The notes make a point of not installing Libero to /usr/local/Microchip/Libero_SoC_v2022.3 but to a non root directory such as ~/microchip. I used ~/opt/microchip.
What the note does not mention is that the license Deamon expects to find the license file at /usr/local/flexlm/licenses. That fact is completely hidden and not mentioned anywhere. Fortunately, running lmgrd generates an error message that gives the fact away.
The second frustration is that the following the very last note in the install process is
I was unable to make this work. May I suggest that this, no doubt very resource, be made more accessible.
As a general note, having to use Ubuntu 20.04 which is very much out of date is a royal pain!
I tried 23.04, 22.04 very time consuming and a complete waste of time. Don’t do it!!
Finally, I accepted the reality of Ubuntu 20.04 and installed it. In total, I’ve spent a total of three solid days to get to what appears to be a working system. I wonder what challenges are ahead.
Also, the license can be altered too in location of itself. /microchip/license/license.dat or whatever location.
Seth
P.S. The one thing I messed up during he install was forgetting the . before ./setup-microchip-tools.sh to make it look like this idea: . ./setup-microchip-tools.sh
Oh! I am using the 22.04.3 versioning of Ubuntu! It needs to be set up correctly but it works…
Looks like they are now all about selling subscriptions and not chips. Years ago we used several of their products and they did have the best tools. However if memory (long time ago) is correct the tools even for commerical use were free since you need good tools to build products with their silicon.
Regarding Libero license, you only need the Silver license which gives you the same FPGA tools as the other licenses (including Synopsis Identify).
I always find the place to request a free Silver license difficult to find but it is on the Microchip FPGA licensing page: https://www.microchip.com/en-us/products/fpgas-and-plds/fpga-and-soc-design-tools/fpga/licensing. Look for the words “free license” on that page. This will bring you to the MicrochipDirect site where you have to register before getting your free license.
Only made sure a loopback at 5Gbps works. I had misread the SYZYGY spec and thought the connector was rated for only 5Gbps but it does not look like there is a limitation on the connector. Worth a try at 12Gbps. SmartDebug can generate an eye for these signals. The eye looked good at 5Gbps.
Not sure how to get an accurate measurement with that. Seems like a stretch, even Pi is going out on limb with Pi5 and claiming, I believe dual 10Gbs with a HAT. That funky connector (pi board) does not seem like it could even handle 1 Gbs as in sustained and no errors… That would be nice if that V board really did hit the 12Gbs with extremely high reliability.
For the last week or so I’ve been busy with Kicad laying out a ZYSYGY to USB-C interface. While an inch may not seem like a lot, at 12.5 GHz every mil matters. So does, ESD protection and so on…
About 1/2" could be saved by extending the interface beyond the present PCB outline … the gods shudder.
Looks like licensing problem. Can you start Libero and open the libero project that was created in the work directory and try to run synthesis from the GUI. That will tell us if this is a general Libero license issue or something more specific with running the scripts in command line.
Well, getting solid data is more important than speed.
Would like to know what the test setups are for the the claims. Not not picking on this board, its a fairly common issue with others.
I got ripped on some fiber networking stuff, bought it on amazon. Dual channel that is not even close to what they claim. It gets even worse, it sat out in the shop way past the return date. By the time it was tested it was too late to return it, about all it is good for is emi, emp and lightening protection, it sure was not speed.
Until the manufactures can post the exact test setup its all a pile …
The quality of the communications channel for a wired interface should be at least 10exp-9. with the necessary test mechanism built into the device.
Knowing your bit error rate is essential. However that alone only tells you there is a problem. Not where. Having access to the eye patterns is really helpful.
I had no issues with the licensing (other than the fact that only one copy of libero can run at a time). I couldn’t get Microchip-FPGA-Tools-Setup to work either. So I ended up hacking it by putting in my bashrc:
and appended the LM_LICENSE_FILE with the location of license.dat from Microchip (it’s in LICENSE_FILE_DIR and the binaries are in LICENSE_DAEMON_DIR). Then I run “lmgrd -c $LICENSE_FILE_DIR/license.dat -l /tmp/uchip.log”. lmdown (which is not provided by Microchip but available from licensing of questa/modelsim) is useful for shutting down the microchip daemons.
I did get to install libero and the other software under a root directory (/opt) but then installing cores is a pain as those need to be at a non-user location and one has to run everything with “sudo libero LM_LICENCE_FILE=$LM_LICENSE_FILE” to update the cores. Plus not sure it is useful for a multiuser environment since only one copy of libero can run at a time anyway (for the free license).
