NVIDIA cuDSS

https://docs.nvidia.com/cuda/cudss/
Admin
NVIDIA cuDSS is a GPU-accelerated CUDA library providing high-performance linear solvers for sparse...
Tokens:50,536
Snippets:487
Trust Score:10
Update:3 months ago
Show doc for...
Tokens:
Raw
### Basic Host Timer Implementation

Source: https://docs.nvidia.com/cuda/cudss/performance_code.html

A simple host-side timer function using `gettimeofday` to measure elapsed time in seconds. This is a crude example and may not be suitable for high-precision benchmarking.

```c
static double second(void) {
 struct timeval tv;
 gettimeofday(&tv, NULL);
 return (double)tv.tv_sec + (double)tv.tv_usec / 1000000.0;
}
```

--------------------------------

### Update CSR Matrix Creation for offsetType

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Examples showing the transition from a single indexType to separate offsetType and indexType parameters, including support for mixed-width indexing.

```c
/* Single indexType covers both row offsets and column indices */
cudssMatrixCreateCsr(&A, nrows, ncols, nnz,
                     rowOffsets, /*rowEnd=*/NULL, colIndices, values,
                     CUDA_R_32I,   /* indexType */
                     CUDA_R_64F,   /* valueType */
                     CUDSS_MTYPE_GENERAL, CUDSS_MVIEW_FULL,
                     CUDSS_BASE_ZERO);
```

```c
/* offsetType and indexType are separate; for uniform 32-bit indexing
   both are set to CUDSS_R_32I */
cudssMatrixCreateCsr(&A, nrows, ncols, nnz,
                     rowOffsets, /*rowEnd=*/NULL, colIndices, values,
                     CUDSS_R_32I,  /* offsetType: type of rowOffsets / rowEnd */
                     CUDSS_R_32I,  /* indexType:  type of colIndices */
                     CUDSS_R_64F,  /* valueType */
                     CUDSS_MTYPE_GENERAL, CUDSS_MVIEW_FULL,
                     CUDSS_BASE_ZERO);
```

```c
cudssMatrixCreateCsr(&A, nrows, ncols, nnz,
                     rowOffsets, /*rowEnd=*/NULL, colIndices, values,
                     CUDSS_R_64I,  /* offsetType: 64-bit row offsets */
                     CUDSS_R_32I,  /* indexType:  32-bit column indices */
                     CUDSS_R_64F,  /* valueType */
                     CUDSS_MTYPE_GENERAL, CUDSS_MVIEW_FULL,
                     CUDSS_BASE_ZERO);
```

--------------------------------

### Compile cuDSS Sample Code (Shared Library) on Windows

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Compile a C++ sample file using nvcc, linking against cuDSS and cuBLAS as shared libraries. Ensure cuDSS and CUDA Toolkit are installed and their paths are correctly set.

```bash
nvcc.exe cudss_simple.cpp -I "%CUDSS_DIR%\include" -lcudss -lcublas -o cudss_simple.exe
```

--------------------------------

### cudssMatrixGetFormat

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the format of a matrix.

```APIDOC
## cudssMatrixGetFormat

### Description
Gets the format of a matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetFormat(
    cudssMatrix_t matrix,
    cudssMatrixFormat_t *format
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `format` (*cudssMatrixFormat_t *)

```

--------------------------------

### Compile cuDSS Sample with Shared Library

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Compile a C++ sample using nvcc, linking against the shared cuDSS and cuBLAS libraries. Ensure cuDSS and CUDA Toolkit are installed and their paths are correctly set.

```bash
nvcc cudss_simple.cpp -I${CUDSS_DIR}/include -L${CUDSS_DIR}/lib -lcudss -lcublas -o cudss_simple
```

--------------------------------

### cudssMatrixGetDn

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the dense data for a matrix.

```APIDOC
## cudssMatrixGetDn

### Description
Gets the dense data for a matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetDn(
    cudssMatrix_t matrix,
    int *rows,
    int *cols,
    void **values
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `rows` (*int*)
- `cols` (*int*)
- `values` (*void **)
```

