Stéphane Guillard - stephane.guillard@steria.com

• If you have a 3D gfx board (and presumably already have Warp3D installed) :
• In order to use the CPU driver, remove the real 3D hwdrivers you have in LIBS:Warp3D/hwdriver (move them somewhere).
• copy the W3D_CPU_PPC.library to libs:Warp3D/hwdrivers,
• setenv SAVE warp3D/cpu/chip n according to your chip as described below as follows from warp3D.h :
• W3D_CHIP_VIRGE 2 // Use for S3 ViRGE boards (like CV64/3D)
• W3D_CHIP_PERMEDIA2 3 // Use for PM2 boards (like CyberVisionPPC or BlizzardVisionPPC)
• W3D_CHIP_AVENGER_LE 5 // Use for Voodoo 3
• Sample : to render onto a Permedia2 BVisionPPC or CVisionPPC you would enter setenv SAVE Warp3D/CPU/Chip 2
• setenv SAVE Warp3D/CPU/Tile n where n is the tiling factor (from 1 to 9). 1 gives accuracy but is slow, 9 gives huge blocks but is fast. My choice is 2.
• enter "avail flush" in a shell to make sure all drivers are removed from memory.
• start your 3D app and the software render is on the job !
• If you don't do the setenv, the swrender WILL NOT WORK. This is due to the way the Warp3D libraries manage the drivers.
• If you haven't got a 3D Gfx board (and presumably don't have Warp3D installed) :
• install Warp3D, pretend you have a CV64/3D even if it is not true.
• copy the W3D_CPU_PPC.library to libs:Warp3D/hwdrivers,
• start your 3D app and the software render is on the job !

• supports all Warp3D features except mainly fog, stencil, alpha blending, filtering, mipmapping, antialiasing
• support for all chunky and planar screenmodes (inc. OCS/ECS/AGA)
• successfully tested with all current Warp3D apps and demos (warptest, gears, engine, minigl demos, mesa demos, dsoab, skulpt, wipeout etc.)
• no enforcer hits, should be fully stable, runs for hours
• configurable to lower resolution and increase speed (setenv warp3D/cpu/tile n where n is the tile size between 1 and 9)
• configurable to force hardware chip id, to enable rendering on a 3D gfx board (setenv warp3D/cpu/chip n where n is chip id)
• Textures are used from app. memory (no buffer allocated ; fast but disables updating texture images asynchronously)
• ZBuffer goes in fast RAM (less VRAM usage)
• All texture bitmap and screen bitmap modes supported.