Spent 6 months last year ripping out an abstraction layer that made every request 40ms slower. We profiled, found the hot path, couldn't fix it without a rewrite. The "optimize later" school never tells you later sometimes means never

▲

tombert 2 hours ago | parent [-]

I'd say it usually means "never".

I also find it a bit annoying is that most people just make shit up about stuff that is "faster". Instead of measuring and/or looking at the compiled bytecode/assembly, people just repeat tribal knowledge about stuff that is "faster" with no justification. I find that this is common amongst senior-level people at BigCos especially.

When I was working in .NET land, someone kept telling me that "switch statements are faster" than their equivalent "if" statements, so I wrote a very straightforward test comparing both, and used dotpeek to show that they compile to the exact same thing. The person still insisted that switch is "faster", I guess because he had a professor tell him this one time (probably with more appropriate context) and took whatever the professor said as gospel.

▲

bluGill an hour ago | parent [-]

I've seen a lot of requests to obtimize code where we can measure the optimal versions saves a few nanoseconds. I just deleted some 'optimal code' that took a lot of mutexes and so was only faster when there is no contetion but in real world multi-writer situations the easy code wins. (Shared memory vs local socket for ipc)

▲

tombert an hour ago | parent [-]

I don't write a lot of super low level stuff, so maybe things are different there, but at least in the normal user space level I've found it pretty rare that explicit mutexes ever beat the performance of an (in my opinion) easier design using queues and/or something like ZeroMQ.

Generally I've found that the penalty, even without contention, is pretty minimal, and it almost always wins under contention.

	▲	bluGill an hour ago \| parent [-]
		To be fair the code in question was written many years ago - before anyone I know had heard of zeromq (it existed but wasn't known). it would be possible to optimize the mutexs out I'm sure - but the big problem wasn't speed it was complexity of code that I now maintain. Since local sockets are easier and faster I'm safe deleting the more complex code that should have never been written.