docs(benchmarks): add benchmark results and cross-runtime fib(35) comparison#34
Merged
Merged
Conversation
…parison Measure threaded interpreter throughput across all opcode groups on linux/amd64 (Intel Xeon @ 2.10 GHz). Add cross-runtime fibonacci(20) comparison against native Go, goja, and tengo. Key findings: - Simple i32/i64/f32/f64 ops: ~20–22 ns/op, 0 allocs - Function calls (bytecode): ~26–29 ns/op, 0 allocs - Host calls: ~36 ns/op, 1 alloc - Heap objects (array/struct): ~90–140 ns/op - Map operations: ~420–535 ns/op - fib(20): minivm 968 µs, tengo 1729 µs (1.8×), goja 2958 µs (3×) Add interp/compare_bench_test.go for the cross-runtime fib(20) benchmark. Add goja and tengo as benchmark-only test dependencies. Update AGENTS.md Documentation Index to include benchmarks.md. https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
Remove interp/compare_bench_test.go (goja/tengo/native-Go fib comparison). Run go mod tidy to drop goja and tengo from go.mod/go.sum. Results are preserved in docs/benchmarks.md. https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
…sults Highlight fib(20) comparison (minivm vs tengo, goja, native Go) and single-instruction throughput table. Link to docs/benchmarks.md for full results. Added to both README.md and README_kr.md. https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
Add wazero (WASM JIT) and gopher-lua (register VM) as comparison runtimes. Embed hand-encoded fib.wasm binary; validate result before timing loop. Results on linux/amd64 (Intel Xeon @ 2.80 GHz): native Go: 37,968 ns/op 0 allocs wazero: 62,219 ns/op 2 allocs (WASM JIT) minivm: 1,157,136 ns/op 0 allocs (threaded interpreter) tengo: 2,000,364 ns/op 28657 allocs gopher-lua: 2,942,015 ns/op 2 allocs goja: 3,964,702 ns/op 39 allocs Update docs/benchmarks.md and both READMEs with new runtime table and execution-model column. Clarify wazero's JIT advantage vs minivm's threaded interpreter, and ARM64 JIT context. https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
fib(35) results on linux/amd64 (Intel Xeon @ 2.80 GHz): native Go: 51,947,220 ns/op 0 B/op 0 allocs wazero: 84,807,148 ns/op 16 B/op 2 allocs (WASM JIT) minivm: 1,672,707,295 ns/op 288 B/op 1 alloc (threaded) tengo: 2,665,298,176 ns/op 312M B/op 39M allocs gopher-lua: 4,081,167,978 ns/op 971K B/op 3793 allocs goja: 5,427,175,850 ns/op 383K B/op 46384 allocs Remove compare_bench_test.go and external dependencies after measurement. https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
- benchmarks.md: consolidate repetitive arithmetic tables into summary rows; tighten section flow; remove duplicate methodology note - README.md / README_kr.md: cleaner intro, consistent section structure, tighter prose, fix instruction set table header https://claude.ai/code/session_01WKDZezhPTdEpczDB2rySPY
Owner
Author
PR Review — docs(benchmarks): add benchmark results and cross-runtime fib(20) comparisonDecisionRequest Changes Merge Readiness SummaryThe PR adds comprehensive and well-structured benchmark documentation with clear methodology and comparative analysis. However, there is a critical discrepancy between the PR description and the actual changes: the description states that goja and tengo dependencies were added, but the final state shows they were not included. This mismatch must be clarified before merge. Blocking Findings1. PR Description Does Not Match Final StateEvidence:
Why it matters: Minimal fix: Option A (if removal was intentional): Option B (if addition was intended): Important FindingsNone — the documentation itself is high-quality. Suggestions
Questions
Readiness Checklist
Minimal Path to Merge
Final RecommendationUpdate PR description to match final state, then merge. Generated by Claude Code |
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
2 participants
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.
Measure threaded interpreter throughput across all opcode groups on linux/amd64 (Intel Xeon @ 2.80 GHz, Go 1.26.2). Add cross-runtime fib(35) comparison against native Go, wazero, gopher-lua, tengo, and goja.
Key findings:
Changes
docs/benchmarks.mdwith full opcode throughput tables and cross-runtime comparisonREADME.mdandREADME_kr.mdwith performance section, new tagline, and readability improvementsAGENTS.mdDocumentation Index to includebenchmarks.mdBenchmark comparison code (
interp/compare_bench_test.go) and external dependencies (wazero, gopher-lua, tengo, goja) were used for measurement and removed after results were recorded indocs/benchmarks.md.