Re: optimization/8873: Optimization on alpha (ev5)

Re: optimization/8873: Optimization on alpha (ev5)

[Falk Hueffner]

The following reply was made to PR optimization/8873; it has been noted by GNATS.

From: Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
To: martin@bruli.net
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: optimization/8873: Optimization on alpha (ev5)
Date: 09 Dec 2002 00:40:54 +0100

 martin@bruli.net writes:
 
 > 	Compiling glibc-2.3.1 using -mcpu=ev5 generates code that does
 > 	not run correctly on alpha/ev5. 
 
 In which way?
 
 > 	Compiling ssh-3.1 using -mcpu=ev5 generates code that does not
 > 	run at all.
 
 Could you be more specific? Does it crash? With what signal?
 
 -- 
 	Falk

Re: optimization/8873: Optimization on alpha (ev5)

[martin]

The following reply was made to PR optimization/8873; it has been noted by GNATS.

From: <martin@bruli.net>
To: <gcc-gnats@gcc.gnu.org>
Cc:  
Subject: Re: optimization/8873: Optimization on alpha (ev5)
Date: Fri, 7 Mar 2003 10:30:18 +0100

  > The last gcc version you can compile and run all off
  > - linux-2.4
  > - glibc-2.2.5
  > - modutils-2.4
  > is 2.95.3. All later version will generate invalid (not runnable
  > code) on my alphas. And I have tested _all_ versions after 2.95.3.
  > No I went back to 2.95.3 with the disadvantage that the most
  > recent glibc version that can be compiled is 2.2.5.
  
  Which versions did you try?
 
 All.
 
 - 3.0.1
 - 3.0.2
 - 3.0.3
 - 3.0.4
 - 3.1
 - 3.1.1
 - 3.2
 - 3.2.1
 
 This took me about one year! Because after compile and install
 of the compile I rebuilt or at least I tried to rebuild all of
 
 1. binutils
 2. gcc again, with the new binutils from step 1
 3. linux kernel
 4. modutils
 5. glibc-2.2.5
 
 After the rebuild of all this I run the new installation and
 tested out some applications including
 
 - apache
 - mysql
 - squid
 - ssh
 - wu-ftpd
 - openssl
 - ISC bind
 - gdb
 
 The test consisted of compile/build, install and run some test
 scripts comparing the results to the correct results generated
 on a gcc-2.95.3 based installation.
 
 When ever I found fatal errors I went to the next gcc version
 starting at the beginning of the build process but it got worse
 from version to version.
 
 At the end I gave up and reinstalled the system totally based on
 gcc-2.95.3. - That's the story.
 
 
 Regards,
 
 -- Martin
 

Re: optimization/8873: Optimization on alpha (ev5)

[Eric Botcazou]

The following reply was made to PR optimization/8873; it has been noted by GNATS.

From: Eric Botcazou <ebotcazou@libertysurf.fr>
To: <martin@bruli.net>
Cc: <gcc-bugs@gcc.gnu.org>,
 <nobody@gcc.gnu.org>,
 <gcc-gnats@gcc.gnu.org>
Subject: Re: optimization/8873: Optimization on alpha (ev5)
Date: Wed, 19 Feb 2003 09:55:42 +0100

 > No I can't.
 
 Then this will be very tricky to sort out the problem.
 
 > The last gcc version you can compile and run all off
 > - linux-2.4
 > - glibc-2.2.5
 > - modutils-2.4
 > is 2.95.3. All later version will generate invalid (not runnable
 > code) on my alphas. And I have tested _all_ versions after 2.95.3.
 > No I went back to 2.95.3 with the disadvantage that the most
 > recent glibc version that can be compiled is 2.2.5.
 
 Which versions did you try?
 
 -- 
 Eric Botcazou

RE: optimization/8873: Optimization on alpha (ev5)

[martin]

The following reply was made to PR optimization/8873; it has been noted by GNATS.

From: <martin@bruli.net>
To: <ebotcazou@gcc.gnu.org>, <gcc-bugs@gcc.gnu.org>, <gcc-prs@gcc.gnu.org>,
       <martin@bruli.net>, <nobody@gcc.gnu.org>, <gcc-gnats@gcc.gnu.org>
Cc:  
Subject: RE: optimization/8873: Optimization on alpha (ev5)
Date: Tue, 18 Feb 2003 18:14:32 +0100

 Hi,
 
 No I can't.
 The last gcc version you can compile and run all off
 - linux-2.4
 - glibc-2.2.5
 - modutils-2.4
 is 2.95.3. All later version will generate invalid (not runnable
 code) on my alphas. And I have tested _all_ versions after 2.95.3.
 No I went back to 2.95.3 with the disadvantage that the most
 recent glibc version that can be compiled is 2.2.5.
 
 Regards,
 Martin
 
 > -----Original Message-----
 > From: ebotcazou@gcc.gnu.org [mailto:ebotcazou@gcc.gnu.org]=20
 > Sent: Tuesday, February 18, 2003 2:40 PM
 > To: gcc-bugs@gcc.gnu.org; gcc-prs@gcc.gnu.org;=20
 > martin@bruli.net; nobody@gcc.gnu.org
 > Subject: Re: optimization/8873: Optimization on alpha (ev5)
 >=20
 >=20
 > Synopsis: Optimization on alpha (ev5)
 >=20
 > State-Changed-From-To: open->feedback
 > State-Changed-By: ebotcazou
 > State-Changed-When: Tue Feb 18 13:40:08 2003
 > State-Changed-Why:
 >     We need more information to identify the problem. Could=20
 > you narrow it down to a specific chunk of code?
 >=20
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=3Dview%20audit-trail&database=3D=
 gcc&p
 r=3D8873
 

Re: optimization/8873: Optimization on alpha (ev5)

[ebotcazou]

Synopsis: Optimization on alpha (ev5)

State-Changed-From-To: open->feedback
State-Changed-By: ebotcazou
State-Changed-When: Tue Feb 18 13:40:08 2003
State-Changed-Why:
    We need more information to identify the problem. Could you narrow it down to a specific chunk of code?

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8873

optimization/8873: Optimization on alpha (ev5)

[martin]

>Number:         8873
>Category:       optimization
>Synopsis:       Optimization on alpha (ev5)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Sun Dec 08 07:16:00 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Martin Brulisauer
>Release:        3.2.1
>Organization:
MBE
>Environment:
System: Linux alps 2.4.19 #3 Mon Nov 25 10:57:02 MET 2002 alpha unknown
Architecture: alpha

host: alphaev5-bruli-linux-gnu
build: alphaev5-bruli-linux-gnu
target: alphaev5-bruli-linux-gnu
configured with: ./configure --prefix=/usr --exec-prefix=/usr --host=alphaev5-bruli-linux --enable-threads=posix --enable-shared --disable-static
>Description:

	Compiling glibc-2.3.1 using -mcpu=ev5 generates code that does
	not run correctly on alpha/ev5. 
	Compiling ssh-3.1 using -mcpu=ev5 generates code that does not
	run at all.

>How-To-Repeat:

	Compile the source using -mcpu=ev5 on a ev5 machine and try
	to run this code.
>Fix:

	Using -mcpu=ev4 generates runnable code. (The advantages of
	the ev5 over the ev4 is lost).

>Release-Note:
>Audit-Trail:
>Unformatted: