aboutsummaryrefslogtreecommitdiff
path: root/llvm/test/Analysis/CostModel/AMDGPU
AgeCommit message (Expand)AuthorFilesLines
6 daysIR: Promote "denormal-fp-math" to a first class attribute (#174293)Matt Arsenault1-2/+2
2025-11-21AMDGPU: Improve getShuffleCost accuracy for 8- and 16-bit shuffles (#168818)Nicolai Hähnle1-488/+498
2025-11-20AMDGPU: Expand cost model shufflevector test (#168816)Nicolai Hähnle1-0/+50
2025-09-25[AMDGPU] Fix vector legalization for bf16 valu ops (#158439)Giuseppe Rossini5-447/+608
2025-07-15[CostModel] Handle all cost kinds in getCmpSelInstrCost (#148233)David Green2-4/+4
2025-07-07[BasicTTIImpl] Add cost entries for ldexp, [l]lround (#146373)Ramkumar Ramachandra1-50/+50
2025-06-26[AMDGPU] Enable vectorization of i8 values. (#134934)Gheorghe-Teodor Bercea3-274/+274
2025-06-26[CostModel] getInstructionCost - match SK_InsertSubvector shuffle patterns be...Simon Pilgrim1-24/+24
2025-06-23[CostModel] improveShuffleKindFromMask - recognise a SK_PermuteSingleSrc inco...Simon Pilgrim1-24/+24
2025-06-18AMDGPU: Report special input intrinsics as free (#141948)Matt Arsenault1-28/+28
2025-06-18AMDGPU: Add baseline cost model tests for special argument intrinsics (#141947)Matt Arsenault1-0/+202
2025-06-18AMDGPU: Cost model for minimumnum/maximumnum (#141946)Matt Arsenault2-528/+636
2025-06-18AMDGPU: Reduce cost of f64 copysign (#141944)Matt Arsenault1-16/+16
2025-06-18AMDGPU: Fix cost model for 16-bit operations on gfx8 (#141943)Matt Arsenault2-26/+26
2025-06-18AMDGPU: Add cost model tests for minimumnum/maximumnum (#141904)Matt Arsenault2-0/+904
2025-05-09[AMDGPU] Update cost model gfx950 min/max tests. NFC. (#139310)Stanislav Mekhanoshin2-0/+38
2025-05-09[AMDGPU][NFC] Add tests in preparation for i8 vectorization (#138801)Gheorghe-Teodor Bercea2-0/+820
2025-05-08[AMDGPU] Legalize vector fminimum and fmaximum with VOP3P (#138971)Stanislav Mekhanoshin2-24/+24
2025-03-06AMDGPU: Add gfx950 cost model tests for minimum and maximum (#130029)Matt Arsenault2-0/+74
2025-02-26[CostModel] Handle vector struct results and cost `llvm.sincos` (#123210)Benjamin Maxwell1-28/+28
2024-12-24[SLP]Add cost estimation for gather node reshufflingAlexey Bataev1-8/+8
2024-11-06[LLVM][IR] Use splat syntax when printing Constant[Data]Vector. (#112548)Paul Walker6-736/+736
2024-08-09AMDGPU: Correct costs of saturating add/sub intrinsics (#100808)Matt Arsenault2-336/+336
2024-08-09TTI: Check legalization cost of abs nodes (#100523)Matt Arsenault3-216/+216
2024-08-08TTI: Check legalization cost of min/max ISD nodes (#100514)Matt Arsenault2-212/+116
2024-08-01AMDGPU: Improve cost handling of canonicalize (#101479)Matt Arsenault1-102/+102
2024-07-30AMDGPU: Enable vectorization of v2f16 copysign (#100799)Matt Arsenault1-128/+128
2024-07-30AMDGPU: Improve cost handling of fma/fmuladd (#100798)Matt Arsenault2-70/+70
2024-07-29AMDGPU: Add some baseline cost model tests (#100797)Matt Arsenault24-414/+4616
2024-07-26AMDGPU: Add baseline test for cost of abs intrinsics (#100522)Matt Arsenault1-0/+310
2024-05-21[AMDGPU] Vectorize more 16 bit shuffles (#90648)Jeffrey Byrnes1-542/+1079
2024-03-07[CostModel] getInstructionCost - improve estimation of costs for length chang...Simon Pilgrim1-184/+184
2023-11-14[AMDGPU] Generic lowering for rint and nearbyint (#69596)Acim-Maravic1-24/+92
2023-10-27[AMDGPU] make v32i16/v32f16 legal (#70484)Changpeng Fang9-40/+40
2023-10-26[opt] Infer DataLayout from triple if not specifiedAlex Richardson2-8/+8
2023-10-15[AMDGPU] Fix cost of fast unsafe f32 fdiv (#68988)Jay Foad1-0/+62
2023-09-12AMDGPU: Correctly lower llvm.sqrt.f32Matt Arsenault1-8/+8
2023-07-25AMDGPU: Correctly expand f64 sqrt intrinsicMatt Arsenault1-8/+8
2023-05-17[NFC][Py Reformat] Reformat lit.local.cfg python files in llvmTobias Hieta1-1/+1
2022-12-15[CostModel] Convert some tests to opaque pointers (NFC)Nikita Popov4-229/+193
2022-11-29[AMDGPU] Add support for new LLVM vector typesMateja Marjanovic9-42/+42
2022-08-10[CostModel] Update RUN -passes=* to double quotes to appease update scripts o...Simon Pilgrim34-199/+199
2022-06-30[AMDGPU] Make v16i16/v16f16 legalPiotr Sobczak11-44/+44
2022-02-09[test] Migrate '-analyze -cost-model' to '-passes=print<cost-model>'Arthur Eubanks34-199/+199
2022-01-24[AMDGPU] Make v8i16/v8f16 legalStanislav Mekhanoshin12-164/+136
2022-01-12[CostModel] Use cost of target trunc type when only it is the only use of a n...Andrew Litteken1-0/+27
2021-12-22[NFC][AMDGPU][CostModel] Add tests for AMDGPU cost model, part 2.Daniil Fukalov11-0/+4333
2021-12-22[NFC][AMDGPU][CostModel] Add tests for AMDGPU cost model.Daniil Fukalov4-0/+1097
2021-12-13[CostModel][AMDGPU] Fix intrinsics costs estimations.Daniil Fukalov5-82/+876
2021-12-03[CostModel][AMDGPU] Fix instructions costs estimation for vector types.Daniil Fukalov9-117/+258
generated by cgit v1.1 at 2026-02-11 15:19:03 +0000