 andrew
|
593657c9aa
|
Remove duplicate code
|
2024-02-09 15:32:06 -08:00 |
|
|
andrew
|
d1ee3d8b69
|
Bring back nextSibling
|
2024-02-09 15:26:54 -08:00 |
|
|
andrew
|
80a79aab1f
|
Still need a better way identify left/right sidedness
|
2024-02-09 15:17:37 -08:00 |
|
|
andrew
|
f3e7279c2c
|
Do less work
|
2024-02-09 13:10:40 -08:00 |
|
|
andrew
|
403608c794
|
WIP just need to identify nodes on begin -> end path
|
2024-02-09 12:51:09 -08:00 |
|
|
andrew
|
8e918acc05
|
Remove dead code
|
2024-02-09 11:51:51 -08:00 |
|
|
andrew
|
d320fb6eee
|
Remove dead code
|
2024-02-08 14:59:27 -08:00 |
|
|
andrew
|
2bd1bf93a9
|
Check indices between left and right after they diverge
|
2024-02-08 14:50:53 -08:00 |
|
|
andrew
|
b018ccc3ae
|
Add TODO
|
2024-02-08 10:16:31 -08:00 |
|
|
andrew
|
d388f6488d
|
Remove dead code
|
2024-02-08 10:13:53 -08:00 |
|
|
andrew
|
968e03403d
|
Revert to linear range checking, but keep short-circuiting
|
2024-02-08 10:05:22 -08:00 |
|
|
andrew
|
2cbb6e0170
|
Short-circuiting, efficient checkRangeRead
|
2024-02-07 16:31:54 -08:00 |
|
|
andrew
|
9363d7866c
|
Specify maxVersion meaning
|
2024-02-07 15:33:15 -08:00 |
|
|
andrew
|
855b7fa069
|
Prepare for new invariant
|
2024-02-07 13:44:46 -08:00 |
|
|
andrew
|
98124d5493
|
Sublinear read range checking
Still slow though
|
2024-02-06 16:52:20 -08:00 |
|
|
andrew
|
6be23803a3
|
Add standalone bench target
|
2024-02-06 15:03:54 -08:00 |
|
|
andrew
|
9a80c96533
|
Short circuit point reads based on maxVersion
|
2024-02-06 14:38:39 -08:00 |
|
|
andrew
|
68239cd80b
|
Maintain nextSib in firstGeq
This is slower, but I think it's closer to what we want for our eventual bespoke
"fused-check-and-search" implementation.
|
2024-02-06 12:37:02 -08:00 |
|
|
andrew
|
03e9dc01fd
|
Implement firstGeq directly
|
2024-02-06 12:11:50 -08:00 |
|
|
andrew
|
d2c89f605f
|
Use rangeVersion indicating < instead of >
This should allow us to use firstGeq (which should be possible to make
more efficient), and generally make things nicer
|
2024-02-06 11:11:45 -08:00 |
|
|
andrew
|
0aa2f67f88
|
Check that all children have reachable present entries
|
2024-02-05 16:59:21 -08:00 |
|
|
andrew
|
d78afe0823
|
Interleave checks for point reads
This doesn't actually seem faster, but it should prepare us to implement
range reads non-naively. It probably should be faster. To be
investigated.
|
2024-02-05 16:44:57 -08:00 |
|
|
andrew
|
57ec97f2ee
|
Naive range reads implementation and test
|
2024-02-05 13:59:10 -08:00 |
|
|
andrew
|
a632d0cbbb
|
gcc fixes
|
2024-02-03 13:16:13 -08:00 |
|
|
andrew
|
81f178face
|
Remove unused code
|
2024-02-01 18:31:36 -08:00 |
|
|
andrew
|
20bd5c2fa6
|
More corpus tinkering. Missing coverage for Node256
|
2024-02-01 18:22:45 -08:00 |
|
|
andrew
|
22daf0b9cb
|
Improve range write performance
|
2024-02-01 16:52:22 -08:00 |
|
|
andrew
|
0eff5628bd
|
Fix lastLeq bug
|
2024-02-01 11:24:57 -08:00 |
|
|
andrew
|
96aca4a1da
|
Range write WIP
|
2024-01-31 17:06:59 -08:00 |
|
|
andrew
|
812eb27708
|
Return inserted physical node from insert
|
2024-01-31 16:22:34 -08:00 |
|
|
andrew
|
31a555c44c
|
Use bitset index for Node48 too
|
2024-01-31 16:04:03 -08:00 |
|
|
andrew
|
bafe1edfa4
|
Remove call to memcmp for partial keys
|
2024-01-31 14:32:25 -08:00 |
|
|
andrew
|
5a69ec82df
|
Partial keys seem to be working
|
2024-01-31 11:10:41 -08:00 |
|
|
andrew
|
220c5caf27
|
WIP compressed key. Tests fail
|
2024-01-30 17:31:44 -08:00 |
|
|
andrew
|
ecd549857b
|
Add printTree
|
2024-01-30 17:16:46 -08:00 |
|
|
andrew
|
906a007c3a
|
Fix use of uninitialized memory
|
2024-01-30 14:50:26 -08:00 |
|
|
andrew
|
f2c9267b48
|
Use a bitset to speed up Node256 scans
|
2024-01-30 14:21:30 -08:00 |
|
|
andrew
|
8edac4d811
|
Use c++20 std::count{l,r}_zero
|
2024-01-30 13:20:26 -08:00 |
|
|
andrew
|
3c9c5bd9fb
|
Add benchmarks for getChild{L,G}eq
|
2024-01-30 13:08:01 -08:00 |
|
|
andrew
|
eca1b9993a
|
Improve symbols test
Remove seed argument and check for disallowed undefined symbols
|
2024-01-30 11:46:03 -08:00 |
|
|
andrew
|
3735a43553
|
Add TestDriver class
|
2024-01-30 11:13:21 -08:00 |
|
|
andrew
|
47faa51a35
|
Add Internal.h
|
2024-01-30 10:49:49 -08:00 |
|
|
andrew
|
831088aad9
|
Leave in the faster version for now
|
2024-01-26 10:25:48 -08:00 |
|
|
andrew
|
b8de633ca1
|
Update maxVersion to max(maxVersion, writeVersion)
|
2024-01-26 10:25:32 -08:00 |
|
|
andrew
|
270f887584
|
Annotate unused functions
|
2024-01-25 15:31:48 -08:00 |
|
|
andrew
|
4d71cf5b75
|
Implement x86 getChildLeq/getChildGeq
|
2024-01-25 14:57:03 -08:00 |
|
|
andrew
|
b15bec6b38
|
SIMD for Node16 in lastLeq and firstGeq
|
2024-01-25 12:40:11 -08:00 |
|
|
andrew
|
35cf3f3132
|
MacOS fixes
|
2024-01-24 15:55:44 -08:00 |
|
|
andrew
|
d76ac2a254
|
Try exported symbols list for Apple
visibility annotations seem to be working for gcc
|
2024-01-24 15:18:00 -08:00 |
|
|
andrew
|
a6a2ad61d9
|
Export shared and static libraries
|
2024-01-24 15:10:06 -08:00 |
|