We now do certain image layout transitions in the render pass instead of a barrier in EndRenderPass.
There is also an additional barrier on buffer uploads to prevent write-after-write hazard. It's a kludge on the fact that we're only tracking the most recent resource access.
Reviewed-on: #49
continuous-integration/drone/push Build is passingDetails
This should fix an issue where draw calls could flicker if a defrag was in progress and a uniform buffer was being used.
Uniform buffer "pools" are now just a single dedicated VulkanBuffer, and the uniform buffer objects are offsets into that buffer.
Reviewed-on: #48
continuous-integration/drone/push Build is passingDetails
Some minor stuff that's cropped up from the D3D11 work so far. This PR updates the Driver_Template with the latest API, and also fixes a crash in the Vulkan driver -- if you acquired a swapchain texture from a window that had been destroyed, there was no null check before de-referencing the WindowData.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #46
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
continuous-integration/drone/tag Build is passingDetails
continuous-integration/drone Build is passingDetails
Various changes to reduce and optimize memory usage.
- Defragmenter
- Allocate 4 16MB transfer buffers for pool
- If transfer is larger than 16MB, create temporary transfer buffer
- Fixed some issues with CopyTextureToTexture
Reviewed-on: #41
continuous-integration/drone/push Build is passingDetails
Depth textures handle MSAA differently than color textures do, because there's no need for a resolve texture. This means the root VulkanTexture can have the sample count instead of needing an additional `msaaTex`.
This changeset also fixes a bug where the depth buffer wasn't getting cleared if there were MSAA color attachments.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #38
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
This change makes Refresh behave more like FNA and other rendering APIs, where user-side textures have a sample count instead of generating MSAA RTs on the fly.
This should theoretically reduce memory consumption since subresource views no longer generate their own MSAA textures. Instead, each texture with a sample count > 1 now has a reference to an MSAA texture baked into the root texture itself, so views can just reference that.
This also simplifies VulkanRenderTarget significantly, to the point where it's just a wrapper around VkImageView. We could probably remove the abstraction entirely at this point.
Since VkRenderPass objects still require knowing the sample count, we can use the first bound texture's sample count. This means users only have to specify sample count in two places -- the graphics pipeline, and the texture. Easy enough.
I also noticed and fixed a bug with clear values with MSAA levels > 1.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #37
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
Before this change, framebuffers were created based on the attachments' full dimensions, instead of the dimensions of the requested mip level. This fixes validation errors in the RenderTextureMipmaps test program.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #36
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
The image type getting passed to imageCreateInfo was always VK_IMAGE_TYPE_2D, even for 3D textures. This fixes that bug, allowing Vulkan to successfully create 3D textures.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #35
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
This PR adds a shader cross-compiler program (written in C#) that uses glslc and spirv-cross to compile GLSL source files to a custom, Refresh-specific shader blob format that contains shader code for various backends.
The goal is that whenever you want to compile your shaders, you just run this program (either via `refreshc` or `dotnet run`), passing in the path to the GLSL source file. You can specify which backends you want to support via flags (`--vulkan`, for instance), and it will generate a .refresh blob file for Refresh to consume.
This can also be extended via C# defines and the `partial class` system. An example is the PS5 shader implementation, whose logic is squirreled away in the NDA'd repository but can be called in sequence with the rest of the shader cross-compile logic. This way, we don't have to run two separate programs to compile our shaders to all supported platforms.
Refresh handles all the parsing of the file format in Refresh_CreateShaderModule, so individual backends do not need to concern themselves with the particulars.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #34
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
Removes the `fixed` parameter from AcquireCommandBuffer, and removes a couple other unused bools from the VulkanCommandBuffer struct.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #33
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
The validator layer was complaining about us updating bound descriptor sets, this is technically a bug in the validator layer but resetting the command buffer in Cleanup makes more sense anyway. Also changed the wait for timeout call to a fence status query to make it easier to understand.
Reviewed-on: #32
continuous-integration/drone/push Build is passingDetails
Fixes a validation error when attempting to draw multiple primitives via an indirect draw call. Adds some documentation for how indirect draw buffers should be set up.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #30
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
This makes the order of VULKAN_Submit consistent with VULKAN_Wait.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #28
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
**Breaking API Changes**
* Removed `REFRESH_SAMPLECOUNT_16/32/64`, since hardware support for these sample counts is generally poor (and completely non-existent with MoltenVK and certain consoles).
* Removed unused `sampleCount` parameter from `Refresh_TextureCreateInfo`.
* Removed `sampleCount` parameter from `Refresh_ColorAttachmentDescription`. The existence of this parameter meant you had to sync up three different sample count values (render pass, pipeline, and color attachment description) whenever you wanted to use multisampling. However, Vulkan requires that all color attachments in a given pipeline _must_ match the pipeline's sample count anyway, so we can assume all color attachments will use the pipeline's sample count.
* Removed the `renderArea` parameter from `Refresh_BeginRenderPass()` since it didn't serve much practical purpose and slightly complicated things on the MoonWorks managed side.
