The results pose an interesting question: “When can we
expect CINT2000 to be fully used?” Figure 4 projects the
percentage of papers that will fully use CINT2000 in the up-
coming years (nonlinear least-squares (NLLS) Marquardt-
Levenberg algorithm for the parabola ax2 + b). In 2005
we can expect 80% of the papers to fully utilize the suite,
a quite respective number. However, by then papers should
already be making the shift to CINT2004.
The data presents an alarming contradiction: Research of
future architectures is being performed with the benchmarks
of the past.