Mixed-Precision Implementation of the Density Matrix Renormalization Group
Benchmark (surveying)
Accuracy and precision
Matrix (chemical analysis)
Double-precision floating-point format
DOI:
10.1021/acs.jctc.2c00632
Publication Date:
2022-11-02T16:12:42Z
AUTHORS (4)
ABSTRACT
The mixed-precision optimization is an effective emerging technique for quantum chemistry methods to obtain better computational performance and maintain the chemical accuracy. Here, we developed a two-level implementation density matrix renormalization group (DMRG) method. This based on idea that DMRG process iterative process. Therefore, first several iteration steps can be executed in single precision. A feasible single-precision may generate moderate accuracy, when few double-precision cleanup sweeps are added, accuracy will recovered. In sweeps, diagonalization method run most time-consuming step precision By combining these two schemes, implemented our benchmark result shows achieved good performance. speed-up of up 2.31 achieved, preserved within 0.01 kcal/mol.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (52)
CITATIONS (6)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....