mm256_set1_pd(2.02020656693165307700e1);
const __m256d P0 = _mm256_set1_pd(1.51390680115615096133e3);
/* Q2 == 1.0 */
const __m256d Q1 = _mm256_set1_pd(2.33184211722314911771e2);
const __m256d Q0 = _mm256_set1_pd(4.36821166879210612817e3);
const __m256d one = _mm256_set1_pd(1.0);
const __m256d two = _mm256_set1_pd(2.0);
__m256d valuemask;
__m256i iexppart;
__m128i iexppart128a, iexppart128b;
__m256d fexppart;
__m256d intpart;
__m256d z, z2;
__m256d PolyP, PolyQ;
iexppart128a = _mm256_cvtpd_epi32(x);
intpart = _mm256_round_pd(x, _MM_FROUND_TO_NEAREST_INT);
/* Add exponent bias */
iexppart128a = _mm_add_epi32(iexppart128a, expbase);
/* We now want to shift the exponent 52 positions left, but to achieve this we need
* to separate the 128-bit register data into two registers (4x64-bit > 128bit)
* shift them, and then merge into a single __m256d.
* Elements 0/1 should end up in iexppart128a, and 2/3 in iexppart128b.
* It doesnt matter what we put in the 2nd/4th position, since that data will be
* shifted out and replaced with zeros.
*/
iexppart128b = _mm_shuffle_epi32(iexppart128a, _MM_SHUFFLE(3, 3, 2, 2));
iexppart128a = _mm_shuffle_epi32(iexppart128a, _MM_SHUFFLE(1, 1, 0, 0));
iexppart128b = _mm_slli_epi64(iexppart128b, 52);
iexppart128a = _mm_slli_epi64(iexppart128a, 52);
iexppart = _mm256_castsi128_si256(iexppart128a);
iexppart = _mm256_insertf128_si256(iexppart, iexppart128b, 0x1);
valuemask = _mm256_cmp_pd(arglimit, gmx_mm256_abs_pd(x), _CMP_GE_OQ);
fexppart = _mm256_and_pd(valuemask, _mm256_castsi256_pd(iexppart));
z = _mm256_sub_pd(x, intpart);
z2 = _mm256_mul_pd(z, z);
PolyP = _mm256_mul_pd(P2, z2);
PolyP = _mm256_add_pd(PolyP, P1);
PolyQ = _mm256_add_pd(z2, Q1);
PolyP = _mm256_mul_pd(PolyP, z2);
PolyQ = _mm256_mul_pd(PolyQ, z2);
PolyP = _mm256_add_pd(PolyP, P0);
PolyQ = _mm256_add_pd(PolyQ, Q0);
PolyP = _mm256_mul_pd(PolyP, z);
z = _mm256_mul_pd(PolyP, gmx_mm256_inv_pd(_mm256_sub_pd(PolyQ, PolyP)));
z = _mm256_add_pd(one, _mm256_mul_pd(two, z));
z = _mm256_mul_pd(z, fexppart);
return z;
}
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
__m128i ewitab;
__m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
__m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
real *ewtab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
jindex = nlist->jindex;
jjnr = nlist->jjnr;
shiftidx = nlist->shift;
gid = nlist->gid;
shiftvec = fr->shift_vec[0];
fshift = fr->fshift[0];
facel = _mm256_set1_pd(fr->epsfac);
charge = mdatoms->chargeA;
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m128i ewitab;
__m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
__m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
real *ewtab;
__m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
real rswitch_scalar,d_scalar;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m128i vfitab;
__m128i ifour = _mm_set1_epi32(4);
__m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
real *vftab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
Color3 sampleFourier3(float * const coeffs[3], const double *recip, size_t nCoeffs,
Float sample, Float &pdf, Float &phi) {
bool flip = false;
if (sample < 0.5f) {
sample *= 2.0f;
} else {
sample = 1.0f - 2.0f * (sample - 0.5f);
flip = true;
}
int iterations = 0;
double a = 0.0,
c = math::Pi_d,
coeff0 = coeffs[0][0],
y = coeff0*math::Pi_d*sample,
deriv = 0.0,
b = 0.5 * math::Pi_d,
cosB = 0,
sinB = 1;
if (nCoeffs > 10 && sample != 0 && sample != 1) {
float stddev = std::sqrt(2.0f / 3.0f * std::log(coeffs[0][1] / coeffs[0][2]));
if (std::isfinite(stddev)) {
b = std::min(c, (double) math::normal_quantile(0.5f + sample / 2) * stddev);
cosB = std::cos(b);
sinB = std::sqrt(1 - cosB * cosB);
}
}
#if FOURIER_SCALAR != 1
__m256d factorB_prev, factorB_cur;
#endif
while (true) {
#if FOURIER_SCALAR == 1
double cosB_prev = cosB,
sinB_prev = -sinB,
sinB_cur = 0.0,
cosB_cur = 1.0,
value = coeff0 * b;
deriv = coeff0;
for (size_t j=1; j<nCoeffs; ++j) {
double sinB_next = 2.0*cosB*sinB_cur - sinB_prev,
cosB_next = 2.0*cosB*cosB_cur - cosB_prev,
coeff = (double) coeffs[0][j];
value += coeff * recip[j] * sinB_next;
