Question About $CPU Metastatement

[Clay Clear]

If I use $CPU 80386 in a program, will the program automatically
abort with a runtime error if the program is run on a machine less
than a '386? I am trying to figure out whether I really need the
code for testing for the 8086/8088/80286/80386 before finally
testing for the 80486. *IF* the PB EXE's will automatically abort
on too low of a processor, then my OBJ file should be able to go
directly to testing for a 80486. Doing so requires using the registers
in 32-bit mode. I have no end machines to determine the answer
myself.

Thank you for any replies.


------------------

[Clay Clear]

Disregard. I will stick to my original plan of using pbvCpu & pbvNpx.
I did not want to rely on them in case they changed in future compiler
versions. I do not have to rely on them NOW, but the code is much shorter
with them.


------------------

[Michael Mattias]

From PB/DOS 3.5 help file: (Help, Language Reference, $CPU, link to "Generated Code")

You can mix CPU code. For example, you can generate 8086/8088 code in
your main program and 80286 code in a unit; when you do this, though,
the final program won't run on a CPU that's at a lower level than the
highest-level CPU specified in the entire program. Thus the example
here would require at least an 80286 machine to run a program made from
the main program and the unit.

[Tom Hanlin]

I am not at work right now, but I would doubt that the program goes
so far as to check whether a specific processor is available. If you
specify a '386, the program probably assumes that you know what you're
doing-- i.e., that a 386 is available. Plus, there are few enough
optimizations that would benefit from 386-level coding anyway, so
running that code on a lesser processor would be "iffy"-- it would
probably usually work, but it might drop dead at any moment.

Short version? If you expect customers will use less than a 386, don't
use 386 optimizations, or put in your own prevention code.

------------------
Tom Hanlin
PowerBASIC Staff

[Clay Clear]

Thanks for the info, Tom. I will go ahead and keep my CPU-testing
code as-is, then (except that I replaced the testing for the values in
pbvCpu & pbvNpx with ASM code that manually checks for 8086/8088/80286
CPU's - do not want to have to recode in future compiler versions if those
internal vars change ).


------------------

[Stavros A Petridis]

I have tested it.
If for example you compile with $CPU 386 meta and you run the program
in a lower CPU the program prints in the screen NO 386 and ends.


------------------
Byte Hunters of the World Unite.

[Clay Clear]

Thanks, Stavros!


------------------