Re: 64bit changes, compiler issues

[John W. Eaton]

On 27-Apr-2005, Clinton Chee <address@hidden> wrote:

| What version of ifort and icc/icpc do you use in Itanium, for your
| successful build?

The following is installed on the system I have access to:

  $ ifort -V
  Intel(R) Fortran Itanium(R) Compiler for Itanium(R)-based applications
  Version 8.0   Build 20040716 Package ID: l_fc_pc_8.0.046_pl050.1
  Copyright (C) 1985-2004 Intel Corporation.  All rights reserved.

  $ uname -a
  Linux XXX 2.4.21-sgi304r1 #2 SMP Wed Mar 30 16:12:04 EST 2005 ia64 ia64 ia64 
GNU/Linux

I built from the current CVS files earlier today with the commands

  configure --enable-64 F77=ifort FFLAGS="-i8 -O"
  make

This is on a system without ATLAS or other fast blas/lapack library
installed.  The resulting binary is linked against the following
libraries:

  liboctinterp.so => /usr/lib/liboctinterp.so (0x2000000000058000)
  liboctave.so => /usr/lib/liboctave.so (0x2000000000564000)
  libcruft.so => /usr/lib/libcruft.so (0x200000000082c000)
  libreadline.so.4 => /usr/lib/libreadline.so.4 (0x2000000000900000)
  libncurses.so.5 => /usr/lib/libncurses.so.5 (0x2000000000974000)
  libdl.so.2 => /lib/libdl.so.2 (0x2000000000a14000)
  libz.so.1 => /usr/lib/libz.so.1 (0x2000000000a2c000)
  libm.so.6.1 => /lib/libm.so.6.1 (0x2000000000a58000)
  libifport.so.6 => /usr/local/intel-8.0-20040412/lib/libifport.so.6 
(0x2000000000aec000)
  libifcoremt.so.6 => /usr/local/intel-8.0-20040412/lib/libifcoremt.so.6 
(0x2000000000b34000)
  libimf.so.6 => /usr/local/intel-8.0-20040412/lib/libimf.so.6 
(0x2000000000d14000)
  libcxa.so.6 => /usr/local/intel-8.0-20040412/lib/libcxa.so.6 
(0x2000000000ea8000)
  libunwind.so.6 => /usr/local/intel-8.0-20040412/lib/libunwind.so.6 
(0x2000000000f14000)
  libpthread.so.0 => /lib/libpthread.so.0 (0x2000000000f48000)
  libstdc++.so.5 => /usr/lib/libstdc++.so.5 (0x2000000000ffc000)
  libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x200000000115c000)
  libc.so.6.1 => /lib/libc.so.6.1 (0x2000000001190000)
  libgpm.so.1 => /usr/lib/libgpm.so.1 (0x2000000001418000)
  /lib/ld-linux-ia64.so.2 => /lib/ld-linux-ia64.so.2 (0x2000000000000000)

It passed failed 4 of the 1207 tests run by make check:

  FAIL: octave.test/io/load-save.m
  FAIL: octave.test/linalg/qr-7.m
  FAIL: octave.test/matrix/rand-1.m
  FAIL: octave.test/matrix/randn-1.m

It's not surprising that it failed the load-save test since load and
save are not completely working for 64-bit systems.  Probably rand
failed because ints in the Fortran random number generator are
expected to be 32-bits wide, or because of some other initialization
problem (I see that it works when I run the test interactively).  The
The qr test is failing the accuracy test at the end of the qr-7.m
file.

jwe