[Jikesrvm-researchers] SPECjvm98 -s1000 ?

[Martin H. <hi...@gm...>]

Hello Jikes RVM researchers:

The command line for SPECjvm98 benchmarks allows specifying workloads
-s1, -s10, or -s100, and specifying multiple runs, for example, with
"-a -m10 -M10".

But I think it would be useful to have a -s1000 option, which would
use a bigger program input than -s100, just like -s100 uses a bigger
input than -s10. That would lead to more live heap data (say, hundreds
of MB) and more computation (say, 90 seconds), hopefully not just by
doing the same thing more often.

One motivation is that for some of the benchmarks, the application
data fits completely into the on-chip cache, which I feel is not
representative of real-world situations.

Since this is a fairly obvious idea, I was wondering whether maybe
somebody has already done it, and would be willing to share it. That
way, we could all make our research a little more relevant, while
still reporting reproducible and comparable results.

Anyone?

Martin