m128_f32 support in SSE intrinsics

[pt]

I have an application written for MS Visual studio 2008 that I would like to compile with Intel C++. I use the type __m128 for intrinsics and the union member m128_f32 to access the individual fields such as here:

toItem->_t.m128_f32[toIndex]=fromItem->_t.m128_f32[fromIndex];

However the Intel C++ compiler does not seem to support the use of m128_f32. I have hundreds of statements using m128_f32 to initialise a float (sub-)item from either another __m128 or from a scalar float variable, how should I port the code to work with the Intel C++ compiler?

- Thanks, Peter

[barsmonster]

I had the similar problem. These inner values should not be used for optimized code(cause optimizer does not "see" dependence).

Use SSE instructions to convert data to 32 bits and vice versa.

[Quoc-An_L_Intel]

Accessing members of an __m128 datatype is now supported with the Intel compiler that comes with Intel Parallel Studio 2011.