Commit Graph

24 Commits

Author SHA1 Message Date
TheTurtle
93767ae31b shader_recompiler: Rework sharp tracking for robustness (#3327)
* shader_recompiler: Remove remnants of old discard

Also constant propagate conditional discard if condition is constant

* resource_tracking_pass: Rework sharp tracking for robustness

* resource_tracking_pass: Add source dominance analysis

When reachability is not enough to prune source list, check if a source dominates all other sources

* resource_tracking_pass: Fix immediate check

How did this work before

* resource_tracking_pass: Remove unused template type

* readlane_elimination_pass: Don't add phi when all args are the same

New sharp tracking exposed some bad sources coming on sampler sharps with aniso disable pattern that also were part of readlane pattern, fix tracking by removing the unnecessary phis inbetween

* resource_tracking_pass: Allow phi in disable aniso pattern

* resource_tracking_pass: Handle not valid buffer sharp and more phi in aniso pattern
2025-07-28 13:32:16 -07:00
TheTurtle
19c3d05ac1 shader_recompiler: Use VM bit for conditional discard (#3306) 2025-07-23 20:58:09 +03:00
TheTurtle
7502739425 control_flow_graph: Treat empty conditional branch as noop (#3296) 2025-07-23 06:09:14 +03:00
TheTurtle
7d4b875ee3 Random fixes (#3216)
* buffer_cache: Handle inline data to flexible memory

* control_flow: Fix single instruction scopes edge case

Fixes the following pattern

v_cmpx_gt_u32 cond
buffer_store_dword value
.LABEL:

Before
buffer[index] = value;

After
if (cond)
{
    buffer[index] = value;
}

* vector_memory: Handle soffset when offen is false

When offen is not used we can substitute the offset argument with soffset and have it handled correctly

* scalar_alu: Handle sharp moves with S_MOV_B64

This fixes unable to track sharp errors when this pattern is used in a shader

* emulator: Add log

* video_core: Bump binary info search range and buffer num
2025-07-09 17:00:06 +03:00
Fire Cube
0c86c54d48 Implement SET_PC_B64 instruction (#2823)
* basic impl

* minor improvements

* clang

* more clang

* improvements requested by squidbus
2025-04-21 14:25:15 -07:00
TheTurtle
1f9ac53c28 shader_recompiler: Improve divergence handling and readlane elimintation (#2667)
Some checks are pending
Build and Release / reuse (push) Waiting to run
Build and Release / clang-format (push) Waiting to run
Build and Release / get-info (push) Waiting to run
Build and Release / windows-sdl (push) Blocked by required conditions
Build and Release / windows-qt (push) Blocked by required conditions
Build and Release / macos-sdl (push) Blocked by required conditions
Build and Release / macos-qt (push) Blocked by required conditions
Build and Release / linux-sdl (push) Blocked by required conditions
Build and Release / linux-qt (push) Blocked by required conditions
Build and Release / linux-sdl-gcc (push) Blocked by required conditions
Build and Release / linux-qt-gcc (push) Blocked by required conditions
Build and Release / pre-release (push) Blocked by required conditions
* control_flow_graph: Improve divergence handling

* recompiler: Simplify optimization passes

Removes a redudant constant propagation and cleans up the passes a little

* ir_passes: Add new readlane elimination pass

The algorithm has grown complex enough where it deserves its own pass. The old implementation could only handle a single phi level properly,
however this one should be able to eliminate vast majority of lane cases remaining. It first performs a traversal of the phi tree to ensure
that all phi sources can be rewritten into an expected value and then performs elimintation by recursively duplicating the phi nodes at each step,
in order to preserve control flow.

* clang format

* control_flow_graph: Remove debug code
2025-03-23 00:35:42 +02:00
squidbus
460c266e04 fix: Restore previous version of divergence PR. 2025-02-02 15:37:17 -08:00
squidbus
831903799b shader_recompiler: Insert end of divergence scope at last relevant instruction. (#2325) 2025-02-02 14:51:45 -08:00
squidbus
65f9bbbfed shader_recompiler: Ignore exec mask for scalar instructions. (#2097) 2025-01-08 19:14:06 +02:00
Vladislav Mikhalin
876445faf1 recompiler: emit a label right after s_branch to prevent dead code interferrence (#1785) 2024-12-14 22:46:55 +02:00
TheTurtle
e9ede8d627 Revert "DmaData and Recompiler fixes (#1775)" (#1784)
This reverts commit cafd40f2c2.
2024-12-14 16:17:14 +02:00
Vladislav Mikhalin
cafd40f2c2 DmaData and Recompiler fixes (#1775)
* liverpool: fix dmadata packet handling

