From 83f139642eb638e73dd74a89b58cbcbc38d72ca5 Mon Sep 17 00:00:00 2001
From: stamoor <stamoor@f3b2605a-c512-4ea7-a41b-209d697bcdaa>
Date: Tue, 6 Sep 2016 22:09:41 +0000
Subject: [PATCH] Reverting optimizations that hurt performance on some
compilers
git-svn-id: svn://svn.icms.temple.edu/lammps-ro/trunk@15551 f3b2605a-c512-4ea7-a41b-209d697bcdaa
---
src/KOKKOS/pair_eam_kokkos.cpp | 137 ++++++++++++++-------------------
src/KOKKOS/pair_eam_kokkos.h | 16 ++--
2 files changed, 65 insertions(+), 88 deletions(-)
diff --git a/src/KOKKOS/pair_eam_kokkos.cpp b/src/KOKKOS/pair_eam_kokkos.cpp
index 37a26145e0..d91da280ac 100755
--- a/src/KOKKOS/pair_eam_kokkos.cpp
+++ b/src/KOKKOS/pair_eam_kokkos.cpp
@@ -340,81 +340,64 @@ void PairEAMKokkos<DeviceType>::array2spline()
rdr = 1.0/dr;
rdrho = 1.0/drho;
- tdual_ffloat4 k_frho_spline_a = tdual_ffloat4("pair:frho_a",nfrho,nrho+1);
- tdual_ffloat4 k_rhor_spline_a = tdual_ffloat4("pair:rhor_a",nrhor,nr+1);
- tdual_ffloat4 k_z2r_spline_a = tdual_ffloat4("pair:z2r_a",nz2r,nr+1);
- tdual_ffloat4 k_frho_spline_b = tdual_ffloat4("pair:frho_b",nfrho,nrho+1);
- tdual_ffloat4 k_rhor_spline_b = tdual_ffloat4("pair:rhor_b",nrhor,nr+1);
- tdual_ffloat4 k_z2r_spline_b = tdual_ffloat4("pair:z2r_b",nz2r,nr+1);
-
- t_host_ffloat4 h_frho_spline_a = k_frho_spline_a.h_view;
- t_host_ffloat4 h_rhor_spline_a = k_rhor_spline_a.h_view;
- t_host_ffloat4 h_z2r_spline_a = k_z2r_spline_a.h_view;
- t_host_ffloat4 h_frho_spline_b = k_frho_spline_b.h_view;
- t_host_ffloat4 h_rhor_spline_b = k_rhor_spline_b.h_view;
- t_host_ffloat4 h_z2r_spline_b = k_z2r_spline_b.h_view;
+ tdual_ffloat_2d_n7 k_frho_spline = tdual_ffloat_2d_n7("pair:frho",nfrho,nrho+1);
+ tdual_ffloat_2d_n7 k_rhor_spline = tdual_ffloat_2d_n7("pair:rhor",nrhor,nr+1);
+ tdual_ffloat_2d_n7 k_z2r_spline = tdual_ffloat_2d_n7("pair:z2r",nz2r,nr+1);
+
+ t_host_ffloat_2d_n7 h_frho_spline = k_frho_spline.h_view;
+ t_host_ffloat_2d_n7 h_rhor_spline = k_rhor_spline.h_view;
+ t_host_ffloat_2d_n7 h_z2r_spline = k_z2r_spline.h_view;
for (int i = 0; i < nfrho; i++)
- interpolate(nrho,drho,frho[i],h_frho_spline_a,h_frho_spline_b,i);
- k_frho_spline_a.template modify<LMPHostType>();
- k_frho_spline_a.template sync<DeviceType>();
- k_frho_spline_b.template modify<LMPHostType>();
- k_frho_spline_b.template sync<DeviceType>();
+ interpolate(nrho,drho,frho[i],h_frho_spline,i);
+ k_frho_spline.template modify<LMPHostType>();
+ k_frho_spline.template sync<DeviceType>();
for (int i = 0; i < nrhor; i++)
- interpolate(nr,dr,rhor[i],h_rhor_spline_a,h_rhor_spline_b,i);
- k_rhor_spline_a.template modify<LMPHostType>();
- k_rhor_spline_a.template sync<DeviceType>();
- k_rhor_spline_b.template modify<LMPHostType>();
- k_rhor_spline_b.template sync<DeviceType>();
+ interpolate(nr,dr,rhor[i],h_rhor_spline,i);
+ k_rhor_spline.template modify<LMPHostType>();
+ k_rhor_spline.template sync<DeviceType>();
for (int i = 0; i < nz2r; i++)
