Carry over some correctness fixes from recent changes to vis into the
qbsp versions of ClipWinding and make the implementations more
consistent overall.
Most importantly:
* Ensure we always have allocated one extra dists/sides slot for wrap around
* Check we have space for a new vertex every time we try to add one
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Just in case unsigned long isn't big enough to capture the pointer
offsets. Also, cast STACK_WINDINGS to the correct type before doing the
comparison.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
No longer update the globals in this function - make the callers do it
instead.
bspinfo no longer refers to the globals at all.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Just having one base pointer was kind of annoying because you have to cast
to get to the lump offsets, then cast to byte * to find the miptex, etc.
Now we can access the same pointer using the correct types in each case.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
First pass at detail support as implemented in tyr-qbsp, using the PRT2
file format (hopefully nobody else already decided to use that?)
Because the "leaf" terminology is pervasive across the source, I've left
it there for now (so "leaf" usually now means "cluster") and added a few
variables with the "_real" suffix when they actually mean the real leafs
and not the clusters which vis still calls leafs. A bit disgusting, but it
actually works so I'll go through and clean up naming, etc. later.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Inspired by the implementation of Benjt Jardrup, save the current state of
the vis job to disc at regular intervals so that we can resume from that
point in case the computer needs to be rebooted/shutdown or crash/power
issues occur.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Importing parts of this from TyrQuake - means that the base type used for
managing the bit vectors is unsigned long. Not really a performance win or
anything here, just a bit nicer to not have to cast from (byte *) to
(long *) - just adds a little extra work before writing out the compressed
data.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Make the base portal vis thread safe and run on multiple threads. Use the
common thread progress indicator code.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Add -vv option to hide extra verbose messages behing. Move a couple of
debug messages to the extra verbose level.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
When a portal is completed, check the updated vis information for leafs
that can no longer be seen from the leaf the portal is attached to. Update
the portals on these no-longer-visible leaves to indicated that they can
no longer see our leaf.
Seems to be at least a small gain on my test maps. I suspect there is
some further improvement to be gained by taking better advantage of this
optimisation by changing out choice of portals in GetNextPortal. Couldn't
find anything trivial that worked though.
Signed-off-by: Kevin Shanahan <kmshanah@disenchant.net>
Kevin Shanahan
Tyrann