Dude, optimization is saved until near the very end of development, that's practically software design 101; optimizing prematurely makes it significantly harder to make further changes without screwing things up.
Knuth needs to be keel-hauled for saying that. I wish he'd publicly apologise so people understand it just doesn't apply to modern cpu architecture.
For example, if DF had been designed from the start to be multithreaded that wouldn't be premature optimisation it would be a solid design decision taken at the correct time to hit the performance goals. Optimisation in modern software isn't unrolling loops or adding hex float tricks, it's laying out memory efficiently so everything always fits in the cache and minimising branches so the cpu never predicts incorrectly, and using memory on one core at a time. The only way to do those things is from the start deep in the design of the software, and it doesn't make it hard to make further changes it makes it easy to add new code that will perform well.
If you leave it to the end you have dozens of systems that all need to be relaid out and read each other's memory and write to each other's memory and it's effectively a total rewrite. And adding multithreading at the end is always, always, always the wrong decision. Toady has said he won't add it so that's not relevant to DF, but it does illustrate the point that "making things faster" is not what Knuth was talking about when he used the word "optimisation".
yeah ok, derailed. I have a thing about people quoting knuth.
Abalieno has said his bit, some people agree, some don't, some use passive aggressive comments to ask the thread to be locked. I feel I should reply to footkerchief as he spent the time replying to my post earlier. I think the gist of it was that we shouldn't say there was nothing done on fort mode, and that the mind rewrite was needed for job priorities.
Well yes, things were done on fort mode and the short hand of "nothing new in fort mode" is not true. But after two years having trees that drop inedible and unusable fruit is a bit of a disappointment still. And again on the job priorities, it's a disappointment because however much work went into the backend the feature isn't there. I personally find it hard to believe the entire mind needs rewriting to allow dwarves to pick their next job from a prioritised list rather than a hardcoded ordered list . But in any case that's two features that (to return to the OP) are noise in the game as they're confusingly incomplete.
This returns to the point raised earlier weighing lots of quick releases against one long release, in these cases we have the worst of both worlds where the release took forever but we have two incomplete half done features. You can see the questions in the forum already about "how do I eat apples?". As I've said, if Toady enjoys it and is making money that's all cool and I'll do something else until there's a release I'll enjoy playing. But raising the point that the dev priorities are giving the game a confusing feature set seems fair.
Author
Topic: Toady, a little rant on modern Dwarf Fortress (Read 51348 times)