[hacknews] Floating point benchmarks

dwt (05/14/82)

I ran a suite of 16 benchmark programs before and after
the installation of the floating point hardware.
An attempt was made to eliminate the effect of overhead,
by subtracting the time for a ``null'' program (one that
was similar to the benchmark, but executed zero iterations
of the main loop).
   The results are as follows:

	PROGRAM		Before time	After time	Speedup factor
	*******         ***********     **********      ***************

	APL		   374		  9.2		  41
	Basic		   184		 28		   7
	C  #1		    40		  0.6		  67
	C  #2		    67		  1		  67

Subjectively, the effect on using APL interactively is
quite spectacular.