Irix6.5 profiling

Irix6.5 profiling

[Tim C Prince]

I note that the -g compiler option enables profiling on Irix6.5.  Is
this an accident or has it been documented some place which I
missed?  A useful capability which I had thought didn't exist.
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

Irix6.5 profiling

[Tim C Prince]

I note that the -g compiler option enables profiling on Irix6.5.  Is
this an accident or has it been documented some place which I
missed?  A useful capability which I had thought didn't exist.
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

g77/x86 unroll-loops performance

[Tim C Prince]

The performance of -O2 has improved significantly in recent
snapshots, emphasizing some of the cases where the compiler
stumbles over unrolling at -O2.  Is there an adjustable switch
which would reduce the tendency to unroll when good
performance is obtained without unrolling?  Kernel 18 is really
more extreme than it appears, as big gains from unrolling one
part are offset by big losses in the other parts.  Pentium II shows
similar comparisons, except that it saturates L1 cache
performance in several kernel tests.

 *********************************************
 THE LIVERMORE  FORTRAN KERNELS "MFLOPS" TEST:
 *********************************************
Run on Pentium Pro 200 Mhz
Compiled on egcs-19990502/g77 -O2 -malign-double
-march=pentiumpro

     -funroll-loops  not unrolled

 KERNEL MFLOP/SEC
------ ---------
  1      86.081       82.372
  2      83.863       77.847
  3     104.491       93.196
  4     132.464       94.817
  5      47.104       46.898
  6      78.903       61.202
  7      74.986       94.291
  8      74.830       79.672
  9      71.458       87.320
 10      18.145       22.260
 11      59.176       61.011
 12      71.013       62.006
 13       6.022        6.012
 14      15.843       17.106
 15      20.360       20.658
 16      34.316       34.711
 17      44.569       40.450
 18      58.901       60.349
 19      48.101       47.506
 20      25.388       27.017
 21     107.685       85.799
 22       9.608        9.631
 23      99.967      100.153
 24      20.315       21.297
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

Limitations on number of open files

[Tim C Prince]

I have ported a legacy application successfully to egcs-1.1.2 g77
under NT4/cygwin, which keeps over 35 data files open and
exercises all conceivable combinations of
write/rewind/read/append/rewind/read.  This program fails the
append part under HPUX10.20, but works correctly on cygwin.
The original depended on non-portable CDC/VAX style data file
handling, and g77 comes closer to supporting this than do
commercial Windows compilers. g77 also runs much faster on
Intel than other compilers available to me.  In spite of my
correcting numerous flagged un-initialized variables, the
application still depends on -fno-automatic, so that is definitely a
useful option.  Evidently, there is no fixed limit of less than 30
files here, so I apologize for any remarks I made to the contrary.
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

Re: Limitations on number of open files

[craig]

>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files

[craig]

>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

The first version of append.f wasn't exactly what I meant, but
close enough.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

The first version of append.f wasn't exactly what I meant, but
close enough.  Well, in the second version I left out an endfile,
but putting it in doesn't make any difference, but now it's in.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

The first version of append.f wasn't exactly what I meant, but
close enough.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Limitations on number of open files

[Tim C Prince]

I have ported a legacy application successfully to egcs-1.1.2 g77
under NT4/cygwin, which keeps over 35 data files open and
exercises all conceivable combinations of
write/rewind/read/append/rewind/read.  This program fails the
append part under HPUX10.20, but works correctly on cygwin.
The original depended on non-portable CDC/VAX style data file
handling, and g77 comes closer to supporting this than do
commercial Windows compilers. g77 also runs much faster on
Intel than other compilers available to me.  In spite of my
correcting numerous flagged un-initialized variables, the
application still depends on -fno-automatic, so that is definitely a
useful option.  Evidently, there is no fixed limit of less than 30
files here, so I apologize for any remarks I made to the contrary.
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

g77/x86 unroll-loops performance

[Tim C Prince]

The performance of -O2 has improved significantly in recent
snapshots, emphasizing some of the cases where the compiler
stumbles over unrolling at -O2.  Is there an adjustable switch
which would reduce the tendency to unroll when good
performance is obtained without unrolling?  Kernel 18 is really
more extreme than it appears, as big gains from unrolling one
part are offset by big losses in the other parts.  Pentium II shows
similar comparisons, except that it saturates L1 cache
performance in several kernel tests.

