Cell Messaging Layer / News: Recent posts

Version 2.6 of the Cell Messaging Layer (CML) is now available from
SourceForge. CML is an extremely fast, MPI-like communication library
for clusters of Cell Broadband Engine processors. With it, any Cell
synergistic processing element (SPE) can communicate directly with any
other SPE, even across a network. The latest release fixes a number
of bugs in the code, so if you had been having trouble with previous
versions of CML, please give v2.6 a shot. The Cell Messaging Layer is
available for download from
http://sourceforge.net/projects/cellmessaging/files/ or by following
the links from the project's home page,
http://cellmessaging.sourceforge.net/.

The Cell Messaging Layer is an extremely fast, MPI-like communication library for clusters of Cell Broadband Engine processors. With it, any Cell synergistic processing element (SPE) can communicate directly with any other SPE, even across a network.

Version 2.4 of the Cell Messaging Layer (CML) is now available from
SourceForge. CML is a message-passing library that simplifies
programming clusters of Cell processors (as used, for example, in the
PlayStation 3 and in LANL's Roadrunner supercomputer). CML 2.4 fixes
a sporadic hang-on-exit bug, simplifies SPE-to-PPE RPC calls by
allowing CML_BYTE_SWAP_NOT_NEEDED as the unit of byte-swapping,
cooperates better with applications by using mfc_tag_reserve() to
allocate DMA tags, and refactors the collective-communication code to
use less memory in some cases.... read more

Version 2.3 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

Version 2.2 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

Version 2.1 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

The latest release of the Cell Messaging Layer (2.0) has run
successfully on Roadrunner, the world's fastest supercomputer and the
first supercomputer to reach a sustained petaflop/s of performance on
the LINPACK benchmark. A CML-based implementation of the Sweep3D
neutron-transport code ran without a hitch on 69,120 SPE cores (12
compute units * 180 nodes/CU * 2 Cell blades/node * 2 Cells/blade * 8
SPEs/Cell), demonstrating that the Cell Messaging Layer scales to
clusters of thousands of Cells or tens of thousands of SPE cores.... read more

Version 1.3 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs. ... read more

Version 1.2 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

Version 1.1 of the Cell Messaging Layer is now available from
SourceForge. The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

Version 1.0 of the Cell Messaging Layer is now available from
SourceForge! The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more

The first prerelease of the Cell Messaging Layer is now available from
SourceForge! The Cell Messaging Layer is an MPI-like
messaging-passing library for clusters of Cell Broadband Engine
processors that makes the entire system look like a homogeneous
cluster of SPE (synergistic processing element) cores. Any SPE can
communicate directly with any other SPE, regardless of physical
location. At the time of this writing, the Cell Messaging Layer is
the world's fastest MPI implementation for the Cell SPEs.... read more