We now sample textures from the device owned image samplers.
Passing of enabled texture units has been simplified by only passing a
list of texture unit indices.
This introduces a new library, LibSoftGPU, that incorporates all
rendering related features that formerly resided within LibGL itself.
Going forward we will make both libraries completely independent from
each other allowing LibGL to load different, possibly accelerated,
rendering backends.
This is a deprecated feature that is still in use in some older games,
most notably Grim Fandango in ScummVM makes heavy use of it.
Luckily, since you can only draw convex polygons, the end result is
exactly the same as when you would have used `glBegin(GL_TRIANGLE_FAN)`
- so we just reuse that code as-is.
Implement support for the `GL_TEXTURE` matrix mode, the texture matrix
stack and texture coordinate matrix transformation.
Also, an unused `m_current_matrix` was removed to make room for
`m_texture_matrix`.
In OpenGL, texture coordinates can have up to 4 values. This change
will help with easy application of texture coordinate matrix
transformations in the future.
Additionally, correct the initial value for texture coordinates to
`{ 0.f, 0.f, 0.f, 1.f}`.
Before, `SoftwareRasterizer` was iterating over all 32 possible texture
units for each fragment and checking each if they're bound to a texture.
After this change, an intrusive list containing only texture units with
bound textures is passed to the rasterizer. In GLQuake, this results in
a performance improvement of ~30% (from 12 to 16 FPS in the first demo)
on my machine.
This smaller block size allows us to use an `u8` for the pixel mask
during triangle rasterization. These changes result in better FPS in
3DFileViewer, which periodically shoots up to 250-300 FPS on my
machine. Before, the peaks were somewhere in the range of 160-170 FPS.
Previously we multiplied the interpolated texture coordinates by
width - 1 and height - 1 instead of width and height which resulted in
some wrongly mapped textures, especially visible in the glquake light
maps.
This also corrects the wrap mode being wrongly swapped for s/t
coordinates.
Since we do not have texture borders implemented yet we always use
GL_CLAMP_TO_EDGE for all clamping wrap modes for the time being.
If an unsupported drawing mode was configured, `glEnd` would not reach
the line where `m_in_draw_state` was set to `false`. This caused all
subsequent invocations to `glBegin` to fail.
Most resources seem to suggest that points on the clip planes are also
inside the frustum, e.g.: https://www.cubic.org/docs/3dclip.htm
This seems to resolve some rendering issues in Grim Fandango in OpenGL
mode where the screen remained black. This was because of quads being
drawn with their vertex positions exactly on the clip planes.
In its current state, ScummVM seems to invoke these methods just after
destroying the current GL context. According to the OpenGL spec:
"Issuing GL commands when the program does not have a current
context results in undefined behavior, up to and including program
termination."
Our old behavior was to deref a `nullptr`, which isn't that great. For
now, protect these two methods. If other ports seem to misbehave as
well, we can always expand the check to other methods.
The `glGet*` family of functions requires that all parameters of
different types are transparently converted into each other. For
example, you can request a boolean parameter as a float or a list of
double values as an integer. It might be considered bad practice to
request parameters through the wrongly-typed function, but to be spec-
compliant we need to implement this.
Introduce a new `::get_context_parameter()` to obtain a parameter
value, which is then converted to the right type by the respective
`::gl_get_*()` functions.