- interpolate(nr,dr,z2r[i],h_z2r_spline_a,h_z2r_spline_b,i);
- k_z2r_spline_a.template modify<LMPHostType>();
- k_z2r_spline_a.template sync<DeviceType>();
- k_z2r_spline_b.template modify<LMPHostType>();
- k_z2r_spline_b.template sync<DeviceType>();
-
- d_frho_spline_a = k_frho_spline_a.d_view;
- d_rhor_spline_a = k_rhor_spline_a.d_view;
- d_z2r_spline_a = k_z2r_spline_a.d_view;
- d_frho_spline_b = k_frho_spline_b.d_view;
- d_rhor_spline_b = k_rhor_spline_b.d_view;
- d_z2r_spline_b = k_z2r_spline_b.d_view;
-
+ interpolate(nr,dr,z2r[i],h_z2r_spline,i);
+ k_z2r_spline.template modify<LMPHostType>();
+ k_z2r_spline.template sync<DeviceType>();
+ d_frho_spline = k_frho_spline.d_view;
+ d_rhor_spline = k_rhor_spline.d_view;
+ d_z2r_spline = k_z2r_spline.d_view;
}
/* ---------------------------------------------------------------------- */
template<class DeviceType>
-void PairEAMKokkos<DeviceType>::interpolate(int n, double delta, double *f, t_host_ffloat4 h_spline_a, t_host_ffloat4 h_spline_b, int i)
+void PairEAMKokkos<DeviceType>::interpolate(int n, double delta, double *f, t_host_ffloat_2d_n7 h_spline, int i)
{
- for (int m = 1; m <= n; m++) h_spline_b(i,m).w = f[m];
+ for (int m = 1; m <= n; m++) h_spline(i,m,6) = f[m];
- h_spline_b(i,1).z = h_spline_b(i,2).w - h_spline_b(i,1).w;
- h_spline_b(i,2).z = 0.5 * (h_spline_b(i,3).w-h_spline_b(i,1).w);
- h_spline_b(i,n-1).z = 0.5 * (h_spline_b(i,n).w-h_spline_b(i,n-2).w);
- h_spline_b(i,n).z = h_spline_b(i,n).w - h_spline_b(i,n-1).w;
+ h_spline(i,1,5) = h_spline(i,2,6) - h_spline(i,1,6);
+ h_spline(i,2,5) = 0.5 * (h_spline(i,3,6)-h_spline(i,1,6));
+ h_spline(i,n-1,5) = 0.5 * (h_spline(i,n,6)-h_spline(i,n-2,6));
+ h_spline(i,n,5) = h_spline(i,n,6) - h_spline(i,n-1,6);
for (int m = 3; m <= n-2; m++)
- h_spline_b(i,m).z = ((h_spline_b(i,m-2).w-h_spline_b(i,m+2).w) +
- 8.0*(h_spline_b(i,m+1).w-h_spline_b(i,m-1).w)) / 12.0;
+ h_spline(i,m,5) = ((h_spline(i,m-2,6)-h_spline(i,m+2,6)) +
+ 8.0*(h_spline(i,m+1,6)-h_spline(i,m-1,6))) / 12.0;
for (int m = 1; m <= n-1; m++) {
- h_spline_b(i,m).y = 3.0*(h_spline_b(i,m+1).w-h_spline_b(i,m).w) -
- 2.0*h_spline_b(i,m).z - h_spline_b(i,m+1).z;
- h_spline_b(i,m).x = h_spline_b(i,m).z + h_spline_b(i,m+1).z -
- 2.0*(h_spline_b(i,m+1).w-h_spline_b(i,m).w);
+ h_spline(i,m,4) = 3.0*(h_spline(i,m+1,6)-h_spline(i,m,6)) -
+ 2.0*h_spline(i,m,5) - h_spline(i,m+1,5);
+ h_spline(i,m,3) = h_spline(i,m,5) + h_spline(i,m+1,5) -
+ 2.0*(h_spline(i,m+1,6)-h_spline(i,m,6));
}
- h_spline_b(i,n).y = 0.0;
- h_spline_b(i,n).x = 0.0;
+ h_spline(i,n,4) = 0.0;
+ h_spline(i,n,3) = 0.0;
for (int m = 1; m <= n; m++) {
- h_spline_a(i,m).z = h_spline_b(i,m).z/delta;
- h_spline_a(i,m).y = 2.0*h_spline_b(i,m).y/delta;
- h_spline_a(i,m).x = 3.0*h_spline_b(i,m).x/delta;
+ h_spline(i,m,2) = h_spline(i,m,5)/delta;
+ h_spline(i,m,1) = 2.0*h_spline(i,m,4)/delta;
+ h_spline(i,m,0) = 3.0*h_spline(i,m,3)/delta;
}
}
