Re: [Gems-users] Multi-processor cache statistics


Date: Thu, 10 Dec 2009 03:35:05 -0500
From: Vishal Oliyil Kunnil <vishal.ok@xxxxxxxxx>
Subject: Re: [Gems-users] Multi-processor cache statistics
Thanks, I was looking at the L1 / L2 statistics, I get the total misses per processor.
protocol - MOSI_SMP_bcast

Vishal

On Thu, Dec 10, 2009 at 3:07 AM, Philip Garcia <pcgarcia@xxxxxxxx> wrote:
what protocol are you using?  In some of the SMP protocols, L1 cache
statistics aren't known by Ruby (as L1 requests don't get forwarded to
Ruby).   In some of the other protocols, the statistics can be
lacking.  I use MOESI_CMP_directory_m for all my simulations, and when
reading back the statistics I get back the following for a 4 processor
CMP containing a coprocessor 5th node:  total_misses: 44345395
[ 5451569 5242672 5426253 5324446 22900455 ], where the numbers are
broken down based on per node statistics.  I am not 100% sure if those
statistics are in the vanilla version of Ruby or just my own (I've
made a number of modifications to it), but often times, I've found the
easiest way to get statistics is to add them myself to Ruby, as every
simulation run tends to have different needs.  For instance, the
protocol I use doesn't give me per processor L1 statistics, despite
all L1 accesses going to Ruby.  Figuring them out tends to be fairly
simple though.

Phil

On Dec 10, 2009, at 1:35 AM, Vishal Oliyil Kunnil wrote:

> Hi all,
>
> I am trying to simulate a two processor  serengeti target, along
> with Ruby.
> I am not sure i interpret correctly, a ruby0.dump-stat does not seem
> to provide
> statistics for the individual processor caches. Is there any way to
> dump per processor
> cache statistics ?
>
> Thanks,
>
> Regards,
> Vishal
>
> --------snip ---------------
> The configuration file settings  are:
>
> L1_CACHE_ASSOC: 4
> L1_CACHE_NUM_SETS_BITS: 12
> L2_CACHE_ASSOC: 4
> L2_CACHE_NUM_SETS_BITS: 18
>
>
> g_MEMORY_SIZE_BYTES: 536870912
> g_DATA_BLOCK_BYTES: 64
> g_PAGE_SIZE_BYTES: 8192
> g_REPLACEMENT_POLICY: PSEDUO_LRU // currently, only other option is
> LRU
>
> g_PROCS_PER_CHIP: 1
>
>
> // set automatically
> g_NUM_PROCESSORS: 2
> g_NUM_L2_BANKS: 0
> g_NUM_MEMORIES: 0
>
> -------snip-------------------
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> Use Google to search the GEMS Users mailing list by adding
> "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
>

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.


[← Prev in Thread] Current Thread [Next in Thread→]