Dear All, I have updated the serial link model to be able to regulate bandwidth and make sure it does not exceed its theoretical limit (Number of Lanes X Clock Frequency). I measure peak bandwidth over epochs of T (periodically), and if the peak bandwidth exceeds the theoretical max value, I stall the incoming transactions. The value of T determines the accuracy of this operation. If it is too small, it will degrade the average bandwidth delivered by the serial link, and if it is too large, it will have no effect and occasionally, peak bandwidth will exceed the theoretical limit.
This mechanism is currently implemented on the request path only. Sincerely, On Mon, Sep 14, 2015 at 3:17 PM, Erfan Azarkhish <[email protected]> wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.gem5.org/r/2993/ > ----------------------------------------------------------- > > (Updated Sept. 14, 2015, 1:17 p.m.) > > > Review request for Default. > > > Repository: gem5 > > > Description (updated) > ------- > > Changeset 11096:5ded5036f584 > --------------------------- > Serial Link model for the HMC > > > Diffs (updated) > ----- > > src/mem/SConscript 1a9ecb4fe05e > src/mem/SerialLink.py PRE-CREATION > src/mem/serial_link.hh PRE-CREATION > src/mem/serial_link.cc PRE-CREATION > > Diff: http://reviews.gem5.org/r/2993/diff/ > > > Testing > ------- > > > Thanks, > > Erfan Azarkhish > > _______________________________________________ > gem5-dev mailing list > [email protected] > http://m5sim.org/mailman/listinfo/gem5-dev > -- Erfan Azarkhish Micrel Lab - Viale Carlo Pepoli 3/2 - 40123, Bologna DEI - University of Bologna, Italy https://www.linkedin.com/in/erfanazarkhish _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