--------------------------------

### cudssMatrixGetBatchDn

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the dense data for a batch matrix.

```APIDOC
## cudssMatrixGetBatchDn

### Description
Gets the dense data for a batch matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetBatchDn(
    cudssMatrix_t matrix,
    int *rows,
    int *cols,
    int *batch_size,
    void **values
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `rows` (*int*)
- `cols` (*int*)
- `batch_size` (*int*)
- `values` (*void **)
```

--------------------------------

### Check Threading Layer in cuDSS MT Mode

Source: https://docs.nvidia.com/cuda/cudss/tips_and_tricks.html

This example demonstrates how to check the threading layer for cuDSS MT mode. It is useful when dealing with multi-threaded issues.

```c++
#include <omp.h>
#include <stdio.h>

int main() {
    int num_threads = omp_get_num_threads();
    printf("Number of threads: %d\n", num_threads);
    return 0;
}
```

--------------------------------

### Basic cuDSS Workflow for Solving Sparse Linear Systems

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Demonstrates the fundamental steps to solve a sparse linear system using cuDSS. This includes initialization, configuration, analysis, factorization, and solving phases. Ensure necessary headers and data structures are defined.

```c++
#include <cudss.h> // cuDSS header

// Device pointers and scalar shape parameters, matrix properties
int* rowOffsets = ...
int* colIndices = ...
double* values = ...
double* bvalues = ...
double* xvalues = ...

//---------------------------------------------------------------------------------
// cuDSS data structures and handle initialization
cudssHandle_t handle;
cudssConfig_t config;
cudssData_t data;
cudssMatrix_t A;
cudssMatrix_t b;
cudssMatrix_t x;

cudssCreate(&handle);

cudssConfigCreate(&config);
cudssDataCreate(handle, &data);
cudssMatrixCreateCsr(&A, ... rowOffsets, colIndices, values, ...);
cudssMatrixCreateDn(&b, ... bvalues, ...);
cudssMatrixCreateDn(&x, ... xvalues, ...);

//---------------------------------------------------------------------------------
// (optional) Modifying solver settings, e.g., reordering algorithm
cudssAlgType_t reorder_alg = CUDSS_ALG_DEFAULT;
cudssConfigSet(config, CUDSS_REORDERING_ALG, &reorder_alg, sizeof(cudssAlgType_t));

//---------------------------------------------------------------------------------
// Reordering & symbolic factorization
cudssExecute(handle, CUDSS_PHASE_ANALYSIS, config, data, A, x, b);

//---------------------------------------------------------------------------------
// Numerical factorization
cudssExecute(handle, CUDSS_PHASE_FACTORIZATION, config, data, A, x, b);

//---------------------------------------------------------------------------------
// Solving the system
cudssExecute(handle, CUDSS_PHASE_SOLVE, config, data, A, x, b);

//---------------------------------------------------------------------------------
// (optional) Extra data can be retrieved from the cudssData_t object
// For example, diagonal of the factorized matrix or the reordering permutation

//---------------------------------------------------------------------------------
// Destroy the opaque objects
cudssConfigDestroy(config);
cudssDataDestroy(handle, data);
cudssMatrixDestroy(A);
cudssMatrixDestroy(x);
cudssMatrixDestroy(b);
cudssDestroy(handle);

//---------------------------------------------------------------------------------
// The solution of the system can now be accessed via the user-allocated device
// pointer xvalues

// ...