@@ -558,13 +541,12 @@ void PairEAMKokkos<DeviceType>::operator()(TagPairEAMKernelA<NEIGHFLAG,NEWTON_PA
p -= m;
p = MIN(p,1.0);
const int d_type2rhor_ji = d_type2rhor(jtype,itype);
- const F_FLOAT4 rhor = d_rhor_spline_b(d_type2rhor_ji,m);
- rhotmp += ((rhor.x*p + rhor.y)*p + rhor.z)*p + rhor.w;
-
+ rhotmp += ((d_rhor_spline(d_type2rhor_ji,m,3)*p + d_rhor_spline(d_type2rhor_ji,m,4))*p +
+ d_rhor_spline(d_type2rhor_ji,m,5))*p + d_rhor_spline(d_type2rhor_ji,m,6);
if (NEWTON_PAIR || j < nlocal) {
const int d_type2rhor_ij = d_type2rhor(itype,jtype);
- const F_FLOAT4 rhor = d_rhor_spline_b(d_type2rhor_ij,m);
- rho[j] += ((rhor.x*p + rhor.y)*p + rhor.z)*p + rhor.w;
+ rho[j] += ((d_rhor_spline(d_type2rhor_ij,m,3)*p + d_rhor_spline(d_type2rhor_ij,m,4))*p +
+ d_rhor_spline(d_type2rhor_ij,m,5))*p + d_rhor_spline(d_type2rhor_ij,m,6);
}
}
@@ -593,11 +575,10 @@ void PairEAMKokkos<DeviceType>::operator()(TagPairEAMKernelB<EFLAG>, const int &
p -= m;
p = MIN(p,1.0);
const int d_type2frho_i = d_type2frho[itype];
- const F_FLOAT4 frho = d_frho_spline_a(d_type2frho_i,m);
- d_fp[i] = (frho.x*p + frho.y)*p + frho.z;
+ d_fp[i] = (d_frho_spline(d_type2frho_i,m,0)*p + d_frho_spline(d_type2frho_i,m,1))*p + d_frho_spline(d_type2frho_i,m,2);
if (EFLAG) {
- const F_FLOAT4 frho_b = d_frho_spline_b(d_type2frho_i,m);
- F_FLOAT phi = ((frho_b.x*p + frho_b.y)*p + frho_b.z)*p + frho_b.w;
+ F_FLOAT phi = ((d_frho_spline(d_type2frho_i,m,3)*p + d_frho_spline(d_type2frho_i,m,4))*p +
+ d_frho_spline(d_type2frho_i,m,5))*p + d_frho_spline(d_type2frho_i,m,6);
if (d_rho[i] > rhomax) phi += d_fp[i] * (d_rho[i]-rhomax);
if (eflag_global) ev.evdwl += phi;
if (eflag_atom) d_eatom[i] += phi;
@@ -651,8 +632,8 @@ void PairEAMKokkos<DeviceType>::operator()(TagPairEAMKernelAB<EFLAG>, const int
p -= m;
p = MIN(p,1.0);
const int d_type2rhor_ji = d_type2rhor(jtype,itype);
- const F_FLOAT4 rhor = d_rhor_spline_b(d_type2rhor_ji,m);
- rhotmp += ((rhor.x*p + rhor.y)*p + rhor.z)*p + rhor.w;
+ rhotmp += ((d_rhor_spline(d_type2rhor_ji,m,3)*p + d_rhor_spline(d_type2rhor_ji,m,4))*p +
+ d_rhor_spline(d_type2rhor_ji,m,5))*p + d_rhor_spline(d_type2rhor_ji,m,6);
}
}
@@ -669,11 +650,10 @@ void PairEAMKokkos<DeviceType>::operator()(TagPairEAMKernelAB<EFLAG>, const int
p -= m;
p = MIN(p,1.0);
const int d_type2frho_i = d_type2frho[itype];
- const F_FLOAT4 frho = d_frho_spline_a(d_type2frho_i,m);
- d_fp[i] = (frho.x*p + frho.y)*p + frho.z;
+ d_fp[i] = (d_frho_spline(d_type2frho_i,m,0)*p + d_frho_spline(d_type2frho_i,m,1))*p + d_frho_spline(d_type2frho_i,m,2);
if (EFLAG) {
- const F_FLOAT4 frho_b = d_frho_spline_b(d_type2frho_i,m);
- F_FLOAT phi = ((frho_b.x*p + frho_b.y)*p + frho_b.z)*p + frho_b.w;
+ F_FLOAT phi = ((d_frho_spline(d_type2frho_i,m,3)*p + d_frho_spline(d_type2frho_i,m,4))*p +
+ d_frho_spline(d_type2frho_i,m,5))*p + d_frho_spline(d_type2frho_i,m,6);
if (d_rho[i] > rhomax) phi += d_fp[i] * (d_rho[i]-rhomax);
if (eflag_global) ev.evdwl += phi;
if (eflag_atom) d_eatom[i] += phi;
