.. meta:: :description: rocSPARSE user guide and documentation :keywords: rocSPARSE, ROCm, API, documentation, user guide .. _rocsparse_docs: ******************************************************************** rocSPARSE user guide ******************************************************************** This topic discusses how to use rocSPARSE, including a discussion of device and stream management, storage formats, pointer mode, and how hipSPARSE interacts with rocSPARSE. HIP device management ===================== Before starting a HIP kernel, you can call :cpp:func:`hipSetDevice` to set the device to run the kernel on, for example, device ``2``. Unless you explicitly specify a different device, HIP kernels always run on device ``0``. This HIP (and CUDA) device management approach is not specific to the rocSPARSE library. rocSPARSE honors this approach and assumes you have already set the preferred device before a rocSPARSE routine call. After you set the device, you can create a handle with :ref:`rocsparse_create_handle_`. Subsequent rocSPARSE routines take this handle as an input parameter. rocSPARSE only queries the specified device (using :cpp:func:`hipGetDevice`) and does not set the device for users. It's your responsibility to provide a valid device to rocSPARSE and ensure device safety. If it's not a valid device, rocSPARSE returns an error message. The handle should be destroyed at the end using :ref:`rocsparse_destroy_handle_` to release the resources consumed by the rocSPARSE library. You **cannot** switch devices between :ref:`rocsparse_create_handle_` and :ref:`rocsparse_destroy_handle_`. To change the device, you must destroy the current handle and create another rocSPARSE handle on a new device. .. note:: :cpp:func:`hipSetDevice` and :cpp:func:`hipGetDevice` are not part of the rocSPARSE API. They are part of the `HIP Device Management API `_. HIP stream management ===================== HIP kernels are always launched in a queue, which is also known as a stream. If you don't explicitly specify a stream, the system provides and maintains a default stream, which you cannot create or destroy. However, you can freely create a new stream using :cpp:func:`hipStreamCreate` and bind it to a rocSPARSE handle using :ref:`rocsparse_set_stream_`. The rocSPARSE routines invoke HIP kernels. A rocSPARSE handle is always associated with a stream, which rocSPARSE passes to the kernels inside the routine. One rocSPARSE routine only takes one stream in a single invocation. If you create a stream, you are responsible for destroying it. See the `HIP Stream Management API `_ for more information. Asynchronous execution ====================== All rocSPARSE library functions are non-blocking and execute asynchronously with respect to the host, except for functions which allocate memory themselves, preventing asynchronicity. These functions might return immediately or before the actual computation has finished. To force synchronization, use either :cpp:func:`hipDeviceSynchronize` or :cpp:func:`hipStreamSynchronize`. This ensures all previously executed rocSPARSE functions on the device or the stream have completed. Multiple streams and multiple devices ===================================== If a system has multiple HIP devices, you can run multiple rocSPARSE handles concurrently. However, you **cannot** run a single rocSPARSE handle concurrently on multiple discrete devices. Each handle can only be associated with a single device, and a new handle should be created for each additional device. Graph support for rocSPARSE =========================== Many of the rocSPARSE functions can be captured