bfwolf Senior Member Joined: 03/01/2025 Location: GermanyPosts: 243
Posted: 08:33am 03 Apr 2026
I completely agree with you! AI is indeed "quite naive" in many respects.
To manage large datasets, it would certainly be better to keep only a portion of the data in RAM — for example, in a "window" buffering the current position of the table being processed. Or a more intelligent cache that can buffer multiple areas using an LRU strategy. Nevertheless, it's crucial to remember that the number of write operations on flash filesystems must be minimized. Unlike magnetic storage media or RAM, they don't survive an unlimited number of writes — not to mention that writes on flash are very slow. With magnetic storage media, writes are only "imperceptibly slower" and practically unlimited.
And of course, there are other conceivable use cases where a lot of information needs to be stored in RAM. Therefore, it would be fantastic to be able to control whether slow or fast RAM should be prioritized for specific variables. As an application programmer, you can assess whether certain variables are speed-critical or not.
Code overlays were one of the reasons why many programs were so terribly slow back then. I'll just mention the C64 with its floppy drive and the sluggish IEC serial bus. Of course, you were often just glad it worked at all.