- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Following compiles with gcc but fails with Intel compiler, any idea?
typedef float __attribute__ ( ( vector_size( 4*sizeof(float) ) ) ) Vec4F; float x(Vec4F const & v){return v[0];}
>icpc -v icpc version 15.0.0 (gcc version 4.8.1 compatibility) >icpc -c foo.cc Y/src/foo.cc(2): error: expression must have pointer-to-object type float x(Vec4F const & v){return v[0];} ^ compilation aborted for foo.cc (code 2)
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello Shahzad,
we've a feature request currently pending for __attribute__ ((vector_size(...
))).
Right now it's not supported by the Intel(R) C++ Compiler. The ticket is DPD200173341 DPD200526864 and I'll tell our engineers once more about the importance of this.
For the CMS code, I've already created a workaround, which I'm not very keen to propose (but could share with you). It simply wraps the vector to a class and overloads various operators. It's very invasive and not my favorite choice.
Maybe you could use the SIMD intrinsics class as a more appropriate workaround?
https://software.intel.com/en-us/node/524399
As soon as the feature request is implemented I'll update the thread.
Best regards,
Georg Zitzlsberger

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page