diff options
author | Matthew Bentham <Matthew.Bentham@arm.com> | 2023-05-31 13:18:33 +0000 |
---|---|---|
committer | Matthew Bentham <matthew.bentham@arm.com> | 2023-06-26 12:44:15 +0000 |
commit | 7d9a78ebfb3553b95421a0da5e2686a3923748db (patch) | |
tree | 666915383e31a9da81f2605f7023c0e85e8226b3 /arm_compute/core/Validate.h | |
parent | a2bb80ea7111509c24caad8629533089decef430 (diff) | |
download | ComputeLibrary-7d9a78ebfb3553b95421a0da5e2686a3923748db.tar.gz |
Remove dependency on fp16 definitions from some core include files
This significantly improves the compilation times for parts of the core library that just need
a definition of float16_t rather than access to all of the fp16 intrinsics.
Signed-off-by: Matthew Bentham <Matthew.Bentham@arm.com>
Change-Id: I5da1c6b0df0dd87d1d17948cd2e9b7375874f455
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/529385
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9781
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/core/Validate.h')
-rw-r--r-- | arm_compute/core/Validate.h | 54 |
1 files changed, 54 insertions, 0 deletions
diff --git a/arm_compute/core/Validate.h b/arm_compute/core/Validate.h index 7e387c9cc1..ff4540cd8a 100644 --- a/arm_compute/core/Validate.h +++ b/arm_compute/core/Validate.h @@ -27,6 +27,7 @@ #include "arm_compute/core/Error.h" #include "arm_compute/core/IKernel.h" #include "arm_compute/core/ITensor.h" +#include "arm_compute/core/QuantizationInfo.h" #include "arm_compute/core/Utils.h" #include "arm_compute/core/Window.h" @@ -285,6 +286,28 @@ arm_compute::Status error_on_mismatching_dimensions(const char *function, const #define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(...) \ ARM_COMPUTE_RETURN_ON_ERROR(::arm_compute::error_on_mismatching_dimensions(__func__, __FILE__, __LINE__, __VA_ARGS__)) +/** Return true if the given format has horizontal subsampling. + * + * @param[in] format Format to determine subsampling. + * + * @return True if the format can be subsampled horizontaly. + */ +inline bool has_format_horizontal_subsampling(Format format) +{ + return (format == Format::YUYV422 || format == Format::UYVY422 || format == Format::NV12 || format == Format::NV21 || format == Format::IYUV || format == Format::UV88) ? true : false; +} + +/** Return true if the given format has vertical subsampling. + * + * @param[in] format Format to determine subsampling. + * + * @return True if the format can be subsampled verticaly. + */ +inline bool has_format_vertical_subsampling(Format format) +{ + return (format == Format::NV12 || format == Format::NV21 || format == Format::IYUV || format == Format::UV88) ? true : false; +} + /** Adjust tensor shape size if width or height are odd for a given multi-planar format. No modification is done for other formats. * * @note Adding here a few links discussing the issue of odd size and sharing the same solution: @@ -349,6 +372,37 @@ arm_compute::Status error_on_tensors_not_even(const char *function, const char * #define ARM_COMPUTE_RETURN_ERROR_ON_TENSORS_NOT_EVEN(...) \ ARM_COMPUTE_RETURN_ON_ERROR(::arm_compute::error_on_tensors_not_even(__func__, __FILE__, __LINE__, __VA_ARGS__)) +/** Calculate subsampled shape for a given format and channel + * + * @param[in] shape Shape of the tensor to calculate the extracted channel. + * @param[in] format Format of the tensor. + * @param[in] channel Channel to create tensor shape to be extracted. + * + * @return The subsampled tensor shape. + */ +inline TensorShape calculate_subsampled_shape(const TensorShape &shape, Format format, Channel channel = Channel::UNKNOWN) +{ + TensorShape output{ shape }; + + // Subsample shape only for U or V channel + if(Channel::U == channel || Channel::V == channel || Channel::UNKNOWN == channel) + { + // Subsample width for the tensor shape when channel is U or V + if(has_format_horizontal_subsampling(format)) + { + output.set(0, output.x() / 2U); + } + + // Subsample height for the tensor shape when channel is U or V + if(has_format_vertical_subsampling(format)) + { + output.set(1, output.y() / 2U); + } + } + + return output; +} + /** Return an error if the passed tensor objects are not sub-sampled. * * @param[in] function Function in which the error occurred. |