KnowledgeBase Archive

An Archive of Early Microsoft KnowledgeBase Articles

View on GitHub

Q42762: Patches of Floating-Point Instructions at Run Time Are Normal

Article: Q42762
Product(s): See article
Version(s): 5.10
Operating System(s): MS-DOS
Keyword(s): ENDUSER | SR# G890217-12193 | mspl13_c
Last Modified: 30-MAY-1989

Question:

I have noticed that memory overwrites occur in the library routine
i8_input. While debugging the program under CodeView with the
assembler listing, I noticed that after executing into the code, the
first 2 bytes of each instruction beginning with a hex CD are
overwritten. The C statement that generated the assembler code with
the problem in it was an "fscanf" function call.

Below is an listing example of what is occurring:

Before
------

7E1F:4266 CD35C0       INT 35 ; FLD     ST(0)
7E1F:4269 CD35E1       INT 35 ; FABS
7E1F:426C CD372E7425   INT 37 ; FLD     TByte Ptr [__chbuf+5C (2574)]

After
-----

7E1F:4266 90           NOP
7E1F:4267 D9C0         FLD      ST(0)
7E1F:4269 90           NOP
7E1F:426A D9E1         FABS
7E1F:426C 90           NOP
7E1F:426D DB2E7425     FLD      TByte Ptr [__chbuf+5C (2574)]

Response:

The behavior you describe is quite normal and nothing to worry about.
Our floating-point package works by generating the INT instructions
you noticed. When these instructions are executed, the routine they
call replaces the INT instructions with the either library calls to
the emulator library or the actual 80 x 87 floating-point
instructions, depending on whether or not a coprocessor is installed.

When the instructions are executed again, there is no overhead for
determining whether on not a coprocessor is installed: the proper
instructions have been patched into place already. This patching
occurs even if the -FPi87 option has been selected.

You can force in-line 8087 instructions to be put into your code.
For information on this technique, query on the following keywords:

   in-line 8087 instructions

There is no need to do this if your code is going to run under DOS or
OS/2. Although this article mainly discusses FORTRAN, it applies to C
as well because the two languages use the same floating-point library.

This code modification occurs only under DOS. Under OS/2, coprocessor
instructions (such as in the second listing) are always generated. If a
coprocessor is not present at execution, the instructions cause
exceptions that are handled by the floating-point emulator software.

Note: Such code modification is impossible under OS/2 because there is
no way to dynamically change a code segment under OS/2, although it is
possible to cause a data segment to be executed.

THE INFORMATION PROVIDED IN THE MICROSOFT KNOWLEDGE BASE IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND. MICROSOFT DISCLAIMS ALL WARRANTIES, EITHER EXPRESS OR IMPLIED, INCLUDING THE WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT SHALL MICROSOFT CORPORATION OR ITS SUPPLIERS BE LIABLE FOR ANY DAMAGES WHATSOEVER INCLUDING DIRECT, INDIRECT, INCIDENTAL, CONSEQUENTIAL, LOSS OF BUSINESS PROFITS OR SPECIAL DAMAGES, EVEN IF MICROSOFT CORPORATION OR ITS SUPPLIERS HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. SOME STATES DO NOT ALLOW THE EXCLUSION OR LIMITATION OF LIABILITY FOR CONSEQUENTIAL OR INCIDENTAL DAMAGES SO THE FOREGOING LIMITATION MAY NOT APPLY.