---
include: hipsparselt_common.yaml
include: known_bugs.yaml
include: spmm_strided_batched_common.yaml

Definitions:
  - &transA_transB_range
    - { transA: N, transB: N }
    - { transA: N, transB: T }
    - { transA: T, transB: N }
    - { transA: T, transB: T }

  - &alpha_beta_range
    - { alpha:  1.0, beta:  0.0 }
    - { alpha: -2.0, beta: -3.0 }
    - { alpha:  0.0, beta:  1.0 }

  - &alpha_beta_range_small
    - { alpha: 2, alphai: 2, beta: -1.0, betai: 2.0 }

Tests:

- name: spmm_strided_batched_small_stride_zero
  category: quick
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_small_matrix_size_stride_a_range
  alpha: 2.0
  alphai: 1.0
  beta: 3.0
  betai: -1.0
  transA: N
  transB: N
  batch_count: [ 1, 3 ]
  sparse_b: [true, false]
  alpha_vector_scaling: [true, false]

- name: spmm_strided_batched_small
  category: quick
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_small_matrix_size_range
  alpha_beta: *alpha_beta_range
  transA_transB: *transA_transB_range
  batch_count: 3
  bias_vector: true
  bias_stride: [0, -1, 256]
  bias_type: [f32_r, f16_r, bf16_r]
  sparse_b: [true, false]
  alpha_vector_scaling: true

- name: spmm_strided_batched_small
  category: quick
  function:
    spmm_strided_batched: *hpa_half_precision
  matrix_size: *strided_batched_small_matrix_size_range
  alpha_beta: *alpha_beta_range
  transA_transB: *transA_transB_range
  batch_count: 3
  bias_vector: true
  sparse_b: [true, false]
  alpha_vector_scaling: true
  gate_residual: [true]
  stride_gate: [0, -1, 16384]
  gpu_arch: '942'

- name: spmm_strided_batched_small_stride_zero
  category: quick
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_small_matrix_size_stride_a_range
  alpha: 2.0
  alphai: 1.0
  beta: 3.0
  betai: -1.0
  transA: N
  transB: N
  batch_count: 3
  bias_vector: true
  bias_stride: [0, -1, 256]
  bias_type: [f32_r, f16_r, bf16_r]
  sparse_b: [true, false]
  alpha_vector_scaling: true

- name: spmm_strided_batched_medium
  category: pre_checkin
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_medium_matrix_size_range
  transA_transB: *transA_transB_range
  alpha_beta: *alpha_beta_range
  batch_count: 3
  sparse_b: [true, false]
  orderA: [C, R]
  orderB: [C, R]
  orderC: [C]
  orderD: [C]

- name: spmm_strided_batched_medium_alt
  category: pre_checkin
  function:
    spmm_strided_batched: *hpa_half_precision
  matrix_size: *strided_batched_medium_matrix_size_range
  transA_transB: *transA_transB_range
  alpha: 1
  beta: 0
  initialization: special
  batch_count: 3
  sparse_b: [true, false]

- name: spmm_strided_batched_medium_stride_zero
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_medium_matrix_size_stride_a_range
  alpha: 2.0
  alphai: 1.0
  beta: 3.0
  betai: -1.0
  transA: N
  transB: N
  batch_count: 3
  sparse_b: [true, false]

- name: spmm_strided_batched_medium_stride_zero_alt
  category: nightly
  function:
    spmm_strided_batched: *hpa_half_precision
  matrix_size: *strided_batched_medium_matrix_size_stride_a_range
  alpha: 1
  beta: 0
  initialization: special
  transA: N
  transB: N
  batch_count: 3
  sparse_b: [true, false]

- name: spmm_strided_batched_large
  category: pre_checkin
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_large_matrix_size_range
  transA_transB: *transA_transB_range
  alpha_beta: *alpha_beta_range
  batch_count: 3
  sparse_b: [true, false]

- name: spmm_strided_batched_large_stride_zero
  category: pre_checkin
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_large_matrix_size_stride_a_range
  alpha: 2.0
  alphai: 1.0
  beta: 3.0
  betai: -1.0
  transA: N
  transB: N
  batch_count: 3
  sparse_b: [true, false]

- name: spmm_strided_batched_deepbench
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_deepbench_vec
  alpha: 1.0
  beta: 0.0
  transA: N
  transB: T
  batch_count: [ 2 ]
  sparse_b: [true, false]

- name: spmm_strided_batched_deepbench_alt
  category: nightly
  function:
    spmm_strided_batched: *hpa_half_precision
  matrix_size: *strided_batched_deepbench_vec
  alpha: 1.0
  beta: 0.0
  initialization: special
  transA: N
  transB: T
  batch_count: 2
  sparse_b: [true, false]

- name: spmm_strided_batched_conv_resnet50_fwd
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_conv_resnet50_fwd
  alpha: 1.0
  beta: 0.0
  transA: N
  transB: N
  batch_count: 4
  sparse_b: [true, false]
  orderA: [C, R]
  orderB: [C, R]

- name: spmm_strided_batched_conv_resnet50_bwddata
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_conv_resnet50_bwddata
  alpha: 1.0
  beta: 0.0
  transA: N
  transB: T
  batch_count: 4
  sparse_b: [true, false]

- name: spmm_strided_batched_conv_inception4_bwddata
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_conv_inception4_bwddata
  alpha: 1.0
  beta: 0.0
  transA: N
  transB: T
  batch_count: 4
  sparse_b: [true, false]

- name: spmm_strided_batched_conv_ctest_fwd
  category: nightly
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size: *strided_batched_conv_ctest_fwd
  alpha: 1.0
  beta: 0.0
  transA: N
  transB: N
  batch_count: 1
  sparse_b: [true, false]

- name: spmm_strided_batched_smoke
  category: smoke
  transA_transB: *transA_transB_range
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size:
    - { M: 8, N: 8, K: 16 }
    - { M: 8, N: 8, K: 32 }
    - { M: 8, N: 8, K: 48 }
    - { M: 8, N: 8, K: 64 }
    - { M: 8, N: 8, K: 128 }
    - { M: 8, N: 8, K: 192 }
    - { M: 8, N: 8, K: 256 }
    - { M: 128, N: 128, K: 16 }
    - { M: 128, N: 128, K: 24 }
    - { M: 128, N: 128, K: 32 }
    - { M: 128, N: 128, K: 48 }
    - { M: 128, N: 128, K: 64 }
    - { M: 128, N: 128, K: 128 }
    - { M: 128, N: 128, K: 192 }
    - { M: 128, N: 128, K: 256 }
  alpha_beta: *alpha_beta_range
  sparse_b: [ true, false]
  batch_count: 3
  alpha_vector_scaling: true
  bias_vector: true
  bias_stride: [0, -1, 256]
  bias_type: [f32_r, f16_r, bf16_r]

- name: spmm_strided_batched_activation_smoke
  category: smoke
  transA_transB: *transA_transB_range
  function:
    spmm_strided_batched: *real_precisions_2b
  matrix_size:
    - { M: 8, N: 8, K: 256 }
    - { M: 128, N: 128, K: 256 }
  alpha: 1.0
  beta: 0.0
  sparse_b: [ true, false]
  batch_count: 3
  activation_type: clippedrelu
  activation_arg1 : [-1.0, 0.0, 0.5]
  activation_arg2 : [-1.0, 0.0, 0.5, 1.0, 3.0]
...