diff options
| author | Narumol Prangnawarat <narumol.prangnawarat@arm.com> | 2020-03-30 16:11:04 +0100 |
|---|---|---|
| committer | Narumol Prangnawarat <narumol.prangnawarat@arm.com> | 2020-03-31 09:29:40 +0100 |
| commit | 250d3927b16abe4d6932cd5dce1184bd7026a2b7 (patch) | |
| tree | f73603873c0fbd692fbcbbd242d2a45cef6dc890 /src/backends/neon/NeonLayerSupport.cpp | |
| parent | e2062cdf1eb31b87860f9889f0e799e89f0dfa30 (diff) | |
| download | armnn-250d3927b16abe4d6932cd5dce1184bd7026a2b7.tar.gz | |
IVGCVSW-4633 Add conversion of BF16 support to Neon
* Add NeonConvertBf16ToFp32Workload
* Add NeonConvertFp32ToBf16Workload
* Add BFloat16 type support to NeonConstantWorkload and NeonTensorHandle
* Add ConvertBf16ToFp32Weight when ConvertBf16ToFp32Layer is added
* Unit tests
Signed-off-by: Narumol Prangnawarat <narumol.prangnawarat@arm.com>
Change-Id: Id5b44a203add5e0c98c1ca4e2162115741b56644
Diffstat (limited to 'src/backends/neon/NeonLayerSupport.cpp')
| -rw-r--r-- | src/backends/neon/NeonLayerSupport.cpp | 20 |
1 files changed, 20 insertions, 0 deletions
diff --git a/src/backends/neon/NeonLayerSupport.cpp b/src/backends/neon/NeonLayerSupport.cpp index c01a178f18..44e84fb974 100644 --- a/src/backends/neon/NeonLayerSupport.cpp +++ b/src/backends/neon/NeonLayerSupport.cpp @@ -259,6 +259,16 @@ bool NeonLayerSupport::IsConstantSupported(const TensorInfo& output, &TrueFunc<>); } +bool NeonLayerSupport::IsConvertBf16ToFp32Supported(const TensorInfo& input, + const TensorInfo& output, + Optional<std::string&> reasonIfUnsupported) const +{ + armnn::IgnoreUnused(input); + armnn::IgnoreUnused(output); + armnn::IgnoreUnused(reasonIfUnsupported); + return true; +} + bool NeonLayerSupport::IsConvertFp16ToFp32Supported(const TensorInfo& input, const TensorInfo& output, Optional<std::string&> reasonIfUnsupported) const @@ -269,6 +279,16 @@ bool NeonLayerSupport::IsConvertFp16ToFp32Supported(const TensorInfo& input, return true; } +bool NeonLayerSupport::IsConvertFp32ToBf16Supported(const TensorInfo& input, + const TensorInfo& output, + Optional<std::string&> reasonIfUnsupported) const +{ + armnn::IgnoreUnused(input); + armnn::IgnoreUnused(output); + armnn::IgnoreUnused(reasonIfUnsupported); + return true; +} + bool NeonLayerSupport::IsConvertFp32ToFp16Supported(const TensorInfo& input, const TensorInfo& output, Optional<std::string&> reasonIfUnsupported) const |