# A config to generate a fast kernel(s) for  BDAS workload problem sizes
# BDAS workloads are kmeans , pca SVD algorithm solvers. usually consists of 2 DGEMM problem sizes
# Here problem size K=500000 represents DGEMM2

GlobalParameters:
  MinimumRequiredVersion: 4.4.0
  PrintLevel: 1
  ForceRedoBenchmarkProblems: True
  ForceRedoLibraryLogic: True
  ForceRedoLibraryClient: True
  CMakeBuildType: Release
  EnqueuesPerSync: 1
  NumBenchmarks: 1
  SyncsPerBenchmark: 1
  LibraryPrintDebug: False
  NumElementsToValidate: 100
  ValidationMaxToPrint: 4
  ValidationPrintValids: False
  ShortNames: False
  MergeFiles: True
  Platform: 0
  Device: 0
  KernelTime: True
  PinClocks: True
  SleepPercent: 0
  PrintSolutionRejectionReason: True
  DataInitTypeBeta : 0
  DataInitTypeAlpha : 1
  NewKernel: 2

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

    - # BenchmarkProblemSizeGroup - Standard
      InitialSolutionParameters:
      BenchmarkCommonParameters:
        - KernelLanguage: ["Assembly"]
        - EdgeType: ["ShiftPtr"]
        - LoopTail: [True]
      ForkParameters:
        - PrefetchLocalRead: [True]
        - PrefetchGlobalRead: [True]
        - ThreadTile:
          - [ 4, 8 ]
          - [ 4, 4 ]
        - WorkGroup:
          - [ 32, 16, 1 ]
        - WorkGroupMapping: [1]
        #- AssertFree0ElementMultiple: [2]
        - GlobalSplitU: [16,32]
        - DepthU: [ 8,16,32 ]
        - VectorWidth: [2]
        - FractionalLoad: [1]
        - GlobalReadVectorWidth: [2,4]
        - OptNoLoadLoop: [2]
        - StaggerUStride: [256,512]
        - StaggerUMapping: [0]
      BenchmarkForkParameters:
      JoinParameters:
      BenchmarkJoinParameters:
      BenchmarkFinalParameters:
        - ProblemSizes:
          - Exact: [ 250, 250, 1, 500000,266,266,500016,500032 ]