summaryrefslogtreecommitdiff
path: root/src/amd/compiler
AgeCommit message (Expand)AuthorFilesLines
2021-06-22aco/gfx10: Emit barrier at the start of NGG VS and TES.Timur Kristóf1-0/+7
2021-06-22aco/gfx10: NGG zero output workaround for conservative rasterization.Timur Kristóf1-2/+8
2021-06-21nir,amd: Suffix nir_op_cube_face_coord/index with _amdJason Ekstrand2-4/+4
2021-06-21aco: Fix checking if load_shared is used by cross lane instructions.Timur Kristóf1-20/+38
2021-06-14aco/lower_phis: don't allocate unused temporary idsRhys Perry1-14/+34
2021-06-14aco/lower_phis: fix undef_operands initialization with >32 predecessorsRhys Perry1-1/+1
2021-06-14aco: move VMEM instructions below descriptor loadsRhys Perry1-1/+2
2021-06-14aco: don't move descriptor loads below buffer loadsRhys Perry1-0/+3
2021-06-14aco/ra: use adjust_max_used_regs() in compact_relocate_vars()Rhys Perry1-0/+2
2021-06-14aco: adjust the condition for expanding vertex fetch data formatRhys Perry1-8/+5
2021-06-14radv,aco: use all attributes in a binding to obtain an alignment for fetchRhys Perry1-8/+14
2021-06-11aco/ra: refactor register assignment for vector operandsDaniel Schürmann1-40/+58
2021-06-11aco/ra: refactor affinity coalescingDaniel Schürmann1-6/+25
2021-06-11aco/ra: only create phi-affinities for killed operandsDaniel Schürmann1-1/+1
2021-06-10aco: fix emit_mbcnt() with a VGPR maskRhys Perry1-1/+2
2021-06-10aco: Use as_vgpr for the second source of mbcnt_amd.Timur Kristóf1-1/+1
2021-06-09amd: Add extra source to the mbcnt_amd NIR intrinsic.Timur Kristóf1-1/+2
2021-06-09aco: Keep VGPR destinations for uniform shared loads when beneficial.Timur Kristóf1-1/+35
2021-06-09aco: Implement byte and lane permute intrinsics.Timur Kristóf2-0/+27
2021-06-09aco: Add validation for v_permlane instructions.Timur Kristóf2-1/+15
2021-06-09aco: Implement nir_op_sad_u8x4.Timur Kristóf3-2/+12
2021-06-09aco: Add note about v_alignbyte in the ISA README.Timur Kristóf1-0/+12
2021-06-09aco: use ds_read_{u8,u16}_d16Rhys Perry2-11/+10
2021-06-09aco: don't ever widen 8/16-bit sgpr load_sharedRhys Perry1-3/+3
2021-06-09aco: use v1b/v2b for ds_read_u8/ds_read_u16Rhys Perry2-4/+6
2021-06-09aco: fix emitting literal offsets with SMEM on GFX7Samuel Pitoiset1-2/+4
2021-06-09aco: fix range checking for SSBO loads/stores with SGPR offset on GFX6-7Samuel Pitoiset2-0/+19
2021-06-08nir: Move workgroup_size and workgroup_variable_size into common shader_infoCaio Marcelo de Oliveira Filho1-3/+3
2021-06-08aco/ra: Split print_regs by lines of 64 registersTony Wasserka1-38/+56
2021-06-08aco/ra: Clean up print_regs output and support byte-allocated variablesTony Wasserka1-32/+58
2021-06-08aco/ra: Fix off-by-one-error in print_regsTony Wasserka1-3/+2
2021-06-08aco/tests: add SDWA testsRhys Perry7-13/+561
2021-06-08aco/tests: add tests for p_extract/p_insert loweringRhys Perry1-0/+174
2021-06-08aco: disallow literals with some instruction formatsRhys Perry1-4/+3
2021-06-08aco: make validate_ir() output usable in testsRhys Perry3-6/+16
2021-06-08aco: optimize 32-bit extracts and inserts using SDWARhys Perry4-23/+271
2021-06-08aco: use byte/word extract pseudo-instructionsRhys Perry1-20/+13
2021-06-08aco: implement nir_op_extract/nir_op_insertRhys Perry1-0/+74
2021-06-08aco: add p_extract/p_insertRhys Perry4-7/+207
2021-06-08aco: disallow SDWA for instructions with 64-bit definitions/operandsRhys Perry1-0/+7
2021-06-07nir: Rename WORK_GROUP (and similar) to WORKGROUPCaio Marcelo de Oliveira Filho2-11/+11
2021-06-07compiler: Rename local_size to workgroup_sizeCaio Marcelo de Oliveira Filho1-3/+3
2021-06-07aco/scheduler: Move cursor handling state to dedicated interfacesTony Wasserka1-113/+145
2021-06-07aco/scheduler: Clean up register demand trackingTony Wasserka1-25/+30
2021-06-04aco: simplify Phi RegClass selectionDaniel Schürmann2-22/+11
2021-06-04radv,aco: scalarize all phis via nir_lower_phis_to_scalar()Daniel Schürmann2-36/+1
2021-06-04nir/unsigned_upper_bound: don't require dominance metadataRhys Perry1-3/+1
2021-06-04aco: don't use nir_block_is_unreachable()Rhys Perry1-9/+32
2021-06-03aco: don't create 4 and 5 dword NSA instructions on GFX10Rhys Perry2-1/+10
2021-06-03aco/tests: improve reporting of failed code checksRhys Perry1-10/+12
generated by cgit v1.2.3 (git 2.39.1) at 2024-04-28 17:22:09 +0000