mitchellhansen
|
314fa3e4af
|
Workgroup sizes of 8 along with reducing the amount of work groups improves performance 20x
|
5 years ago |
mitchellhansen
|
711e678969
|
That's it. Fifo present mode halts until the next time the screen is refreshed. Mailbox will replace the lined up image if the refresh period has not been reached
|
5 years ago |
mitchellhansen
|
9135d9ec84
|
minimal example of 15ms lag
|
5 years ago |
mitchellhansen
|
3f14555c99
|
texturing works, next will be images
|
5 years ago |
mitchellhansen
|
a2de3557f0
|
removed old shader_kernel
|
5 years ago |
mitchellhansen
|
ebf2cb5322
|
running. Rust is absolutley amazing. I haven't run this in ~20 commits, yet it only took a few unwrap() fixes and it's back working
|
5 years ago |
mitchellhansen
|
0357296771
|
more refactoring, compiles.
|
5 years ago |
mitchellhansen
|
1737319fc5
|
lots of refactoring
|
5 years ago |
mitchellhansen
|
878b37c3e0
|
cleaned loop. Compiles. Refactoring Canvas next
|
5 years ago |
mitchellhansen
|
66d11e22c0
|
just doing a naive layered compute for meow
|
5 years ago |
mitchellhansen
|
885e19fb64
|
starting to hook up the compute frame translation.
|
5 years ago |
mitchellhansen
|
a3607ebc7d
|
Most of the compute side is mocked up and compiling.
|
5 years ago |
mitchellhansen
|
9b8d5cd828
|
have the api of the compu somewhat figured out. Need to weigh my options on texture, buffer, and image handles and how I'm going to get them from the vkprocessor
|
5 years ago |
mitchellhansen
|
e0e9218d6e
|
brainstorming
|
5 years ago |
mitchellhansen
|
c39994a7ae
|
more brainstorming on how spriting and computing is going to work
|
5 years ago |
mitchellhansen
|
be20f3ae2a
|
sprites are working, but the implementation is pretty bad
|
5 years ago |
mitchellhansen
|
e351e36c92
|
pulled out the draw call. forgot to allocate buffers
|
5 years ago |
mitchellhansen
|
0be20eb959
|
removed the extra stuff from the kernels. Having some issues with the layouts not agreeing with the set size...
|
5 years ago |
mitchellhansen
|
dfd4cbb6a2
|
Moved the framebuffer gen to the canvas. Cleaned up a lot of unused imports
|
5 years ago |
mitchellhansen
|
8cd5e3e562
|
Finally appeased the compiler. Need to do some real cleaning up on the relationship between canvas and vk processor
|
5 years ago |
mitchellhansen
|
f06adc9a55
|
ugga ugga. Lifetime problems with the Canvas
|
5 years ago |
mitchellhansen
|
a7e79668eb
|
very close to the skeleton of the canvas written. Still haven't tested it but it's looking good. Need to figure out how I'm goign to iterate shader_kernels without borrowing from self
|
5 years ago |
mitchellhansen
|
2207c41956
|
Brainstorming the design.
|
5 years ago |
mitchellhansen
|
c5b3c29ad4
|
in the midst of a very messy refactor of the way i build the command buffer
|
5 years ago |
mitchellhansen
|
56455774bc
|
cleaning up the shader hoohaw
|
5 years ago |
mitchellhansen
|
bb1f782168
|
pulled swapchain out of shaderkernels
|
5 years ago |
mitchellhansen
|
28878dc345
|
fixed minor regression, cleaned up source tree
|
5 years ago |
mitchellhansen
|
1f33d96ae5
|
Added sprite and better vertex format. works, but now I lost texturing
|
5 years ago |
mitchellhansen
|
cce893a0c8
|
saving working on key entry. going to do pallette now
|
5 years ago |
mitchellhansen
|
a91e071909
|
refactored the compute image
|
5 years ago |
mitchellhansen
|
a8679459ea
|
I have kinda given up trying to wrangle this descriptorsetbuilder. Going to read more on how sprites are best handled in regards to their textures
|
5 years ago |
mitchellhansen
|
9951afd5a5
|
.
|
5 years ago |
mitchellhansen
|
982e2ad517
|
initial impl of get image set
|
5 years ago |
mitchellhansen
|
77ccd415d0
|
easiest solution there was just to remove the unclonable objects from ShaderKernels
|
5 years ago |
mitchellhansen
|
db06459bd6
|
still working on getting ownership of everything squared away
|
5 years ago |
mitchellhansen
|
321f30b4cc
|
The entry point takes a borrow of the shader module which is then cloned by the pipeline. So I gotta keep the shader, entry, and modules in a high scope
|
5 years ago |
mitchellhansen
|
2327a7f05f
|
added compute kernel helper
|
5 years ago |
mitchellhansen
|
e476cb2e4e
|
It was the gamma being wrong. Interesting how much different this is working than SFML+OpenCL
|
5 years ago |
mitchellhansen
|
58f5289a3b
|
wasn't the layout. It was the colors being bound between 0-1 instead of 255 like I was expecting
|
5 years ago |
mitchellhansen
|
5928eb5dde
|
getting close, something wrong with copying the buffer. Probably the format that I've selected
|
5 years ago |
mitchellhansen
|
26410b78a2
|
Got something working. Need to test and verify I'm getting the data from the compute kernel
|
5 years ago |
mitchellhansen
|
4d5b09e9ca
|
adding a second image buffer for GPU only access. Immutable buffer is write once for textures only. No sampler for regular image buffer though??
|
5 years ago |
mitchellhansen
|
937b4fc8ca
|
added texture. looks like copying the one from the compute output isn't going to totally work out of the box
|
5 years ago |
mitchellhansen
|
5751965ce3
|
got everything worked out in the loop. I guess next will be to swap the compute buffer to the image buffer
|
5 years ago |
mitchellhansen
|
3f79b276a9
|
got it working again. Finally clicked with me what was going on with the functions & how to specify their generic types
|
5 years ago |
mitchellhansen
|
a71958e815
|
working on porting. Got some stuff cleaned up but I've broken the example I had running at one point
|
5 years ago |
mitchellhansen
|
86b4f7d9a3
|
woweee, vkprocessor compiles now
|
5 years ago |
mitchellhansen
|
f5bf07085d
|
About 1/3 of the way done encapsulating all this vulkan boilerplate into a struct
|
5 years ago |
mitchellhansen
|
5da2f9f0b5
|
had a bit of a time finding the problem with the specialization constants being unimplemented. Adding an impl lets rust elicit the pipeline type so I can store it, thank god
|
5 years ago |
mitchellhansen
|
c8a968f4f6
|
porting over the graphics example
|
5 years ago |