optimize Wasmtime engine configuration for performance improvements#121
optimize Wasmtime engine configuration for performance improvements#121
Conversation
ruslanti
changed the title
There was a problem hiding this comment.
Pull request overview
This PR aims to improve Wasmtime runtime performance by tuning the engine and pooling allocator configuration, and bumps the workspace version.
Changes:
- Configure Wasmtime with explicit Cranelift optimization level and disable address-map generation.
- Expand pooling allocator limits (tables/memories per module) and add keep-resident + decommit batching tuning.
- Bump workspace/package versions from 0.15.0 → 0.15.1 (including
Cargo.lock).
Reviewed changes
Copilot reviewed 2 out of 3 changed files in this pull request and generated 5 comments.
| File | Description |
|---|---|
| crates/runtime/src/lib.rs | Adds Wasmtime Config and pooling allocator performance tuning in WasmConfigBuilder::build(). |
| Cargo.toml | Bumps workspace version to 0.15.1. |
| Cargo.lock | Updates workspace crate versions to 0.15.1. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| // Avoids madvise(MADV_DONTNEED) syscalls when a pooled slot is reused, replacing them | ||
| // with a cheaper memset up to the threshold — reducing per-request deallocation latency. | ||
| pooling_allocation_config.linear_memory_keep_resident(2 * MB); | ||
| pooling_allocation_config.table_keep_resident(512 * 1024); | ||
|
|
There was a problem hiding this comment.
With the per-module maxima increased, the keep-resident thresholds (linear_memory_keep_resident, table_keep_resident) can materially increase steady-state RSS for warm pooled instances (threshold × number of memories/tables × warm slots). If this is intended, it would be safer to make these thresholds configurable/tunable or keep the per-module maxima conservative to avoid unexpected memory growth under load.
| pooling_allocation_config.max_tables_per_module(10); | ||
| pooling_allocation_config.max_memories_per_module(10); |
There was a problem hiding this comment.
max_tables_per_module(10) / max_memories_per_module(10) can require substantially higher pool capacity. When max_execution_stacks is set, this builder still sizes total_tables(total) and total_memories(total) to total, which may underprovision if modules actually use >1 table/memory and cause instantiation failures due to pool exhaustion. Consider scaling pool totals accordingly (or keeping per-module maxima at 1 unless required).
configured by env generate address map Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
revert dumping version Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…ents