Re: [Gems-users] results of microbenchmark deque-tm

[Jayaram Bobba <bobba@xxxxxxxxxxx>]

'Workload transactions' refer to transactions at a higher semantic level 
(for ex. database transactions)
and do not have any relation to TM transactions that are used for 
synchronization/atomicity.
They are often used in our simulations as a measure of work done by the 
program. Please refer to
http://www.cs.wisc.edu/multifacet/papers/ieeemicro03_variability.pdf
for more info on these 'workload transactions'.
The overlapping nomenclature is unfortunate and we will attempt to address
it in future releases.
Workload transactions are begun by calling ruby magic call #3 and ended 
with magic call #5.
LogTM transactions use a different set of magic call numbers.

Setting PROFILE_XACT to true should give you some information about 
LogTM execution. I cannot comment
on your particular case without more information.

I would also recommend that you upgrade to GEMS 2.0, if possible.

Jayaram

Irene-Aggeliki Chounta wrote:
> hi Jayaram and thanks for the reply.
>  
> Would you kindly help me out with some questions I have? forgive my 
> ignorance, I'm new to these all.
>  
> as you said:
>  
> > Note that the 'transaction' you see on the screen output refers to a
> > workload transaction which is
> > different from LogTM transactions.
> >
>  
> to be honest, that's not clear to me at all. In hpca's 2006 paper of 
> transactional memory is stated that: 
> /"support for the LogTM interface was added using simics magic 
> instructions........To implement the "begin" instruction the memory 
> simulator uses a simics call to create a checkpoint. During a 
> transaction, the memory simulator models the log updates"./
>  
> so when we test a microbenchmark that makes use of BEGIN_TRANSACTION 
> we imply a magic call and then the log updates of LogTM are modeled by 
> the memory system. doesn't this mean that a workload transaction 
> corresponds to a logTM transaction?
> And since they are different, how can I watch the LogTM transactions 
> that took place during the execution? (I already tried to print out 
> some stuff from TransactionManager.C and other programs but with no 
> luck...)
>  
> so, thank you very much for the help.
>  
> --
> Irene,
>  
>  
> > Jayaram
> >
> > Irene-Aggeliki Chounta wrote:
> >> hi everyone,
> >> 
> >> I'm experimenting with protocol MESI_SMP_LogTM and currently running
> >> deque-tm to check out how it works.
> >> 
> >> I make the needed changes to simics script, copy the binary to
> >> simulated machine and execute and then I get in simics host the 
> following:
> >> 
> >> end_transaction_magic_call: Unexpected magic call number 4
> >> end_transaction_magic_call: transaction started: 0, transaction
> >> completed: 0, transaction_limit: 4, None
> >> [cpu2] v:0x0000000000010f58 p:0x00000003f95b4f58  magic (sethi
> >> 0x40000, %g0)
> >> simics>
> >> 
> >> so from what i can understand there's no transaction started due to
> >> this unexpected magic call number 4, is that correct????
> >> 
> >> why is this happening anyway????
> >> 
> >> thanks in advance.
> >> 
> >> ps. happy new year all ;)
> >> 
> >> 
> >> -------------------------------------------------------------------
> >> irene
> >> ......in accordance with the prophecy.
> >> 
> ------------------------------------------------------------------------
> >>
> >> _______________________________________________
> >> Gems-users mailing list
> >> Gems-users@xxxxxxxxxxx <mailto: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 <mailto: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.
>
>