So we should be able to use Chrome devtools protocol to get that, there we have JSHeapUsedSize and .JSHeapTotalSize. Or what do you think @Krinkle ?

I was thinking about window.performance.memory, but that's only because I'm not very familiar with all the great stuff the devtools protocol offers. Probably better indeed. Nice!

I've been tracking JSHeap for a while (but I haven't looked at it) and it looks like it will hard to use to see any changes because the metrics is going up and down. Here's two examples one for mobile and one desktop and I guess if we do a change it needs to big for us to see it.

This is collected when the page has finished loading. But it could also be less noisy for fresnel running on pages with less content?

I wonder if we need something like https://github.com/paulirish/memory-stats.js and measure more over a user journey or something like that?

Something like this would've been useful during the various improvements @DannyS712 and I were doing over the past year in mw.loader. Without it, it wasn't worth it I found to measure it locally every time and with multiple runs.

I ran into it again today as part of https://gerrit.wikimedia.org/r/c/mediawiki/core/+/758549/ which I suspect may reduce the memory size a bit more in the registry, but again takes a fair amount of effort to confirm and seems useful to try to incorporate if it's stable.

I suspect it will be more stable than in our synthetic test given that we have a relatively simple install in CI is, and its a very local fast, and we use Headless Chrome. (And still multiple runs.)

I suspect it will be more stable than in our synthetic test given that we have a relatively simple install in CI

I don't think the tool will matter, I think for all our measurements should run the tests on dedicated hardware, I think that is the crowd-pleaser we are looking for with all our measurements. I looked at dedicated Mac mini m1 and the metric look better but I'm not sure exactly how Chrome handles it internally:

We could also use the raw Chrome trace log. We do that today and on a dedicated machine the numbers looks good:

There's also a new thing in Firefox that the Mozilla performance team added memoryReport , I know they use it in their tests but I haven't tried it so much, You get data from memory usage in the main process but its low level and I'm not sure how usable it would be for us.