d702044bcd
merge
2015-07-15 23:52:29 -07:00
a3626bc5bd
Issue #24583 : Fix crash when set is mutated while being updated.
2015-07-15 23:50:14 -07:00
5d2385ff6f
Neaten-up a little bit.
2015-07-08 11:52:27 -07:00
9632a7d735
Issue 24581: Revert c9782a9ac031 pending a stronger test for mutation during iteration.
2015-07-07 15:29:24 -07:00
11ce8e6c37
Minor bit of factoring-out common code.
2015-07-06 19:08:49 -07:00
3dbc11cadd
Tighten-up code in the set iterator to use an entry pointer rather than indexing.
2015-07-06 19:03:01 -07:00
ef6bd7d963
Tighten-up code in set_next() to use an entry pointer rather than indexing.
2015-07-06 08:43:37 -07:00
b48d6a63ff
Bring related functions add/contains/discard together in the code.
2015-07-05 16:27:44 -07:00
73799b181e
Change add/contains/discard calls to pass the key and hash instead of an entry struct.
2015-07-05 16:06:10 -07:00
dc28d5a198
Clean-up call patterns for add/contains/discard to better match the caller's needs.
2015-07-05 10:03:20 -07:00
ac2ef65c32
Make the unicode equality test an external function rather than in-lining it.
...
The real benefit of the unicode specialized function comes from
bypassing the overhead of PyObject_RichCompareBool() and not
from being in-lined (especially since there was almost no shared
data between the caller and callee). Also, the in-lining was
having a negative effect on code generation for the callee.
2015-07-04 16:04:44 -07:00
e186c7674c
Make sure the dummy percentage calculation won't overflow.
2015-07-04 11:28:35 -07:00
c2480dc0c4
Minor cleanup.
2015-07-04 08:46:31 -07:00
b322326f48
Minor nit: Make the style of checking error return values more consistent.
2015-07-03 23:37:16 -07:00
4897300276
Minor factoring: move redundant resize scaling logic into the resize function.
2015-07-03 20:00:03 -07:00
3c1f52e829
Call set_lookkey() directly to avoid unnecessary memory spills and reloads.
2015-07-03 18:31:09 -07:00
15f0869609
Move insertion resize logic into set_insert_key().
...
Simplifies the code a little bit and does the resize check
only when a new key is added (giving a small speed up in
the case where the key already exists).
Fixes possible bug in set_merge() where the set_insert_key()
call relies on a big resize at the start to make enough room
for the keys but is vulnerable to a comparision callback that
could cause the table to shrink in the middle of the merge.
Also, changed the resize threshold from two-thirds of the
mask+1 to just two-thirds. The plus one offset gave no
real benefit (afterall, the two-thirds mark is just a
heuristic and isn't a precise cut-off).
2015-07-03 17:21:17 -07:00
2eff9e9441
Minor refactoring. Move reference count logic into function that adds entry.
2015-06-27 22:03:35 -07:00
91672617d5
Minor tweeak to tighten the inner-loop.
2015-06-26 02:50:21 -07:00
38bb95e49d
Minor code cleanup.
2015-06-24 01:22:19 -07:00
7e3592dca6
Harmonize the bottom of the outer loop with its entry point
...
giving a small simplification. Timings show that hash
pre-check seems only benefit the inner-loop (the linear probes).
2015-06-21 10:47:20 -07:00
6ee588f14e
Restore quick exit (no freeslot check) for common case (found null on first probe).
2015-06-20 21:39:51 -07:00
7aa5341164
Reverting my previous commit.
...
Something went horribly wrong when I was doing `hg rebase`.
2015-05-30 10:57:56 -04:00
e90982111a
Issue #24115 : Update uses of PyObject_IsTrue(), PyObject_Not(),
...
PyObject_IsInstance(), PyObject_RichCompareBool() and _PyDict_Contains()
to check for and handle errors correctly.
2015-05-30 17:48:54 +03:00
ac5569b1fa
Issue #24115 : Update uses of PyObject_IsTrue(), PyObject_Not(),
...
PyObject_IsInstance(), PyObject_RichCompareBool() and _PyDict_Contains()
to check for and handle errors correctly.
2015-05-30 17:48:19 +03:00
fa494fd883
Issue #24115 : Update uses of PyObject_IsTrue(), PyObject_Not(),
...
PyObject_IsInstance(), PyObject_RichCompareBool() and _PyDict_Contains()
to check for and handle errors correctly.
