Performance difference trilogy and pg on TechEmpower benchmarks

[p8]

For sequel there is a noticable difference in performance result for the TechEmpower benchmark between the pg and trilogy adapters. This difference doesn't seem to occur for the sinatra tests with ActiveRecord (the bottom two):

https://www.techempower.com/benchmarks/#section=test&runid=cda95417-edbb-4da3-ac87-10431d28b020&w=zik0zj-zik0zj-zik0xr-zik0zj-1&f=zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-hra0hr-yyku7z-zik0zj-zik0zj-zik0zj-zik0zj-1

To Reproduce:

Checkout the repo at: https://github.com/TechEmpower/FrameworkBenchmarks

Run the test with:

./tfb --test roda-sequel --type fortune
./tfb --test roda-sequel-postgres --type fortune

[jeremyevans]

I don't have any ideas off the top of my head. You'd probably have to profile to get a better idea. However, it's unsurprising to me that there are performance differences. Considering it's a different database and different adapter code, I would not expect the same performance. Is there a reason you would expect the same performance?

I don't know why ActiveRecord has similar performance between the two cases, but it's substantially lower than Sequel's performance, so maybe there is a bottleneck in ActiveRecord that slows down both cases.

[p8]

Thanks for the suggestions.
I'll try to profile to maybe get a better idea.
With the performance enhancements in sequel_pg I was expecting postgres to be the faster option.
Also with Postgres being the database of choice for the top entries in the benchmark.

Active Record having similar performance being the outlier is an interesting take as well!