--- include: rocblas_common.yaml include: known_bugs.yaml Definitions: - &special_case_range - { N: 0, lda: 1, K: 0, incx: 1, incy: 1, batch_count: 1 } - { N: 1, lda: 1, K: 0, batch_count: 0 } - { N: -1, lda: 1, K: 0, incx: 1, incy: 1, batch_count: 0 } - { N: 0, lda: 1, K: -1, incx: 1, incy: 1, batch_count: 0 } - { N: 0, lda: 1, K: 1, incx: 1, incy: 1, batch_count: 0 } - { N: 0, lda: 1, K: 0, incx: 0, incy: 1, batch_count: 0 } - { N: 0, lda: 1, K: 0, incx: 1, incy: 0, batch_count: 0 } - { N: 0, lda: 1, K: 0, incx: 1, incy: 1, batch_count: -1 } - &tiny_matrix - { N: 33, lda: 33, K: 3} - &medium_matrix_size_range - { N: -1, lda: -1, K: -1 } - { N: 10, lda: 2, K: -1 } - { N: 33, lda: 33, K: 1} - { N: 33, lda: 33, K: 32 } - { N: 300, lda: 600, K: 99 } - &large_matrix_size_range - { N: 4011, lda: 4011, K: 53 } - { N: 8000, lda: 8000, K: 129 } - &incx_incy_range - { incx: 1, incy: 1 } - { incx: 0, incy: -1 } - { incx: 2, incy: 1 } - { incx: -1, incy: -1 } - &incx_incy_range_small - { incx: -2, incy: 2 } - &alpha_beta_range - { alpha: 1.0, beta: 0.0 } - { alpha: -1.0, beta: -1.0 } - { alpha: 2.0, beta: 1.0 } - { alpha: 0.0, beta: 1.0 } # quick success - &alpha_beta_range_small - { alpha: 2.0, beta: 1.5 } - &alpha_beta_range_nan - { alpha: .NaN, beta: 2.0 } - { alpha: 1.5, beta: .NaN } # alpha value changed from 1.2 -> 1.5 to overcome the equality issue when comparing results of rocBLAS with the host - &alpha_beta_range_size_t_inc - { alpha: 0, alphai: 0, beta: 0.5, betai: 0 } # only incy used - { alpha: 1, alphai: 0, beta: 0.5, betai: 0 } - &size_t_index - { N: 46976, lda: 46976, K: 45777, incx: 1, incy: 1 } - &size_t_incx_incy - { N: 3, lda: 3, K: 2, incx: *c_pos_x2_overflow_int32, incy: 1 } - { N: 3, lda: 3, K: 2, incx: 1, incy: *c_pos_x2_overflow_int32 } Tests: - name: sbmv_bad category: quick function: - sbmv_bad_arg - sbmv_batched_bad_arg - sbmv_strided_batched_bad_arg precision: *single_precision api: [ C, FORTRAN, FORTRAN_64 ] - name: sbmv_arg_check category: quick function: - sbmv - sbmv_batched - sbmv_strided_batched precision: *single_double_precisions uplo: L matrix_size: *special_case_range api: [ C, FORTRAN ] # non-batched - name: sbmv_NaN category: pre_checkin function: sbmv precision: *single_double_precisions uplo: [ U ] matrix_size: *tiny_matrix incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_nan - name: sbmv_medium category: pre_checkin function: sbmv precision: *single_double_precisions uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range - name: sbmv_large category: nightly function: sbmv precision: *single_double_precisions uplo: [ U, L ] matrix_size: *large_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small pointer_mode_device: false - name: sbmv_size_t category: stress function: sbmv precision: *single_precision uplo: [ U, L ] matrix_size: *size_t_index incx: 1 incy: 1 alpha_beta: *alpha_beta_range_small pointer_mode_host: false os_flags: LINUX gpu_arch: ['90a','942','950'] - name: sbmv_size_t_inc category: stress function: sbmv precision: *single_precision uplo: [ U, L ] arguments: *size_t_incx_incy # zero alpha case only uses inc not lda alpha_beta: *alpha_beta_range_size_t_inc pointer_mode_device: false os_flags: LINUX gpu_arch: ['90a','942','950'] - name: sbmv_ILP64 category: stress arguments: - { N: 2, lda: 2147483649, K: 2, incx: 2, incy: 1, batch_count: 