```

--------------------------------

### cudssMatrixGetCsr

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the CSR data for a matrix.

```APIDOC
## cudssMatrixGetCsr

### Description
Gets the CSR data for a matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetCsr(
    cudssMatrix_t matrix,
    int *rows,
    int *cols,
    void **values,
    int **row_ptr,
    int **col_ind
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `rows` (*int*)
- `cols` (*int*)
- `values` (*void **)
- `row_ptr` (*int **)
- `col_ind` (*int **)
```

--------------------------------

### Compile cuDSS Sample with Static Library

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Compile a C++ sample linking against static versions of cuDSS and cuBLAS libraries. This requires specifying the full path to the static library files using -Xlinker flags.

```bash
nvcc cudss_simple.cpp -I${CUDSS_DIR}/include \
 -Xlinker=${CUDSS_DIR}/lib/libcudss_static.a \
 -Xlinker=${CTK_DIR}/lib64/libcublas_static.a \
 -Xlinker=${CTK_DIR}/lib64/libcublasLt_static.a \
 -Xlinker=${CTK_DIR}/lib64/libculibos.a \
 -o cudss_simple_static
```

--------------------------------

### cudssConfigCreate

Source: https://docs.nvidia.com/cuda/cudss/functions.html

Initializes the cuDSS config object (`cudssConfig_t`) which holds the settings of the solver related to solving a specific linear system. It allocates light resources on the host.

```APIDOC
## cudssConfigCreate

### Description
Initializes the cuDSS config object (`cudssConfig_t`) which holds the settings of the solver related to solving a specific linear system. It allocates light resources on the host.

### Signature
`cudssStatus_t cudssConfigCreate(cudssConfig_t* config)`

### Parameters
*   **config** - [out] [host] cuDSS config object

### Returns
*   [out] The error status of the invocation. Must return `CUDSS_STATUS_SUCCESS` on success and some other status code otherwise.

### Notes
To release the allocated memory, [`cudssConfigDestroy()`](https://docs.nvidia.com/cuda/cudss/functions.html.md#c.cudssConfigDestroy "cudssConfigDestroy") must be called.
```

--------------------------------

### cudssMatrixGetDistributionRow1d

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the 1D distribution of rows for a matrix.

```APIDOC
## cudssMatrixGetDistributionRow1d

### Description
Gets the 1D distribution of rows for a matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetDistributionRow1d(
    cudssMatrix_t matrix,
    int *row_dist
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `row_dist` (*int *)

```

--------------------------------

### Compile cuDSS Sample Code (Static Library) on Windows

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Compile a C++ sample file using nvcc, linking against cuDSS and cuBLAS as static libraries. This requires specific linker flags to ensure all necessary code is included.

```bash
nvcc.exe cudss_simple.cpp -I %CUDSS_DIR%\include \
 -Xlinker=/WHOLEARCHIVE:"%CUDSS_DIR%\lib\cudss.lib" \
 -Xlinker=/WHOLEARCHIVE:"%CTK_DIR%\lib\x64\cublas.lib" \
 -Xlinker=/FORCE -o cudss_simple_static.exe
```

--------------------------------

### cudssMatrixGetBatchCsr

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the CSR data for a batch matrix.

```APIDOC
## cudssMatrixGetBatchCsr

### Description
Gets the CSR data for a batch matrix.

### Function Signature
```c
nvStatus_t cudssMatrixGetBatchCsr(
    cudssMatrix_t matrix,
    int *rows,
    int *cols,
    int *batch_size,
    void **values,
    int **row_ptr,
    int **col_ind
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `rows` (*int*)
- `cols` (*int*)
- `batch_size` (*int*)
- `values` (*void **)
- `row_ptr` (*int **)
- `col_ind` (*int **)
```

--------------------------------

### cudssDistributedInterface_t.cudssCommSize

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the size (number of processes) of a communicator.

```APIDOC
## cudssDistributedInterface_t.cudssCommSize

### Description
Gets the size of a communicator.

### Function Signature
`cudssError_t cudssDistributedInterface_t::cudssCommSize(cudssComm_t comm, int *size)`

### Parameters
*   **comm** (cudssComm_t) - The communicator.
*   **size** (int *) - Pointer to store the size.
```

--------------------------------

### cudssMatrixCreateBatchDn

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates a batch of dense matrices.

