Re: New assembly language instructions to support OO languages?

Hans-Peter Diettrich wrote:
> Michael Tiomkin schrieb:
>
>
>> I understand that some ancient computers (in the 2nd millennium)
>>used segmented memory. If we could implement this and show that it
>>runs most of the programs faster, somebody would be able to build such
>>a processor.
>
>
> A segmented model requires possibly huge address translation tables,
> which have to reside in the cache. Unfortunately loading these tables
> on context switches (threads...) or on loading segment registers
> resulted in very poor runtime behaviour. The page concept (flat model)
> reduces the reloads to situations which require I/O (swap file), so
> that adjustments to the tables are neglectable.


The only way I can make sense of these (dubious) assertions is to assume
that you have a specific implementation in mind.


The system I work on is both segmented and paged. The Page Table
(note singular) consumes only a small fraction of memory. Being a
global Page Table, there is no penalty incurred when switching
threads. The time to load a segment register is a few cycles, and
doesn't seem to normally impact performance.


I can't make any sense of your last sentence no matter how I try.


David W. Schroth
[I expect he's thinking of the segmentation on x86 chips, which has rather
bad performance. Segment register loads are very slow, and since there
aren't very many seg registers, there were a lot of loads. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.