[libre-riscv-dev] [Bug 173] dynamic partitioned "shift"

[bugzilla-daemon at libre-riscv.org]

http://bugs.libre-riscv.org/show_bug.cgi?id=173

--- Comment #6 from Luke Kenneth Casson Leighton <lkcl at lkcl.net> ---
wait... no... yes, it looks like you've already done that (to some extent),
at least limited the range of the matrix partial-results.

however what's missing is the truncation of B to match the length of the
output:

                comb += matrix[i][j].eq(a << b)
                start = end

should be something like:

                outwid = matrix[i][j].shape()[0]
                bwid = math.ceil(math.log2(outwid + 1))
                comb += matrix[i][j].eq(a << b[:bwid])
                start = end

or something like that.  there's almost certainly at least two off-by-one
bugs in that :)

-- 
You are receiving this mail because:
You are on the CC list for the bug.