2015-05-30 17:45:22 +03:00
8651a50475
Issue #23359 : Specialize set_lookkey intoa lookup function and an insert function.
2015-05-27 10:37:20 -07:00
5af9e13c18
Minor stylistic and consistency cleanup.
2015-05-13 01:44:36 -07:00
1bd8d75be3
Issue #23290 : Optimize set_merge() for cases where the target is empty.
...
(Contributed by Serhiy Storchaka.)
2015-05-13 01:26:14 -07:00
438f9134cf
Mirco-optimizations to reduce register spills and reloads observed on CLANG and GCC.
2015-02-09 06:48:29 -06:00
8249282622
Minor code clean up.
2015-02-04 08:37:02 -08:00
06bb1226d1
Issue 23359: Reduce size of code in set_lookkey. Only do linear probes when there is no wrap-around.
...
Nice simplification contributed by Serhiy Storchaka :-)
2015-02-03 08:15:30 -08:00
c658d85487
Issue 23359: Tighten inner search loop for sets (don't and-mask every entry lookup).
2015-02-02 08:35:00 -08:00
59ecabd12a
Keep the definition of i consistent between set_lookkey() and set_insert_clean().
2015-01-31 02:45:12 -08:00
9edd753229
Minor tweak to improve code clarity.
2015-01-30 20:09:23 -08:00
06a1c8dfa0
Fix typo in a comment.
2015-01-30 18:02:15 -08:00
f8d1a31e70
Revert unintended part of the commit (the key==dummy test wasn't supposed to change).
2015-01-26 22:06:43 -08:00
a5ebbf6295
Remove unneeded dummy test from the set search loop (when the hashes match we know the key is not a dummy).
2015-01-26 21:54:35 -08:00
3037e84ad1
Issue #23269 : Tighten search_loop in set_insert_clean()
...
Instead of masking and shifting every loopup, move the wrap-around
test outside of the inner-loop.
2015-01-26 21:33:48 -08:00
b335dfe7fa
Set the hash values of dummy entries to -1. Improves quality of entry->hash == hash tests.
2015-01-25 16:38:52 -08:00
4d45c1069b
Update out-of-date comments.
2015-01-25 16:27:40 -08:00
93035c44fd
Issue #23119 : Simplify setobject by inlining the special case for unicode equality testing.
2015-01-25 16:12:49 -08:00
ed741d4ff0
A hybrid of and-masking and a conditional-set-to-zero produce even faster search loop.
2015-01-18 21:25:15 -08:00
bd9b200b87
Update copyright for 2015 updates.
2015-01-18 16:10:30 -08:00
9cd6a789c6
Clean-up, simplify, and slightly speed-up bounds logic in set_pop().
...
Elsewhere in the setobject.c code we do a bitwise-and with the mask
instead of using a conditional to reset to zero on wrap-around.
Using that same technique here use gives cleaner, faster, and more
consistent code.
2015-01-18 16:06:18 -08:00
1202a4733e
Issue 23261: Clean-up the hack to store the set.pop() search finger in a hash field instead of the setobject.
2015-01-18 13:12:42 -08:00
8edf27c134
Small clean-up. Factor-out common code for add, contains, and discard function pairs.
2014-12-26 23:08:58 -08:00
08e3dc0ad6
Issue #23107 : Tighten-up loops in setobject.c
...
* Move the test for an exact key match to after a hash match
* Use "used" as a loop counter instead of "fill"
* Minor improvements to variable names and code consistency
2014-12-26 20:14:00 -08:00
12174a5dca
Issue #22156 : Fix "comparison between signed and unsigned integers" compiler
...
warnings in the Objects/ subdirectory.
PyType_FromSpecWithBases() and PyType_FromSpec() now reject explicitly negative
slot identifiers.
2014-08-15 23:17:38 +02:00
426d9958a2
Add development comments to setobject.c
2014-05-18 21:40:20 +01:00
6ba5665fc7
Fix typo in comment.
2014-01-14 08:15:03 -05:00
74fc8c47f6
Add comments to frozenset_hash().
...
Also, provide a minor hint to the compiler on how to group the xors.
2014-01-05 12:00:31 -08:00
e259f13874
Minor code clean-up. Keep the C-API all in one section.
2013-12-15 11:56:14 -08:00
710a67edfc
Note that LINEAR_PROBES can be set to zero.
2013-09-21 20:17:31 -07:00
4ef0528b97
Minor beautification. Put updates and declarations in a more logical order.
