On 08/02/10 20:18, john gilmore wrote:
Several--well 4--people have asked me off-list how the assembly-time
construction of the table of one-bit counts can be automated. Its manual
construction by inspection is possible, though not I think advisable, for a
table of the one-bit element counts for 0(1)255, which contains only 256
values. For a table of these counts for 0(1)65535, which contains 65536
values, something else is required. ...
Here's mine; macro-free:
BITTAB CSECT
TABLE DC 65536AL1(*-TABLE-(*-TABLE)/2-(*-TABLE)/4-(*-TABLE)/8-(*-x
TABLE)/16-(*-TABLE)/32-(*-TABLE)/64-(*-TABLE)/128-(*-TABx
LE)/256-(*-TABLE)/512-(*-TABLE)/1024-(*-TABLE)/2048-(*-Tx
ABLE)/4096-(*-TABLE)/8192-(*-TABLE)/16384-(*-TABLE)/3276x
8)
END
This [snipped] macro is "very inefficient", but notional compile-time
inefficiency is a small price to pay for ensuring that a generated execution-time table
is correct. Readers who are unfamiliar with created set symbols will find it opaque but
not unintelligible. It is also an illustration of macro of the sort that mathematicians
call a lemma: it is of no independent interest, and it does not itself generate code.
Its function is instead to facilitate and simplify the work of another macro or macros
that do generate code.
Which is opaquer is in the eye of the beholder.
-- gil
