HP MLIB User's Guide Vol. 1 7th Ed.

364 HP MLIB User’s Guide

F_SGEMM/F_DGEMM/F_CGEMM/F_ZGEMM General matrix-matrix multiply

BLAS_TRANS k-by-m matrix A

BLAS_CONJ_TRANS k-by-m matrix A*

LDA Leading dimension of array A.

Error conditions for lda depend on the value of transa.

Each of the following conditions generates an error ﬂag

that is passed to the error handler:

• lda < 1

• TRANSA = BLAS_NO_TRANS and lda < m

• TRANSA = BLAS_TRANS and lda < k

• TRANSA = BLAS_CONJ_TRANS and lda < k

B Array containing the matrix B, whose size is indicated

by TRANSB:

BLAS_NO_TRANS k-by-n matrix B

BLAS_TRANS n-by-k matrix B

BLAS_CONJ_TRANS n-by-k matrix B*

LDB Leading dimension of array B.

Error conditions for ldb depend on the value of transb.

Each of the following conditions generates an error ﬂag

that is passed to the error handler:

• ldb < 1

• TRANSB = BLAS_NO_TRANS and ldb < k

• TRANSB = BLAS_TRANS and ldb < n

• TRANSB = BLAS_CONJ_TRANS and ldb < n

BETA The scalar BETA.

C The m-by-n matrix operand C. The representation of

the matrix entry c

i,j

in C is denoted by C(i, j) for all (i, j)

in the interval [0... m - 1] x [0... n - 1].

LDC Leading dimension of array C. If ldc < 1 or ldc < m, an

error ﬂag is generated and passed to the error handler.

Output C The updated m-by-n matrix C replaces the input.

C αop A()op B() βC+←