GL_CLAMP (clamp to border) was changed to GL_CLAMP_TO_EDGE in 2008
(f2baf359). In 2018 (ce1d5406) I made OpenGL 1.2 be required since
GL_CLAMP_TO_EDGE is used.
Restore support for GL_CLAMP in order to support OpenGL 1.1 like vanilla
Quake 3 does. This should allow using the default Microsoft Windows
GDI Generic OpenGL 1.1 driver (untested but it won't fail the version
check at least).
From gpuinfo.org, it looks like drivers stopped advertising support for
GL_SGIS_texture_edge_clamp so use a version check in addition to the
extension check.
r_allowExtensions 0 disables using GL_CLAMP_TO_EDGE in the opengl1
renderer. GL_CLAMP support wasn't added to the opengl2 renderer.
For lerped frames (refEntity_t frame not equal oldframe) IQM joint
matrices may have incorrect axis scale. This can cause significant model
distortion. The matrix lerp is linear causing each vector to move in a
straight line between frames instead of arcing like a circle. Each joint
frame can have a different scale so can't just normalize the joint
matrix.
Store joints as quaternions and spherical lerp between them and then
convert to a matrix. For my test model, setting up the skeleton is four
times slower now but it still seems to be fast enough to be usable.
Bounds are optional for animated IQM models but are not possible to
include with unanimated models (seems intended for use with separate
model containing animations and bounds). Calculating bounds for
unanimated IQM models fixes culling and head model on HUD which
calculates position from model bounds.
The axis returned for IQM tag was the animation's joint rotation without
the base frame joint rotation. It only worked correct for models that
did not rotate the base frame joints.
Only calculate vertex blend matrix for each unique bone indexes/weights
combination once per-surface instead of recalculating for each vertex.
For best performance the model surfaces needs to use few vertex bone
indexes and weights combinations.
Unroll loops so GCC better optimizes them.
In my tests drawing animated IQM may take 50% as long in opengl1 and
70% as long in opengl2. It will vary by model though and might not
help much at all.
Made unanimated IQM models skip matrix math altogether.
- Only allocate memory for vertex arrays that are present in the IQM
file and are actually used (may not have colors or blend index/weights,
don't load tangents in opengl1). (Colors is fixed to next commit.)
- Explicitly handle loading IQM files without meshes (bones only).
- Better IQM validation. Header data offset 0 mean data is not present
in file. Check if required vertex arrays are present.
This involved a lot of white space changes and moving code around.
Load functions procs supported by OpenGL ES 2.0, though there is not a
compatible renderer yet. Change argument for GLimp_Init from coreContext
to fixedFunction.
Also declare the GL functions in tr_local.h so there is compile error
for non-core GL functions instead of SEGFAULT from dereferencing a NULL
pointer.
Disable the non-functional stencil shadow code that hasn't been updated
to use OpenGL 3.2 core compatible drawing.
If renderer is compiled into client (USE_RENDERER_DLOPEN=0) and after
start up set r_allowExtension to 0 and run vid_restart, some extension
were still used.
Moved all the code using Altivec intrinsics to separate files. This
means we can optionally use GCC's -maltivec on just these files, which
are chosen at runtime if the CPU supports Altivec, and compile the rest
without it, making a single binary that has Altivec optimizations but
can still work on G3.
Unlike SSE and similar extensions on x86, there does not seem to be
a way to enable conditional, targeted use of Altivec based on runtime
detection (which is what ioquake3 wants to do) without also giving the
compiler permission to use Altivec in code generation; so to not crash
on CPUs that do not implement Altivec, we'll have to turn it off
altogether, except in translation units that are only entered when
runtime Altivec detection is successful.
This has been tested on Linux PPC (on an Altivec-enabled CPU),
but we may need further work after testing trickles out to other
PowerPC devices and ancient Mac OS X builds.
I did a little work on this patch, but the majority of the effort belongs
to Simon McVittie (thanks!).
- Parse OpenGL version in sdl_glimp.c to share with both renderers.
- Add GL_VERSION_ATLEAST(major, minor) macro.
- Get address of glGetStringi if using OpenGL 3.
- Fix glConfig.extensions_string when using GL3 core context in
opengl2 renderer.
- Make opengl1 renderer's gfxinfo support qglGetStringi too.
After 'Fix floating point precision loss in renderer', Windows x86
client won't load the renderer DLLs. The problem is a 64 bit modulus.
I couldn't find any reports of this online. However, client with
built-in renderer worked with the 64 bit modulus.
Only tested with mingw-w64.
Fix floatTime using float precision instead of double using GCC.
Fix R_BindAnimatedImage to be in sync with function table.
Fix vertexDeform bulge, vertexDeform normals, noise wave function
at high level time.
Revert unnecessary float -> double conversions.
Patch for https://bugzilla.icculus.org/show_bug.cgi?id=5931 by
Eugene C. from 2013 plus recent fix for tcMod rotate.
I merged the changes into the OpenGL2 renderer though the fix for
tcMod turb doesn't translate.
Models don't have a surface limit; skins shouldn't either. Some player
models require more than 32 surfaces since vanilla Quake 3 did not
enforce the limit.
Skins are now limited to 256 surfaces because having no limit would
require parsing the skin file twice. The skin surfaces are dynamically
allocated so it doesn't increase memory usage when less surfaces
are used.
gcc 6 with -Wall -Wextra warns:
code/botlib/l_precomp.c: In function ‘PC_NameHash’:
code/botlib/l_precomp.c:551:2: warning: ‘register’ is not at beginning of declaration [-Wold-style-declaration]
int register hash, i;
^~~
Modern compilers either ignore the register storage class when
generating code, or generate better code without it, so just remove
most of them.
The remaining uses are in third-party bundled libraries (libjpeg, zlib),
and in a PowerPC-specific inline function consisting of inline
assembler (because I'm not 100% confident that it doesn't have
some practical use there).
Explicitly set cull type for skybox to front, instead of using whatever
cull type the previous shader used (which could result in the skybox
not being visible due to only drawing back faces). The sky cloud stages
set the cull type so they are not affected by previous cull type.
The number of draw surfaces was range checked against number of surfaces for
the current view but needs to check total for the frame otherwise can read
past the end of the tr.refdef.drawSurfs array when there are multiple views.
Reserve space for end of list and swap buffer commands. These are absolutely
required and cannot be dropped. Dropping swap buffer command causes screen
to not update and possible crash from drawsurf buffer overflow if not enough
cmd buffer space for many continous frames.
Stencil shadow is not drawn if a mesh, or multiple meshes with the same
entity and shader, have more than 500 vertexes. The issue is caused by storing
the projected positions in the tess vertex buffer. Use a new array instead.
