dgmm_batch¶

Computes a group of (diagonal matrix-matrix product (dgmm) operations.

Description¶

The dgmm_batch routines perform multiple diagonal matrix-matrix product (dgmm) operations in a single call. The diagonal matrices are stored as dense vectors and the operations are performed with groups of matrices and vectors.

dgmm_batch supports the following precisions:

T
`float`
`double`
`std::complex<float>`
`std::complex<double>`

dgmm_batch (Buffer Version)¶

Buffer version of dgmm_batch supports only strided API.

Strided API¶

Strided API operation is defined as:

for i = 0 … batch_size – 1
    A and C are matrices at offset i * stridea in a, i * stridec in c.
    X is a vector at offset i * stridex in x
    if (left_right == side::left)
        C = diag(X) * A
    else
        C = A * diag(X)
end for

where:

A is a matrix
X is a diagonal matrix stored as a vector

For strided API, all matrices A and C and vector X have the same parameters (size, increments) and are stored at a constant stride given by stridea, stridec and stridex from each other.

The a and x buffers contain all the input matrices. Total number of matrices in a and x are given by batch_size parameter.

Syntax¶

namespace oneapi::mkl::blas::column_major {
    void dgmm_batch(sycl::queue &queue,
                    oneapi::mkl::side left_right,
                    std::inte64_t m,
                    std::int64_t n,
                    sycl::buffer<T,1> &a,
                    std::int64_t lda,
                    std::int64_t stridea,
                    sycl::buffer<T,1> &x,
                    std::int64_t incx,
                    std::int64_t stridex,
                    sycl::buffer<T,1> &c,
                    std::int64_t ldc,
                    std::int64_t stridec,
                    std::int64_t batch_size);
}

namespace oneapi::mkl::blas::row_major {
    void dgmm_batch(sycl::queue &queue,
                    oneapi::mkl::side left_right,
                    std::inte64_t m,
                    std::int64_t n,
                    sycl::buffer<T,1> &a,
                    std::int64_t lda,
                    std::int64_t stridea,
                    sycl::buffer<T,1> &x,
                    std::int64_t incx,
                    std::int64_t stridex,
                    sycl::buffer<T,1> &c,
                    std::int64_t ldc,
                    std::int64_t stridec,
                    std::int64_t batch_size);
}

Input Parameters¶

queue: The queue where the routine should be executed.
left_right: Specifies the position of the diagonal matrix in the product. See Data Types for more details.
m: Number of rows of matrix A and matrix C. Must be at least zero.
n: Number of columns of matrix A and matrix C. Must be at least zero.
a: Buffer holding input matrices A. Size of the buffer must be at least lda * k + stridea * (batch_size - 1) where k is n if column major layout or m if row major layout is used.
lda: Leading dimension of matrices A. Must be at least m if column major layout or n if row major layout is used. Must be positive.
stridea: Stride between two consecutive A matrices. Must be at least zero. See Matrix Storage for more details.
x: Buffer holding input matrices X. Size of the buffer must be at least (1 + (len - 1)*abs(incx)) + stridex * (batch_size - 1) where len is n if the diagonal matrix is on the right of the product or m otherwise.
incx: Stride between two consecutive elements of the X vectors.
stridex: Stride between two consecutive X vectors. Must be at least zero. See Matrix Storage for more details.
c: Buffer holding input/output matrices C. Size of the buffer must be at least batch_size * stridec.
ldc: Leading dimension of matrices C. Must be at least m if column major layout or n if row major layout is used. Must be positive.
stridec: Stride between two consecutive C matrices. Must be at least ldc * n if column major layout or ldc * m if row major layout is used. See Matrix Storage for more details.
batch_size: Number of dgmm computations to perform. Must be at least zero.

Output Parameters¶

c: Buffer holding output matrices C overwritten by batch_size dgmm operations.

dgmm_batch (USM Version)¶

USM version of dgmm_batch supports group API and strided API.

Group API¶

Group API operation is defined as:

idx = 0
for i = 0 … group_count – 1
     for j = 0 … group_size – 1
         A and C are matrices at a[idx] and c[idx]
         X is a vector at x[idx]
         if (left_right[idx] == side::left)
             C = diag(X) * A
         else
             C = A * diag(X)
         idx = idx + 1
     end for
end for

where:

A is a matrix
X is a diagonal matrix stored as a vector

For group API, each group contain matrices and vectors with the same parameters (size, increment). The a and x arrays contain the pointers for all the input matrices. Total number of matrices in a and x are given by:

$total\_batch\_count = \sum_{i=0}^{group\_count-1}group\_size[i]$

Syntax¶

namespace oneapi::mkl::blas::column_major {
    sycl::event dgmm_batch(sycl::queue &queue,
                           oneapi::mkl::side *left_right,
                           std::int64_t *m,
                           std::int64_t *n,
                           const T **a,
                           std::int64_t *lda,
                           const T **x,
                           std::int64_t *incx,
                           T **c,
                           std::int64_t *ldc,
                           std::int64_t group_count,
                           std::int64_t *group_size,
                           const std::vector<sycl::event> &dependencies = {})
}

namespace oneapi::mkl::blas::row_major {
    sycl::event dgmm_batch(sycl::queue &queue,
                           oneapi::mkl::side *left_right,
                           std::int64_t *m,
                           std::int64_t *n,
                           const T **a,
                           std::int64_t *lda,
                           const T **x,
                           std::int64_t *incx,
                           T **c,
                           std::int64_t *ldc,
                           std::int64_t group_count,
                           std::int64_t *group_size,
                           const std::vector<sycl::event> &dependencies = {})
}

