Dynamic rendering #92
Conversation
Changing local variables naming to lowerCamelCase Co-authored-by: Johannes Unterguggenberger <[email protected]>
| // which would denote how to use the attachment - use this attachment as stencil, depth or color | ||
| if(is_depth_format(currAttachment.format())) | ||
| { | ||
| if(depthAttachment.has_value()) |
There was a problem hiding this comment.
One more issue I wanted to discuss. There is no way (at least if I understood the code correctly) to have separate depth and stencil attachments.
For example suppose I have two D24S8 attachments and I'd like to use the first one as depth and the second one as stencil. Because the stencil/depth attachment state is directly deduced from the format of the attachment and both of them have combined format there is no way to know which one to use as which.
I was thinking of adding an optional flag that would mark the attachment as being either depth only or stencil only, do you agree with that?
There was a problem hiding this comment.
Yeah, more options are always a good thing. Please proceed as you have planned!
There was a problem hiding this comment.
This would get resolved by the change I propose in this comment. I could use the dynamic_usage of each attachment to alleviate this issue.
…to dynamic_rendering
- removed dynamic_rendering_attachment - added declare_dynamic(_for) functions to avk::attachment which create dynamic rendering attachments - changed graphics_pipeline_t::renderpass_pointer() back to renderpass_reference() which now returns a reference wrapper
…eeds to be discussed further)
|
So I added an initial implementation of MSAA, however I am not very happy with it as is. I tried to repurpose subpass_usage for this, however I believe it should be it's own configure structure. Subpass usage(s) have a bunch of other stuff that are not relevant at all to what I am trying to use it for. Instead I would propose something like "dynamic_usage" which would store usage flag (ie. color, depth, stencil + resolve information). What is your opinion on this? I feel (again) that the usage of subpasses might be misleading, however you might like it as it is in the spirit of what we ended up with in regards to the attachments. Lastly I have a question, why does subpass_usage not have color resolve mode? It only has depth or stencil ones. |
…modified the dynamic functionality to conform to these usages
#include <vulkan/vulkan_enums.hpp> #include <vulkan/vulkan_structs.hpp>
…ImageLayout::eAttachmentOptimal
Opened a preliminary pull request for reviewing