Performance testing of ABACUS #1948
Replies: 10 comments 10 replies
|
Version: v3.1.3 dataset1:ABACUS-PW (CG)dataset2:ABACUS-LCAO (gen-elpa) |
|
Version: v3.1.4 dataset1:ABACUS-PWdataset2:ABACUS-LCAO |
|
Version: v3.2.0 dataset1:ABACUS-PWdataset2:ABACUS-LCAO |
|
Is this test using MPI or OpenMP? |
|
Machine: |
|
maybe a test run using OpenMP is also helpful. I am not sure how to choose |
|
version: v3.2.3
version: v3.2.3
version: v3.2.3
version: v3.2.3
|
|
@Satinelamp The above are some results of testing on v3.2.3, which is a reference. |
|
Is there any tests of performance comparison between ABACUS and VASP? In my experience, even when I use LCAO basis, ABACUS is obviously slower than VASP. |
@Satinelamp Hi! Can you share the output files of VASP. There are only input files in the zip file. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Since v3.1.3, along with the release of new version, we will report the performance testing.
The geometric mean of the inverse of total time is used as the metric, and the value of v3.0.0 is set to 1.0 as the baseline.
ks_solver.ks_solver.c32_m64_cpu aliin Bohrium platform to do the calculationOMP_NUM_THREADS=1 mpirun -np 16 abacusPW samples:
LCAO samples:
All reactions