Hello,

I am interested in tracing what C++ code is causing requests to the cache.
For example I have developed a L1 and L2 cache memory  trace, telling me,
using the request packets what physical memory addresses are being
accessed/requested.

So using ALPHA FS mode we have C++ code being compiled into Alpha ASM.
(this is the issue!)

(1)From my understanding I can do for each memory request:
packet->request->pc which will give me the pc address of the instruction
that requested the memory access. Is this correct?

(2) I can then use this pc address to trace the code of C++ that requested
this memory access? Is this correct?

_______________________________________
The second part is a bit hazy to me.
Basically each benchmark that I create a memory trace from the linux disk
image, I need to re-compile the benchmarks using -g, which is a flag for
debugging info. Using the DWARFS format I can use a program like readelf
to read the debugging info. Using readelf (any better programs) I will be
able to see what sections of C++ code are requesting memory accesses using
the pc addresses from the 1st part.

Is this correct?Any better ideas or methods? Any help in giving me
additional details or advice is greatly appreciated. As the second part is
pretty hazy to me.


Thanks,
EF
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to