Bug 77258
Summary: | run-perf-test should report memory usage | ||
---|---|---|---|
Product: | WebKit | Reporter: | Ryosuke Niwa <rniwa> |
Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
Status: | NEW | ||
Severity: | Normal | CC: | abarth, cmarcelo, eric, kling, koivisto, mitz, morrita, tony |
Priority: | P2 | ||
Version: | 528+ (Nightly build) | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Bug Depends on: | |||
Bug Blocks: | 77378 |
Ryosuke Niwa
Right now, performance tests can only measure run-time performance. We should provide a way to measure memory usage as well.
For example, Chromium pert bots report various peak/average memory usage: http://build.chromium.org/f/chromium/perf/xp-release-dual-core/dom_perf/report.html?history=150&rev=-1&graph=vm_peak_b
Attachments | ||
---|---|---|
Add attachment proposed patch, testcase, etc. |
Hajime Morrita
(In reply to comment #0)
> Right now, performance tests can only measure run-time performance. We should provide a way to measure memory usage as well.
>
> For example, Chromium pert bots report various peak/average memory usage: http://build.chromium.org/f/chromium/perf/xp-release-dual-core/dom_perf/report.html?history=150&rev=-1&graph=vm_peak_b
Interesting. I think not all tests need this.
But having HTML5 standard in our portfolio for this purpose will help recent memory hunting effort.
Ryosuke Niwa
(In reply to comment #1)
> Interesting. I think not all tests need this.
> But having HTML5 standard in our portfolio for this purpose will help recent memory hunting effort.
There are many other pages with interesting characteristics (e.g. lots of javascirpt, lots of text, lots of bidi-text, etc...) we may want to keep track of. The most important question is which value we measure/report.