Some return values can be assigned to a `const auto`, somewhere `dir` is
mutated so the whole result has to be assigned to a non `const` variable
and once only dir is used, so structured binding isn't used there.
Avoid cases of filenames
Update years in terms and other legal(ish) documents
Don't update years in license headers, since change is not meaningful
Will add linter rule in seperate commit
Happy recompiling everyone!
Original Patch By: Nescio
Comment By: Gallaecio
Differential Revision: D2620
This was SVN commit r27786.
Two changes to the vertex pathfinder that should generally improve
performance:
- Sort unaligned edges by distance like we did aligned edges. This
isn't very scientific, but the idea is generally sound, and empirically
it seems to do OK.
- Relax the optimality by using a weighted heuristic, with weight 4/3
or 5/3 depending on the number of obstacles around. In the worst cases,
A* will return paths that are this many times less optimal, but search
should be faster in general (and sometimes much faster).
Both of these optimisations might increase the constant-cost slightly,
but should decrease the worst-case runtimes.
Differential Revision: https://code.wildfiregames.com/D5034
This was SVN commit r27724.
6581796103 removed the ability for terrain to affect movement speed. The
JPS pathfinder cannot support it, and the approach was poor anyways,
coupling rendering data with simulation data.
This lets us remove the dependency on CTerrainTextureManager everywhere.
Tested by: langbart
Differential Revision: https://code.wildfiregames.com/D4459
This was SVN commit r26269.
The pathfinder computations are run asynchronously (and potentially on
the main thread) in-between simulation turns, thus reducing
pathfinder-related lag considerably in common cases.
To make this most efficient, the number of paths computed during a turn
via MaxSameTurnMoves is reduced from 64 to 20.
This has a hard dependency on the obstruction manager (via the vertex
pathfinder) not being modified in-between simulation turn (or to put it
more generally on the simulation state not changing outside of turn
computation), otherwise results will be non-deterministic and go OOS.
This is currently entirely safe (as in, it indeed does not happen that
the simulation state changes in-between turn), but future work towards
improving simulation sandboxing would be good.
Thanks to Kuba386 for maintaining & improving the patch in 2020
Thanks to everyone who tested the various iterations of this patch.
Fixes#4324
Differential Revision: https://code.wildfiregames.com/D14
This was SVN commit r25657.
TERRAIN_TILE_SIZE is now only used in relevant places in the simulation.
This also makes it mostly possible to use CFixed in constexpr contexts.
Refs #5566
Differential Revision: https://code.wildfiregames.com/D3078
This was SVN commit r25360.
- 25332f9b86 added 'override' to a method in ShaderProgram.cpp. This
adds it to the rest.
- BinarySerializer.cpp had custom JS includes, because putting those in
ScriptExtraHeaders failed. The cause is that gloox leaves "#define
lookup" up. This has been reported upstream, and fixed here.
- VS17 complains about an uninitialised variable, though the usage is
safe.
Differential Revision: https://code.wildfiregames.com/D3916
This was SVN commit r25359.
This changes ParamNode to use UTF8 values internally (XMB files are UTF8
since cb9d0733ef).
This removes the need for a lot of conversions, speeding things up and
allows cleaning up the validator interface & a few other callsites.
ConstructJSVal could be a tad slower because of UTF8->16 conversions
within Spidermonkey; but the difference is unlikely to be noticeable in
practica.
Also:
- Changes `ToXML` to `ToXMLString` for clarity.
- Add a simple "op" test & show a particular behaviour of merge nodes
that I intend to change somewhat in D3830.
- Remove Component.h from simulation2 PCH - brought in too much.
Tested by: langbart
Differential Revision: https://code.wildfiregames.com/D3834
This was SVN commit r25228.
This implements a form of crowd movement that I've generally called
'unit pushing' in the last few years.
Essentially, any two units will push each other away when they're too
close. This makes it possible to ignore unit-unit obstructions, and thus
makes movement much smoother in crowds.
This first iteration of this system only allows pushing between idle
units and between moving units (i.e. a moving unit does not affect an
idle one).
