Date: Tue, 10 Sep 1996 11:26:52 +0200 (SAT) From: rv@groa.uct.ac.za (Russell Vincent) To: smp@csn.net Cc: peter@spinner.dialix.com, freebsd-smp@freebsd.org Subject: Re: Intel XXpress - some SMP benchmarks Message-ID: <m0v0P5p-0004vdC@groa.uct.ac.za> In-Reply-To: <m0v0NvE-0004vdC@groa.uct.ac.za> from "Russell Vincent" at Sep 10, 96 10:11:52 am
next in thread | previous in thread | raw e-mail | index | archive | help
I wrote: [ With regard to the write-back L2 cache affecting the SMP code ] > Sorry, I thought I had mentioned it to you. I did eventually find > an option to disable the cache completely, but it didn't help with > the problems we were having. Aaaaaarrggh! I still happen to have the machine, so just double checked this with the latest code and changing the write-back to write-through cache has fixed the problem. The machine now runs fine. The debug options are set to: # define TRACE_CPU_SWITCH_NOT # define FAKE_MP_NOT /* do nothing but return */ # define REAL_MP_NOT /* print data */ # define DO_MP_CALL_NOT /* push regs, call mpX(), pop regs */ # define PUSH_REGS_NOT /* push regs, pop regs */ I am not sure what happened the last time I tried - perhaps I had something else wrong in the code. If the machine doesn't get removed under my feet, I will post a new set of benchmarks, including this new breakthrough. -Russell
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?m0v0P5p-0004vdC>