[Calibration] LuIndex
Description
Part of #12.
- adjusted
default
andlarge
size to feed Wikipedia archive to Lucene - add
--linedoc
and--dirwalk
option to specify doc-per-line format and directory walk; - add uncompressed Wikipedia archive used for Lucene benchmark;
- modified benchmark harness;
- download and prepare Wikipedia data at build time (if checksum failed)
- currently referencing the Wikipedia data relative to scratch, this needs to be changed later with external data feature.
Performance Numbers
machines | default (sec) | large (sec) |
---|---|---|
Intel i7-6700k w. 16GB RAM (fisher.moma) | 10.115 | 469.390 |
AMD FX-8320 w. 8GB RAM (lynx.moma) | 17.808 | 845.633 |
Checklist before merge
-
what to do with the big data (8.5G); leave it as it is, or use Git LFS, or merge #11 in. -
squash into 1 comment.