```APIDOC
## cudssMatrixCreateBatchDn

### Description
Creates a batch of dense matrices.

### Function Signature
`cudssError_t cudssMatrixCreateBatchDn(cudssHandle_t handle, int n, int m, int batch_size, cudssLayout_t layout, cudssMatrix_t *matrix)`

### Parameters
*   **handle** (cudssHandle_t) - The handle of the CUDSS library.
*   **n** (int) - The number of rows in each matrix.
*   **m** (int) - The number of columns in each matrix.
*   **batch_size** (int) - The number of matrices in the batch.
*   **layout** (cudssLayout_t) - The memory layout (CUDSS_LAYOUT_ROW_MAJOR or CUDSS_LAYOUT_COL_MAJOR).
*   **matrix** (cudssMatrix_t *) - Pointer to store the created dense matrix.
```

--------------------------------

### cudssDistributedInterface_t.cudssCommRank

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Gets the rank of the calling process within a communicator.

```APIDOC
## cudssDistributedInterface_t.cudssCommRank

### Description
Gets the rank of the calling process within a communicator.

### Function Signature
`cudssError_t cudssDistributedInterface_t::cudssCommRank(cudssComm_t comm, int *rank)`

### Parameters
*   **comm** (cudssComm_t) - The communicator.
*   **rank** (int *) - Pointer to store the rank.
```

--------------------------------

### Get Number of Perturbed Pivots with cudssDataGet

Source: https://docs.nvidia.com/cuda/cudss/tips_and_tricks.html

Use cudssDataGet with CUDSS_DATA_NPIVOTS to retrieve the count of perturbed pivots during factorization. This can indicate if a matrix is not well-conditioned.

```c
cudssDataGet(handle, CUDSS_DATA_NPIVOTS, &npivots);
```

--------------------------------

### cudssCreate

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates an instance of the cuDSS library. This is typically the first step before using other cuDSS functionalities.

```APIDOC
## cudssCreate

### Description
Creates an instance of the cuDSS library.

### Function Signature
```c

```

### Parameters
(No specific parameters are detailed in the source for this function.)

### Return Value
(The return value is not explicitly detailed in the source, but likely returns a handle to the cuDSS instance.)

### Usage Example
(No usage example is provided in the source.)
```

--------------------------------

### Enable Pivoting Epsilon Scaling with cudssConfigSet

Source: https://docs.nvidia.com/cuda/cudss/tips_and_tricks.html

Enable automatic scaling of the pivoting epsilon for better accuracy with matrices of varying magnitudes by using cudssConfigSet with CUDSS_CONFIG_PIVOT_EPSILON_ALG.

```c
cudssConfigSet(handle, CUDSS_CONFIG_PIVOT_EPSILON_ALG, &pivot_epsilon_alg);
```

--------------------------------

### Migrate Solve Algorithm Configuration

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Update the solve algorithm configuration to use the new cudssSolveAlg_t enum.

```c
cudssAlgType_t solve_alg = CUDSS_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_SOLVE_ALG, &solve_alg,
               sizeof(cudssAlgType_t));
```

```c
cudssSolveAlg_t solve_alg = CUDSS_SOLVE_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_SOLVE_ALG, &solve_alg,
               sizeof(cudssSolveAlg_t));
```

--------------------------------

### Update Renamed Parameters in cuDSS

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Demonstrates the transition from old parameter names to the updated 0.8.0 naming convention for configuration and data handling.

