This should be less of a drag on the playsim than having each light a separate actor. A quick check with ZDCMP2 showed that the light processing time was reduced to 1/3rd from 0.5 ms to 0.17 ms per tic.
It's also one native actor class less.
This will mostly ensure that each patch used for composition is only loaded once and automatically unloaded once no longer needed.
So far only for paletted rendering, but the same logic can be used for true color as well.
In ZDoom this would affect everything using a patch that got used in a front sky layer, even if the texture was totally unrelated. It is only owed to the low usability of such patches for other purposes that this hasn't caused problems.
Previously it tried to copy all patches of composite sub-images directly onto the main image.
This caused massive complications throughout the entire true color texture code and made any attempt of caching the source data for composition next to impossible because the entire composition process operated on the raw data read from the texture and not some cacheable image. While this may cause more pixel data to be processed, this will be easily offset by being able to reuse patches for multiple textures, once a caching system is in place, which even for the IWADs happens quite frequently.
Removing the now unneeded arguments from the implementation also makes things a lot easier to handle.
* it's no longer the main texture objects managing the pixel buffer but FSoftwareTexture.
* create proper spans for true color textures. The paletted spans only match if the image does not have any translucent pixels.
* create proper warp textures instead of working off the paletted variants.
As a side effect, caching of pixel buffers for texture composition is temporarily disabled, as it management of texture redirections. These things will be reimplemented once things progress further. The existing methods here had their share of serious issues that should be fixed.
This was done to make reviewing easier, again because it is virtually impossible to search for the operators in the code.
Going through this revealed quite a few places where texture animations were on but shouldn't and even more places that did not check PASLVERS, although they were preparing some paletted rendering.
* it was never saved in savegames, leaving the state of dead bodies undefined
* it shouldn't be subjected to pointer substitution because all it contains is old dead bodies, not live ones.
Since the only thing it gets used for is swapping out PlayerPawns it can safely skip all global variables that never point to a live player, which allowed to remove quite a bit of code here that stood in the way of scriptifying more content
This allows retaining the functionality, even if for the JIT compiler's benefit all default arguments are pushed onto the stack instead of reading them from the defaults array.
This will now both exclude floor caps when only ceiling elements are used and everything outside the bounding box of active portal lines.
Hopefully this is enough to fix the issues with portal caps but of course it is not foolproof if someone just makes the right setup.
Apparently they can indeed happen with broken map setups like isolated linedefs somewhere in the wild (see Strife MAP08.)
Although they are a problem for triangulation, this isn't what sections get used for currently so it's of no real concern.
In case this is needed later their work data gets marked as 'bad' for the time being.
A section is a continuous part of a sector or in some case of several nearby continuous parts. For sectors with far away parts multiple sections will be created, especially when they lie in disjoint parts of the map.
This is mainly supposed to cut down on time for linking dynamic lights. Since they need to traverse subsectors to find all touching sidedefs a more coarse data structure that only contains the info needed for this is more suitable. In particular, this does not contain any intra-sector lines, i.e. those with both sides in the same sector.
Setting the buffer and its attribute bindings must be one step, not twp. With Vulkan this is a single API call.
This removes the now obsolete SetVertexBuffer method from the model renderer's interface.
Add smoothGroup member to OBJFace struct, and assign the current smooth group number to it
Move face normal calculation code to CalculateNormalFlat
Add AddVertFaces method, which initializes and populates the vertFaces array of arrays, which holds references to triangle references per vertex
Only initialize and populate vertFaces if the model has missing normals and smooth groups
Assign smooth groups to triangle data
Add CalculateNormalSmooth method, which calculates the normals for each face the vertex is attached to, depending on whether or not the faces are part of the given smooth group, and averages them out
Add OBJTriRef struct, which holds references to triangles on OBJ surfaces
Make {agg,cur}SurfFaceCount unsigned ints
Change nvec to a value instead of a pointer
(This clearly shows that using 'long' as parameters in any interface must be stopped. It is fundamentally unsafe to have a type whose size is not reliable - it's either an int-sized nor a pointer sized value, depending on the platform, and essentially worthless.)
Pass surf->vbStart to SetupFrame instead of 0, and pass 0 to DrawArrays instead of surf->vbStart.
Use a potentially faster method of modifying the OBJ file text buffer by modifying it directly.
Add RealignVector and FixUV methods to begin work on re-aligning OBJ models to the same orientation as MD3 models.
Re-align OBJ models to match MD3 models
Fix normal calculation for re-aligned OBJs
Ensure AddSkins does not go out of bounds of surfaceskinIDs
Do not precache skins that were replaced by the user.
Fix OBJs with a large number of materials not being fully rendered
Print a warning message if a material referenced by the OBJ could not be found.
Free surface triangles once they are no longer needed
Also, use continue instead of return so that surfaces after those with missing materials are still rendered.
Fail if a face side has no vertex reference. Vertex references are required for a valid OBJ.
Clean up OBJ model code
Remove commented code, mainly Printf's that aren't used any more.
Add more documentation comments, and tweak existing documentation comments
Replace ParseVector2 and ParseVector3 with a template ParseVector function
Create a new surface for each 'usemtl' statement in the OBJ file, and fix memory errors caused by TriangulateQuad.
Calculate missing normals, and fix incorrect UV coordinates
Fix construction of vertex buffer for objects with multiple surfaces
Localize curMtl, curSurface, aggSurfFaceCount, and curSurfFaceCount to FOBJModel::Load(), since they are not used anywhere else.
Fix parsing of OBJs without UV references
Internally, I replaced hashtag line comments with C-style line comments, and I replaced each forward slash with newSideSep.
If no UV coordinates are available, add a default vector of (0,0).
