Lines Matching full:aco
155 - radv: ACO miscompiles a specific DIRT 5 shader
1120 - aco: Add load_sbt_amd intrinsic implementation.
1127 - aco: Implement bvh64_intersect_ray_amd intrinsic.
1551 - aco: fix additional register requirements for spilling
1552 - aco: relax validation rules for p_reduce dst RegType
1554 - aco/ra: prevent underflow register for p_create_vector operands
1556 - aco/ra: also prevent overflow register for p_create_vector operands
1557 - aco: remove condition operand from branch in invert block
1558 - radv,aco: scalarize all phis via nir_lower_phis_to_scalar()
1559 - aco: simplify Phi RegClass selection
1560 - aco/ra: only create phi-affinities for killed operands
1561 - aco/ra: refactor affinity coalescing
1562 - aco/ra: refactor register assignment for vector operands
1564 - aco/ra: handle copies of definition registers
1565 - aco/ra: handle copies of copies better
1566 - aco/util: replace DIV_ROUND_UP(n+1,m) by n/m+1
1567 - aco: reorder and cleanup #includes
1568 - aco: add missing Licenses and remove Authors from files
1569 - aco: add 'common/' and 'llvm/' prefix to #includes
1570 - aco/meson: remove unnecessary dependencies
1571 - aco: refactor SDWA opcode validation
1572 - aco: remove (wrong) GCC array-bounds warning
1574 - aco: add .clang-format file
1575 - aco: Format.
1576 - aco/meson: remove inc_gallium from include_directories
1577 - aco: fix self-intersecting register swaps
1578 - aco: fix extract_vector optimization
1579 - aco/isel: avoid unnecessary calls to nir_unsigned_upper_bound()
1580 - aco/insert_waitcnt: Remove many unnecessary wait_imm.combine()
1581 - aco/live_var_analysis: change worklist to a single integer
1582 - aco/optimizer: ensure to not erase high bits when propagating packed constants
1583 - aco: include <cstddef> in aco_util.h
3651 - android: aco: add aco_optimizer_postRA.cpp to Makefile.sources
4390 - aco/ra: use original names when renaming loop carried phi operands
4391 - aco/ra: remove live-in temporary from live_out_per_block when moving it
4394 - aco: set TRUNC_COORD=0 for nir_texop_tg4
4396 - aco: remove image parameter from get_sampler_desc()
4398 - aco: don't update register demand during RA validation
4399 - aco: allow SDWA sels smaller than the operand size
4400 - aco: add and use Program::progress
4411 - radv,aco: use nir_address_format_vec2_index_32bit_offset
4416 - aco: group loads from the same vertex binding into the same clause
4417 - radv,aco: use per-attribute vertex descriptors for robustness
4418 - Revert "radv,aco: don't use MUBUF for multi-channel loads on GFX8 with robustness2"
4419 - radv,aco: compact vertex buffer descriptors
4421 - aco/ra: initialize temp_in_scc earlier
4422 - aco/ra: fix get_reg_for_operand() with no free registers
4423 - aco/ra: fix get_reg_for_operand() when the blocking var is a vector
4424 - aco/ra: fix get_reg_for_operand() with vector operands
4425 - aco/ra: use flags instead of booleans for update_renames()
4426 - aco: disallow SGPRs on DPP instructions
4432 - aco: do not clause NSA instructions
4433 - aco/tests: add tests for form_hard_clauses()
4434 - aco/tests: improve reporting of failed code checks
4435 - aco: don't create 4 and 5 dword NSA instructions on GFX10
4436 - aco: don't use nir_block_is_unreachable()
4440 - aco: disallow SDWA for instructions with 64-bit definitions/operands
4441 - aco: add p_extract/p_insert
4442 - aco: implement nir_op_extract/nir_op_insert
4443 - aco: use byte/word extract pseudo-instructions
4446 - aco: optimize 32-bit extracts and inserts using SDWA
4447 - aco: make validate_ir() output usable in tests
4448 - aco: disallow literals with some instruction formats
4449 - aco/tests: add tests for p_extract/p_insert lowering
4450 - aco/tests: add SDWA tests
4451 - aco: use v1b/v2b for ds_read_u8/ds_read_u16
4453 - aco: don't ever widen 8/16-bit sgpr load_shared
4454 - aco: use ds_read_{u8,u16}_d16
4455 - aco: fix emit_mbcnt() with a VGPR mask
4461 - radv,aco: use all attributes in a binding to obtain an alignment for fetch
4462 - aco: adjust the condition for expanding vertex fetch data format
4463 - aco/ra: use adjust_max_used_regs() in compact_relocate_vars()
4464 - aco: don't move descriptor loads below buffer loads
4465 - aco: move VMEM instructions below descriptor loads
4466 - aco/lower_phis: fix undef_operands initialization with >32 predecessors
4467 - aco/lower_phis: don't allocate unused temporary ids
4473 - aco/tests: fix 32-bit build
4475 - aco: remove resource flags
4476 - aco: handle NIR loops without breaks
4481 - aco: don't create v_madmk_f32/v_madak_f32 from v_fma_legacy_f16
4721 - aco: fix opquantize2f16 on GFX6-7
4736 - radv: remove warnings about RADV_PERFTEST=aco,llvm
4743 - aco: adjust NGG if provoking vertex mode is last
4767 - aco: fix derivatives/intrinsics with SGPR sources
4782 - aco: fix emitting discard when the program just ends
4783 - radv: stop reporting ACO from the device name
4802 - aco: fix range checking for SSBO loads/stores with SGPR offset on GFX6-7
4804 - aco: fix emitting literal offsets with SMEM on GFX7
4834 - aco: fix emitting d16 for MIMG instructions on GFX9+
4835 - aco: fix emitting a16 for MIMG instructions on GFX10+
4836 - aco: fix shared_atomic_comp_swap if the second source isn't a VGPR
4838 - aco: use nir_ssa_def_is_unused() to determine if atomic dest is used
4973 - aco: Mark VCC clobbered for iadd8 and iadd16 reductions on GFX6-7.
