Michael.Witbrock@cs.cmu.edu (03/02/90)
I do neural network research. That means I to an awful lot of calculations of the form x= sum(over i) w(i)*o(i) i.e. multiplies and sums. I don't care about savages or mandlebrot tests, and I'm willing to write assembly code for my inner loops. I just found out that I can buy an unbundled 2630 at edu discount if I act within the next month. EXACTLY HOW MANY FLOATING POINT OPERATIONS PER SECOND CAN A 2630 BOARD DO, IF THEY ARE EQUAL NUMBERS OF MULTIPLYS AND ADDS (they can be short floats, not doubles, if this helps). I need to know whether my typical run would be a week (acceptable) or a month (unacceptable) (and you thought ray tracing was bad!). Please reply with as much detail as possible on the floating point capabilities of the 2630, and please do so soon. regards. Michael p.s. What I really desire is one of these new IBM workstations - 25MFlops, $12000. pretty hot (and I never thought I'd say that about an IBM thing). But I haven't got $12000, so lets see what I can do with my amiga.
portuesi@tweezers.esd.sgi.com (Michael Portuesi) (03/02/90)
> p.s. What I really desire is one of these new IBM workstations - > 25MFlops, $12000. pretty hot (and I never thought I'd say that about an > IBM thing). But I haven't got $12000, so lets see what I can do with my > amiga. you mean that you can't get the CS department to buy one of the new IBM machines for you? Does the new IBM machine really do 25 MFlops, or is it just 25 MIPS? --M -- __ \/ Michael Portuesi / Silicon Graphics, Inc. / portuesi@sgi.com
aburto@marlin.NOSC.MIL (Alfred A. Aburto) (03/03/90)
In article <4ZvUba200XNI88ZUU8@cs.cmu.edu> Michael.Witbrock@cs.cmu.edu writes: >I do neural network research. That means I to an awful lot of >calculations of the form > >x= sum(over i) w(i)*o(i) > >i.e. multiplies and sums. I don't care about savages or mandlebrot >tests, and I'm willing to write >assembly code for my inner loops. > >I just found out that I can buy an unbundled 2630 at edu discount if I >act within the next month. > >EXACTLY HOW MANY FLOATING POINT OPERATIONS PER SECOND CAN A 2630 BOARD >DO, IF THEY ARE EQUAL NUMBERS OF MULTIPLYS AND ADDS (they can be short >floats, not doubles, if this helps). I need to know whether my typical >run would be a week (acceptable) or a month (unacceptable) (and you >thought ray tracing was bad!). > >regards. Michael > ----------- The 68882 @25 MHz can do typical Double Precision Scalar FADD/FSUB/FMUL and FDIV's at a rate of approximately 0.46 MFLOPS Peak --- This is for scalar operations with variables in FPU registers and fast 32-bit memory. When working with arrays, as you indicated, where the variables can not be assigned to FPU registers then the MFLOPS rating goes down (perhaps alot depending upon the code a compiler generates). The Livermore Loops program (LLOOPS) can give you a good estimate of MFLOPS rating in this case. I think the A2630 with 32-bit fast memory and a good 020 compiler (Like Lattice C 5.04) will achieve 0.3 MFLOPS or so when working with FP arrays. If you want high performance with FP arrays then you will need to use an 'Array Processor' which can do A[I] * B[I] in one operation for all elements for example --- this is the way to get the big MFLOPS ratings ... There might be a way to interface an 'Array Processor' to the Amiga, but I'm not aware of such a system ... Al Aburto aburto@marlin.nosc.mil
hue@netcom.UUCP (Jonathan Hue) (03/03/90)
In article <PORTUESI.90Mar2142138@tweezers.esd.sgi.com> portuesi@sgi.com (Michael Portuesi) writes: > >Does the new IBM machine really do 25 MFlops, or is it just 25 MIPS? They claim 27 to 41 MIPS based upon Dhrystone 1.1, a fairly useless benchmark. They also say 22.3 to 34.7 Specmarks, which I think is a useful number. As far as FP goes, they say 7.5 MFLOPs for the slowest boxes, 13 for the fastest. There's an RISC System/6000 in the corner where I work, I think it's a 520 (the cheapest, slowest deskside model) but there aren't any labels on it. I've been running rayshade (ray tracer posted to comp.sources.unix) on it and the NeXT on my desk. The NeXT is probably comparable to a 2630 (25MHz '882) and took 13 hours to render a scene. The IBM box took 2hours and 39 minutes for the same picture. I'm sure the OS and compiler we have is much older than what was used to run the benchmarks, and will be much improved when the OS finally ships, so these numbers will probably go up (who knows how much). Anyone porting rayshade to the Amiga? It should be pretty easy. I bet the same image takes about a week to render on a Amiga 500 with s/w floating point. I've already got some of the images converted to HAM for the Amiga, they are really very nice. -Jonathan
raz@picowatt.Sun.COM (Stephen -Raz- Berry) (03/05/90)
Someone who's name I lost writes: >> p.s. What I really desire is one of these new IBM workstations - >> 25MFlops, $12000. pretty hot (and I never thought I'd say that about an >> IBM thing). But I haven't got $12000, so lets see what I can do with my >Does the new IBM machine really do 25 MFlops, or is it just 25 MIPS? The benchmarks that were just released say 27Mips and 7Mflops for their base machine. Of course IBM still doesn't know what a real operating system is, so they gave it AIX. >\/ Michael Portuesi / Silicon Graphics, Inc. / portuesi@sgi.com ------ Stephen -Raz- Berry BIX: razberry Project Engineer - Sun Microsystems raz@sun.com
sparks@corpane.UUCP (John Sparks) (03/07/90)
raz@picowatt.Sun.COM (Stephen -Raz- Berry) writes: > Of course IBM still doesn't know what a real operating >system is, so they gave it AIX. A friend of mine jokes that they need a windowing system under AIX. He proposes calling it PANES. AIX and PANES. <groan> -- John Sparks | D.I.S.K. 24hrs 1200bps. Accessable via Starlink (Louisville KY) sparks@corpane.UUCP <><><><><><><><><><><> D.I.S.K. ph:502/968-5401 thru -5406 Build something that even a fool can use, and only a fool will want to use it.