1. Add one more command buffer to accommodate uploading stuff inbetween frames when there are already 2 frames in flight on GPU.
2. Remove committed from to-upload buffer/image tracking, as it is completely unnecessary.
- make staging/prep phase use separate command buffer
- flush this command buffer early if needed
- move command pool to its own module
- move shader loading stuff to pipeline module
- cleanup lots of command buffer passing for model loading, it can use staging explicitly now
- Clear split between static and dynamic geometry within the same buffer
- Store previous frame geometry data longer
- Use simpler ring buffer for allocation
- Add waf unit tests for alolcator
Known issues:
- ray tracing still glitches a lot
adds multiple command buffers and bits of synchronization.
is not correct at all:
- no depth buffer barrier
- no ubo slots in render
- no proper render memory management
- no ray tracing support whatsoever
- ... etc
Allocate device memory in big 128Mb chunks. Sub-allocate buffers, images
and other resources within those chunks.
This uses a dumb grow-only bump allocator for now. Will not work for
more than a couple maps.
now vulkan can do screenshots too. however, it depends on blits being
available between swapchain and rgba8 formats, so if it's not the case
for some reason, the resulting image will have mixed components.
this basically checks that acceleration structurs are building without
any vulkan errors (we still don't know whether they are bulding with
correct geometry though).
we also have compute shader pipeline ready for ray tracing (yay)
but it doesn't do any raytracing yet
Now each geometry is first allocated a slot with VK_RenderBufferAlloc,
and then VK_RenderBufferLock/Unlock are used to upload buffer contents.
This allow for two things:
1. Uploading buffer data to GPU memory on/after Unlock.
2. (Re)building BLAS for RTX on/after Unlock.
These buffers are now directly referenced by render_draw_t, which also
will be helpful in the future for both renderers.
1. Improve performance by coalescing multiple surfaces/draw calls with
the same textures (within a single brush model).
2. Add lightmaps. We don't support lightstyles yet, though, so they seem
a bit wrong at times. Also gamma handling is not great.
Oh well.
it's now *very slow* because every surface (which is just a few polys)
gets its own draw call and descriptor bind. F
we also had to make default sampler repeating, so now 2d textures in
menu have weird edge artifacts. also F
making progress i guess
