diff options
author | Keith Davis <keith.davis@arm.com> | 2019-11-04 08:58:33 +0000 |
---|---|---|
committer | Keith Davis <keith.davis@arm.com> | 2019-11-04 16:46:35 +0000 |
commit | 5236e1d6bcff6ebec7ec10d7d416cc6ead5482dd (patch) | |
tree | 4152c5fcd6b9c11848a02dfa4ff8705a2cfae0a5 /src/armnn | |
parent | f71079328ae72a65c91e410b2bd35eabb67cb6d1 (diff) | |
download | armnn-5236e1d6bcff6ebec7ec10d7d416cc6ead5482dd.tar.gz |
IVGCVSW-3835 Create Encoder and Decoder for QSymm8PerAxis
* Add QuantizedSymm8PerAxis to armnn DataType (types.hpp) and
* Add Quantize and Dequantize template for int8 in TypeUtils to be able to compute QSymm8 of the weight
* Create PerAxisIterator for per-axis quantization
* Create QSymm8PerAxisDecoder
* Create QSymm8PerAxisEncoder
Signed-off-by: Keith Davis <keith.davis@arm.com>
Change-Id: Ibcfe0288a197b7ee50b543bdbd77b7edb8a547c2
Diffstat (limited to 'src/armnn')
-rw-r--r-- | src/armnn/TypesUtils.cpp | 8 |
1 files changed, 8 insertions, 0 deletions
diff --git a/src/armnn/TypesUtils.cpp b/src/armnn/TypesUtils.cpp index cdc30da8ca..83c56c491c 100644 --- a/src/armnn/TypesUtils.cpp +++ b/src/armnn/TypesUtils.cpp @@ -33,6 +33,10 @@ float armnn::Dequantize(QuantizedType value, float scale, int32_t offset) return dequantized; } +/// Explicit specialization of Quantize for int8_t +template +int8_t armnn::Quantize<int8_t>(float value, float scale, int32_t offset); + /// Explicit specialization of Quantize for uint8_t template uint8_t armnn::Quantize<uint8_t>(float value, float scale, int32_t offset); @@ -45,6 +49,10 @@ int16_t armnn::Quantize<int16_t>(float value, float scale, int32_t offset); template int32_t armnn::Quantize<int32_t>(float value, float scale, int32_t offset); +/// Explicit specialization of Dequantize for int8_t +template +float armnn::Dequantize<int8_t>(int8_t value, float scale, int32_t offset); + /// Explicit specialization of Dequantize for uint8_t template float armnn::Dequantize<uint8_t>(uint8_t value, float scale, int32_t offset); |