what's the fastest way to get CUDA result to screen ?
category: general [glöplog]
hello,
i wonder what's the fastest way to get a framebuffer, which is stored in CUDA-globalmem, onto screen using DX9 ? i've seen an OGL example, but with DX ?
i wonder what's the fastest way to get a framebuffer, which is stored in CUDA-globalmem, onto screen using DX9 ? i've seen an OGL example, but with DX ?
of course i wanna avoid any device2host-memcpy-calls ;)
lol
i'm sure he's calling the host, too...
iirc cuda to texture or cuda to framebuffer device-device copies are not possible. i recommend asking the people at official cuda forums on nvidia website.
the fastest way would probably be to start with one of the sample projects, no?
why CUDA anyways? it is vendor-dependent. besides os x snowleopard has drivers for opencl with which I am fiddling around. so it shouldn't be so long until we see opencl up and running widespread.
why CUDA? well, if you're doing raymarching and is feeling lazy, there's one reason. But there are also more valid reasons, like e.g. if you want to do something else than filling triangles. Then the two most popular gpu-apis are just in the way.
hyde I mean, why specifically CUDA :). I mean he can as well just wait for the newcomers
xortdsc, yes - please stop what you're doing immediately until Decipher gives you the green light.
lol
haha, rright...
i'm sure it is possible, since global-mem to vertex-buffer works perfectly, i dont see any reason why it shouldn't work for framebuffers. anyway, i'll try the cuda-forums...
i'm sure it is possible, since global-mem to vertex-buffer works perfectly, i dont see any reason why it shouldn't work for framebuffers. anyway, i'll try the cuda-forums...