... mechanisms^1.1

Even in this approach it was necessary to make all code developers agree on a common format for data exchange, specifications for invoking various components, and had a glue code that binds all these components together in a single ``application''.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... group.^3.1

This was called a Branch Office Chare, or Branched Chare in earlier versions of Charm++.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... enabled.^3.2

Enabling an object method is different from executing it. Execution occurs under the control of a scheduler, whereas a method is enabled upon availability of its input data. This separation of execution and enabling objects methods is crucial to our interface model, as it provides a control-point for the Charisma runtime system to effectively utilize computational resources.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... Solver^5.1

The Conjugate Gradient Solver was written by Orion Lawlor as part of a class project at the University of Illinois.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...AMPIzer^5.2

Karthikeyan Mahesh developed AMPIzer working with me.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...^5.3

Eric deSturler and Jay Hoeflinger, our colleagues at CSAR, converted Rocflo to AMPI, while I converted Rocsolid along with a research assistant.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... quicker^5.4

Working with Jim Jiao, the original developer of Rocface, I converted it to AMPI in 45 minutes.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... NCSA^5.5

This experiment was carried out by Sameer Kumar with the Conjugate Gradient Solver code provided by Orion Lawlor.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.