Re: [Gems-users] Time spent in synchronization

[Hemayet Hossain <hossain@xxxxxxxxxxxxxxxx>]

Hi Mike,
Thanks a lot.

The info about the real machine I used are given below. Actually I am using a modified version of MSI_MOESI_CMP_directory protocol (tuned for single CMP) and for interconnection I am using 4x4 mesh having 3 cycle hop latency. 16MB shared L2 (16 slices of 1 MB each) and 64KBL1D and 64KB L1I.

Thanks,
Hemayet
-------------------------------------
SunOS swym.cs.rochester.edu 5.10 Generic_118833-33 sun4u sparc SUNW,Sun-Fire
0       on-line   since 06/13/2007 09:32:40
Status of virtual processor 0 as of: 12/13/2007 11:48:56
  on-line since 06/13/2007 09:32:40.
  The sparcv9 processor operates at 1200 MHz,
        and has a sparcv9 floating point processor.
1       on-line   since 06/13/2007 09:32:41
2       on-line   since 06/13/2007 09:32:41
3       on-line   since 06/13/2007 09:32:41
8       on-line   since 06/13/2007 09:32:41
9       on-line   since 06/13/2007 09:32:41
10      on-line   since 06/13/2007 09:32:41
11      on-line   since 06/13/2007 09:32:41
12      on-line   since 06/13/2007 09:32:41
13      on-line   since 06/13/2007 09:32:41
14      on-line   since 06/13/2007 09:32:41
15      on-line   since 06/13/2007 09:32:41
20      on-line   since 06/13/2007 09:32:41
21      on-line   since 06/13/2007 09:32:41
22      on-line   since 06/13/2007 09:32:41
23      on-line   since 06/13/2007 09:32:41
-------------------------------------

Mike Marty wrote:

What real machine did you use?

Running Simics without Ruby is cheating since there is no synchronization penalty (everything is 1 IPC).

MESI_SCMP_directory is a blocking directory protocol, so any lock handoffs are essentially 4-hop.  What interconnect topology were you using? 

--Mike

On Dec 13, 2007 10:01 AM, Hemayet Hossain <hossain@xxxxxxxxxxxxxxxx> wrote:

Hi All,
I am simulating some splash2 benchmarks by using ruby with simics 2.2.19
(Solaris 10) and to characterize the time spent in synchronization, I
have instrumented the synchronization calls like locks and barrier. I
have binded each thread to a specific processor (one-to-one) and
collecting the time by calling high resolution timer gethrtime().  In
real machine run (having 16 processors) for 16 threads I get around 19%
time spent on synchronization for a program. If I run the same program
in simics without ruby, I also get similar percentage of time spent in
synchronization.

But If I run the same program in simics with ruby, the time spent in
synchronization is much higher (goes around 75% of total).  I have
collected the time from both programs and from ruby. Both are getting
almost same percentage number. I am using MESI_SCMP_directory like
protocol having 2 cycles for L1 and 14 cycles for L2 access.

Does anyone have any idea what's going on? What wrong with my setup? I
would really appreciate your reply.
Thanks,
Hemayet

_______________________________________________
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.