2013-09-21 15:39:49 -07:00
0ce1953bf7
When LINEAR_PROBES=0, let the compiler remove the dead code on its own.
2013-09-21 14:07:18 -07:00
c70a2b7bb9
Make the linear probe sequence clearer.
2013-09-21 14:02:55 -07:00
8408dc581e
Issue 18771: Make it possible to set the number linear probes at compile-time.
2013-09-15 14:57:15 -07:00
742d8716ff
Put the defines in the logical section and fix indentation.
2013-09-08 00:25:57 -07:00
583cd03fd1
Minor code beautification.
2013-09-07 22:06:35 -07:00
4ea9080da9
Improve code clarity by removing two unattractive macros.
2013-09-07 21:01:29 -07:00
8f8839e10a
Remove the freelist scheme for setobjects.
...
The setobject freelist was consuming memory but not providing much value.
Even when a freelisted setobject was available, most of the setobject
fields still needed to be initialized and the small table still required
a memset(). This meant that the custom freelisting scheme for sets was
providing almost no incremental benefit over the default Python freelist
scheme used by _PyObject_Malloc() in Objects/obmalloc.c.
2013-09-07 20:26:50 -07:00
04fd9dd52b
Small rearrangement to bring together the three functions for probing the hash table.
2013-09-07 17:41:01 -07:00
ae7b00e2d3
Move the overview comment to the top of the file.
2013-09-07 15:05:00 -07:00
c56e0e3980
Minor touchups.
2013-09-02 16:32:27 -07:00
69492dab07
Factor-out the common code for setting a KeyError.
2013-09-02 15:59:26 -07:00
a35adf5b09
Instead of XORed indicies, switch to a hybrid of linear probing and open addressing.
...
Modern processors tend to make consecutive memory accesses cheaper than
random probes into memory.
Small sets can fit into L1 cache, so they get less benefit. But they do
come out ahead because the consecutive probes don't probe the same key
more than once and because the randomization step occurs less frequently
(or not at all).
For the open addressing step, putting the perturb shift before the index
calculation gets the upper bits into play sooner.
2013-09-02 03:23:21 -07:00
6c3c1ccd1b
Update copyright.
2013-08-31 21:34:24 -07:00
95c0d67581
Further reduce the cost of hash collisions by inspecting an additional nearby entry.
2013-08-31 21:27:08 -07:00
afe890923f
Tighten-up the lookkey() logic and beautify the code a bit.
...
Use less code by moving many of the steps from the initial
lookup into the main search loop.
Beautify the code but keep the overall logic unchanged.
2013-08-28 20:59:31 -07:00
9d95254bb7
Issue #18772 : fix the gdb plugin after the set implementation changes
2013-08-24 21:07:07 +02:00
bfc1e1a9cd
Add the same dummy type that is used in dictionaries.
2013-08-23 03:22:15 -05:00
fcf3b500ba
Issue 18797: Remove unneeded refcount adjustments for dummy objects.
...
It suffices to keep just one reference when the object is created.
2013-08-22 08:20:31 -07:00
5bb1b1dd6f
Hoist the global dummy lookup out of the inner loop for set_merge().
2013-08-21 01:34:18 -07:00
929cbac307
Remove a redundant hash table probe (this was artifact from an earlier draft of the patch).
2013-08-20 23:03:28 -07:00
ae9e616a00
Issue 18772: Restore set dummy object back to unicode and restore the identity checks in lookkey().
...
The Gdb prettyprint plugin depended on the dummy object being displayable.
Other solutions besides a unicode object are possible. For now, get it
back up and running.
The identity checks in lookkey() need to be there to prevent the dummy
object from leaking through Py_RichCompareBool() into user code in the
rare circumstance where the dummy's hash value exactly matches the hash
value of the actual key being looked up.
2013-08-20 22:28:24 -07:00
3c0a4f5def
Issue18771: Reduce the cost of hash collisions for set objects.
2013-08-19 07:36:04 -07:00
07351a0449
Remove the else-clause because the conditions are no longer mutually exclusive.
2013-08-17 02:39:46 -07:00
237b34b074
Use a known unique object for the dummy entry.
...
This lets us run PyObject_RichCompareBool() without
first needing to check whether the entry is a dummy.
2013-08-17 02:31:53 -07:00
8ad3919577
Hoist the global "dummy" lookup outside of the reinsertion loop.
