ARSC T3D Users' Newsletter 13, December 2, 1994

Upcoming Software Upgrade on T3D

ARSC will be upgrading the T3D software to CrayLib_M 1.1.1.2, MAX 1.1.0.4 and SCC_M 4.0.2.11 on December 4th.

By now, users should have run their programs and saved the executables and results. These executables and results should be compared to the results after the new products are installed. If there are erroneous differences then Mike Ess should be contacted.

Using FLUSH to Keep I/O Synchronized

Coert Olmsted of ARSC brought in a puzzler that is illustrated by the following code and results: 

      cat writepenum.f
      intrinsic my_pe
      integer mype

      mype = my_pe()

      write(6,111) mype
      call barrier
      if(mype.eq.0) write(6,222)
      call barrier
      write(6,111) mype

  111 format(1x,i2,$)
  222 format(/)
      end
  /mpp/bin/cf77 writepenum.f
  a.out
  2  2  4  4  3  3  7  7  5  5  6  6  1  1  0

  0
It was intended that the use of the format character "$" would prevent new lines from each PE to the screen output and there would be only two lines of output, with each line having each PE report once. What actually happens is that the writes are executing correctly, obeying the barrier calls but, each PE does writes to its own buffer which is then dumped to the screen when all the PEs terminate.

The solution to get the desired results is to not let results accumulate in the individual buffers but to flush them out at the right time. Now the source looks like: 


      intrinsic my_pe
      integer mype

      mype = my_pe()

      write(6,111) mype
      call flush( 6 )
      call barrier()
      if(mype.eq.0) write(6,222)
      call barrier()
      write(6,111) mype
      call flush( 6 )

  111 format(1x,i2,$)
  222 format(/)
      end
and the expected results are: 

  7  0  5  4  2  6  1  3

  0  1  4  2  6  5  7  3
Using the FLUSH routine, the I/O buffers do not hide the order of events.

Reminders

List of differences between T3D and Y-MP:

The current list of differences between the T3D and the Y-MP is:

  1. Data type sizes are not the same (Newsletter #5)
  2. Uninitialized variables are different (Newsletter #6)
  3. The effect of the -a static compiler switch (Newsletter #7)
  4. There is no GETENV on the T3D (Newsletter #8)
  5. Missing routine SMACH on T3D (Newsletter #9)
  6. Different Arithmetics (Newsletter #9)
  7. Different clock granularities for gettimeofday (Newsletter #11)
I encourage users to e-mail in differences that they have found, so we all can benefit from each other's experience.

PE Limits

At ARSC, most production runs can be handled by the 32 PE NQS queue. The use of large partitions of 64 or 128 PEs seems restricted to research problems of the "what if..." type. Basically, development is done on smaller partitions and then to complete a table of performance or problem sizes, the larger partitions are used.

We have been accommodating this situation by individually changing a user's interactive permissions with the understanding that users would share their results with us. We are in the process of notifying users of the change in this policy. In the future we will reset everyone back to the default settings and then make it easy for users to request changes that will be in effect for a short time period (perhaps a week per request).

Future Newsletter

While at the past Supercomputing '94 conference I got a lot of information on the T3D. It will take me a while to outline it, but eventually you'll see all of it in the upcoming newsletters.

Supplemental Update - Sunday, Dec 4

Upcoming upgrade on ARSC T3D software

ARSC will be upgrading the T3D software to CrayLib_M 1.1.1.2, MAX 1.1.0.4 and SCC_M 4.0.2.11 in the near future.

We attempted this upgrade Sunday, December 4th, but were unsuccessful. We will try again in the near future and ARSC will notify its T3D users through this newsletter distribution when the upgrade is complete.

Second Call for Test Programs

In putting together a regression test for upgrades of T3D software, I found two areas where I need more coverage with user-supplied programs. I would like to ask again for users to submit test programs on:
  1. PVM called from Fortran
  2. Shmem routines called from either Fortran or C
Ideally each program would exercise many different routines, run for only minutes, produce a text file that provides timing information and could be diffed with previous results to show correct results. That's a lot to ask but that's what I need.
Current Editors:
Ed Kornkven ARSC HPC Specialist ph: 907-450-8669
Kate Hedstrom ARSC Oceanographic Specialist ph: 907-450-8678
Arctic Region Supercomputing Center
University of Alaska Fairbanks
PO Box 756020
Fairbanks AK 99775-6020
E-mail Subscriptions: Archives:
    Back issues of the ASCII e-mail edition of the ARSC T3D/T3E/HPC Users' Newsletter are available by request. Please contact the editors.