diff options
author | Gunes Bayir <gunes.bayir@arm.com> | 2022-09-11 15:59:19 +0100 |
---|---|---|
committer | Gunes Bayir <gunes.bayir@arm.com> | 2022-09-16 12:38:59 +0000 |
commit | c4f2743951473f8d97f5a43767fdbb31a4df967c (patch) | |
tree | 26c49b1af8113bb169931f3af5e502904d455a25 /src/cpu/operators | |
parent | 0d05b6690fe69c57f63ca43d59b551f074613062 (diff) | |
download | ComputeLibrary-c4f2743951473f8d97f5a43767fdbb31a4df967c.tar.gz |
Optimize Quantized/Integer Bilinear Scale for Neon™
This patch introduces several performance optimizations regarding the Bilinear Scale operator with REPLICATE Border mode. Changes apply only to NHWC.
This patch
- Reduces the memory footprint by disabling precomputation of indices and weights when they're not used
- Rewrites the kernels for QASYMM8/QASYMM8_SIGNED/U8(Uint8)
- Adds S8(Int8) Bilinear Scale for Border mode REPLICATE
- Removes Bilinear Scale SVE kernels for Quantized and Integer types and adjust the heuristics to choose the Neon™ implementation
- Adds new test cases where the input and output of the Bilinear Scale operator have different quantization scale and offset
Resolves: COMPMID-5453, COMPMID-5454
Change-Id: I3d251e76e0c6978fd5a0a1795ec62ab536bec93c
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8250
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/operators')
-rw-r--r-- | src/cpu/operators/CpuScale.cpp | 4 | ||||
-rw-r--r-- | src/cpu/operators/CpuScale.h | 4 |
2 files changed, 5 insertions, 3 deletions
diff --git a/src/cpu/operators/CpuScale.cpp b/src/cpu/operators/CpuScale.cpp index fdb52e5ede..a13a0f56a2 100644 --- a/src/cpu/operators/CpuScale.cpp +++ b/src/cpu/operators/CpuScale.cpp @@ -215,9 +215,9 @@ void CpuScale::prepare(ITensorPack &tensors) _scale_info.interpolation_policy; const SamplingPolicy sampling_policy = _scale_info.sampling_policy; - bool precompute_indices_weights = arm_compute::scale_utils::is_precomputation_required(_data_layout, src->info()->data_type(), policy_to_use); + bool precompute_indices_weights = arm_compute::scale_utils::is_precomputation_required(_data_layout, src->info()->data_type(), policy_to_use, _scale_info.border_mode); - if(precompute_indices_weights == true) + if(precompute_indices_weights) { switch(policy_to_use) { diff --git a/src/cpu/operators/CpuScale.h b/src/cpu/operators/CpuScale.h index f605af6712..ee7c523bad 100644 --- a/src/cpu/operators/CpuScale.h +++ b/src/cpu/operators/CpuScale.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021 Arm Limited. + * Copyright (c) 2021-2022 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -45,6 +45,8 @@ public: * @param[in, out] src Source tensor info. Data type supported: QASYMM8/QASYMM8_SIGNED/U8/S16/F16/F32. (Written to only for @p border_mode != UNDEFINED) * @param[out] dst Destination tensor info. Data type supported: Same as @p src. All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane. * @param[in] info @ref ScaleKernelInfo to be used for configuration + * + * @note Using S8 data type only supports NHWC, @p border_mode Replicate, and @p policy Bilinear */ void configure(ITensorInfo *src, ITensorInfo *dst, const ScaleKernelInfo &info); /** Static function to check if given info will lead to a valid configuration |