commit 3883b2776cec33f16f0ea9a2d795de2b7c766df7
author Cathal Corbett <cathal.corbett@arm.com> Fri Jul 22 16:03:36 2022 +0100
committer Nikhil Raj <nikhil.raj@arm.com> Fri Aug 05 15:50:57 2022 +0100
tree 6842e15904037d73426d814d5751945b3d9c2376
parent 9d63fee68081b65bd72de3a70da76c2696c6c6ed

GitHub #667: Neon fold padding into average pool 2D quantization bug fix.

  * Originated from a GitHub issue: https://github.com/ARM-software/armnn/issues/667
  * Initially, Arm NN supports the pool 2D operation because there is no padding
    on the pool2d. Neon failure occurs when padding is followed by average pool 2D
    due to folding optimization.
  * Here we prevent the folding optimization from happening for the above special case
    and add it in as a backend specific optimization.

Signed-off-by: Cathal Corbett <cathal.corbett@arm.com>
Change-Id: Ia0fd90c3a6b4b9d29c81106f154617d2e893e26b

CMakeLists.txt

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 1d8ebe2..7061c60 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -544,6 +544,7 @@
         src/armnn/test/optimizations/ConvertConstantsBFloatTests.cpp
         src/armnn/test/optimizations/ConvertConstantsFloatToHalfTests.cpp
         src/armnn/test/optimizations/ConvertConstantsHalfToFloatTests.cpp
+        src/armnn/test/optimizations/FoldPadIntoQuantizedAveragePooling2DTests.cpp
         src/armnn/test/optimizations/FoldPadTests.cpp
         src/armnn/test/optimizations/Fp32NetworkToBf16ConverterTests.cpp
         src/armnn/test/optimizations/Fp32NetworkToFp16ConverterTests.cpp