Re: Not enough parallelism in programming

From: Andrew Reilly <andrew-newspost@xxxxxxxxxxxxxxxxxxxxx>
Date: Fri, 02 Sep 2005 10:58:47 +1000

On Fri, 02 Sep 2005 10:07:51 +1000, Andrew Reilly wrote:
> Doesn't it rely on there being "spare" processors sitting in idle states,
> with cache hot on the same data set, for this to be a win? How often is
> that likely?

Please pardon my stupidity: of course there are processors sitting idle.
The whole purpose of this search for extra parallelism is to make use of
them. Duh.

Cheers,

--
Andrew

.

References:
- Re: Not enough parallelism in programming
  - From: Scott A Crosby
- Re: Not enough parallelism in programming
  - From: Andrew Reilly

Prev by Date: Re: Not enough parallelism in programming
Next by Date: Re: Not enough parallelism in programming
Previous by thread: Re: Not enough parallelism in programming
Next by thread: Re: Not enough parallelism in programming
Index(es):
- Date
- Thread

Relevant Pages

Re: Track size and maximum single volume data set size
... Track size and maximum single volume data set size ... > From: IBM Mainframe Discussion List ... > to the PHYSICAL geometry even with emulation, RAID, striping, cache, ... cache that the physical I/O is not significant to the response time of ...
(bit.listserv.ibm-main)
Re: "A foolish consistancy" or "3390 cyl/track architecture"
... The track index has a pointer to the physical location of the ... cache is managed in full tracks. ... When Snapshot is used to copy a volume or a data set, ...
(bit.listserv.ibm-main)
Re: "A foolish consistancy" or "3390 cyl/track architecture"
... The track index has a pointer to the physical location of the ... cache is managed in full tracks. ... When Snapshot is used to copy a volume or a data set, ...
(bit.listserv.ibm-main)
Re: [PATCH 1/5] cpuset memory spread basic implementation
... If the job has a number of threads hitting the same data set and: ... If the cache is one that goes with a data set, ... current focus is on filesystem related objects: ... Is there anything else that needs to be spread? ...
(Linux-Kernel)
Re: [PATCH 1/5] cpuset memory spread basic implementation
... and not really a workload-specific access-pattern thing. ... It's job specific, and cache specific. ... If the job has a number of threads hitting the same data set and: ... then for that job spreading makes sense. ...
(Linux-Kernel)