```c
int hybrid_mode = 1;
cudssConfigSet(config, CUDSS_CONFIG_HYBRID_MODE, &hybrid_mode, sizeof(int));

/* ... */

/* Retrieving the partition tree after analysis */
void *tree_data = NULL;
size_t tree_size = 0;
cudssDataGet(handle, data, CUDSS_DATA_ELIMINATION_TREE,
             &tree_data, sizeof(void*), &tree_size);

/* Supplying the partition tree to a subsequent run */
cudssDataSet(handle, data, CUDSS_DATA_USER_ELIMINATION_TREE,
             tree_data, tree_size);
```

```c
int hybrid_mode = 1;
cudssConfigSet(config, CUDSS_CONFIG_HYBRID_MEMORY_MODE, &hybrid_mode, sizeof(int));

/* ... */

/* Retrieving the partition tree after analysis */
void *tree_data = NULL;
size_t tree_size = 0;
cudssDataGet(handle, data, CUDSS_DATA_ND_PARTITION_TREE,
             &tree_data, sizeof(void*), &tree_size);

/* Supplying the partition tree to a subsequent run */
cudssDataSet(handle, data, CUDSS_DATA_USER_ND_PARTITION_TREE,
             tree_data, tree_size);
```

--------------------------------

### Configure CUDSS Matching (0.8.0+)

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Updated method for managing matching using only the CUDSS_CONFIG_MATCHING_ALG parameter.

```C
/* Enabling matching: choose an algorithm (any value except NONE) */
cudssMatchingAlg_t matching_alg = CUDSS_MATCHING_ALG_AUTO;
cudssConfigSet(config, CUDSS_CONFIG_MATCHING_ALG, &matching_alg,
               sizeof(cudssMatchingAlg_t));

/* Disabling matching (this is the default, but can be set explicitly) */
cudssMatchingAlg_t matching_alg = CUDSS_MATCHING_ALG_NONE;
cudssConfigSet(config, CUDSS_CONFIG_MATCHING_ALG, &matching_alg,
               sizeof(cudssMatchingAlg_t));
```

--------------------------------

### Migrate Pivot Epsilon Algorithm Configuration

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Update the pivot epsilon algorithm configuration to use the new cudssPivotEpsilonAlg_t enum.

```c
cudssAlgType_t pivot_eps_alg = CUDSS_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_PIVOT_EPSILON_ALG, &pivot_eps_alg,
               sizeof(cudssAlgType_t));
```

```c
cudssPivotEpsilonAlg_t pivot_eps_alg = CUDSS_PIVOT_EPSILON_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_PIVOT_EPSILON_ALG, &pivot_eps_alg,
               sizeof(cudssPivotEpsilonAlg_t));
```

--------------------------------

### cudssCreate

Source: https://docs.nvidia.com/cuda/cudss/functions.html

Initializes the cuDSS library handle, which holds the library context. This function must be called before any other cuDSS library calls.

```APIDOC
## cudssCreate

### Description
Initializes the cuDSS library handle (`cudssHandle_t`) which holds the cuDSS library context. It allocates light hardware resources on the host, and must be called prior to making any other cuDSS library calls. The cuDSS library context is tied to the current CUDA device. To use the library on multiple devices, one cuDSS handle should be created for each device.

### Parameters
#### Path Parameters
- **handle** (cudssHandle_t*) - Output - cuDSS library handle

### Returns
- **cudssStatus_t** - The error status of the invocation. Must return `CUDSS_STATUS_SUCCESS` on success and some other status code otherwise.
```

--------------------------------

### Matrix Creation and Manipulation Functions

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Functions for creating, destroying, and modifying sparse and dense matrices in various formats.

```APIDOC
## cudssMatrixCreateBatchCsr

### Description
Creates a batch of CSR matrices.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixCreateBatchDn

### Description
Creates a batch of dense matrices.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixCreateCsr

### Description
Creates a CSR matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixCreateDn

### Description
Creates a dense matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixDestroy

### Description
Destroys a cuDSS matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetBatchCsr

### Description
Retrieves the CSR data from a batch matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetBatchDn

### Description
Retrieves the dense data from a batch matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetCsr

### Description
Retrieves the CSR data from a matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetDistributionRow1d

