Commit Graph

1942 Commits

Author SHA1 Message Date
Billy Laws
b86dd99e1a Align all SSBOs to 0x40 bytes
Required by Adreno GPUs
2022-12-03 22:50:56 +00:00
Billy Laws
bfae292fb0 Make executor slot count setting exponential 2022-12-03 22:50:56 +00:00
Billy Laws
e0ae94be9d Enable robustness1 Vulkan feature 2022-12-03 22:50:56 +00:00
Billy Laws
e8ef2d80af CMake build file updates 2022-12-03 22:50:56 +00:00
Billy Laws
bf03f945ee Implement the Kepler compute engine
This can reuse a fair bit of the now-commonised Maxwell 3D code and mostly consists of compute-specific pipeline code which was deemed not suitable for being commonised (e.g. descriptor update code is somewhat duplicated). Of note is how compute lacks any active state at all de to its use of QMDs which bundle up all state into a single object in memory.
2022-12-03 22:50:56 +00:00
Billy Laws
4bc81f007f Add some convinience helpers to compute engine regs 2022-12-03 22:50:56 +00:00
Billy Laws
4267a6af36 Add support for parsing and compiling compute shaders to the shader manager 2022-12-03 22:50:56 +00:00
Billy Laws
86dab65af4 Commonise maxwell3d state updater 2022-12-03 22:50:56 +00:00
Billy Laws
a0b81d54d6 Use pitch layout for linear RTs
More likely to match in the texture cache when being sampled.
2022-12-03 22:50:56 +00:00
Billy Laws
ac85df7b7a Start transition cache lookup with most recent one 2022-12-03 22:50:56 +00:00
Billy Laws
62c86b7690 Move maxwell3d to common constant buffer code 2022-12-03 22:50:56 +00:00
Billy Laws
8f0a6e78c5 Add Vulkan stride dynamic state and robustness support
Fixes the waterfall in SMO by specifying vertex buffer bounds.
2022-12-03 22:50:56 +00:00
Billy Laws
23a7f70a8e Commonise maxwell3d guest shader caching code 2022-12-03 22:50:56 +00:00
Billy Laws
6f6a312692 Commonise maxwell3d pipeline binding handling code
A lot of pipeline code is difficult to commonise due to the inherent difference between compute and graphics pipelines, however the binding layout is shared so we can at least commonise that
2022-12-03 22:50:56 +00:00
Billy Laws
be8cbabd97 Commonise maxwell3d texture code
This will be shared with the compute engine implementation.
2022-12-03 22:50:56 +00:00
Billy Laws
61e95c4b2c Commonise maxwell3d sampler code
This will be shared with the compute engine implementation, the only thing of note with this is that the binding register is now passed as a param since it is part of the compute QMD which can't be dirty tracked.
2022-12-03 22:50:56 +00:00
Billy Laws
7f93ec3df6 Commonise maxwell3d interconnect common code for use by other engines
The compute engine will require most of this for basic functionality.
2022-12-03 22:50:56 +00:00
Billy Laws
281838fde1 Apply GPU readback hack to both buffers and textures
And rename as appropriate.
2022-12-03 22:50:56 +00:00
Billy Laws
f358c4517e Update edge credits 2022-12-03 22:50:56 +00:00
Billy Laws
eb00dc62f8 Implement support for 36 bit games by using split code/heap mappings
Although rtld and IPC prevent TLS/IO and code from being above the 36-bit AS limit, nothing depends the heap being below it. We can take advantage of this by stealing as much AS as possible for code in the lower 36-bits.
2022-12-02 22:10:03 +00:00
Dima
e8e1b910c3 Add possibility to disable audio output 2022-12-02 00:33:28 +01:00
lynxnb
70109f8fbd Work around invalid values in CNTFRQ_EL0 register
Exynos SoCs have a bug where the `CNTFRQ_EL0` register is either set to 0 or contain incoherent values. With this patch, the frequency value is loaded into a static variable and used instead of reading the register. The value will be initialised to the correct value for affected SoCs, while unaffected ones will use the value from the register.
2022-12-02 00:23:28 +01:00
lynxnb
54d0246ca6 Tweak GpuDriverActivity FAB padding 2022-11-28 00:06:07 +01:00
lynxnb
2e8d7b559c Use the original view padding/margin when applying window insets
Adding to the current view padding/margin values results in applying the insets over and over again as insets listeners can be called multiple times.
