Leopold Toetsch <[EMAIL PROTECTED]> wrote:
> OTOH I'm thinking of adding _keyed_str variants to used _keyed vtables.
> We are already constructing too much temporary key PMCs from a string
> just to pass a string to an hash lookup. Object ops and functions do
> this all over the place repeatedly.
And here is a profile timing comparison for doing 1E6 findclass ops [1]:
CODE OP FULL NAME CALLS TOTAL TIME AVG T. ms
---- ----------------- ------- ---------- ----------
743 findclass_i_sc 1000000 1.356876 0.0014
-3 DOD 1058 0.211712 0.2001
...
---- ----------------- ------- ---------- ----------
11 3001066 1.666512 0.0006
Doing the same with a direct hash_get() [2]:
1257 findclass_str_i_sc 1000000 0.462577 0.0005
...
---- ----------------- ------- ---------- ----------
11 3000009 0.564461 0.0002
This is 3 times faster and saves ~1000 DOD runs.
[1]
newclass P1, "Foo"
loadlib P2, "myops_ops"
set I0, 1000000
time N0
loop:
findclass I1, "Foo" # or findclass_str in myops.ops
dec I0
if I0, loop
time N1
sub N1, N0
print N1
print "\n"
end
unoptimized build, Athlon 800.
[2] Of course calling hash_get() directly isn't the way to go, but
VTABLE_exists_keyed_str() is just one vtable call slower, which isn't
much.
# from dynoplibs/myops.ops
# make -C dynoblibs clean all
inline op findclass_str(out INT, in STR) {
Hash *h = (Hash*)PMC_ptr1v(interpreter->class_hash);
if (hash_get(interpreter, h, $2)) {
$1 = 1;
} else {
$1 = 0;
}
goto NEXT();
}
leo
