Event-driven DPD timings, temperature, power and migration data

Citation Author(s):
Jonathan
Beaumont
Imperial College London
Submitted by:
Jonathan Beaumont
Last updated:
Tue, 05/17/2022 - 22:17
DOI:
10.21227/n1fh-m125
Data Format:
Excel Spreadsheets
Research Article Link:
A Parallel Mode Optimized GPU Accelerated Monte Carlo Model for Light Propagation in 3-D Voxelized Bio-Tissues
License:
Creative Commons Attribution
87 Views
Categories:
Other
0
0 ratings - Please login to submit your rating.
ACCESS DATASET CITE SHARE/EMBED

Abstract 

This dataset is the recorded data for a Dissipative Particle Dynamics simulator which is designed to run on an event-driven high performance computing platform, POETS (poets-project.org). 

This is in support of an article submitted to IEEE Transactions on Parallel and Distributed Systems, entitled "An event-driven approach to Dissipative Particle Dynamics".

Instructions: 

"dpd-performance-results.xls" contains the wallclock timing results of POETS DPD simulator when run on 1, 2, 4 and 8 POETS machines, as well as result for the same simulations when run with a serial DPD simulator, an MPI based DPD simulator (with varying numbers of procesors) and a few results for a GPGPU based DPD simulator.

"Migration averages.xls" contains data which makes up the average number of beads per timestep which are migrated from one cell to another on a POETS event-driven DPD simulator and an MPI based DPD simulator running on 64 cores.

"MIPS-FLOPS-power.xlsx" contains the data on MIPS (Millions of Instructions Per Second) and FLOPS (FLoating-point Operations Per Second) of the POETS system when producing the results for "dpd-performance-results.xls". This is used with the power per FPGA board (50 Watts) to provide another performance metric for POETS of how many instructions and floating-point operations per second it performs per watt of power. 

The remaining files contain RDF data for a simulation of a system of volume 25x25x25, and 3 beads per unit volume. The makeup of these beads is 60% water, 30% Oil type 1 and 10% Oil type 2. The values are of RDF over time, indicating how likely it is that a bead of type A (given by the first type in the filename) to a bead of type B (the second type in the filename). This gives an indication of, over time, how the beads interact and separate, as oil and water is expected to do, and form structures.

Dataset Files

LOGIN TO ACCESS DATASET FILES

 

QUESTIONS?