efficiency change by dynamic allocation in microprocessor

Question>
Hi,
I should make scientific program which use linea alebra with so many
matrixs.

I would like to know whether it is good to make matrix class using dynamic
memory allocation (new/delete or malloc/free) or using two-D arrays with
fixed (big enough)size?

of course, my focus is efficiency in terms of speed.
How much I would experience speed down if I use matrix class using dynamic
allocation?

If the quantity of speed down is big enough, where can I get some useful
information about matrix computation in embedded system?

c.f.> I already made matrix class using new/delete and am using it for my
project. however I think it does not have enough performance.
It seems to be slower than I thought.
and I am using samsung S3c2410 (ARM920T core with 16K chache and mmu).


This message was sent using the comp.arch.embedded web interface on
www.EmbeddedRelated.com
.