Thanks Anthony,

If we suppose that all instructions of a given application can be entirely cached in i-cache, does the number of read request misses in i-cache will correspond to the number of line fills in i-cache ? Also, do you know how can I see in stats file if all instructions of the application are entirely cached in L1 i-cache only, or if some instructions are cached in L1 i-cache and others are cached in L2 ? If i am not wrong, in the first case, read misses in i-cache will correspond to compulsory cache misses which will occur only at the beginning of the execution. Whereas in the second case, some read misses will occur during the execution (not only at the beginning).

For instance, I have simulated the benchmark BARNES of SPLASH-2 in FS mode with --caches and --l2cache options. In stats file, I can see that results :

system.physmem.num_reads::cpu.inst            22321
system.l2.overall_misses::cpu.inst              22321
system.cpu.icache.ReadReq_misses::cpu.inst        98236

Does it mean that instructions of the application is not entirely cached in L1 i-cache ?

Cordialement / Best Regards

SENNI Sophiane
Ph.D. candidate - Microelectronics
LIRMM - www.lirmm.fr

Le 08/03/2014 15:28, Anthony Gutierrez a écrit :
In most CPUs the i-cache is read only. Nothing should be explicitly written to the i-cache, it will be populated when a fetch req misses, causing a line fill. Even in self-modifying code you don't write explicitly to an executable page, you first remove execute permissions and mark it as data.


Anthony Gutierrez
http://web.eecs.umich.edu/~atgutier <http://web.eecs.umich.edu/%7Eatgutier>


On Sat, Mar 8, 2014 at 9:23 AM, senni sophiane <[email protected] <mailto:[email protected]>> wrote:

    Hi everybody,

    In stats file, I can find "dcache.WriteReq_accesses" corresponding
    to the number of write request accesses in D-Cache. However, there
    is no such results for I-Cache. As far as I know, a number of
    writes has to occur to cache instructions of the application from
    Main Memory to I-Cache. So, am I missing something ? Or are there
    other results in stats file that can inform us about I-Cache write
    accesses ?

    Thanks

-- Cordialement / Best Regards

    SENNI Sophiane
    Ph.D. candidate - Microelectronics
    LIRMM - www.lirmm.fr <http://www.lirmm.fr>

    _______________________________________________
    gem5-users mailing list
    [email protected] <mailto:[email protected]>
    http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users




_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to