[parsec-users] Mispredictiona and IPC relation

Mahmood Naderan nt_mahmood at yahoo.com
Thu Jun 16 02:07:27 EDT 2011


Hi
Recently in my simulations (4-core cmp) I found that for some 
PARSEC
applications

1- as the mispredictions are increased (which is 
bad), IPC is
increased too (which is good) !

2- as the mispredictions 
are decreased (which is good), IPC is
decreased too (which is bad) !

I 
have to say, it is not a general observation. I mean only one
benchmark 
(bodytrack) fits '1' and one benchmark (streamcluster) fits
'2'.

Do 
you think it is normal? till now I could find a reason for them.

Thanks 
for any suggestion

// Naderan *Mahmood;


More information about the parsec-users mailing list