Drew Northup wrote:
>
> Hey Kevin,
> Do you think that maybe looking into somehow optimising the resultant
> assembler code from the plex86 codebase would help speed up the final
> virtualization code? I know that you make your actual assembly work as fast
> as possible, but could we get any observable gains from optimising the C/C++
> portions of this project in a more "hands-on" way?
There is a lot of room in the algorithms I use for
speed-ups. Keep watch of the file 'PERFORMANCE', which
I use to keep a list of things todo in this area. There
are a bunch more things to add on that list.
I'm doing some things now which are _quite_ horrific for
performance. Once Windows runs, I'm going to begin
implementing some of the ideas in PERFORMANCE. I really
hate to do this before then.
Additionally, some of the special guest OS drivers and
such will boost performance.
Likely, performance will be nothing like it is currently.
It should be _much_ faster. Then when an option is added,
to turn of SBE altogether for well-behaved ring3 code,
it will be even faster. If you come across other ideas,
post them too. Anyways, still in the "get it working" stage.
We're still compiling all code with "-g". ;^)
-Kevin
