RE: Perlin Noise - Early demo release was ( - Texture Generator)
- Posted by rforno at tutopia.com Nov 25, 2002
- 416 views
Thanks. It will suffice being 1 channel. The range may be -32767 to 32767 *or* greater, if this eases things. Anyway, I'd be grateful in case you explain the basics of Perlin noise generation, for me to try. ----- Original Message ----- From: <dm31 at uow.edu.au> Subject: RE: Perlin Noise - Early demo release was ( - Texture Generator) |The problem with your approach, at least for me, is that it is too slow and |requires a lot of code. I'd like to attain the same results, but at least |two orders of magnitude faster. Any chance? One, it is already around 6 times faster then it was. :S Two, in pure Eu, it will be hard getting it too much faster. Of course I always try optimising code thou. Too much code? *me* looks at latest version here.... uh oh. Its already double since that release, and once 3D(Volumetric/animated 2D) and 4D(animated 3D) is done, it will be over double again... OK, here is what can be done. First, if you where using PTexture.ew, the blitfast/surface_flip bits of code can be deleted and bitmap_(un)lock moved to outside the for loops. (Its like it is atm for testing purposes). One thing I was going to do, is see if the random generator code I had in C would be faster then the rand function if I wrote a dll to use it. Has anyone got a random number generator in Eu/C/ASM that IS faster then rand()? Now, I was going to try to keep this code pure Eu, HOWEVER, once I get a stable version going, I might start writing some ASM to speed it up in a alternative version. Eu is just not suited to fast, intensive work. So, it is PSound.ew not PTexture that you will be using right? If I can get some details I'll see if I can speed it up by making a specific version that only meets THOSE requirements. Output Range: -32567 <=> 32567 ? Dimensions: 1 (2? If you require a slightly different 2 channel 4 stero) Anyothers that I should know? (I take it that it is producing appropriate data atm) Cheers, Dan McG ==^^=============================================================== This email was sent to: rforno at tutopia.com