Charm++ parallel programming system

[Xuehan Xu <xxhdx1985126 AT gmail.com>]

Hi, sir:-)
     Sorry for keep on asking, I have no means to bother you, but I'm
still confused about the record-replay feature of bigsim.
     I read your papers----"Simulating Large Scale Parallel
Applications using Statistical Models for Sequential Execution Blocks"
and "Robust Non-Intrusive Record-Replay with Processor Extraction".
According to these two papers, bigsim emulator should be able to
"replay the application and record the detailed traces". I read the
source code of bgMain, and find the two command-line arguments:
"+bgrecord" and "+bgreplay", according to the description of both
arguments, it seems that "+bgrecord" is to let bigsim record the
message ordering of the simulated application, and "+bgreplay" is to
replay the process that ran on the specific target processor. But it
seems that there is no arguments to let bigsim emulator record the
detailed traces.
     I'm very confused, sir. Please help me, thank you very much:-)

On 15/10/2011, Xuehan Xu 
<xxhdx1985126 AT gmail.com>
 wrote:
> Thank you very much, sir:-)
>
> On 15/10/2011, Gengbin Zheng 
> <zhenggb AT gmail.com>
>  wrote:
>> Xuehan,
>>
>>  If you are using startTraceBigSim and endTraceBigSim already, you
>> don't need to call BgMark(). Those two functions already call it.
>>
>> Gengbin
>>
>> On Fri, Oct 14, 2011 at 8:07 PM, Xuehan Xu 
>> <xxhdx1985126 AT gmail.com>
>> wrote:
>>> Thank you, sir:-)
>>> So, I have to call BgMark in the code when running the emulator, and
>>> call
>>> startTraceBigSim() and endTraceBigSim() in the same code when running it
>>> on
>>> a real machine or some cycle accurate simulator, so that the
>>> interpolationtool can rewrite the bgTrace log correctly. Is that
>>> correct?
>>>
>>> In addition, sir, in the BigSim Emulator Mannual, it says that
>>> startTraceBigSim() and endTraceBigSim() can be called at most only once
>>> in
>>> any SEB. What if I have an SEB with a loop? For example:
>>>
>>> while(....){
>>> recv();
>>> some sequential execution code;
>>> send();
>>> }
>>>
>>> What should I do to log the execution time between the recv() call and
>>> the
>>> send() call?
>>>
>>> On 15 October 2011 03:28, Gengbin Zheng 
>>> <zhenggb AT gmail.com>
>>>  wrote:
>>>>
>>>> You need to insert call:
>>>>  BgMark("a marker string");
>>>>
>>>> in your code to insert the markings.
>>>>
>>>> Gengbin
>>>>
>>>> On Fri, Oct 14, 2011 at 12:02 PM, Xuehan Xu 
>>>> <xxhdx1985126 AT gmail.com>
>>>> wrote:
>>>> > Dear Sirs:
>>>> >        Sorry for keep on asking... But I'm really confused
>>>> >
>>>> >        I read the interpolatelog.C file in the directory
>>>> > "example/bigsim/tools/rewritelog/", it seems that it can only
>>>> > interpolate
>>>> > the bgTrace when the bgTrace files themselves have bracket timings.
>>>> > How
>>>> > to
>>>> > make the emulator log the bracket timings?
>>>> >
>>>> > _______________________________________________
>>>> > charm mailing list
>>>> > charm AT cs.uiuc.edu
>>>> > http://lists.cs.uiuc.edu/mailman/listinfo/charm
>>>> >
>>>> > _______________________________________________
>>>> > ppl mailing list
>>>> > ppl AT cs.uiuc.edu
>>>> > http://lists.cs.uiuc.edu/mailman/listinfo/ppl
>>>> >
>>>> >
>>>
>>>
>>
>