java layer performance comparisions

[Marcus G. Daniels]

[for "javaswarm StartHeatbugs -b" (250 iterations)]

Debian 2.2, AMD-K6 3D 444 MHz, 160MB    (884 Bogomips)
  Kaffe 1.0b4, Swarm 2.0.1               70 secs
  Sun JDK 1.2.2, Swarm 1999-12-16        36 secs 
  Kaffe 1.0b4, Swarm 1999-12-16          28 secs  

Solaris 7, UltraSparc 5 333 MHz, 128MB  (662 Bogomips)
  Kaffe 1.0b4, Swarm 2.0.1               63 secs
  Kaffe 1.0b4, Swarm 1999-12-16          43 secs
  Kaffe 1.0.5, Swarm 1999-12-16          33 secs
  Objective C Heatbugs, Swarm 2.0.1      13 secs

Redhat 6.1, Sparc 10 RT620X4, 64MB      (55 Bogomips, per) 
  Kaffe 1.0.5, Swarm 1999-12-16          318 secs 

Redhat 6.1, Pentium MMX 200 MHz, 64MB   (399 Bogomips)
  Kaffe 1.0.5, Swarm 2.0.1               115 secs
  Kaffe 1.0.5, Swarm 1999-12-16          50 secs

Windows 98, Pentium II 400 MHz, 128MB
  JDK 1.2.2, Swarm 2.0.1                 35 secs
  JDK 1.2.2, Swarm 1999-12-16            16 secs
  Objective C heatbugs, Swarm 2.0.1       9 secs

These performance improvements are a result of caching
Java method implementations for Actions, and from
tightening up the multilanguage object directory routines.

Further work: no real effort has been made to generate fast JNI stubs,
and we haven't yet looked at building Swarm models with a native code
compiler like gcj (http://sourceware.cygnus.com/java).

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.