About the Benchmarks

The benchmarks are taken from the Gambit-C distribution. For basic information, see http://www.iro.umontreal.ca/~gambit/bench.html, but we have used the following compact mode names for the table:

• base = safe with generic arithmetic, roughly R6RS semantics
• base_!_ = unsafe with generic arithmetic
• fxfl = safe with fixnum- and flonum-specific arithmetic
• fxfl_!_ = unsafe with fixnum- and flonum-specific arithmetic

• Bigloo 3.3a
• Gambit-C 4.6.0
• PLT Scheme 4.2.4 using modules that import the r5rs library

• MacBook Pro, 2.53 GHz, Mac OS X 10.6.2, compiling to 32-bit programs

All results are based on CPU time, and the results are normalized to one of the times. By default, the results in a given row are normalized to the fastest time in the row, which is shown in the leftmost column. The cell for the fastest time in a row is always in bold green; by default, a white background also highlights the fastest time, but clicking one of the group titles restricts the white background to the fastest time within the group, while aslo normalizing all columns to the white-cell time. Clicking the title of an individual column normalizes all rows to that column's times (and disables all white backgrounds).

Click mode in the upper-left corner of the benchmark page to group the columns by compilation mode, and click impl to group the benchmarks by implementation.

For similar benchmarks with more compilers, see this other page.