Split hardware mapping from `fifo.ys` into `fifo_map.ys`. Reduces size of `fifo.out` log and allows separate yosys calls in the makefile.
Some tidy up and minor changes in `fifo.ys` for better discussion.
Filled out note on `clean` (changed from `opt_clean`) and introduced `;;`.
Highlighted `$memrd` and added a paragraph about it.
More detail on the flatten and merging of `fifo_reader` block.
Brief discussion on the changes from `$memrd` to `$memrd_v2`.
Split coarse grain representation into 4 parts, loosely: fsm/opt, other optimizations/techmap/memory_dff, DSPs, alumacc/memory -nomap.
Split hardware mapping into subsections as well: memory blocks (map_ram and map_ffram), arithmetic (map_gates), FFs (map_ffs), LUTs (map_luts and briefly abc), and other (map_cells and a note on hilomap and iopadmap).
Also add `-T` flag to Yosys call to remove footer from log output.
Added highlighting in (most) schematics.
Written down to end of coarse-grain, with a couple of TODOs for filling in gaps.
Includes `techmap_synth.rst` stub.
Fifo code based on SBY quick start.
Instead of showing the full design we are (currently) focusing on a single output (rdata), using `%ci*` to get the subcircuit it relies on.
Overall structure in place to match the iCE40 flow.
Still needs a new example design, and more text for the later sections (which the counter doesn't cover).
Updates code examples, removing `counter_outputs.ys` in favour of a single script. Also adds a .gitignore for the output file `synth.v`.
`example_synth.rst` still pending updated example.
Moved the last files out of the resources directory.
Some tidy up/reformatting of the extensions to allow literalincludes from `my_cmd.cc`.
Most (all?) of the getting started guidelines file is either in the quick guide section, or sections referenced by it. Instead of including it verbatim, we'll instead just leave a reference to it but then jump straight into the quick guide.
Include an image for the absval generated module. Still needs more surrounding text but it's good enough for now.
Also includes some other minor tidying, including removing the no longer used abc_01 code example.
Includes comparison of `abc` v `abc9`. Also creates a new subsection of the
yosys internals for extending yosys (moving the previous extensions.rst into it).
Co-authored-by: Lofty <dan.ravensloft@gmail.com>
Blocking tasks are now capital TODO (compared to non-blocking todo).
Updated some of the todos.
Added note about which intel synth does which families.
Rename extended Yosys universe to Yosys family.
Added brief text to landing page, and also a note about the restructure and where to find old docs.
Moved todolist above ToC in preparation for disabling it in the config (so that it doesn't need it's own header).
Fixed pdf build, was previously breaking on trying to include the svg badges.
The $shift and $shiftx cells perform a left logical shift if the second
operand is negative. This change passes the sign of the second operand
of AST_SHIFT and AST_SHIFTX into $shift and $shiftx cells, respectively.
Includes CAD suite info and details on the OSS CAD suite nightly build targets.
Instructions for building from source, largely based on the readme but with some minor modifications.
Tests are still WIP, but we replaced the old test suites with a brief comment on the github workflow tests. Still needs more on the tests themselves and how to run them locally.
Also an extra todo on the index page.
When we are iterating over a `SigSpec`, the visited values will be of
type `SigBit` (as is the return type of `operator*()`). Account for that
in the publicly declared types.
`std::iterator` has been deprecated in C++17. Yosys is being compiled
against the C++11 standard but plugins can opt to compile against a
newer one. To silence some deprecation warnings when those plugins are
being compiled, replace the `std::iterator` inheritance with the
equivalent type declarations.
Based on the Ignite presentation and github.
Adds links for the extended Yosys universe.
Moves the original thesis stuff further down (and labels it as such).
When a plugin is being loaded from Python source, the binding will be
available under
import libyosys
That is unfortunately different from how a self-standing Python program
would import the Yosys interface, which is
from pyosys import libyosys
Until that is made consistent, at least fix the example to have it
working as is.
Moved remaining content into relevant places.
Added `load_design.rst` to more scripting.
Split fsm handling and abc out of optimization passes. Also moved things around to match the general flow previously described.
Changed generic `synth` for `prep` instead.
Replace `typical_phases.rst` and `examples.rst` with a single `example_synth.rst`.
Also updating the counter example to match.
Aims to reduce redundancy, and simplify the getting started section.
Details on things like `proc`, `memory` and `fsm` should instead be in the advanced section (under the new `synth` subsection).
The previous version could easily generate a large amount of padding
when the constant factor was significantly larger than the width of the
shift data input. This could lead to huge amounts of logic being
generated before then being optimized away at a huge performance and
memory cost.
Additionally and more critically, when the input width was not a
multiple of the constant factor, the input data was padded with 'x bits
to such a multiple before interspersing the 'x padding needed to align
the selectable windows to power-of-two offsets.
Such a final padding would not be correct for shifts besides $shiftx,
and the previous version did attempt to remove that final padding at the
end so that the native zero/sign/x-extension behavior of the shift cell
would be used, but since the last selectable window also got
power-of-two padding appended after the padding the code is trying to
remove got added, it did not actually fully remove it in some cases.
I changed the code to only add 'x padding between selectable windows,
leaving the last selectable window unpadded. This omits the need to add
final padding to a multiple of the constant factor in the first place.
In turn, that means the only 'x bits added are actually impossible to
select. As a side effect no padding is added when the constant factor is
equal to or larger than the width of the shift data input, also solving
the reported performance bug.
This fixes#4056
Remove duplicate %.pmg -> %_pm.h pattern. One of the duplicates overrode
the other, and in some conditions there were build races as to whether
the target directory for the generated header would exist. Instead have
a single rule which is properly generalized.
Krystine Sherwin
Martin Povišer
Dag Lem
Jannis Harder
N. Engelhardt
github-actions[bot]
Merry
Miodrag Milanović