Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

mvn optimization for 1D and 2D input #5414

Merged
merged 4 commits into from
May 12, 2021
Merged

mvn optimization for 1D and 2D input #5414

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
57189ab
optimized 1D and 2D execution
chenhu-wang Apr 27, 2021
b75ee1f
shape fine tune for unified 5d precedure
chenhu-wang Apr 28, 2021
7607cb6
update per review
chenhu-wang May 4, 2021
a5c38d1
rebase on master conflict resolve
chenhu-wang May 7, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
57 changes: 38 additions & 19 deletions inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -506,7 +506,7 @@ struct jit_uni_mvn_kernel_f32 : public jit_uni_mvn_kernel, public jit_generator
Xbyak::Reg64 reg_d_bias = rdx;

Xbyak::Reg64 reg_load_table = r15;
Xbyak::Reg64 reg_load_store_mask = rcx;
Xbyak::Reg64 reg_load_store_mask = rbp;

Vmm vmm_val = Vmm(1);
Vmm vmm_mean = Vmm(0);
Expand Down Expand Up @@ -672,6 +672,7 @@ MKLDNNMVNNode::MKLDNNMVNNode(const std::shared_ptr<ngraph::Node>& op, const mkld
IE_THROW(NotImplemented) << errorMessage;
}

const ngraph::Shape& inDataShape = op->input_value(0).get_shape();
if (auto mvnOp = ngraph::as_type_ptr<ngraph::op::v6::MVN>(op)) {
normalizeVariance_ = mvnOp->get_normalize_variance();
epsValue_ = mvnOp->get_eps();
Expand All @@ -681,7 +682,7 @@ MKLDNNMVNNode::MKLDNNMVNNode(const std::shared_ptr<ngraph::Node>& op, const mkld
}

acrossChannels_ = false;
const auto& inDataShapeSize = op->input_value(0).get_shape().size();
const auto& inDataShapeSize = inDataShape.size();
if (inDataShapeSize == mvnOp->input_value(1).get_shape()[0] + 1 || inDataShapeSize == 1)
acrossChannels_ = true;
} else if (auto mvnOp = ngraph::as_type_ptr<ngraph::op::v0::MVN>(op)) {
Expand All @@ -690,6 +691,37 @@ MKLDNNMVNNode::MKLDNNMVNNode(const std::shared_ptr<ngraph::Node>& op, const mkld
epsMode_ = INSIDE_SQRT;
acrossChannels_ = mvnOp->get_across_channels();
}

transformTo5DCase(inDataShape);
}

void MKLDNNMVNNode::transformTo5DCase(const ngraph::Shape& shape) {
switch (shape.size()) {
// for 1 and 2 rank, if acrossChannels_ is true, adjust shape to fully vectorize under unified 5d procedure.
// otherwise there are not enough data in spatial dimension to process in one kernel.
case 1 : // C
if (acrossChannels_) {
shape5D = std::make_tuple(1, 1, 1, 1, shape[0]);
acrossChannels_ = false;
break;
} else {
shape5D = std::make_tuple(1, shape[0], 1, 1, 1);
break;
}
case 2 : // NC
if (acrossChannels_) {
shape5D = std::make_tuple(1, shape[0], 1, shape[1], 1);
acrossChannels_ = false;
break;
} else {
shape5D = std::make_tuple(shape[0], shape[1], 1, 1, 1);
break;
}
case 3 : { shape5D = std::make_tuple(shape[0], shape[1], 1, shape[2], 1); break; }
case 4 : { shape5D = std::make_tuple(shape[0], shape[1], 1, shape[2], shape[3]); break; }
case 5 : { shape5D = std::make_tuple(shape[0], shape[1], shape[2], shape[3], shape[4]); break; }
default : { IE_THROW() << "MVN layer with name '" << getName() << "' doesn't support planar layout with rank: " << shape.size(); }
}
}

void MKLDNNMVNNode::getSupportedDescriptors() {
Expand Down Expand Up @@ -798,19 +830,6 @@ void MKLDNNMVNNode::initSupportedPrimitiveDescriptors() {
pushDesc(MKLDNNMemory::GetPlainFormat(getChildEdgeAt(0)->getDims()), impl_type);
}

