Previously we had a single culling frustum based on the main camera, and
any object outside the frustum would never get rendered, even if it
should actually contribute to shadows or reflections/refractions. This
caused ugly pop-in effects in the shadows and reflections while
scrolling.
Extend the renderer to support multiple cull groups, each with a
separate frustum and with separate lists of submitted objects, so that
shadows and reflections will render the correctly culled sets of
objects.
Update the shadow map generation to compute the (hopefully) correct
bounds and matrices for this new scheme.
Include terrain patches in the shadow bounds, so hills can cast shadows
correctly.
Remove the code that tried to render objects slightly outside the camera
frustum in order to reduce the pop-in effect, since that was a
workaround for the lack of a proper fix.
Remove the model/patch filtering code, which was used to cull objects
that were in the normal camera frustum but should be excluded from
reflections/refractions, since that's redundant now too.
Inline DistanceToPlane to save a few hundred usecs per frame inside
CCmpUnitRenderer::RenderSubmit.
Fixes#504, #579.
This was SVN commit r15445.
Backport os_getversion from premake4.4-beta4 and extend it to support
POSIX-compilant systems.
Add some missing newlines at the end of file as these cause warnings on
OpenBSD.
Update the config paths.
This was SVN commit r11970.
the old debug_assert always ran and tested the expression, which slows
down release builds. wrapping them in #ifndef NDEBUG is clumsy. the new
ASSERT behaves like assert and ENSURE like the old debug_assert. Let's
change any time-critical but not-super-important ENSURE to ASSERT to
speed up release builds. (already done in bits.h and unique_range.h)
This was SVN commit r9362.
this snowballed into a massive search+destroy of the hodgepodge of
mostly equivalent types we had in use (int, uint, unsigned, unsigned
int, i32, u32, ulong, uintN).
it is more efficient to use 64-bit types in 64-bit mode, so the
preferred default is size_t (for anything remotely resembling a size or
index). tile coordinates are ssize_t to allow more efficient conversion
to/from floating point. flags are int because we almost never need more
than 15 distinct bits, bit test/set is not slower and int is fastest to
type. finally, some data that is pretty much directly passed to OpenGL
is now typed accordingly.
after several hours, the code now requires fewer casts and less
guesswork.
other changes:
- unit and player IDs now have an "invalid id" constant in the
respective class to avoid casting and -1
- fix some endian/64-bit bugs in the map (un)packing. added a
convenience function to write/read a size_t.
- ia32: change CPUID interface to allow passing in ecx (required for
cache topology detection, which I need at work). remove some unneeded
functions from asm, replace with intrinsics where possible.
This was SVN commit r5942.
- u_anim_name disambiguation
- fix implementation of finite (use our fpclassify instead of compiler's
routine)
- รค -> ae
- workaround for MAP_ANONYMOUS
- fix GLint in ogl.cpp
- add include for SIZE_MAX in string_s
- avoid PIC clobbered error and speed up rdtsc a bit
- add include for stat
This was SVN commit r4170.
and a fix for the auto-build: #define enums from EXT_framebuffer_object
* clip shadow bounds to frustum
* add CBrush, a class representing a convex object
This was SVN commit r3695.
