diff options
author | Omar Al Khatib <omar.alkhatib@arm.com> | 2022-12-20 14:36:45 +0000 |
---|---|---|
committer | Omar Al Khatib <omar.alkhatib@arm.com> | 2022-12-29 09:14:15 +0000 |
commit | 939b21ad4b9ed15d43b4ee8b17484e57ed55a01f (patch) | |
tree | 6c90f78ad73161ffd0f54956a77fa50ef1916819 /src/cpu/kernels/sub/neon/list.h | |
parent | bb1ab0505452691d0e536921cd4c4e2ce32e40ed (diff) | |
download | ComputeLibrary-939b21ad4b9ed15d43b4ee8b17484e57ed55a01f.tar.gz |
Use CPU quantized addition kernel for quantized subtraction
Resolves : [COMPMID-5629]
Signed-off-by: Omar Al Khatib <omar.alkhatib@arm.com>
Change-Id: I061ea5bdafa3a01e66ff869d158f26a38d19e125
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8835
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/kernels/sub/neon/list.h')
-rw-r--r-- | src/cpu/kernels/sub/neon/list.h | 10 |
1 files changed, 7 insertions, 3 deletions
diff --git a/src/cpu/kernels/sub/neon/list.h b/src/cpu/kernels/sub/neon/list.h index ac1346001a..f7e1a040bd 100644 --- a/src/cpu/kernels/sub/neon/list.h +++ b/src/cpu/kernels/sub/neon/list.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021 Arm Limited. + * Copyright (c) 2021-2022 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -35,6 +35,8 @@ namespace cpu #define DECLARE_SUB_KERNEL(func_name) \ void func_name(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) +DECLARE_SUB_KERNEL(sub_qasymm8_neon_fixedpoint); +DECLARE_SUB_KERNEL(sub_qasymm8_signed_neon_fixedpoint); DECLARE_SUB_KERNEL(sub_qasymm8_neon); DECLARE_SUB_KERNEL(sub_qasymm8_signed_neon); DECLARE_SUB_KERNEL(sub_qsymm16_neon); @@ -81,7 +83,8 @@ void sub_same_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); Iterator output(dst, win); - execute_window_loop(win, [&](const Coordinates &) + execute_window_loop( + win, [&](const Coordinates &) { const auto non_broadcast_input_ptr = reinterpret_cast<const T *>(non_broadcast_input.ptr()); const auto output_ptr = reinterpret_cast<T *>(output.ptr()); @@ -127,7 +130,8 @@ void sub_same_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const Iterator input2(src1, input2_win); Iterator output(dst, win); - execute_window_loop(win, [&](const Coordinates &) + execute_window_loop( + win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const T *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const T *>(input2.ptr()); |