One more step towards BSP thread-safety. This one brought with it a very
noticeable speed boost (ie, not lost in the noise) thanks to the face
visframes being in tightly packed groups instead of 128 bytes apart,
though the sw render's boost is lost in the noise (but it's very
fill-rate limited).
This is next critical step to making BSP rendering thread-safe.
visframe was replaced with cluster (not used yet) in anticipation of BSP
cluster reconstruction (which will be necessary for dealing with large
maps like ad_tears).
The main goal was to get visframe out of mnode_t to make it thread-safe
(each thread can have its own visframe array), but moving the plane info
into mnode_t made for better data access patters when traversing the bsp
tree as the plane is right there with the child indices. Nicely, the
size of mnode_t is the same as before (64 bytes due to alignment), with
4 bytes wasted.
Performance-wise, there seems to be very little difference. Maybe
slightly slower.
The unfortunate thing about the change is the plane distance is negated,
possibly leading to some confusion, particularly since the box and
sphere culling functions were affected. However, this is so point-plane
distance calculations can be done with a single 4d dot product.
It was added only because FitzQuake used it in its pre-bsp2 large-map
support. That support has been hidden in bspfile.c for some time now.
This doesn't gain much other than having one less type to worry about.
Well tested on Conflagrant Rodent (the map that caused the need for
mclipnode_t in the first place).
While gcc was quite correct in its warning, all I needed was to
explicitly truncate the string. I don't remember why I didn't do that
back when I made the changes in 4f58429137, but it works now, and the
surrounding code does expect the string to be no more than 15 chars
long. This fixes yet another memory leak (but timedemo over multiple
runs still leaks like a sieve).
This gives a rather significant speed boost to timedemo demo1: from
about 2300-2360fps up to 2520-2600fps, at least when using
multi-texture.
Since it was necessary for testing the scrap, gl got the ability to set
the console background texture, too.
It's down to 128 bytes from 184, which fits nicely in two cache lines.
This made a nice difference to glsl, unknown to vulkan (it crashed after
about 31/51 timedemo loops), and was a was for sw and gl.
While it takes one extra step to grab the marksurface pointer,
R_MarkLeaves and R_MarkLights (the two actual users) seem to be either
the same speed or fractionally faster (by a few microseconds). I imagine
the loss gone to the extra fetch is made up for by better bandwidth
while traversing the leafs array (mleaf_t now fits in a single cache
line, so leafs are cache-aligned since hunk allocations are aligned).
The model system is rather clunky as it is focused around caching, so
unloading is more of a suggestion than anything, but it was good enough
for testing loading and unloading of IQM models in Vulkan.
This is an extremely extensive patch as it hits every cvar, and every
usage of the cvars. Cvars no longer store the value they control,
instead, they use a cexpr value object to reference the value and
specify the value's type (currently, a null type is used for strings).
Non-string cvars are passed through cexpr, allowing expressions in the
cvars' settings. Also, cvars have returned to an enhanced version of the
original (id quake) registration scheme.
As a minor benefit, relevant code having direct access to the
cvar-controlled variables is probably a slight optimization as it
removed a pointer dereference, and the variables can be located for data
locality.
The static cvar descriptors are made private as an additional safety
layer, though there's nothing stopping external modification via
Cvar_FindVar (which is needed for adding listeners).
While not used yet (partly due to working out the design), cvars can
have a validation function.
Registering a cvar allows a primary listener (and its data) to be
specified: it will always be called first when the cvar is modified. The
combination of proper listeners and direct access to the controlled
variable greatly simplifies the more complex cvar interactions as much
less null checking is required, and there's no need for one cvar's
callback to call another's.
nq-x11 is known to work at least well enough for the demos. More testing
will come.
While I doubt the difference is all that significant, this should speed
up entity rendering because it cuts out a lot of branching, and
eliminates scanning the same list multiple times only to not do anything
for large chunks of the list.
As much as it can be since the texture data is interleaved with the
model data in the files (I guess not that bad a design for 25 years ago
with the tight memory constraints), but this paves the way for
supporting sprites in Vulkan.
The fact that numleafs did not include leaf 0 actually caused in many
places due to never being sure whether to add 1. Hopefully this fixes
some of the confusion. (and that comment in sv_init didn't last long :P)
Modern maps can have many more leafs (eg, ad_tears has 98983 leafs).
Using set_t makes dynamic leaf counts easy to support and the code much
easier to read (though set_is_member and the iterators are a little
slower). The main thing to watch out for is the novis set and the set
returned by Mod_LeafPVS never shrink, and may have excess elements (ie,
indicate that nonexistent leafs are visible).
Getting close to understanding (again) how it all works. I only just
barely understood when I got vulkan's renderer running, but I really
need to understand for when I modify things for shadows. The main thing
hurdle was tinst, but that was dealt with in the previous commit, and
now it's just sorting out the mess of elechains and elementss.
Its sole purpose was to pass the newly allocated instsurf when chaining
an instance model (ammo box, etc) surface, but using expresion
statements removes the need for such shenanigans, and even makes
msurface_t that little bit smaller (though a separate array would be
much better for cache coherence).
More importantly, the relevant code is actually easier to understand: I
spent way too long working out what tinst was for and why it was never
cleared.
Double benefit, actually: faster when building a fat PVS (don't need to
copy as much) and can be used in multiple threads. Also, default visiblity
can be set, and the buffer size has its own macro.
The node struct was 72 bytes thus two cache line. Moving the pointer
into the brush model data block allows nodes to fit in a single cache
line (not that they're aligned yet, but that's next). It doesn't seem to
have made any difference to performance (at least in the vulkan
renderer), but it hasn't hurt, either, as the only place that needed the
parent pointer was R_MarkLeaves.
This is a big step towards a cleaner api. The struct reference in
model_t really should be a pointer, but bsp submodel(?) loading messed
that up, though that's just a matter of taking more care in the loading
code. It seems sensible to make that a separate step.
The dynamic array macros made this much easier than last time I looked
at it, especially when it came to figuring out the bad memory accesses
that I seem to remember from my last attempt 9 years ago.
This cleans up texture_t and possibly even improves locality of
reference when running through texture chains (not profiled, and not
actually the goal).
The depth limits in the gl and glsl renderers and in the trace code really
bothered me, but then the fix hit me: at load-time, recurse the trees
normally and record the depth in the appropriate place. The node stacks can
then be allocated as necessary (I chose to add a paranoia buffer of 2, but
I expect the maximum depth will rarely be used).
All of the nastiness is hidden in bspfile.c (including the old bsp29
specific data types). However, the conversions between bsp29 and bsp2 are
implemented but not yet hooked up properly. This commit just gets the data
structures in place and the obvious changes necessary to the rest of the
engine to get it to compile, plus a few obvious "make it work" changes.
For now, only the glsl loader disables caching, but it stores the frame
vertices in GL memory, so its hunk usage is relatively lower (and will be
lower still when I get skins sorted out).
Unfortunately, the intel driver on my eeepc doesn't like the mipmas for
plat_top2 or +2floorsw. If I either don't load their mipmaps, or skip
drawing them, things seem to work nicely.
It turns out my complicated plan was just that: complicated. Although there
are currently some bugs, the method I used to build the VBO in the first
place will work equally well for building the index lists.
The entire vertex set from every model is put into one list (not yet
uploaded). chains of elements arrays are build for non-instanced models
(instanced models will have their chains built each frame).
Still nothing being rendered: still in the process of building the display
lists, but I'm making good progress. Get this into git before something
goes wrong :)
