diff options
| author | SiCong Li <sicong.li@arm.com> | 2020-06-08 17:30:51 +0100 |
|---|---|---|
| committer | SiCong Li <sicong.li@arm.com> | 2020-06-10 17:33:25 +0000 |
| commit | a32e2aef81cfcba9f5ae1770ceeb4a8d26fdc1f4 (patch) | |
| tree | b791e466df9dff4032be409d6765cb8a9af9319d /tests/validation/NEON/ReductionOperation.cpp | |
| parent | 8aa8764982d23ed8b8c8810bbfda30542f21e034 (diff) | |
| download | ComputeLibrary-a32e2aef81cfcba9f5ae1770ceeb4a8d26fdc1f4.tar.gz | |
COMPMID-3523: Fix validation fails on armv8.2-a
* Fix neon sqrt activation delta(epsilon)
* Fix NEON Hard Swish validation tolerance
* Fix NEON FP16 LogSoftmaxLayer validation test typo
* Raise NEON reduction (sum) f16 tolerance
Change-Id: Ia33d69ce5f0b78be1893fb8e13d2761a8e7fceff
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3318
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/validation/NEON/ReductionOperation.cpp')
| -rw-r--r-- | tests/validation/NEON/ReductionOperation.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tests/validation/NEON/ReductionOperation.cpp b/tests/validation/NEON/ReductionOperation.cpp index f155e97d0d..4c99daab77 100644 --- a/tests/validation/NEON/ReductionOperation.cpp +++ b/tests/validation/NEON/ReductionOperation.cpp @@ -46,7 +46,7 @@ namespace AbsoluteTolerance<float> tolerance_f32(0.0001f); RelativeTolerance<float> rel_tolerance_f32(0.0001f); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC -AbsoluteTolerance<float> tolerance_f16(0.1f); +AbsoluteTolerance<float> tolerance_f16(0.2f); RelativeTolerance<float> rel_tolerance_f16(0.1f); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC /** Tolerance for quantized operations */ |