### Description
Gets the 1D row distribution of a matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetDn

### Description
Retrieves the dense data from a matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixGetFormat

### Description
Gets the format of a cuDSS matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixSetBatchCsrPointers

### Description
Sets the CSR data pointers for a batch matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

```APIDOC
## cudssMatrixSetBatchValues

### Description
Sets the values for a batch matrix.

### Method
C function

### Endpoint
n/a

### Parameters
None explicitly documented.

### Request Example
None provided.

### Response
None explicitly documented.
```

--------------------------------

### Migrate Factorization Algorithm Configuration

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Update the factorization algorithm configuration to use the new cudssFactorizationAlg_t enum.

```c
cudssAlgType_t factor_alg = CUDSS_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_FACTORIZATION_ALG, &factor_alg,
               sizeof(cudssAlgType_t));
```

```c
cudssFactorizationAlg_t factor_alg = CUDSS_FACTORIZATION_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_FACTORIZATION_ALG, &factor_alg,
               sizeof(cudssFactorizationAlg_t));
```

--------------------------------

### Update threading interface implementation

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Comparison of the threading interface structure before and after adding the required property query callback.

```c
cudssThreadingInterface_t cudssThreadingInterface = {
    .cudssGetMaxThreads = my_get_max_threads,
    .cudssParallelFor   = my_parallel_for,
    /* no cudssThreadingGetProperty field */
};
```

```c
cudssThreadingInterface_t cudssThreadingInterface = {
    .cudssGetMaxThreads          = my_get_max_threads,
    .cudssParallelFor            = my_parallel_for,
    /* New: provide a property query callback */
    .cudssThreadingGetProperty   = my_threading_get_property,
};
```

--------------------------------

### Define GetProperty function pointers

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

New function pointer signatures required for custom communication and threading interface structures.

```c
/*  For cudssDistributedInterface_t:  */
int (*cudssDistributedGetProperty)(libraryPropertyType propertyType, int *value);

/*  For cudssThreadingInterface_t:  */
int (*cudssThreadingGetProperty)(libraryPropertyType propertyType, int *value);
```

--------------------------------

### Enable Matching and Scaling with cudssConfigSet

Source: https://docs.nvidia.com/cuda/cudss/tips_and_tricks.html

Improve accuracy by enabling matrix matching and scaling. Use cudssConfigSet with CUDSS_CONFIG_USE_MATCHING and CUDSS_CONFIG_MATCHING_ALG.

```c
cudssConfigSet(handle, CUDSS_CONFIG_USE_MATCHING, &use_matching);
```

```c
cudssConfigSet(handle, CUDSS_CONFIG_MATCHING_ALG, &matching_alg);
```

--------------------------------

### Compile cuDSS Sample with Shared Library (pip wheels)

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

When using pip wheels, replace -lcudss with -l:libcudss.so.0 due to limitations with symlinks in wheels. This command links the shared cuDSS library.

```bash
nvcc cudss_simple.cpp -I${CUDSS_DIR}/include -L${CUDSS_DIR}/lib -l:libcudss.so.0 -lcublas -o cudss_simple
```

