User guide

2-26
Modeling Your Design
In simulations, this memory model used 4 MB of machine memory
with the /*sparse*/ pragma, 81 MB without it. There is a small
runtime performance cost to sparse memory models: the simulation
of the memory with the /*sparse*/ pragma took 64 seconds, 56
seconds without it.
The larger the memory, and the fewer elements in the memory that
your design reads or writes to, the more machine memory you will
save by using this feature. It is intended for memories that contain at
least a few MBs. If your design accesses 1% of its elements you could
save 97% of machine memory. If your design accesses 50% of its
elements, you save 25% of machine memory. Don’t use this feature
if your design accesses more than 50% of its elements because using
the feature in these cases may lead to more memory consumption
than not using it.
Using sparse memory models does not increase the memory size
limits described in the previous section.
Note:
• Sparse memory models cannot be manipulated by PLI
applications through tf calls (the tf_nodeinfo routine issues a
warning for sparse memory and returns NULL for the memory
handle).
• Sparse memory models cannot be used as a personality matrix
in PLA system tasks.