850102

DFRAG (Disk FRAGmentation) is a program that reads BITMAP.SYS
from a disk drive and prints a histogram of its analysis in terms
of lengths of runs of 1's (available blocks) and 0's (in-use).

The main FORTRAN program reads the entire bitmap into a byte array,
then counts runs of bytes of all zeroes, all ones, and mixed.
The current version (2.0G) attempts to extract information
from mixed bytes adjacent to 00 or FF bytes, and then does a
quick analysis of the remainder to extract information about
some combinations of single and double bits. The pseudo-log (base 2)
of the length of the run is used to establish the histogram bucket to be
incremented.  After the bitmap has been scanned, the histograms
are printed (individually scaled), in conjunction with arithmetic and
logarithmic means*, and the arithmetic median.  A blank line between histogram
lines indicates that one or more empty histogram buckets was skipped.
(* Logarithmic mean is the anti-log of the arithmetic mean of the pseudo-logs
of the run-lengths; as such it is a little more useful than the arithmetic mean
of the run-lengths, but has a quirk such that its value may be less than,
equal to, or greater than the median, depending on how many runs of length 1
were encountered (log(1) === 0), whereas arithmetic mean will always be
greater than or equal to the median.)  Output is paginated and time-stamped.

Two MACRO routines complete the package:

1) ACCBITMAP assigns a channel, and accesses BITMAP.SYS by its file-id (2,2,0);
   the statements
	; ***	.ASCII	/_/
		.ASCII	/$255$/
   should be alternately activated or commented out depending on operation
   in or out of a cluster.

2) L2X2 performs the pseudo-log2 function with a scaling factor of 4 (2**2):
	L2X2(I) = log2(I)*4 + [(I - (exp2(log2(I)))/(log2(I)*4)]
   but it's really not that complicated: it floats its integer parameter,
   subtracts the exponent bias (to regain the "hidden" bit), and shifts
   to return the scaled value.

   Examples:	 I   L2X2(I)
		 1	 0
		 2	 4
		 3	 6
		 4	 8
		 5	 9
		 6	10
		 7	11
		 8	12
		 9	12
		10	13

   It has potential application for histogram scaling also, where the
   bins are linear but have an exponential range of values.

Generation

	FORTRAN DFRAG
	MACRO ACCBITMAP
	MACRO L2X2
	LINK DFRAG,L2X2,ACCBITMAP	[ L2X2 first due to frequency of usage ]

Execution

	DFR*AG :== $dev:[directory]DFRAG
	DFRAG Dxyz options		[ produces output file Dxyz.DAT ]
					[ with FORTRAN carriage-control ]
		options:
			/TRace	produce Dxyz.TRC showing runs and
				adjacent-bit adjustment
			/SHort	suppress output line for empty bins (default)
			/LOng	print bin-number, with blank value,
				for empty bins

   or

	RUN DFRAG		[ prompts for Dxyz ]
			/SHort	option is the default, no /TRace

Caveats, problems, etc

The current version does not evaluate all run lengths in "mixed" bytes;
that is planned for version 2.1, after I figure out whether to do it by
table lookup or LIB$FF{S/C} calls.

The output (and trace) file(s) will be incorrect by the amount of
extent-caching current for the drive.  I have not been able to determine
how to (temporarily) force extent-caching to be disabled for a drive.
The following sequence will usually yield two different successive
versions of Dxyz.DAT:
	DFRAG Dxyz
	ANA/DISK Dxyz:		[ let run for a few seconds, then ^Y or ^C ]
	DFRAG Dxyz
If anyone can tell me how the VERIFY utility does it (I tried to figure it out
from the microfiche), the information would be greatly appreciated,
and quickly incorporated into the next release.

Device name MUST be 4 characters, and options MUST follow the device name
(spaces are not required, however).

User must have read-access to [000000]BITMAP.SYS in the disk in question,
either directly, or by SYSPRV or READALL privileges.

Charles Garman		609-866-8821
RCA MSR			(soon to be GE?)
Borton Landing Rd, POW-2
Moorestown NJ 08057