-mfpmath=sse -mno-80387 troubles

[Richard Henderson]

On Fri, Jan 28, 2005 at 05:23:37PM +0100, tbp wrote:
> With -mno-80387 there's some pushing and then a call to floatdidf
> which returns the result on the top of the fpu stack.

Without amd64 instructions, there's no way to directly convert a
64-bit value into floating point.  We *could* expand that sort of
thing inline, with SSE2 using cvtpi2pd, but no one's written the
pattern for it.  And it wouldn't help SSE1 anyway; using cvtpi2ps
would result in double rounding errors.


r~