Seems the headline isn't accurate as it implied he is against liquid. May be better to use - Shopify/liquid: Performance: 53% faster parse+render, 61% fewer allocations.
-The autoresearch pattern; Is this like a brute force AI trial and error to performance improvement for agent?
Surprisingly he got the fewer object allocation right, but it isn't 53% faster as it would implied 1.53x the speed. In reality it is 53% reduction ( or half ) in rendering time which is close to 2x faster.
At the scale of Shopify, I wonder how much saving would make with a 2x faster rendering software. I am thinking if it would hit a million dollar per year.
>With GC consuming 74% of total CPU time
I guess there is still plenty of room for improvement.
-The autoresearch pattern; Is this like a brute force AI trial and error to performance improvement for agent?
- 53% faster combined parse+render time, , 61% fewer object allocations[1]
Surprisingly he got the fewer object allocation right, but it isn't 53% faster as it would implied 1.53x the speed. In reality it is 53% reduction ( or half ) in rendering time which is close to 2x faster.
At the scale of Shopify, I wonder how much saving would make with a 2x faster rendering software. I am thinking if it would hit a million dollar per year.
>With GC consuming 74% of total CPU time
I guess there is still plenty of room for improvement.
[1] https://github.com/Shopify/liquid/pull/2056