GlobalParameters:
  MinimumRequiredVersion: 4.4.0
  PrintLevel: 1
  LibraryPrintDebug: True
  PrintSolutionRejectionReason: True
  ForceRedoBenchmarkProblems: True
  ForceRedoLibraryLogic: True
  ForceRedoLibraryClient: True
  CMakeBuildType: Release
  EnqueuesPerSync: 1
  SyncsPerBenchmark: 1
  NumElementsToValidate: -1
  ValidationMaxToPrint: 4
  ValidationPrintValids: False
  ShortNames: False
  MergeFiles: True
  Platform: 0
  Device: 0
  KernelTime: True
  PinClocks: False
  SleepPercent: 200
  DataInitTypeBeta : 0
  PrintTensorA: 1
  PrintTensorB: 1
  PrintTensorC: 3
  PrintTensorD: 3

BenchmarkProblems:
  ########################################
  # NN - standard
  ########################################
  -
    - # ProblemType
      OperationType: GEMM
      DataType: s
      TransposeA: False
      TransposeB: False
      UseBeta: True
      Batched: True

    - # BenchmarkProblemSizeGroup - Standard
      InitialSolutionParameters:
      BenchmarkCommonParameters:
        - KernelLanguage: ["Assembly"]
        - EdgeType: ["ShiftPtr"]
        - LoopTail: [True]
        - PrefetchLocalRead: [True]
      ForkParameters:
        - PrefetchGlobalRead: [True]
        - ThreadTile:
          - [ 8, 8 ]
        - WorkGroup:
          - [ 16, 16, 1 ]
        - WorkGroupMapping: [8]
        - GlobalSplitU: [1]
        - DepthU: [ 8 ]
        - VectorWidth: [-1]
      BenchmarkForkParameters:
      JoinParameters:
        - MacroTile
        - GlobalSplitU
      BenchmarkJoinParameters:
      BenchmarkFinalParameters:
        - ProblemSizes:
            #- Range: [ [5760], [5760], [1], [5760] ]
          - Range: [ [4], [4], [1], [6] ]

  ########################################
  # NT - standard
  ########################################
  -
    - # ProblemType
      OperationType: GEMM
      DataType: s
      TransposeA: False
      TransposeB: True
      UseBeta: True
      Batched: True

    - # BenchmarkProblemSizeGroup - Standard
      InitialSolutionParameters:
      BenchmarkCommonParameters:
        - KernelLanguage: ["Assembly"]
        - EdgeType: ["ShiftPtr"]
        - LoopTail: [True]
        - PrefetchLocalRead: [True]
      ForkParameters:
        - PrefetchGlobalRead: [True]
        - ThreadTile:
          - [ 8, 8 ]
        - WorkGroup:
          - [ 16, 16, 1 ]
        - WorkGroupMapping: [8]
        - GlobalSplitU: [1]
        - DepthU: [ 8 ]
        - VectorWidth: [-1]
      BenchmarkForkParameters:
      JoinParameters:
        - MacroTile
        - GlobalSplitU
      BenchmarkJoinParameters:
      BenchmarkFinalParameters:
        - ProblemSizes:
            #- Range: [ [5760], [5760], [1], [5760] ]
          - Range: [ [4], [4], [1], [6] ]

  ########################################
  # TN - standard
  ########################################
  -
    - # ProblemType
      OperationType: GEMM
      DataType: s
      TransposeA: True
      TransposeB: False
      UseBeta: True
      Batched: True

    - # BenchmarkProblemSizeGroup - Standard
      InitialSolutionParameters:
      BenchmarkCommonParameters:
        - KernelLanguage: ["Assembly"]
        - EdgeType: ["ShiftPtr"]
        - LoopTail: [True]
        - PrefetchLocalRead: [True]
      ForkParameters:
        - PrefetchGlobalRead: [True]
        - ThreadTile:
          - [ 8, 8 ]
        - WorkGroup:
          - [ 16, 16, 1 ]
        - WorkGroupMapping: [8]
        - GlobalSplitU: [1]
        - DepthU: [ 8 ]
        - VectorWidth: [-1]
      BenchmarkForkParameters:
      JoinParameters:
        - MacroTile
        - GlobalSplitU
      BenchmarkJoinParameters:
      BenchmarkFinalParameters:
        - ProblemSizes:
            #- Range: [ [5760], [5760], [1], [5760] ]
          - Range: [ [4], [4], [1], [6] ]

  ########################################
  # TT - standard
  ########################################
  -
    - # ProblemType
      OperationType: GEMM
      DataType: s
      TransposeA: True
      TransposeB: True
      UseBeta: True
      Batched: True

    - # BenchmarkProblemSizeGroup - Standard
      InitialSolutionParameters:
      BenchmarkCommonParameters:
        - KernelLanguage: ["Assembly"]
        - EdgeType: ["ShiftPtr"]
        - LoopTail: [True]
        - PrefetchLocalRead: [True]
      ForkParameters:
        - PrefetchGlobalRead: [True]
        - ThreadTile:
          - [ 8, 8 ]
        - WorkGroup:
          - [ 16, 16, 1 ]
        - WorkGroupMapping: [8]
        - GlobalSplitU: [1]
        - DepthU: [ 8 ]
        - VectorWidth: [-1]
      BenchmarkForkParameters:
      JoinParameters:
        - MacroTile
        - GlobalSplitU
      BenchmarkJoinParameters:
      BenchmarkFinalParameters:
        - ProblemSizes:
            #- Range: [ [5760], [5760], [1], [5760] ]
          - Range: [ [4], [4], [1], [6] ]
          - Range: [ [6], [4], [1], [4] ]

  ########################################