I guess there's pros and cons to both methods and I'll concede that a badly programmed automated system could certainly interfere with your usual program flow. However, I still don't think I'm entirely incorrect.
I don't see much difference between stalling a program during AI calculations, or stalling a program after a loop has finished (and waiting for a VWait, or whatever). So long as the whole loop is still done within one 'frame' (whatever that may represent in any program's case), it really doesn't matter where in the loop the calculations were done.
Automated garbage collection has a much better idea of exactly how hard your computer is being worked at any given moment. If memory were to be freed up during some intense AI/physics calculations, it could well mean the difference between very smooth performance and one more type instance pushing vmem to extend itself. Or, using Mark's own example, for speed ups during load routines. Which I guess isn't so important really, unless streaming is involved.
I'm sorry to have to dispute, but I believe the opposite.
Say the garbage collector realises when the CPU is practically idleing and then does it's work. In fact, the most efficient gc would kick in when the sum of CPU usage for all threads in your program is less than 100%. That kind of accuracy can't be matched by human intervention.
I don't see much difference between stalling a program during AI calculations, or stalling a program after a loop has finished (and waiting for a VWait, or whatever). So long as the whole loop is still done within one 'frame' (whatever that may represent in any program's case), it really doesn't matter where in the loop the calculations were done.
Automated garbage collection has a much better idea of exactly how hard your computer is being worked at any given moment. If memory were to be freed up during some intense AI/physics calculations, it could well mean the difference between very smooth performance and one more type instance pushing vmem to extend itself. Or, using Mark's own example, for speed ups during load routines. Which I guess isn't so important really, unless streaming is involved.
It knows WHAT to flush, but not ideally when.
I'm sorry to have to dispute, but I believe the opposite.
Say the garbage collector realises when the CPU is practically idleing and then does it's work. In fact, the most efficient gc would kick in when the sum of CPU usage for all threads in your program is less than 100%. That kind of accuracy can't be matched by human intervention.