This was quite nasty, a multi-day effort to trim this down and then
increasingly growing disappointment as I discovered it was affecting
basically any use of the API.
Ew. At first I tried to just port a growable Array of StringViews (which
would already save quite a lot), but then I realized I have a clear
upper bound on the extensions and so can use a "counting sort" without
having to deduplicate anything after.
After the previous (rather minimal) reduction by the Context cleanup,
this reduces the size of magnum-gl-info.wasm from 245 to 237 kB. Quite
significant, I'd say!
These are in most cases the only strings that are used, and I don't
think having to call std::strlen() for each of them is a good idea if
we don't need to.
We're going to eventually include this class in all Application classes
(need that in order to inherit a to-be-created Configuration class) and
the <string> and <vector> would be just too much. This change caused
magnum-gl-info.wasm (WebGL 2 build) to go down from 247 to 245 kB. Not
much, but that's I guess because there's still a lot other vectors of
strings elsewhere.
There's a lot more places to clean up, will do those in separate
commits. This change is the most atomic I could do, and it introduces a
breaking change to all APIs that returned a std::vector or a
std::string. Fortunately (or as I hope) those weren't used that much, so
it shouldn't cause build breakages for that many people.
Quite a lot of the optimization ideas is borrowed from the new Vk
library -- such as "interning" the driver workaround strings to avoid
allocating their copies.
Not sure what I thought here, but using GL_IMPLEMENTATION_COLOR_READ_*
with glGetFramebufferParameteriv() is possible only since GL 4.5
(there's no extension adding it) and there's no equivalent for it in
GLES 3.1. I assumed that since the entry point is defined in 3.1 then it
would also accept the same enum, but apparently not -- and this
inconsistency is not fixed in 3.2 either.
Funnily enough, all drivers I tried (Mesa ES, SwiftShader, ANGLE) were
accepting this. The first driver that complained was ARM Mali on Huawei
P10 and there I assumed a driver bug -- but on the contrary, this is the
only driver following the spec properly :)
Now the glGetFramebufferParameteriv() API (and the DSA equivalent) is
used only on GL 4.5 contexts and up, everywhere else it's done using
the old-style glGetInteger(). I also renamed the implementations to not
misleadingly mention ES 3.1, since it has nothing to do with this. The
code path dispatch code is a bit entangled due to all the workarounds,
hopefully not breaking any behavior that worked before.
Was Magnum::GL::Extensions::GL before and the redundancy was completely
unnecessary. Potential future extensions coming from GLX, EGL or whatnot
will most probably be in the Platform namespace in a completely separate
file, so this is not a problem.
All code internal to the GL library is affected, not much the outside,
as that is handled by the compatibility alias.
At the moment just the GL library itself w/o the tests, and without
backwards compatibility aliases. The following types were left in the
root namespace, despite being in the GL/ directory, as they will get
moved back soon:
* Image, CompressedImage and their dimensional typedefs
* ImageView, CompressedImageView and their dimensional typedefs
* PixelStorage
Not PixelFormat etc., that one will stay in the GL namespace and a
completely new PixelFormat enum will be provided in the root namespace.
Minimal updates (just the include guards) so Git is hopefully able to
detect the rename and track the history properly.
Everything except Magnum::GL doesn't compile now.
The DSA function does not accept any texture target parameter so the
cube map texture was bound as a whole (and thus behaving as a layered
attachment) instead of just a single face. Thanks to @chpatrick for the
report.
In OpenGL ES 2.0 there is EXT_draw_buffers, which I overlooked somehow,
so I added it to extension list and included in the implementation. It
combines NV_draw_buffers and NV_fbo_color_attachments, so the
implementation now selects one of the two based on which extension is
supported, preferring the EXT one. Updated the documentation to be
less confusing, fixed extension links. Also the single-output
mapForDraw() is not handled separately on ES anymore and just calls
DrawBuffers implementation with single parameter, resulting in less
generated code.
EXT_draw_buffers can also be called on default framebuffer and
apparently in ES there is no way to map front framebuffer for drawing,
so I removed it from the DefaultFramebuffer::DrawAttachment enum.
This was a leftover from some not-well-thought-out design decision. The
function is now used exclusively for binding for draw, as all
framebuffer reading functions (blit(), read()) are doing the read
binding internally. Moreover it required the user to be extra careful on
ES2, because in many cases there are no separate binding points for
reading and drawing.
The function is now parameter-less and always bind the framebuffer for
drawing. The logic for internal binding was also simplified and on ES2
there are separate implementations for single/separate binding points.
For *Framebuffer::checkStatus() the documentation was updated to explain
the meaning of the parameter on ES2 implementation. Also removed the
need for FramebufferTarget::ReadDraw binding, as it was rather
confusing.
Old *Framebuffer::bind(FramebufferTarget) is now just an alias to the
parameter-less function, ignoring the parameter. Along with
FramebufferTarget::ReadDraw it is marked as deprecated and will be
removed in some future release.
The documentation of ARB_invalidate_subdata mentions that all the
functions are really just a hint for the implementation to make some
performance optimizations and they are not affecting behavior at all. So
it's perfectly fine to do nothing if the extension is not supported.
I didn't do this originally as I mistakenly thought that invalidating
depth buffer would somehow behave the same as clearing it, but that's
not the case.
Vladimír Vondruš