Investigate the bottleneck in the search benchmark using toplev

[fulmicoton]

Toplev seems very interesting to identify bottlenecks.

Let's use it and compare the results to our current champion PISA. We want different report for intersections, union, and phrase queries.

[scampi]

Can you develop what you expect in this issue ?

• What is toplev ? A web search didn't turn up anything obvious
• What is the bottleneck in question to be investigated ?
• What is PISA ? I see it mentioned in the changelog, is that the same thing ?

[fmassot]

@fulmicoton can you add a link for Toplev?

@scampi : PISA project is here https://github.com/pisa-engine/pisa

Concerning bottlenecks can be IO or CPU or both. Let's take the example of union term queries you will see that PISA is way faster in the benchmark https://tantivy-search.github.io/bench/ I believe this comes from the fact that their WAND algorithm uses a better data structure to exclude documents that will not make it to the top 10. But this will need to be confirmed for example.

[fmassot]

@scampi for toplev, see the manual here: https://github.com/andikleen/pmu-tools/wiki/toplev-manual