Date: Thu, 10 Jul 2003 15:03:41 -0700 (PDT) From: Julian Elischer <julian@elischer.org> To: Marcel Moolenaar <marcel@xcllnt.net> Cc: FreeBSD current users <current@freebsd.org> Subject: Re: small scheduler hack/patch Message-ID: <Pine.BSF.4.21.0307101445250.40558-100000@InterJet.elischer.org> In-Reply-To: <20030710214339.GA48547@ns1.xcllnt.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 10 Jul 2003, Marcel Moolenaar wrote: > On Fri, Jul 11, 2003 at 07:21:16AM +1000, Bruce Evans wrote: > > have MD definitions. Its first arg has type u_int64_t on ia64's and > > u_int on other arches. This is bogus for ia64's since subr_smp.c uses > > u_int for all bitmaps of CPUs, so systems with more than 32 CPUs cannot > > actually work. > > The bogosity is in MI code. Not being able to support 64-way (or higher) > XYZ machines because MI code uses 32-bit bitmaps is wrong. Both the type > and the access to it should be abstracted in MI code to allow for > compound types. Much akin to sigset_t. I think that the fact that it is even a bitmap should be hidden it comes I think from the fact that some hardware treats things as bitmaps. (?) there are lots of cases where the code is doing foreach cpu if (cpu->mask & our_mask) continue; /* skip ourself */ which could easly be if (cpu->number == PCPU_GET(cpu_number)) i.e there are a lot of cases where a mask is not the natural way to express the information being used.. abstracting it to cpu_is_us(pcpu) makes it more readable and more MI > > -- > Marcel Moolenaar USPA: A-39004 marcel@xcllnt.net >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0307101445250.40558-100000>