Closed gojimmypi closed 5 years ago
FYI, this is a long known issue in the placement validity checks, where arachne's placer will sometimes end up using more than 32 local inputs in a logic tile. It is properly fixed in nextpnr.
Is this issue documented somewhere? I spent a ton of time trying to figure it out. Certainly googling "fatal error: failed to route" nextpnr
and related terms did not seem to help. Like most things - easy when you know the answer, eh! lol :)
As for nextpnr, I would have tried, but I've not yet been able to get it to compile as noted in https://github.com/YosysHQ/nextpnr/issues/214 and https://github.com/YosysHQ/nextpnr/issues/129#issuecomment-455707572
Golly! Its so fixed, except I get it after 1) following this install and then 2) building from this directory, see the error dump of the following.
I have to also note that my third part directory (nextpnr/3rdparty/abseil-cpp) directory was empty and was causing a build fail. So I changed directory to my nextpnr/3rdparty and then cloned abeil-cpp using:
git clone https://github.com/abseil/abseil-cpp.git
Note, I have just installed fresh today and so cannot vouch for a fix in nextpnr, if that was supposed to remove the problem.
I am on Debian 10, running Python 3.7.3.
On the assumption the original install and having to fix that install prompts the problem, I will need rip the lot out and try a fresh install of symbiflow instead.
@gojimmypi the compile problem is probably because the nextpnr distro is missing 3rd party dependencies.
Probably won't help Makefile of nextpnr does not have an "uninstall" target ... bad form.
I've been working with the RISCV example project on TinyFPGA BX and have documented my experiences here.
One the the things that really tripped me up was the
fatal error: failed to route
message with arachne-pnr. (full details shown below).I just wanted to document here that the resolution to this problem for me was to use the
-r
option to "randomize the seed". Specifically like this:arachne-pnr -r -d 8k -P cm81 -o hardware.asc -p hardware.pcf hardware.blif
Hopefully having this explicit GitHub issue will help others avoid the problems I had with this.
Here's the output from arachne-pnr without the
-r
, showing the error.