more code for tracing and other shit; NOTHING IS WORKING YET
more grid code uglification -- should be a little faster now
uglified grid code even more: each proxy now has "enabled" flag,
and traces without callbacks (most "did collide with anything?"
traces aren't interested in exact collision info) will return
simple true/false flag, so we can avoid expensive callback
invocation.
not working yet, tho: i fucked up flag checking somewhere
uglified grid code even more: each proxy now has "enabled" flag,
and traces without callbacks (most "did collide with anything?"
traces aren't interested in exact collision info) will return
simple true/false flag, so we can avoid expensive callback
invocation.
not working yet, tho: i fucked up flag checking somewhere
no more "correct offsets" in particle engine
previous ray tracer was bugged: it could return position inside wall,
so particles were trying to "compensate" for that (wrongly). and they
were doing it on *each*(!) think step.
as particles are pure eye candy (and i rewrote ray tracer), i removed
that "compensator" altogether.
previous ray tracer was bugged: it could return position inside wall,
so particles were trying to "compensate" for that (wrongly). and they
were doing it on *each*(!) think step.
as particles are pure eye candy (and i rewrote ray tracer), i removed
that "compensator" altogether.
more particle control options
alternate sky scaling
more fixes to grid; dunno, seems to work again
grid code uglification; particles are great again (i hope)
no more tree for map
map ray tracer now using grid instead of tree
faster hash clears
removed some unused code
removed some unused code
restored particles
better wall tracer
better hitscan tracer; no more level trace bitmap (but no more particles too, alas)
new tree-based weapon hitscan tracer (sometimes it is faster than the old one, sometimes it is slower...)
xprofiler: made curtime getters public
g_weapons.g_Weapon_gun: faster traces (i hope)
added simple hash table tester
hashtable: resize bugfix; `forEach()` iterator
console commands can have help strings (somebody, please, add 'em! ;-)
better console completion
"mon_sq_enabled" console command
removed unnecessary indirection in hash table
some optimizations in hash table
hashtable cosmetix
added robin hood hash table implementation
removed "monidx" argument from monster spatial query callback (each monster has UIDs, use UID to get monster object later!)
cosmetix
dyntree cosmetix
rewritten dyntree visitor; seems to fix segfaults (but i don't know why)
replaced some var args to constref args
more monster tree cosmetix
monster tree cosmetix
accelerated explosion sq
use tree to check collisions with shot projectiles
faster `g_weapons.CheckTrap()`
uidmap for monsters, so we don't have to do linear searches
some tree code for monsters
no more tree for items (we don't need it)
removed e_writelog from g_netmaster
replaced e_fixedbuffer with e_msg
no more public `gMonsters`
removed unused function
g_language: Fix some interface strings
converted grid and tree to generics (fuck you, FPC! your generics fuckin' sux fuckin' balls!)
no more global `gItems[]` array; created DynTree for items (not used yet); also, fixed FATAL bug in dyntree node manager
added `.positionChanged()` to (almost) all entities; don't forget to call it after changing entity coords!
Revert "don't flush enet every message"
This reverts commit 42057f398fd17568bb3dbc00b6ffd8c97a99c5e1.
This reverts commit 42057f398fd17568bb3dbc00b6ffd8c97a99c5e1.
`g_basic.g_CollideLevel()` now using accelerated coldet
render code cleanup; collision code cleanup; reverted useless "speedup"
profiler cosmetix
slightly better memory locality
intermediate commit (faster render and coldet; this commit won't build!)
pointer taking is typed now (it is safer this way)
changed mode from DELPHI to OBJFPC (with heavy customization, see "shared/a_modes.inc")
experimental grid with buckets
tree: desperate attempts to win several microseconds...
shit; "--dbg-mainwan" doesn't work; changed to "-dbg-mainwad"
"--dbg-mainwad-restore" and "--dbg-mainwad wad"
cosmetix
removed some "try/finally" (sorry, it is FUCKIN' SLOW!)
cosmetic variable renamings
added small speedup trick to gird (tree already has it)
more tree/grid debug commands
README update
cosmetix
README repo URL fix
don't use tree for drawing by default
cosmetix
"t_dump_node_queries" (temp debug, DO NOT TOUCH)
tree: use integers instead of floats (we don't really need floats, except for segment queries)
cosmetix
cosmetix
tree seems to work now
"-dbg-mainwad" cli arg
more tree code; still not working
g_triggers.pas: fixed "Warning: Function result variable does not seem to initialized"
hidden "Game/DefaultMegawadStart" option to change default playing wads
some cosmetix in common compiler flags
some statistics
tree can render things (buggy)
dynamic aabb tree implementation (ported from D, completely untested)
more SAP code; still not working right
Sweep-And-Prune broad phase implementation; not working yet
added README
don't flush enet every message
better memory locality and access patterns for grid (it is IMPORTANT for old atoms, for example)
adaptive profiler history length (not really)
"r_draw_grid" -> "r_gridrender"
cosmetix
more profiler fixes
more profiling code
new console commands: "r_draw_grid" and "dbg_coldet_grid"
more profiler code; smoothing values, history
more profiler code
profiler
moved compiler options to shared/a_modes.inc; turned on advanced records
flamer projectiles don't push
downloading resources...