4980 - aco: Allow workgroup barrier and shared scope for NGG shaders.
4981 - aco: Fixup the NIR metadata after sanitize_cf_list.
4982 - aco: Split ngg_emit_sendmsg_gs_alloc_req from the wave0 check.
4985 - aco: Implement new NGG specific NIR intrinsics.
4987 - radv: Use new NGG NIR lowering for VS/TES when ACO is used.
4989 - radv: Use new NIR lowering of NGG GS when ACO is used.
4990 - aco: Determine whether a few more instructions need exec.
4991 - aco: Use Operand instead of Temp for the exec mask stack.
4992 - aco: Remember when exec mask is const, and restore the const then.
4993 - aco: Don't use s_and_saveexec with branches when exec is constant.
4994 - aco: Refactor SSA elimination phi info to use vector instead of map.
4995 - aco: Eliminate useless exec writes in jump threading.
4996 - aco/insert_exec_mask: Fixed unused variable warning in release build.
4997 - aco/util: Initialize IDSet::bits_set to zero.
5003 - aco: Don't eliminate exec write when it's used by a copy later.
5004 - aco: Don't DCE instructions that write non-temps, eg. exec.
5005 - aco: Add Operand(Temp, PhysReg) constructor.
5006 - aco: New writeout overloads for the test framework.
5007 - aco: Introduce a new, post-RA optimizer.
5008 - aco: Use s_cbranch_vccz/nz in post-RA optimization.
5009 - aco: Eliminate SALU comparison when SCC can be used instead.
5013 - aco: Add note about v_alignbyte in the ISA README.
5015 - aco: Implement nir_op_sad_u8x4.
5016 - aco: Add validation for v_permlane instructions.
5018 - aco: Implement byte and lane permute intrinsics.
5019 - aco: Keep VGPR destinations for uniform shared loads when beneficial.
5022 - aco: Use as_vgpr for the second source of mbcnt_amd.
5024 - aco: Fix checking if load_shared is used by cross lane instructions.
5026 - aco/gfx10: NGG zero output workaround for conservative rasterization.
5027 - aco/gfx10: Emit barrier at the start of NGG VS and TES.
5038 - aco: Implement NGG culling related intrinsics.
5044 - radv, aco, ac/nir: Tweak position export scheduling for NGG culling.
5105 - aco/spill: Fix improper handling of exec phis
5106 - aco/scheduler: Fix register demand computation for downwards moves
5107 - aco/scheduler: Fix register demand computation for upwards moves
5108 - aco/scheduler: Verify register demand invariants in debug mode
5111 - aco/scheduler: Clean up register demand tracking
5112 - aco/scheduler: Move cursor handling state to dedicated interfaces
5113 - aco/ra: Fix off-by-one-error in print_regs
5114 - aco/ra: Clean up print_regs output and support byte-allocated variables
5115 - aco/ra: Split print_regs by lines of 64 registers
5116 - aco: Replace Operand literal constructors with factory member functions
5117 - aco: Remove use of deprecated Operand constructors in test_to_hw_instr.cpp
5118 - aco: Remove use of deprecated Operand constructors in aco_builder.h
5119 - aco: Remove use of deprecated Operand constructors
5120 - aco: Clean up unneeded literal casts
5121 - aco: Remove deprecated Operand constructors