| 1 |
largeNbDicts |
| 2 |
===================== |
| 3 |
|
| 4 |
`largeNbDicts` is a benchmark test tool |
| 5 |
dedicated to the specific scenario of |
| 6 |
dictionary decompression using a very large number of dictionaries. |
| 7 |
When dictionaries are constantly changing, they are always "cold", |
| 8 |
suffering from increased latency due to cache misses. |
| 9 |
|
| 10 |
The tool is created in a bid to investigate performance for this scenario, |
| 11 |
and experiment mitigation techniques. |
| 12 |
|
| 13 |
Command line : |
| 14 |
``` |
| 15 |
largeNbDicts [Options] filename(s) |
| 16 |
|
| 17 |
Options : |
| 18 |
-z : benchmark compression (default) |
| 19 |
-d : benchmark decompression |
| 20 |
-r : recursively load all files in subdirectories (default: off) |
| 21 |
-B# : split input into blocks of size # (default: no split) |
| 22 |
-# : use compression level # (default: 3) |
| 23 |
-D # : use # as a dictionary (default: create one) |
| 24 |
-i# : nb benchmark rounds (default: 6) |
| 25 |
--nbBlocks=#: use # blocks for bench (default: one per file) |
| 26 |
--nbDicts=# : create # dictionaries for bench (default: one per block) |
| 27 |
-h : help (this text) |
| 28 |
|
| 29 |
Advanced Options (see zstd.h for documentation) : |
| 30 |
--dedicated-dict-search |
| 31 |
--dict-content-type=# |
| 32 |
--dict-attach-pref=# |
| 33 |
``` |