Nicholas Clark wrote:
> On Fri, Oct 11, 2002 at 05:01:49PM -0400, Dan Sugalski wrote:
>
>>At 9:02 PM +0100 10/11/02, Nicholas Clark wrote:
>>
>
>>>I would like to kill all generated variants of all the 3 argument opcodes
>>>where both input arguments are constants. They truly are superfluous.
>>Where both operands are ints or nums, I think it's a good idea. I'm
>>less sure in the case where there's a PMC or string involved, as
>>there may be some assumption of runtime behaviour (in the case of
>>constant PMCs that might have some methods overloaded) or strings
>>where the compiler is expecting runtime conversion to happen before
>>whatever gets done.
Ok, we need 2 + 3 args PMC ops becaus of overloading.
> I think I agree with this reasoning. I was really thinking of the ints
> as being easiest to bump off, providing we can be sure that things will
> consistently for bytecode compile on a 32 bit parrot, but run by a 64
> bit parrot (or likewise for different length floating point)
By preprocessing ints and nums, we would move overflow/precision issues
from the running machine to the compiling machine. But the issue itself
would remain.
So we could end up with:
add_i_ic add_n_nc
add_i_i_i add_n_n_n
add_i_i_ic add_n_n_nc
add_i_ic_i add_n_nc_n
where the latter could be tossed, when the assebler/imcc swaps
arguments. This saves 6 opcodes for each <add> and <mul> and ~2 opcodes
for each <sub>, <div> and <mod>. <sub> with constants could be rewritten
to <add>. Saving a total of ~20 ops.
Some 2 operand bitwise operations are there some not - there is no
consistency - I would remove the to operand bitwise integer ops.
> More free speedup. I had this crazier idea - experiment with splitting
> every parrot function out into its own object file, and see what happens
> with permuting the order of all of them.
I have s small shell script (attached), which generates opcode coverage
for all PBC files in the parrot tree:
total ops 866
ops types 177
op usage stat
540 op-list-all
161 op-list-s
So only ~2/3 opcodes are currently used/tested (ops types/op-list-s are
ops short names w/o variants).
To run the script do:
make test
cd languages/perl6
make
perl6 --test
cd -
make disassemble
./op-stat
> But I think I need a lot of tuits, and a decent way of doing permutations
> with genetic algorithms. (I've got access to a fast machine, but it will
> have to stop smoking perl5 for the duration).
This would keep the machine busy for some time ;-) Interesting idea.
> Nicholas Clark
leo
#!/bin/sh
DIS=disassemble
[ -e .op-list1 ] && rm .op-list1
[ -e op-list-all ] && rm op-list-all
[ -e op-list-s ] && rm op-list-s
find . -name '*.pbc' -fprint /dev/stderr -exec $DIS {} \; |
sed -e's/^[[:blank:]]*L[0-9]*://' | \
sed -e's/^[[:blank:]]+//' | cut -d\ -f1 >> .op-list1
sort < .op-list1 | uniq -c | sort -rg > op-list-all
sort < .op-list1 | sed -e's/_.*//' | uniq -c | sort -rg > op-list-s
echo
T=`grep NAME lib/Parrot/OpLib/core.pm | wc -l`
S=`grep NAME lib/Parrot/OpLib/core.pm | sort | uniq | wc -l`
echo total ops $T
echo ops types $S
echo op usage stat
wc -l op-list* | head -2
