From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 5514 invoked by alias); 4 Jan 2002 01:59:44 -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 5452 invoked from network); 4 Jan 2002 01:59:41 -0000 Received: from unknown (HELO crocodile-dances.eterna.com.au) (203.15.111.99) by sources.redhat.com with SMTP; 4 Jan 2002 01:59:41 -0000 Received: by crocodile-dances.eterna.com.au (Postfix, from userid 127) id 15F5517EEB; Fri, 4 Jan 2002 12:59:21 +1100 (EST) Received: from cygnus.com (localhost [127.0.0.1]) by crocodile-dances.eterna.com.au (Postfix) with ESMTP id C334A5B83D; Fri, 4 Jan 2002 12:59:21 +1100 (EST) To: Andrew Cagney Cc: gdb-patches@sources.redhat.com subject: re: [patch] AltiVec support for PSIM. in-reply-to: your message of "Thu, 03 Jan 2002 16:28:33 +0930." <3C340119.70201@cygnus.com> organisation: Red Hat, Asia-Pacific. Date: Thu, 03 Jan 2002 17:59:00 -0000 Message-ID: <10879.1010109556@cygnus.com> From: matthew green X-SW-Source: 2002-01/txt/msg00019.txt.bz2 That just leaves the problem of how to generate an altivec enabled simulator from two files. I can think of two alternatives: steal code from the more up-to-date sim/igen generator as that supports a ``:include:'' directive; pull a nasty hack such as `cat ppc-instructions altivec.igen > tmp.igen` comes to mind (it does defeat igen's ability to refer back to source code lines though). yup. i had a quick look at supporting :include: but that has a few issues with $(builddir) != $(srcdir) that don't look trivial to solve and the sim/igen code has changed somewhat... Given that I suspect you're going to need to be able to generate multiple simulators - with / without Altivec, stealing code from sim/igen might be the better medium to long term option. Since, as illustrated by MIPS, sim/igen supports this. for the short term, i'm going to go with the "cat" hack, but i will work on making :include: work in the longer term. as you note, the cat solution loses the line number info. a third solution would be to permit multiple "-i" switches to be passed to igen, but that also looks non-trivial. at least the call to insn_table_expand_insns() would need to be modified somehow, but i could be wrong. thanks, .mrg.