Lines Matching refs:aco

48 -  New compiler backend "ACO" for RADV (RADV_PERFTEST=aco)
58 -  radv/aco Jedi Fallen Order hair rendering buggy
484 -  amd: Build aco only if radv is enabled
782 -  aco: Initial commit of independent AMD compiler
783 -  radv/aco: Setup alternate path in RADV to support the experimental
786 -  radv/aco: enable VK_EXT_shader_demote_to_helper_invocation
788 -  aco: only emit waitcnt on loop continues if we there was some load or
793 -  radv/aco: Don't lower subtractions
794 -  aco: call nir_opt_algebraic_late() exhaustively
796 -  aco: re-use existing phi instruction when lowering boolean phis
797 -  aco: don't reorder instructions in order to lower boolean phis
798 -  aco: don't combine minmax3 if there is a neg or abs modifier in
800 -  aco: ensure that uniform booleans are computed in WQM if their uses
802 -  aco: refactor value numbering
803 -  aco: restrict scheduling depending on max_waves
804 -  aco: only skip RAR dependencies if the variable is killed somewhere
805 -  aco: add can_reorder flags to load_ubo and load_constant
806 -  aco: don't schedule instructions through depending VMEM instructions
807 -  aco: Lower to CSSA
808 -  aco: improve live variable analysis
809 -  aco: remove potential critical edge on loops.
810 -  aco: fix live-range splits of phis
811 -  aco: fix transitive affinities of spilled variables
812 -  aco: don't insert the exec mask into set of live-out variables when
814 -  aco: consider loop_exit blocks like merge blocks, even if they have
816 -  aco: don't add interferences between spilled phi operands
817 -  aco: simplify calculation of target register pressure when spilling
818 -  aco: ensure that spilled VGPR reloads are done after p_logical_start
819 -  aco: omit linear VGPRs as spill variables
820 -  aco: always set scratch_offset in startpgm
821 -  aco: implement VGPR spilling
823 -  aco: fix immediate offset for spills if scratch is used
824 -  aco: only use single-dword loads/stores for spilling
825 -  aco: fix accidential reordering of instructions when scheduling
826 -  aco: workaround Tonga/Iceland hardware bug
827 -  aco: fix invalid access on Pseudo_instructions
828 -  aco: preserve kill flag on moved operands during RA
829 -  aco: don't split live-ranges of linear VGPRs
830 -  aco: fix a couple of value numbering issues
2539 -  android: aco: fix undefined template 'std::__1::array' build errors
2541 -  android: aco: add support for libmesa_aco
2543 -  android: aco: fix Lower to CSSA
2554 -  aco: Cleanup insert_before_logical_end
2798 -  aco: run nir_lower_int64() before nir_lower_idiv()
2799 -  aco: implement 64-bit ineg
2800 -  aco: fix GFX9 opcode for v_xad_u32
2801 -  aco: fix v_subrev_co_u32_e64 opcode
2802 -  aco: fix opcode for s_mul_hi_i32
2803 -  aco: check for duplicate opcode numbers
2804 -  radv/aco: actually disable ACO when unsupported
2805 -  aco,radv/aco: get dissassembly for release builds if requested
2806 -  aco: store printed backend IR in binary
2807 -  radv/aco: return a correct name and description for the backend IR
2808 -  aco,radv: rename record_llvm_ir/llvm_ir_string to record_ir/ir_string
2809 -  aco: don't CSE v_readlane_b32/v_readfirstlane_b32
2810 -  aco: CSE readlane/readfirstlane/permute/reduce with the same exec
2812 -  aco: set loop_info::has_discard for demotes
2813 -  aco: don't remove the loop exec mask in transition_to_Exact()
2814 -  radv/aco,aco: set lower_fmod
2816 -  aco: fix load_constant with multiple arrays
2819 -  aco: move s_andn2_b64 instructions out of the p_discard_if
2820 -  aco: enable nir_opt_sink
2821 -  aco: Allow literals on VOP3 instructions.
2822 -  aco: Assemble opsel in VOP3 instructions.
2823 -  aco: workaround GFX10 0x3f branch bug
2824 -  aco: pad code with s_code_end on GFX10
2825 -  aco: Initial work to avoid GFX10 hazards.
2826 -  aco: Use the VOP3-only add/sub GFX10 instructions if needed.
2827 -  aco: Have s_waitcnt_vscnt write to NULL.
