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.
Bug: https://bugs.dolphin-emu.org/issues/13404
On macOS 13.6 / Intel HD 5000, Dolphin crashes with this message:
> -[MTLIGAccelDevice setShouldMaximizeConcurrentCompilation:]: unrecognized selector
This should be available on all macOS 13.3+ systems – but when using OCLP drivers,
some devices use an older version of Metal.framework, which doesn't expose the selector.
This concerns Intel Ivy Bridge, Haswell and Nvidia Kepler when using OCLP on macOS 13.3
or newer.
(See
34676702f4/docs/PATCHEXPLAIN.md (L354C1-L354C83))
As the behavior is an optional optimization anyway, perform a dynamic
detection to avoid crashing if the feature is not available.
AMD Metal drivers have a goofy bug where the bbox buffer stops being coherent with the cpu if you copy to it from a private (gpu) buffer and don't do anything else with it in that command buffer.
While the NV extension is totally fine, the KHR extension should be able to support more hardware.
For NVIDIA, the hardware either supports both or neither, it just needs a driver from the last two years.
For AMD, the drivers from late 2022-12 seems to bring support for the KHR extension.
For Intel, the KHR is also supported for some years.
