home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Usenet 1994 October
/
usenetsourcesnewsgroupsinfomagicoctober1994disk2.iso
/
misc
/
volume9
/
proft.ms
/
README
< prev
Wrap
Text File
|
1989-11-26
|
2KB
|
40 lines
PROFILER READ-ME
================
This software is an execution time profiler for programs developed
with Turbo C and Microsoft C, or any other program that is linked
with Microsoft or Turbo LINK. PROFILE will read the linker's output
map file, and then run the executable program N times (default 1),
using the timer interrupt to gather statistical data about where
execution is mostly spent. After the executable has been run, the
data is formatted in different ways and sent to a file.
As is mentioned in the MAN file, there is a constant (ADJ_CONST) in
the source that is dependent on the compiler used. This is a correc-
tion constant to compensate the actual load address for the executing
program when it resides in primary memory. The supplied values work
for MicroSoft C v.5.1, and for TurboC v.2.0. Probably other values
need to be used for other compilers, and maybe also for other ver-
sions of DOS (the supplied values work for PCDOS 3.30). Unfortunately,
I know of no way to know for certain where a program executed by
spawn..() will come down into memory. On the other hand PROFILE con-
tains two command line switches ('-?' and '-#') to check that the
constant is correct. In other words, to install this you need a
little more than just compile it and go... it works well in my
installation. Though I'd pass it along - hopefully it is of use
to at least someone.
A word of the program's heritage - I started to write it after
collecting Diomidis Spinellis' profiler for UseNet. Although his
profiler works as advertised (and is well written) I think my
version is somewhat improved, as discussed in the MAN file. At
least, it will handle CTRL-C in an orderly manner...
To compile, you need a UNIX compatible make (in other words,
MicroSoft MAKE will NOT work). Anyway the makefiles are so simple
you could just as well compile 'by hand'. Who knows, maybe some
of you can remove the need of AJD_CONST and post the result (or
at least mail it to me)?
Bjorn