Commit Graph

208 Commits

Author SHA1 Message Date
ReinUsesLisp
82c2601555 video_core: Reimplement the buffer cache
Reimplement the buffer cache using cached bindings and page level
granularity for modification tracking. This also drops the usage of
shared pointers and virtual functions from the cache.

- Bindings are cached, allowing to skip work when the game changes few
  bits between draws.
- OpenGL Assembly shaders no longer copy when a region has been modified
  from the GPU to emulate constant buffers, instead GL_EXT_memory_object
  is used to alias sub-buffers within the same allocation.
- OpenGL Assembly shaders stream constant buffer data using
  glProgramBufferParametersIuivNV, from NV_parameter_buffer_object. In
  theory this should save one hash table resolve inside the driver
  compared to glBufferSubData.
- A new OpenGL stream buffer is implemented based on fences for drivers
  that are not Nvidia's proprietary, due to their low performance on
  partial glBufferSubData calls synchronized with 3D rendering (that
  some games use a lot).
- Most optimizations are shared between APIs now, allowing Vulkan to
  cache more bindings than before, skipping unnecesarry work.

This commit adds the necessary infrastructure to use Vulkan object from
OpenGL. Overall, it improves performance and fixes some bugs present on
the old cache. There are still some edge cases hit by some games that
harm performance on some vendors, this are planned to be fixed in later
commits.
2021-02-13 02:17:22 -03:00
ReinUsesLisp
75ccd9959c gpu: Report renderer errors with exceptions
Instead of using a two step initialization to report errors, initialize
the GPU renderer and rasterizer on the constructor and report errors
through std::runtime_error.
2021-02-13 02:16:19 -03:00
ReinUsesLisp
80a673a27f maxwell_3d: Silence array bounds warnings 2021-01-24 04:31:41 -03:00
ReinUsesLisp
9764c13d6d video_core: Rewrite the texture cache
The current texture cache has several points that hurt maintainability
and performance. It's easy to break unrelated parts of the cache
when doing minor changes. The cache can easily forget valuable
information about the cached textures by CPU writes or simply by its
normal usage.The current texture cache has several points that hurt
maintainability and performance. It's easy to break unrelated parts
of the cache when doing minor changes. The cache can easily forget
valuable information about the cached textures by CPU writes or simply
by its normal usage.

This commit aims to address those issues.
2020-12-30 03:38:50 -03:00
Lioncash
4c5f5c9bf3 video_core: Remove unnecessary enum class casting in logging messages
fmt now automatically prints the numeric value of an enum class member
by default, so we don't need to use casts any more.

Reduces the line noise a bit.
2020-12-07 00:41:50 -05:00
ReinUsesLisp
622830f4e1 maxwell_3d: Use insert instead of loop push_back
This reduces the overhead of bounds checking on each element.
It won't reduce the cost of allocation because usually this vector's
capacity is usually large enough to hold whatever we push to it.
2020-11-11 19:52:19 -03:00
ReinUsesLisp
9ea8cffe35 maxwell_3d: Move code to separate functions
Deduplicate some code and put it in separate functions so it's easier to
understand and profile.
2020-11-11 19:52:19 -03:00
Lioncash
ff45c39578 General: Make use of std::nullopt where applicable
Allows some implementations to avoid completely zeroing out the internal
buffer of the optional, and instead only set the validity byte within
the structure.

This also makes it consistent how we return empty optionals.
2020-09-22 17:32:33 -04:00
ReinUsesLisp
da53bcee60 video_core: Initialize renderer with a GPU
Add an extra step in GPU initialization to be able to initialize render
backends with a valid GPU instance.
2020-08-22 01:51:45 -03:00
David Marcec
6ce5f3120b Macro HLE support 2020-06-24 12:09:01 +10:00
bunnei
c2ea1e1bcb
Merge pull request #4049 from ReinUsesLisp/separate-samplers
shader/texture: Join separate image and sampler pairs offline
2020-06-13 13:48:27 -04:00
ReinUsesLisp
5b2b6d594c shader/texture: Join separate image and sampler pairs offline
Games using D3D idioms can join images and samplers when a shader
executes, instead of baking them into a combined sampler image. This is
also possible on Vulkan.

