|
07:43 pm

|
| |
Thursday, March 27th, 2008 |
Good news: people are still updating and bumping my thread on CUDA performance problems. Bad news: still no word from NVIDIA.Update 2008-03-28 10:49: NVIDIA's Simon Green responded, promising improvement in CUDA 2.0! On a completely different subject: Poll #1161673 Programs
Open to: All, detailed results viewable to: AllWhat's better? |
|
| |
| Comments |
Depends on the program. :P
Let's say programs that regular people use: no lives hang in the balance, and no expensive space probes are in peril.
I would actually say a blend: don't just rush it out the door, but try to work a few design kinks out. It doesn't need to be perfect on the first iteration, especially if it's either (a) backend stuff or (b) going to be auto-patching itself over time. There is something to be said for being first to market.
However, once it's out there, you should focus more on the design and making sure that's right. It'll save you grief and bug reports over the course of the software lifecycle.
laaaaaaaaaaaaaaaaaaaaaaaaaaaaazy
For an audience of programmers or even just tech-savvy people, a fast solution is probably fine. But in cases where you expect your users to include people who wouldn't be comfortable with unintuitive behavior, you want to get it right. You want to hide the computery guts from people if they don't want to see it. Electronic gaming is a good example: many PC games are released with bugs or even just naiive design decisions (to be fixed in patches, hopefully?), while console games are (ideally) complete out of the box. PC games often have a level of customizability or user-generated content that only "in-depth" players will take advantage of, while console games are usually a packaged experience. PC games draw from a market of people who have powerful computers and are willing to fiddle with settings, console games should be usable by anybody. | |