deriv += coeff * cosB_next;
sinB_prev = sinB_cur; sinB_cur = sinB_next;
cosB_prev = cosB_cur; cosB_cur = cosB_next;
}
#else
initializeRecurrence(cosB, factorB_prev, factorB_cur);
__m256d
sinB_prev = _mm256_set1_pd(-sinB),
sinB_cur = _mm256_set1_pd(0.0),
cosB_prev = _mm256_set1_pd(cosB),
cosB_cur = _mm256_set1_pd(1.0),
value_vec = _mm256_set_sd(coeff0 * b),
deriv_vec = _mm256_set_sd(coeff0);
for (size_t j=1; j<nCoeffs; j+=4) {
__m128 coeff_vec_f = _mm_load_ps(coeffs[0]+j);
__m256d recip_vec = _mm256_load_pd(recip+j);
__m256d coeff_vec = _mm256_cvtps_pd(coeff_vec_f);
__m256d sinB_next = _mm256_add_pd(
_mm256_mul_pd(factorB_prev, sinB_prev),
_mm256_mul_pd(factorB_cur, sinB_cur));
__m256d cosB_next = _mm256_add_pd(
_mm256_mul_pd(factorB_prev, cosB_prev),
_mm256_mul_pd(factorB_cur, cosB_cur));
value_vec = _mm256_add_pd(value_vec, _mm256_mul_pd(
_mm256_mul_pd(recip_vec, coeff_vec), sinB_next));
deriv_vec = _mm256_add_pd(deriv_vec, _mm256_mul_pd(coeff_vec, cosB_next));
sinB_prev = _mm256_splat2_pd(sinB_next);
cosB_prev = _mm256_splat2_pd(cosB_next);
sinB_cur = _mm256_splat3_pd(sinB_next);
cosB_cur = _mm256_splat3_pd(cosB_next);
}
double value = simd::hadd(value_vec);
deriv = simd::hadd(deriv_vec);
#endif
value -= y;
if (std::abs(value) <= 1e-5 * coeff0 || ++iterations > 20)
break;
else if (value > 0.0)
c = b;
else
a = b;
b -= value / deriv;
if (!(b >= a && b <= c))
b = 0.5f * (a + c);
cosB = std::cos(b);
sinB = std::sqrt(1-cosB*cosB);
}
double Y = deriv;
if (flip)
b = 2.0*math::Pi_d - b;
pdf = (Float) (math::InvTwoPi_d * Y / coeff0);
phi = (Float) b;
#if FOURIER_SCALAR == 1
double cosB_prev = cosB,
cosB_cur = 1.0;
double R = coeffs[1][0];
double B = coeffs[2][0];
for (size_t j=1; j<nCoeffs; ++j) {
double cosB_next = 2.0*cosB*cosB_cur - cosB_prev,
coeffR = (double) coeffs[1][j],
coeffB = (double) coeffs[2][j];
R += coeffR * cosB_next;
B += coeffB * cosB_next;
cosB_prev = cosB_cur; cosB_cur = cosB_next;
}
#else
__m256d
cosB_prev = _mm256_set1_pd(cosB),
cosB_cur = _mm256_set1_pd(1.0),
R_vec = _mm256_set_sd(coeffs[1][0]),
B_vec = _mm256_set_sd(coeffs[2][0]);
for (size_t j=1; j<nCoeffs; j+=4) {
__m128 coeff_R_vec_f = _mm_load_ps(coeffs[1]+j);
__m128 coeff_B_vec_f = _mm_load_ps(coeffs[2]+j);
__m256d coeff_R_vec = _mm256_cvtps_pd(coeff_R_vec_f);
__m256d coeff_B_vec = _mm256_cvtps_pd(coeff_B_vec_f);
__m256d cosB_next = _mm256_add_pd(
_mm256_mul_pd(factorB_prev, cosB_prev),
_mm256_mul_pd(factorB_cur, cosB_cur));
R_vec = _mm256_add_pd(R_vec, _mm256_mul_pd(coeff_R_vec, cosB_next));
B_vec = _mm256_add_pd(B_vec, _mm256_mul_pd(coeff_B_vec, cosB_next));
cosB_prev = _mm256_splat2_pd(cosB_next);
cosB_cur = _mm256_splat3_pd(cosB_next);
}
double R = simd::hadd(R_vec);
double B = simd::hadd(B_vec);
#endif
double G = 1.39829 * Y - 0.100913 * B - 0.297375 * R;
return Color3((Float) R, (Float) G, (Float) B)
* (2 * math::Pi) * (Float) (coeff0 / Y);
}
BI_FORCE_INLINE inline avx_double operator*(const avx_double& o1,
const double& o2) {
avx_double res;
res.packed = _mm256_mul_pd(o1.packed, _mm256_set1_pd(o2));
return res;
}
real *shiftvec,*fshift,*x,*f;
real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
real scratch[4*DIM];
__m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
real * vdwioffsetptr0;
__m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
__m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
__m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
__m256d velec,felec,velecsum,facel,crf,krf,krf2;
real *charge;
int nvdwtype;
__m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
int *vdwtype;
real *vdwparam;
__m256d one_sixth = _mm256_set1_pd(1.0/6.0);
__m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
__m128i ewitab;
__m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
__m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
real *ewtab;
__m256d dummy_mask,cutoff_mask;
__m128 tmpmask0,tmpmask1;
__m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
__m256d one = _mm256_set1_pd(1.0);
__m256d two = _mm256_set1_pd(2.0);
x = xx[0];
f = ff[0];
nri = nlist->nri;
iinr = nlist->iinr;
