ModelSim 5.5e crashes (read failure / vlm process)
Wolfram Stumpf
I'm simulating a VHDL design (not a netlist) with ModelSim SE 5.5e
(Sun). From time to time, the simulation runs without any problems.
But most of the time, ModelSim crashes and writes the following
message to the shell:
# ERROR: Read failure in vlm process (0,0)
Problem with simulator... vsim U/I closing. (1)
Problem with simulator... vsim U/I closing. (2)
It happens completly indeterministic but only with this testbench /
design combination. But as ModelSim crashes, I can't get any
information about the reason. Has anybody had the same problem?
Regards
Wolfram Stumpf
Narendran K Nathan
gives another error
SIGPIPE not seen!
and exits. Always 25% of my runs used to exit like this. There will be
another 5% due to
licence problems. This is really frustrating!
But I noted one phenomenon. This problem decreases by running the same
simulation in the
same machine continuously. Please try it.
Srinivasan Venkataramanan
There was a post (or a couple of them) in this Group a while back about
"good practices with Modelsim", (Do a search on google groups for "modelsim
crash fli" ), I am "re-using" that post here :-)
HTH,
Srinivasan
---- Old Post(s) -----------
it only seems to happen when the simulation is running
and you try to scroll, cut/paste, etc. in the waves window.
I always try to stop the simulator before working with the waves
window.
===========
However, there are things you can do to reduce the probability of crashing:
- Ensure that your FLI routines are perfectly coded, with no segfaults or
memory leaks or file descriptor leaks. ModelSim is extremely intolerant
of
bad arguments passed to its FLI functions.
- Do not have the signals or structure windows open while the simulator is
running or restarting. It's OK to have the waveform window open. If you
need to use the signals or structure windows, then close them before
rerunning the simulation.
- Do not use mixed Verilog/VHDL simulation.
- Occasionally quit and re-invoke from scratch.
If I heed the above advice, ModelSim will crash very rarely (roughly once
per
24 hours of continuous use - I run jobs overnight).
--------------
--
Srinivasan Venkataramanan
ASIC Design Engineer
Software & Silicon Systems India Pvt Ltd. - an Intel company
Bangalore, India
I don't speak for Intel
"Wolfram Stumpf" <wolfram...@diehl-avionik.de> wrote in message
news:e9445532.02061...@posting.google.com...
Mike Treseler
Wolfram Stumpf wrote:
> # ERROR: Read failure in vlm process (0,0)
>
Utku Ozcan
When I kill the Modelsim processes, the messages above come out.
I think Modelsim quits abnormally because of a reason. This might be
RAM problem or harddisk problem or whatever. Maybe you are running
too many applications on your PC / workstation, and operating system
tries to kill it.
Utku
Christian
I read about you rather disgusting Modelsim Problem.
I encountered similar errors when dragging and dropping signals into wave
window.
That was hardware- and envorironmental independent... on both Unix and
Windows platforms Modelsim crashed even if no single delta was simulated.
I found out that older Modelsim versions like 5.4 don't have this problem.
However, I'd like to continue to use Modelsim 5.5e because it is part of
FPGA-Adv.5.2 which I'm using.
The guys at Modelsim aren't really caring about that, I fear.
so I wonder if other users have probably a solution on that.
regards
Chris
Wolfram Stumpf
set the stack size to 100 MB, which didn't help. It doesn't even use
up all the memory on our server. It seems to me, that ModelSim isn't
programmed very clean because the simulation runs for a few time and
then it crashes. If I do a logout and log in again, it is possible to
run the simulation again until the next crash. Maybe it reserves
memory, which it doesn't release properly. The main thing, that
confuses me is the fact, that I had no problems with Modelsim so far.
I could swear, the whole problem is related to my VHDL testbench. But
Modelsim doesn't raise an error to give me the chance to solve the
problem. Maybe I did the one thing, no one thought about at Modeltech
Company. By the way, I do only use the command line. My testbench
produces nice images to look at. At the moment I don't have to look at
the waveforms. But doing the simulation via command line and have a
look at the waveform afterwards seems to be a good way.
Once again the error message:
Problem with simulator... vsim U/I closing. (2)
Problem with simulator... vsim U/I closing. (1)
# ERROR: Read failure in vlm process (0,0)
Does anyone now what the numbers (2),(1),(0,0) tell me?
Mike Treseler
> Problem with simulator... vsim U/I closing. (2)
> Problem with simulator... vsim U/I closing. (1)
> # ERROR: Read failure in vlm process (0,0)
>
> Does anyone now what the numbers (2),(1),(0,0) tell me?
>
Any chance you have more than one instance of vsim GUI running?
Check by doing a: ps -ef
Try running vsim with a script from the command line.
vsim test_design -do test_design.tcl
Or make a pass/fail testbench that will run
without the GUI
vsim -c test_closed_loop -do "run -all;exit -f"
-- Mike Treseler
Bill Austin
> Does anyone now what the numbers (2),(1),(0,0) tell me?
Have you considered asking this and your other questions directly to
Model Technology Support,
http://www.model.com/support/email_support.asp?
Kim Enkovaara
Wolfram Stumpf wrote:
> Thanks for the answers so far! But none of them solved the problem. I
> set the stack size to 100 MB, which didn't help. It doesn't even use
> up all the memory on our server. It seems to me, that ModelSim isn't
> programmed very clean because the simulation runs for a few time and
> then it crashes. If I do a logout and log in again, it is possible to
Usually when I see problems like that they are on the FLI/PLI
libraries that corrupt the simulator memory. Fortunately the problems
are easy to pinpoint from the stack traces in crash.
> run the simulation again until the next crash. Maybe it reserves
> memory, which it doesn't release properly. The main thing, that
> confuses me is the fact, that I had no problems with Modelsim so far.
> I could swear, the whole problem is related to my VHDL testbench. But
Does the simulator produce stack traces to transcript or to
vsim_stacktrace.vstf file. At least all my crashes produce the
stacktraces and the traces contain even line numbers of the failing
code that were in the stack during execution.
> Modelsim doesn't raise an error to give me the chance to solve the
> problem. Maybe I did the one thing, no one thought about at Modeltech
> Company. By the way, I do only use the command line. My testbench
Why don't you call to the Mentor support or send e-mail. If you have
some support contracts they are very helpful. I have debugged many
problems with them and we have usually found good solutions. If the
case is very difficult they might need the design files so be prepared
to some NDA papers :)
--
=============================================================================
Mr. Kim Enkovaara | kim.en...@iki.fi | Microelectronic Riemannian
Vasamatie 1 C 16 | IRC: embo | curved-space fault in
02630 Espoo | | write-only file system
Wolfram Stumpf
I'm working *without* the GUI. I wrote a little script, which compiles
my design and runs the simulation afterwards.
Regards
Wolfram Stumpf
David Bishop
Wolfram Stumpf wrote:
> Once again the error message:
>
> Problem with simulator... vsim U/I closing. (2)
> Problem with simulator... vsim U/I closing. (1)
> # ERROR: Read failure in vlm process (0,0)
>
> Does anyone now what the numbers (2),(1),(0,0) tell me?
My group hit these errors as well. Modeltech told us the standard
answer, which was to upgrade to 5.6. I doubt that they even looked
at the VLM logs we sent them. We found 5.6 to be even worse
than 5.5e, so we dropped back to 5.4c.
Makes one want to look at Aldec....
--
NAME: David W. Bishop INTERNET: dbi...@vhdl.org