aboutsummaryrefslogtreecommitdiff
path: root/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp')
-rw-r--r--src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp11
1 files changed, 9 insertions, 2 deletions
diff --git a/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp b/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
index 810efe539f..a56a73c44a 100644
--- a/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
+++ b/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
@@ -33,8 +33,8 @@
using namespace arm_compute;
-NEDirectConvolutionLayer::NEDirectConvolutionLayer()
- : _accumulate_bias_kernel(), _conv_kernel(), _input_border_handler(), _accumulator()
+NEDirectConvolutionLayer::NEDirectConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
+ : _memory_group(std::move(memory_manager)), _accumulate_bias_kernel(), _conv_kernel(), _input_border_handler(), _accumulator()
{
}
@@ -46,6 +46,9 @@ void NEDirectConvolutionLayer::configure(ITensor *input, const ITensor *weights,
_accumulator.allocator()->free();
}
+ // Manage intermediate buffers
+ _memory_group.manage(&_accumulator);
+
// Allocate the intermediate accumulator tensor in case of fixed point input
switch(output->info()->data_type())
{
@@ -87,6 +90,10 @@ void NEDirectConvolutionLayer::run()
{
NEScheduler::get().schedule(&_input_border_handler, Window::DimZ);
+ _memory_group.acquire();
+
NEScheduler::get().schedule(&_conv_kernel, Window::DimZ);
NEScheduler::get().schedule(&_accumulate_bias_kernel, Window::DimY);
+
+ _memory_group.release();
}