magnum

Commit Graph

Author	SHA1	Message	Date
Vladimír Vondruš	a2554cb656	TextureTools: don't clear the output in DistanceField. The test passes now. This made the benchmark run significantly faster. ~200 µs instead of 670. Are clears really taking up more time than all those texture fetches? Strange.	3 years ago
Vladimír Vondruš	abf8a759f2	TextureTools: verify DistanceField doesn't overwrite the whole output. It does that, yeah. Another reason why incremental filling couldn't have ever worked.	3 years ago
Vladimír Vondruš	c92e1b55e5	TextureTools: properly test the DistanceField framebuffer overload. Yeah, it doesn't work with offsets. Of course.	3 years ago
Vladimír Vondruš	ada0645f34	TextureTools: make DistanceField working with subrectangles. Just cannot use gl_FragCoord in here. That's it, that's the fix. What's however COMPLETELY unexpected is that this simple change made the process significantly faster on my Intel GPU, from ~815 µs to 670! I can't even pretend I understand what's going on here, but maybe doing less math in the fragment shader when calculating the texture coordinates (and thus possibly the driver having a better idea how to prefetch or schedule?) is what made this faster? Or maybe it's due to one uniform input less, and two interpolated values instead of four on the way from the vertex shader?	3 years ago
Vladimír Vondruš	89af6ff82e	TextureTools: fix DistanceField to not be slightly shifted. The original implementation wrongly assumed that the input and output pixel centers align, which would only be a case if the ratio of the input and output sizes would be odd. Which it in practice isn't, usually it's a 1024x1024 texture scaled down to 128x128 or something like that. The flipped test cases added in the previous commit now pass. According to the benchmark, the new code is very slightly slower (~815 µs vs ~805 before). The new code isn't really more complex than the old one, it just does slightly different work -- there are new corner case in the initial logic for marking the pixel inside or outside, on the other hand some corner cases that had to be handled in the previous case are no longer a thing.	3 years ago
Vladimír Vondruš	93137eb025	TextureTools: test axis-flipped DistanceField processing. It should result in exactly the same output when flipped back, i.e. no accidental shifts happening.	3 years ago
Vladimír Vondruš	6357ed52c0	TextureTools: assert that DistanceField output is framebuffer-drawable. If it's not, it's a programmer error (i.e., don't use Luminance or packed formats, won't work), and since there's no way for the API to report a failure in a programmatic way, this was causing hard-to-track errors.	3 years ago
Vladimír Vondruš	94c31b13e2	TextureTools: it's not SwiftShader, it's Luminance not being renderable. And the thing that changed for SwiftShader 4.1 was added EXT_texture_rg support, which made the test work. So no, this was not a SwiftShader bug at all.	3 years ago
Vladimír Vondruš	f39beda8f1	TextureTools: remove test code that assumed GL debug output is always on. It's not the case for quite a while, and especially for GPU benchmarks it'd be harmful.	3 years ago
Vladimír Vondruš	f2dd0c5d75	TextureTools: make the DistanceField benchmark more stable. Use a different overload to not have to create a temp framebuffer every time, move also the error checking outside the loop and increase the iteration count to actually have a result that isn't wildly different every time.	3 years ago
Vladimír Vondruš	c60a1a98b9	TextureTools: no, it's no longer creating a shader every time. This was fixed back in `f0bb710cd3`, shortly after adding the benchmark, but the comment wasn't removed by accident.	3 years ago
Vladimír Vondruš	f091b80fae	TextureTools: make the DistanceField class movable. No reason not to, even though the move is destructive. Also unblocks Text::DistanceFieldGlyphCache which wasn't movable due to this and one other problem.	3 years ago
Vladimír Vondruš	15d0727096	TextureTools: remove redundant namespace prefix from test code.	3 years ago
Vladimír Vondruš	484596f936	TextureTools: remove redundant include from a test.	3 years ago
Vladimír Vondruš	78a93734dd	TextureTools: pass the whole test dir path to the test.	3 years ago