2013-08-15 02:18:55 -07:00
9ed5f27266
Issue #18722 : Remove uses of the "register" keyword in C code.
2013-08-13 20:18:52 +02:00
c629d4c9a2
Replace outdated optimization with clearer code that compiles better.
...
Letting the compiler decide how to optimize the multiply by five
gives it the freedom to make better choices for the best technique
for a given target machine.
For example, GCC on x86_64 produces a little bit better code:
Old-way (3 steps with a data dependency between each step):
shrq $5, %r13
leaq 1(%rbx,%r13), %rax
leaq (%rax,%rbx,4), %rbx
New-way (3 steps with no dependency between the first two steps
which can be run in parallel):
leaq (%rbx,%rbx,4), %rax # i*5
shrq $5, %r13 # perturb >>= PERTURB_SHIFT
leaq 1(%r13,%rax), %rbx # 1 + perturb + i*5
2013-08-05 22:24:50 -07:00
c86d7e989c
Silence compiler warning for an unused declaration
2013-08-04 12:00:36 -07:00
5e946bacef
Fix compilation warning with gcc 4.8 (unused typedef)
2013-06-18 23:28:18 +02:00
c2176e46d7
Fix the internals of our hash functions to used unsigned values during hash
...
computation as the overflow behavior of signed integers is undefined.
NOTE: This change is smaller compared to 3.2 as much of this cleanup had
already been done. I added the comment that my change in 3.2 added so that the
code would match up. Otherwise this just adds or synchronizes appropriate UL
designations on some constants to be pedantic.
In practice we require compiling everything with -fwrapv which forces overflow
to be defined as twos compliment but this keeps the code cleaner for checkers
or in the case where someone has compiled it without -fwrapv or their
compiler's equivalent.
Found by Clang trunk's Undefined Behavior Sanitizer (UBSan).
Cleanup only - no functionality or hash values change.
2012-12-10 18:32:53 -08:00
27cbcd6241
Fix the internals of our hash functions to used unsigned values during hash
...
computation as the overflow behavior of signed integers is undefined.
In practice we require compiling everything with -fwrapv which forces overflow
to be defined as twos compliment but this keeps the code cleaner for checkers
or in the case where someone has compiled it without -fwrapv or their
compiler's equivalent.
Found by Clang trunk's Undefined Behavior Sanitizer (UBSan).
Cleanup only - no functionality or hash values change.
2012-12-10 18:15:46 -08:00
0e1af282b8
Fix typo.
2012-09-28 16:43:40 +03:00
49526f48fc
Issue #14785 : Add sys._debugmallocstats() to help debug low-level memory allocation issues
2012-06-22 14:55:41 -04:00
a701388de1
Rename _PyIter_GetBuiltin to _PyObject_GetBuiltin, and do not include it in the stable ABI.
2012-04-05 00:04:20 +02:00
31668b8f7a
Issue #14288 : Serialization support for builtin iterators.
2012-04-03 10:49:41 +00:00
093ce9cd8c
Issue #6695 : Full garbage collection runs now clear the freelist of set objects.
...
Initial patch by Matthias Troffaes.
2011-12-16 11:24:27 +01:00
1cebc207ea
merge 3.2
2011-10-30 14:24:59 -04:00
2b50a01d11
remove unused variable
2011-10-30 14:24:44 -04:00
c34f5c256a
Fix the return value of set_discard (issue #10519 )
2011-10-30 14:35:39 +02:00
e0aa803714
Fix the return value of set_discard (issue #10519 )
2011-10-30 14:35:12 +02:00
7c5e34d8a3
Avoid unnecessary recursive function calls (#closes #10519 )
2011-10-30 13:57:45 +02:00
5acc27ebe4
Avoid unnecessary recursive function calls ( closes #10519 )
2011-10-30 13:56:41 +02:00
bd928fef42
Rename _Py_identifier to _Py_IDENTIFIER.
2011-10-14 10:20:37 +02:00
1ee1b6fe0d
Use identifier API for PyObject_GetAttrString.
2011-10-10 18:11:30 +02:00
d63a3b8beb
Implement PEP 393.
2011-09-28 07:41:54 +02:00
Raymond Hettinger
Yury Selivanov
Serhiy Storchaka
Victor Stinner
Eric V. Smith
Antoine Pitrou
Gregory P. Smith
Ezio Melotti
David Malcolm
Kristján Valur Jónsson
Benjamin Peterson
Petri Lehtinen
Martin v. Löwis