By: Jan Wassenberg (jan.wassenberg.delete@this.gmail.com), June 4, 2022 11:23 pm
Room: Moderated Discussions
Jukka Larja (roskakori2006.delete@this.gmail.com) on June 4, 2022 12:15 am wrote:
> Can there be any larger hindrance than having to rewrite/redesign? Most of the stuff (by lines of code or time
> spent writing the code) doesn't have enough data items to even theoretically be data-parallel. High percentage
> of code also doesn't live more than few days or weeks, so spending even little more time to write it than is
> necessary is generally a bad idea (this is likely a somewhat unusual aspect of game development though).
It seems to be a very different situation, so I understand your perspective that a compiler flag is the most that could be done.
From my perspective, it's fairly common that hot and fairly long-lived code is rewritten, because we have a lot of it, and software rarely gracefully scales beyond 10x, or when requirements change. And whenever such a rewrite happens, it's a good time to modernize code from 1990s style heavy on branches/lookup tables/virtual functions to data parallel.
> Can there be any larger hindrance than having to rewrite/redesign? Most of the stuff (by lines of code or time
> spent writing the code) doesn't have enough data items to even theoretically be data-parallel. High percentage
> of code also doesn't live more than few days or weeks, so spending even little more time to write it than is
> necessary is generally a bad idea (this is likely a somewhat unusual aspect of game development though).
It seems to be a very different situation, so I understand your perspective that a compiler flag is the most that could be done.
From my perspective, it's fairly common that hot and fairly long-lived code is rewritten, because we have a lot of it, and software rarely gracefully scales beyond 10x, or when requirements change. And whenever such a rewrite happens, it's a good time to modernize code from 1990s style heavy on branches/lookup tables/virtual functions to data parallel.