This is because the unitMotion logic to detect it is stuck & needs to
use the pathfinders starts breaking: units can fail to move because they
are pushed away from their intended movement, and the current logic
fails to handle this gracefully.
Thankfully, the most value of this patch in terms of player experience
is found in the improvements to group movements and shuttling.
Other impacts:
- As the short pathfinder is called less often, we can increase the
starting search range & reduce the # of max turns, both improving
collision recovery.
- The performance of idle units is slightly worsened, as they must be
checked for idle-idle collisions. If needed a 'sleeping' system, as used
in physics engine, could be implemented.
- In general, however, expect slight performance improvements, as fewer
short paths are computed.
- Gathering efficiency should increase slightly, since shuttling times
are likely reduced slightly.
- As a sanity change to improve some edge cases (units that say they're
moving, i.e. pushable, but don't actually move), the 'going straight'
logic is turned off if a short path has been computed. This requires a
few cascading changes to work correctly.
Technical notes:
- To reduce the cost of the n^2 comparisons that pushing requires, units
are only compared within a small square on a grid which is lazily
reconstructed each turn. The overhead seems rather small, and this is
much simpler than keeping an up-to-date grid.
- The design is intended to be parallelisable if needed someday.
- The pathfinder's CheckMovement ignores moving units in UnitMotion, as
that is now the spec. Idle units are not ignored, which is required for
the 'collision' detection to work correctly (see above).
Refs #3442 (not fixed - idle units are not pushed by moving units).
Fixes#5084 (the overlap can still happen, but units will push each
other away).
Differential Revision: https://code.wildfiregames.com/D1490
This was SVN commit r25182.
Introduce a LOS_TILE_SIZE, to replace usage of TERRITORY_TILE_SIZE in
the LOS code.
This makes it possible to change the resolution of LOS/Terrain without
affecting the other component.
Additional refactoring:
- LosTile has been renamed LosRegion (it's more comparable to the
hierarchical pathfinder regions/spatial subdivisions)
- LosState explicitly refers to "los vertices" instead of terrain
vertices.
Refs #5566
Differential Revision: https://code.wildfiregames.com/D3076
This was SVN commit r24980.
This improves behaviour when units need to go around a concave obstacle.
They would tend to clump inside the 'dead-end' before realising they
needed to go around. This was rather easy to trigger on Acropolis. See
included Unit Motion Integration Test.
The cause is the logic that removed the next long waypoint when
obstructed. While that behaviour is desirable, removing too many
waypoints means the unit tries to short-path, using a small domain
range, to a goal that's impassable, meaning they go as close as they can
in Euclidian distance, i.e. towards the dead end.
This changes that behaviour by only deleting waypoints within a certain
distance from the entity, scaling with search-space range. It's tricky
to find a good compromise between performance and behaviour here, but
the values I've picked seem OK.
However, the fact that the entity would ultimately remove all waypoints
and thus trigger a full path recomputation was actually a feature,
inherited from D2754 / 892f97743b. This diff therefore handles that
explicitly, doing so on a more regular basis to behave better overall.
As a further cleanup, "m_FailedPathComputations" is incremented in
HandleObstructedMove, as it is quite possible to never increment it in
PathResult despite not getting actionnable paths. This thus renames it
to m_FailedMovements, and uses the opportunity to clean up PathResult(),
by only having one path for both short and long-range paths. Further,
PathResult now does not immediately request new paths, leaving that to
Move(), to avoid requesting transient paths that aren't actionnable.
This also makes it possible to revert 9e41ff39fc. It requires increasing
the MAX_FAILED variable, or more units get stuck as they reach the max
more often.
The search-space expansion is slightly slowed, and with a little more
delay, as a performance optimisation. From testing, this doesn't impact
real movement much as units short paths tend to be invalidated by the
next turn, as other units move, anyways.
Clarify comment around the vertex-pathfinder search-space bounds hack,
and ensure it isn't used for the very worst cases of units being stuck,
as it could be a pessimisation then.
