This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
c/3917: gcc for IA-64 displays erroneous assembler timing information
- To: gcc-gnats at gcc dot gnu dot org
- Subject: c/3917: gcc for IA-64 displays erroneous assembler timing information
- From: gbv at ctv dot es
- Date: Thu, 2 Aug 2001 18:54:35 +0200
- Reply-To: gbv at ctv dot es
>Number: 3917
>Category: c
>Synopsis: IA-64 assembler output shows erroneous cycle counting
>Confidential: no
>Severity: non-critical
>Priority: medium
>Responsible: unassigned
>State: open
>Class: pessimizes-code
>Submitter-Id: net
>Arrival-Date: Thu Aug 02 10:06:00 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator: Guillermo Ballester Valor
>Release: 3.0
>Organization:
>Environment:
System: Linux mozart 2.2.19 #2 Sat Jun 2 21:32:31 CEST 2001 i586 unknown
Architecture: i586
host: i586-pc-linux-gnu
build: i586-pc-linux-gnu
target: i586-pc-linux-gnu
configured with: ../gcc-3.0/configure
>Description:
This is no a problem of my i586 system. I'm working with telnet to a machine with RedHat v.7.1 for IA-64 (gcc v.3.0). I've observed assembler output wich differs what I read in manuals.
The bundles with only float point code displays a cycle per boundle.
I thought there is two FPU units and in most single cases the processor can schedule two FPU operation per cycle.
This should be not important whether is only an error on the comment
added to assembler output, and the instructions were scheduled correctly. But
the problem is more important if the schedule parameters are incorrect.
Finally, I would like to say there is a big increased perfomence for
IA-64, from gcc 2.95. In the software I'm developing the speed now almost twice
faster :)
>How-To-Repeat:
Simply compile a program with heavy FP work with -O3 -S option and
see the output.
>Fix:
I don't know.
>Release-Note:
>Audit-Trail:
>Unformatted: