| Age | Commit message (Collapse) | Author |
|---|
|
VideoBackends:Metal: Log file on failed pipeline compile
|
|
Ranges Algorithms Modernization - Trivial
|
|
Fixes LIT (https://bugs.dolphin-emu.org/issues/13635). The text does not include normals, but has lighting enabled. With the previous default of (0, 0, 0), lighting was always black (as dot(X, (0, 0, 0)) is always 0). It seems like the normal from the map in the background (0, 0, 1) is re-used.
LIT also has the vertex color enabled while vertex color is not specified, the same as SMS's debug cubes; the default MissingColorValue GameINI value of solid white seems to work correctly in this case.
|
|
Ranges Algorithms Modernization - Find
|
|
The inequality operator is automatically generated by the compiler if `operator==` is defined.
|
|
In BTEmu.cpp, `std::mem_fn` was not necessary for the predicate to compile.
|
|
|
|
In OGLConfig.cpp, `std::views::reverse` is used rather than sorting using `std::ranges::greater` in order to parallel other instances of reverse iteration in the function.
|
|
|
|
|
|
|
|
We now use Vulkan 1.2 if available
|
|
Vulkan: Don't do MoltenVK things on Asahi Linux
|
|
VideoBackends:Metal: Fix bbox on newer iOS devices
|
|
AcquireNextImage before marking semaphores used
|
|
|
|
SIMD-scoped permute operations are Apple6, but reduction operations are Apple7
|
|
It defaults to trying to emulate simdgroup functions on iOS
|
|
|
|
|
|
Previously we'd assume all Apple GPUs were MoltenVK, including those running on Asahi Linux with open source Honeykrisp drivers.
|
|
|
|
Add support for libfmt-11
|
|
This is a minor improvement to add line numbers to the LOG_VULKAN_ERROR
define. Basically error logs for Vulkan will now look like:
```
// This
25:03:347 VideoBackends/Vulkan/VulkanLoader.cpp:247 E[Video]: (WaitForCommandBufferCompletion:278) vkWaitForFences failed: (2: VK_TIMEOUT)
// Instead of
15:45:154 VideoBackends/Vulkan/VulkanLoader.cpp:247 E[Video]: (WaitForCommandBufferCompletion) vkWaitForFences failed: (2: VK_TIMEOUT)
```
|
|
fmt::join was moved into fmt/ranges.h
Signed-off-by: Alfred Wingate <parona@protonmail.com>
|
|
|
|
Things using dolphin_find_optional_system_library need to link against the name used there or they won't work with both the system and bundled cases
|
|
|
|
VideoCommon: More specific subgroup op bugs
|
|
|
|
Now only VertexLoader remains... But that one might be tricky.
|
|
To ensure memory safety, callers of GetPointer have to perform a bounds
check. But how is this bounds check supposed to be performed?
GetPointerForRange contained one implementation of a bounds check, but
it was cumbersome, and it also isn't obvious why it's correct.
To make doing the right thing easier, this commit changes GetPointer to
return a span that tells the caller how many bytes it's allowed to
access.
|
|
On all platforms, this would result in out of bounds accesses when getting the component sizes (which uses stuff from VertexLoader_Position.h/VertexLoader_TextCoord.h/VertexLoader_Normal.h). On platforms other than x64 and ARM64, this would also be out of bounds accesses when getting function pointers for the non-JIT vertex loader (in VertexLoader_Position.cpp etc.). Usually both of these would get data from other entries in the same multi-dimensional array, but the last few entries would be truly out of bounds. This does mean that an out of bounds function pointer can be called on platforms that don't have a JIT vertex loader, but it is limited to invalid component formats with values 5/6/7 due to the size of the bitfield the formats come from, so it seems unlikely that this could be exploited in practice.
This issue affects a few games; Def Jam: Fight for New York (https://bugs.dolphin-emu.org/issues/12719) and Fifa Street are known to be affected.
I have not done any hardware testing for this PR specifically, though I *think* I previously determined that at least a value of 5 behaves the same as float (4). That's what I implemented in any case. I did previously determine that both Def Jam: Fight for New York and Fifa Street use an invalid normal format, but don't actually have lighting enabled when that normal vector is used, so it doesn't change rendering in practice.
The color component format also has two invalid values, but VertexLoader_Color.h/.cpp do check for those invalid ones and return a default value instead of doing an out of bounds access.
|
|
|
|
|
|
|
|
We now use subgroup ops for more than just a minor performance optimization
|
|
Allows non-constant simd broadcast, which is used by the ascii art shader
|
|
|
|
devices."
This reverts commit 79648e1c24eac81f54365bbcb6dbc49ad6b16b1c.
|
|
It was reported that some games (Zelda Wind Waker and Zelda Twilight Princess but others may also exhibit the issue) have graphical issues with the max pixel samplers set to 16 on some Android devices (ex: Pixel6); since this was increased for a performance heavy feature (custom shaders) just disable it for now. In the future, this could be handled more elegantly
|
|
VideoBackends / VideoCommon: update max pixel shader samplers from 8 to 16
|
|
VideoBackends / VideoCommon: add type enum to dictate how the texture is used; support texture 2d
|
|
2D texture, a texture array, or a cube map; support 2D texture type across backends
Co-authored-by: TellowKrinkle <tellowkrinkle@gmail.com>
|
|
|
|
Fixes a build failure with clang 17.
The destructor needs to be in the cpp file, since we have a forward
declared std::unique_ptr type as part of the class. So technically the
default inline destructor could invoke without seeing the full data type
definition.
|
|
this allows us to support more samplers than the native Wii/GC
|
|
Crosses off a lingering TODO.
Also amends a few nearby cases where a u32 cast was being repromoted to
size_t.
|
|
|
|
|