2828 -  radv/aco: disable NGG when ACO is used
2829 -  aco/gfx10: fix inline uniform blocks
2830 -  aco/gfx10: disable GFX9 1D texture workarounds
2831 -  aco: rework scratch resource code
2832 -  aco: update print_ir
2835 -  aco: don't apply sgprs/constants to read/write lane instructions
2836 -  aco: use can_accept_constant in valu_can_accept_literal
2837 -  aco: readfirstlane vgpr pointers in convert_pointer_to_64_bit()
2838 -  aco: implement divergent vulkan_resource_index
2839 -  aco: don't use p_as_uniform for vgpr sampler/image indices
2840 -  aco: fix scheduling with s_memtime/s_memrealtime
2841 -  aco: don't CSE s_memtime
2842 -  aco: emit_split_vector() s_memtime results
2844 -  aco: use nir_lower_idiv_precise
2845 -  aco: run opt_algebraic in a loop
2846 -  aco: small stage corrections
2847 -  aco: fix 64-bit p_extract_vector on 32-bit p_create_vector
2848 -  aco: create load_lds/store_lds helpers
2849 -  aco: fix sparse store_lds()
2850 -  aco: properly combine additions into ds_write2_b64/ds_read2_b64
2851 -  aco: use ds_read2_b64/ds_write2_b64
2852 -  aco: add a few missing checks in value numbering
2853 -  aco: keep can_reorder/barrier when combining addition into SMEM
2854 -  aco: add missing bld.scc()
2855 -  Revert "aco: only emit waitcnt on loop continues if we there was some
2858 -  aco: increase accuracy of SGPR limits
2859 -  aco: take LDS into account when calculating num_waves
2860 -  aco: Fix reductions on GFX10.
2861 -  aco: Remove dead code in reduction lowering.
2862 -  aco: try to group together VMEM loads of the same resource
2863 -  aco: a couple loop handling fixes for GFX10 hazard pass
2864 -  aco: rename README to README.md
2865 -  aco: fix new_demand calculation for first instructions
2866 -  aco: fix shuffle with uniform operands
2867 -  aco: fix read_invocation with VGPR lane index
2868 -  aco: don't propagate vgprs into v_readlane/v_writelane
2869 -  aco: don't combine literals into v_cndmask_b32/v_subb/v_addc
2870 -  aco: fix 64-bit fsign with 0
2871 -  aco: propagate p_wqm on an image_sample's coordinate p_create_vector
2872 -  aco: fix i2i64
2873 -  aco: add v_nop inbetween exec write and VMEM/DS/FLAT
3280 -  aco: Set +wavefrontsize64 for LLVM disassembler in GFX10 wave64 mode.
3281 -  aco: Add missing GFX10 specific fields and some README notes.
3282 -  aco: Support GFX10 SMEM in aco_assembler.
3283 -  aco: Support GFX10 VINTRP in aco_assembler.
3284 -  aco: Support GFX10 DS in aco_assembler.
3285 -  aco: Support GFX10 MUBUF in aco_assembler.
3287 -  aco: Link ACO with amd/common.
3288 -  aco: Support GFX10 MTBUF in aco_assembler.
3289 -  aco: Support GFX10 MIMG and GFX9 D16 in aco_assembler.
3290 -  aco: Fix GFX9 FLAT, SCRATCH, GLOBAL instructions, add GFX10 support.
3291 -  aco: Support GFX10 EXP in aco_assembler.
3292 -  aco: Support GFX10 VOP3 and VOP1 as VOP3 in aco_assembler.
3293 -  aco: Set GFX10 DLC bit properly.
3294 -  aco: Use ac_get_sampler_dim, delete duplicate code.
3295 -  aco: Set GFX10 dimensionality on the instructions that need it.
3296 -  aco: Support subvector loops in aco_assembler.
3297 -  aco: Fix VS input VGPRs on GFX10.
3298 -  aco: Fix s_dcache_wb on GFX10.
3299 -  aco: Add extra assertion for number of FS input VGPRs.
3300 -  aco: Clean up usages of PhysReg::reg from aco_assembler.
3301 -  aco/gfx10: Wait for pending SMEM stores before loads
3302 -  aco/gfx10: Fix PS exports for SPI_SHADER_32_AR.
3303 -  aco/gfx10: Update constant addresses in fix_branches_gfx10.
3304 -  aco/gfx10: Add notes about some GFX10 hazards.
3305 -  aco/gfx10: Mitigate VcmpxPermlaneHazard.
3306 -  aco/gfx10: Mitigate VcmpxExecWARHazard.
3307 -  aco/gfx10: Mitigate SMEMtoVectorWriteHazard.
3308 -  aco/gfx10: Mitigate LdsBranchVmemWARHazard.
3309 -  aco/gfx10: Fix mitigation of VMEMtoScalarWriteHazard.
3310 -  aco: Refactor hazard mitigations, separate pass for GFX10.
3313 -  aco: Implement subgroup shuffle in GFX10 wave64 mode.
3314 -  aco: Introduce vgpr_limit to keep track of available VGPRs.