std::tuple<size_t, size_t, size_t, size_t, size_t> MKLDNNMVNNode::get5dShapes(const SizeVector& dims) {
std::tuple<size_t, size_t, size_t, size_t, size_t> shapes;
switch (dims.size()) {
case 1 : { shapes = std::make_tuple(1, dims[0], 1, 1, 1); break; }
case 2 : { shapes = std::make_tuple(dims[0], dims[1], 1, 1, 1); break; }
case 3 : { shapes = std::make_tuple(dims[0], dims[1], 1, dims[2], 1); break; }
case 4 : { shapes = std::make_tuple(dims[0], dims[1], 1, dims[2], dims[3]); break; }
case 5 : { shapes = std::make_tuple(dims[0], dims[1], dims[2], dims[3], dims[4]); break; }
default : { IE_THROW() << "MVN layer with name '" << getName() << "' doesn't support planar layout with rank: " << dims.size(); }
}
return shapes;
}

void MKLDNNMVNNode::createPrimitive() {
auto& dstMemPtr = getChildEdgeAt(0)->getMemoryPtr();
auto& srcMemPtr = getParentEdgeAt(0)->getMemoryPtr();
Expand All @@ -832,7 +851,7 @@ void MKLDNNMVNNode::createPrimitive() {
jcp.across_channels = acrossChannels_;
SizeVector in_dims = getParentEdgeAt(0)->getDims().ToSizeVector();
int N = 0;
std::tie(N, jcp.C, jcp.D, jcp.H, jcp.W) = get5dShapes(in_dims);
std::tie(N, jcp.C, jcp.D, jcp.H, jcp.W) = shape5D;

if (mayiuse(cpu::x64::avx512_common)) {
mvn_kernel.reset(new jit_uni_mvn_kernel_f32<cpu::x64::avx512_common>(jcp, *attr.get()));
Expand Down Expand Up @@ -926,7 +945,7 @@ void MKLDNNMVNNode::mvn_pln(const uint8_t* src_data, uint8_t* dst_data, const Si
}

size_t N = 0; size_t C = 0; size_t D = 0; size_t H = 0; size_t W = 0;
std::tie(N, C, D, H, W) = get5dShapes(dims);
std::tie(N, C, D, H, W) = shape5D;

size_t C1 = H * W;
size_t C2 = C1 * D;
Expand Down Expand Up @@ -1054,7 +1073,7 @@ void MKLDNNMVNNode::mvn_ref(const uint8_t* src_data, uint8_t* dst_data, const Si
const float *src_data_ptr = reinterpret_cast<const float *>(src_data);
float *dst_data_ptr = reinterpret_cast<float *>(dst_data);
size_t N = 0; size_t C = 0; size_t D = 0; size_t H = 0; size_t W = 0;
std::tie(N, C, D, H, W) = get5dShapes(dims);
std::tie(N, C, D, H, W) = shape5D;

size_t C1 = H * W;
size_t C2 = C1 * D;
Expand Down Expand Up @@ -1157,7 +1176,7 @@ void MKLDNNMVNNode::mvn_blk(const uint8_t* src_data, uint8_t* dst_data, const Si
}

size_t N = 1; size_t C = 1; size_t D = 1; size_t H = 1; size_t W = 1;
std::tie(N, C, D, H, W) = get5dShapes(dims);
std::tie(N, C, D, H, W) = shape5D;

bool is_nhwc = false;
Layout layout = getParentEdgeAt(0)->getDesc().getLayout();
Expand Down
4 changes: 3 additions & 1 deletion inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,7 +104,9 @@ class MKLDNNMVNNode : public MKLDNNNode {

void setPostOps(mkldnn::primitive_attr &attr, bool initWeights = false);

std::tuple<size_t, size_t, size_t, size_t, size_t> get5dShapes(const InferenceEngine::SizeVector& dims);
void transformTo5DCase(const ngraph::Shape& shape);

std::tuple<size_t, size_t, size_t, size_t, size_t> shape5D;

bool acrossChannels_ = false;
bool normalizeVariance_ = true;
Expand Down