... M. Kuhn¹

supported in part by the National Science Foundation under grant CCR-9411861

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

... level²

The amount of interprocedural optimization actually performed by gcc on the Intel machine is suspect since the -O2 optimization level produced the same object code as the -O3 optimization level for each of our test programs.

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

... intentional.³

Execution time can be computed as the product of instruction count, average cycles-per-instruction, and cycle-time.

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

... lookups⁴

We use the C++ terminology in this paper rather than introduce the corresponding SELF terminology.

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