andersom@boulder.Colorado.EDU (ANDERSON MARC O) (08/15/90)
[1] - Is it just me (or my machine) or does it take all of you 3b1'ers
a half an hour to compile small stuff with g++? (gnu c++)
It took about 30 minutes to compile a simple program like:
main()
{
cout << "anarchy!!!\n";
}
which doesn't seem right... (oh yeah... with an #include <stream.h> at
the beginning)....
[2] - I have a 3b1 with a 20 meg miniscribe harddrive.... I just figured
out how to leach my brains out with FTP, and now have about 75 blocks
remaining on my system.... Anyway, would it be true to say that "getting
a faster hard drive would increase the speed on my unix system" since most
of unix is hard drive access rather than processing time? (I believe my
hd has about a 60 ms access time which is disgusting!!!).
Supposing I get say an 80 meg hard drive with 28 ms access time (any
recommendations for good 80 meggers that aren't too cheap, that would
work easily with the 3b1?), I still can run my 20 megger for mabey news
or arhives or something, correct? How would I do this? Would I need any of
those hardware upgrades that people have been talking about? Which ones would
I need?? (Does the one advertised by Brian support more than one drive,
does it only allow you to use more than eight heads?.. Or do you not even
need special hardware to run just two harddrives?).
Anyway, please bear with me... Although I have v3.51 of the operating
system, I have no documentation AT ALL. I was capable of replacing the
RTC battery in my system, but haven't exposed myself to much else of the
3b1 internal workings....
Any help would be very appreciated....
Marc
andersom@tramp.Colorado.EDU
lenny@icus.ICUS.COM (Lenny Tropiano) (08/16/90)
In article <24776@boulder.Colorado.EDU> andersom@tramp.Colorado.EDU (ANDERSON MARC O) writes: |>[1] - Is it just me (or my machine) or does it take all of you 3b1'ers |>a half an hour to compile small stuff with g++? (gnu c++) |> |>It took about 30 minutes to compile a simple program like: |> |>main() |>{ |> cout << "anarchy!!!\n"; |>} |> |>which doesn't seem right... (oh yeah... with an #include <stream.h> at |>the beginning).... |> |>[2] - I have a 3b1 with a 20 meg miniscribe harddrive.... I just figured |>out how to leach my brains out with FTP, and now have about 75 blocks |>remaining on my system.... Anyway, would it be true to say that "getting |>a faster hard drive would increase the speed on my unix system" since most |>of unix is hard drive access rather than processing time? (I believe my |>hd has about a 60 ms access time which is disgusting!!!). [...] Well the problem is two fold. So here's your two answers. Yes, GCC and G++ compiler is quite LARGE. Check out that binary size ... 1039 -rwxr-xr-x 1 gnu gnu 527024 Feb 17 12:43 gcc-cc1 That's >1/2mb executable! gcc-ccplus (or whatever it's called) has to be as large, if not larger. Now, that takes a long time to be called, and with your 20mb (~65ms seek time) hard drive that's a long time to be read in. Now if my assumptions are right, the 20mb drive probably comes on your 3B1 with either 512K (eeek) or 1MB of memory. Either way that's major swapping ... yes, more disk writing! |>Supposing I get say an 80 meg hard drive with 28 ms access time (any |>recommendations for good 80 meggers that aren't too cheap, that would [...] Yes, a faster drive WILL help, in fact you'll feel like you have an all new machine after it! :-) Another help would be memory, to cut down on the swapping. With my 4MB machine and 22ms seek time drive things hummm right along. |>work easily with the 3b1?), I still can run my 20 megger for mabey news |>or arhives or something, correct? How would I do this? Would I need any of |>those hardware upgrades that people have been talking about? Which ones would [...] I'll mail you information on this... The net has seen it many times over! :-) For others who want it ... you can request the information on the ICUS 2nd HD drive plans from OSU-CIS in ~/att7300/ICUS_HD2.Z or use the BITFTP server to request it by mail... |>andersom@tramp.Colorado.EDU -Lenny -- | Lenny Tropiano ICUS Software Systems lenny@icus.ICUS.COM | | {ames,pacbell,decuac,sbcs,hombre,rayssd}!icus!lenny attmail!icus!lenny | +------ ICUS Software Systems -- PO Box 1; Islip Terrace, NY 11752 ------+
