Compared to Corrade, the improvement in compile time is about a minute
cumulative across all cores, or about 8 seconds on an 8-core system (~2
minutes before, ~1:52 after). Not bad at all. And this is with a
deprecated build, the non-deprecated build is 1:48 -> 1:41.
This caused MeshToolsCompileGLTest to fail in a strange way, and
PhongGLTest::renderLowLightAngle() as well. Which looked rare enough that
I first suspected some random driver bug, but apparently it was all
caused by these using the default infinity range instead of explicitly
calling setLightRanges() on the shader.
The test is now updated to explicitly verify the default value when a
setter isn't called, to catch this problem better in case it reappears in
a different form elsewhere.
The output is now *much* closer to the Vector shader output (mean
difference went from ~6 to 1.79), and in the multi-draw case it's
clearly visible that it's no longer slightly weirdly uneven.
The test files are now also much smaller as the originals were created
before RLE was implemented in TgaImageConverter.
Not that C++ STL and exceptions would be anything to take inspiration
from, but there's std::out_of_range. Python IndexError is also specified
as "index out of range", not "bounds".
Partially needed to avoid build breakages because Corrade itself
switched as well, partially because a cleanup is always good. Done
except for (STL-heavy) code that's deprecated or SceneGraph-related APIs
that are still quite full of STL as well.
Interesting, didn't know this kind of feature support was possible. I
guess it's still better than no GLES3.2 at all. Also, it's a phone from
2017, so probably not all that important to care about anymore anyway.
Most of the testing scaffolding here is a preparation for the actually
complex formats like BC6/7 or ASTC. Also, it's great to be able to use
Magnum from Python to prepare data for testing the C++ Magnum APIs.
So it's possible to have light culling enabled on, say, 64 lights, but
with only at most 3 applied each draw, allowing the shader compiler to
unroll the loop if it makes sense. This also better prepares for SSBO
support where the total light count would be unbounded and thus the
value ignored, and thus the value can be 0.
This prepares for SSBO support where the total count is unbounded (and
thus the value is ignored, thus it can be 0).
Also regroup the doc paragraphs so it's clear what's related to UBO
usage and what applies to classic uniforms as well.
Currently just the bare minimum, more features such as handling
multiple contiguous strips and loops inside a single mesh or an
overlapping layout will come later.
The shader requires the input data to be laid out in a rather specific
way, and there will be a dedicated MeshTools utility for it in the
following commits. For independence though, the shader tests use a
custom helper.
The initial implementation has certain corner cases which will be
eventually resolved. For now they are pinned down with repro cases in
the test. But apart from that, it's pretty much usable in practice.
Remaining join styles (round and miter-clip) as well as stipple support
will eventually follow as well.
Also fixes a WebGL error in the tests. I suspect it might fail elsewhere
as well. Nevertheless, this still doesn't mean that it would be
impossible to use dynamic joint count with UBOs -- simply pad the UBO in
that case.
Apparently a framebuffer attachment can be bound only if it's actually
written to by a shader, thus instanced test cases that had it set up
always were failing with a GL error if the shader didn't have ObjectId
output enabled.
I unified both into a single renderSetup()/renderTeardown() routine,
where the ObjectId renderbuffer is attached always, but then only the
test cases that actually use it are mapping it explicitly. And clearing
as well, because apparently it can be cleared only if it's mapped. Huh.
Given that I made a breaking change by returning Containers::String
instead of a std::string, I can take it further and replace also
std::pair with Containers::Pair -- it won't bring any more pain to the
users, they have to change their code anyway.
Co-authored-by: Hugo Amiard <hugo.amiard@wonderlandengine.com>
It's not a GL error, and allows the application to compile just a single
shader for all skinned meshes, not one for each skeleton size. Together
with the dynamic per-vertex joint count this means the app only needs a
single shader for all skinned meshes, which is nice.
Those were temporarily added until Trade::AbstractImporter is ported
away from std::string, and that's done for quite some time already, so
these are no longer needed.
High-level docs with examples will be written once there's corresponding
support in MeshTools::compile() *and* in importer plugins, as skinned
meshes are usually brought in from files, never set up directly.
Co-authored-by: Squareys <squareys@googlemail.com>
Vladimír Vondruš