The techmap rules for this target do not work in the first place (note
lack of >2-input LUT mappings), and if proper support is ever added,
it'd be better placed in the synth_intel_alm backend.
By operating at a layer of abstraction over the rather clumsy Intel primitives,
we can avoid special hacks like `dffinit -highlow` in favour of simple techmapping.
This also makes the primitives much easier to manipulate, and more descriptive
(no more cyclonev_lcell_comb to mean anything from a LUT2 to a LUT6).
Use GNU make's foreach iterator and remove nonexistent files. Gmake is
already a requirement of the build system.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
This regresses Cyclone V and Cyclone 10 substantially, but these
numbers were artificial, targeting a BRAM that they did not contain.
Amusingly, synth_intel still does better when synthesizing PicoSoC
than Quartus when neither are inferring block RAM.
Marcelina Kościelnicka
Dan Ravensloft
Ben Widawsky
dh73
Larry Doolittle