VOGONS

Common searches


Reply 20 of 23, by wd

User metadata
Rank DOSBox Author
Rank
DOSBox Author

> Using a weird program doesn't mean anything else will be slow either,
> because much of the optimization will still be common for any program.

Was thinking about the memory access functions, that is if you got one
that largely uses custom functions (several graphics modes do) this might
outweight everything and inverse the condition, which is quite bad for
most other games.

Reply 21 of 23, by `Moe`

User metadata
Rank Oldbie
Rank
Oldbie

Well, ideally you'd choose a handful of games, and only one of them used these modes. But as I already said, doing PGO with just one typical program that doesn't do anything weird will be better than not doing PGO at all. One such program shouldn't be too hard to find.

Reply 22 of 23, by frobme

User metadata
Rank Member
Rank
Member
`Moe` wrote:

Well, I think even using just a text-mode program for PGO is better than none at all, since there's a lot in DOSBox that can be optimized and which is not graphics mode dependent.

Definitely agreed. Having written and optimized a couple of emulators myself, it is rare that the code base is breaking performance on a sub-system of the emulated machine, but more often the problems in performance are per-tick cycle costs that you pay regardless of code execution. Not always, but it would certainly benefit you to run a general simulation of a DOS program no matter how simple it is, so you can get the optimizer thinking about basic instruction scheduling costs and the like.

Of course, you can still build a pathologically bad set of code for a specific case (we had a couple of these in Atari800) but they are usually easily isolated and identified "hey, when I run in graphics mode 7, my frame rate drops 50%!".

-Frob

Reply 23 of 23, by frobme

User metadata
Rank Member
Rank
Member

Thought I would update this with the later code bases..my latest build (as of 02/15/2007) which is Visual Studio 2005, program guided optimization, yields a result of average 474 ticks in Doom, which is the equivalent of 157.6 fps.

This puts Dosbox (when running Doom) in the neighborhood of a high end P3 machine in speed, host hardware being my Intel Core Duo 2.93Ghz. This is quite an impressive increase in speed over the last couple of months, so congratulations to the Dosbox team.

-Frob