aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp
diff options
context:
space:
mode:
authorGeorgios Pinitas <georgios.pinitas@arm.com>2021-07-16 16:16:43 +0100
committerGeorgios Pinitas <georgios.pinitas@arm.com>2021-07-22 02:25:50 +0000
commit4ee8b1599dbaf7634d25607fa5ac96ba3dc6b0f2 (patch)
tree2f8362d33cdad4212f4b96995681c68184c759e1 /src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp
parent59fd7a722e5bc7e85309d6200bc37a772721a719 (diff)
downloadComputeLibrary-4ee8b1599dbaf7634d25607fa5ac96ba3dc6b0f2.tar.gz
Update GEMM assembly kernels
- Introduce Fp32 kernels with internal calculations in Bfloat16 when fast_mode is enabled - Improve kernel selection heuristics Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I68a9e7e862b6fd2721b46e0d7cc791091c4ab279 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5965 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp')
-rw-r--r--src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp27
1 files changed, 26 insertions, 1 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp b/src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp
index 854b6751c1..07c4769479 100644
--- a/src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp
+++ b/src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4.hpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017-2020 Arm Limited.
+ * Copyright (c) 2017-2021 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -25,6 +25,7 @@
#ifdef __aarch64__
+#include "../performance_parameters.hpp"
#include "../std_transforms_fixed.hpp"
namespace arm_gemm {
@@ -66,6 +67,30 @@ public:
StdTransformsFixed<operand_type, result_type, 4, 4, 16> transforms = {};
StdTransformsFixed<operand_type, result_type, 4, 4, 16, true> transforms_quantized = {};
+ template<typename T>
+ static PerformanceParameters get_performance_parameters(const CPUInfo *ci) {
+ if (std::is_same<T, int32_t>::value) {
+ switch (ci->get_cpu_model()) {
+ case CPUModel::A510:
+ return { 2.64, 2.72, 2.64 };
+
+ default:
+ return { 7.95, 3.76, 7.27 };
+ }
+ }
+
+ if (std::is_same<T, int8_t>::value) {
+ switch(ci->get_cpu_model()) {
+ case CPUModel::A510:
+ return { 2.64, 1.79, 0.10 };
+ default:
+ return { 7.95, 4.09, 0.33 };
+ }
+ }
+
+ return { 0.0 };
+ }
+
kern_type kernel = a64_gemm_u8_4x4;
cls_a64_gemm_u8_4x4(const CPUInfo *) { }