02-19-2010 12:58 PM
I am running Redhat Linux Enterprise 5 using ISE/EDK/SDK 11.4. We are currently trying to use the Virtex5 XUP Board, and have downloaded the BSB from the Xilinx Website. Through the BSB we create a very basic setup (just microblaze plus BRAM memory controllers), During the mapping phase (typyically phase 5.3-5.5) mapping receives a segmentation fault. We have created our own mapping script which contains only the basic options as well and the same errors occur. By removing all UCF constraints except for clk and reset pin mappings we are able to get to mapping phase 10.3 however it still fails with a segfualt after that. The exact same system mapped to the ml505 maps (and creates a bitstream) with no problem, so it is not a tool setup issue on our end.
We have taken the exact same system (V5 XUP BSB) and tried this on an equivalent windows based system and mapping works fine with no changes to the UCF. I am inclined to believe this an issue with V5lx110T mapping and has nothing to do with the V5 XUP Board in particular.
Has anyone had this problem and/or know of any fix? Or is this an internal problem.
02-19-2010 01:28 PM
I am running ISE 11.4 (XPS, ISE, etc.) Red Hat Enterprise Linux WS Release 4, update 7, with no problems...
Do you have enough RAM on your workstation?
02-19-2010 01:44 PM
The Machine has 4GB of RAM with around 2.5-3GB available to applications at any given time. In addition if we watch the memory utilization of the map process during execution it only reaches around 10% of available memory (however it does consume significant CPU 99%, but that is to be expected)
Have you specifically tried to map to the LX110t device using EDK and the BSB for V5? I also created the system from scratch (without BSB) and the same issue occurs. I have not tried using ISE standalone to create a bitstream for the device. But i can see what happens if we do that as well.
02-19-2010 02:08 PM
I mapped a very simple design (2 dip switches controlling an led) and tested it with ISE. and still mapping fail in phase 14.4 with a segmentation fault.
02-19-2010 02:25 PM
Yes. I just tried all the zip files this morning, and made sure I could compile all three of the reference designs. The xupv5_bsb_std_ip design fails with two bad pins for the I2C controller, so it is suggested to comment out those two pins in the ucf.
With those two pins commented out, then this, too, compiles without errors all the way to making a bitstream.
02-19-2010 02:44 PM
After some more digging I do believe this is a memory problem at some level (not exactly sure where yet) because when using smaller devices (LXT30 or 50) there are no problems. Anything bigger than a 50 (85, 110, 220) and we get the same behavior. The same thing happens when also trying to use the FX series (70 and above).
I am guessing the map process is not being allowed to allocate enough memory in the first place (maybe a quota is in place) and it is exiting unexpectedly. I believe finding this will solve the problem.
02-19-2010 02:59 PM
lists the memory requirements...
Sounds like there is something odd going on. Of course, I have a ton of memory, as I use many tools that are just insane hogs of memory (not Xilinx tools).
Perhaps someone will comment?
03-12-2010 10:43 AM
You don't explicitly state this, but I presume that the SegFault is happening during MAP (I want to clarify because a lot of folks don't distinguish between NGDBuild, MAP, and PAR).
We definitely take an interest in these SegFault issues. Normally, when we run out of memory in a machine we throw a Portability error and report the out-of-memory condition.
We don't consider SegFaults acceptable in any scenario.
I highly recommend that you file a WebCase with us a tech support so that we can triage this in a more official manner (and file any needed defect reports).
03-12-2010 11:11 AM
Yes, it is indeed happening during MAP (after NGDBuild). We are currently in the process of making sure this is indeed a tool issue and isn't due to an improper installation of Redhat/Xilinx Tools.
I should know exactly what is causing this problem (Map process, Redhat/Xilinx install) mid-next week and will act upon it accordingly. Thank you for your help. As of know, I also do not believe it is a memory problem, but something else.
03-14-2010 01:38 PM
What phase of MAP is it failing it? That might give us some more data to triage with.