Re:Help the Graphics team track down an interesting WebRende...
by Arelius on Tuesday February 18, 2020 @11:43PM
(#22363091)
Have you spoken to baldurk about this? Since RenderDoc has GL support and it captures with Detours, in theory with some work it should be able to capture at a level above ANGLE. Not sure if that'd be helpful, but I expect maybe.
Alternatively, could you emit markers at or above the Angle level to provide an easier to understand RenderDoc trace?
Re:Help the Graphics team track down an interesting WebRende...
by jonahrd on Tuesday February 18, 2020 @09:47PM
(#22362663)
You can compile ANGLE with ANGLE_ENABLE_DEBUG_TRACE=1 to get a log of all GLES calls to ANGLE's frontend. It's not the same level of detail as RenderDoc, but it's something.
Re:Help the Graphics team track down an interesting WebRende...
by Const-me on Tuesday February 18, 2020 @09:14PM
(#22362513)
> it doesn't capture ANGLE GL function usage
It's opensource with MIT license. I wouldn't expect too many issues supporting such use case, given RD's support for normal OS-supplied GLES. I would only expect complications if you'll try capturing both layers at the same time, GLES between your app and angle, and D3D between angle and d3d11.dll.
Re:Help the Graphics team track down an interesting WebRende...
by jonahrd on Tuesday February 18, 2020 @09:51PM
(#22362673)
RenderDoc does have some issues tracking calls from ANGLE's backend, but workarounds are documented here for some platforms: https://chromium.googlesource.com/angle/angle/+/master/doc/D...
It would be nice to have a native D3D11 backend (and D3D12, and Metal, and Vulkan) someday, but that day isn't today. gfx-rs, or wgpu-rs, looks promising as an abstraction layer over all of these APIs.