Unofficial VirtualDub Support Forums [Powered by IB Forums]

Printable Version of Topic
Click here to view this topic in its original format

Unofficial VirtualDub Support Forums > VirtualDub Filters and Filter Development > Intel Compiler Error

Posted by: jpsdr Oct 15 2011, 10:49 AM

I've the following problem with last Intel compiler :

>int128.cpp(40): error : no suitable user-defined conversion from "vduint128" to "const vdint128" exists
1> return (q[1]^x.q[1])<0 ? -(const vdint128&)bd : (const vdint128&)bd;

CODE

const vdint128 vdint128::operator*(const vdint128& x) const {
vdint128 X = x.abs();
vdint128 Y = abs();

vduint128 bd(VDUMul64x64To128(X.q[0], Y.q[0]));

bd.q[1] += X.q[0]*Y.q[1] + X.q[1]*Y.q[0];

return (q[1]^x.q[1])<0 ? -(const vdint128&)bd : (const vdint128&)bd;
}

Posted by: jpsdr Oct 17 2011, 07:32 AM

So, of course, question is : How correct it ?
Strange thing is that until the last version, it compiled without problem...

Posted by: IanB Oct 17 2011, 09:10 PM

The compiler is right but is being overly pedantic. You can probably deceive it with some & and * magic.

(const vdint128&)(*((vdint128*)&bd))

Posted by: jpsdr Oct 19 2011, 07:33 AM

Thanks, it worked.

Posted by: phaeron Oct 23 2011, 07:31 PM

I don't understand why the Intel compiler is trying to issue a value cast here. C-style casts are allowed to do a reinterpret_cast and a diagnostic can only be issued if the cast is impossible or ambiguous. There are no conversions between vdint128 and vduint128 that would allow an ambiguous cast situation.