pouët.net

i admit the lack of the half-texel-shift is pretty tempting - but that's just a gripe, really.

... and "not doable" != "using dx9 would have made my dx10 4k into a dx9 5k"

kb: well if thats the criteria then no, because you can always dig out that old software rasteriser.. :)

but anyway, what about e.g. manually resolving msaa buffers and being able to write different values to different msaa samples in the shader. that'd be extremely useful to me.
id love to move to dx10 but a) half my machines have xp and b) most of the world's machines have xp. if other people start though, ill surely follow. :)

gargaj: that half texel shift makes porting from 9 to 10 a bit more annoying tho :)

Manual resolve of MSAA buffers is one of the primary reasons to me. Finally, proper HDR + MSAA.

smash: come on, it's not like FLT demos aren't flawed in that aspect already even in DX9 ;)

smash: nah, you just have to undefine all the "HACK" parts :P

What features in DX10 are there that aren't in GL? (Not a troll, genuinely interested)

kusma: no, its just that we worked around dx9's problems so effectively that it takes a bit more effort to unwork it.. :)

chaos, then I have a good example I think. :) You could use a geometry shader to cast a frustum into your spacial acceleration structure and spawn only the visible geometry. Wait yes people already do this but, well, i'm sure it's a lot faster on the GPU anyway. ;) and err, by much more than 30% :) because you can only act at the batch level on the CPU else it's too slow.

nystep: correct me if i got something wrong here, but doesn't that geometry shader you mention only speed up the rasterizing process a little (because after projection transformation this geometry gets thrown away anyway) and in fact not at all deal with the problem of "not transferring too much data over PCIe bus"?

and still having not done anything with DX10, i am also very interested in things that might be possible with DX10, yet not in OpenGL, so add 1 to DJSelwynGummer's request count...

@DJSelwynGummer: I'd say "drivers that actually work and not crap out on you randomly"

@DJSelwynGummer: how about "api not written by chipmunks"; "no extensions or caps". :)

NeARAZ, well I'm used to that by now ;) That's not the API's fault though.

Does anyone who's not massively inbred (ie. not smash) want to contribute?

dx9 for now, skip dx10 totally, move to dx11 when it's officially out

yeah probably going to wait til adoption rates go up, the new features are slightly tempting, mostly the ones that make coding easier

re: dx11
dx11 is probably going to be 90% dx10, same way windows 7 will be 90% vista

@Hyde. This is an unwritten rule for 4k-Intros ;) Of course they are 4k in size but have a library with >30MB in the background. So, complaining that with DX10 a Intro would be 5kB in size is ok. but useless...

mop: Considering that DX11 is a strict super-set of DX10, that doesn't make much sense to me...

Since most people are not even using DX 8 in their prods on PC, i'm not slightly tempted to "upgrade" for demos....

I'll stay with dx9 for a while. for the audience and because i hate vista.

are geometry shaders fast nowadays? i head they suck on most DX10 hardware.

Kusma: sure DX11 is a beefed up DX10, but Vista will be short lived once Win7 is officially released, so I presume DX10 will become extinct like Vista by next year.

Mop: this reminds me of the old discussion if one should upgrade to AGA or should stay with ECS :-)

it reminds me of theparty1993! please, tell us a story about that, magic!!!

maali: Oh, The Party 1993 was awezum!