> On Fri, Mar 14, 2008 at 2:21 PM, William Kucharski > <William.Kucharski at sun.com> wrote: >> ken mays wrote: >> >> > The Sony PS3 can be obtained for <=$399 USD and you >> > get a nice Blu-ray player tossed in for free. The >> > 'free' Blu-Ray player and Nvidia RSX GPU are more than >> > worth the price of admission here in the long term and >> > as a feasible solution. Both the old and new Mac Mini >> > runs about $499 USD and has far less capabilities in >> > key areas. >> >> I've mentioned this in the past, but one of the biggest challenges with a >> port >> to PS3, Power6, or any other multi-CPU/multi-core PPC environment is that >> Solaris at present assumes a processor with strong memory ordering. >> >> The current PPC work doesn't run into issues because the ODW is single >> CPU, >> single threaded but to go beyond that a fair amount of work would need to >> be >> done to add appropriate memory synchronization instructions to Solaris >> (isync/sync/eieio), particularly in the case of DMA transfers. > > Not sure how PPC handles multicore, but can we not just pretend that > it's single ( ie, not activate the other processors ) until we're at > the point where that seems like a good idea ?
I don't think that it can be that trivial. The multi-core processors and the memory architecture issues alone are considerable in complexity. see : TSOtool: A Program for Verifying Memory Systems Using the Memory Consistency Model Written by Sudheendra Hangal, Durgam Vahia, Chaiyasit Manovit, Juin-Yeu Joseph Lu and Sridhar Narayanan IEEE Int. Symp. on Computer Architecture (ISCA04), 2004. In this paper, we describe TSOtool, a program to check the behavior of the memory subsystem in a shared memory multiprocessor. TSOtool runs pseudo-randomly generated programs with data races on a system compliant with the Total Store Order (TSO) memory consistency model; it then checks the results of the program against the formal TSO specification. Such analysis can expose subtle memory errors like data corruption, atomicity violation and illegal instruction ordering. http://www.opensparc.net/cgi-bin/goto.php?w=/pubs/papers/tsotool_isca04.pdf and my personal favorite : Memory Model = Instruction Reordering + Store Atomicity Arvind MIT CSAIL 32-G866, 32 Vassar St. Cambridge, MA 02139 arvind at csail.mit.edu Jan-Willem Maessen Sun Microsystems Laboratories UBUR02-311, 1 Network Dr. Burlington, MA 01803 JanWillem.Maessen at sun.com see http://csg.csail.mit.edu/pubs/memos/Memo-493/memo-493.pdf There is nothing remotely trivial in the issues. Dennis