--------------------------------

### Update Matrix Creation to cuDSS 0.8.0

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Demonstrates the transition from cudaDataType_t to cudssDataType_t and the inclusion of the new offsetType parameter in CSR matrix creation.

```C
#include <library_types.h> /* or cuda_runtime_api.h */
#include <cudss.h>

cudssMatrix_t A, x, b;

/* Sparse CSR system matrix with double-precision values, int32 indices */
cudssMatrixCreateCsr(&A, nrows, ncols, nnz,
                     rowOffsets, /*rowEnd=*/NULL, colIndices, values,
                     CUDA_R_32I, CUDA_R_64F,
                     CUDSS_MTYPE_GENERAL, CUDSS_MVIEW_FULL,
                     CUDSS_BASE_ZERO);

/* Dense right-hand side and solution */
cudssMatrixCreateDn(&b, nrows, 1, nrows, bvalues, CUDA_R_64F, CUDSS_LAYOUT_COL_MAJOR);
cudssMatrixCreateDn(&x, nrows, 1, nrows, xvalues, CUDA_R_64F, CUDSS_LAYOUT_COL_MAJOR);
```

```C
#include <cudss.h>

cudssMatrix_t A, x, b;

/* Replace CUDA_R_* with CUDSS_R_* equivalents; also add the new offsetType
   argument (here the same as indexType for uniform 32-bit indexing) */
cudssMatrixCreateCsr(&A, nrows, ncols, nnz,
                     rowOffsets, /*rowEnd=*/NULL, colIndices, values,
                     CUDSS_R_32I, CUDSS_R_32I, CUDSS_R_64F,
                     CUDSS_MTYPE_GENERAL, CUDSS_MVIEW_FULL,
                     CUDSS_BASE_ZERO);

cudssMatrixCreateDn(&b, nrows, 1, nrows, bvalues, CUDSS_R_64F, CUDSS_LAYOUT_COL_MAJOR);
cudssMatrixCreateDn(&x, nrows, 1, nrows, xvalues, CUDSS_R_64F, CUDSS_LAYOUT_COL_MAJOR);
```

--------------------------------

### Migrate Reordering Algorithm Configuration

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Update the reordering algorithm configuration to use the new cudssReorderingAlg_t enum.

```c
cudssAlgType_t reorder_alg = CUDSS_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_REORDERING_ALG, &reorder_alg,
               sizeof(cudssAlgType_t));
```

```c
cudssReorderingAlg_t reorder_alg = CUDSS_REORDERING_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_REORDERING_ALG, &reorder_alg,
               sizeof(cudssReorderingAlg_t));
```

--------------------------------

### Configure CUDSS Matching (Pre-0.8.0)

Source: https://docs.nvidia.com/cuda/cudss/migration_guide.html

Legacy method for enabling or disabling matching by setting both a boolean flag and an algorithm type.

```C
/* Enabling matching required setting two parameters */
int use_matching = 1;
cudssAlgType_t matching_alg = CUDSS_ALG_DEFAULT;

cudssConfigSet(config, CUDSS_CONFIG_USE_MATCHING, &use_matching,
               sizeof(int));
cudssConfigSet(config, CUDSS_CONFIG_MATCHING_ALG, &matching_alg,
               sizeof(cudssAlgType_t));

/* Disabling matching */
int use_matching = 0;
cudssConfigSet(config, CUDSS_CONFIG_USE_MATCHING, &use_matching,
               sizeof(int));
```

--------------------------------

### Set cuDSS and CUDA Library Paths

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Update LD_LIBRARY_PATH to include cuDSS and CUDA Toolkit library directories for runtime access. Ensure CUDSS_DIR and CTK_DIR environment variables are set.

```bash
export LD_LIBRARY_PATH=${CUDSS_DIR}/lib:${CTK_DIR}/lib64:${LD_LIBRARY_PATH}
```

--------------------------------

### cudssConfigCreate

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates a configuration object for cuDSS operations. This object holds various settings that control the behavior of cuDSS algorithms.

```APIDOC
## cudssConfigCreate

### Description
Creates a configuration object for cuDSS operations.

### Method
C Function

### Endpoint
n/a

### Parameters
None explicitly documented in the provided text.

### Request Example
```c
// Example usage not provided in source.
```

### Response
None explicitly documented in the provided text.
```

--------------------------------

### cudssMatrixCreateDn

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates a dense matrix descriptor.

```APIDOC
## cudssMatrixCreateDn

### Description
Creates a dense matrix descriptor.

### Function Signature
```c
nvStatus_t cudssMatrixCreateDn(
    cudssMatrix_t *matrix,
    int rows,
    int cols,
    cudssMatrixFormat_t format
);
```

### Parameters
#### Path Parameters
- `matrix` (*cudssMatrix_t*)
- `rows` (*int*)
- `cols` (*int*)
- `format` (*cudssMatrixFormat_t*)
```

