Recent Posts

Make sure you haven't maximized the Assembly view, preventing source view from being visible.  I.E. click the little unmaximize box to the upper-right of the Assembly code view.

Hi Adam,

Sorry for the delay in getting back to you - we don't always get a notification when a post is made to forums.

I assume you're measuring the memory usage of the GUI? How are you measuring the heap usage? The GUI is partially implemented in Java/SWT - so memory released as java objects may not be reflected in the space shown in top for example.

If memory usage is critical, you may want to take profiles using the zoom command line tool and then later open them up with the Zoom UI on a machine that is less memory constrained. Alternatively, the command line can output a basic summary of the call graph.

I'm working with Zoom 1.6.6 under Ubuntu 10.04, right now it looks like the GUI never releases memory associated with profiles.

After 1 profile Zoom is using 204M, after a second identical run 282M, if I close (without saving) one of the profiles the memory usage remains unchanged. This trend continues until around 10 profiles or so, at which point I have to restart the GUI to free up some memory.

I'm getting the profiles using zoomscript exec to launch my program, it profiles the first 30 seconds using a time profile with a sampling interval of 400us. Because the program is running in a loop over a large quantity of data I tend to Ctrl+C it once I've obtained the profile samples.

That aside, I think Zoom is hands down the best Linux profiler I've used, and would heartily recommend it to anyone looking for one.

The Xeon in this system is not a PIII of course. The problem is that RHEL 4.x doesn't parse the extended model field. So it's listed as model 7 when the CPU is model 23.

i friends i am BE in electronics in communication.i have recently joined as tech support associate in a mnc. since there are no IT jobs in market right now. i am satisfied.but i cant make any career out of this job

Can you send a profile that shows this problem? Mail it to support at rotateright.com.

Thanks!

Hi!

I sampled my app with Zoom and get the list of symbols in the UI, but when I double click on a symbol Zoom jumps into the asm code instead of the C++ code. The program was compiled and linked with -g and -fno-omit-frame-pointer under x64. The binary contains the sourcecode information, I can dump it with objdump. The "Source-to-Assembly" windows says: "No debug information available.".
I tried -g, -gdwarf-2 and -gwarf-4 as compileoptions, none of them work.

System:
Linux xen 2.6.34-ARCH #1 SMP PREEMPT (x64)
g++ (GCC) 4.5.0 20100610

I tried NILFS on a USB-stick and its peformance is better than ext2 and ext3. Btrfs is currently (kernel 2.6.34) still very unstable for me (lost free space, corrupt fs) and as far as I remember it wasn't as fast as NILFS(2).
Here is my NILFS-diary:
http://www.blah-blah.ch/Mra/Nilfs2performance
Bye

At this point, we do not support Pentium III based systems. Therefore a compatible event list wasn't included.

Hi,

I'm testing Zoom on a different machine, a RHEL 4 64-bit install.  This is Redhat Update 5 though.

When starting the Zoom GUI I get the error:

"Unable to connect to local daemon.

Reason: Failed to launch Zoom daemon.  The reason for failure has been recorded in the log file at "~/.zoom/zoom-run.log".  Please report the bug using the 'Report Bug' function accessible from the 'Help' menu.  Thank you."

However, the referenced zoom-run.log file is empty (although it does exist).

Running the zoom daemon manually results in:

/opt/rotateright/bin/zoom run
2010-Jun-14 15:34:59 WARNING: failed to read CPU max frequency from /dev/rrprofile/cpu_khz  [Linux/RRDriverInfo.cpp:318]
-t: No such file or directory
2010-Jun-14 15:34:59 WARNING: failed to read CPU max frequency from /dev/rrprofile/cpu_khz  [Linux/RRDriverInfo.cpp:318]
ERROR: Unable to find the event list for i386:piii.

Any ideas?

Cheers,

James Bray

PS: The kernel module you sent for my previous issue (http://rotateright.com/forum/index.php?topic=153.0) worked fine.