 *********************************************
 THE LIVERMORE  FORTRAN KERNELS "MFLOPS" TEST:
 *********************************************
Run on Pentium Pro 200 Mhz
Compiled on egcs-19990502/g77 -O2 -malign-double
-march=pentiumpro

     -funroll-loops  not unrolled

 KERNEL MFLOP/SEC
------ ---------
  1      86.081       82.372
  2      83.863       77.847
  3     104.491       93.196
  4     132.464       94.817
  5      47.104       46.898
  6      78.903       61.202
  7      74.986       94.291
  8      74.830       79.672
  9      71.458       87.320
 10      18.145       22.260
 11      59.176       61.011
 12      71.013       62.006
 13       6.022        6.012
 14      15.843       17.106
 15      20.360       20.658
 16      34.316       34.711
 17      44.569       40.450
 18      58.901       60.349
 19      48.101       47.506
 20      25.388       27.017
 21     107.685       85.799
 22       9.608        9.631
 23      99.967      100.153
 24      20.315       21.297
Dr. Timothy C. Prince
Consulting Engineer
Solar Turbines, a Caterpillar Company
alternate e-mail: tprince@computer.org

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

The first version of append.f wasn't exactly what I meant, but
close enough.  Well, in the second version I left out an endfile,
but putting it in doesn't make any difference, but now it's in.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Limitations on number of open files -Reply

[Tim C Prince]

I have attached a script and fortran source which shows this.
Maybe you could tell me if I'm trying something which is not
supported.  I have run it with egcs-1.1.2 and egcs-19990502 on
Irix6.5, hpux10.20, and NT4SP3/cygwin.  The fortran runs silently
(successfully) using the vendor's compilers on the Unix
systems, as well as with g77 on cygwin (yes, over 90 open files).
The failing appending writes don't return any error codes but the
stuff never shows up in the data file.  I imagine this is too much
stuff to send to egcs-bugs, nor would it be relevant without lower
level diagnostics.

>>> craig@jcb-sc.com 05/12/99 06:34am >>>
>This program fails the
>append part under HPUX10.20, but works correctly on cygwin.

Definitely submit the best bug report you can on this problem, in
case we can fix it before the release!

Thanks for the good news in the rest of your email!

        tq vm, (burley)

Re: Results for gcc-2.95 19990608 on hppa1.1-hp-hpux10.20 -Reply

[Jeffrey A Law]

  In message <"376533EB.6E8C.2C8E.000*/c=US/admd= /prmd=Cat/o=GWise/s=Prince/g=
Tim/i=C/"@MHS>you write:
  > Is there a patch for dejagnu which should make it suppress this
  > warning, which appears to be the cause of the Failure
  > indication?
OK.  There's a new dejagnu snapshot being uploaded to the egcs machine
as I type.  I haven't tested it in any way shape or form though.  Feedback
would be appreciated.

Thanks,
jeff

Re: Results for gcc-2.95 19990608 on hppa1.1-hp-hpux10.20 -Reply

[Jeffrey A Law]

  In message <"376533EB.6E8C.2C8E.000*/c=US/admd= /prmd=Cat/o=GWise/s=Prince/g=
Tim/i=C/"@MHS>you write:
  > Is there a patch for dejagnu which should make it suppress this
  > warning, which appears to be the cause of the Failure
  > indication?
OK.  There's a new dejagnu snapshot being uploaded to the egcs machine
as I type.  I haven't tested it in any way shape or form though.  Feedback
would be appreciated.

Thanks,
jeff

Re: Results for gcc-2.95 19990608 on hppa1.1-hp-hpux10.20 -Reply -Reply

[Tim C Prince]

I ran this new dejagnu without any difficulty on Irix6.5, and on
hpux1.1 it is eliminating this objectionable failure indication.
So it seems to be doing the job.

My understanding of tcl is not great enough to figure out what it
means by missing a package path on those targets where the
native dejagnu requires only that TCL_LIBRARY be set and the
runtest.exp have the whoami calls removed.

>>> law@cygnus.com 06/14/99 12:49pm >>>

  In message <"376533EB.6E8C.2C8E.000*/c=US/admd=
/prmd=Cat/o=GWise/s=Prince/g=
Tim/i=C/"@MHS>you write:
  > Is there a patch for dejagnu which should make it suppress
this
  > warning, which appears to be the cause of the Failure
  > indication?
OK.  There's a new dejagnu snapshot being uploaded to the
egcs machine
as I type.  I haven't tested it in any way shape or form though.
Feedback
would be appreciated.

Thanks,
jeff

Re: Results for gcc-2.95 19990608 on hppa1.1-hp-hpux10.20 -Reply -Reply

[Tim C Prince]

I ran this new dejagnu without any difficulty on Irix6.5, and on
hpux1.1 it is eliminating this objectionable failure indication.
So it seems to be doing the job.

My understanding of tcl is not great enough to figure out what it
means by missing a package path on those targets where the
native dejagnu requires only that TCL_LIBRARY be set and the
runtest.exp have the whoami calls removed.

>>> law@cygnus.com 06/14/99 12:49pm >>>

  In message <"376533EB.6E8C.2C8E.000*/c=US/admd=
/prmd=Cat/o=GWise/s=Prince/g=
Tim/i=C/"@MHS>you write:
  > Is there a patch for dejagnu which should make it suppress
this
  > warning, which appears to be the cause of the Failure
  > indication?
OK.  There's a new dejagnu snapshot being uploaded to the
egcs machine
as I type.  I haven't tested it in any way shape or form though.
Feedback
would be appreciated.

Thanks,
jeff