One approach to this solution would be to use separate samplers on
Vulkan and leave this unimplemented on OpenGL, but we can't do this
because there's no consistent way of determining which constant buffer
holds a sampler and which one an image. We could in theory find the
first bit and if it's in the TIC area, it's an image; but this falls
apart when an image or sampler handle use an index of zero.

The used approach is to track for a LOP.OR operation (this is done at an
IR level, not at an ISA level), track again the constant buffers used as
source and store this pair. Then, outside of shader execution, join
the sample and image pair with a bitwise or operation.

This approach won't work on games that truly use separate samplers in a
meaningful way. For example, pooling textures in a 2D array and
determining at runtime what sampler to use.

This invalidates OpenGL's disk shader cache :)

- Used mostly by D3D ports to Switch
2020-06-05 00:24:51 -03:00
bunnei
34d4abc4f9
Merge pull request #4009 from ogniK5377/macro-jit-prod
video_core: Implement Macro JIT
2020-06-04 11:40:52 -04:00
David Marcec
eca3d16e54 Default init labels and use initializer list for macro engine 2020-06-04 22:23:07 +10:00
David Marcec
411f5527d4 Mark parameters as const 2020-06-03 16:33:38 +10:00
David Marcec
3a20e74f40 Pass by reference instead of copying parameters 2020-06-02 16:37:06 +10:00
bunnei
bb6d93630f
Merge pull request #3998 from ReinUsesLisp/init-3d
maxwell_3d: Initialize more registers to their expected value
2020-06-01 16:11:56 -04:00
David Marcec
b032ebdfee Implement macro JIT 2020-05-30 11:40:04 +10:00
ReinUsesLisp
9b06e823ee maxwell_3d: Reduce severity of logs that can be spammed
These logs were killing performance on some games when they were
spammed. Reduce them to Debug severity.
2020-05-28 18:23:25 -03:00
ReinUsesLisp
f3f056c3b6 maxwell_3d: Initialize line widths
Initialize line widths to avoid setting a line width of zero.
2020-05-27 16:53:43 -03:00
ReinUsesLisp
31eb658fea maxwell_3d: Initialize polygon modes
NVN expects this to be initialized as Fill, otherwise games that never
bind a rasterizer state will log an invalid polygon mode.
2020-05-27 16:52:52 -03:00
bunnei
50c27d5ae1
Merge pull request #3885 from ReinUsesLisp/viewport-swizzles
video_core: Implement viewport swizzles with NV_viewport_swizzle
2020-05-08 15:16:53 -04:00
bunnei
41682e0888
Merge pull request #3815 from FernandoS27/command-list-2
GPU: More optimizations to GPU Command List Processing and DMA Copy Optimizations
2020-05-05 17:12:42 -04:00
ReinUsesLisp
9b8e962368 maxwell_3d: Add viewport swizzles 2020-05-04 17:50:59 -03:00
ReinUsesLisp
fe931ac976 {maxwell_3d,buffer_cache}: Implement memory barriers using 3D registers
Drop MemoryBarrier from the buffer cache and use Maxwell3D's register
WaitForIdle.

To implement this on OpenGL we just call glMemoryBarrier with the
necessary bits.

