Sender: itm AT dra DOT hmg DOT gb
Message-ID: <33097AB8.41C67EA6@dra.hmg.gb>
Date: Tue, 18 Feb 1997 09:47:36 +0000
From: Ian Miller <itmiller AT dra DOT hmg DOT gb>
Organization: DRA
MIME-Version: 1.0
To: DJGPP workers <djgpp-workers AT delorie DOT com>
Subject: Re: Exact timming II
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

*My* problem with uclock() is illustrated by the following C++ program

------------------------------------
#include <stdlib.h>
#include <time.h>
#include <iostream.h>

int main(int argc, char** argv)
{
	if (argc != 2)
	{
		cerr << "usage: test <number of iterations>\n";
		exit(1);
	}

	int n = atoi(argv[1]);

	uclock_t t0 = uclock();
	uclock_t t  = 0;
	for (int i = 0; i < n; i++)
	{
		t = uclock();
		if (t < t0)
		{
			cerr << i << " t/t0: "
			     << t << "/" << t0 << "\n";
		}
		t0 = t;
	}

	exit(0);
} // main(int, char**)
------------------------------------

Some examples of its output:

------------------------------------
E:\home\t2>uctest 1200
200 t/t0: -59013/6432
E:\home\t2>uctest 1200
287 t/t0: -55922/9218
E:\home\t2>uctest 1200
183 t/t0: -59539/5887
E:\home\t2>uctest 1200
E:\home\t2>uctest 1200
E:\home\t2>uctest 1200
E:\home\t2>uctest 1200
12 t/t0: -64616/411
------------------------------------

That was all in a Win95 DOS box. On further investigation
I found that there was no such problem in Win95's DOS mode
nor in my previous version of DOS. Hmmm.

So far as investigating why this is so, I tried the
following little program (heavily based on uclock.c)

------------------------------------
/* Copyright (C) 1996 DJ Delorie, see COPYING.DJ for details */
/* Copyright (C) 1995 DJ Delorie, see COPYING.DJ for details */
/* DJ's copyright acknowledged :-), ITM 1997 */
#include <time.h>
#include <pc.h>
#include <libc/farptrgs.h>
#include <go32.h>

int 
main(int argc, char** argv)
{
#define N 7
  unsigned char lsb, msb;
  unsigned long tics, otics;
  uclock_t rv[N];
  int i;

  for (i = 0; i < N; i++)
  {
    /* Make sure the numbers we get are consistent */
    do {
      otics = _farpeekl(_dos_ds, 0x46c);
      outportb(0x43, 0x00);
      lsb = inportb(0x40);
      msb = inportb(0x40);
      tics = _farpeekl(_dos_ds, 0x46c);
    } while (otics != tics);

    /* calculate absolute time */
    msb ^= 0xff;
    lsb ^= 0xff;
/*     rv[i] = ((uclock_t)tics << 16) | (msb << 8) | lsb; */
    rv[i] = (msb << 8) | lsb;
  }

  printf("%u\n", rv[0]);
  for (i = 1; i < N; i++)
  {
    printf("%+d\n", rv[i] - rv[i - 1]);
  }

  uclock();

  for (i = 0; i < N; i++)
  {
    /* Make sure the numbers we get are consistent */
    do {
      otics = _farpeekl(_dos_ds, 0x46c);
      outportb(0x43, 0x00);
      lsb = inportb(0x40);
      msb = inportb(0x40);
      tics = _farpeekl(_dos_ds, 0x46c);
    } while (otics != tics);

    /* calculate absolute time */
    msb ^= 0xff;
    lsb ^= 0xff;
/*     rv[i] = ((uclock_t)tics << 16) | (msb << 8) | lsb; */
    rv[i] = (msb << 8) | lsb;
  }

  printf("%u\n", rv[0]);
  for (i = 1; i < N; i++)
  {
    printf("%+d\n", rv[i] - rv[i - 1]);
  }

  return 0;
#undef N
}
------------------------------------

Win95 DOS box:

E:\home\t2>uc2
46271
+124
+60
+60
+60
+60
+60
58976
+43
+30
+29
+30
+29
+30

(very occasionally I get a large negative number instead of +30 or +29
 but that's probably just wrapping)

Reboot.
DOS prompt only:

40813
+38
+16
+18
+16
+16
+16
35
+19
+8
+9
+8
+8
+8

Reboot.
Previous version of DOS (6.x):

24023
+40
+16
+18
+16
+16
+18
36
+20
+8
+8
+8
+8
+8


It looks to me like the timer mode is changed in all three cases.
Does this get me anywhere?

Regards,
-- 
Ian Miller, Dorset, UK
DJGPP 2.01, Win95 DOS box (LFN undefined, FNCASE=y)