Opcode unsupported

[Ioannis_P_]

Hello,

I have installed Intel Compiler 15.0.1.148 and MPSS 3.4.7 on Windows.

I try to compile my application natively for Xeon Phi (using /Qmic) but I get lots of errors like this:

C:/Program Files/Intel/MPSS//k1om-mpss-linux/usr/include/c++/iostream(75): (col. 25) error #13393: Opcode unsupported on target architecture: movdqa

Note: I get similar "Opcode unsupported" error when compiling on Linux for native Xeon Phi build (-mmic) using Intel Compiler 15 6th update and MPSS 3.7.

Any ideas how to solve this?

[Loc_N_Intel]

Hi,

It means the operation MOVDQA in the machine  language is not supported. Is it possible that you post your code and how do you compile it? Thanks.

[Ioannis_P_]

Hi and thanks,

the code is pretty big to show for this case but I can show something small that gives the same problem:

#include <float.h>

template <class T>
class limits
{
    public:
        static inline T Min() { return 0; }
        static inline T Max() { return 0; }
        static inline T Epsilon() { return 0; }
};

template <>
class limits<float>
{
    public:
        static inline float Min() { return FLT_MIN; }
        static inline float Max() { return FLT_MAX; }
        static inline float Epsilon() { return FLT_EPSILON; }
};

static const float PrecisionMax=limits<float>::Max();

I get the same Opcode unsupported error at the line that defines PrecisionMax.

The program is compiled with /Qmic & intel compiler.

Best wishes

[Kevin_D_Intel]

I cannot reproduce this. I am not exactly aligned with your MPSS but given you reproduced this across Windows/Linux with different MPSS and compilers makes me think I should be able to reproduce this with combinations I have. On Windows under MPSS 3.5.37182.0, with PSXE 2015 Update 1, I cannot reproduce the error.

C:\u629126> icl /Qmic -c  u629126.cpp
Intel(R) C++ Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 15.0.1.148 Build 20141023
Copyright (C) 1985-2014 Intel Corporation.  All rights reserved.

Can you show us the actual compile line or at least the compiler options used. Is anything more than /Qmic (-mmic) used?

[Ioannis_P_]

To be absolutely honest with you, I reduced the code to only a small part but didn't check whether there was still error emitted. I am actually glad that you couldn't reproduce it meaning that I must be doing something wrong on my side.

Let me test to see if I can find a small code sample that does not compile.