Finally, this explicits a 2011 hack where if the long-pathfinder fails
to return a valid path the goal's center is used directly. This happens
when the goal is unreachable to the long-pathfinder, which may be
because it is actually unreachable or because only the short-pathfinder
can reach it. In those situations, the hack allows a last-ditch attempt
at reaching it before failing to move entirely. Performance wise, this
is faster overall for actually unreachable goals, since it skips all the
intermediate steps. For reachable goals, it might be occasionally
slower, but that case is quite rare (certainly rarer than unreachable
goals).
Reported By: Angen
Fixes#5795
Differential Revision: https://code.wildfiregames.com/D3203
This was SVN commit r24429.
By using templates appripriately we can remove the need for explicit
specification of serializers, making it easier to serialize container
types and to write new serialization helpers.
Direct serialization calls haven't been replaced in this diff.
Comments by: vladislavbelov
Differential Revision: https://code.wildfiregames.com/D3207
This was SVN commit r24427.
GUIObjectBase is made a IGUIObject* to avoid including those headers
un-necessarily. Subsequent diffs ought to clean up the various of
pointers for that with a similar type with reference semantics.
Also:
- Add standard C and C++ headers (mostly cstring for memcpy, string and
vector) where needed.
- Swap out some includes for forward declarations
- Clean up un-necessary boost includes in precompiled and other headers.
- Clean up precompiled headers, including fewer things.
- Move ACPI to the windows-specific folder as it's included there only
and mostly specific to that platform.
Thanks Stan for the testing.
Differential Revision: https://code.wildfiregames.com/D3129
This was SVN commit r24352.
This cleans up many un-necessary header includes, either simply
providing nothing or forward declarations in their place.
No major compilation time change here, though this does reduce depencies
in some headers.
Also fix up old MacOS STL-include fixes that are no longer relevant.
Differential Revision: https://code.wildfiregames.com/D3128
This was SVN commit r24227.
ComputeTargetPosition called Dot() with large enough vectors that it
overflowed. Avoid that by not actually doing the full dot product.
Reported by: Itms
Fixes#5852
Differential Revision: https://code.wildfiregames.com/D3061
This was SVN commit r24152.
939002f0dc changed from vectors to grid which broke resetting when
terrain size changed.
Also use SAFE_ARRAY_DELETE for simplicity.
Reported by: vladislavbelov
Reviewed By: vladislavbelov
Differential Revision: https://code.wildfiregames.com/D2961
This was SVN commit r24117.
Improve unitAI: don't move if the requester can reach us and we are
close enough. This avoids an issue where ships moved more than necessary
when picking up many units.
Also improve requester UnitAI -> retry pickup if the target entity is
Idle.
Improve unitMotion: periodically recompute paths in "known bad path"
mode to adapt to moving targets.
Expose UnitMotion reachability to scripts and other code.
This adds a test map for some common and some tricky pickup cases, using
triggers.
Based on a patch by: causative
Reviewed By: Freagarach
Fixes#3472
Differential Revision: https://code.wildfiregames.com/D665
This was SVN commit r23925.
Patch By: pcpa and wraitii
Tested By: Nescio, andy5995 and others
Differential Revision: https://code.wildfiregames.com/D2745
This was SVN commit r23794.
Changing Grid.h should recompile faster, as it is now included in fewer
TUs.
Differential Revision: https://code.wildfiregames.com/D2784
This was SVN commit r23774.
Range manager has several `std::vector` for fixed-size arrays and 2D
grids. By using proper data structures, the code readability is
improved.
This also moves around the LosVisibility enum.
Comments by: Stan`, nani
Differential Revision: https://code.wildfiregames.com/D2770
This was SVN commit r23769.
These variables might be used uninitialised in very odd cases that don't
happen in svn.
Reported By: elexis
Differential Revision: https://code.wildfiregames.com/D2250
This was SVN commit r22877.
By leveraging custom ordering in std::set, MakeGoalReachable and
FindNearestNavcellInRegions can be optimised and the code simplified.
Differential Revision: https://code.wildfiregames.com/D1882
This was SVN commit r22817.
