andrew
5a132799a4
Add cycles_total
2024-08-15 15:13:00 -07:00
andrew
72469ebb6e
Erase along left spine. Not faster
2024-08-15 15:07:44 -07:00
andrew
6c79847a42
Add instructions_total for linux
2024-08-15 15:06:53 -07:00
andrew
405a2ca161
Fix typo
2024-08-15 13:52:51 -07:00
andrew
f93466316a
Pass in-tree reference to mergeWithChild
2024-08-15 13:52:06 -07:00
andrew
5626cd09d9
Add to corpus
2024-08-15 11:50:04 -07:00
andrew
41840220c3
Optimize version handling in mergeWithChild
2024-08-15 11:49:13 -07:00
andrew
7ff00e7846
Extract mergeWithChild to function
2024-08-15 11:40:52 -07:00
andrew
6242f40d48
Require that eraseBetween leave at least one child or entryPresent
2024-08-15 11:37:36 -07:00
andrew
403d70a1d3
Prefer not copying node in eraseBetween
...
If numChildren + entryPresent is enough, we don't have to copy even if
it would fit in a smaller node.
If we have to copy, we might as well use the smallest acceptable node
type.
2024-08-15 11:33:16 -07:00
andrew
9763452713
Separate beginIsPrefix path and simplify slightly
2024-08-15 11:29:15 -07:00
andrew
73d0593fca
Remove separate prefix write codepath for now
2024-08-14 21:29:43 -07:00
andrew
23c2a3e1c6
SIMD for eraseBetween (Node16)
2024-08-14 18:12:46 -07:00
andrew
a64e792964
Remove unused function
2024-08-14 17:40:04 -07:00
andrew
5e362d5330
Add to corpus
2024-08-14 17:37:18 -07:00
andrew
cc526cb6ba
Call eraseBetween on useAsRoot in addWriteRange
2024-08-14 17:08:55 -07:00
andrew
7e49888bec
More eraseBetween optimizations
2024-08-14 16:40:29 -07:00
andrew
e64ebabced
eraseBetween optimizations
2024-08-14 16:13:37 -07:00
andrew
1e34951a77
Fix use-of-uninit in eraseBetween (Node256)
2024-08-14 15:25:10 -07:00
andrew
baf64520d6
Have eraseBetween take in-tree node by reference
2024-08-14 15:04:11 -07:00
andrew
3499626127
Fix potential strict aliasing issues
2024-08-14 15:01:34 -07:00
andrew
b7f9084694
destroyTree -> eraseTree. Use freelist
2024-08-14 14:47:22 -07:00
andrew
4b82502946
Accept node by ref for eraseBetween again
2024-08-14 14:43:19 -07:00
andrew
68bbacb69a
Use getInTree in eraseBetween
2024-08-14 14:43:19 -07:00
andrew
3078845673
Fix nodes_released accounting
2024-08-14 14:43:19 -07:00
andrew
43f6126cc4
Add a missing assert, call to removeNode
2024-08-14 14:43:19 -07:00
andrew
b911d87d55
eraseBetween bug fixes
2024-08-14 14:43:19 -07:00
andrew
0c65a82b78
Separate codepath for prefix writes
...
Uses the newly-added eraseBetween
2024-08-14 14:43:19 -07:00
andrew
e024cb8291
Track entriesErased in destroyTree
2024-08-14 14:43:19 -07:00
andrew
0740dcad43
Strengthen checkMemoryBoundInvariants check for Node0
...
I think this is probably checked elsewhere, but let's check here too for
consistency
2024-08-14 14:28:28 -07:00
andrew
176df61321
Simplify handling of beginNode invalidation if begin is prefix of end
2024-08-14 14:26:13 -07:00
andrew
0a850f22e9
Update README benchmarks
2024-08-12 17:33:45 -07:00
andrew
479b39d055
Add to corpus
2024-08-12 17:28:19 -07:00
andrew
482408d725
Add GCOVR_EXCL_LINE to __builtin_unreachable
2024-08-12 17:07:55 -07:00
andrew
45995e3307
Update comments
2024-08-12 17:05:21 -07:00
andrew
359b0b29ff
Avoid function call if no partial key
2024-08-12 16:33:03 -07:00
andrew
54e47ebd40
Simplify insert loop
2024-08-12 16:25:12 -07:00
andrew
1c9dda68a6
Call consumePartialKey from getOrCreateChild
2024-08-12 16:24:05 -07:00
andrew
142455dd28
Move consumePartialKey, and allow empty partial key
2024-08-12 16:18:58 -07:00
andrew
567d385fbd
WIP create child in getOrCreateChild
2024-08-12 16:11:16 -07:00
andrew
8a44055533
Consume first byte in insert iteration
2024-08-12 15:39:09 -07:00
andrew
62516825d1
Try specifying filter multiple times
2024-08-12 15:26:36 -07:00
andrew
3d592bd6a9
Move longestCommonPrefix to its own file
2024-08-12 15:10:05 -07:00
andrew
f5f5fb620b
Run gc at 200%
...
150% pessimized the "real data" benchmark
2024-08-12 10:48:24 -07:00
andrew
e3d1b2e842
Add to corpus
...
After a bunch of fuzzing on "udon" (my new zen4 machine)
2024-08-09 19:48:22 -07:00
andrew
9f8800af16
Add more to corpus (from fuzzing on osx)
...
Maybe we should just accumulate the corpus instead of replacing it? That
should be easier on git right?
2024-08-09 18:06:54 -07:00
andrew
182c065c8e
Insert common prefix in addWriteRange
...
This allows us to use our optimized implementation for setting max
version along the search path instead of a one-off loop
2024-08-09 14:57:04 -07:00
andrew
2dba0d5be3
Have insert return a pointer to the in-tree pointer
2024-08-09 13:58:31 -07:00
andrew
a1dfdf355c
Use metrics to count change in entry count
...
This lets us run gc slower safely
2024-08-09 13:44:49 -07:00
andrew
15919cb1c4
Add range writes to server_bench
2024-08-09 13:43:24 -07:00
andrew