Speed up matrix operation

Hi guys,

i am calculating one resulting matrix by multiplying three matrices in the form

result=M1*M2/M3

All matrices are sparse. The dimensions are

M1: [71560x71560]
M2: [71560x23814]
M3: [23814x23814]

The calculation doesnt finish on my personal computer in a time of 12 hours. On my server it isnt finished after 12 hours too. Are there hints or other functions to improve the speed of calculation?

Regards Marc

[Bruno Luong]

Possibly not without further knowledge about the matrices.

[Marc Gebhardt]

Which kind of information do you want to know?

[madhan ravi]

try

whos M1
whos M2
whos M3

in command window and paste it here

[Marc Gebhardt]

Name          Size                 Bytes  Class     Attributes

M1        71560x71560            4885064  double    sparse    

Name          Size                 Bytes  Class     Attributes

M2       71560x23814            1335512  double    sparse    

Name          Size                    Bytes  Class     Attributes

 M3         23814x23814            1531064008  double    sparse

M2 only contains ones. The other two matrices are filled with doubles.

[Walter Roberson]

It is common for sparse divided by sparse to yield a dense matrix, which might lead to a lot of swapping along the way.

[Bruno Luong]

"Which kind of information do you want to know?"

Sparseness. Real? Complex? Symmetric? Hermitian? Definite Positive? Who knows else?

For sparse matrix, it is essential to have to correct ordering of unknowns/equations to avoid filling issue when doing inversion, if it's possible.