This is basically the same issue as in tests/various/plugin.sh,
which uses yosys-config to compile a plugin. `yosys-config --cxxflags`
points to `$PREFIX/share/` (/usr/local/share by default), which might
not exist yet or might be out of date. Building directly from the
headers in ./share/ avoids this.
The $div and $mod cells use truncating division semantics (rounding
towards 0), as defined by e.g. Verilog. Another rounding mode, flooring
(rounding towards negative infinity), can be used in e.g. VHDL. The
new $divfloor cell provides this flooring division.
This commit also fixes the handling of $div in opt_expr, which was
previously optimized as if it was $divfloor.
The $div and $mod cells use truncating division semantics (rounding
towards 0), as defined by e.g. Verilog. Another rounding mode, flooring
(rounding towards negative infinity), can be used in e.g. VHDL. The
new $modfloor cell provides this flooring modulo (also known as "remainder"
in several languages, but this name is ambiguous).
This commit also fixes the handling of $mod in opt_expr, which was
previously optimized as if it was $modfloor.
Before this patch, the code passed around std::string objects by
value. It's probably not a hot-spot, but it can't hurt to avoid the
copying.
Removing the copy and clean-up code means the resulting code is ~6.1kb
smaller when compiled with GCC 9.3 and standard settings.
- Pass a string argument by reference
- Avoid multiple calls to IdString::str and IdString::c_str
- Avoid combining checks for size > 0 and first char (C strings are
null terminated, so foo[0] != '\0' implies that foo has positive
length)
With GCC 9.3, at least, compiling select.cc spits out a warning about
an implausible bound being passed to strncmp. This comes from inlining
IdString::compare(): it turns out that passing std::string::npos as a
bound to strncmp triggers it.
This patch replaces the compare call with a memcmp with the same
effect. The repeated calls to IdString::c_str are slightly
inefficient, but I'll address that in a follow-up commit.
The chosen value shouldn't have any effect. I considered something
clearly wrong like -1, but there's no checking inside the generated
lexer, and I suspect this will cause even weirder bugs if triggered
than just setting it to INITIAL.
This code originally comes from commit 458a940. When an interface is
used via a modport, code in genrtlil.cc sets '\\interface_type' and
'\\interface_modport' properties on the wire.
In hierarchy.cc, we pick up the modport name and add it to a dict
called modports_used_in_submodule (that maps connection source to
modport name).
Before this patch, the modport name is retrieved as a strpool and then
iterated over in an arbitrary order, discarding all entries but the
last. In practice, the pool will always have 0 or 1 entries because
the string used to construct it is a valid identifier, so doesn't
contain any pipe symbols.
This patch changes the code to retrieve the modport name as just a
string. This will have the same effect in practice, but may be a bit
less confusing!
The code also gets moved down closer to where the result is used,
which might be a bit more efficient since we won't always get as far
as the check.
The patch also removes some commented-out code, which I think was
intended to add some typechecking at some point, but was never
implemented. Since this dates back to October 2018, I think it makes
more sense to just take it out.
The existing code does a search to figure out whether id is in the
dict (with the call to count()), and then looks it up again to get the
result (with the call to at()). This version calls find() instead,
avoiding the double lookup.
Code size increases slightly (6kb). I think this is because the
contents of find() are getting inlined, and then inlined into lots of
the callsites for cell() and wire().
Looking at the compiled code before this patch, you just get
a (non-inlined) call to count() followed by a call to at(). After the
patch, the contents of find() have been inlined (so you see do_hash,
then do_lookup). The result for each function is about 30 bytes / 40%
bigger, which presumably also enlarges call-sites that inline it.
There was a handwritten copy constructor, which I'm not sure was
actually legal C++ (it unconditionally read from the 'data' member of
a union, which wouldn't have been written if wire was true). It was
also a bit less efficient than the constructor you get from the
compiler by default (which is allowed to just copy the memory).
This gives a marginal (~0.25%) decrease in code size when compiled
with GCC 9.3.
Miodrag Milanovic
clairexen
Eddie Hung
Xiretza
Alberto Gonzalez
whitequark
Rupert Swarbrick