Results

Organization of Data

We ordered the results chronologically. In order to understand the temporal evolution of our results refer to timeline.

Applications

The applications used belong to Sandia National Laboratories except in the case of dense matrix multiplication using MPI. In each report we will explain which package was used.

Since T2000 is designed to run applications that execute mainly integer operations, we only tested the machine in this feature. In other words, all the performed test are done using integer operations.

Parameters

We used some parameters in our test. In this section you can find the definitions.

	MIOPS: This parameter means millions of integer operations per second.
	NEQ: Number of Equations. We are solving a system *Ax** = b, were A is a matrix and x and b are vectors. The number of equations is equal to the number of rows on A. Because A is a square matrix, the number of rows is equal to the number of columns.
	Speedup: The concept is defined as the time required by a serial algorithm divided by the time required by a parallel algorithm to perform the same task in the same machine. We can express this relation as:

Efficiency: This metric is defined as the speedup divided by the number of parallel tasks (or threads). The value of the efficiency varies between 0 and 1. We can express the concept as:

SWaP: (New !) Is defined as Space, Watts and Performance. We can express it as:

Results April 21, 2006

We tested the machine doing Dense Matrix Multiplication and Sparse Matrix Multiplication.

You can find the report of Dense Matrix Multiplication using MPI here.

You can find the report of Dense Matrix Multiplication using OpenMP here.

You can find the report of Sparse Matrix Multiplication using MPI here.

You can find the report of Sparse Matrix Multiplication using OpenMP here.

Results April 6, 2006

You can find the report here.