ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Using NEIGHMASK bitmasking for array indices to improve vector gather performance with nextgen compiler.

Browse Source
This commit is contained in:
W. Michael Brown
2022-08-22 15:57:02 -07:00
parent 66bbfa67dc
commit f7cf5b6751
25 changed files with 83 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/INTEL/pppm_intel.cpp
View File

@ -403,7 +403,6 @@ void PPPMIntel::particle_map(IntelBuffers<flt_t,acc_t> *buffers)
// (nx,ny,nz) = global coords of grid pt to "lower left" of charge
// current particle coord can be outside global and local box
// add/subtract OFFSET to avoid int(-0.75) = 0 when want it to be -1
int nx = static_cast<int> ((x[i].x-lo0)*xi+fshift) - OFFSET;
int ny = static_cast<int> ((x[i].y-lo1)*yi+fshift) - OFFSET;
int nz = static_cast<int> ((x[i].z-lo2)*zi+fshift) - OFFSET;
@ -941,6 +940,7 @@ void PPPMIntel::fieldforce_ad(IntelBuffers<flt_t,acc_t> *buffers)
#endif
#endif
for (int i = ifrom; i < ito; i++) {
i = IP_PRE_dword_index(i);
particle_ekx[i] *= hx_inv;
particle_eky[i] *= hy_inv;
particle_ekz[i] *= hz_inv;