jejones@mcrware.UUCP (James Jones) (01/19/88)
Well...I took my own advice and instrumented the program to announce each step (creating/writing the file, reading the file, and finishing). (This is for that program that writes a 100K file and then reads it.) Results are: version with SS_SSIZ version without SS_SSIZ write 2:43 1:45 read 0:14 0:14 (This is done with a CoCo 3, using a Sardis Tech floppy controller and 80-track 6ms floppy.) 14 seconds to read 100K comes to a tad over 7K bytes/second. So...it appears to take *far* longer to write a file than to read it. The SS_SSIZ cuts out a big hunk by avoiding repeated trips to the allocation bitmap and the file descriptor sector, but I don't know where the rest of the time is going. It will be interesting to find out. James Jones [This message has been treated with lemon-freshened DISCLAIMO!]
jejones@mcrware.UUCP (James Jones) (01/19/88)
One big factor in the difference between read and write throughput is probably write verification. (Whether you turn it off is a matter for you to decide; I'm a bit too chicken to do it. <grin>) One should also probably experiment with interleave factors. James Jones
jejones@mcrware.UUCP (James Jones) (01/19/88)
Taking the bit between my teeth, the bull by the horns, the nail by the head, and so forth, but most importantly loading up the "dmode" utility from CIS (an RBF equivalent of "xmode") to make it easy to do, I formatted a fresh disk, turned verify off, and tried the by now very familiar program. Results, with the program set to do an SS_SSIZ setstat: it took 24 seconds to write the file and 18 seconds to read. (Why 18 instead of 14? I don't know; it may have had something to do with precisely where things happened to fall on the disk. With dmode it will be easy to turn verify off and then back on, so I can run the program on the very same floppy I used the other day. No, I won't post another message when I do. :-) So--hypothesis verified. (Recall that with verify on, it took 1:44 to write the file, a tad over four times as long.) The next question, which I don't know the answers to (so I'll shut up after asking them), is: in practice, how dangerous is it to run with verify turned off? James Jones (The above, as far as I can tell, doesn't contain any opinions, so how can they be the opinions of anybody or any organization? Also, if I strike you as highly ignorant of floppy disk operation, keep in mind that I'm a software person. :-)