3 ms·

> Wouldn't it be great if they consolidate with time on common api? Given how intrusive this one is compared to a conservative Boehm GC? No, especially not whe

by josefx 5y ago

> Wouldn't it be great if they consolidate with time on common api? Given how intrusive this one is compared to a conservative Boehm GC? No, especially not when every GC ends up inheriting design warts like finalize(). Moved my entire Java code away from finalize to the half dozen dedicated cleanup classes Java provided (that was before try with resources). > wouldn't it be beneficial to have system (kernel) provided gc? I can only think of giving the kernel the ability to force cleanup in low memory situations as a positive and for that you probably wouldn't have to put the entire GC into the kernel.

mirekrusin 5y ago
> I can only think of giving the kernel the ability to force cleanup in low memory situations as a positive and for that you probably wouldn't have to put the entire GC into the kernel. ...but isn't "low memory situation" the only situation when you'd have to actually do it in the first place (at least thoroughly/more expensively but bail once enough reclaimed)?
- dragontamer 5y ago
  > ...but isn't "low memory situation" the only situation when you'd have to actually do it in the first place Only in a throughput based collector. This here is a latency-optimized collector. It wants to run many, many small cleanups so that the main-application thread continues forward with the smallest amount of delay. Big cleanups require bigger delays.
  - mirekrusin 5y ago
    Yes, makes sense.