2022-11-28 00:04:39 +01:00
Billy Laws
b2384e83f5 Add prepo:a service 2022-11-25 16:26:00 +00:00
Billy Laws
736216a6f4 Stub OpenPatchDataStorageByCurrentProcess 2022-11-25 16:26:00 +00:00
Billy Laws
44033d7f8d Adjust CalendarTime year to be relative to 0AD 2022-11-25 16:26:00 +00:00
Billy Laws
2ce2604421 Implement VFS file deletion 2022-11-25 16:26:00 +00:00
Billy Laws
6c968e0357 Fix GetEntryType IPC return type 2022-11-25 16:26:00 +00:00
lynxnb
ec220c8ea9 Use an extended FAB in GpuDriverActivity 2022-11-23 19:49:42 +05:30
lynxnb
163f4f2014 Fix window insets handling when in landscape mode
To avoid code duplication, insets handling has been moved to a separate interface.
2022-11-23 19:49:42 +05:30
lynxnb
ab6c5f4c50 Improve robustness of KeyReader.import
* Close the input and output file streams before moving the output file to the final destination
* Clean up the destination path before moving the new file
* Introduce a `ImportResult` return value to differentiate between the possible causes of import errors
* Display more meaningful error messages in the UI
2022-11-23 19:49:42 +05:30
lynxnb
38129d9dc3 Mark some strings as non-translatable 2022-11-23 19:49:42 +05:30
lynxnb
ee8c055641 Make GpuDriverInstallResult PascalCase 2022-11-23 19:49:42 +05:30
Billy Laws
7f1667de82 Avoid using trapping for frequently trapped shaders
Fall back to hashing for every shader access as that ends up being faster than applying traps for every execution.
2022-11-19 12:49:05 +00:00
Billy Laws
06095918a9 Introduce per-channel sequence number for invalidation tracking
For cases like shaders, which may be uploaded through I2M (which no longer causes an execution) we need a way to cause an invalidation on all writes
2022-11-19 12:49:05 +00:00
Billy Laws
97e3f7fd34 Increase max swapchain image count 2022-11-19 12:49:05 +00:00
Billy Laws
c49119f5ef Fixup depth bounds register arguments 2022-11-19 12:49:05 +00:00
Billy Laws
db3c5c33c4 Clamp depth bounds into 0-1 range 2022-11-19 12:49:05 +00:00
Billy Laws
e1bbd521d9 Fix potential circular queue submission race
If a producer thread was waiting for the queue to have free space and the consumer thread hadn't yet acquired the production mutex a deadlock could occur
2022-11-19 12:49:05 +00:00
Billy Laws
13baf2312f Add a workaround for sampling BGRA textures with a swizzle 2022-11-19 12:49:05 +00:00
Billy Laws
13a96c5aba Implement a helper shader for partial clears
These are not natively supported by Vulkan, so use a helper shader and colorWriteMask for the same behaviour.
2022-11-19 12:49:05 +00:00
Billy Laws
ac0e225114 Use vkCmdBlit for texture copies when formats dont match 2022-11-19 12:49:05 +00:00
Billy Laws
c8fc8f84ec Fallback to RGBA888 for unsupported swapchain formats as opposed to swizzle 2022-11-19 12:49:05 +00:00
Billy Laws
e0bc0d3a97 Avoid megabuffering buffers larger than the chunk size 2022-11-19 12:49:05 +00:00
Billy Laws
b6f49884b3 Use lower_bound to speedup texture hostMapping lookup 2022-11-19 12:49:05 +00:00
Billy Laws
e7fda28ac6 Skip over textures in cache which have been replaced with a layer/mip match 2022-11-19 12:49:05 +00:00
Billy Laws
88cc696c7f Only use 2D array depth targets when depth > 1 2022-11-19 12:49:05 +00:00
Billy Laws
7fed971b2d Take firstIndex into account when calculating index (quad) buffer size
Without this we would miss any elements beyond indexCount in the index buffer and they would be filled with random garbage causing vertex bombs
2022-11-19 12:49:05 +00:00
Billy Laws
1f9de17e98 Begin command buffers asynchronously in command executor
vkBeginCommandBuffer can take quite some time on adreno, move it to the cycle waiter thread where it won't block GPFIFO.
2022-11-19 12:49:05 +00:00