From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 5334 invoked by alias); 16 Apr 2004 19:15:46 -0000 Mailing-List: contact gdb-patches-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sources.redhat.com Received: (qmail 5313 invoked from network); 16 Apr 2004 19:15:44 -0000 Received: from unknown (HELO localhost.redhat.com) (66.30.197.194) by sources.redhat.com with SMTP; 16 Apr 2004 19:15:44 -0000 Received: from gnu.org (localhost [127.0.0.1]) by localhost.redhat.com (Postfix) with ESMTP id 8E2FB2B9D; Fri, 16 Apr 2004 15:15:39 -0400 (EDT) Message-ID: <408030DB.3080708@gnu.org> Date: Fri, 16 Apr 2004 19:15:00 -0000 From: Andrew Cagney User-Agent: Mozilla/5.0 (X11; U; NetBSD macppc; en-GB; rv:1.4.1) Gecko/20040217 MIME-Version: 1.0 To: Orjan Friberg Cc: Daniel Jacobowitz , gdb-patches@sources.redhat.com Subject: Re: Multiplexed registers and invalidating the register cache References: <407D242B.109@axis.com> <20040414144607.GA5700@nevyn.them.org> <407E67EA.80701@axis.com> <407E70FD.7060709@axis.com> <407EAA4B.7000500@gnu.org> <407FD693.2080804@axis.com> In-Reply-To: <407FD693.2080804@axis.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-SW-Source: 2004-04/txt/msg00371.txt.bz2 > Andrew Cagney wrote: > >> >> Consider the effect of modifying the $sp. >> >> While it might in theory be possible to implement some sort of complicated look-aside cache schema, in reality there is zero return on investment. Since recovery from the flush can't be slower than recovery from single-step, and single step is way way more critical, we should focus on single step. > > > Ok, I'm convinced that the whole frame cache should be discarded, except I must be missing something since calling flush_cached_frames doesn't refetch the current set of registers (like calling registers_changed does). (mumble something about doco) two fyis: - the frame cache is built on-demand, hence the absence of any explict rebuild call (one characteristic of a frame ID is that it survives frame cache flushes) - there shouldn't be separate register and frame flush calls, combining the two into a single observer call is a thing-to-do-today enjoy, Andrew