Found when running in valgrind looking at another issue.
- RenderPass' depth_stencil_target
Tripped in SDL_BindGPUFragmentSamplers when not binding a DS target
- VulkanCommandBuffer's swapchainRequested
Tripped in VULKAN_Submit for the end transition barrier when
creating an image. The field is only reset when reused, not on first
use
(cherry picked from commit 265236d952)
We previously thought this wasn't possible because constant buffer offsets and
partial updates were unavailable, but we were reading the wrong table - this is
only the case for D3D11...
https://learn.microsoft.com/en-us/windows/win32/direct3d11/overviews-direct3d-11-devices-downlevel-intro
... while 12 doesn't list this feature at all:
https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-levels
We double checked and Jesse Natalie confirmed that this feature is required for
D3D12 even for 11_0 drivers. (Thanks Jesse!)
Additionally, D3D12 requires that UAVs are accessible from all shader stages,
meaning Tier 2 is enough to support the number of UAVs we need. Tier 1 could be
a property to lower the requirements, but that can be done later.
When skipping presentation due to the window being hidden, presentDataCount is not incremented on the command buffer, and subsequently the submitted command buffers will not be cleaned up as long as the window is hidden. This results in a lag spike when showing the window due to all previously submitted command buffers suddenly being cleaned up at once, and lag at shutdown due to an equivalent number of fences needing to be destroyed.
Instead of relying on presentDataCount to determine whether a command buffer should be cleaned up, use a flag, which is set under the appropriate circumstances.
(cherry picked from commit 42463569d5)
For a depth buffer in D3D12 that is also going to be used in a texture sampler, the creation (on an Intel HD 5500 igpu) fails. e.g. SDL_GPUTextureCreateInfo type = TEXTURE_2D, format = D32_FLOAT, usage = DEPTH_STENCIL_TARGET | SAMPLER
The error messages are:
D32_FLOAT
D3D12 ERROR: ID3D12Device::CreateShaderResourceView: The Format (0x29, R32_FLOAT) is invalid when creating a View; the
Resource was already created with a fully qualified Format, which is not castable (0x28, D32_FLOAT).
D24_UNORM
D3D12 ERROR: ID3D12Device::CreateShaderResourceView: For the resource format D24_UNORM_S8_UINT, when making a D3D view, the format name for the view can't be R24_UNORM_X8_TYPELESS.
I found this is because the texture format needs to be created as _TYPELESS, then the views (depth stencil view, shader resource view) should then be created as their respective types - e.g. texture = R32_TYPELESS, dsv = D32_FLOAT, srv = R32_FLOAT
Tested and working on:
NVidia RTX 3050 (D3D12 feature set 12_2)
Intel HD 5500 (D3D12 feature set 11_1)
(cherry picked from commit c81b62293a)
The default should be the entrypoint generated by SDL_shadercross. That way it doesn't need to be hand-specified in the common workflow.
(cherry picked from commit dcb97a5f49)
this was causing problems when i was specifying non-zero buffer slots in
`SDL_BindGPUVertexBuffers`, `SDL_GPUVertexAttribute` and
`SDL_GPUVertexBufferDescription`.
`firstSlot + 1` is simply copied over from metal and d3d12 backends.
(cherry picked from commit 512485869c)
The GPU API doesn't currently support transparent windows (transparent swapchain effects doesn't seem possible on D3D12) so we should explicitly fail so users don't expect transparency and then not get it.
Fixes https://github.com/libsdl-org/SDL/issues/12410
(cherry picked from commit 2fbb583290)
This fixes Nintendo Switch where the identity bit isn't available. We only needed this on Android to automatically handle device user orientation.
(cherry picked from commit 2c7c3d4d7b)
Since there is no option to change it this seems like a much better default value. The stretching behaviour is really off-putting.
The new behavior shows a small black border when resizing the windows. This makes it more in line with the other backends and it's what you would expect to happen as a user.
I changed it so that the OS does the orientation change itself with a potential performance penalty.
This makes it automatically do the right thing, just like on iOS which would make the orientation change behaviour more consistent across different platforms.
But without adding an option to the user, this would disallow the user solving the problem in his in the app/shaders and saving some performance.
It's up to you to decide what to do. But I changed this in my local copy of the source for my use case.
But this also
