Stream: git-wasmtime

Topic: wasmtime / PR #3736 Add a simple "server" benchmark.


view this post on Zulip Wasmtime GitHub notifications bot (Jan 27 2022 at 22:58):

cfallin opened PR #3736 from server-benchmark to main:

This benchmark repeatedly instantiates and invokes the start function on
a module, from multiple threads in parallel. It is meant to be a simple
test harness for optimizing e.g. instance allocators and other core
runtime functionality.

Split off from #3697.

<!--

Please ensure that the following steps are all taken care of before submitting
the PR.

Please ensure all communication adheres to the code of conduct.
-->

view this post on Zulip Wasmtime GitHub notifications bot (Jan 27 2022 at 22:58):

cfallin requested alexcrichton for a review on PR #3736.

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

Each iteration can take a variable amount of time since it goes to a different module, so would it be better to ensure that iters is rounded up/down to a multiple of the number of modules to ensure we don't see effects where "the really slow module" is run one less time than the others?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

Is this generally intended to limit the number of instances to ensure instantiation always succeeds? For example without this do too many instances get spawned?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

For benchmarking I think it might be better to use get_typed_func here which is more expected in server-like situations.

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

Perhaps instead of a "make engine" function could this be a "make config" function? Dealing with boolean parameters typically requires commenting at each use-site what the parameter means and it's not all that hard to create an Engine from a config so dealing with just configs here might be easier?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

I actually naively thought that block_on only waited for the future specified, but to confirm this does indeed wait for all the spawned tasks to return?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

Could this use if cfg!(...) instead of #[cfg]?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

Cut-off comment?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton created PR review comment:

To avoid confusing this with an actual benchmark, could this be placed at benches/common/mod.rs?

view this post on Zulip Wasmtime GitHub notifications bot (Jan 28 2022 at 14:33):

alexcrichton submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (Feb 09 2022 at 22:00):

cfallin closed without merge PR #3736.


Last updated: Dec 23 2024 at 12:05 UTC