Understanding...

- Client: The BOINC software that manages and controls the Science Applications (SA) available in 32 and 64 bit compile:
- Project: The Distributed Computing Grid that sends the SA for processing to a Client, some offering a 32 bit SA, others a 64 bit SA or both.

- 32 bit SA: What most floating point, double precision computations run at. WCG SA's are intense on these.
- 64 bit SA: An advanced compile for faster processing, but due to the fact that the floating point, double precision computations still run in 32 bit, not very much faster than 32 bit. Gains reported and measured are ~15%.

What the 64 bit client does is blow up the benchmark, Dhrystone (integer) test to double or more. The Whetstone (floating point) part remains almost the same as the 32 bit client which is foretelling why the real gain is not enormous. So, Is Milky Way FP or Int Intensive?

So much for my simplified understanding.

Answers are No and Maybe.
The science applications that we run at WCG are provided by the project scientists and audited and certified by WCG. At this stage if an application were so poorly written that it could be made 48 times faster by writing it correctly I think that it would be detected and that the project would be delayed until an acceptable version is provided. I doubt that WCG takes the risk to make us waste our CPU cycles by crunching lousy code.

There could possibly be a more efficient version only if the scientists provide one through the same path, which most of them are not really motivated to do.
However there are some projects like HCC which are already much more efficient (about 25 % faster for HCC) when running in 64-bit mode without any other modification.

Cheers. Jean.

. . . not a pool but a spring!