* vector_alu: Improve handling of mbcnt append/consume patterns
The existing implementation was written to handle a single pattern of mbcnt before the DS_APPEND instruction
v_mbcnt_hi_u32_b32 vX, exec_hi, 0
v_mbcnt_lo_u32_b32 vX, exec_lo, vX
ds_append vY offset:4 gds
v_add_i32 vX, vcc, vY, vX
In this case however the DS_APPEND is before the mbcnt pattern
ds_append vX gds
v_mbcnt_hi_u32_b32 vY, exec_hi, vX
v_mbcnt_lo_u32_b32 vZ, exec_lo, vY
The mbcnt instructions are always in pairs of hi/lo and in general are quite flexible. But they assume the subgroup size is 64 so they are not recompiled literally. Together with DS_APPEND they are used to derive a unique per thread index in a buffer (different from using thread_id as order could be random). DS_APPEND instruction works on per subgroup level, by adding number of active threads of subgroup to the GDS counter, essentially giving a multiple-of-64 base index to all threads. Then each thread executes the mbcnt pair which returns the number of active threads with id less than the itself and adds it with the base.
The recompiler translates DS_APPEND into an atomic increment of a storage buffer counter, which already gives the desired unique index, so this pattern is a no-op. On main it was set to zero as per the first pattern to avoid altering the DS_APPEND result. The new handling passes through the initial value of the pattern instead, which has the same effect but works on either case.
* vk_rasterizer: Always sync DMA buffers
* Added V_CMP_EQ_U64 shader opcode support and added 64-bit relational operators (<,>,<=,>=)
* Fixed clang-format crying because I typed xargs clang-format instead of xargs clang-format-19
* Replaced V_CMP_EQ_U64 code to match V_CMP_U32 to test
* Updated V_CMP_U64 for future addons
* shader_recompiler: Replace texel buffers with in-shader buffer format interpretation
* shader_recompiler: Move 10/11-bit float conversion to functions and address some comments.
* vulkan: Remove VK_KHR_maintenance5 as it is no longer needed for buffer views.
* shader_recompiler: Add helpers for composites and bitfields in pack/unpack.
* shader_recompiler: Use initializer_list for bitfield insert helper.
* shader_recompiler: Account for instruction array flag in image type.
* shader_recompiler: Check da flag for all mimg instructions.
* shader_recompiler: Convert cube images into 2D arrays.
* shader_recompiler: Move image resource functions into sharp type.
* shader_recompiler: Use native AMD cube instructions when possible.
* specialization: Fix buffer storage mistake.
* shader_recompiler: Tessellation WIP
* fix compiler errors after merge
DONT MERGE set log file to /dev/null
DONT MERGE linux pthread bb fix
save work
DONT MERGE dump ir
save more work
fix mistake with ES shader
skip list
add input patch control points dynamic state
random stuff
* WIP Tessellation partial implementation. Squash commits
* test: make local/tcs use attr arrays
* attr arrays in TCS/TES
* dont define empty attr arrays
* switch to special opcodes for tess tcs/tes reads and tcs writes
* impl tcs/tes read attr insts
* rebase fix
* save some work
* save work probably broken and slow
* put Vertex LogicalStage after TCS and TES to fix bindings
* more refactors
* refactor pattern matching and optimize modulos (disabled)
* enable modulo opt
* copyright
* rebase fixes
* remove some prints
* remove some stuff
* Add TCS/TES support for shader patching and use LogicalStage
* refactor and handle wider DS instructions
* get rid of GetAttributes for special tess constants reads. Immediately replace some upon seeing readconstbuffer. Gets rid of some extra passes over IR
* stop relying on GNMX HsConstants struct. Change runtime_info.hs_info and some regs
* delete some more stuff
* update comments for current implementation
* some cleanup
* uint error
* more cleanup
* remove patch control points dynamic state (because runtime_info already depends on it)
* fix potential problem with determining passthrough
---------
Co-authored-by: IndecisiveTurtle <47210458+raphaelthegreat@users.noreply.github.com>
* 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
* shader_recompiler: Add more format swap modes
* texture_cache: Handle stencil texture reads
* emulator: Support loading font library
* readme: Add thanks section
* shader_recompiler: Constant buffers as integers
* shader_recompiler: Typed buffers as integers
* shader_recompiler: Separate thread bit scalars
* We can assume guest shader never mixes them with normal sgprs. This helps avoid errors where ssa could view an sgpr write dominating a thread bit read, due to how control flow is structurized, even though its not possible in actual control flow
* shader_recompiler: Implement data append/consume operations
* clang format
* buffer_cache: Simplify invalidation scheme
* video_core: Remove some invalidation remnants
* adjust
* shader_recompiler: Implement V_MOVRELS_B32, V_MOVRELD_B32,
V_MOVRELSD_B32
Generates a ton of OpSelects to hardcode reading or writing from each
possible vgpr depending on the value of m0
Future work is to do range analysis to put an upper bound on m0 and
check fewer registers.
* fix runtime info after rebase
* shader_recompiler: Use null image when shader is compiled with unbound sharp
* video_core: Refactor and render target swizzles
* liverpool_to_vk: Add missing swap format from RDR
* video_core: Refactor shader recompiler interface
* Makes it much easier to pass runtime information to the recompiler and have it treated as part of the shader key. Also pulls out most runtime state from Info struct
* shader_recompiler: Avoid some asserts
* video_core: Compile shader permutations
* spirv: Only specific storage image format for atomics
* ir: Avoid cube coord patching for storage image
* spirv: Fix default attributes
* data_share: Add more instructions
* video_core: Query storage flag with runtime state
* kernel: Use std::list for semaphore
* video_core: Use texture buffers for untyped format load/store
* buffer_cache: Limit view usage
* vk_pipeline_cache: Fix invalid iterator
* image_view: Reduce log spam when alpha=1 in storage swizzle
* video_core: More features and proper spirv feature detection
* video_core: Attempt no2 for specialization
* spirv: Remove conflict
* vk_shader_cache: Small cleanup
