Fuse batch normalization changes to enable fp16 in armv8a multi_isa builds
* Code guarded with __ARM_FEATURE_FP16_VECTOR_ARITHMETIC needs
to be moved to an fp16.cpp file to allow compilation with
-march=armv8.2-a+fp16
* fp16.cpp needs to use the template fused_batch_normalization_dwc_nhwc() that
had to be moved from impl.cpp to impl.h
* Removed impl.cpp
* Partially resolves MLCE-1102
Change-Id: Idaaa113c71729e32e565acf5fb5694c76c36d76d
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10308
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/filelist.json b/filelist.json
index d3cf984..2a88aec 100644
--- a/filelist.json
+++ b/filelist.json
@@ -973,8 +973,7 @@
],
"neon": {
"common": [
- "src/cpu/kernels/fuse_batch_normalization/nchw/all.cpp",
- "src/cpu/kernels/fuse_batch_normalization/nhwc/neon/impl.cpp"
+ "src/cpu/kernels/fuse_batch_normalization/nchw/all.cpp"
],
"fp16": [
"src/cpu/kernels/fuse_batch_normalization/generic/fp16.cpp",