By default, latest development versions (branches master | main) of tools are built.
Then, smoke tests are executed and, if successful, the corresponding images are updated in the registries.
However, some specific tools are not built from sources, but installed through (system) package managers.
Those are marked with ! in the table below.
Note
Package images are not to be used for executing the tools, but for composing images including multiple resources. See :ref:`Development:package-images` for further details.
- Simulation:
- |SHIELD:Image:sim| GHDL + NVC + Verilator + Icarus Verilog
- |SHIELD:Image:sim/osvb| CoCoTb, OSVVM and VUnit; on top of
sim. - |SHIELD:Image:sim/scipy-slim| matplotlib and numpy; on top of
sim. - |SHIELD:Image:sim/scipy|
osvb; on top ofsim/scipy-slim. - |SHIELD:Image:sim/octave-slim|
octave; on top ofsim. - |SHIELD:Image:sim/octave|
osvb; on top ofsim/octave-slim.- |SHIELD:Image:sim/octave/gnuplot| gnuplot; on top of
sim/octave.
- |SHIELD:Image:sim/octave/gnuplot| gnuplot; on top of
- Implementation: GHDL + Yosys + nextpnr
- |SHIELD:Image:impl/ice40| nextpnr-ice40 only, and |SHIELD:Image:impl/icestorm| including icestorm.
- |SHIELD:Image:impl/ecp5| nextpnr-ecp5 only, and |SHIELD:Image:impl/prjtrellis| including prjtrellis.
- |SHIELD:Image:impl/nexus| nextpnr-nexus only, and |SHIELD:Image:impl/prjoxide| including prjoxide.
- |SHIELD:Image:impl/generic| nextpnr-generic only.
- |SHIELD:Image:impl/pnr| all nextpnr targets (ecp5, ice40, nexus, and generic).
- |SHIELD:Image:impl| impl:pnr, including icestorm, prjtrellis and prjoxide.
- Formal:
- |SHIELD:Image:formal| all solvers depending on Python 3.
- |SHIELD:Image:formal/min| Z3 only.
- |SHIELD:Image:formal/all| all solvers, depending on either Python 2 or Python 3.
- Programming: |SHIELD:Image:prog|
:awesome:`F4PGA <f4pga>` is an ecosystem of EDA tools for the development of FPGAs of multiple vendors. The tools used in the F4PGA ecosystem are packaged by Antmicro and contributors using Conda. The sources of Conda packages for EDA tooling are found in |SHIELD:HDL:conda-ci|, |SHIELD:HDL:conda-compilers|, |SHIELD:HDL:conda-eda|, |SHIELD:HDL:conda-misc| and |SHIELD:HDL:conda-prog|.
In HDL Containers, as a complement to container images based on tools built on other container images, a set of
Conda-based F4PGA images is provided.
The base is |SHIELD:Image:conda|, which includes a Miniconda setup.
On top of that, the following ready-to-use images allow following the guidelines in
- |SHIELD:Image:conda/f4pga/xc7/toolchain|: the toolchain for
FPGA_FAM=xc7, on top ofconda. The following images are based on this:- |SHIELD:Image:conda/f4pga/xc7/a50t|: includes architecture definitions for xc7a50t.
- |SHIELD:Image:conda/f4pga/xc7/a100t|: includes architecture definitions for xc7a100t.
- |SHIELD:Image:conda/f4pga/xc7/a200t|: includes architecture definitions for xc7a200t.
- |SHIELD:Image:conda/f4pga/xc7/z010|: includes architecture definitions for xc7z010.
- |SHIELD:Image:conda/f4pga/xc7/z020|: includes architecture definitions for xc7z020.
- |SHIELD:Image:conda/f4pga/xc7|: includes all the architecture definitions for the xc7 family, except xc7a200t (due to hard disk limits on GitHub Actions).
- |SHIELD:Image:conda/f4pga/eos-s3|: the toolchain for
FPGA_FAM=eos-s3and the architecture definitions, on top ofconda.
Important
The compression ratio of these images is very high compared to other container images. That is because these include a significant amount of data in text format, which is much better compressed than binaries. As a result, the size of the images when pulled is approximately as follows:
- xc7/toolchain: 3 GB
- xc7/a50t: 6 GB
- xc7/a100t: 8.5 GB
- xc7/a200t: 15 GB
- xc7/z010: 5 GB
- xc7/z020: 7.5 GB
- xc7: 18 GB (would be 30 GB if a200t was included)
- eos-s3: 2 GB