Enable acl neon backend mul operation (#7105)

Update acl neon backend ShapeFixer's mul
Remove exception in kernel generator for mul

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

diff --git a/runtimes/neurun/backend/acl_neon/KernelGenerator.cc b/runtimes/neurun/backend/acl_neon/KernelGenerator.cc
index 4a5b741..2862392 100644 (file)
--- a/runtimes/neurun/backend/acl_neon/KernelGenerator.cc
+++ b/runtimes/neurun/backend/acl_neon/KernelGenerator.cc
@@ -583,9 +583,6 @@ void KernelGenerator::visit(const model::operation::FullyConnectedNode &node)
 
 void KernelGenerator::visit(const model::operation::MulNode &node)
 {
-  // TODO: fix, tests are failing
-  throw std::runtime_error("NYI");
-
   const auto ofm_index{node.getOutputs().at(0)};
   const auto lhs_index{node.getInputs().at(model::operation::MulNode::Input::LHS)};
   const auto rhs_index{node.getInputs().at(model::operation::MulNode::Input::RHS)};
@@ -600,8 +597,9 @@ void KernelGenerator::visit(const model::operation::MulNode &node)
 
   auto l = nnfw::cpp14::make_unique<::arm_compute::NEPixelWiseMultiplication>();
 
+  // RoundingPolicy for scale:1.0 is only allowed RoundingPolicy::TO_ZERO
   l->configure(lhs_alloc->handle(), rhs_alloc->handle(), ofm_alloc->handle(), 1.0, // scale
-               arm_compute::ConvertPolicy::SATURATE, arm_compute::RoundingPolicy::TO_NEAREST_EVEN);
+               arm_compute::ConvertPolicy::SATURATE, arm_compute::RoundingPolicy::TO_ZERO);
 
   fn = std::move(l);
 

diff --git a/runtimes/neurun/backend/acl_neon/ShapeFixer.cc b/runtimes/neurun/backend/acl_neon/ShapeFixer.cc
index 90753c6..7901ba0 100644 (file)
--- a/runtimes/neurun/backend/acl_neon/ShapeFixer.cc
+++ b/runtimes/neurun/backend/acl_neon/ShapeFixer.cc
@@ -105,17 +105,11 @@ void ShapeFixer::visit(const model::operation::MulNode &node)
   {
     const auto broadcast_rank =
         std::max(_ctx.at(lhs_index).shape().rank(), _ctx.at(rhs_index).shape().rank());
-    auto lhs_shape = _ctx.at(lhs_index).shape();
-    auto rhs_shape = _ctx.at(rhs_index).shape();
 
-    lhs_shape.extendRank(broadcast_rank);
-    rhs_shape.extendRank(broadcast_rank);
-  }
-
-  // Nontrivial broadcasting isn't supported yet
-  if (!(_ctx.at(lhs_index).shape() == _ctx.at(rhs_index).shape()))
-  {
-    throw std::runtime_error("NYI");
+    // TODO remove const_cast later. For example, _ctx may need to be a non const variable or
+    //      a node to extend shape may be inserted in front of this operation
+    const_cast<::neurun::model::Shape &>(_ctx.at(lhs_index).shape()).extendRank(broadcast_rank);
+    const_cast<::neurun::model::Shape &>(_ctx.at(rhs_index).shape()).extendRank(broadcast_rank);
   }
 }
 

diff --git a/tests/nnapi/nnapi_gtest.skip.armv7l-linux.acl_neon b/tests/nnapi/nnapi_gtest.skip.armv7l-linux.acl_neon
index 86daa12..be5a45a 100644 (file)
--- a/tests/nnapi/nnapi_gtest.skip.armv7l-linux.acl_neon
+++ b/tests/nnapi/nnapi_gtest.skip.armv7l-linux.acl_neon
@@ -20,7 +20,6 @@ GeneratedTests.logistic*
 GeneratedTests.lsh_projection*
 GeneratedTests.lstm*
 GeneratedTests.mobilenet*
-GeneratedTests.mul*
 GeneratedTests.neg*
 GeneratedTests.notequal*
 GeneratedTests.prelu_ex*