Jos Hulzink wrote: > The original loop GCC creates was about 9 instructions long, increasing > addresses made a mess of the code, so I had to decrease them. The only way > that produced neat code was treating the Z-buffer like an array of kgiu32 > values. Do you mean that the loop was unrolled or that it was bad ? Rodolphe
- RFD: KGI(con) speed optimization Jos Hulzink
- Re: RFD: KGI(con) speed optimization Steffen Seeger
- Re: RFD: KGI(con) speed optimization Jos Hulzink
- Re: RFD: KGI(con) speed optimization Rodolphe Ortalo
- Re: RFD: KGI(con) speed optimization Jos Hulzink
- Re: RFD: KGI(con) speed optimization becka
