⚡ Bolt: Remove clone on VM instructions in execute loop - #18
Conversation
💡 What: Changed the VM's main `execute_loop` to avoid cloning `BytecodeInstruction` structs on every iteration, passing `op` and `operands` slice to match block and debugger. 🎯 Why: `BytecodeInstruction` can be expensive to clone because it contains a `Vec<Operand>`. The executor loop evaluates millions of instructions per second, so eliminating this clone significantly reduces memory allocations and garbage collection overhead. Added an explicit `is_enabled()` check before `self.debugger.trace_instruction` to avoid lazy allocation of `self.stack.get_dump()` during execution. 📊 Impact: Expected to reduce execution time for compute-heavy TechScript programs by ~45%. On a naive fibonacci(28) test, execution time dropped from 1.55s to 0.81s. 🔬 Measurement: Verified with `time tsc run fib.ts`. Ran full test suite to ensure correctness. Co-authored-by: Tcode-Motion <188012755+Tcode-Motion@users.noreply.github.com>
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
1 participant
I've removed the
clone()operation that happened on every single iteration of the VMexecute_loopwhen reading instructions. This operation was extremely expensive becauseBytecodeInstructionwraps a heap-allocatedVec<Operand>.Instead, I decoupled it by directly copying the
openum and slicingoperands. In addition, I guarded thedebugger.trace_instructioncall behind a boolean condition. This preventsself.stack.get_dump()from unneccesarily building a copy of the stack on every iteration.Benchmarks showed an execution time drop from 1.55s to 0.81s (~45% speedup) on a naive Fibonacci implementation. The tests run cleanly and no semantic behaviors were altered.
I've also logged this finding in
.jules/bolt.mdto prevent similar performance regressions in the future.PR created automatically by Jules for task 9953887423283783221 started by @Tcode-Motion