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 : Silicon Graphics, Inc. (SGI) -- Ignore unavailable to you. Want to Upgrade?

To: Stormweaver who wrote (3975)	1/14/1998 3:22:00 AM
From: Alexis Cousein	Respond to of 14451

> - most stable and standard UNIX, NCR to standardize on > Solaris now But recently Sequent switched to DEC Unix, probably becasue of 64-bit issues. >SGI - > - 64 bit already > - good graphic software + grahpic capability > - good low end price points > - SVR4 yet?? Since release 5.2 (about 3 years ago). - binary compat between arch's ??? Yes. I'm still running some apps I compiled on a Personal IRIS (1990) running 5.3 on my O2 (1997) ;). - binary compat between OS releases ??? Yes. All SVR4-based releases, that is. 4.0.5 used COFF binary format binaries; there was compatibility support for these up to 6.1. - threaded kernel ? Yes. - what thread standard supported?? That's question is, I think, obviously for threaded user apps, not kernel. POSIX pthreads, sequent-like sproc() based calls, and heavier process- based models with thread-like characteristics using pcreatel(). Some of the libraries (like BLAS et al) can automatically be threaded on MP systems without touching the application, and there have been autoparallelizers for C and Fortran for some time. - ? any good mid-range boxes (between Cray and o2000?) O2000 already expands a long way into what you would call Cray space (hence the branding Cray Origin2000 for these larger systems). You basically use 8-CPU building blocks ans stack them together; some customers are already using single-system-image 128 CPU boxes. Cray has a box (T3E) that's quite similar, but has CPUs and CPU to memory subsystem well tuned to the specific spaces of hand-tuned MPP number-crunching programs (different from the usual cache-based approach), up to 2048 CPUs, the interconnect is different (so that you can fit many CPUs in a smaller space). And Cray has small and large vector-processors, but they're different beasts altogether (the class of codes optimized for general-purpose CPUs and vector processors are different, though many codes are migrating to use general-purpose CPUs on SMP/ccNUMA).