On 05 Jun 08, at 22:58, Graham Cox wrote:
call L_objc_msgSend$stub
leal LC17-"L00000000012$pb"(%ebx), %ecx
movl %eax, %edx
andl $65535, %edx
shrl $16, %eax
cvtsi2ss %eax, %xmm0
cvtsi2ss %edx, %xmm1
mulss (%ecx), %xmm0
addss %xmm1, %xmm0
movss %xmm0, -44(%ebp)
This'll take the return value of objc_msgSend (in eax) and convert it
to floating-point. The shift/multiply/add bit is a bit strange, but
the cvtsi2ss is a clear sign of an integer->float conversion. GCC is
assuming that the return value should be interpreted as an integer,
then cast to a float.
call L_objc_msgSend_fpret$stub
fstps -16(%ebp)
As the _fpret suggests, this is behaving correctly.
Keep in mind that, in the x86 architecture (along with most other
modern architectures), floating-point values are kept in separate
registers from other values. As the Mach-O x86 ABI mandates that (non-
struct) return values are returned in registers, rather than on the
stack, it's therefore crucial that functions which return floating-
point values have explicit prototypes.
_______________________________________________
Cocoa-dev mailing list (Cocoa-dev@lists.apple.com)
Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com
This email sent to [EMAIL PROTECTED]
