Clarification of the leading dimension in CUBLAS when transposing

c matrix cuda gpgpu cublas
mchen picture mchen · May 4, 2013 · Viewed 9.9k times · Source

For a matrix A, the documentation only states that the corresponding leading dimension parameter lda refers to the:

leading dimension of two-dimensional array used to store the matrix A

Thus I presume this is just the number of rows of A given CUBLAS' column major storage format. However, when we consider Op(A), what does the leading dimension refer to now?

Answer

talonmies picture talonmies · May 4, 2013

Nothing changes. The leading dimension always refers to the length of the first dimension of the array. The data order flags (normal, transpose, conjugate) only indicate to BLAS how the data within the array is stored. They have no effect on the array itself, which is always column major ordered and requires an LDA value for indexing in 2D.

So whether the matrix data is stored in transposed form or not, an m x n array always has LDA>=m.

Related questions

undefined reference to `std::ios_base::Init::Init()'

I write this code to read 3 files, TM is the size of square matrix, LER the No. of rows of an array and from last value define a non-square matrix of (ler/2)*2 Then... the code read a file with some …

Read More
How to map the indexes of a matrix to a 1-dimensional array (C++)?

I have an 8x8 matrix, like this: char matrix[8][8]; Also, I have an array of 64 elements, like this: char array[64]; Then I have drawn the matrix as a table, and filled the cells with numbers, each number being incremented from …

Read More
C libraries for mathematical matrix operations

I know there are some optimized algorithms around for all kind of matrix decompositions (QR decomposition, SVD,...), multiplications and the likes. Yet, I couldn't find a good overview. For C++, there is quite some useful information in this question, but …

Read More