aboutsummaryrefslogtreecommitdiff
path: root/arm_compute/runtime/NEON/functions/NEGEMM.h
diff options
context:
space:
mode:
Diffstat (limited to 'arm_compute/runtime/NEON/functions/NEGEMM.h')
-rw-r--r--arm_compute/runtime/NEON/functions/NEGEMM.h27
1 files changed, 15 insertions, 12 deletions
diff --git a/arm_compute/runtime/NEON/functions/NEGEMM.h b/arm_compute/runtime/NEON/functions/NEGEMM.h
index d4a9f68beb..9df2e08956 100644
--- a/arm_compute/runtime/NEON/functions/NEGEMM.h
+++ b/arm_compute/runtime/NEON/functions/NEGEMM.h
@@ -41,12 +41,15 @@ class NEGEMMInterleave4x4Kernel;
class NEGEMMMatrixAdditionKernel;
class NEGEMMMatrixMultiplyKernel;
class NEGEMMTranspose1xWKernel;
-class NEGEMMAssemblyDispatch;
+namespace cpu
+{
+class CpuGemmAssemblyDispatch;
+}
/** Basic function to execute GEMM. This function calls the following kernels:
*
* If optimized assembly is available:
- * -# @ref NEGEMMAssemblyDispatch
+ * -# @ref cpu::CpuGemmAssemblyDispatch
* -# @ref NEActivationLayer (if alpha != 1.0)
* Else:
* -# @ref NEGEMMInterleave4x4Kernel (if the output tensor is a matrix)
@@ -119,16 +122,16 @@ public:
void prepare() override;
private:
- MemoryGroup _memory_group;
- IWeightsManager *_weights_manager;
- std::unique_ptr<NEGEMMInterleave4x4Kernel> _interleave_kernel;
- std::unique_ptr<NEGEMMTranspose1xWKernel> _transpose_kernel;
- std::unique_ptr<NEGEMMMatrixMultiplyKernel> _mm_kernel;
- std::unique_ptr<NEGEMMAssemblyDispatch> _asm_glue;
- std::unique_ptr<NEGEMMMatrixAdditionKernel> _ma_kernel;
- NEActivationLayer _alpha_scale_func;
- NEArithmeticAddition _add_bias;
- NEActivationLayer _activation_func;
+ MemoryGroup _memory_group;
+ IWeightsManager *_weights_manager;
+ std::unique_ptr<NEGEMMInterleave4x4Kernel> _interleave_kernel;
+ std::unique_ptr<NEGEMMTranspose1xWKernel> _transpose_kernel;
+ std::unique_ptr<NEGEMMMatrixMultiplyKernel> _mm_kernel;
+ std::unique_ptr<cpu::CpuGemmAssemblyDispatch> _asm_glue;
+ std::unique_ptr<NEGEMMMatrixAdditionKernel> _ma_kernel;
+ NEActivationLayer _alpha_scale_func;
+ NEArithmeticAddition _add_bias;
+ NEActivationLayer _activation_func;
Tensor _tmp_a;
Tensor _tmp_b;