[PATCH] D38417: [test-suite] Adding HACCKernels app

[Hal Finkel via llvm-commits]

On 11/08/2017 08:15 AM, Sanjay Patel via Phabricator wrote:
> spatel added a comment.
>
> In https://reviews.llvm.org/D38417#918942, @zvi wrote:
>
>> Since the final commit of this patch, https://reviews.llvm.org/rL317483, the AVX2 buildbot is broken: http://lab.llvm.org:8011/builders/clang-cmake-x86_64-avx2-linux/builds/1402
>>
>> Brian, will you be able to look into this failure right away? If not, please consider reverting the commit until we get this sorted out. If you believe the failure is due to a bug in LLVM, please create a bug report. Thanks.
>
> As noted in off-list email, on x86 I'm seeing this output:
> $ ./317576fastbroadwell 450
> Iterations: 450
> Gravity Short-Range-Force Kernel (4th Order): 34376.3 **689.585** -2378.97: 0.088137 s
> Gravity Short-Range-Force Kernel (5th Order): 34361.8 689.281 -2378.1: 0.089248 s
> Gravity Short-Range-Force Kernel (6th Order): 34360.9 689.252 -2378.1: 0.091363 s
>
> While the HACCKernels.reference_output is:
> Iterations: 450
> Gravity Short-Range-Force Kernel (4th Order): 34376.3 **689.584** -2378.97
> Gravity Short-Range-Force Kernel (5th Order): 34361.8 689.281 -2378.1
> Gravity Short-Range-Force Kernel (6th Order): 34360.9 689.252 -2378.1
>
> If this test is being built with -ffast-math, we should add an FP tolerance to the output verification?

Yes, that would be reasonable. When the running time of the test was 
reduced, it made it a bit more sensitive to these kinds of things.

  -Hal
>
>
> Repository:
>    rL LLVM
>
> https://reviews.llvm.org/D38417
>
>
>

-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory