I'm currently doing work related to isolating compiler optimization bugs. However in the course of testing with Intel C++ 13, I've found a number testcases where the compiler crashes before even emitting code, or in one case generates unlinkable code for a valid testcase (gcc and clang accept).
Where would be the appropriate place to post these bugs so somebody at Intel can have a look at them? All the testcases are fairly small - 10-30 lines of C code. Thanks.
Sergey Kostrov wrote:
>>...Where would be the appropriate place to post these bugs so somebody at Intel can have a look at them?..
It is absolutely the right forum to report any problems or issues with Intel C++ compiler. Please try to group these test cases in different posts or threads. Thanks in advance.
Excellent. I see that Intel C++ 13.1 was recently released, so I'll re-test with the newer version before posting. Would it be better to group them all under a single thread? Thanks.
okay we have no problem with the C++ 13.0 under openSUSE 12.2 x86_64 linux, we use many programssource for test the compiletime, codequality , optimize and so far, we scimark2c, mpich2-3.0.2, linpack , whetstone, dhrystone and much more no compiler crash at any time
Georg Zitzlsberger (Intel) wrote:
the first place to report problems and get help quickly is here, not the forum:
There, you also have a guaranteed response time.
That makes sense. However, I don't seem to be able to log in there. We're using the non-commercial licensed version (we're researchers at UC Davis) - perhaps this is why we don't have access to the Intel Premier site? Thanks,
We've continued to give advice indicating that holders of non-commercial licenses could register them at https://registrationcenter.intel.com and thus create a premier account. Has this changed? The priority on issues filed by the non-commercial license would not be under "guaranteed response."
Hello Sergey, Hello Georg
now we have update from C++ 13.0 to C++ 13.1 on the same operating System openSUSE 12.2 64 Bit Linux with KDE 4.10.0
we use the same programs for test such as scimark2c, flops.c, dhrystone, whetstone, we compile the complete CUDA 5.0 SDK Sample Suite, we recompile the LLNL Kernel ( Loop-test ) and many other test sourcecodes , no problems no compiler crashes, we use different options to compile, we link with the standard linux libs, and the Intel Libs, and much more, the same result, no compiler crash, the cuda samples nbody, matrixmul are 8 to 10 % faster as gcc compilation, we reach 1.2 TFlops/sec with CUDA nbody on a midrange SLI configuration with GeForce GTX560Ti graphic cards, we like the ICC 13.1
we have install ParallelStudio XE 2013 update 1 on openSUSE 12.3 RC1 and test with the same programs such as scimark2, linpack, whetsone, dhrystone and many other, no problems no compiler crash ( C++ 13.0 ),
works fine no problem on openSUSE 12.3 RC 1