Wanted: Concurrent benchmarks

[Luke Gorrie]

Richard Carlsson <richardc@REDACTED> writes:

> It would be really helpful if we could get our hands
> on some more real-world examples (but we would also
> be happy to get some more toy benchmarks, just as
> long as they do a lot of message passing).

In the Jungerl there is lib/ermacs/src/edit_bench.erl. This includes a
synthetic benchmark which models the editor redisplay code, which
wants to send a large data structure (tree of binaries representing a
buffer) between two processes.

In the editor proper I have a kludge to prune the buffer before
sending it to the redisplay process (since only so much can fit on the
screen), but I didn't want to do it this way. IIRC sending the whole
tree was just fine with the shared-heap emulator but otherwise too
slow on large (e.g. tens or hundreds of megabytes) buffers.

I wrote this benchmark to submit to you guys as a shared-heap test
case, as I had promised at an EUC a couple of years ago. I don't
remember if I ever mentioned it was written though :-)

Cheers,
Luke