COMPMID-3599: Disable direct gemm convolution testing for armv7 for 8bit

NEGEMMConv2d does not support armv7 kernels for uint8/int8 input.
Restrict int8/uint8 testing to aarch64 only.

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I3ccbed64b3c7c834f9f9377f48c427f1139d6a26
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4405
Reviewed-by: Sang-Hoon Park <sang-hoon.park@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

diff --git a/tests/validation/NEON/ConvolutionLayer.cpp b/tests/validation/NEON/ConvolutionLayer.cpp
index 112188fdfa3f7a813ea031ded91d9dc2081e5108..9dff36b13924d76c7423be4ea325f6710ee7c9df 100644 (file)
--- a/tests/validation/NEON/ConvolutionLayer.cpp
+++ b/tests/validation/NEON/ConvolutionLayer.cpp
@@ -514,6 +514,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, NEDirectGEMMConv2dLayerFixture<float>, framewor
 TEST_SUITE_END() // FP32
 TEST_SUITE_END() // Float
 
+#ifdef __aarch64__
 template <typename T>
 using NEDirectGEMMConv2dLayerQuantizedFixture = ConvolutionValidationQuantizedFixture<Tensor, Accessor, NEGEMMConv2d, T>;
 
@@ -568,6 +569,7 @@ FIXTURE_DATA_TEST_CASE(RunSmallSigned, NEDirectGEMMConv2dLayerQuantizedPerChanne
 }
 TEST_SUITE_END() // QSYMM8_PER_CHANNEL
 TEST_SUITE_END() // Quantized
+#endif           // __aarch64__
 
 TEST_SUITE_END() // DirectGEMMConv2d