While the profiler described here may still work, Venkman is now the preferred way to profile JavaScript. You don't need a special build to profile with Venkman. See the FAQ entry for more information.
javascript runtime profiler
What is it?
Runtime JavaScript profiler. Shows call paths w/ execution times, and number of compiles. (syd, jband)How do I build it?
Set environment variable:- Win32:
XPC_TOOLS_SUPPORT=1(Switch toMOZ_XPC_TOOLS_SUPPORT?) - Linux: Add
--enable-xpctoolsto your configure line. - Mac: ? Probably needs a little build love.
mozilla/js/src/xpconnect.
How do I run it?
Set the environment variable MOZILLA_JS_PROFILER_OUTPUT to the filename you want data dumped to, run the build, and exit mozilla. Note, data collection will affect mozilla speed.Interpreting profiler output
Here is a sample clause from a JS Profiler log file:
chrome://global/content/strres.js
[2,3] srGetStrBundleWithLocale() {27-46} {min 31, max 70 avg 50}
[2,3] srGetStrBundle() {50-52} {min 37, max 76 avg 55}
[2,2] () {1-162} {min 1, max 3 avg 2}
[2,0] localeSwitching() {57-134}
[2,3] srGetAppLocale() {5-23} {min 1, max 6 avg 3}
[2,0] localeTo() {138-158}
[2,0] selectLocale() {162-182}
The first line specifies the js file from which times were gathered.
Each line below that is of the format:
[A, B] C {D-E} F {G, H, I}
where:
A -> The function compile count
B -> The function call count
C -> The function name
D -> Base line number of the function in the JS file
E -> Extent (last) line number of the function in the JS file
F -> total size (net of malloc overhead) of JS function data structures
G -> minimum time of execution in milliseconds
H -> maximum time of execution, in milliseconds
I -> average execution time of function, in milliseconds
[#compiles, #calls] function {lines in file} {Tmin, Tmax, Tavg}
If the function call count (B) is zero, then times (F, G, and H) are not printed, this makes the output a bit easier to read.Unnamed function (lambda) expressions (e.g. JS object methods) will be listed as "anonymous" functions.
Bugs/Issues
- JS Profiler also doesn't appear to count loading/compilation time.
Brendan notes that that XUL JS is compiled at load time (startup or first new window for a given XUL application), and never recompiled. So for profiling operations such as deleting 30 selected mail messages, compile time is 0. Also, JS compile time is better profiled with quantify, to show hotspots in the C code implementing the JS compiler (as opposed to script hotspots found by the JS profiler). - The collection code is rather primitive, so execution times will not be accurate for functions that executed recursively within themselves (syd).