**Behavior Changes**
* When creating a render pass or graphics pipeline, the requested multisample count will be converted into a sample count that's actually supported by the GPU. For example, if you request 8x MSAA on a device that only supports up to 4x MSAA, it will silently fall back to 4x MSAA.
* All color attachments are now forced to have an internal store op of `STORE`, even if `REFRESH_STORE_OP_DONTCARE` is specified. The one exception is internal multisample textures -- if `DONTCARE` is used, those textures will be discarded to save on bandwidth. (Their resolve textures will still be stored.)
* The RenderPass hashing logic was updated so that it would still work correctly with the removal of `Refresh_ColorAttachmentDescription.sampleCount`.
**Bug Fixes**
* Fixed bugs where multisampling logic wasn't kicking in for certain sample counts due to incorrect enum comparisons.
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #26
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
- Fixed bug where multisample enum values were not being translated into their equivalent Vulkan enums
- Fixed bug where MSAA attachments in transient render pass creation were using a sample count of 1
- Fixed bug where the clearValues array in BeginRenderPass was not being populated with clear values for multisample attachments. (Modeled the fix after https://github.com/FNA-XNA/FNA3D/blob/master/src/FNA3D_Driver_Vulkan.c#L8723)
- Fixed bug where the multisample texture was not being transitioned when beginning a render pass
Reviewed-on: #25
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
- The application must now call `Refresh_SetBackend` before creating the device
- `Refresh_SetBackend` can set a preferred backend, but will fall back if it is not available
- Device creation no longer takes presentation parameters
- Windows must now be explicitly claimed by the device
- Windows can be unclaimed by the device
- Windows can have their swapchain present mode changed after creation
Reviewed-on: #23
continuous-integration/drone/push Build is passingDetails
This PR contains the general changes needed for the PS5 driver (being developed separately, outside this repository).
Co-authored-by: Caleb Cornett <caleb.cornett@outlook.com>
Reviewed-on: #19
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
Fixes various errors related to the swapchain being recreated.
Additionally, `BeginRenderPass` now allows a NULL `renderArea` parameter. If NULL, Refresh will select a sensible default render area.
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
continuous-integration/drone/push Build is passingDetails
Removes `Refresh_RenderTarget`, `Refresh_CreateRenderTarget` and `Refresh_QueueDestroyRenderTarget`. Render targets are now managed by the implementation.
Adds `REFRESH_TEXTUREFORMAT_B8G8R8A8`.
Adds `Refresh_AcquireSwapchainTexture`. Returns a swapchain texture for the requested window.
Removes `Refresh_QueuePresent`. It is now assumed that the texture returned by `Refresh_AcquireSwapchainTexture` will be presented. This texture can be manipulated like any other texture.
Adds `Refresh_GetSwapchainFormat`. Returns the swapchain format for the requested window.
Reviewed-on: #15
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
continuous-integration/drone/push Build is passingDetails
Removes `Refresh_RenderPass` and `Refresh_Framebuffer` objects.
`Refresh_BeginRenderPass` now takes a set of `Refresh_ColorAttachmentInfo` structs and an optional `Refresh_DepthStencilAttachmentInfo` struct that describe the render pass. The render pass and framebuffer objects are now managed by the implementation instead of the application.
Accordingly, `Refresh_GraphicsPipelineCreateInfo` now takes a `Refresh_GraphicsPipelineAttachmentInfo` struct that describes render passes that may be used with the pipeline. It is an error to bind a pipeline during an incompatible render pass.
Reviewed-on: #14
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
continuous-integration/drone/push Build is passingDetails
Rework submission flow to reduce waiting and improve the resource cleanup process.
This patch also gets rid of descriptor set caching.
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
continuous-integration/drone/push Build is passingDetails
This allows us to avoid monotonous boilerplate when bringing up new backends.
Co-authored-by: TheSpydog <thespydog@noreply.example.org>
Co-committed-by: TheSpydog <thespydog@noreply.example.org>
continuous-integration/drone/push Build is passingDetails
Buffer uploads now require a command buffer, like textures do. Additionally it is now regarded as an error to upload data in the middle of a render pass.
Reviewed-on: #6
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
continuous-integration/drone/push Build is passingDetails
We used to have monolithic uniform buffers on the VulkanRenderer object, but this was inefficient in the case of threaded usage.
Now, we have a pool of uniform buffers. A uniform buffer is selected from the pool when a pipeline is bound. The uniform buffers are rotated upon presentation.
Now pushing uniforms is now a concern of the command buffer instead of the pipeline. The pipeline should just always have been a static object anyway.
Additionally, we now do extra buffer record-keeping so that buffer data can be updated after a bind/draw.
Fence submission has also been restructured so that submissions don't cause unnecessary blocks. Now we assign one fence per submission, and we don't wait for fences until it's time to present.
Reviewed-on: #1
Co-authored-by: cosmonaut <evan@moonside.games>
Co-committed-by: cosmonaut <evan@moonside.games>
cosmonaut
TheSpydog
Hazel Stagner
David Weil
thatcosmonaut