mirrors/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython.git synced 2025-12-04 16:43:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 10

Aggressive reordering of dict comparisons. In case of collision, it stands

Browse source 
to reason that me_key is much more likely to match the key we're looking
for than to match dummy, and if the key is absent me_key is much more
likely to be NULL than dummy:  most dicts don't even have a dummy entry.
Running instrumented dict code over the test suite and some apps confirmed
that matching dummy was 200-300x less frequent than matching key in
practice.  So this reorders the tests to try the common case first.
It can lose if a large dict with many collisions is mostly deleted, not
resized, and then frequently searched, but that's hardly a case we
should be favoring.
This commit is contained in:
Tim Peters 2001-05-13 06:43:53 +00:00
parent 2f228e75e4
commit 342c65e19a
1 changed files with 21 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

51
Objects/dictobject.c
View file

@ -219,26 +219,21 @@ lookdict(dictobject *mp, PyObject *key, register long hash)
incr = (hash ^ ((unsigned long)hash >> 3)) & mask; incr = (hash ^ ((unsigned long)hash >> 3)) & mask;
if (!incr) if (!incr)
incr = mask; incr = mask;
/* In the loop, me_key == dummy is by far (factor of 100s) the
least likely outcome, so test for that last. */
for (;;) { for (;;) {
ep = &ep0[(i+incr)&mask]; ep = &ep0[(i+incr)&mask];
if (ep->me_key == NULL) { if (ep->me_key == NULL) {
if (restore_error) if (restore_error)
PyErr_Restore(err_type, err_value, err_tb); PyErr_Restore(err_type, err_value, err_tb);
if (freeslot != NULL) return freeslot == NULL ? ep : freeslot;
return freeslot;
else
return ep;
} }
if (ep->me_key == dummy) { if (ep->me_key == key) {
if (freeslot == NULL)
freeslot = ep;
}
else if (ep->me_key == key) {
if (restore_error) if (restore_error)
PyErr_Restore(err_type, err_value, err_tb); PyErr_Restore(err_type, err_value, err_tb);
return ep; return ep;
} }
else if (ep->me_hash == hash) { else if (ep->me_hash == hash && ep->me_key != dummy) {
if (!checked_error) { if (!checked_error) {
checked_error = 1; checked_error = 1;
if (PyErr_Occurred()) { if (PyErr_Occurred()) {
@ -257,11 +252,12 @@ lookdict(dictobject *mp, PyObject *key, register long hash)
else if (cmp < 0) else if (cmp < 0)
PyErr_Clear(); PyErr_Clear();
} }
else if (ep->me_key == dummy && freeslot == NULL)
freeslot = ep;
/* Cycle through GF(2^n)-{0} */ /* Cycle through GF(2^n)-{0} */
incr = incr << 1; incr <<= 1;
if (incr > mask) if (incr > mask)
incr ^= mp->ma_poly; /* This will implicitly clear incr ^= mp->ma_poly; /* clears the highest bit */
the highest bit */
} }
} }
@ -314,28 +310,23 @@ lookdict_string(dictobject *mp, PyObject *key, register long hash)
incr = (hash ^ ((unsigned long)hash >> 3)) & mask; incr = (hash ^ ((unsigned long)hash >> 3)) & mask;
if (!incr) if (!incr)
incr = mask; incr = mask;
/* In the loop, me_key == dummy is by far (factor of 100s) the
least likely outcome, so test for that last. */
for (;;) { for (;;) {
ep = &ep0[(i+incr)&mask]; ep = &ep0[(i+incr)&mask];
if (ep->me_key == NULL) { if (ep->me_key == NULL)
if (freeslot != NULL) return freeslot == NULL ? ep : freeslot;
return freeslot; if (ep->me_key == key
else || (ep->me_hash == hash
return ep; && ep->me_key != dummy
} && compare(ep->me_key, key) == 0))
if (ep->me_key == dummy) {
if (freeslot == NULL)
freeslot = ep;
}
else if (ep->me_key == key
|| (ep->me_hash == hash
&& compare(ep->me_key, key) == 0)) {
return ep; return ep;
} else if (ep->me_key == dummy && freeslot == NULL)
freeslot = ep;
/* Cycle through GF(2^n)-{0} */ /* Cycle through GF(2^n)-{0} */
incr = incr << 1; incr <<= 1;
if (incr > mask) if (incr > mask)
incr ^= mp->ma_poly; /* This will implicitly clear incr ^= mp->ma_poly; /* clears the highest bit */
the highest bit */
} }
} }