Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : All About Sun Microsystems -- Ignore unavailable to you. Want to Upgrade?

To: QwikSand who wrote (44356)	8/3/2001 9:31:54 AM
From: Charles Tutt	Respond to of 64865

I'm not an expert on the subject, but as I understand it parallelism can be exploited at the process, thread, and/or instruction levels (the sub-instruction level can also be done in certain contexts, particularly vector processing, but that's a whole different can of worms -- e.g. instruction pipelines; super-process parallelism leads to "server farms" and load leveling systems). Traditional SMP addresses primarily the process level, while the Itanium puts its focus on instruction level parallelism (and I personally think it will have a tough time of it, because even if an individual process or thread can be optimized by the compiler, that does nothing to address a mix of processes or threads, so that context switching can gum up the compiler's best intent, especially in systems with a heavy load of I/O interrupts; in other words, there's no way the compiler can get its "arms" around the whole dynamic system context in most systems of interest). Thread level parallelism has been handled in the past similarly to process level parallelism via "lightweight" context switching, but without much special recognition of such multithreading by the hardware. As I read the article, Sun will increasingly hone the processor to help with the thread level. Of course, they say it better. <g> All JMHO. Charles Tutt (TM)

To: QwikSand who wrote (44356)	8/3/2001 11:02:27 AM
From: Prognosticator	Read Replies (1) \| Respond to of 64865

Point B: Java & Solaris. P. (Clarification for those who are interested: thread level parallelism, which UltraSparc V, the result of MAJC, will make possible with up to 1024 processing cores per CPU, will drastically speed up Java and Solaris, both of which are inherently multi-threaded). In a typical web-server written in Java, there can be thousands of threads running on a single CPU. In a typical server installation of Solaris, there are thousands of threads performing O/S tasks on behalf of the other programs. Most of them will be blocked awaiting I/O, but those that contend for cycles will suddenly speed up to 100% of their CPU. No new programming skills required, especially for Java programmers. IMO USV is Sun's masterstroke, that will finally dispatch Itanic/CISC architectures to the dustbin of history. I don't care if it takes them until 2010 to fully realize the 1024 core chip, because once they do, they are talking about a 10THz CPU, thats Terra-Hertz, 1000 GHz. Intel will barely be doing 10GHz by that point The potential of this alone is why I am still buying SUNW. ).