Self-compile on rabbit and snake
Stephen Weeks
sweeks@research.nj.nec.com
Thu, 4 Mar 1999 22:01:19 -0500
Here is a comparison of the running times of three generations (G0,
G1, G2) of a polyvariant self-compile on both rabbit and snake. For
G1 and G2, I also give the GC%. G1 and G2 both used a 950m heap. As
you can see, there was no improvement on rabbit in going from G0 to
G1, but there was a big (expected) improvement on snake. That is, we
expected G1 (and G2, which should be identical to G1) to run about
twice as fast as G0. The only explanation I see for the numbers is
that the non-speedup is due to the memory card in rabbit.
rabbit snake
G0(s) 1147.99 2003.99
G1(s) 1104.06 919.82
%GC 20.9% 17.9%
G2(s) 1108.09 915.34
%GC 21.0% 18%
Here are some other (machine invariant) statistics about the run.
code size(bytes): 16,141,623
stack size(bytes): 16,777,216
max semispace size(bytes): 471,859,200
number of GCs: 43
bytes allocated: 8,797,345,976
bytes copied: 2,090,201,096
max live(bytes): 111,347,692