They show that there is difference aginst other shadowmapping algorithms however Zed is also doing the PSSM so there should be no difference.
Wait. So, when Zed’s saying “there is no difference”, he’s talking about his particular implementation of PSSM (1 texture with multiple passes) vs. the “stock” PSSM case? I thought he was talking about non-PSSM cases.
Hmm. Looking back, it seems I misunderstood a couple of statements from him. My appologies for the mixup.
Originally posted by Mars_9999:
Hey Zed, with the image you posted for me to look at each cube there is a shadowmap? So you have 5 shadowmaps if I am seeing correctly? The first 4 are from the origin but only allowed out small amounts and get larger when you get to 4 and the 5th starts where the 3rd ends? BTW could I email you my SM code to see if it will work with your idea? Thanks
that image was a quick drawing of the general idea, viewed from top down the blue lines are view area
u will want to position the frustums depending on the sun direction etc
yes each cube is a frustum (containing a SM)
personally if youre doing a RTS which seem to be viewed from a topdown 2.5D viewpoint i wouldnt worry about using CSM, its more suitable for something where the camera is level with the ground eg a FPS
btw wolfgang engel is meant to be writing a paper on the subject (dont know when its due), he asked me to check it, which aint a good idea as apparent by the confusion on this thread due to my lack of explaining capabilities 
Yeah its a RTS but the viewing is not always top down, and IMO the quality of the shadows is poor even with 1024x1024 due to I have to cover such a large area with one shadowmap…
So what would you suggest Zed if not CSM? For a RTS games…
IMO the quality of the shadows is poor even with 1024x1024 due to I have to cover such a large area with one shadowmap…
What resolution are you trying to render at that a 1024x1024 looks “poor”?
You could try a CSM/PSSM system, using 2x512x512's . Those look pretty decent, compared to 1x1024x1024.
Though for an RTS, it’s probably not a good idea, considering the number of units and so forth you’re probably trying to render. You’ll probably kill the GPU on vertex transfer and/or vertex T&L.
Personally, though, if your RTS doesn’t involve getting close to units/terrain, I wouldn’t bother with shadows at all. And if it does, I’d probably consider reconsidering whether the gameplay works adequately from such a perspective.
Originally posted by Korval:
Though for an RTS, it’s probably not a good idea, considering the number of units and so forth you’re probably trying to render. You’ll probably kill the GPU on vertex transfer and/or vertex T&L.
The units in RTS are likely to have significantly less polygons than say the FPS characters and much of the details can be done with normal maps, especially when the camera is designed to display a big area from top down view. Additionally for rendering of the shadowmaps, position only vertices and simple vertex shaders are sufficient so vertex transfer bandwith and vertex processing power used to render several shadowmaps might be lower than bandwith&power used to render the resulting shaded view.
Imho the PSSM/CSM is not that usefull for the top down camera because it tries to distribute the camera frustum into the shadowmaps based on distance from the camera and in the top down view there is ussualy nothing near the camera and most of the things are in similiar distance from the camera. For that case it might be better to split the ground plane within the camera view to several areas and cover each of them with one shadowmap.
1600x1200
Number of units shouldn’t be out of control when used with culling?
Why not use shadows? Most RTS games these days have them…
Komat, what would a normal map do for me when it comes to shadows on the terrain? Never heard of this before.
Originally posted by Mars_9999:
Komat, what would a normal map do for me when it comes to shadows on the terrain? Never heard of this before.
Directly nothing. It only allows to use lower number of small triangles on the objects and thus it can decrease the cost of rendering of the shadowmap, if the same geometry is used for both shadowmap rendering and scene rendering.
Originally posted by Komat:
[quote]Originally posted by Mars_9999:
Komat, what would a normal map do for me when it comes to shadows on the terrain? Never heard of this before.
Directly nothing. It only allows to use lower number of small triangles on the objects and thus it can decrease the cost of rendering of the shadowmap, if the same geometry is used for both shadowmap rendering and scene rendering. [/QUOTE]Like Komat said, it does nothing, except for one tiny thing, a normal map creates a multitude of tiny shadows due to variations in the diffuse lighting, these shadows should be at most as dark as the shadowmap shadows, it’s not a big issue, but you have to keep that in mind.
keep in mind with this
http://appsrv.cse.cuhk.edu.hk/~fzhang/pssm_vrcia/comp_knights.htm
is the limited draw distance compareed to a more typical scene
http://s24.photobucket.com/albums/c26/zz…mgAnch=imgAnch1
im guessing the difference is 50meters vs 1km ie 20:1
eg take the PSSM(4; 512x512) photo it looks ok in the above screenshot but thats cause its shrunk down to 8x6cm onscreen blow that image up to fullscreen + each of those pixels will be a huge blocky ugly thing (and this is with a limited view distance) expand the view distance out to 1km + u see for decent results i believe (multiple maps) 2048x2048 is minimum (or 1024x1024 with blurring)
im guessing the difference is 50meters vs 1km ie 20:1
I’m not seeing a kilometer in that screen shot. I’m seeing maybe a 4:1, and that’s being generous.
Your point on screen resolution vs. shadowmap resolution is valid. But, then again, I’ve never been one to feel that high resolutions are a necessary construct. I’d rather use strong antialiasing and anisotropic filtering than raise the resolution much past maybe 720p or so.
i admit youre right u vulcan bastard, the landarea was 1km not the view dist, though that is more than 200m (the grass blades are extra high due to the fact i had to ‘sex them up’ as at 20cm high u could hardly see them unless there were 20billion of the f’ers in the 1m^2 region before the the camera)
comeon man AA is a poor mans high res aka supersampling, aka have the cake + eat it too
in a sense youre right look at tv its at what in the US640x480 or PAL like here 720x576 hardly hires but why does it look so good?
is it possible to grad looks? it must be
i have the follwoing
area on screen / draw distance 10
physics 8
resolution 10
framerate 5
AA 3
lighting / shading 40
texture quality / variety 6
mesh quality / polygon size 6
animation quality 4
particle effects quality 4
post processing 3
…
environment interactivity
num interactive characters onscreen
is this a fair rating?
like other’s said, using CSM/PSSMs in an almost top-down RTS is pretty useless
you might want to have a look at variance shadow maps though
those are pretty good at hiding aliasing artifacts at lower
shadow map resolutions
