Hi Tristan,
> clearly (and as you have already predicted), most of the time is spent in
> next compute. I have an idea to improve that.
the stats look pretty extreme - not sure it is really that extreme in
practice, valgrind profiling shows more time spent running processes.
Anyway, there are a couple of wired-or busses that use the following
resolution function:
function or_reduce(arg : std_ulogic_vector) return std_ulogic is
variable a : std_ulogic_vector(arg'length-1 downto 0);
variable h : integer;
begin
if arg'length < 1 then
return '0';
end if;
a := to_UX01(arg);
if a'length = 1 then
return a(a'left);
elsif arg'length = 2 then
return a(a'left) or a(a'right);
else
h := (a'length+1)/2+a'right;
return or_reduce(a(a'left downto h)) or or_reduce(a(h-1 downto a'right));
end if;
end or_reduce;
Not sure where the runtime of this function shows up - in next compute?
Happy new year!
Tom
_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss
