Run attached test on Solaris Sparc (I run on SB2000, USIII 2x1.2Ghz, 2G RAM),
with -server, and notice a significant regression in performance of
some trigonometric functions since 1.4.2:
Total benchmark time:
  1.5b33 119351 ms
  1.4.2_04 85418 ms
  1.3.1_11 99391 ms

In my case trigonometric calculations were almost 2x slower
in 1.5 than it was in 1.4.2:
  131: 45976 ms
  142: 40344 ms
  150: 74498 ms

###@###.### 2004-01-09

The performance regression is largely in sin and cos. and appears to orginate with our previous work on 4738699. Reassigning to the RE for that bug.


###@###.### 2004-01-09