This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: x86 patch: SSE-based FP<=>int conversions, round 3
On Dec 22, 2006, at 12:45 PM, Eric Christopher wrote:
I tried this; Darwin's assembler doesn't like that. I guess this
means that Darwin's assembler isn't really an AT&T assembler (?).
It looks to me like the Right Thing To Do is to create a third, non-
AT&T, non-Intel/Microsoft assembler variant "ATT_Darwin."
Do you want a new assembler type added to my patch?
Ick, no. I'd rather have our as fixed. We've fixed a number of
things in our assembler to match gas so far.
-eric
Likewise, on Dec 21, 2006, at 7:02 PM, Mike Stump wrote:
Ick. :-( I'd rather get our gas and gas on the same page and go
from there... You can put in a config check for support of the form
you want, and enable the pattern when the assembler supports it.
See gcc_GAS_CHECK_FEATURE in configure for an example of how to do
this.
Relax guys. I looked at it again, and I respectfully disagree with
Jan about the {w}; the instruction returns a 32-bit result, so if it
has any suffix, I think it should be "l" (or "{l}" in the pattern).
But I doubt that anyone expects a suffix. The instruction has no
variants, so there's no confusion about the width of the operation or
the result:
cvttsd2si - Convert with Truncation Scalar Double-Precision Floating-
Point Value to Signed Doubleword Integer
stuart