In trying to use transparency shadow maps in Pixie 2.2.2, however, whenever I replace
sColor = shadow(shadowmap, Ps);
which compiles fine, with
sColor = (color)shadow(shadowmap, Ps);
to try to get it to sample transparency information as well, I get a Parse error from
sdrc. Has anyone else seen this behavior from version 2.2.2. I am running the pre-compiled
Also, am I right in assuming that without this cast, the shadow function returns an ordinary
in/out shadow sample even if it is then stored in a color value? That seems to be the behavior
that I am seeing.
sColor = color shadow(shadowmap, Ps);
Should work. I think technically (color) shadow is a cast of the function, rather than specifying the color variant of shadow... I'll investigate if we can still support the (color) shadow as a syntax.
That makes sense, and compiles perfectly. However it still doesn't seem to be giving me a
transparency lookup. I'll try it on some simpler scenes, and see if I can get it to work.
Are there any other differences in the lookup other than adding the color keyword before
the shadow lookup call? If I cannot get it to work, I'll post a scene.
Some of the traffic on this board had given the syntax for tsm lookups as the (color) syntax,
perhaps it would be good to put the correct syntax on the transparency shadow map wiki page.
One of the reasons that I went to tsm's is that I've been having problems with Pixie 2.2.2
crashing when rendering ordinary shadow maps. No crashes with TSM's though even when used
as an ordinary shadow lookup. I'm working on getting a stripped down scene to demonstrate
the problem and I'll post it shortly.
Thanks so much for the quick response,
I am getting an intermittent crash of Pixie 2.2.2 when using ordinary (non tsm) shadowmaps.
rndr crashes about 4 times in 5. I've created a test scene that demonstrates the crash on
my system, which can be downloaded at
I've included the shadowmap rib file and the shaders that I'm using. For reference, my system
has the following characteristics
Standard Pixie 2.2.2 Win32 build from the Pixie website
Intel core 2 duo processor
If you need any more information about my platform please let me know. I am currently trying
to compile the latest build from the svn (looking for a Win32 distribution of flex/bison to use)
That's a little alarming. Pixie definitely shouldn't crash doing something as simple as rendering a shadow map.
I can't reproduce this on my OSX machine with 2.2.2 source base or the current source base - so it could be platform specific. Unfortunately, I don't have a Win32 machine to test with.
I'm doing some tests to make sure we didn't miss anything stupid - and the bug is almost certainly simple.
If you do
Display "file.tx" "shadow" "z"
then you don't have to do the MakeShadow yourself.
I'll let you know if I get anywhere. Okan may have more success with Win32 debugging.
I just wanted to let you know that I did confirm that the error happens on a 2nd
Window's box, I had a colleague render the image and they experienced the same
I have compiled both the 2.2.2 and SVN versions of Pixie and will try to get you
some Debug information as soon as I can.
Thanks so much for the help,
Dao Hung Huynh
here it works well. I tested on my machine with Windows XP, I tried my scene with shadow map by using the MakeShadow method. Though I have not used the current SVN sources. I will update and test the lastest build and let you know if there got some crash.
Did you try the scene that I posted? There's a link in an earlier message in this thread.
It is entirely possible that the problem is in the formatting of either my RIB file or my
light shader. If other scenes that use shadow maps work fine, then it is likely that the
problem either lies in my scene or is manifest because of something that my exporter is
doing in this scene.
Thanks again for the help,
Yes I did, it seemed to be taking a lot of memory, with more progressively as the scene rendered (which is unusal).
I was tied up this weekend, so I haven't got an answer yet as to whether there's a bug / what the cause is... but I have a couple of guesses.
I'll post again when I know more.
Sorry about that, I seem to have replied to the wrong post.
Yes I did render the scene, but I can't recreate the issue (though as noted above, it may be win32 specific).
Could someone try it on Win32 in leu of not having a box to test on?
Sorry for the confusion, my question about trying my scene was in reply to another
poster who had indicated that they had tried ordinary shadow maps on their Win32 box,
but their post indicated that they were using one of their own scenes, that apparently
worked just fine. My intention was to ask that poster if they had tried my scene on
their box. Sorry for the confusion.
This seems to be a rather intermittent problem, and on my machine (and my colleague's
machine) it will turn up more in certain scenes than in others. The only thing that
I can think of (because we are using the same shaders and the same exporter) is that it
depends on the level of tesselation or the specific arrangements of the polygons as
to whether the problem shows up often, seldom or not at all. Given the experience of
the last person who posted, it may very well be something unorthodox that I am doing
in my shader (or at least something that pixie is having trouble with) and if so, it'd
be great to find out what that is.
I did compile the most recent svn build and found that when debugging, the problem
never shows up, however, as soon as I render it outside of the debugger, the problem will
show up quite often, at least with the scene I posted. As I'm not terribly familiar with
Pixie's codebase, I haven't tried to dig into it any more than that.
I have a work-around for now by using TSM's as regular shadow maps, but obviously that's
overkill and a waste of disk space :^)
Thanks so much for your time and attention to these issues,