Re: [Gems-users] question about XACT_LOG_BUFFER_SIZE parameter

[cge@xxxxxxxx]

I add beginEscapeAction in hardwareAbort.
Thanks for the advice.

Does XACT_LOG_BUFFER refer to a cache in the processor?
It will store the logs in the buffer cache without accessing memory when
writing logs as long as the log size is smaller than the buffer size.
Is my understanding correct?
If so, it will also improve the performance when logging by avoiding  
accessing memory.
By the way, are there any overheads by setting the buffer size too large ?

Thanks,
Cen

Quoting Jayaram Bobba <bobba@xxxxxxxxxxx>:

> Ruben's approach seems better.
> Add a beginEscapeAction() call in hardwareAbort()...
>
> Jayaram Bobba wrote:
> > Simplest fix would be to replace
> >
> > assert(m_escapeDepth[thread] >= 1)
> > with
> > if (m_escapeDepth[thread] < 1) return;
> >
> > in TransactionInterfaceManager.C:335
> >
> > Better fix would be to not implicitly end an escape action when using
> > hardware rollback (in restartTransactionCallback()).
> > We will try to fix it in the next release. Thanks for pointing it out.
> >
> > Jayaram
> >
> > cge@xxxxxxxx wrote:
> > > Thank you for the response.
> > >
> > > When I set it to 256.
> > > I got the following exception:
> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > simics-common: log_tm/TransactionInterfaceManager.C:335: void
> > > TransactionInterfaceManager::endEscapeAction(int): Assertion
> > > `m_escapeDepth[thread] >= 1' failed.
> > > Abort (SIGABRT) in main thread
> > > The simulation state has been corrupted. Simulation cannot continue.
> > > Please restart Simics.
> > > Starting command line. (May have skipped commands in script files.)
> > > [cpu3] v:0x0000000000011814 p:0x0003cf95814  magic (sethi 0x801, %g0)
> > > Setting new inspection cpu: cpu3
> > > Traceback (most recent call last):
> > >    File "../../../gen-scripts/mfacet.py", line 308, in
> > > console_branch_internal
> > >      wait_for_string(get_console(), __prompt)
> > >    File
> > > "/s/csls05/csl/virtutech/simics-3.0.22/x86-linux/lib/python/text_console_common.py",
> > > line 10, in  wait_for_string
> > >      wait_for_obj_hap("Xterm_Break_String", obj, break_id)
> > >    File
> > > "/s/csls05/csl/virtutech/simics-3.0.22/x86-linux/lib/python/cli_impl.py",
> > > line  3371, in wait_for_obj_hap
> > >      return wait_for_hap_common([hap_name, name, idx0])
> > >    File
> > > "/s/csls05/csl/virtutech/simics-3.0.22/x86-linux/lib/python/cli_impl.py",
> > > line  3349, in wait_for_hap_common
> > >      raise SimExc_Break, "Script branch interrupted"
> > > sim_core.SimExc_Break: Script branch interrupted
> > > Exception in python branch
> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > >
> > > There is an assertion fail in end_escape_action.
> > > It does not happen if I set the buffer size to be small like 0 or 16.
> > >
> > > In ruby/simics/commands.C
> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > if ((val == 0x122) && xact_mgr->shouldTrap(smt_thread_num)){
> > >      // use software handler
> > >      if (xact_mgr->shouldUseHardwareAbort(smt_thread_num)){
> > >          xact_mgr->hardwareAbort(smt_thread_num);
> > >      } else {
> > >        xact_mgr->trapToHandler(smt_thread_num);
> > >      }
> > >    }
> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > I think it is because setting the buffer size big enough will make
> > > shouldUseHardwareAbort return true.
> > > However, hardwareAbort does not call beginEscapeAction
> > > while trapToHandler does.
> > >
> > > Any comments how to correct it?
> > >
> > > Thanks,
> > > Cen
> > >
> > > Quoting Jayaram Bobba <bobba@xxxxxxxxxxx>:
> > >
> > >
> > > >  From the wiki,
> > > > *
> > > > XACT_LOG_BUFFER_SIZE*: (>=0) The size of a magical hardware log buffer.
> > > > Used to perform 0-cycle restoration of memory values for any
> > > > transaction
> > > > having logs not exceeding this buffer size.
> > > >
> > > > Basically, it allows zero-cycle aborts for small transactions.
> > > >
> > > > You would need a bigger log buffer for transactions to be able to use
> > > > it. In LogTM-SE, each transaction starts
> > > > off by writing a transaction header which itself is 68 bytes. So you
> > > > would probably need something around 256 bytes
> > > > to be useful...
> > > >
> > > > What kind of exception are you seeing when value is set to 1024?
> > > >
> > > > Jayaram
> > > >
> > > > cge@xxxxxxxx wrote:
> > > >
> > > > > Hi,
> > > > >
> > > > > Does anyone know how can XACT_LOG_BUFFER_SIZE parameter enhance the
> > > > > performance in
> > > > > MESI_CMP_filter_directory protocol.
> > > > > I tried to set the value to 16 or 32, and see no performance
> > > > > difference in terms of cycles when running micro-program.
> > > > >
> > > > > Also when I set the parameter to a large value like 1024, an exception
> > > > > will be thrown. Is there an allowable maximum value?
> > > > >
> > > > > Thanks,
> > > > > Cen
> > > > > _______________________________________________
> > > > > 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.
> > >
> > > _______________________________________________
> > > 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.