Related articles |
---|
Need long memory trace rbjones6@hotmail.com (Rob) (2003-11-21) |
Re: Need long memory trace tmk@netvision.net.il (2003-12-03) |
From: | tmk@netvision.net.il (Michael Tiomkin) |
Newsgroups: | comp.compilers |
Date: | 3 Dec 2003 20:20:39 -0500 |
Organization: | http://groups.google.com |
References: | 03-11-093 |
Keywords: | storage, testing |
Posted-Date: | 03 Dec 2003 20:20:39 EST |
Rob <rbjones6@hotmail.com> wrote in message news:03-11-093...
> I need some help trying to find a memory address trace of length 10M
> or more, It can be from any application program, but should be
> addresses captured with caches disabled. I have been searching google
> for an example with no luck.
I did this many years ago, for trace-based compiler optimizations.
I think I used a single-step trace, analyzing the instructions
and computing the addresses of loads/stores. This was very easy.
It was a RISC machine, and the instructions were very simple,
but I'm sure that for x86 you can find an external program
for analyzing instructions.
Disabling the caches doesn't matter anything in this case,
unless you want to use hardware tracing. And hardware
tracing is usually very expensive.
BTW, 10M is a very short memory trace. If your program accesses
memory every 5 instructions (for x86 this should be 2!), it represents
several milliseconds of CPU time.
Return to the
comp.compilers page.
Search the
comp.compilers archives again.