Also, remove "this->" from ResolveIndex to make the code a bit cleaner, and fix a minor garbage issue I failed to notice earlier (normref would pick up garbage if there was no normal reference).
Ensure usemtl statements remain intact
It may be a bit inefficient, but I tried modifying the buffer directly, and I got memory corruption errors. In this case, it's a lot better to be safe than sorry.
What works:
- Parsing the model
- Constructing geometry (surfaces) for triangulated models
What doesn't:
- Rendering the model
- Building the vertex buffer
- Triangulating quads
UE1: Add support for Deus Ex format vertex data.
UE1: Group triangles by skin index AND type/flags (preparation for per-surface render style support).
UE1: Add handling of Weapon Triangle (preparation for model attachment support).
UE1: Support flat shaded triangle flag.
This removes 3 uniforms, consisting of 9 floats. Those were merged into other values that never get used at the same time.
It also moves the costly setup of the fixed colormap out of the render state into the 2D processing code.
Since 3D forces use of render buffers now, it is no longer necessary to draw the entire scene with the colormap active, meaning it can be handled more efficiently.
src/gl/scene/gl_flats.cpp:215:3: error: cannot jump from this goto statement to its label
src/r_data/models/models.cpp💯18: error: no member named 'floor' in namespace 'std'
- rename gl_InitModels to InitModels
- add commented out support for #include in modeldefs (blocked by gene tech having broken #include statements in its modeldef files)
src/r_data/models/models.cpp:418:33: warning: comparison of integers of different signs: 'long' and 'unsigned long' [-Wsign-compare]
src/r_data/models/models.cpp:427:38: warning: comparison of integers of different signs: 'long' and 'unsigned long' [-Wsign-compare]
src/r_data/models/models_ue1.cpp:49:37: warning: comparison of integers of different signs: 'long' and 'unsigned long' [-Wsign-compare]
(concatenated "UMSH" signature + datafile + anivfile)
This is pretty much 100% functional by now.
Hasn't been tested on platforms other than Linux yet, though.
Code definitely deserves some cleaning.
This is for mitigating the recently discovered problem with attenuated lights getting reduced in size, even on OpenGL 3+. The intent of the shrinking was to account for higher brightness of non-attenuated lights on OpenGL 2 and was never meant to be active on more modern versions.
The factor will apply to any attenuated light defined after it and will be inherited by included sub-lumps, but it will only last for the lunp it is set in.
If you have a definition for the broken behavior, AddLightAssociation
'lightsizefactor 0.667' at the top of your GLDEFS.
- with renderers freely switchable, some shortcuts in the 3D floor code had to be removed, because now the hardware renderer can get FF_THISINSIDE-flagged 3D floors.
- changed handling of attenuated lights in the legacy renderer to be adjusted when being rendered instead of when being spawned. For the software renderer the light needs to retain its original values.
It may use the same calculations as the hardware renderer but must use the 2D drawer for display.
It should be investigated if the hardware renderer can do this as well.
This was done mainly to reduce the amount of occurences of the word FTexture but it immediately helped detect two small and mostly harmless bugs that were found due to the stricter type checks.
* Instead of using the red channel it now uses the grayscale value. While slower in a few situations, it is also more precise and makes the feature more useful.
* For paletted textures do not use the index as alpha anymore but the actual grayscaled color. This is again to make the feature more consistent and useful.
* To compensate for the above there is now a list of hashes for known alpha textures in patch format, so that they don't get broken.
* IMGZ is now considered a grayscale format. There's only two known textures that use IMGZ for something else than crosshairs and those are explicitly handled.
* several smaller fixes.
* the actual color conversion functions for paletted output are now consolidated in a small number of inlines so that future changes are easier to do.
Note: This hasn't been tested yet and will need further changes in the hardware rendering code. As it is it is not production-ready.
Now it is no longer necessary to provide specially set up textures for rendering shaded decals, they can use any PNG texture now that contains a proper red channel.
Handling of the alPh chunk has been removed as a result as it in no longer needed.
Even though unlikely, this should work as a regular texture because it can be used as such.
As a result of the above, true color generation needs to be done explicitly now.
Until now each subclass of FTexture had to implement the entire span generation itself, presumably so that a few classes can use simpler structures.
This does not work if a texture can have more than one pixel buffer as is needed for alpha textures.
Even though it means that some classes will allocate more data now, it's the only way to do it properly.
In addition this removes a significant amount of mostly redundant code from the texture classes.
- added alpha texture processing to all converted classes
As of now this is not active and not tested.
Note that as part of the conversion even those textures that were working as alphatextures will not look correct until the higher level code gets adjusted.
- a bit of header cleanup.
* moved <zlib.h> and <bzlib.h> from files.h to files_decompress.cpp because they are no longer needed for defining the interface.
* added <functional> to the precompiled header
As nice as the automatic is, this will trigger far too many cases where it will disable translucency for mods that only change some texts. Dehacked is very often only used for non-actor related modifications.
If the automatic is supposed to be the default it needs to do a lot more thorough checks to avoid bug reports due to misunderstanding the feature.
It now works the following way:
(0) - Force off (ZDoom defaults)
(1) - Force on (Doom defaults)
(2) - Auto off (Prefer ZDoom defaults - if DEHACKED is detected with no ZSCRIPT it will turn on) (default)
(3) - Auto on (Prefer Doom defaults - if DECORATE is detected with no ZSCRIPT it will turn off)
This was very poorly done without ever addressing the issues a composite render style can bring, it merely dealt with the known legacy render styles.
The same, identical code was also present in two different places.
The oversight that AlterWeaponSprite overrode even forced styles was also fixed.
OpenGL is not implemented yet but with the problems eliminated should be doable now.
