by tanelpoder on 4/22/25, 2:45 AM with 3 comments
by kristianp on 4/22/25, 3:51 AM
> I was honestly shocked when this simple change gave us an 8% speedup right off the bat.
Is that all they did to get 8% speedup? Could be a measurement error?
by rvz on 4/22/25, 3:44 AM
Also, If you needed to re-architect the entire codebase to solve a performance issue, either you chose one of the most inefficient technologies / languages or the code itself was badly architected in the first place or both.
Before any architectural changes to the codebase first check if you can get performance gains from the compiler flags and measure it. That should be the industry standard practice for high quality efficient software.
We must learn from excellent SWEs teams such as DeepSeek which frankly embarrassed the entire AI industry due to their performance optimizations and savings in inference usage.
by potato-peeler on 4/22/25, 3:01 PM
Young devs ignoring their seniors is a tale as old as time