Date: Tue, 23 Oct 2001 08:55:49 +0100 (BST) From: Doug Rabson <dfr@nlsystems.com> To: Marcel Moolenaar <marcel@xcllnt.net> Cc: Peter Wemm <peter@wemm.org>, <ia64@FreeBSD.ORG> Subject: Re: Hazards [was: Re: cvs commit: src/sys/ia64/ia64 sal.c] Message-ID: <20011023085414.H549-100000@salmon.nlsystems.com> In-Reply-To: <20011022200334.B566@dhcp01.pn.xcllnt.net>
index | next in thread | previous in thread | raw e-mail
On Mon, 22 Oct 2001, Marcel Moolenaar wrote: > On Mon, Oct 22, 2001 at 04:57:27PM -0700, Peter Wemm wrote: > > Peter Wemm wrote: > > > as (2.9 and 2.11.2) report: > > > pmap.s:1230: Warning: Use of 'ld8' violates RAW dependency 'DTC' (data) > > > pmap.s:1230: Warning: Only the first path encountering the conflict is report > > ed > > > pmap.s:1213: Warning: This is the location of the conflicting usage > > > pmap.s:3463: Warning: Use of 'ld8' violates RAW dependency 'DTC' (data) > > > pmap.s:3463: Warning: Only the first path encountering the conflict is report > > ed > > > pmap.s:3457: Warning: This is the location of the conflicting usage > > > pmap.s:3467: Warning: Use of 'st8' violates RAW dependency 'DTC' (data) > > > pmap.s:3467: Warning: Only the first path encountering the conflict is report > > ed > > > pmap.s:3457: Warning: This is the location of the conflicting usage > > > > [..] > > > > > I haven't finished looking for the info about this.. Is this real? > > > It seems that gas is complaining about an instruction that it thinks > > > should be there but is missing. > > > > It seems gas is expecting a srlz.d, but it looks like it isn't necessary in > > these cases.. The first bit of code is straight from 16.2.2.2.2 in > > volume 2, and the final critical_exit() has the srlz.d. I assume this isn't > > necessary since we know that we're not going to be accessing the user > > areas during this loop. > > I don't think dependencies depend on user areas or not. There > simply is a load following a ptc and according to Appendix A, > volume 2 (page A-7) this means that there must be data > serialization (srlz.d). I took this to mean that there should be serialization between a ptc and a reader of the va that the ptc covered. I don't think serialization is really needed here but it should be harmless to add a couple of srlz.d instructions to ia64_cpu.h. -- Doug Rabson Mail: dfr@nlsystems.com Phone: +44 20 8348 6160 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-ia64" in the body of the messagehelp
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011023085414.H549-100000>