a0f3a51

WIP: robin hood hashing

rciorba Aug 2, 2016

141c729

WIP: disable failing selftests because they break other tests

rciorba Aug 2, 2016

60f7501

cleanup + fix offset tracking on swap

rciorba Aug 4, 2016

1387b2d

update capacity based on new HASH_MAX_LOAD

rciorba Aug 21, 2016

e9d9f1a

use .99 HASH_MAX_LOAD

rciorba Sep 7, 2016

0ba49eb

add benchmark tests for ChunkIndex

rciorba Sep 7, 2016

689f93a

avoid creation of very large byte strings in benchmark

rciorba Sep 7, 2016

82d8ba0

extract slow parts of benchmark

rciorba Sep 8, 2016

e1a5c12

set MAX_HASH_LOAD to 0.93

rciorba Sep 8, 2016

21f881b

update hardcoded hashes in testsuite/hashindex.py

rciorba Sep 8, 2016

92bd12f

Merge remote-tracking branch 'origin/master' into robin_hood

rciorba Sep 8, 2016

71a4669

appease flake8

rciorba Sep 8, 2016

163c3a2

benchmark hash get/set with 2**23 keys

rciorba Sep 8, 2016

c07fcf1

add 1/3 misses in getitem benchmark

rciorba Sep 15, 2016

2d7925a

shortcut hashindex_set by having hashindex_lookup hint about address

rciorba Sep 26, 2016

8b01b32

fix compilation on arm

rciorba Dec 21, 2016

c11dda6

WIP: more rounds, fewer keys in benchmarks

rciorba Dec 17, 2016

075f8f8

WIP: move most of benchmark code in C reduce inconsistencies

rciorba Dec 19, 2016

a9390b1

actually fill the hashindex close to 93%

rciorba Dec 27, 2016

cc001c7

lower key count and iterations since the benchmark is taking to long

rciorba Dec 27, 2016

261f758

fix bug in hashindex_set that never triggered RH swapping

rciorba Dec 29, 2016

a197a13

separate benchmarks for inserting and updating values

rciorba Dec 29, 2016

7909b82

make the RH bucket swap thread safe

rciorba Dec 30, 2016

6e3ece1

remove commented out code

rciorba Dec 30, 2016

d7b4cd8

re-add missing keys to getitem benchmark

rciorba Dec 30, 2016

498a82f

rename all c benchmarks so it's easier to run them exclusively

rciorba Dec 31, 2016

63e7dbb

separate test for getitem without lookups for missing keys

rciorba Dec 31, 2016

7e0e8e5

benchmark multiple fill rates

rciorba Jan 8, 2017

e032ac4

WIP

rciorba Jan 8, 2017

13a8faa

measure more fill rates at 85 and 90%

rciorba Jan 8, 2017

4de8578

implement deletion for robin hood hashing

rciorba Jan 13, 2017

94781af

add benchmark for deletion + churn benchmark

rciorba Jan 15, 2017

151de70

actually call the churn function in the benchmark

rciorba Jan 15, 2017

64cedf6

fix segfault in c_delete benchmark

rciorba Jan 16, 2017

8031a55

move tmp_entry + entry_to_insert from stack to single alloc on heap

rciorba Jan 24, 2017

e6e6b32

only check for offset shortcut periodically when setting value

rciorba Jan 28, 2017

a8b528a

replace multiple memswap calls with single call to memmove

rciorba Feb 16, 2017

10314aa

add extra test for the hashindex

rciorba Feb 20, 2017

b66be6a

init the hashindex with same capacity in c benchmarks

rciorba Feb 20, 2017

4d8ab86

don't leak memory with the entry_to_insert+tmp_entry

rciorba Feb 21, 2017

340a624

fix bad skiphint in hashindex_lookup

rciorba Feb 21, 2017

3f3ab2d

fix bad setup in c benchmarks

rciorba Feb 21, 2017

e5092bc

delete commented out code in benchmark.py

rciorba Feb 22, 2017

91bec5e

run missing key lookup shortcut every 64 buckets

rciorba Feb 22, 2017

4ab6309

fix setitem benchmark to use an empty index

rciorba Feb 22, 2017

a968a62

implement shifting for deletion

rciorba Feb 25, 2017

a0c000d

wrap up delete memmove

rciorba Feb 27, 2017

4d1303a

remove a bunch of debug code

rciorba Feb 27, 2017

36bdbad

get rid of unused index->entry_to_insert

rciorba Feb 27, 2017

78be711

cleanup the benchmark code

rciorba Feb 27, 2017