We present a new parallel computational model, named Log-GPS, which capture
s synchronization.
The LogGPS model is an extension of the LogGP model, which abstracts commun
ication on parallel platforms. Although the LogGP model captures long messa
ges with one bandwidth parameter (G), it does not capture synchronization t
hat is needed before sending a long message by high-level communication lib
raries. Our model has one additional parameter, S, defined as the threshold
for message length, above which synchronous messages are sent.
We also present some experimental results using both models. The results in
clude (1) a verification of the LogGPS model, (2) an example of synchroniza
tion analysis using an MPI program and (3) a comparison of the models. The
results indicate that the LogGPS model is more accurate than the LogGP mode
l, and analyzing synchronization costs is important when improving parallel
program performance.