Skip to content

bpo-41972: Tweak fastsearch.h string search algorithms #27091

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ambv merged 21 commits into from
Jul 19, 2021
Merged

bpo-41972: Tweak fastsearch.h string search algorithms #27091

Changes from 1 commit
Commits
Show all changes
21 commits
Select commit Hold shift + click to select a range
90a6a0c
Don't force inlining
sweeneyde Jul 6, 2021
19fa661
Use a Boyer-Moore skip table
sweeneyde Jul 7, 2021
d2219d5
Switch to two-way in more cases
sweeneyde Jul 8, 2021
3b8760f
Tighter comments
sweeneyde Jul 9, 2021
5d86b8d
Fix the direction of the cutoff inequality :)
sweeneyde Jul 10, 2021
2e9a9b1
Separate a gap loop
sweeneyde Jul 10, 2021
c84314c
update cutoffs
sweeneyde Jul 10, 2021
7517c60
tweak cutoffs, fix i - cut + 1 jump
sweeneyde Jul 10, 2021
e5ef33b
Refactor into smaller functions
sweeneyde Jul 11, 2021
a01fb70
Focus the table on the last character of the pattern, giving a tighte…
sweeneyde Jul 11, 2021
6be0c32
tweak cutoffs
sweeneyde Jul 11, 2021
f9fe807
📜🤖 Added by blurb_it.
blurb-it[bot] Jul 12, 2021
25f9b56
Fix sanitizer warnings
sweeneyde Jul 12, 2021
d3f8438
Merge branch 'fastersearch' of https://github.com/sweeneyde/cpython i…
sweeneyde Jul 12, 2021
4379ea6
Refactor for sanitizer
sweeneyde Jul 12, 2021
e336b79
Don't goto declarations
sweeneyde Jul 12, 2021
b8df3e1
Emphasize to the compiler that the loop is a loop
sweeneyde Jul 13, 2021
869f7d0
Use Py_LOCAL_INLINE
sweeneyde Jul 15, 2021
747a39a
Use Py_LOCAL_INLINE again
sweeneyde Jul 15, 2021
72143ef
Tweak thresholds based on most recent comparison.
sweeneyde Jul 16, 2021
4d7d102
Fix greater/less than sign.
sweeneyde Jul 16, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Tighter comments
  • Loading branch information
@sweeneyde
sweeneyde committed Jul 9, 2021
commit 3b8760f8d1b07f94e2c6f0bad1d1587a5259c112
30 changes: 11 additions & 19 deletions Objects/stringlib/fastsearch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -362,7 +362,7 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
// See http://www-igm.univ-mlv.fr/~lecroq/string/node26.html#SECTION00260
const Py_ssize_t len_needle = p->len_needle;
const Py_ssize_t cut = p->cut;
const Py_ssize_t period = p->period;
Py_ssize_t period = p->period;
const STRINGLIB_CHAR *const needle = p->needle;
const STRINGLIB_CHAR *window_last = haystack + len_needle - 1;
const STRINGLIB_CHAR *const haystack_end = haystack + len_haystack;
Expand All @@ -387,9 +387,8 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
memory = 0;
shift = table[(*window_last) & TABLE_MASK];
}
// In most cases, this "Horspool" loop is the hot loop.
while (shift > 0 && window_last < haystack_end) {
LOG("Fast Horspool skip.\n");
LOG("Horspool skip.\n");
window_last += shift;
shift = table[(*window_last) & TABLE_MASK];
LOG_LINEUP();
Expand All @@ -401,39 +400,36 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
Py_ssize_t i = Py_MAX(cut, memory);
for (; i < len_needle; i++) {
if (needle[i] != window[i]) {
LOG("Right half does not match. Jump ahead by %zd.\n",
i - cut + 1);
LOG("Right half does not match.\n");
window_last += i - cut + 1;
memory = 0;
goto periodicwindowloop;
}
}
for (i = memory; i < cut; i++) {
if (needle[i] != window[i]) {
LOG("Left half does not match. Jump ahead by period %zd.\n",
period);
LOG("Left half does not match.\n");
window_last += period;
memory = len_needle - period;
goto periodicwindowloop;
}
}
LOG("Left half matches. Returning %zd.\n",
window - haystack);
LOG("Found a match!\n");
return window - haystack;
}
}
else {
Py_ssize_t gap = p->gap;
period = Py_MAX(gap, period);
LOG("Needle is not periodic.\n");
assert(cut < len_needle);
STRINGLIB_CHAR needle_cut = needle[cut];
windowloop:
while (window_last < haystack_end) {
LOG_LINEUP();
Py_ssize_t shift = table[(*window_last) & TABLE_MASK];
// In most cases, this "Horspool" loop is the hot loop.
while (shift > 0 && window_last < haystack_end) {
LOG("Fast Horspool skip.\n");
LOG("Horspool skip.\n");
window_last += shift;
shift = table[(*window_last) & TABLE_MASK];
LOG_LINEUP();
Expand All @@ -445,23 +441,19 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
for (Py_ssize_t i = cut; i < len_needle; i++) {
if (needle[i] != window[i]) {
Py_ssize_t two_way_shift = i - cut + 1;
LOG("Right half does not match. two-way: %zd, gap: %zd "
"--> Advance by %zd\n",
two_way_shift, gap, Py_MAX(two_way_shift, gap));
LOG("Right half does not match.\n");
window_last += Py_MAX(two_way_shift, gap);
goto windowloop;
}
}
for (Py_ssize_t i = 0; i < cut; i++) {
if (needle[i] != window[i]) {
LOG("Left half does not match. period: %zd, gap: %zd)"
"--> Advance by %zd\n",
period, gap, Py_MAX(period, gap));
window_last += Py_MAX(period, gap);
LOG("Left half does not match.\n");
window_last += period;
goto windowloop;
}
}
LOG("Left half matches. Returning %zd.\n", window - haystack);
LOG("Found a match!\n");
return window - haystack;
}
}
Expand Down