Date: Sun, 25 Apr 1999 19:12:14 +0100 (BST) From: Doug Rabson <dfr@nlsystems.com> To: Dmitrij Tejblum <dima@tejblum.dnttm.rssi.ru> Cc: freebsd-alpha@freebsd.org Subject: Re: "I-stream memory barrier" Message-ID: <Pine.BSF.4.05.9904251900440.28665-100000@herring.nlsystems.com> In-Reply-To: <199904251359.RAA02496@tejblum.dnttm.rssi.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 25 Apr 1999, Dmitrij Tejblum wrote: > It looks to me that our pmap take alpha_pal_imb() wrong. Our pmap > apparently assume that I-cache like TLB is virtually indexed and > tagged with the ASN. So we call alpha_pal_imb() only when start new ASN > generation. NetBSD have completely different idea (I-cache is indexed > by the physical address?) and call alpha_pal_imb() in every second pmap > function. I suspect they are right... > > Comments? I'm pretty sure that I-cache lines are tagged with ASNs. The Linux kernel only calls imb() when it runs out of ASNs and if it didn't work properly, I doubt that a system could survive even a single threaded make world, let alone -j20. This is only relavent to user code since the ASNs are used to distinguish between different user address spaces. I don't think we properly flush the I-cache when kernel executable code changes which is an issue for KLD. Probably KLD should arrange for an IMB on the alpha by calling some machine dependant pmap function. -- Doug Rabson Mail: dfr@nlsystems.com Nonlinear Systems Ltd. Phone: +44 181 442 9037 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9904251900440.28665-100000>