Hi!  I've got code that looks like this:
__asm
{
	L4:		FCHS;
	L5:		FADD     ST(3),ST(0);
	L6:		FXCH     ST(1);
	L15:		FSUB     ST(0),ST(3);
}
and the DMC (v849) compiler reports the errors:
ad_high.cpp(101) : Error: Illegal type/size of operands for the
fadd instruction
ad_high.cpp(103) : Error: Illegal type/size of operands for the
fsub instruction
This compiles fine with Microsoft C++ Professional 6.0.
Any ideas what's wrong?
Thanks!
-- Glenn Lewis

Glenn Lewis wrote:
> Hi!  I've got code that looks like this:
> __asm
> {
> 	L4:		FCHS;
> 	L5:		FADD     ST(3),ST(0);
> 	L6:		FXCH     ST(1);
> 	L15:		FSUB     ST(0),ST(3);
> }
> and the DMC (v849) compiler reports the errors:
> ad_high.cpp(101) : Error: Illegal type/size of operands for the
> fadd instruction
> ad_high.cpp(103) : Error: Illegal type/size of operands for the
> fsub instruction
> This compiles fine with Microsoft C++ Professional 6.0.
> Any ideas what's wrong?

The ST(0) register value is implied, so just use:
	FADD ST(3);
for example.

OK, so I'm assuming that the ST(0) is implied as well in FSUB, to
give me:
FSUB ST(3);
Thanks.
-- Glenn

Unfortunately, I'm still stuck... I'm told that this is compilable by any ANSI Standard C compiler... maybe the company that generated this code is
wrong.  Here's what I tried:

        ruby -p -e "if ($_ =~ /FSUB/); $_.gsub!(/,ST\(0\)/, ''); end; if ($_ =~ /FADD/); $_.gsub!(/ST\(0\),/, ''); end" ad_close.cpp > tmp.cpp
        dmc -c tmp.cpp -o ad_close.obj
tmp.cpp:
        L4:             FSUB     ST(1);
        ^
tmp.cpp(91) : Warning 13: Illegal type/size of operands for the fsub instruction
        L5:             FADD     dword ptr [ESI+20];
        ^
tmp.cpp(92) : Warning 13: Illegal type/size of operands for the fsub instruction
        L9:             FMUL     dword ptr [ESI+200];
        ^
tmp.cpp(96) : Error: Illegal type/size of operands for the fadd instruction
        L12:            FADD     ST(1),ST(0);
        ^
tmp.cpp(99) : Warning 13: Illegal type/size of operands for the fadd instruction
        L13:            FSUB     dword ptr [ESI+16];
        ^
tmp.cpp(100) : Error: Illegal type/size of operands for the fadd instruction
        L15:            FDIV     dword ptr [ESI+176];
        ^
tmp.cpp(102) : Error: Illegal type/size of operands for the fdiv instruction
        L19:            FCOMI    ST(0),ST(2);
        ^
tmp.cpp(106) : Warning 13: Illegal type/size of operands for the fadd instruction
        L20:            FCMOVNB  ST(0),ST(3);
        ^
tmp.cpp(107) : Error: Illegal type/size of operands for the fcomi instruction
        L23:            FADD     ST(1);
        ^
tmp.cpp(110) : Warning 13: Illegal type/size of operands for the fsub instruction
        L24:            FADD     ST(0);
        ^
tmp.cpp(111) : Warning 13: Illegal type/size of operands for the fadd instruction
        L25:            FCOMI    ST(0),ST(3);
        ^
tmp.cpp(112) : Warning 13: Illegal type/size of operands for the fadd instruction
        L26:            FSUB     ST(0);
        ^
tmp.cpp(113) : Error: Illegal type/size of operands for the fcomi instruction

Unfortunately, the compiler spits out the line *after* the error, but you get the general idea. -- Glenn Lewis

Glenn Lewis wrote:
> Unfortunately, I'm still stuck... I'm told that this is compilable by any ANSI Standard C compiler... maybe the company that generated this code is
> wrong.

Whoever told you that was in error, Standard C doesn't specify inline assembler. Try this:

> __asm
> {
>         L4:             FCHS
>         L5:             FADD     ST,ST(3)
>         L6:             FXCH     ST(1)
>         L15:            FSUBR    ST(3),ST
> }

Since I continue to get more and more errors with the __asm section
from DMC, and since it compiles fine with MSVC++6.0, is it possible
to go and ahead compile this code with MS Visual C++ 6.0 and then
link it into my D program?
I can't seem to get that to work either.
Thanks!
-- Glenn Lewis

Glenn Lewis wrote:
> Since I continue to get more and more errors with the __asm section
> from DMC, and since it compiles fine with MSVC++6.0, is it possible
> to go and ahead compile this code with MS Visual C++ 6.0 and then
> link it into my D program?
> I can't seem to get that to work either.

The inline asm errors are all fixable, they are probably only one or two things going awry. Easier to address them one by one.