[DotGNU]a review of runtime performance across 0.3.6 release

[Gopal V]

Hi All,
        Following up on rhys's optimisations on ilrun , I decided to
take a closer look at how much he has done across the 0.3.6 release.

----------------------VERSION 0.3.5------------------------
Compiled from the CVS snapshot of 11-May , for ilrun

Sieve  ... 336
Loop   ... 363
Logic  ... 251
String ... 30
Float  ... 13
Method ... 313

PNetMark rating is 124
-----------------------------------------------------------

----------------------VERSION 0.3.5------------------------
Compiled from the CVS snapshot of 11-May , for ilrun_opt

Sieve  ... 577
Loop   ... 538
Logic  ... 530
String ... 42
Float  ... 18
Method ... 562

PNetMark rating is 202
-----------------------------------------------------------

----------------------VERSION 0.3.7------------------------
Compiled from today's CVS , for ilrun

Sieve  ... 754
Loop   ... 653
Logic  ... 1209
String ... 72
Float  ... 26
Method ... 786

PNetMark rating is 309
-----------------------------------------------------------

So obviously this shows a performance boost of more than 100% 
over the past fortnight ?. And he isn't finished with optimisations.
If I'm to believe him , "Optimizing direct threaded code by selective
inlining" will improve performance further ....

But still true to its name rhys's comments were "the primary selling 
point is portability, not speed - getting into a speed war is a fool's game."
(maybe I'm a fool, but does it matter much ?....)

I wish someone with lotsa bandwidth and cpu cycles to spare would 
download the set of snapshots and draw a "manhattan"..(bar chart)
And a portability matrix won't hurt either .... as the MAC OSX,
HPUX performance points come in ..... :)

But as someone said "there are lies , damned lies ,statistics,
benchmarks and release dates".....

Gopal
-- 
The difference between insanity and genius is measured by success