diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2020-07-02 20:02:20 +0100 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2020-07-06 16:51:32 +0000 |
commit | 5aa1a0b7ca5eed010e4b297a95b1c4851f741328 (patch) | |
tree | ba882de9e86589dfdd33937d538a89bbdf01c40e /src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp | |
parent | 42550c039105597ff6acd4e5efc0ee3c7c20b08e (diff) | |
download | ComputeLibrary-5aa1a0b7ca5eed010e4b297a95b1c4851f741328.tar.gz |
COMPID-3324: Clean GEMM kernels
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I170de1671e061a78740caee31fb4a1b8642c1369
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3505
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp | 21 |
1 files changed, 7 insertions, 14 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp b/src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp index 4897bedf47..f572f7940b 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_interleaved.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 ARM Limited. + * Copyright (c) 2017-2020 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -386,8 +386,8 @@ public: // not multi for now (as this would cause problems with the buffer // manager). ndrange_t get_window_size() const override { - auto m_win_size = (_Mround / strategy::out_height()) * _nbatches; - return { m_win_size, 1u, 1u, 1u, 1u, 1u }; + // _Mround is a multiple of out_height by definition. + return { (_Mround / strategy::out_height()) * _nbatches }; } // set_nthreads: pass on to buffer manager to avoid it waiting for non-existant threads. @@ -399,7 +399,10 @@ public: } // Execute - void execute_1d(unsigned int start, unsigned int end, int threadid) { + void execute(const ndcoord_t &work_range, const ndcoord_t &, int threadid) override { + const auto start = work_range.get_position(0); + const auto end = work_range.get_position_end(0); + if (_pretransposed) { execute_internal<true>(start, end, threadid); } else { @@ -407,16 +410,6 @@ public: } } - //Execute - void execute(const ndcoord_t& work_range, const ndcoord_t& thread_locator, int threadid) override { - UNUSED(thread_locator); - - const auto start = work_range.get_position(0); - const auto stop = work_range.get_position_end(0); - - execute_1d(start, stop, threadid); - } - // Interface implementation - working space size_t get_working_size() const override { // In all cases, we need one A buffer plus a C buffer per thread. |