Vulkan lacks this synchronization primitive, so we set an event and
immediately wait for it. This is not a pretty solution, but it's what
Vulkan can do without submitting the current command buffer to the queue
(which ends up being more expensive on the CPU).
2020-04-28 02:18:12 -03:00
Fernando Sahmkow
90e5694230 VideoCore/Engines: Refactor Engines CallMethod. 2020-04-27 21:47:58 -04:00
Fernando Sahmkow
18a88d19dc Maxwell3D: Process Macros on MultiMethod. 2020-04-23 08:52:56 -04:00
Fernando Sahmkow
3fedcc2f6e DMAPusher: Propagate multimethod writes into the engines. 2020-04-23 08:52:55 -04:00
Fernando Sahmkow
b7bc3c2549 FenceManager: Manage syncpoints and rename fences to semaphores. 2020-04-22 11:36:16 -04:00
Fernando Sahmkow
4adfc9bb08 Rasterizer: Document SignalFence & ReleaseFences and setup skeletons on Vulkan. 2020-04-22 11:36:14 -04:00
Fernando Sahmkow
a081a7c855 GPU: Fix rebase errors. 2020-04-22 11:36:13 -04:00
Fernando Sahmkow
487379c593 OpenGL: Implement Fencing backend. 2020-04-22 11:36:10 -04:00
Fernando Sahmkow
339d0d9d6c GPU: Delay Fences. 2020-04-22 11:36:08 -04:00
Fernando Sahmkow
da8f17715d GPU: Refactor synchronization on Async GPU 2020-04-22 11:36:06 -04:00
ReinUsesLisp
238c6016f9 maxwell_3d: Initialize format attributes constant as one
nouveau expects this to be true but it doesn't set it.
2020-04-16 21:15:07 -03:00
namkazy
fc37672f26 apply replay logic to all writes. remove replay from MacroInterpreter::Send (@fincs) 2020-03-22 22:25:44 +07:00
namkazy
d4e93cf38c maxwell_3d: init shadow_state 2020-03-22 13:35:11 +07:00
namkazy
22f4268c2f maxwell_3d: this seem more correct. 2020-03-22 12:02:54 +07:00
namkazy
7051dc1902 maxwell_3d: update comments for shadow ram usage 2020-03-22 11:35:26 +07:00
Nguyen Dac Nam
63c2635e6f maxwell_3d: track shadow ram ctrl and hw reg value 2020-03-22 10:53:41 +07:00
ReinUsesLisp
978172530e const_buffer_engine_interface: Store component types
This is required for Vulkan. Sampling integer textures with float
handles is illegal.
2020-03-09 18:40:53 -03:00
ReinUsesLisp
9e74e6988b maxwell_3d: Flatten cull and front face registers 2020-02-28 17:56:41 -03:00
ReinUsesLisp
eed789d0d1 video_core: Reintroduce dirty flags infrastructure 2020-02-28 17:56:41 -03:00
ReinUsesLisp
96ac3d518a gl_rasterizer: Remove dirty flags 2020-02-28 16:39:27 -03:00
bunnei
b2bc7682b4
Merge pull request #3414 from ReinUsesLisp/maxwell-3d-draw
maxwell_3d: Unify draw methods
2020-02-19 16:13:50 -05:00
ReinUsesLisp
91aa58e410 maxwell_3d: Unify draw methods
Pass instanced state of a draw invocation as an argument instead of
having two separate virtual methods.
2020-02-14 18:09:40 -03:00
ReinUsesLisp
73d2d3342d gl_query_cache: Optimize query cache
Use a custom cache instead of relying on a ranged cache.
2020-02-14 17:38:27 -03:00
ReinUsesLisp
aae8c180cb gl_query_cache: Implement host queries using a deferred cache
Instead of waiting immediately for executed commands, defer the query
until the guest CPU reads it. This way we get closer to what the guest
program is doing.

To archive this we have to build a dependency queue, because host APIs
(like OpenGL and Vulkan) use ranged queries instead of counters like
NVN.

Waiting for queries implicitly uses fences and this requires a command
being queued, otherwise the driver will lock waiting until a timeout. To
fix this when there are no commands queued, we explicitly call glFlush.
2020-02-14 17:33:13 -03:00
ReinUsesLisp
2b58652f08 maxwell_3d: Slow implementation of passed samples (query 21)
Implements GL_SAMPLES_PASSED by waiting immediately for queries.
2020-02-14 17:27:17 -03:00
Fernando Sahmkow
d6ed31b9fa GPU: Address Feedback. 2020-02-13 18:16:07 -04:00