Bitwise reproducibility

Contents

Bitwise reproducibility#

Some routines do not produce deterministic results from run to run. This is typically the case when HIP atomics are used. This page catalogues the run-to-run reproducibility of each routine.

Sparse level 1 functions#

Function name	yes	no
`rocsparse_Xaxpyi()`	x
`rocsparse_Xdoti()`	x
`rocsparse_Xdotci()`	x
`rocsparse_Xgthr()`	x
`rocsparse_Xgthrz()`	x
`rocsparse_Xroti()`	x
`rocsparse_Xsctr()`	x

Sparse level 2 functions#

Function name	yes	no
`rocsparse_Xbsrmv_analysis()`	x
`rocsparse_bsrmv_clear()`	x
`rocsparse_Xbsrmv()`	x
`rocsparse_Xbsrxmv()`	x
`rocsparse_Xbsrsv_buffer_size()`	x
`rocsparse_Xbsrsv_analysis()`		x
`rocsparse_bsrsv_zero_pivot()`	x
`rocsparse_bsrsv_clear()`	x
`rocsparse_Xbsrsv_solve()`		x
`rocsparse_Xcsrmv_analysis()`	x
`rocsparse_csrmv_clear()`	x
`rocsparse_Xcsrsv_buffer_size()`	x
`rocsparse_Xcsrsv_analysis()`		x
`rocsparse_csrsv_zero_pivot()`	x
`rocsparse_csrsv_clear()`	x
`rocsparse_Xcsrsv_solve()`		x
`rocsparse_Xcsritsv_buffer_size()`	x
`rocsparse_Xcsritsv_analysis()`		x
`rocsparse_csritsv_zero_pivot()`	x
`rocsparse_csritsv_clear()`	x
`rocsparse_Xcsritsv_solve()`		x
`rocsparse_Xcsritsv_solve_ex()`		x
`rocsparse_Xgemvi_buffer_size()`	x
`rocsparse_Xgemvi()`	x

The reproducibility of rocsparse_Xbsrmv(), rocsparse_Xbsrxmv(), rocsparse_Xcoomv(), rocsparse_Xcsrmv(), rocsparse_Xellmv(), rocsparse_Xhybmv(), and rocsparse_Xgebsrmv() is more complicated, depending on whether A is transposed or not. See the chart below to determine whether these routines are deterministic.

Routine	A non-transpose		A transpose
Routine	Yes	No	Yes	No
rocsparse_Xbsrmv	x		N/A	N/A
rocsparse_Xbsrxmv	x		N/A	N/A
rocsparse_Xcoomv	x			x
rocsparse_Xcsrmv	x			x
rocsparse_Xcsrmv (info != NULL)		x		x
rocsparse_Xellmv	x			x
rocsparse_Xhybmv	x			x
rocsparse_Xgebsrmv	x		N/A	N/A

Sparse level 3 functions#

Function name	yes	no
`rocsparse_Xcsrsm_buffer_size()`	x
`rocsparse_Xcsrsm_analysis()`		x
`rocsparse_csrsm_zero_pivot()`	x
`rocsparse_csrsm_clear()`	x
`rocsparse_Xcsrsm_solve()`		x
`rocsparse_Xbsrsm_buffer_size()`	x
`rocsparse_Xbsrsm_analysis()`		x
`rocsparse_bsrsm_zero_pivot()`	x
`rocsparse_bsrsm_clear()`	x
`rocsparse_Xbsrsm_solve()`		x
`rocsparse_Xgemmi()`	x

The reproducibility of rocsparse_Xbsrmm(), rocsparse_Xgebsrmm(), and rocsparse_Xcsrmm() is more complicated, depending on whether A is transposed or not. See the chart below to determine whether these routines are deterministic.

Routine	A non-transpose		A transpose
Routine	Yes	No	Yes	No
rocsparse_Xbsrmm	x		N/A	N/A
rocsparse_Xgebsrmm	x		N/A	N/A
rocsparse_Xcsrmm	x			x

Sparse extra functions#

Function name	yes	no
`rocsparse_bsrgeam_nnzb()`	x
`rocsparse_Xbsrgeam()`	x
`rocsparse_Xbsrgemm_buffer_size()`	x
`rocsparse_bsrgemm_nnzb()`	x
`rocsparse_Xbsrgemm()`	x
`rocsparse_csrgeam_nnz()`	x
`rocsparse_Xcsrgeam()`	x
`rocsparse_Xcsrgemm_buffer_size()`	x
`rocsparse_csrgemm_nnz()`		x
`rocsparse_csrgemm_symbolic()`		x
`rocsparse_Xcsrgemm()`		x
`rocsparse_Xcsrgemm_numeric()`		x

