technical:whitepaper:darwin_ucx_openmpi

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision		 Previous revision
technical:whitepaper:darwin_ucx_openmpi [2021-02-12 17:07] – [Older Open MPI Releases] freytechnical:whitepaper:darwin_ucx_openmpi [2021-02-12 17:15] (current) – frey
Line 240: Line 240:
 End Module End Module
 </file> </file>
 +The ''MaxStride8Byte'' constant represents 1.0 GiB worth of 8-byte entities (a double-precision real on x86_64 is 8 bytes wide).
 +
 +Using this module, the previous code is transformed to:
 +<code fortran>
 +      Use mpi_utils
 +        :
 +      Allocate(Rmat(2,144259970)
 +        :
 +      Call BroadcastR(Rmat, 2*144259970, 0, 0, MPI_COMM_WORLD, mpierr)
 +</code>
 +The data is now broadcast as a 268435456-element chunk followed by a 20084484-element chunk, both of which are well below the 1.0 GiB limit associated with the OFI MTL.  The efficiency of MTL should be well in excess of the overhead involved in the "chunking," making this an efficient broadcast mechanism on older Open MPI releases on DARWIN.
  • technical/whitepaper/darwin_ucx_openmpi.1613167645.txt.gz
  • Last modified: 2021-02-12 17:07
  • by frey