Here's some data on the webserver, OSX 10.6.8, clang 3.3 svn (old)
and clang's C++ libraries.
RPRVT RSHRD RSIZE VPRVT VSIZE STATE
4084K 244K 6304K 30M 2394M sleeping
3764K 244K 6304K 30M 2394M sleeping
14M 244K 27M 45M 2410M sleeping
11M 244K 25M 43M 2407M sleeping
39M+ 244K 41M+ 61M+ 2425M+ running
14M 244K 36M 50M 2415M sleeping
26M- 244K 38M- 53M- 2418M- running
32M+ 244K 44M+ 61M+ 2425M+ running
40M 244K 48M 66M 2430M sleeping
14M 244K 38M 52M 2417M sleeping
HAMMER
23M+ 244K 41M+ 55M+ 2420M+ running
50M+ 244K 53M+ 70M+ 2434M+ running
22M+ 244K 41M+ 55M+ 2420M+ running
17M 244K 40M 54M 2418M sleeping
MANY HAMMER
32M+ 244K 45M+ 60M+ 2425M+ running
50M+ 244K 53M+ 71M+ 2435M+ running
48M 244K 52M 69M 2433M sleeping
18M 244K 40M 53M 2418M sleeping
ULTRA HAMMER
34M+ 244K 46M+ 61M+ 2426M+ running
50M 244K 53M 71M 2435M sleeping
21M 244K 40M 53M 2418M sleeping
The program hammer:
for var i in 0 upto 1000 do
var result = System::system ("curl
http://localhost:1235/$/usr/local/lib/felix/felix-1.1.9dev/share/src/lib/std/strings/string.fdoc
>/dev/null");
println$ "result=" + result.str;
done
is used once under HAMMER, several times concurrently under
MANY HAMMER, and even more times under ULTRA HAMMER
(using ./hammer & ./hammer & etc on the command line).
As you can see, memory usage creeps up, but the GC is bringing
it down again. The final line is a bit higher than initially, but less than
peak.
Considering on Linux the webserver memory use creeps up monotonically
just from manual page loads, I'd say this version on OSX is not leaking,
though it's hard to tell.
On the felix-lang.org slice, I dare not run hammer from my local
machine for long or I'll kill my bandwidth quota. Using Ubuntu 12.04 LTS
and gcc. These are remote page loads.
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
4617 skaller 20 0 216m 18m 3060 S 0.0 3.9 0:02.34 webserver
4617 skaller 20 0 216m 18m 3060 S 10.0 3.9 0:04.15 webserver
4617 skaller 20 0 231m 33m 3060 S 10.0 6.9 0:09.21 webserver
4617 skaller 20 0 259m 60m 3064 R 16.3 12.4 0:14.76 webserver
4617 skaller 20 0 261m 63m 3064 S 13.0 12.9 0:31.81 webserver
Accidental kill .. new process:
6146 skaller 20 0 253m 55m 3044 S 0.0 11.4 0:08.49 webserver
Now running hammer on the slice ..
6146 skaller 20 0 260m 62m 3044 R 92.3 12.8 0:24.38 webserver
6146 skaller 20 0 261m 63m 3044 S 90.2 13.0 1:51.39 webserver
6146 skaller 20 0 261m 63m 3044 S 0.0 13.0 4:49.62 webserver
minutes later and many page loads .. its still stable at 13% memory.
And with the many hammer ..
6146 skaller 20 0 261m 63m 3052 R 87.6 13.0 6:10.61 webserver
6146 skaller 20 0 261m 63m 3052 R 87.6 13.0 8:32.52 webserver
No change. No leak.
The webserver runs the GC every 60 seconds.
I'm surprised and very pleased you can actually use the Async I/O system
(timer)
to run the collector at regular intervals inside an fthread!!!
--
john skaller
[email protected]
http://felix-lang.org
------------------------------------------------------------------------------
Sponsored by Intel(R) XDK
Develop, test and display web and hybrid apps with a single code base.
Download it for free now!
http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk
_______________________________________________
Felix-language mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/felix-language
