wasm-runner: add parallel_map/4 for concurrent collection processing#13
Open
wasm-runner: add parallel_map/4 for concurrent collection processing#13
Conversation
Adds the missing WasmRunner.run/2 and run!/2 functions for ergonomic
one-shot WASM execution with a keyword-list calling convention:
{:ok, 8} = WasmRunner.run("math.wasm", add: [5, 3])
{:ok, [8, 55]} = WasmRunner.run("math.wasm", add: [5, 3], fibonacci: [10])
8 = WasmRunner.run!("math.wasm", add: [5, 3])
Features:
- Single call returns unwrapped value: {:ok, 8} not {:ok, [8]}
- Multiple calls return list: {:ok, [8, 55]}
- Options (wasi, cache) mixed into keyword list
- Works with precompiled modules for 53x speedup
- Auto WASI detection for Go modules
- Instance always cleaned up (even on error)
Performance characteristics (from benchmark):
- Cold run/2: ~2ms (dominated by WASM compilation)
- Precompiled run/2: ~40μs (53x faster)
- Pre-loaded call_single: ~15μs (call overhead only)
- WASM fibonacci(30) is 5x faster than pure Elixir
Includes:
- 22 tests covering single/multi calls, errors, cleanup, precompiled
- Benchmark: bench/wasm_runner_run.exs
- Performance docs in docs/PERFORMANCE_GUIDE.md
Adds Pool.call_many/2, call_many!/2, and call_many_unwrapped/2 for executing multiple WASM function calls in a single checkout/checkin cycle. This eliminates N-1 GenServer round-trips when batching calls. Performance (fixtures/math.wasm, pool size 4): - 5 calls: 1.27x faster than repeated Pool.call - 20 calls: 1.18x faster - 100 calls: 1.12x faster Includes: - 16 tests covering correctness, ordering, errors, concurrency - Benchmark comparing call_many vs repeated call vs native Elixir - Performance documentation in docs/POOL_CALL_MANY.md
Add WasmRunner.pipe/3 and pipe!/3 that chain WASM function calls,
feeding the output of each call as input to the next. The module is
loaded once, all calls execute on a single instance, and cleanup is
automatic.
Features:
- Output of each call becomes first arg of the next by default
- :pipe placeholder for explicit argument positioning
- Works with precompiled modules, WASI auto-detection
- Includes 18 tests covering chaining, placeholders, errors, cleanup
- Benchmark script comparing pipe vs separate run calls
Example:
{:ok, 42} = WasmRunner.pipe("math.wasm", [
{:add, [5, 3]}, # => 8
{:fibonacci, []}, # => fibonacci(8) = 21
{:multiply, [:pipe, 2]} # => multiply(21, 2) = 42
])
- Add Firebird.call/4 accepting opts (timeout) forwarded to Runtime.call - Add call_single/4 and call_single!/4 with optional timeout kwarg - Thread :timeout through run/2, pipe/3, run_batch/3, run_concurrent/4 - Thread :timeout through benchmark/4 and benchmark_compare/5 - Add extract_call_opts/1 to cleanly separate call opts from start opts - Add :timeout to @known_opts so run/2 keyword API recognizes it - Add 12 tests covering timeout threading for all WasmRunner functions
- Register @wasm attribute with accumulate: true in wasm_modules - Remove unused generate_expr/3 in wat_gen.ex (use /5 directly) - Prefix unused module_name param in firebird.compile.ex
Add functional collection primitives to WasmRunner that load a WASM module once, apply a function over an enumerable, and auto-cleanup the instance: - map/4, map!/4: apply a WASM function to each element - reduce/5, reduce!/5: fold a collection through a WASM binary op - filter/4, filter!/4: keep elements where WASM predicate returns non-zero - each/4: apply for side-effects only Input normalization handles integers, tuples, and lists automatically. Works with file paths, raw bytes, and precompiled modules. Includes 34 tests and a Benchee benchmark (bench/wasm_collection.exs).
Add WasmRunner.parallel_map/4 and parallel_map!/4 that distribute work across multiple WASM instances using Task.async_stream. The module is compiled once and shared across all instances for efficiency. Key features: - Preserves input order in output - Concurrency auto-capped to collection size - Defaults to System.schedulers_online() instances - Works with precompiled modules for zero-compilation overhead - Extracts start_parallel_instances/3 helper (shared with run_concurrent) Includes 13 tests covering ordering, edge cases, precompiled modules, and consistency with sequential map/4.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
By Sleepy
Adds
WasmRunner.parallel_map/4andparallel_map!/4— concurrent versions ofmap/4that distribute work across multiple WASM instances usingTask.async_stream.Key features:
System.schedulers_online()parallel instancesstart_parallel_instances/3helper (reusable byrun_concurrent/4)When to use:
map/4→ small collections or fast functions (avoids instance overhead)parallel_map/4→ large collections with expensive per-element work (crypto, sorting, recursive algorithms)Tests: 13 tests covering ordering, edge cases, precompiled modules, and consistency with sequential
map/4.