Optimise MakeGoalReachable using global regions, by leveraging the fact
that we can easily know which goal regions we can reach. Particularly
for point-goals, it becomes almost instant.
This does not change hashes.
Differential Revision: https://code.wildfiregames.com/D1835
This was SVN commit r22653.
Following 3a839517ce, SparseGrid::Reset() replaces the value of m_Data
(a C-style dynamically sized array) with a new value-initialised array,
by using the new operator. Previous code simply did a memset.
This means that when m_Data was not null, it leaked memory as the
previous array was not deleted.
SparseGrid::Reset is called when destroying the sparse grid, and a
SparseGrid is used by the long pathfinder when computing JPS paths. As
that is called rather often, it resulted in a relatively serious memory
leak that could make very long games use several GB of memory.
This fixes the leak by using placement new, which re-uses the memory,
effectively doing the same as the previously existing memset.
Profiling by historic_bruno and elexis showed that performance from
value-initialising with placement new was at worst similar to memset,
and testing shows this was compiled in a memset call in several cases
anyways.
Reviewed By: elexis
Fixes#5522
Differential Revision: https://code.wildfiregames.com/D2121
This was SVN commit r22545.
As reported by #5521, Ordering units to walk to a point in a forest can
lag terribly, as units will end up computing long short paths in the
forest, which can result in `ComputeShortPath` calls that individually
take upwards of 80ms.
This can be alleviated by allowing units to stop a bit earlier. A23
handled this in UnitMotion directly, but it's better to handle this in
UnitAI to avoid surprises and to make it customisable on a per-state
level.
This diff further speeds up using the short pathfinder by starting with
a smaller search range, limiting the max-range more and moving the range
slightly towards the goal.
This also refactors UM sending messages to UnitAI so that we may in the
future push more information (in particular, the entity_id that a unit
was obstructed by could be interesting for COMBAT).
This doesn't fix the possibility of lag, but it reduces its occurrence
to levels that should be similar to A23 and thus acceptable enough.
Tested By: Freagarach
Fixes#5521
Differential Revision: https://code.wildfiregames.com/D2105
This was SVN commit r22526.
5c642611c4 introduces undefined behaviour: for vertices that are part of
non-AA squares, the value of "quadOutward" is undefined. This fixes that
by again setting it to QUADRANT_ALL by default, as it was pre
5c642611c4. This led to OOS.
Reviewed By: elexis
Fixes#5508
Differential Revision: https://code.wildfiregames.com/D2082
This was SVN commit r22483.
Compute the outward quadrants once and for all instead of setting them
dynamically, because there is no reason why we should always arrive from
the same quadrant as the first time we see a vertex.
Don't consider quadrants for the start-vertex, because of the edge
expansion (which can put us in illegal quadrants)
These result in (much) better paths, the tradeoff being that we now look
at some more vertices.
Fixes#5476
Differential Revision: https://code.wildfiregames.com/D1908
This was SVN commit r22473.
This adds back the "domain" edges to the short/vertex pathfinder.
Without these edges, units could use points far away from the path
search that they inaccurately thought were reachable, leading to some
pathing oddities in rare cases.
Reviewed By: temple
Differential Revision: https://code.wildfiregames.com/D443
This was SVN commit r22465.
D981/c219ee54b2 changed IsInXChecks to use edge-to-edge distance instead
of centre-to-edge, which broke UnitMotion's min-range movement, which
assumed distance to the center.
Since units are represented as squares, the diagonal point may be closer
to the target than the "real" clearance by a factor √2, so the delta
between minimum range and maximum range should be at least `(√2 - 1) *
clearance` to be safe in all situations (this is generally not a problem
for regular units which have a clearance of 0.8, but could be one for
catapults or elephants).
Differential Revision: https://code.wildfiregames.com/D1969
This was SVN commit r22422.
These functions were placed in UnitMotion, which had nothing to do with
range checks and made them available only to moving entities for no
particular reason.
This patch also adds support for square-square range checks and
shape-shape distance checks.
Modified from a patch by bb on top of work from wraitii.
Differential Revision: https://code.wildfiregames.com/D981
This was SVN commit r22345.
