some profiling numbers

[graydon hoare]

Daniel Berlin wrote:

>> Well, it can output info about macros (you need -g3), but only to the 
>> extent that you can do something like p AMACRO(5) in gdb, and have it 
>> give you the right answer. It can't stop the preprocessor from 
>> expanding them in the source, obviously 
>
well, stopping them from being expanded would be very artificial 
anyways. but dealing with macros correctly would help with a lot of 
programs, not just gcc. can dwarf tell me which VMA extents are the 
results of macro expansions? I can bin samples by any partition 
available, but the samples themselves are just "so-and-so-many overflows 
at VMA foo".

> Also, what type of machine was this done on?
> I don't remember noticing it written somewhere.
> The evidence David Edelsohn gave was for an AIX machine, i forget 
> which POWER processor model it was using.
>
> I've also gotten similar numbers to Dave on a g4 running OSX.

a consumer level pentium 4 (1.6ghz, non-hyperthreaded), with rather 
small L1 data cache (8kb, 4-way, 64 byte lines).  I'd suspect quite 
different results on other machines, but I'm not sure they'd be worse in 
the cache department. maybe. really the whole data set needs to be 
several more dimensions (hardware, release, optimization settings, and 
counter frequencies / ratios) but it would take months to collect that 
data set, if only one person is working on it.

-graydon