tgingold on master
Add missing file for previous c… (compare)
tgingold on master
synth: minor refactoring about … synth: add synth-flags, add deb… synth: fold addition on constan… and 6 more (compare)
--enable-synthonly, so you built GHDL with
--enable-libghdl(which is enabled by default), isn't it? I.e. the same you reported here but with a more recent version of GNAT.
When I now want to use the --synth option with ghdl, I only get
ghdl:error: unknown option '--synth' for command '-e'
This was an issue until last week or so, when a fix was added. Might it be that you weren't using the latest commit from master at first?
Ah, it seems that the reason was I forgot to do
xcode-select -s /Applications/Xcode.app/Contents/Developerafter installing the new gnat-2019
I think I did try this in #867, but I will try again. For now, building on macOS with
--enable-libghdl fails on Travis CI or GitHub Actions. Therefore, I think this is an issue with how we are (programatically) installing GNAT.
Apart from that, on GNU/Linux there is a kind of hackish but very powerful feature: GHDL can generate PIE binaries of your design, which can be dynamically loaded as a shared library. This allows co-execution from any foreign tool (Python, C/C++, etc.).this is interesting ... I've done some QEMU + FPGA co-sim in the past
We are working on providing better packages (i.e., PKGBUILD artifacts, rpms, debs, etc.) for Fedora/CentOS/RHEL, Debian/Ubuntu, ArchLinux, MSYS/MINGW... Which flavour do you need?I generally use Ubuntu, CentOS if I have to (i.e. to run diamond)
Apart from that, if you use docker containers, I believe that you don't need to have both GHDL and Lattice Diamond in the same imageI need the Diamond files to build the Lattice libraries. Unfortunately I have hit a "brick wall" on this since the Lattice libraries for the MACHXO2 are not complete (some parts are in Verilog)
How do you handle licensing on CI systems? Do you use private CI servers which can access private license servers?so far file based licences (i.e Lattice Diamond, Xilinx Vivado) and I put the licence file in the docker image
I've done some QEMU + FPGA co-sim in the past
If the target device is ARM, with GHDL you can avoid QEMU. E.g., in Zynq devices, GHDL simulations can be executed. Depending on the complexity, simulation can be slow, tho. It is then possible to execute the software in the target ARM device and the simulation in a workstation.
I need the Diamond files to build the Lattice libraries.
You can share just the directories containing those files.
I have hit a "brick wall" on this since the Lattice libraries for the MACHXO2 are not complete (some parts are in Verilog)
Oh. I'm not using tar anymore. I forgot that. Neither mkdir. The directory is implicitly created by
COPY in the dockerfiles:
Nevertheless, I'm in the middle of the migration to GitHub Actions. I will open an issue to track this, and I will fix it in the following days. The fix itself is easy, but I need to trigger several jobs in a chain, and I don't know how to force it semi-automatically yet.
base: ERROR: No such command: ghdl (type 'help' for a command overview)when running symbiyosys
ghdl/synth:formalimage, in order to ensure that whatever we push does not need additional tweaks.