3 ms·
That said, it's 2018 and you should be using the GPU for all rendering. The only compositor I know of that actually works without one is plain old Weston, so it
by bitwize 8y ago
That said, it's 2018 and you should be using the GPU for all rendering. The only compositor I know of that actually works without one is plain old Weston, so it's not like falling back to CPU-munged bitmaps is some sort of compatibility mode. If you want to do pixel-level operations on your output framebuffer, write a fragment shader.
- emersion 8y agoSometimes there are tasks that are not really suited for the GPU. For instance, drawing text is much easier to do on the CPU. Also for clients that only draw a few frames occasionally (status bars, background images) drawing on the CPU is totally fine.
- abainbridge 8y agoYeah, that's a thing I should learn to do. When I last tried, about 10 years ago, the OpenGL drivers on my then Windows system used to decide to make my app take 20 seconds to start every now and again. The OpenGL init function would just block for that long. That was a show stopper for me. Also, my machine uses the graphics hardware embedded in the processor. That shares system DRAM. The DRAM bandwidth is already the bottleneck with my CPU routines, so it's not clear how much performance benefit the GPU would offer (although there could be power savings, I guess). But basically, my CPU based rendering code is already fast enough and I'm pretty sure it's simpler than getting the GPU involved and easier to debug and more portable (to, say, an Amiga 1200). Therefore I consider writing a fragment shader to be a form of premature optimization. I realize this makes me a mad man :-)