diff options
| author | Aleksandr Nikolaev <aleksandr.nikolaev@arm.com> | 2020-06-25 12:25:52 +0100 |
|---|---|---|
| committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2020-07-16 16:02:50 +0000 |
| commit | a084b46835d20fdfe6e590b91b7ca64fba3542df (patch) | |
| tree | e4abe49cc10f3406dcaad9d3dfbbc22cc54476a6 /src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp | |
| parent | f738fe6b6e059916294c48b942952c261569df18 (diff) | |
| download | ComputeLibrary-a084b46835d20fdfe6e590b91b7ca64fba3542df.tar.gz | |
[ONCPUML-97]: Implement "int8" support for 2D decomposition at high core counts
Interleaved2d functionality was extended to uint8 and int8 kernels.
Change-Id: If78facbce56e9ec7b2f4c23436af0bd5db7f7b69
Signed-off-by: Aleksandr Nikolaev <aleksandr.nikolaev@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3467
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp | 10 |
1 files changed, 9 insertions, 1 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp b/src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp index b9e2bf6c26..7b08041005 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019 Arm Limited. + * Copyright (c) 2019-2020 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -75,6 +75,14 @@ static const GemmImplementation<uint8_t, uint8_t, Requantize32> gemm_quint8_meth [](const GemmArgs &args, const Requantize32 &) { return args._Nsize<=256 && args._Ksize>128; }, [](const GemmArgs &args, const Requantize32 &qp) { return new GemmHybridQuantized<hybrid_u8u32_dot_16x4, uint8_t, uint8_t>(args, qp); } }, +/** QUANTIZE_WRAPPER_2D enables 2D parallelisation hint for IScheduler in NEGEMMAssemblyDispatch */ +{ + GemmMethod::QUANTIZE_WRAPPER_2D, + "quantized_wrapper_2d", + nullptr, + [](const GemmArgs &args, const Requantize32 &) { return (args._maxthreads >= 8) && (args._Msize >= 8) && (args._Nsize >= 8);}, + [](const GemmArgs &args, const Requantize32 &qp) { return new QuantizeWrapper<uint8_t, uint8_t, uint32_t>(args, qp); } +}, { GemmMethod::QUANTIZE_WRAPPER, "quantized_wrapper", |