into a graph node using the HIP Graph Management APIs. See :ref:`Functions supported with Graph Capture` to determine whether a rocSPARSE routine is supported or not. For a list of graph-related HIP APIs, see the `HIP Graph Management API `_. The following code creates a graph with ``rocsparse_function()`` as the graph node. .. code-block:: c++ CHECK_HIP_ERROR((hipStreamBeginCapture(stream, hipStreamCaptureModeGlobal)); rocsparse_(); CHECK_HIP_ERROR(hipStreamEndCapture(stream, &graph)); The captured graph can be launched as shown below: .. code-block:: c++ CHECK_HIP_ERROR(hipGraphInstantiate(&instance, graph, NULL, NULL, 0)); CHECK_HIP_ERROR(hipGraphLaunch(instance, stream)); Graph support requires asynchronous HIP APIs. .. _Functions Supported with Graph Capture: Functions supported with graph capture ======================================== The following functions support graph capture: Sparse level 1 functions ------------------------ ================================================= === == Function name yes no ================================================= === == :cpp:func:`rocsparse_Xaxpyi() ` x :cpp:func:`rocsparse_Xdoti() ` x :cpp:func:`rocsparse_Xdotci() ` x :cpp:func:`rocsparse_Xgthr() ` x :cpp:func:`rocsparse_Xgthrz() ` x :cpp:func:`rocsparse_Xroti() ` x :cpp:func:`rocsparse_Xsctr() ` x ================================================= === == Sparse level 2 functions ------------------------ ============================================================================= === == Function name yes no ============================================================================= === == :cpp:func:`rocsparse_Xbsrmv_analysis() ` x :cpp:func:`rocsparse_bsrmv_clear` x :cpp:func:`rocsparse_Xbsrmv() ` x :cpp:func:`rocsparse_Xbsrxmv() ` x :cpp:func:`rocsparse_Xbsrsv_buffer_size() ` x :cpp:func:`rocsparse_Xbsrsv_analysis() ` x :cpp:func:`rocsparse_bsrsv_zero_pivot` x :cpp:func:`rocsparse_bsrsv_clear` :cpp:func:`rocsparse_Xbsrsv_solve() ` x :cpp:func:`rocsparse_Xcoomv() ` x :cpp:func:`rocsparse_Xcsrmv_analysis() ` x :cpp:func:`rocsparse_Xcsrmv() ` x :cpp:func:`rocsparse_csrmv_clear` x :cpp:func:`rocsparse_Xcsrsv_buffer_size() ` x :cpp:func:`rocsparse_Xcsrsv_analysis() ` x :cpp:func:`rocsparse_csrsv_zero_pivot` x :cpp:func:`rocsparse_csrsv_clear` x :cpp:func:`rocsparse_Xcsrsv_solve() ` x :cpp:func:`rocsparse_Xcsritsv_buffer_size() ` x :cpp:func:`rocsparse_Xcsritsv_analysis() ` x :cpp:func:`rocsparse_csritsv_zero_pivot` x :cpp:func:`rocsparse_csritsv_clear` x :cpp:func:`rocsparse_Xcsritsv_solve() ` x :cpp:func:`rocsparse_Xcsritsv_solve() ` x :cpp:func:`rocsparse_Xellmv() ` x :cpp:func:`rocsparse_Xgebsrmv() ` x :cpp:func:`rocsparse_Xgemvi_buffer_size() ` x :cpp:func:`rocsparse_Xgemvi() ` x :cpp:func:`rocsparse_Xhybmv() ` x ============================================================================= === == Sparse level 3 functions ------------------------ ========================================================================= === == Function name yes no ========================================================================= === == :cpp:func:`rocsparse_Xcsrmm() ` x :cpp:func:`rocsparse_Xcsrsm_buffer_size() ` x :cpp:func:`rocsparse_Xcsrsm_analysis() ` x :cpp:func:`rocsparse_csrsm_zero_pivot` x :cpp:func:`rocsparse_csrsm_clear` x :cpp:func:`rocsparse_Xcsrsm_solve() ` x :cpp:func:`rocsparse_Xbsrmm() ` x :cpp:func:`rocsparse_Xbsrsm_buffer_size() ` x :cpp:func:`rocsparse_Xbsrsm_analysis() ` x :cpp:func:`rocsparse_bsrsm_zero_pivot` x :cpp:func:`rocsparse_bsrsm_clear` x :cpp:func:`rocsparse_Xbsrsm_solve() ` x :cpp:func:`rocsparse_Xgebsrmm() ` x :cpp:func:`rocsparse_Xgemmi() ` x ========================================================================= === == Sparse extra functions ---------------------- ============================================================================= === == Function name yes no ============================================================================= === == :cpp:func:`rocsparse_bsrgeam_nnzb` x :cpp:func:`rocsparse_Xbsrgeam() ` x :cpp:func:`rocsparse_Xbsrgemm_buffer_size() ` x :cpp:func:`rocsparse_bsrgemm_nnzb` x :cpp:func:`rocsparse_Xbsrgemm() ` x :cpp:func:`rocsparse_csrgeam_nnz` x :cpp:func:`rocsparse_Xcsrgeam() ` x :cpp:func:`rocsparse_Xcsrgemm_buffer_size() ` x :cpp:func:`rocsparse_csrgemm_nnz` x :cpp:func:`rocsparse_csrgemm_symbolic` x :cpp:func:`rocsparse_Xcsrgemm() ` x :cpp:func:`rocsparse_Xcsrgemm_numeric() ` x ============================================================================= === == Preconditioner functions ------------------------ ===================================================================================================================== === == Function name yes no ===================================================================================================================== === == :cpp:func:`rocsparse_Xbsric0_buffer_size() ` x :cpp:func:`rocsparse_Xbsric0_analysis() ` x :cpp:func:`rocsparse_bsric0_zero_pivot` x :cpp:func:`rocsparse_bsric0_clear` x :cpp:func:`rocsparse_Xbsric0() ` x :cpp:func:`rocsparse_Xbsrilu0_buffer_size() ` x :cpp:func:`rocsparse_Xbsrilu0_analysis() ` x :cpp:func:`rocsparse_bsrilu0_zero_pivot` x :cpp:func:`rocsparse_Xbsrilu0_numeric_boost() ` x :cpp:func:`rocsparse_bsrilu0_clear` x :cpp:func:`rocsparse_Xbsrilu0() ` x :cpp:func:`rocsparse_Xcsric0_buffer_size() ` x :cpp:func:`rocsparse_Xcsric0_analysis() ` x :cpp:func:`rocsparse_csric0_zero_pivot` x :cpp:func:`rocsparse_csric0_clear` x :cpp:func:`rocsparse_Xcsric0() ` x :cpp:func:`rocsparse_Xcsrilu0_buffer_size() ` x :cpp:func:`rocsparse_Xcsrilu0_numeric_boost() ` x :cpp:func:`rocsparse_Xcsrilu0_analysis() ` x :cpp:func:`rocsparse_csrilu0_zero_pivot` x :cpp:func:`rocsparse_csrilu0_clear` x :cpp:func:`rocsparse_Xcsrilu0() ` x :cpp:func:`rocsparse_csritilu0_buffer_size` x :cpp:func:`rocsparse_csritilu0_preprocess` x :cpp:func:`rocsparse_Xcsritilu0_compute() ` x :cpp:func:`rocsparse_Xcsritilu0_compute_ex() ` x :cpp:func:`rocsparse_Xcsritilu0_history() ` x :cpp:func:`rocsparse_Xgtsv_buffer_size() ` x :cpp:func:`rocsparse_Xgtsv() ` x :cpp:func:`rocsparse_Xgtsv_no_pivot_buffer_size() ` x :cpp:func:`rocsparse_Xgtsv_no_pivot() ` x :cpp:func:`rocsparse_Xgtsv_no_pivot_strided_batch_buffer_size() ` x :cpp:func:`rocsparse_Xgtsv_no_pivot_strided_batch() ` x :cpp:func:`rocsparse_Xgtsv_interleaved_batch_buffer_size() ` x :cpp:func:`rocsparse_Xgtsv_interleaved_batch() ` x :cpp:func:`rocsparse_Xgpsv_interleaved_batch_buffer_size() ` x :cpp:func:`rocsparse_Xgpsv_interleaved_batch() ` x ===================================================================================================================== === == Conversion functions -------------------- ========================================================================================================================= === == Function name yes no ========================================================================================================================= === == :cpp:func:`rocsparse_csr2coo` x :cpp:func:`rocsparse_csr2csc_buffer_size` x :cpp:func:`rocsparse_Xcsr2csc() ` x :cpp:func:`rocsparse_Xgebsr2gebsc_buffer_size() ` x :cpp:func:`rocsparse_Xgebsr2gebsc() ` x :cpp:func:`rocsparse_csr2ell_width` x :cpp:func:`rocsparse_Xcsr2ell() ` x :cpp:func:`rocsparse_Xcsr2hyb() ` x :cpp:func:`rocsparse_csr2bsr_nnz` x :cpp:func:`rocsparse_Xcsr2bsr() ` x :cpp:func:`rocsparse_csr2gebsr_nnz` x :cpp:func:`rocsparse_Xcsr2gebsr_buffer_size() ` x :cpp:func:`rocsparse_Xcsr2gebsr() ` x :cpp:func:`rocsparse_coo2csr` x :cpp:func:`rocsparse_ell2csr_nnz` x :cpp:func:`rocsparse_Xell2csr() ` x :cpp:func:`rocsparse_hyb2csr_buffer_size` x :cpp:func:`rocsparse_Xhyb2csr() ` x :cpp:func:`rocsparse_Xbsr2csr() ` x :cpp:func:`rocsparse_Xgebsr2csr() ` x :cpp:func:`rocsparse_Xgebsr2gebsr_buffer_size() ` x :cpp:func:`rocsparse_gebsr2gebsr_nnz()` x :cpp:func:`rocsparse_Xgebsr2gebsr() ` x :cpp:func:`rocsparse_Xcsr2csr_compress() ` x :cpp:func:`rocsparse_create_identity_permutation` x :cpp:func:`rocsparse_inverse_permutation` x :cpp:func:`rocsparse_cscsort_buffer_size` x :cpp:func:`rocsparse_cscsort` x :cpp:func:`rocsparse_csrsort_buffer_size` x :cpp:func:`rocsparse_csrsort` x :cpp:func:`rocsparse_coosort_buffer_size` x :cpp:func:`rocsparse_coosort_by_row` x :cpp:func:`rocsparse_coosort_by_column` x :cpp:func:`rocsparse_Xdense2csr() ` x :cpp:func:`rocsparse_Xdense2csc() ` x :cpp:func:`rocsparse_Xdense2coo() ` x :cpp:func:`rocsparse_Xcsr2dense() ` x :cpp:func:`rocsparse_Xcsc2dense() ` x :cpp:func:`rocsparse_Xcoo2dense() ` x :cpp:func:`rocsparse_Xnnz_compress() ` x :cpp:func:`rocsparse_Xnnz() ` x :cpp:func:`rocsparse_Xprune_dense2csr_buffer_size() ` x :cpp:func:`rocsparse_Xprune_dense2csr_nnz() ` x :cpp:func:`rocsparse_Xprune_dense2csr() ` x :cpp:func:`rocsparse_Xprune_csr2csr_buffer_size() ` x :cpp:func:`rocsparse_Xprune_csr2csr_nnz() ` x :cpp:func:`rocsparse_Xprune_csr2csr() ` x :cpp:func:`rocsparse_Xprune_dense2csr_by_percentage_buffer_size() ` x :cpp:func:`rocsparse_Xprune_dense2csr_nnz_by_percentage() ` x :cpp:func:`rocsparse_Xprune_dense2csr_by_percentage() ` x :cpp:func:`rocsparse_Xprune_csr2csr_by_percentage_buffer_size() ` x :cpp:func:`rocsparse_Xprune_csr2csr_nnz_by_percentage() ` x :cpp:func:`rocsparse_Xprune_csr2csr_by_percentage() ` x :cpp:func:`rocsparse_Xbsrpad_value() ` x ========================================================================================================================= === == Reordering functions -------------------- ======================================================= === == Function name yes no ======================================================= === == :cpp:func:`rocsparse_Xcsrcolor() ` x ======================================================= === == Utility functions ----------------- =================================================================================================== === == Function name yes no =================================================================================================== === == :cpp:func:`rocsparse_Xcheck_matrix_csr_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_csr() ` x :cpp:func:`rocsparse_Xcheck_matrix_csc_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_csc() ` x :cpp:func:`rocsparse_Xcheck_matrix_coo_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_coo() ` x :cpp:func:`rocsparse_Xcheck_matrix_gebsr_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_gebsr() ` x :cpp:func:`rocsparse_Xcheck_matrix_gebsc_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_gebsc() ` x :cpp:func:`rocsparse_Xcheck_matrix_ell_buffer_size() ` x :cpp:func:`rocsparse_Xcheck_matrix_ell() ` x :cpp:func:`rocsparse_check_matrix_hyb_buffer_size() ` x :cpp:func:`rocsparse_check_matrix_hyb() ` x =================================================================================================== === == Sparse generic functions ------------------------ ==================================================== === == Function name