--- include: rocblas_common.yaml include: known_bugs.yaml Definitions: - &small_matrix_size_range - { N: 4, lda: 4 } - { N: 9, lda: 9 } - { N: 16, lda: 16 } - { N: 63, lda: 63 } - { N: 65, lda: 65 } - { N: 128, lda: 128 } - { N: 1, lda: 1 } - { N: 10, lda: 9 } - { N: 1, lda: 1 } - { N: 10, lda: 2 } - { N: 200, lda: 200 } - &special_case_range # Quick return - { N: 0, lda: 1, incx: 1, incy: 1, batch_count: 1 } - { N: 1, lda: 1, incx: 1, incy: 1, batch_count: 0 } # invalid_arg checks - { N: -1, lda: 1, incx: 1, incy: 1, batch_count: 0 } - { N: 0, lda: 0, incx: 1, incy: 1, batch_count: 0 } - { N: 1, lda: 0, incx: 1, incy: 1, batch_count: 0 } - { N: 0, lda: 1, incx: 0, incy: 1, batch_count: 0 } - { N: 0, lda: 1, incx: 1, incy: 0, batch_count: 0 } - { N: 0, lda: 1, incx: 1, incy: 1, batch_count: -1 } - &tiny_matrix_size - { N: 127, lda: 127 } - &medium_matrix_size_range - { N: 400, lda: 400 } - { N: 500, lda: 601 } - &large_matrix_size_range - { N: 1000, lda: 1000 } - { N: 2000, lda: 2000 } - { N: 4011, lda: 4011 } - &incx_incy_range - { incx: 1, incy: 1 } - { incx: -1, incy: 3 } - { incx: 3, incy: -1 } - { incx: 10, incy: 100 } - &incx_incy_range_small - { incx: 2, incy: 2 } - &alpha_beta_range - { alpha: 2.0, beta: 0.0, alphai: 1.5, betai: 0.5 } - { alpha: -1.0, beta: -1.0, alphai: 0.5, betai: 1.5 } - { alpha: 2.0, beta: 1.0, alphai: -1.5, betai: 0.5 } - { alpha: 0.0, beta: 1.0, alphai: -0.5, betai: 0 } - &alpha_beta_range_small - { alpha: 2.0, beta: 1.5, alphai: -0.5, betai: 2.5 } - &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, stride_a: 2206744576 } - &size_t_incx_incy - { N: 3, lda: 3, incx: *c_pos_x2_overflow_int32, incy: 1 } # x is m element - { N: 3, lda: 3, incx: 1, incy: *c_pos_x2_overflow_int32 } # y is n element Tests: - name: hemv_bad_arg category: quick function: - hemv_bad_arg - hemv_batched_bad_arg - hemv_strided_batched_bad_arg precision: *single_double_precisions_complex uplo: N api: [ C, FORTRAN, FORTRAN_64 ] - name: hemv_NaN category: pre_checkin function: hemv precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range_small alpha: [ -1.0, .NaN ] # NaN is converted to 0.0 in test code alphai: [ 1.0 ] beta: [ 0.5, 1.0, .NaN ] betai: [ 0.5 ] - name: hemv_fortran category: quick function: hemv precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small api: FORTRAN - name: hemv_small category: quick function: hemv precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range - name: hemv_medium category: pre_checkin function: hemv precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range - name: hemv_large category: nightly function: hemv precision: *single_double_precisions_complex 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: hemv_size_t category: stress function: hemv precision: *single_precision_complex uplo: [ U, L ] matrix_size: *size_t_index incx: 1 incy: 1 alpha_beta: *alpha_beta_range_small pointer_mode_host: false - name: hemv_size_t_inc category: stress function: hemv precision: *single_precision_complex uplo: [ U, L ] arguments: *size_t_incx_incy alpha_beta: *alpha_beta_range_size_t_inc pointer_mode_device: false os_flags: LINUX gpu_arch: ['90a','942','950'] # hemv_batched - name: hemv_batched_NaN category: pre_checkin function: hemv_batched precision: *single_double_precisions_complex uplo: [ L ] matrix_size: *tiny_matrix_size incx_incy: *incx_incy_range alpha: [ 1.5, .NaN ] # NaN is converted to 0.0 in test code beta: [ 0.5, 1.0, .NaN ] batch_count: [ 2 ] - name: hemv_batched_fortran category: quick function: