erict@flatline.UUCP (J. Eric Townsend) (04/12/89)
Ok, here's the situation: Pascal programming assignment: read in a file of around 1400 words, sort them into a binary search tree, with a linked list at each node keeping track of every line that that word exists on. After you've built the tree, read it out in order (from lowest to highest) and print out the linked list associated with each word. Pretty simple to code, actually, and it only took about 10K of source code. (w/ comments). In the time it takes this program to run on a 3b1 (w/ only kermit running in the background) I kermited the source to the vax at UH at 1200bps, compiled it, linked it ran it, and read the output (21K) at 1200 bps. I started writing this when I finished reading the output from UH's vax. The 3b1 just finished with it's output. Would somebody *PLEASE* figure out a way to get more ponies into my box?!?!?!?!? :-) -- "Enter, oh seeker of knowledge... That's *YOU*, fathead!" J. Eric Townsend Inet: cosc5fa@george.uh.edu 511 Parker #2 Houston,Tx,77007 EastEnders Mailing List: eastender@flatline.UUCP
alex@wolf.umbc.edu (Alex Crain) (04/13/89)
In article <509@flatline.UUCP>, erict@flatline.UUCP (J. Eric Townsend) writes: > In the time it takes this program to run on a 3b1 (w/ only kermit running > in the background) I kermited the source to the vax at UH at 1200bps, > compiled it, linked it ran it, and read the output (21K) at 1200 bps. Sounds like your pascal compiler to me. The 3b1 is no speed demon, but I've found that it benchmarks at about 90% of the speed of a VAXstation 2000. I've done supstancial programming in C and Lisp on this machine and found it to be generally faster then an 11/785 with 10 users. The real speed killer with this box is the single DMA line, which means that only one I/O device can happen at any given moment. THis makes the machine I/O bound, a real problem if you have a 9600 baud line on tty000, which will cause the cpu to interrupt itself to death. No fix for this that I know of. THe other problem is usually lack of memory. The difference between .5 meg and 2 meg is pretty amazing. I would encourage anyone to add all the memory that they can afford, and maybe a fast disk (if you don't have one now). More memory will prevent swapping, thus limiting disk I/O to file handleing, thus freeing up the DMA channel for the modem/serial line. :alex Alex Crain Systems Programmer alex@umbc3.umbc.edu Univ Md Baltimore County umbc3.umbc.edu!nerwin!alex
thad@cup.portal.com (Thad P Floryan) (04/14/89)
I concur with Alex' statements. My own tests show the 3b1 to be faster overall than a Mac ][ (except for floating point), and definitely more done per unit time than with the office VAX 11/780 in EVERY respect (running VMS 4.7 and 14MB RAM; the 3B1 is 3.5MB RAM with the 67MB HD). From your description of the speed (or, more properly, the lack thereof), it suggests that your PASCAL is an interpreter rather than a compiler. If you really want a fast PASCAL compiler, contact LPI; email me for the address. LPI also has excellent Fortran and other languages for the 3B1 (as shown at the recent UNIFORUM in San Francisco). Their products are NOT inexpensive; I'm still budgeting to buy their Fortran since I have some 30 years' worth of plotter stuff, math libraries, etc. that I want to run on the 3B1 (which I'm presently running on an Amiga 68020 using AbSoft's Fortran). Thad Floryan [thad@cup.portal.com (OR) ..!sun!portal!cup.portal.com!thad]
root@arakis.UUCP (Hans Jespersen) (04/19/89)
In article <17137@cup.portal.com> thad@cup.portal.com (Thad P Floryan) writes: >If you really want a fast PASCAL compiler, contact LPI; email me for the >address. LPI also has excellent Fortran and other languages for the 3B1 (as Anyone know how the LPI stuff compares to AT&T SVS-Pascal and SVS-Fortran. Considering AT&T often co-labels software, perhaps these products are one and the same ? I have been happy with my SVS-Pascal, but then again I have nothing to compare it with. The cool thing with this baby is that I can compile (COFF) .o object code and link it in with my C programs (and visa-versa). -- Hans Jespersen UUCP: uunet!attcan!nebulus!arakis!hans or ..!attcan!hjespers