abrown opened Issue #2619:
After discussing https://github.com/bytecodealliance/wasmtime/pull/2612, @fitzgen and I agreed that the best way to configure the bench-api shared library for running benchmarks is with a configuration
struct
. This has advantages such as:
- the bench-api signatures explicitly indicate what is required for running a benchmark (e.g. working directory, where to capture stdout, what wasi modules are enabled, etc.)
- we will avoid thrash on the bench-api signatures and confine this type of configuration change to the configuration
struct
(still an API change, but not as drastic)One additional thought is to version the configuration
struct
in some way (first field is ausize
?) to verify that both sightglass and the bench-api shared library are operating on the same understanding of how to configure the benchmark.
fitzgen commented on Issue #2619:
related: https://github.com/bytecodealliance/sightglass/issues/78
abrown commented on Issue #2619:
Another way of doing this would be to use environment variables to communicate from sightglass to the bench-api shared library, but this seems more prone to error.
Last updated: Nov 22 2024 at 16:03 UTC