icc generates endless loop when using avx instructions in nbench

[Jan_K_4]

Hello,

I'm experimenting with icc flags and nbench and it seems that icc is generating wrong code when using avx. All but the assignment test are working fine. If I use -msse4.2 the assignment test is also working. The -mavx flag seems to trigger the bug.

ICC Version : 13.1.0

CFLAGS -mavx -O2 

nbench url: http://www.tux.org/~mayer/linux/bmark.html

[Shenghong_G_Intel]

Hi Jan K,

Thank you for submitting your issue here. I will take a look into it to verify the issue. I will update you my test result soon.

Thanks,

Shenghong

[Shenghong_G_Intel]

Hi Jan K,

I can reproduce your issue and the issue is wrong optimization caused by vectorization. I will submit the issue to our developer to fix.

If you need workaround for the issue, you can have multiple options:

1. as you mentnined, it can work with SSE4.2

2. you can also remove '-vec-' option by setting CFLAGS to "-mavx -O2 -vec-"

3. the issue is caused by the function "first_assignments" in "nbench1.c", so if you disable vectorization for the loop in this function, it can also work. You have many possible solutions:

(1) disable vectorization for "nbench1.c" using "-vec-" by updating the Makefile

(2) add "#pragma novector" for the loops in "first_assignments" function in nbench1.c, you need to update code

I will not list all the possible ways....such as disable optimization for "first_assignments" and so on...but the root cause of this issue is the vectorization for the loop in first_assignments is wrong.

Let me know if you have any concerns. I will notify you here when the issue is fixed in future version.

Thanks,

Shenghong

[Shenghong_G_Intel]

Hi Jan K,

This issue is fixed in latest 2013 update 5 compiler. Please upgrade your compiler.

Thanks,

Shenghong

[Armando_Lazaro_Alami]

I can confirm that update 4 created a wrong executable when using AVX as target. Update 5 is correct.