[parsec-users] PARSEC simulation results

Jim Dempsey jim at quickthreadprogramming.com
Wed May 19 17:20:45 EDT 2010


My guess is as you increase the number of ways you decrease the working data
set size per way for the fixed size L2 cache.
 
Jim

  _____  

From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of sheng qiu
Sent: Wednesday, May 19, 2010 12:48 PM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] PARSEC simulation results


Hi all,

I have saw a weird missrate results while doing simulation on PARSEC
benchmarks. there're two applications called facesim and streamcluster which
have streaming behaviour in these two applications. My simulation config is:
8-core, 2-way L1 private cache(64k InstCache and 64k Dcache), a shared 8M
L2cache.
i set different way-associative in L2 cache: 4-way, 8-way, 16-way. the
results show 4-way L2 has missrate 0.27, 8-way : 0.46, 16-way: 0.699. I
think it's quite weird that the missrate increase so much as the way
increased. Because other factors are not change(i.e. cache size). Do you
have any suggestions what's wrong? the output of these simulations are
verified correct. 
by the way, the simulator i use is M5.

Thanks,
Sheng




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20100519/0bbe2780/attachment.htm>


More information about the parsec-users mailing list