Dear Erik,

This note provides a fix that should eliminate the problem you are
seeing with the XTIME information not advancing in the MM5 V3 model
output.

In the file MPP/RSL/RSL/rsl_ioserve.c, in the routine
RSL_WRITE_MM5V3_SM_HEADER, around line 1205, there is a sequence of
code that reads:

  { float x ;
    x = (double)*xtime_p ;
    bcopy( &x  , &(wbuf[icurs]), rwordsize/2 ) ;
  }

Change this to read:

  { float x ; double y ;
    bcopy( xtime_p, &y, rwordsize ) ;
    x = y ;
    bcopy( &x  , &(wbuf[icurs]), rwordsize/2 ) ;
  }

Then, while still in the MPP/RSL/RSL directory, type 'make t3e'. After
this, cd to the top level directory and 'rm Run/mm5.mpp' and then 'make
mpp'.

Please let me know if this works for you. Thanks for reporting this
error and for your patience.

Best regards,

John Michalakes



At 03:25 PM 2/23/00 +0000, Erik Berge wrote:
Dear Wei

As you might now we are running MM5 for norwegian cities operationally
on CRAY T3E. So far we utlized 2.12, but we are now implementing 3.3
on T3E (We are running 3.3 successfully on SGI already).

We have encountered one problem after a run of the 3.3 MPP on T3E
when processing the output data. Parameter "time" in diagmm5
equals 64.00 for every output time interval except for the initial one
which is correctly = 0.00. As we could not find any soulution to this with
our
own set up we wonder if there could be any problems during writing
of the header in the MPP version? This could maybe be the parameter
"xtime" in this routine.

We appreciate your response to the problem


Best regards


Erik Berge