Re: FIFOs.
Re: FIFOs.
- Subject: Re: FIFOs.
- From: "John C. Randolph" <email@hidden>
- Date: Mon, 20 Aug 2001 07:20:34 -0700
On Monday, August 20, 2001, at 07:00 AM, Ondra Cada wrote:
John,
John C. Randolph (JCR) wrote at Mon, 20 Aug 2001 06:45:10 -0700:
JCR> I want a FIFO implemented as a proper thread-safe ring buffer that
will
JCR> grow as necessary to prevent overflow, and avoid needless
shuffling of
JCR> memory as your example will do. I want this FIFO to be writeable
by one
JCR> thread while being read by another.
The "memory shuffling" is a roaring nonsense.
No it's not, it's a waste of cycles that grows proportionately with the
number of items in the array, and it's a waste of cycles that's likely
to be buried deep in a nested loop (at least in my code.) I've seen
apps at several customer sites that had to do things like finalize
objects for storage in a database before exiting, where there could
easily be a hundred thousand objects that had to be checked. You may
think the memory shuffling cost is trivial, but I assure you, it adds up.
There just might be one
application from 10000 which *MIGHT* have such demands that
NSMutableArray-based FIFO would be unuseable there, but even that I
doubt.
Not speaking of the fact that NSMutableArray *CAN BE* (and, by some
other's
message, actually _IS_) implemented as a ring buffer.
As for thread-safeness, well, add two more lines with locking. Still not
worth an API, by far.
Well then, is NSEnumerator worth having? After all, it's just a wrapper
around an index, isn't it?
-jcr
For every complex problem there is an answer that is clear, simple, and
wrong. -- H L Mencken
| References: | |
| | >Re: FIFOs. (From: Ondra Cada <email@hidden>) |