yes no ==================================================== === == :cpp:func:`rocsparse_axpby()` x :cpp:func:`rocsparse_gather()` x :cpp:func:`rocsparse_scatter()` x :cpp:func:`rocsparse_rot()` x :cpp:func:`rocsparse_spvv()` x :cpp:func:`rocsparse_sparse_to_dense()` x :cpp:func:`rocsparse_dense_to_sparse()` x :cpp:func:`rocsparse_spgemm()` x :cpp:func:`rocsparse_v2_spmv_buffer_size()` x :cpp:func:`rocsparse_spgeam_buffer_size()` x :cpp:func:`rocsparse_spgeam()` x :cpp:func:`rocsparse_sptrsv_buffer_size()` x :cpp:func:`rocsparse_sptrsm_buffer_size()` x :cpp:func:`rocsparse_spilu0_buffer_size()` x :cpp:func:`rocsparse_spic0_buffer_size()` x :cpp:func:`rocsparse_spildlt0_buffer_size()` x :cpp:func:`rocsparse_sddmm_buffer_size()` x :cpp:func:`rocsparse_sddmm_preprocess()` x :cpp:func:`rocsparse_sparse_to_sparse_buffer_size()` x :cpp:func:`rocsparse_sparse_to_sparse()` x :cpp:func:`rocsparse_extract_buffer_size()` x :cpp:func:`rocsparse_extract_nnz()` x :cpp:func:`rocsparse_extract()` x ==================================================== === == For :cpp:func:`rocsparse_spmv()`, :cpp:func:`rocsparse_spmm()`, :cpp:func:`rocsparse_spsv()`, and :cpp:func:`rocsparse_spsm()`, ``hipGraph`` is supported when passing the buffer size or compute stages but is not supported when passing the preprocess stage. For :cpp:func:`rocsparse_v2_spmv()`, :cpp:func:`rocsparse_sptrsv()`, :cpp:func:`rocsparse_sptrsm()`, :cpp:func:`rocsparse_spic0()`, :cpp:func:`rocsparse_spildlt0()`, and :cpp:func:`rocsparse_spilu0()`, ``hipGraph`` is supported when passing the compute stage but is not supported when passing the analysis stage. For :cpp:func:`rocsparse_sddmm()`, ``hipGraph`` is supported only when using the default algorithm. Storage formats =============== For information on the rocSPARSE sparse array storage formats, see :doc:`storage formats <../conceptual/storage-formats-sparse>`. Pointer mode ============ The auxiliary functions :cpp:func:`rocsparse_set_pointer_mode` and :cpp:func:`rocsparse_get_pointer_mode` are used to set and get the value of the state variable :cpp:enum:`rocsparse_pointer_mode`. If :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_host`, then scalar parameters must be allocated on the host. If :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_device`, then scalar parameters must be allocated on the device. There are two types of scalar parameter: #. Scaling parameters, such as ``alpha`` and ``beta``, used, for example, in :cpp:func:`rocsparse_scsrmv` and :cpp:func:`rocsparse_scoomv`. #. Scalar results from functions such as :cpp:func:`rocsparse_sdoti` or :cpp:func:`rocsparse_cdotci`. For scalar parameters such as ``alpha`` and ``beta``, memory can be allocated on the host heap or stack when :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_host`. The kernel launch is asynchronous, and if the scalar parameter is on the heap, it can be freed after the kernel launch returns. When :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_device`, the scalar parameter must not be changed until the kernel completes. For scalar results, when :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_host`, the function blocks the CPU until the GPU has copied the result back to the host. When :cpp:enum:`rocsparse_pointer_mode` is equal to :cpp:enumerator:`rocsparse_pointer_mode_device`, the function returns after the asynchronous launch. Similar to the vector and matrix results, the scalar result is only available when the kernel has completed execution. .. _rocsparse_logging: Activity logging [Deprecated] ============================= Four different environment variables can be set to