It was thus said that the Great Jonathan Katz once stated: If you don't mind. I would like to know the speed, since the original code was run at 33MHz. I need the code below compiled (at the highest optimization setting) and run. I would like the output from time % time a.out >output as well as the output from the program (so I can check it). Warning: The program will take some time to run, but hopefully, it'll finish in less than 10 hours [1]. (Also, if possible, if you could change MAX to 5000 [2] and run it a second time that would also be appreciated). -spc (Thank you very much if you can do this) [1] This is just a fraction of the original program that generates a single image. The full program calculated some 500 images and thus, took around a year to run. [2] The original MAX was 5000. #include <stdio.h> #include <stdlib.h> #include <string.h> #include <math.h> #define MAX 32768 #define DIM 500 int mainloop(const double,const double,const double,const double); int main(void) { double A; double B; for (B = 4.0 ; B >= -4.0 ; B -= 0.016) for (A = -4.0 ; A <= 4.0 ; A += 0.016) printf("%d\n",mainloop(A,B,4.0,4.0)); return(0); } unsigned int serial; unsigned int pix[DIM][DIM]; int mainloop(const double A,const double B,const double C,const double D) { double xn,xn1; double yn,yn1; int ix; int iy; int count; serial++; for (count = 0 , xn = yn = 0.5 ; count < MAX; count++) { xn1 = ((A * yn) + B) * xn * (1.0 - xn); yn1 = ((C * xn) + D) * yn * (1.0 - yn); xn = xn1; yn = yn1; if (xn == HUGE_VAL) return MAX-1; if (xn == -HUGE_VAL) return MAX-1; if (yn == HUGE_VAL) return MAX-1; if (yn == -HUGE_VAL) return MAX-1; if (xn < 0.0) continue; if (xn >= 1.0) continue; if (yn < 0.0) continue; if (yn >= 1.0) continue; ix = (int)(xn * DIM); iy = (int)(yn * DIM); if (pix[ix][iy] == serial) break; pix[ix][iy] = serial; } return(count-1); }

On 08/08/2013 22:19, Eric Smith wrote: On a 150MHz R5000 Indy, using the IRIX C compiler (which is normally a bit better than GCC on IRIX systems) I get the same output result and these times: real 2.47 user 2.35 sys 0.08 I can possibly dig out an R3000 Indigo but it'll be running IRIX 5.3 (IIRC) and I can also run it on an R10k Origin 2000 (with 16 CPUs though it'll only use one) -- but like Eric I wonder if Sean's code isn't doing quite what's wanted. -- Pete Peter Turnbull Network Manager University of York

On Thu, Aug 8, 2013 at 4:21 PM, Sean Conner <spc at conman.org> wrote: So about half of the output lines are 32767. $ time ./prog >output real 0m0.201s user 0m0.196s sys 0m0.003s $ grep -v 32767 output | wc -l 127187 I tried changing MAX to 256, sticking a PGM header on it ("P5 500 500 255"), converting it to a png, and viewing it, but it didn't seem very interesting. Maybe my expectations were too high.

It was thus said that the Great Eric Smith once stated: Sounds about right. The timings seem right for a modern system. The result, when drawn, will look like the image at http://boston.conman.org/2013/08/06.1 It's basically a map of chaotic attractors---more about it at http://boston.conman.org/2013/08/04.1 Generally, when I originally ran this code (or code like it---I'm having to recreate some of this) twenty-some years ago, it took a year to generate the 500 images. I'm just curious if I missed some obvious optimizations back then (and I think I did). -spc

It was thus said that the Great Pete Turnbull once stated: Is that in minutes? Because if it's seconds, I think I'm gonna cry ... -spc (Why do I get the feeling that's seconds?)

It was thus said that the Great Pete Turnbull once stated: Not really, since I didn't have access to such a machine at the time. Not bad at all, considering. -spc

On Thu, Aug 8, 2013 at 8:00 PM, Jonathan Katz <jon at jonworld.com> wrote: I tried a 2nd time compiling with -O3 for more optimization.... cuts the time down somewhat.. sgi 33# cc -o foo -O3 ./foo.c sgi 34# rm output.txt sgi 35# time ./foo > output.txt 37.6u 0.2s 0:37 99% sgi 36# wc -l output.txt 250000 output.txt One more test, writing to /dev/null which would be faster than writing to SCSI-I disks... sgi 37# time ./foo > /dev/null 37.5u 0.1s 0:37 100% Only .1s of a difference; I thought it would be larger. Now for the non-optimized executable... sgi 39# time ./foo.no-opt > /dev/null 56.9u 0.2s 0:57 99% a .2 second difference in CPU time. -- -Jon Jonathan Katz, Indianapolis, IN.

It was thus said that the Great Jonathan Katz once stated: *Sob* Yes. Man, when I wrote that code in the early 90s I really missed some optimizations, since what you just ran in one minute took about ten hours to run. Sigh. Thanks to all that helped. I shall now go sob quietly in the corner, thinking of the year it took that program to run ... -spc (... when it could have finished in ten hours ... )

It was thus said that the Great Jonathan Katz once stated: True, but at the time, writing that program was part of my job ... -spc