--------------------------------

### cudssCreateMg

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates a handle to the cuDSS multi-grid solver. This function is used for initializing the multi-grid solver functionality.

```APIDOC
## cudssCreateMg

### Description
Creates a handle to the cuDSS multi-grid solver.

### Function Signature
```c
void cudssCreateMg(cudssHandle_t *handle)
```

### Parameters
*   **handle** (cudssHandle_t *) - Pointer to a location where the cuDSS multi-grid solver handle will be stored.

### Remarks
This function initializes the multi-grid solver component of the cuDSS library. The returned handle is specific to the multi-grid solver and should be destroyed using `cudssDestroy`.
```

--------------------------------

### cudssMatrixCreateBatchDn

Source: https://docs.nvidia.com/cuda/cudss/functions.html

Creates a matrix object wrapped around a batch of dense matrices. The provided data buffer for the matrix values must contain device-visible pointers to device-visible data.

```APIDOC
## cudssMatrixCreateBatchDn

### Description
Creates a matrix object wrapped around a batch of dense matrices. The provided data buffer for the matrix values must contain device-visible pointers to device-visible data.

Note: cuDSS supports non-uniform batches with varying shapes, e.g `nrows`, `ncols`, `ld` can be different for each batch instance.

Note: MGMN mode does not support matrix batches.

See more limitations for using `cudssMatrix_t` objects in the documentation for the main routine, [`cudssExecute()`](https://docs.nvidia.com/cuda/cudss/functions.html.md#c.cudssExecute "cudssExecute").

### Parameters
*   **matrix** – [out] [host] Created matrix object
*   **batchCount** – [in] [host] Size of the batch
*   **nrows** – [in] [host] Number of rows for each matrix in the batch
*   **ncols** – [in] [host] Number of columns for each matrix in the batch
*   **ld** – [in] [host] Leading dimension for each matrix in the batch
*   **values** – [in] [device] Pointer to values of each dense matrix in the batch
*   **indexType** – [in] [host] Index type for scalar arrays (nrows, ncols, ld)
*   **valueType** – [in] [host] Data type of the matrix
*   **layout** – [in] [host] Memory layout

### Returns
[out] The error status of the invocation. Must return `CUDSS_STATUS_SUCCESS` on success and some other status code otherwise.
```

--------------------------------

### cudssDeviceMemHandler_t.device_alloc

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Allocates memory on the device using the provided handler.

```APIDOC
## cudssDeviceMemHandler_t.device_alloc

### Description
Allocates memory on the device using the provided handler.

### Method
C Function

### Endpoint
N/A

### Parameters
N/A

### Request Example
N/A

### Response
N/A
```

--------------------------------

### cudssCreate

Source: https://docs.nvidia.com/cuda/cudss/genindex.html

Creates a handle to the cuDSS library. This is typically the first function called when using the cuDSS library.

```APIDOC
## cudssCreate

### Description
Creates a handle to the cuDSS library.

### Function Signature
```c
void cudssCreate(cudssHandle_t *handle)
```

### Parameters
*   **handle** (cudssHandle_t *) - Pointer to a location where the cuDSS library handle will be stored.

### Remarks
This function initializes the cuDSS library and returns a handle that must be used in subsequent cuDSS function calls. The handle should be destroyed using `cudssDestroy` when no longer needed.
```

--------------------------------

### Set cuDSS and CUDA Library Paths on Windows

Source: https://docs.nvidia.com/cuda/cudss/getting_started.html

Update the system's PATH environment variable to include the cuDSS and CUDA Toolkit library directories. This is necessary for the compiler and linker to find the required libraries.

```batch
setx PATH "%CUDSS_DIR%\lib:%CTK_DIR%\lib64:%PATH%"
```
Tokens:
Raw