Overview
| Comment: | smaz_generic: optimize compression
For some reason it seems even with -O3, calling Dict_Entry involves a string copy, which makes `memcpy` the larger time consumer of the copmpression algorithm. Inlining it manually improves performance a lot. |
|---|---|
| Timelines: | family | ancestors | descendants | both | trunk |
| Files: | files | file ages | folders |
| SHA1: |
674fadc74b1d6f25dc5e0b6f047e3cfe |
| User & Date: | nat on 2016-12-16 20:55:19 |
| Other Links: | manifest | tags |
Context
|
2016-12-17
| ||
| 22:48 | tools/smaz: add a roundtrip check option to help debug new code check-in: f44feb5e3e user: nat tags: trunk | |
|
2016-12-16
| ||
| 20:55 |
smaz_generic: optimize compression
For some reason it seems even with -O3, calling Dict_Entry involves a string copy, which makes `memcpy` the larger time consumer of the copmpression algorithm. Inlining it manually improves performance a lot. check-in: 674fadc74b user: nat tags: trunk | |
|
2016-12-15
| ||
| 20:12 | tools/smaz: use the new validation function instead of ad-hoc code check-in: 3a95d52c86 user: nat tags: trunk | |
Changes
Modified src/natools-smaz_generic.adb from [328ba1ead5] to [c2e4810a49]. [diff]