hemv_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ -1, 0, 3 ] api: FORTRAN - name: hemv_batched_small category: quick function: hemv_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range batch_count: [ 3 ] - name: hemv_batched_medium category: pre_checkin function: hemv_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range batch_count: [ 3 ] - name: hemv_batched_large category: nightly function: hemv_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *large_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ 3 ] pointer_mode_host: false - name: hemv_batched_grid_z_16bit category: stress arguments: - { N: 3, lda: 3, incx: 1, incy: 1, batch_count: *c_grid_yz_require_passes} #generic kernel uplo: [ U, L ] alpha_beta: *alpha_beta_range_small pointer_mode_host: false function: - hemv_batched: *single_precision_complex - hemv_strided_batched: *single_precision_complex stride_scale: [ 1 ] # hemv_strided_batched - name: hemv_strided_batched_NaN category: pre_checkin function: hemv_strided_batched precision: *single_double_precisions_complex uplo: [ U ] matrix_size: *tiny_matrix_size incx_incy: *incx_incy_range alpha: [ 2.0, .NaN ] # NaN is converted to 0.0 in test code beta: [ 0.5, 1.0, .NaN ] batch_count: [ 2 ] stride_scale: [ 1.5 ] - name: hemv_strided_batched_fortran category: quick function: hemv_strided_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ -1, 0, 3 ] stride_scale: [ 2 ] api: FORTRAN - name: hemv_strided_batched_small category: quick function: hemv_strided_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *small_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range batch_count: [ 3 ] stride_scale: [ 1.5] - name: hemv_strided_batched_medium category: pre_checkin function: hemv_strided_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *medium_matrix_size_range incx_incy: *incx_incy_range alpha_beta: *alpha_beta_range batch_count: [ 3 ] stride_scale: [ 1 ] - name: hemv_strided_batched_large category: nightly function: hemv_strided_batched precision: *single_double_precisions_complex uplo: [ U, L ] matrix_size: *large_matrix_size_range incx_incy: *incx_incy_range_small alpha_beta: *alpha_beta_range_small batch_count: [ 3 ] stride_scale: [ 1 ] pointer_mode_host: false # 64-bit interface - name: hemv_ILP64 category: stress arguments: - { N: 2, lda: 2147483649, incx: 1, incy: 1, batch_count: 1 } - { N: 2, lda: 2, incx: 2147483649, incy: 1, batch_count: 1 } - { N: 2, lda: 2, incx: 1, incy: 2147483649, batch_count: 1 } - { N: 2, lda: 2, 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: - hemv: *single_precision_complex - hemv_batched: *single_precision_complex - hemv_strided_batched: *single_precision_complex - name: hemv_graph_test category: pre_checkin function: - hemv - hemv_batched - hemv_strided_batched precision: *single_double_precisions_complex matrix_size: *tiny_matrix_size uplo: [ U ] alpha: [ 2.0 ] stride_scale: [ 1 ] batch_count: [ 5 ] graph_test: true - name: hemv_repeatability_check category: stress function: - hemv - hemv_batched - hemv_strided_batched precision: *single_double_precisions_complex initialization: hpl uplo: [ U, L ] matrix_size: - { N: 4011, lda: 4011 } alpha_beta: *alpha_beta_range_small atomics_mode: 0 batch_count: [ 2 ] iters: 5 stride_scale: 1 pointer_mode_host: false 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: hemv_stride_alpha_beta category: pre_checkin function: - hemv_batched - hemv_strided_batched precision: *single_double_precisions_complex 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 ...