comp-chem/cp2k
1
0
Fork 0
mirror of https://github.com/cp2k/cp2k.git synced 2026-07-28 14:15:19 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

Adapted RI-HFX single node benchmark input

Browse source
This commit is contained in:
abussy 2022-10-14 10:20:45 +02:00 • committed by Augustin Bussy
parent cea20fb160
commit 7d91dcfa18
1 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
benchmarks/QS_single_node/RI-HFX_H2O-32.inp
View file

@ -63,9 +63,9 @@
RI_FLAVOR RHO !Using the RHO flavor, which involve more sparse contractions than the MO flavor
RI_METRIC IDENTITY !Use the overlap as RI metric for maximum sparsity
MIN_BLOCK_SIZE 4 !This is the default block size. Smaller blocks help with sparsity.
MEMORY_CUT 9 !Memory cut: controls the batching of tensor contractions. With this default, at most
&END RI !1/9 of the tensors are sent to the GPU at a given time (note: rounded to a square number)
&END HF
MEMORY_CUT 3 !Memory cut: controls the batching of tensor contractions. With this default,
&END RI !at most 1/9 of the sparser tensors are sent to the GPU at a given time,
&END HF !and 1/27 of the desner tensors.
&END XC
&END DFT
&SUBSYS