[SDL] Generalised BitBlit proposal

Ryan C. Gordon icculus at clutteredmind.org
Tue Sep 27 22:04:19 PDT 2005


> Your characterisation of blits as nasty code is absolutely right, and to my 
> mind this is the best reason for automatically generating them. They're hard 
> to write, understand, and port, bug-prone, and need to be optimised for 
> several different architectures. These are exactly the same reasons why 
> compilers were invented.

And we use a compiler to generate the machine code already, at least for
the C fallbacks. I don't see a practical use to writing a simple
compiler, and targeting it at all sorts of processors because the
existing blitting code is inelegant. I think it would turn out to be a
lot of work that adds a ton of code to the project, turns out to be
messy and buggy as well, and the final output would underperform
compared to the existing blitters.

I don't mind being proven wrong, though. Feel free to take a shot at it
if you are so inclined and we'll address this anew with working code.

--ryan.





More information about the SDL mailing list