Preconditioner functions#

Function name	yes	no
`rocsparse_Xbsric0_buffer_size()`	x
`rocsparse_Xbsric0_analysis()`		x
`rocsparse_bsric0_zero_pivot()`	x
`rocsparse_bsric0_clear()`	x
`rocsparse_Xbsric0()`		x
`rocsparse_Xbsrilu0_buffer_size()`	x
`rocsparse_Xbsrilu0_analysis()`		x
`rocsparse_bsrilu0_zero_pivot()`	x
`rocsparse_Xbsrilu0_numeric_boost()`		x
`rocsparse_bsrilu0_clear()`	x
`rocsparse_Xbsrilu0()`		x
`rocsparse_Xcsric0_buffer_size()`	x
`rocsparse_Xcsric0_analysis()`		x
`rocsparse_csric0_zero_pivot()`	x
`rocsparse_csric0_clear()`	x
`rocsparse_Xcsric0()`		x
`rocsparse_Xcsrilu0_buffer_size()`	x
`rocsparse_Xcsrilu0_numeric_boost()`		x
`rocsparse_Xcsrilu0_analysis()`		x
`rocsparse_csrilu0_zero_pivot()`	x
`rocsparse_csrilu0_clear()`	x
`rocsparse_Xcsrilu0()`		x
`rocsparse_csritilu0_buffer_size()`	x
`rocsparse_csritilu0_preprocess()`		x
`rocsparse_Xcsritilu0_compute()`		x
`rocsparse_Xcsritilu0_compute_ex()`		x
`rocsparse_Xcsritilu0_history()`		x
`rocsparse_Xgtsv_buffer_size()`	x
`rocsparse_Xgtsv()`	x
`rocsparse_Xgtsv_no_pivot_buffer_size()`	x
`rocsparse_Xgtsv_no_pivot()`	x
`rocsparse_Xgtsv_no_pivot_strided_batch_buffer_size()`	x
`rocsparse_Xgtsv_no_pivot_strided_batch()`	x
`rocsparse_Xgtsv_interleaved_batch_buffer_size()`	x
`rocsparse_Xgtsv_interleaved_batch()`	x
`rocsparse_Xgpsv_interleaved_batch_buffer_size()`	x
`rocsparse_Xgpsv_interleaved_batch()`	x

Conversion functions#

Function name	yes	no
`rocsparse_csr2coo()`	x
`rocsparse_csr2csc_buffer_size()`	x
`rocsparse_Xcsr2csc()`	x
`rocsparse_Xgebsr2gebsc_buffer_size()`	x
`rocsparse_Xgebsr2gebsc()`	x
`rocsparse_csr2ell_width()`	x
`rocsparse_Xcsr2ell()`	x
`rocsparse_Xcsr2hyb()`	x
`rocsparse_csr2bsr_nnz()`	x
`rocsparse_Xcsr2bsr()`	x
`rocsparse_csr2gebsr_nnz()`	x
`rocsparse_Xcsr2gebsr_buffer_size()`	x
`rocsparse_Xcsr2gebsr()`	x
`rocsparse_coo2csr()`	x
`rocsparse_ell2csr_nnz()`	x
`rocsparse_Xell2csr()`	x
`rocsparse_hyb2csr_buffer_size()`	x
`rocsparse_Xhyb2csr()`	x
`rocsparse_Xbsr2csr()`	x
`rocsparse_Xgebsr2csr()`	x
`rocsparse_Xgebsr2gebsr_buffer_size()`	x
`rocsparse_gebsr2gebsr_nnz()`	x
`rocsparse_Xgebsr2gebsr()`	x
`rocsparse_Xcsr2csr_compress()`	x
`rocsparse_create_identity_permutation()`	x
`rocsparse_inverse_permutation()`	x
`rocsparse_cscsort_buffer_size()`	x
`rocsparse_cscsort()`	x
`rocsparse_csrsort_buffer_size()`	x
`rocsparse_csrsort()`	x
`rocsparse_coosort_buffer_size()`	x
`rocsparse_coosort_by_row()`	x
`rocsparse_coosort_by_column()`	x
`rocsparse_Xdense2csr()`	x
`rocsparse_Xdense2csc()`	x
`rocsparse_Xdense2coo()`	x
`rocsparse_Xcsr2dense()`	x
`rocsparse_Xcsc2dense()`	x
`rocsparse_Xcoo2dense()`	x
`rocsparse_Xnnz_compress()`	x
`rocsparse_Xnnz()`	x
`rocsparse_Xprune_dense2csr_buffer_size()`	x
`rocsparse_Xprune_dense2csr_nnz()`	x
`rocsparse_Xprune_dense2csr()`	x
`rocsparse_Xprune_csr2csr_buffer_size()`	x
`rocsparse_Xprune_csr2csr_nnz()`	x
`rocsparse_Xprune_csr2csr()`	x
`rocsparse_Xprune_dense2csr_by_percentage_buffer_size()`	x
`rocsparse_Xprune_dense2csr_nnz_by_percentage()`	x
`rocsparse_Xprune_dense2csr_by_percentage()`	x
`rocsparse_Xprune_csr2csr_by_percentage_buffer_size()`	x
`rocsparse_Xprune_csr2csr_nnz_by_percentage()`	x
`rocsparse_Xprune_csr2csr_by_percentage()`	x
`rocsparse_Xbsrpad_value()`	x

