bnjbvr opened PR #2755 from vcode-debug-in-debug-only
to main
:
This logging step may be quite expensive, since rendering the VCode into strings has never been
optimized at all and is using pretty naive string concatenation. Removing it is a clear compile times win for a large wasm module I'm looking at. These are the results with parallel compilation enabled, so there could be an overall speedup due to less malloc lock contention too.
- x64 (AMD Ryzen 9 3950X): from 5.2 seconds to 1.8 seconds, i.e. 65% speedup
- aarch64 (Mac M1): from 17.6 seconds to 6.4 seconds, i.e. 63% speedup
alexcrichton submitted PR Review.
bjorn3 submitted PR Review.
bjorn3 created PR Review Comment:
// if the Debug level hasn't been statically (through features) or dynamically (through
bnjbvr updated PR #2755 from vcode-debug-in-debug-only
to main
.
bnjbvr merged PR #2755.
Last updated: Nov 22 2024 at 16:03 UTC