[MITgcm-support] RE: conversion from local tile index to global new I/O format index

Chris Hill cnh at mit.edu
Tue Mar 30 15:40:14 EST 2004


 as far as I can see, apart from me being brain dead and copying my comments
rather than the code the formula you have
>     i_global_io_index = exch2_txgloballo(W2_myTileList(bi))+i-1
>     j_global_io_index = exch2_tygloballo(W2_myTileList(bi))+j-1
is correct

 For example tile 216 of the 85x85 tile cube has


 That means that the local coord of tile 216 (85,85) would map to the global
coord (3060,510). Not sure what you are talking about with face corners? 

 Note - W2_myTileList is setup at run time. So maybe that's causing some
confusion. The numbers in W2_mytilelist are simply a mapping between a local
per process tile rank {1:nsx} and a global tile rank {1:nsx*npx}. 

 Does that make sense or am I missing something?

 Also could you send the mails to support too, this is the same information
other people will need in the future and if it goes to support it is all
there to be searched etc...


> -----Original Message-----
> From: Dimitris Menemenlis [mailto:menemenlis at jpl.nasa.gov] 
> Sent: Tuesday, March 30, 2004 3:27 PM
> To: 'Andrew Eichmann'
> Cc: Dimitris Menemenlis; Chris Hill; 'Ichiro Fukumori'
> Subject: Re: conversion from local tile index to global new 
> I/O format index
> Andy, so what is the complete expression needed to obtain a 
> global index in your new I/O format.  Could you please write 
> it out explicitly so that I don't mess up.  Thanks, D.

More information about the MITgcm-support mailing list