This commit reverts PR #2037. Passing `NativeSettings` to emulation code through a member reference, instead of a local variable, caused unpredictable crashes when using custom GPU drivers (v615+) on some Qualcomm SoCs.
The exact cause of the issue remains unknown, my best guess is that it was caused by an incorrect optimization performed on the Kotlin bytecode in release mode, which caused an issue when reading memory that had been forked, because of running emulation in a separate process.
Runtime settings modification will be reimplemented in the future via an alternative method.
Indirect layers are used by the game to render layer on its own, the game allocates a buffer with the size from `GetIndirectLayerImageRequiredMemoryInfo` and uses `GetIndirectLayerImageMap` to draw the applet contents into the buffer.
As we don't LLE applet implementations nor do our HLE implementations draw equivalent applets, we cannot submit this to the guest. As a result, these functions are stubbed with the framebuffer being cleared to red.
Stubbing these functions allows titles such as Dark Souls to not crash while initializing indirect layers.
Accessing the settings class during the execution of the `OnChangedCallback` results in a deadlock, as accesses to values are protected by a mutex. Instead, we now keep a local copy of the relevant settings and update those with the new value.
I missed that addSubpass was only called once-per-subpass, meaning that if a new barrier req was discovered several draws into the RP it wouldn't be applied. Split out barriers into a seperate function to avoid this.
Full pipeline barriers between every RP can be extremely expensive on HW, by analysing the inputs and outputs of a draw it's possible to construct a much more optimal barrier that only syncs what is neccessary.
Sometimes view pointers may change despite the underlying Vulkan image view not actually changing, so use vk::ImageViews for tracking to keep RP breaks to a minimum.
The layer stride provided by the depth register in Maxwell3D needs to be shifted by 2, this caused the stride to be 1/4th of what it needed to be resulting in OOB access.
When calculating mip-level dimensions in terms of GOBs, they need to be divided by 2 while rounding upwards rather than downwards. This fixes corrupted textures and OOB access on lower mip levels across a substantial amount of titles, reducing arbitrary crashes as a result.
The default `compiler_check` strategy used by ccache was causing a 0% hit rate, as it was using the compiler's "modified time" as part of the hash. Since the script installs dependencies from scratch for every run, mtime was always different, causing different hashes even when source files were unchanged. To avoid this, use the NDK version as part of the hash instead.
