[fpc-pascal] DOOM game for FPC

[Vinzent Hoefler]

On Sunday 04 December 2005 02:19, Kornel Kisielewicz wrote:

> No, I think that the error is in the BSP building algorithm.

The BSP-trees are built beforehand. I don't see how this would affect 
game performance unless the built BSP-tree is heavily unoptimized.

> Original
> C-source probably used some float-model-specific optimizations, the
> guy who ported it to Delphi converted them to
> delphi-float-model-specific results, and while porting to FPC those
> things got broken. I know that in the "old-days" there were several
> machine-wise tricks on floats that were used for BSP trees, and I
> imagine that could be the main source of the problem.

The original Doom didn't use floating point at all. Back in the "old 
days" (meaning: 486 at best) it was far too slow (no pipeline and all). 
Maybe the conversion now uses floats, but the original code definitely 
didn't. Quake was the first of Id's games to rely heavily on floating 
point usage (and IIRC also a lot of hand-optimized assembly to get 
those through the pipeline fast enough).


Vinzent.

-- 
public key: http://www.t-domaingrabbing.ch/publickey.asc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: signature
URL: <http://lists.freepascal.org/pipermail/fpc-pascal/attachments/20051205/22ca155c/attachment.sig>