Reordering functions#

Function name	yes	no
`rocsparse_Xcsrcolor()`	x

Utility functions#

Function name	yes	no
`rocsparse_Xcheck_matrix_csr_buffer_size()`	x
`rocsparse_Xcheck_matrix_csr()`	x
`rocsparse_Xcheck_matrix_csc_buffer_size()`	x
`rocsparse_Xcheck_matrix_csc()`	x
`rocsparse_Xcheck_matrix_coo_buffer_size()`	x
`rocsparse_Xcheck_matrix_coo()`	x
`rocsparse_Xcheck_matrix_gebsr_buffer_size()`	x
`rocsparse_Xcheck_matrix_gebsr()`	x
`rocsparse_Xcheck_matrix_gebsc_buffer_size()`	x
`rocsparse_Xcheck_matrix_gebsc()`	x
`rocsparse_Xcheck_matrix_ell_buffer_size()`	x
`rocsparse_Xcheck_matrix_ell()`	x
`rocsparse_check_matrix_hyb_buffer_size()`	x
`rocsparse_check_matrix_hyb()`	x

Sparse generic functions#

Function name	yes	no
`rocsparse_axpby()`	x
`rocsparse_gather()`	x
`rocsparse_scatter()`	x
`rocsparse_rot()`	x
`rocsparse_spvv()`	x
`rocsparse_sparse_to_dense()`	x
`rocsparse_dense_to_sparse()`	x
`rocsparse_spsv()`		x
`rocsparse_spsm()`		x
`rocsparse_spgemm()`		x
`rocsparse_v2_spmv_buffer_size()`	x
`rocsparse_spgeam_buffer_size()`	x
`rocsparse_spgeam()`	x
`rocsparse_sddmm_buffer_size()`	x
`rocsparse_sddmm_preprocess()`	x
`rocsparse_sddmm()`	x
`rocsparse_sparse_to_sparse_buffer_size()`	x
`rocsparse_sparse_to_sparse()`	x
`rocsparse_extract_buffer_size()`	x
`rocsparse_extract()`	x

The reproducibility of rocsparse_spmv() and rocsparse_v2_spmv() is more complicated because these generic routines support multiple sparse matrix formats and algorithms. See the chart below to determine whether a given algorithm is deterministic.

Bit-wise reproducibility of SpMV/v2_SpMV
Algorithm	A non-transpose		A transpose
Algorithm	Yes	No	Yes	No
rocsparse_spmv_alg_csr_stream	x			x
rocsparse_spmv_alg_csr_rowsplit	x			x
rocsparse_spmv_alg_csr_adaptive		x		x
rocsparse_spmv_alg_csr_lrb		x		x
rocsparse_spmv_alg_csr_stream (CSC FORMAT)		x	x
rocsparse_spmv_alg_csr_rowsplit (CSC FORMAT)		x	x
rocsparse_spmv_alg_csr_adaptive (CSC FORMAT)		x		x
rocsparse_spmv_alg_csr_lrb (CSC FORMAT)		x		x
rocsparse_spmv_alg_coo	x			x
rocsparse_spmv_alg_coo_atomic		x		x
rocsparse_spmv_alg_ell	x		N/A	N/A
rocsparse_spmv_alg_bsr	x		N/A	N/A

The reproducibility of rocsparse_spmm() is more complicated because this generic routine supports multiple sparse matrix formats and algorithms. See the chart below to determine whether a given algorithm is deterministic.

Bit-wise reproducibility of SpMM
Algorithm	A non-transpose		A transpose
Algorithm	Yes	No	Yes	No
rocsparse_spmm_alg_csr	x			x
rocsparse_spmm_alg_csr_row_split	x			x
rocsparse_spmm_alg_csr_nnz_split		x		x
rocsparse_spmm_alg_csr_merge_path		x		x
rocsparse_spmm_alg_csr (CSC FORMAT)		x	x
rocsparse_spmm_alg_csr_row_split (CSC FORMAT)		x	x
rocsparse_spmm_alg_csr_nnz_split (CSC FORMAT)		x		x
rocsparse_spmm_alg_csr_merge_path (CSC FORMAT)		x		x
rocsparse_spmm_alg_coo_segmented	x			x
rocsparse_spmm_alg_coo_atomic		x		x
rocsparse_spmm_alg_coo_segmented_atomic		x		x
rocsparse_spmm_alg_bell	x		N/A	N/A
rocsparse_spmm_alg_bsr	x		N/A	N/A