Input Parameters¶

queue: The queue where the routine should be executed.
left_right: Array of group_count parameters. left_right[i] specifies the position of the diagonal matrix in group i. See Data Types for more details.
m: Array of group_count integers. m[i] specifies number of rows of A for every matrix in group i. All entries must be at least zero.
n: Array of group_count integers. n[i] specifies number of columns of A for every matrix in group i. All entries must be at least zero.
a: Array of pointers to input matrices A with size total_batch_count. Size of the array must be at least lda[i] * n[i] if column major layout or at least lda[i] * m[i] if row major layout is used. See Matrix Storage for more details.
lda: Array of group_count integers. lda[i] specifies the leading dimension of A for every matrix in group i. All entries must be positive and at least m[i] if column major layout or at least n[i] if row major layout is used.
x: Array of pointers to input vectors X with size total_batch_count. Size of the array must be at least (1 + len[i] – 1)*abs(incx[i])) where len[i] is n[i] if diagonal matrix is on the right of the product or m[i] otherwise. See Matrix Storage for more details.
incx: Array of group_count integers. incx[i] specifies the stride of X for every vector in group i. All entries must be positive.
c: Array of pointers to input/output matrices C with size total_batch_count. Size of the array must be least ldc[i] * n[i] if column major layout or at least ldc[i] * m[i] if row major layout is used. See Matrix Storage for more details.
ldc: Array of group_count integers. ldc[i] specifies the leading dimension of C for every matrix in group i. All entries must be positive and at least m[i] if column major layout or at least n[i] if row major layout is used.
group_count: Specifies number of groups. Must be at least zero.
group_size: Array of group_count integers. group_size[i] specifies the number of diagonal matrix-matrix product operations in group i. All entries must be at least zero.
dependencies: List of events to wait for before starting computation, if any. If omitted, defaults to no dependencies.

Output Parameters¶

c: Array of pointers to output matrices C overwritten by total_batch_count dgmm operations.

Return Values¶

Output event to wait on to ensure computation is complete.

Strided API¶

Strided API operation is defined as:

for i = 0 … batch_size – 1
    A and C are matrices at offset i * stridea in a, i * stridec in c.
    X is a vector at offset i * stridex in x
    if (left_right == side::left)
        C = diag(X) * A
    else
        C = A * diag(X)
end for

where:

A is a matrix
X is a diagonal matrix stored as a vector

For strided API, all matrices A and C and vector X have the same parameters (size, increments) and are stored at a constant stride given by stridea, stridec and stridex from each other.

The a and x buffers contain all the input matrices. Total number of matrices in a and x are given by batch_size parameter.

Syntax¶

namespace oneapi::mkl::blas::column_major {
    sycl::event dgmm_batch(sycl::queue &queue,
                           oneapi::mkl::side left_right,
                           std::inte64_t m,
                           std::int64_t n,
                           const T *a,
                           std::int64_t lda,
                           std::int64_t stridea,
                           const T *x,
                           std::int64_t incx,
                           std::int64_t stridex,
                           T *c,
                           std::int64_t ldc,
                           std::int64_t stridec,
                           std::int64_t batch_size,
                           const std::vector<sycl::event> &dependencies = {})
}

namespace oneapi::mkl::blas::row_major {
    sycl::event dgmm_batch(sycl::queue &queue,
                           oneapi::mkl::side left_right,
                           std::inte64_t m,
                           std::int64_t n,
                           const T *a,
                           std::int64_t lda,
                           std::int64_t stridea,
                           const T *x,
                           std::int64_t incx,
                           std::int64_t stridex,
                           T *c,
                           std::int64_t ldc,
                           std::int64_t stridec,
                           std::int64_t batch_size,
                           const std::vector<sycl::event> &dependencies = {})
}

Input Parameters¶

queue: The queue where the routine should be executed.
left_right: Specifies the position of the diagonal matrix in the product. See Data Types for more details.
m: Number of rows of matrix A and matrix C. Must be at least zero.
n: Number of columns of matrix A and matrix C. Must be at least zero.
a: Pointer to input matrices A. Size of the array must be at least lda * k + stridea * (batch_size - 1) where k is n if column major layout or m if row major layout is used.
lda: Leading dimension of matrices A. Must be at least m if column major layout or n if row major layout is used. Must be positive.
stridea: Stride between two consecutive A matrices. Must be at least zero. See Matrix Storage for more details.
x: Pointer to input matrices X. Size of the array must be at least (1 + (len - 1)*abs(incx)) + stridex * (batch_size - 1) where len is n if the diagonal matrix is on the right of the product or m otherwise.
incx: Stride between two consecutive elements of the X vectors.
stridex: Stride between two consecutive X vectors. Must be at least zero. See Matrix Storage for more details.
c: Pointer to input/output matrices C. Size of the array must be at least batch_size * stridec.
ldc: Leading dimension of matrices C. Must be at least m if column major layout or n if row major layout is used. Must be positive.
stridec: Stride between two consecutive C matrices. Must be at least ldc * n if column major layout or ldc * m if row major layout is used. See Matrix Storage for more details.
batch_size: Number of dgmm computations to perform. Must be at least zero.
dependencies: List of events to wait for before starting computation, if any. If omitted, defaults to no dependencies.

Output Parameters¶

c: Pointer to output matrices C overwritten by batch_size dgmm operations.

Return Values¶

Output event to wait on to ensure computation is complete.

copy_batch gemm_batch

oneAPI Math Kernel Library (oneMKL) documentation

dgmm_batch¶

Description¶

dgmm_batch (Buffer Version)¶

Strided API¶

Syntax¶

Input Parameters¶

Output Parameters¶

dgmm_batch (USM Version)¶

Group API¶

Syntax¶

Input Parameters¶

Output Parameters¶

Return Values¶

Strided API¶

Syntax¶

Input Parameters¶

Output Parameters¶

Return Values¶