[uClinux-dev] Re: MMU trade-off
First of all: Thanks for the quick, frequent, interesting and informative response!! The numbers mentioned by the developer were a gut feeling and rough estimation, not derived by tests and measurements, but because I had neither I also had no basis to disagree. The CPUs in use are PowerPCs (MPC852 MPC8247) from the PowerQuicc II family with 8MB to 32MB of RAM, thus AFAIK the cache design is descent and the cache has not to be flushed completly on a context switch, right? I'll take your information into account, propose some tests and let you know if we get proper results! Thanks again and best regards, Thomas ___ uClinux-dev mailing list uClinux-dev@uclinux.org http://mailman.uclinux.org/mailman/listinfo/uclinux-dev This message was resent by uclinux-dev@uclinux.org To unsubscribe see: http://mailman.uclinux.org/mailman/options/uclinux-dev
Re: [uClinux-dev] Re: MMU trade-off
Jivin Thomas Häberle lays it down ... First of all: Thanks for the quick, frequent, interesting and informative response!! The numbers mentioned by the developer were a gut feeling and rough estimation, not derived by tests and measurements, but because I had neither I also had no basis to disagree. The CPUs in use are PowerPCs (MPC852 MPC8247) from the PowerQuicc II family with 8MB to 32MB of RAM, thus AFAIK the cache design is descent and the cache has not to be flushed completly on a context switch, right? I'll take your information into account, propose some tests and let you know if we get proper results! I think it might have been posted in one of the threads, but just in case it wasn't, here is a doc that gives numbers to MMU vs. no MMU. http://opensrc.sec.samsung.com/document.html Always good to have some real numbers ;-) Cheers, Davidm -- David McCullough, david_mccullo...@securecomputing.com, Ph:+61 734352815 McAfee - SnapGear http://www.snapgear.comhttp://www.uCdot.org ___ uClinux-dev mailing list uClinux-dev@uclinux.org http://mailman.uclinux.org/mailman/listinfo/uclinux-dev This message was resent by uclinux-dev@uclinux.org To unsubscribe see: http://mailman.uclinux.org/mailman/options/uclinux-dev
Re: [uClinux-dev] Re: MMU trade-off
On Mon, Mar 30, 2009 at 08:10:12PM +1000, David McCullough wrote: Jivin Thomas H?berle lays it down ... First of all: Thanks for the quick, frequent, interesting and informative response!! The numbers mentioned by the developer were a gut feeling and rough estimation, not derived by tests and measurements, but because I had neither I also had no basis to disagree. The CPUs in use are PowerPCs (MPC852 MPC8247) from the PowerQuicc II family with 8MB to 32MB of RAM, thus AFAIK the cache design is descent and the cache has not to be flushed completly on a context switch, right? I'll take your information into account, propose some tests and let you know if we get proper results! I think it might have been posted in one of the threads, but just in case it wasn't, here is a doc that gives numbers to MMU vs. no MMU. http://opensrc.sec.samsung.com/document.html Always good to have some real numbers ;-) Well that's arm. The arm always did seem to have one of the weirdest memory setups. It is the only cpu I have worked with where the OS ran in virtual addresses compared to hardware. Interesting to debug PCI drivers when the addresses the hardware sees and the addresses the OS sees don't match, but have to be mapped through some part of the MMU. The PowerPC seems very straight forward and sane compared to the arm. I hightly doubt the MMU has any significant impact on performance on an PowerPC. I have never tried running a PowerPC without the MMU though, so I have no idea. It runs very will with MMU on plain linux though. -- Len Sorensen ___ uClinux-dev mailing list uClinux-dev@uclinux.org http://mailman.uclinux.org/mailman/listinfo/uclinux-dev This message was resent by uclinux-dev@uclinux.org To unsubscribe see: http://mailman.uclinux.org/mailman/options/uclinux-dev