Talk:Single program, multiple data

[Untitled]

Unlike SPMD, shared memory multiprocessing, also called symmetric multiprocessing

I guess should say:

Unlike distributed, shared memory multiprocessing, also called symmetric multiprocessing

No, SMP is typically not distributed. There used to be research into virtual shared memory on distributed systems, but I haven't heard about that anymore since the 90ies. 89.12.252.29 (talk) 06:18, 4 December 2016 (UTC)

Does "Single Process, Multiple Data" really make sense? IMHO only "Single Program, Multiple Data" makes sense. — Preceding unsigned comment added by 87.182.235.84 (talk) 12:44, 16 June 2011 (UTC)

Yes, it makes sense if you interpret Process as it is used in the Unix context: one address space in which many threads can run in parallel. But the term no longer appears in the page. 89.12.252.29 (talk) 06:18, 4 December 2016 (UTC)

broken external link

The external link to "Single Program Multiple Data stream" is broken. It is redirected to a new home page for Overture. But most of the documentation there is now in PDF format. I have not been able to fathom which part of the documentation the original link refers to. 89.12.252.29 (talk) 06:10, 4 December 2016 (UTC)

Over-defined?

I understand that the Frederica Darema citation is just that, a citation of her original definition. However, it seems to me that the "... but through synchronization directives ... self-schedule themselves" over-defines what SPMD should mean in the context of Flynn's taxonomy, where I think can just be something like "All processes start running the same program. They can independently diverge through control branches and operate on different data.". I don't think that if that is the case for a processor, but the processes do not have synchronization directives and/or inter-process self-scheduling it would not be SPMD. Maybe there could be a paragraph to make that clear. Tikayay (talk) 15:40, 28 July 2025 (UTC)