Vladimír Vondruš	3b32483d0b	TextureTools: (expected failing) test for DistanceField with an offset. The shader uses gl_FragCoord, which doesn't really work with a viewport that doesn't start at origin. Gotta fix this.	3 years ago
Vladimír Vondruš	bbb54a9df8	It's 2023 already!	3 years ago
Vladimír Vondruš	5f55f9e756	TextureTools: add a DistanceField overload taking a framebuffer. So one can directly read it back on GLES without having to wrap the texture in a framebuffer again. This change also puts the framebuffer completeness check before the clear() and bind() which makes it no longer emit a GL error. The error is still silent though, which isn't nice. Gotta fix that eventually as well.	3 years ago
Vladimír Vondruš	5d0454ce98	Remove random unneeded <string> and StringStl.h includes. Or document why is it still needed.	3 years ago
Vladimír Vondruš	0001308d61	Test: port file/image comparison tests to the new Utility::Path.	4 years ago
Vladimír Vondruš	822fa0d644	Updated copyright year.	4 years ago
Vladimír Vondruš	361d87cd9e	Shaders,TextureTools: XFAIL tests that hit weird bugs on SwiftShader 4.0. No patience to investigate.	4 years ago
Vladimír Vondruš	04aeebaf1d	TextureTools: print a note if we're using EXT_texture_rg in a test.	4 years ago
Vladimír Vondruš	ce63928efc	Make plugin existence checks in tests consistent.	4 years ago
Vladimír Vondruš	65a935cedc	Test: adapt to changes in CORRADE_SKIP(). Chose to update the tests instead of adding a compatibility DebugStl.h include to Tester.h, which would have a significant impact on compile times.	5 years ago
Vladimír Vondruš	3d136503d8	Updated copyright year.	5 years ago
Vladimír Vondruš	19e0e96d74	Updated copyright year.	6 years ago
Vladimír Vondruš	65a2ced2a7	Make this compile and test cleanly with CORRADE_NO_ASSERT defined.	6 years ago
Vladimír Vondruš	398988193a	TextureTools: increase fuzzy compare threshold for iOS/WebGL.	7 years ago
Vladimír Vondruš	93789721b2	De-inline Image headers, remove dependency on ImageView. This means users have to add some more #includes on their side. Sorry but it's for your own good :P	7 years ago
Vladimír Vondruš	b97c27ba82	DebugTools,TextureTools: better diagnostics for missing plugins in tests.	7 years ago
Vladimír Vondruš	67cf914864	TextureTools: avoid the test looking for plugins we didn't supply.	7 years ago
Vladimír Vondruš	9e3690670a	TextureTools: explicitly force operation sequencing in the test. On GCC 4.8 at least, release() gets called on the image before size(), causing size() to be zero.	7 years ago
Vladimír Vondruš	4ff67d53b1	TextureTools: don't benchmark GPU if we don't have time queries. Makes a loud BOOM on SwiftShader.	7 years ago
Vladimír Vondruš	49d70377af	Test: fixed various unused function warnings.	7 years ago
Vladimír Vondruš	f9f3df69de	Get rid of std::unique_ptr.	7 years ago
Vladimír Vondruš	2149f7b869	Wrap all tests in unnamed namespaces. This makes the compiler warn about test cases that aren't ever used, which is a good thing.	7 years ago
Vladimír Vondruš	e6b7aa12fa	Updated copyright year.	7 years ago
Vladimír Vondruš	fe6fedf6c8	TextureTools: add some fuzziness to DistanceField test comparison.	8 years ago
Vladimír Vondruš	5cc140d60c	TextureTools: make the test files properly accessible on an iOS device. Testing exclusively on the iOS Simulator isn't a good thing after all.	8 years ago
Vladimír Vondruš	b9739a97ff	TextureTools: make distance field test working on WebGL 1 as well. And remove the implicit assumptions, making it more robust.	8 years ago
Vladimír Vondruš	f0bb710cd3	TextureTools: make distance field processing into a stateful class.	8 years ago
Vladimír Vondruš	1fa1869ba7	TextureTools: add a benchmark for distance field processing. It's a bit shitty because we're recreating the shader and everything every iteration.	8 years ago
Vladimír Vondruš	c983068fcc	TextureTools: initial test for distanceField(). Fully passes only on desktop and ES3 (Mesa), expecting minor differences onother GPUs. ES2 is slightly broken and needs fixing; doesn't even compile on WebGL 1 and causes a serious GPU stall on WebGL 2 -- in both causes caused by the unbounded nested loops. Rendering doesn't work on WebGL 1 at the moment, since luminance formats are not renderable. And for a RGBA output format I would need some utility to get rid of the extra channels in order to pass the comparison. Lots of work to do here.	8 years ago

44 Commits (dcf6652764d199e314f6b4612f2b88271c9e9437)