Modelsim crash (code 211) when using library

I am using a library prepared for testbenching a future core by a core
vendor.  In some simulations, the library works fine, but in others it
causes Modelsim to crash during a vsim.  When it crashes, it gives a code of
211 which is a segmentation fault.  The crash is total - Modelsim exits
immediately, printing a stack trace to its stderr, but nothing to its main
console.

I believe the crashes may have something to do with how large the simulation
is - simulation in a small design doesn't crash, but in a large design it
crashes.  Strangely enough, in one case the determining factor was how I
vmap'd a library - if I had the complete absolute library path,  it would
crash, but if I gave it a relative path, it ran.  I was later able to get
this same testbench running with an absolute library path after eliminating
two libraries I realized weren't being used (which is in part why I believe
that design size is a factor).

Modelsim support wasn't much help (as usual).  Any suggestions for things
that I might try?  Will upgrading the PC (more memory) help the situation?
Are there Modelsim settings that would help?  I can also request that the
core vendor recompile the core with certain options if that is advisable
(although they will want to use the nodebug option to protect their IP).

Thanks,
Doug

Doug Miller wrote:

> I am using a library prepared for testbenching a future core by a core
> vendor.  In some simulations, the library works fine, but in others it
> causes Modelsim to crash during a vsim.  When it crashes, it gives a code
> of
> 211 which is a segmentation fault.  The crash is total - Modelsim exits
> immediately, printing a stack trace to its stderr, but nothing to its main
> console.

Make sure you are using exactly the same revision of modelsim
and OS to use the library that the vendor is using to compile it.

> Any suggestions for things
> that I might try?

If this is a commercial application consider
buying source code (maybe from another vendor)
or writing your own.
The inability to view, edit and trace code may
hamper development and maintenance of your
fpga image.

> Will upgrading the PC (more memory) help the situation? 

More memory might speed up a working sim,
but it is unlikely to fix a crashing one.
The crash is most likely due to a bug
in the model. It could be a windows or modelsim bug.

> Are there Modelsim settings that would help?  I can also request that the
> core vendor recompile the core with certain options if that is advisable
> (although they will want to use the nodebug option to protect their IP).

Send all your files and a crash recipe to the core vendor.
If the vendor needs your business he will make it work for you.

 -- Mike Treseler