* recompiler: emit a label right after s_branch to prevent dead code interferrence

* specialize barriers
2024-12-14 14:33:06 +02:00
TheTurtle
22a2741ea0 shader_recompilers: Improvements to SSA phi generation and lane instruction elimination (#1667)
* shader_recompiler: Add use tracking for Insts

* ssa_rewrite: Recursively remove phis

* ssa_rewrite: Correct recursive trivial phi elimination

* ir: Improve read lane folding pass

* control_flow: Avoid adding unnecessary divergant blocks

* clang format

* externals: Update ext-boost

---------

Co-authored-by: Frodo Baggins <baggins31084@proton.me>
2024-12-05 23:14:16 +02:00
Herman Semenoff
96ea686eb6 Fixed return strict const iterator, replace to range-based loop C++17 and code refactor (#548)
Signed-off-by: Herman Semenov <GermanAizek@yandex.ru>
Co-authored-by: georgemoralis <giorgosmrls@gmail.com>
2024-10-18 11:06:11 +03:00
TheTurtle
ee38eec7fe shader_recompiler: Additional scope handling and user data as push constants (#1013)
* shader_recompiler: Use push constants for user data regs

* shader: Add some GR2 instructions

* shader: Add some instructions

* shader: Add instructions for knack

* touchups

* spirv: Better names

* buffer_cache: Ignore non gpu modified images

* clang format

* Add log

* more fixes
2024-09-23 08:55:43 +02:00
psucien
b687ae5e34 GnmDriver: Clear context support (#567)
* gnmdriver: added support for gpu context reset

* shader_recompiler: minor validation fixes

* shader_recompiler: added `V_CMPX_GT_I32`

* shader_recompiler: fix for crash on inline sampler access

* compilation warnings and dead code elimination

* amdgpu: fix for registers addressing

* libraries: videoout: reduce logging pressure

* shader_recompiler: fix for devergence scope detection
2024-08-25 23:01:05 +03:00
TheTurtle
3f9c86ad33 vk_pipeline_cache: Avoid recompiling new shaders on each new PL (#480)
* cfg: Add one more divergence case

* Seen in RDR shaders

* renderer_vulkan: Reduce number of compiled shaders

* vk_pipeline_cache: Remove some unnecessary checks
2024-08-21 02:00:24 +03:00
TheTurtle
1d1c88ad31 control_flow_graph: Initial divergence handling (#434)
* control_flow_graph: Initial divergence handling

* cfg: Handle additional case

* spirv: Handle tgid enable bits

* clang format

* spirv: Use proper format

* translator: Add more instructions
2024-08-16 20:05:37 +03:00
TheTurtle
381ba8c7a5 video_core: Implement guest buffer manager (#373)
* video_core: Introduce buffer cache

* video_core: Use multi level page table for caches

* renderer_vulkan: Remove unused stream buffer

* fix build

* oops forgot optimize off
2024-08-08 15:02:10 +03:00
psucien
f041276b04 recompiler: added support for discard on export with masked EXEC 2024-07-13 14:57:01 +02:00
TheTurtle
6ceab6dfac shader_recompiler: Implement most integer image atomics, workgroup barriers and shared memory load/store (#231)
* shader_recompiler: Add LDEXP

* shader_recompiler: Add most image integer atomic ops

* shader_recompiler: Implement shared memory load/store

* shader_recompiler: More image atomics

* externals: Update sirit

* clang format

* cmake: Add missing files

* shader_recompiler: Fix some atomic bugs

* shader_recompiler: Vs outputs

* shader_recompiler: Shared mem has side-effects, fix format component order

* shader_recompiler: Inline constant buffer impl

* video_core: Fix regressions

* Work

* Fixup a few things
2024-07-05 00:15:44 +03:00
TheTurtle
998d046210 video_core: Add depth buffer support and fix some bugs (#172)
* memory: Avoid crash when alignment is zero

* Also remove unused file

* shader_recompiler: Add more instructions

* Also fix some minor issues with a few existing instructions

* control_flow: Don't emit discard for null exports

* renderer_vulkan: Add depth buffer support

* liverpool: Fix wrong color buffer number type and viewport zscale

* Also add some more formats
2024-06-07 16:26:43 +03:00
raphaelthegreat
58de7ff55a video_core: Implement basic compute shaders and more instructions 2024-05-30 01:39:24 +03:00
TheTurtle
8730968385 video: Import new shader recompiler + display a triangle (#142) 2024-05-22 01:35:12 +03:00