enable logging in rocSPARSE: ``ROCSPARSE_LAYER``, ``ROCSPARSE_LOG_TRACE_PATH``, ``ROCSPARSE_LOG_BENCH_PATH``, and ``ROCSPARSE_LOG_DEBUG_PATH``. ``ROCSPARSE_LAYER`` is a bit mask that enables logging, where several logging modes for :ref:`rocsparse_layer_mode_` can be specified as follows: ================================ ============================================================== ``ROCSPARSE_LAYER`` not set Logging is disabled. ``ROCSPARSE_LAYER`` set to ``1`` Trace logging is enabled. ``ROCSPARSE_LAYER`` set to ``2`` Bench logging is enabled. ``ROCSPARSE_LAYER`` set to ``3`` Trace logging and bench logging are enabled. ``ROCSPARSE_LAYER`` set to ``4`` Debug logging is enabled. ``ROCSPARSE_LAYER`` set to ``5`` Trace logging and debug logging are enabled. ``ROCSPARSE_LAYER`` set to ``6`` Bench logging and debug logging are enabled. ``ROCSPARSE_LAYER`` set to ``7`` Trace logging and bench logging and debug logging are enabled. ================================ ============================================================== When logging is enabled, each rocSPARSE function call writes the function name and function arguments to the logging stream. The default logging output is streamed to ``stderr``. .. note:: Performance will degrade when logging is enabled. By default, the environment variable ``ROCSPARSE_LAYER`` is not set and logging is disabled. To capture activity logging in a file, set the following environment variables as required: * ``ROCSPARSE_LOG_TRACE_PATH`` specifies a path and file name to capture trace logging streamed to that file. * ``ROCSPARSE_LOG_BENCH_PATH`` specifies a path and file name to capture bench logging. * ``ROCSPARSE_LOG_DEBUG_PATH`` specifies a path and file name to capture debug logging. .. note:: If the file cannot be opened, the logging output is streamed to ``stderr``. .. warning:: Trace, debug, and bench logging is deprecated and will be removed in a future release ROC-TX support in rocSPARSE ============================ The `ROC-TX `_ library contains application code instrumentation APIs to support the high-level correlation of runtime API or activity events. When integrated with rocSPARSE, ROC-TX enables users to view the call stack of rocSPARSE and HIP API functions in profiling tools such as :doc:`rocProfiler `, offering better insights into runtime behavior and performance bottlenecks. To enable ROC-TX profiling, set the environment variable ``ROCSPARSE_ROCTX=1`` when running the program with rocProf: .. code-block:: shell ROCSPARSE_ROCTX=1 /opt/rocm/bin/rocprofv3 --kernel-trace --marker-trace --hip-trace --output-format pftrace -- ./example_program This will generate a ``.pftrace`` file which can then be viewed using the `Perfetto UI `_. .. note:: ROC-TX support in rocSPARSE is unavailable on Windows and is not supported in the static library version on Linux. rocSPARSE Fortran bindings ============================== Fortran functionality for rocSPARSE is offered by :doc:`hipFORT `. For more information, consult the :doc:`hipFORT API file list `. hipSPARSE ========= :doc:`hipSPARSE ` is a SPARSE marshalling library with multiple supported backends. It sits between the application and a "worker" SPARSE library, marshalling inputs into the backend library and marshalling results back to the application. hipSPARSE exports an interface that does not require the client to change, regardless of the chosen backend. hipSPARSE supports rocSPARSE and NVIDIA CUDA cuSPARSE as backends. hipSPARSE focuses on convenience and portability. If performance outweighs these factors, then it's best to use rocSPARSE itself. hipSPARSE can be found on `GitHub `_.