"Rod Pemberton" <spamtrap@[EMAIL PROTECTED]
> wrote:
>
>"Tim Roberts" <spamtrap@[EMAIL PROTECTED]
> wrote:
>> step <spamtrap@[EMAIL PROTECTED]
> wrote:
>> >
>>
>> gcc does not use the size of the operand to figure out which
instruction to
>> use. It can't tell the difference between the single and
double-precision
>> instructions, so you have to tell it, by adding a suffix letter.
>>
>> If you really want singles (which is what you have here with "float"),
use
>> "flds" and "fstps".
>
>Are you sure that's his problem?
>
>gcc 3.4.1 (binutils 2.1.5) compiles the code without error (DJGPP). It
>warns about 'fadd' needing to be 'faddp', 'fsub'...'fsubp' etc.
No, I'm not. I have tried several times to gain a comfortable
understanding of gas syntax and gcc's inline asm, and every time I think I
truly get it, I post an offering and someone offers a counterexample.
Many
web pages on the topic contradict each other, in what I suspect are
version
differences.
Perhaps I should have said "there exists at least one older version of gcc
that required flds and fstps instead of fld and fstp".
--
Tim Roberts, timr@[EMAIL PROTECTED]
& Boekelheide, Inc.
|
