From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 8006 invoked by alias); 10 Aug 2006 21:11:57 -0000 Received: (qmail 7980 invoked by uid 48); 10 Aug 2006 21:11:49 -0000 Date: Thu, 10 Aug 2006 21:11:00 -0000 Message-ID: <20060810211149.7979.qmail@sourceware.org> X-Bugzilla-Reason: CC References: Subject: [Bug target/28623] [4.1/4.2 regression] ICE in extract_insn, at recog.c:2077 (nrecognizable insn) [alpha] In-Reply-To: Reply-To: gcc-bugzilla@gcc.gnu.org To: gcc-bugs@gcc.gnu.org From: "falk at debian dot org" Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org X-SW-Source: 2006-08/txt/msg00836.txt.bz2 List-Id: ------- Comment #2 from falk at debian dot org 2006-08-10 21:11 ------- Confirmed. This is a smaller test case: int vformat(char *buffer) { return buffer[32767]; } This needs -mcpu=ev45. It is triggered by the synthetization of the 8-bit load by 64-bit loads. ldb v0, x(a0) is done as lda a0,x+1(a0) ldq_u v0,-1(a0) extqh v0,a0,v0 sra v0,56,v0 and if x=32767, this fails. (Incidentally, maybe we should do ldq_u t1,x(a0) lda a0,x+1(a0) extqh t1,a0,t1 sra t1,56,v0 and save one cycle at the cost of one extra register.) -- falk at debian dot org changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |rth at gcc dot gnu dot org Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28623