IntInf benchmarks
Matthew Fluet <[email protected]>
| Newsgroups | gmane.comp.lang.ml.mlton.user |
|---|---|
| Message-ID | <CAMrhFL6hMauQ9E-FSzo_XT_KE5t6H8mwG1WibXVjEXWR5tCTkw@mail.gmail.com> |
There has been some progress on ensuring that IntInf-free programs do not have a dependency on GMP (https://github.com/MLton/mlton/issues/59); there has also been some (orthogonal) progress on providing a pure SML implementation of IntInf primitives (https://github.com/MLton/mlton/issues/60), which would provide a path for non-IntInf-free programs to not have a dependency on GMP. In any case, I have a couple of options on how to separate the GMP symbols from the MLton runtime library, but I would like to test their performance impacts. So, I'm looking for some IntInf intensive benchmarks. I've adapted the Shootout pidigits benchmark (https://github.com/MatthewFluet/mlton/commit/3a683cfea6100b3cff914555604fe28baa5e4bef), but am interested in any others. With regards to testing the performance of changes to the GMP interface from the MLton runtime library, a good benchmark would ideally be performing mainly ~128-bit arithmetic -- that is, the arithmetic operands should be larger than can be represented as a native 64-bit integer (so that MLton won't be able to perform the arithmetic without calling the GMP function), but not so large that the time to perform the GMP arithmetic would make a small change in the overhead of calling the GMP functions lost in the noise. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. ------------------------------------------------------------------------------