1 } - { N: 2, lda: 2, K: 2, incx: 2147483649, incy: 1, batch_count: 1 } - { N: 2, lda: 2, K: 2, incx: 1, incy: -2147483649, batch_count: 1 } - { N: 2, lda: 2, K: 2, stride_a: 4, incx: 1, incy: 1, batch_count: *c_grid_yz_require_passes } uplo: [ U, L ] stride_scale: [ 1 ] alpha_beta: *alpha_beta_range_small pointer_mode_host: false api: [ C_64 ] os_flags: [ LINUX ] function: - sbmv: *single_precision - sbmv_batched: *single_precision - sbmv_strided_batched: *single_precision # batched - name: sbmv_batched_NaN category: pre_checkin function: sbmv_batched precision: *single_double_precisions uplo: [ U ] matrix_size: *tiny_matrix incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_nan batch_count: [ 2 ] - name: sbmv_batched_medium category: pre_checkin function: sbmv_batched precision: *single_double_precisions uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range batch_count: [ -1, 0, 1, 257 ] pointer_mode_host: false - name: sbmv_batched_large category: nightly function: sbmv_batched precision: *single_double_precisions uplo: [ U, L ] matrix_size: *large_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ 1, 3 ] pointer_mode_host: false - name: sbmv_batched_grid_z_16bit category: stress arguments: - { N: 3, K: 3, lda: 3, incx: 1, incy: 1, batch_count: *c_grid_yz_require_passes } alpha_beta: *alpha_beta_range_small uplo: [ U, L ] pointer_mode_host: false function: - sbmv_batched: *single_precision - sbmv_strided_batched: *single_precision stride_scale: 1 # strided batched - name: sbmv_strided_batched_NaN category: pre_checkin function: sbmv_strided_batched precision: *single_precision uplo: [ L ] matrix_size: *tiny_matrix incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_nan batch_count: [ 2 ] stride_scale: [ 2 ] - name: sbmv_strided_batched_medium category: pre_checkin function: sbmv_strided_batched precision: *single_double_precisions uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ -1, 0, 1, 257 ] stride_scale: [1] pointer_mode_host: false - name: sbmv_strided_batched_large category: nightly function: sbmv_strided_batched precision: *single_double_precisions uplo: [ U, L ] matrix_size: *large_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ 1, 3 ] stride_scale: [1] pointer_mode_device: false - name: sbmv_graph_test category: pre_checkin function: - sbmv - sbmv_batched - sbmv_strided_batched precision: *single_double_precisions matrix_size: *tiny_matrix uplo: [ U ] incx: [ 1 ] alpha: [ 2.0 ] stride_scale: [ 1 ] batch_count: [ 5 ] graph_test: true - name: sbmv_repeatablity_check category: stress function: - sbmv - sbmv_batched - sbmv_strided_batched precision: *single_double_precisions matrix_size: *tiny_matrix uplo: [ U ] incx: [ 1 ] alpha: [ 2.0 ] stride_scale: [ 1 ] batch_count: [ 2 ] initialization: hpl atomics_mode: 0 pointer_mode_host: false iters: 5 devices: 4 repeatability_check: true # per-batch alpha/beta (rocblas_set_batch_alpha_stride / rocblas_set_batch_beta_stride) # coverage, device pointer mode only - name: sbmv_stride_alpha_beta category: pre_checkin function: - sbmv_batched - sbmv_strided_batched precision: *single_double_precisions uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range_small stride_scale: [ 1 ] batch_count: [ 3 ] alpha_beta_stride: true stride_c: 2 # alpha stride stride_d: 2 # beta stride pointer_mode_host: false ...