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.
Changed SoundGroupTable to use the right string type for the comparison.
Also changed from CStr8 to CStr because I prefer that spelling (and it's
more consistent with most other code).
This was SVN commit r4771.
- Added "Screenshot Mode" and "Fog of War" game attributes. (Screenshot
Mode causes units to be initialized to Hold stance instead of Aggress
and also forces LOS to be All Visible. Atlas turn on Screenshot Mode by
default so units don't try to kill each other in there.)
- Modified LOSManager to allow disabling fog of war.
- Removed some debug message spam.
- Enabled line antialiasing for aura rendering and fixed some bugs that
caused strange effects (color was not set properly for the center point,
and when a unit was both mouseover'ed and selected, the aura was drawn
twice).
- Modified Stand stance to allow retaliation on attacks (normally Stand
will attack any enemy in LOS, but this is useful if a neutral unit is in
LOS).
- Modified pathfinder to not take into account terrain slope, which is
an expensive calculation - we'll eventually take into account terrain
type instead.
This was SVN commit r4527.
Map XML: Store camera position. Stopped using DTDs (because they make it
too hard to change the XML structure without breaking all the old XML
files).
Game.h: Include fewer files, to make compilation sometimes faster.
World: Changed some things to not be singletons, since they were
(ab)used as CWorld members.
This was SVN commit r3670.
minimap, PatchRData, renderer: fix ELOSState comparison vs & and use
SAFE_DELETE
ia32: add rounding control constants (for FISTP)
premake: set /QIfist compiler setting which causes float->int casts to
go through FISTP instruction instead of _ftol2() (much faster, but
requires CPU state to have been set)
LOSManager.cpp: cache m_TilesPerSize; use MIN/MAX; use
CTerrain::CalcFromPosition; clean up GetUnitState a bit. now runs at
203ns, down from 222
This was SVN commit r3099.
ia32: removing misleading and outdated comment
Player: add "LOS token" - speeds up LosManager by caching a temp result
LOSMAnager: speed up by about 9%, more to come. combined 2 DWORDs into
one u16; further optimization pending. old code remains (via #ifdef
frankenstein)
This was SVN commit r3097.
LOS issues still outstanding:
- LOS looks ugly because of quad tesselation into 2 triangles
- Quad tesselation is unspecified by OpenGL (in fact using GL_QUADS on
LOS quads seemed to give different tesselations than it did for the
underlying terrain quads, but terrain rendering also used GL_QUADS).
This should be fixed once we decide on the quad tesselation issue.
- Units with traits.vision.permanent set are visible through FOW even if
you havent seen them before; this should only be true when you have seen
them before. But it gets even more complicated - if a permanent unit
seen through FOW dies or gets upgraded or something, perhaps you should
remember the old version. I'm not completely sure how to do this
(probably involves cloning its actor somehow).
This was SVN commit r2876.