Is it possible to limit CPUtime? about quickjs-emscripten

[justjake]

Hey @BartTactick,

how do you expect this to work? There’s an option to run a function periodically that can interrupt execution. You can probably use that to implement the time-based semantics you need. There’s an example of a deadline based interrupt in the README, see shouldInterrupt: … here: https://github.com/justjake/quickjs-emscripten#safely-evaluate-javascript-code

You can add this to a Runtime using setInterruptHandler: https://github.com/justjake/quickjs-emscripten/blob/main/doc/classes/QuickJSRuntime.md#setinterrupthandler

[BartTactick]

Hey @justjake ,

Basicly the way Cloudflare workers work: https://developers.cloudflare.com/workers/platform/pricing/
So lets say your doing a heavy loop it will count how long you have used the CPU.
But when the CPU is idle (for example waiting for a network response or something) it wont count anything.
In the end you would have a the MS that the script has ran.
And based on that you could dispose the execution/script.

[justjake]

To clarify, you want to count the wall-clock time taken by every call to a VM function that executes code in the VM, and possibly interrupt execution once a certain limit is reached, but when the VM exits because it’s awaiting a host Promise (ie, IO inside an async loop running in the VM) you don’t want to count that time?

I think this is very doable without changes to the library today. The VM does not block the host thread while it is awaiting. And, you always control when the VM resumes execution after a host promise resolves a VM promise. So you can save a time stamp before every call into the VM and after every call into the VM exits, and use those to account for the CPU time spent. Additionally, you could use the interrupt handler to interrupt if (timeSoFar + (Date.now() - startTimeOfThisCall) > MAX_TIME).

[BartTactick]

Im pointing towards CPU time or user-cpu time if your looking at this stackoverflow post:
"The user-cpu time and system-cpu time are pretty much as you said - the amount of time spent in user code and the amount of time spent in kernel code."

https://stackoverflow.com/a/7335965

[justjake]

I think my previous post describes a way to capture user-cpu time if you define user-cpu time as “evaluating code in the vm”, which I think seems reasonable.

[BartTactick]

I think my previous post describes a way to capture user-cpu time if you define user-cpu time as “evaluating code in the vm”, which I think seems reasonable.

I will look into it. Thank you so far!