diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2020-01-22 18:36:27 +0000 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2020-01-31 11:36:14 +0000 |
commit | 94672fb2af6535adc6ea7fe8b8498580ad8cf3f4 (patch) | |
tree | 189c68e7ff2c29d7800e09239da6d9a6364d5b05 /src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp | |
parent | 6a342648ae50beb8457871862f14fc9baef6b74f (diff) | |
download | ComputeLibrary-94672fb2af6535adc6ea7fe8b8498580ad8cf3f4.tar.gz |
COMPMID-3003: Integrate assembly kernels utilizing MMLA instruction.
MMLA is a matrix-multiply instruction introduced on armv8.6-A
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I572a54981d48f5a1e0e9e51102cb7ae28ad87806
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/2663
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp | 72 |
1 files changed, 72 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp b/src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp new file mode 100644 index 0000000000..f669b870c6 --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8.hpp @@ -0,0 +1,72 @@ +/* + * Copyright (c) 2019 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#pragma once + +#ifdef __aarch64__ + +#include <cstdint> +#include "../std_transforms_fixed.hpp" + +namespace arm_gemm { + +// Actual kernel implementations +void a64_interleaved_s8s32_mmla_12x8(const int8_t *, const int8_t *, int32_t *, int, int, int); + +class interleaved_s8s32_mmla_12x8 { +public: + typedef int8_t operand_type; + typedef int32_t result_type; + + typedef void (*kern_type)(const int8_t *, const int8_t *, int32_t *, int, int, int); + + /* Kernel blocking parameters */ + static unsigned int out_width() + { + return 12; + } + + static unsigned int out_height() + { + return 8; + } + + static unsigned int k_unroll() + { + return 8; + } + + // Use the standard fixed size transforms. + StdTransformsFixed<operand_type, result_type, 8, 12, 8> transforms = {}; + + kern_type kernel=a64_interleaved_s8s32_mmla_12x8; + + interleaved_s8s32_mmla_12x8(const CPUInfo *ci) + { + UNUSED(ci); + } +}; + +} // namespace arm_gemm + +#endif // __aarch64__ |