97
edits
Necko/MobileCache/MicroBenchmarks: Difference between revisions
Jump to navigation
Jump to search
Necko/MobileCache/MicroBenchmarks (view source)
Revision as of 11:30, 14 September 2011
, 14 September 2011no edit summary
No edit summary |
|||
| Line 4: | Line 4: | ||
Each benchmark below is described and explained, and we also show output results from selected platforms. Note that results are provided as examples and information only; the benchmarks are meant for evaluating effect of code-changes, i.e. you run the benchmarks without and with your change and compare the results to evaluate the effect of your code. | Each benchmark below is described and explained, and we also show output results from selected platforms. Note that results are provided as examples and information only; the benchmarks are meant for evaluating effect of code-changes, i.e. you run the benchmarks without and with your change and compare the results to evaluate the effect of your code. | ||
== Telemetry vs microbenchmarks == | |||
There has been some discussion about xpcshell-based microbenchmarks vs using telemetry to measure cache-performance. Although IMO these approaches are complementary, there '''may''' be a point to use telemetry-code to capture data from xpcshell-tests (as opposed to using JS Time-functions from the test itself). I've started a pro/con list for each approach - feel free to add and comment. '''Note that lab-experiments in both approaches will use synthetic test-patterns - the difference is in the way we capture data.''' | |||
{| border="1" cellpadding="5" cellspacing="0" align="center" | |||
|+'''Measure cache-performance using JS vs using telemetry-code''' | |||
|- | |||
!| | |||
!| Time-function in JavaScript | |||
!| Telemetry code compiled into the browser | |||
|- | |||
!| Pros | |||
|| | |||
* Add and experiment with measurements without recompiling | |||
|| | |||
* Avoid JS and thread-schedule jitter | |||
* Better granularity than JS-functions (<i>???</i>) | |||
* Verify changes using results from real users | |||
|- | |||
!| Cons | |||
|| | |||
* Lab-tests only, no way to verify for real usage | |||
|| | |||
* Increased code-size (<i>also in releases??</i>) | |||
* Must have an idea up-front about timing-profile (histograms) | |||
|- | |||
|} | |||
== test_timing_cache.js == | == test_timing_cache.js == | ||
| Line 285: | Line 313: | ||
CacheHit (disk-cache) 22626 101 224.02 | CacheHit (disk-cache) 22626 101 224.02 | ||
== Wanted tests == | |||
Unordered and informal list of tests we may also want to create | Unordered and informal list of tests we may also want to create | ||