@@ -742,18 +722,16 @@ void PairEAMKokkos<DeviceType>::operator()(TagPairEAMKernelC<NEIGHFLAG,NEWTON_PA
// hence embed' = Fi(sum rho_ij) rhojp + Fj(sum rho_ji) rhoip
const int d_type2rhor_ij = d_type2rhor(itype,jtype);
- const F_FLOAT4 rhor_ij = d_rhor_spline_a(d_type2rhor_ij,m);
- const F_FLOAT rhoip = (rhor_ij.x*p + rhor_ij.y)*p + rhor_ij.z;
-
+ const F_FLOAT rhoip = (d_rhor_spline(d_type2rhor_ij,m,0)*p + d_rhor_spline(d_type2rhor_ij,m,1))*p +
+ d_rhor_spline(d_type2rhor_ij,m,2);
const int d_type2rhor_ji = d_type2rhor(jtype,itype);
- const F_FLOAT4 rhor_ji = d_rhor_spline_a(d_type2rhor_ji,m);
- const F_FLOAT rhojp = (rhor_ji.x*p + rhor_ji.y)*p + rhor_ji.z;
-
+ const F_FLOAT rhojp = (d_rhor_spline(d_type2rhor_ji,m,0)*p + d_rhor_spline(d_type2rhor_ji,m,1))*p +
+ d_rhor_spline(d_type2rhor_ji,m,2);
const int d_type2z2r_ij = d_type2z2r(itype,jtype);
- const F_FLOAT4 z2r_a = d_z2r_spline_a(d_type2z2r_ij,m);
- const F_FLOAT z2p = (z2r_a.x*p + z2r_a.y)*p + z2r_a.z;
- const F_FLOAT4 z2r_b = d_z2r_spline_b(d_type2z2r_ij,m);
- const F_FLOAT z2 = ((z2r_b.x*p + z2r_b.y)*p + z2r_b.z)*p + z2r_b.w;
+ const F_FLOAT z2p = (d_z2r_spline(d_type2z2r_ij,m,0)*p + d_z2r_spline(d_type2z2r_ij,m,1))*p +
+ d_z2r_spline(d_type2z2r_ij,m,2);
+ const F_FLOAT z2 = ((d_z2r_spline(d_type2z2r_ij,m,3)*p + d_z2r_spline(d_type2z2r_ij,m,4))*p +
+ d_z2r_spline(d_type2z2r_ij,m,5))*p + d_z2r_spline(d_type2z2r_ij,m,6);
const F_FLOAT recip = 1.0/r;
const F_FLOAT phi = z2*recip;
@@ -894,5 +872,4 @@ template class PairEAMKokkos<LMPDeviceType>;
#ifdef KOKKOS_HAVE_CUDA
template class PairEAMKokkos<LMPHostType>;
#endif
-}
-
+}
\ No newline at end of file
diff --git a/src/KOKKOS/pair_eam_kokkos.h b/src/KOKKOS/pair_eam_kokkos.h
index b3d33bb37f..89d1d61b6b 100755
--- a/src/KOKKOS/pair_eam_kokkos.h
+++ b/src/KOKKOS/pair_eam_kokkos.h
@@ -136,14 +136,14 @@ class PairEAMKokkos : public PairEAM {
DAT::t_int_2d_randomread d_type2rhor;
DAT::t_int_2d_randomread d_type2z2r;
- typedef Kokkos::DualView<F_FLOAT4**,Kokkos::LayoutLeft,DeviceType> tdual_ffloat4;
- typedef typename tdual_ffloat4::t_dev_const_randomread t_ffloat4_randomread;
- typedef typename tdual_ffloat4::t_host t_host_ffloat4;
-
- t_ffloat4_randomread d_frho_spline_a, d_frho_spline_b;
- t_ffloat4_randomread d_rhor_spline_a, d_rhor_spline_b;
- t_ffloat4_randomread d_z2r_spline_a, d_z2r_spline_b;
- void interpolate(int, double, double *, t_host_ffloat4, t_host_ffloat4, int);
+ typedef Kokkos::DualView<F_FLOAT**[7],Kokkos::LayoutRight,DeviceType> tdual_ffloat_2d_n7;
+ typedef typename tdual_ffloat_2d_n7::t_dev_const_randomread t_ffloat_2d_n7_randomread;
+ typedef typename tdual_ffloat_2d_n7::t_host t_host_ffloat_2d_n7;
+
+ t_ffloat_2d_n7_randomread d_frho_spline;
+ t_ffloat_2d_n7_randomread d_rhor_spline;
+ t_ffloat_2d_n7_randomread d_z2r_spline;
+ void interpolate(int, double, double *, t_host_ffloat_2d_n7, int);
virtual void file2array();
void array2spline();
--
GitLab