Based on the findings, it is possible to conclude that imagery and chunking instruction significantly improve student performance among high school Latin I students.
In order to fully utilize chunking and imagery techniques, teachers must first be exposed to, and develop an understanding of the two methods.
In all cases, speedups were obtainable on up to 56 processors with at least one of the chunking schemes and the larger problem size.
In all benchmarks, factoring performed as well as, or better than, the other three chunking schemes.
The forward scheduling of iterations poses a challenge to the variable-size chunking schemes, since the inner (sequential) loop is triangular.
Therefore, the fixed-size chunking schemes that we tried (SC and SS) perform relatively poorly.
The adjoint-convolution program showed no performance improvement for any fixed-size chunking scheme.
For the matrix multiplication benchmark, factoring performed better than fixed-size chunking with optimal K.
We thank Fran Allen and Michael Burke for their continuing support of this research, and Ron Cytron for suggesting that chunking be revisited.