This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Some (small) c++ compilation profiling data (oprofile)


I have a Dell Inpiron 4100 with a 1 GHz mobile Pentium III and oprofile 
is using the mobile Pentium III performance monitoring hardware.

What kind of processor does oprofile think the machine has?  The 
following command should return a number:

cat /proc/sys/dev/oprofile/cpu_type

I would expect this return "2" for a Pentium III based on enum for 
op_cpu in oprofile/libop/op_interface.h. oprofile has some code to 
determine the processor type. It is possible that it isn't identifying 
the processor. If oprofile cannot identify the processor, it assumes 
CPU_RTC (should be "4"). The /proc/cpuinfo should provide some clues 
with information about "cpu family" and "model".

-Will


Karel Gardas wrote:

> On Fri, 17 May 2002, Will Cohen wrote:
> 
> 
>>Profiling and characterization of gcc by oprofile would be very useful.
>>
>>The cvs version of oprofile has a "--reverse" option for oprofpp, which
>>
> 
> Thanks, I'll use it in the future reports.
> 
> 
>>What environment (hardware and software) did you use to take these
>>measurements? Ideally, someone else should be able to reproduce and
>>verify the result.  GCC is a moving target, the cvs tag used to check
>>out the code should be included. Is the bootstrapped compiler being used
>>for the tests? If just the stage1 compiler is being used, the compiler
>>used to build the GCC being profiled can influence the measurements.
>>
> 
> I've meassured gcc 3.1 (release) bootstraped with standard 3 stage
> bootstrap. Since I'm learning oprofile now, I've used standard
> configuration and so header of tables look:
> 
> Cpu type: CPU with RTC device
> Cpu speed was (MHz estimation) : 995.685000
> Counter 0 counted RTC_Interrupts events (RTC interrupts/sec (rounded up to power of two)) count 1024
> vma      samples  %-age       symbol name
> 080496f0 0        0           _start
> 08049714 0        0           Letext
> 
> Machine is notebook (apm switched off by 'apm=off' kernel option) 1GHz
> PIII 512MB RAM. It seems mobile P6 doesn't provide perf. counter registers
> and so oprofile used rtc. I haven't found yet, how to configure oprofile
> on my machines to measure other data than time spent in functions.
> 
> Cheers,
> 
> Karel
> --
> Karel Gardas                  kgardas@objectsecurity.com
> ObjectSecurity Ltd.           http://www.objectsecurity.com
> 




Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]