Re: [Starpu-devel] StarPU: Data flow between tasks

[Ankur Sharma <ankur.sharma@mailbox.tu-dresden.de>]

Hi,

Thanks for the reply.

I actually want to access this information from the tasks so that I can write 
them out to some file descriptor. In a format like

taskid
input1:         taskid 1        ->      datasize1       (dependency 1)
                taskid 2        ->      datasize2 (dependency 2)
                …..

output:         taskid3 ->      datasize3
                taskid4 ->      datasize4
                …


Sincere regards
Ankur Sharma

On 30 Sep 2014, at 18:51, Samuel Thibault <samuel.thibault@ens-lyon.org> 
wrote:

> Ankur Sharma, le Thu 25 Sep 2014 08:53:28 +0200, a écrit :
>> Is it possible to measure the size of partitioned data allotted to each 
>> task?
> 
> Where do you want to get the information?  Tasks themselves can access that
> information from the descriptor passed to the codelet.  We haven't 
> implemented
> getting the information on the generated task graph.
> 
>> And is it possible to get the id of task that executes the cpu or cuda 
>> implementation?
> 
> From the codelet you can access the task by using 
> starpu_task_get_current(), and
> thus its ID.
> 
> Samuel