Indeed that's another solution, it might even be more intelligent to do what you propose instead of frequently do a ASID_FLUSH which is basically stalling the CPU pipeline while iterating over the whole TLB which takes something like 2*1024 clock cycles :)
2013/12/6 Werner Almesberger <wer...@almesberger.net> > Yann Sionneau wrote: > > I'm doing a little experiment with LM32 MMU: adding ASID (Address Space > > ID) support to the TLBs > > Sounds cool. > > > This introduces a new TLB operation : LM32_TLB_OP_ASID_FLUSH. > > But do you actually need that ? Unless you expect to cycle through > ASIDs at a very high rate, Couldn't you, when running out of ASIDs, > just garbage-collect all unused ones and then do a complete flush, > at a very low overall cost since it wouldn't happen very often ? > > - Werner > _______________________________________________ > Devel mailing list > Devel@lists.milkymist.org > https://ssl.serverraum.org/lists/listinfo/devel > -- Yann Sionneau
_______________________________________________ Devel mailing list Devel@lists.milkymist.org https://ssl.serverraum.org/lists/listinfo/devel