dts@quad.sialis.mn.org (David T. Sandberg) (08/16/90)
In article <24776@boulder.Colorado.EDU> andersom@tramp.Colorado.EDU (ANDERSON MARC O) writes: :[1] - Is it just me (or my machine) or does it take all of you 3b1'ers :a half an hour to compile small stuff with g++? (gnu c++) And shortly thereafter writes: :[2] - I have a 3b1 with a 20 meg miniscribe harddrive.... I just figured :out how to leach my brains out with FTP, and now have about 75 blocks :remaining on my system.... Your observed slowness of g++ compiles might be directly related to your dearth of free disk. Caveat: I've never used g++, so I don't know what kind of demands it makes on disk space. Nor do I know if this slowness was observed before your disk space was thus reduced. Still, it's a thought... -- \\ \ David Sandberg, consultant \\ // "Small hats!" / Richfield MN // \\ \ dts@quad.sialis.mn.org \\
bamford@cbnewsd.att.com (harold.e.bamford) (08/16/90)
In article <24776@boulder.Colorado.EDU> andersom@tramp.Colorado.EDU (ANDERSON MARC O) writes: > [...] Anyway, would it be true to say that "getting >a faster hard drive would increase the speed on my unix system" since most >of unix is hard drive access rather than processing time? [...] I used to have a 7300 with the stock 20MB hard disk with 80 ms access time. I upgraded to a 67 MB (formatted) with 28 ms and noted that EVERYTHING was faster. 1) ksh used to take about 5 seconds for 2nd instances (ie, not the first invocation of ksh) and after the upgrade, they were practically instantaneous. First instances were still slow, but I didn't care since I eventually linked /bin/ksh to /bin/sh and logging in invoked the first instance. Yes, with both old and new disks I had done a "chmod +t /bin/ksh" 2) large compilations (C) and troff jobs took about 1/3 of the original time. Suspiciously close to the ratio of new vs. old access times. 3) large nroff jobs ran about twice as fast. I was amazed (and pleased). -- Harold
esmith@goofy.apple.com (Eric Smith) (08/18/90)
In article <24776@boulder.Colorado.EDU> andersom@boulder.Colorado.EDU (ANDERSON MARC O) writes: [1] - Is it just me (or my machine) or does it take all of you 3b1'ers a half an hour to compile small stuff with g++? (gnu c++) It took about 30 minutes to compile a simple program like: main() { cout << "anarchy!!!\n"; } which doesn't seem right... (oh yeah... with an #include <stream.h> at the beginning).... Sounds like you don't have enough physical RAM. I hope you have more than 1 MB. One time I accidentally booted my 7300 without the combo board installed, so it had only .5 M. I'm told that earlier releases of the OS ran OK with that little memory, but 3.51 took almost half an hour to boot. The stock disk drives in 3B1s and 7300s are dog slow, so you want as much physical memory as possible to minimize swapping. -- Eric L. Smith Opinions expressed herein do not necessarily reflect those esmith@apple.com of my employer, friends, family, computer, or even me! :-)
andy@juno.caltech.edu (Andy Fyfe) (08/18/90)
In article <24776@boulder.Colorado.EDU> andersom@tramp.Colorado.EDU (ANDERSON MARC O) writes: >[1] - Is it just me (or my machine) or does it take all of you 3b1'ers >a half an hour to compile small stuff with g++? (gnu c++) The problem might well be "malloc". If you want to see what can happen, write a program that does do { p = malloc(4000); printf("malloc returns %x\n", p); } while (p != 0); and compare this program when linked with libc/shlib to this program linked with -lmalloc. At one point I had compiled g++ on my 3b1 (though I got rid of it when I realized it couldn't compile larger programs for lack of virtual address space) and I compared the times for the osu version, and the ones I compiled with the shlib, -lmalloc, and gnu malloc. The difference was, I believe about an hour versus 4 minutes. (Though the quicker ones got through 3 or 4 fewer source lines before giving me a virtual memory exhausted error.) If you want to link with -lmalloc and still use the shared library, you need a program like shcc or the shared ld (that I wrote, but based on shcc) which you can find in the osu archives as well. Andy Fyfe andy@csvax.cs.caltech.edu ....!ames!elroy!cit-vax!andy