Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 2 年 前同步成功

通知 2325
Star 20933
Fork 5424
P
Paddle
提交 ae4e9075 编写于 作者: X xiaoli.liu@intel.com
浏览文件
操作

extract functions for primitive cache

上级 e02a8025
显示空白变更内容
内联 并排
Showing with 178 addition and 196 deletion
paddle/fluid/operators/conv_mkldnn_op.cc
浏览文件 @ ae4e9075
......@@ -33,6 +33,35 @@ using platform::GetMKLDNNFormat;
template <typename T>
class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
public:
struct ConvInfo{
const paddle::framework::Tensor* input;
const paddle::framework::Tensor* bias;
const paddle::framework::Tensor* output;
const paddle::framework::Tensor* weight;
std::vector<int>* strides;
std::vector<int>* paddings;
std::vector<int>* dilations;
std::vector<int>* src_tz;
std::vector<int>* weights_tz;
std::vector<int>* dst_tz;
int g;
};
struct MkldnnInfo{
bool fuse_relu;
bool fuse_residual_conn;
bool force_fp32_output;
bool is_test;
const mkldnn::engine* mkldnn_engine;
std::vector<primitive>* pipeline;
const std::string* key_conv_pd;
std::string* key;
std::shared_ptr<platform::ConvMKLDNNHandler> handler;
std::shared_ptr<mkldnn::memory> src_memory_p;
std::shared_ptr<mkldnn::memory> user_src_memory_p;
std::shared_ptr<mkldnn::memory> dst_memory_p;
std::shared_ptr<mkldnn::convolution_forward> conv_p;
std::shared_ptr<mkldnn::convolution_forward::primitive_desc> conv_pd;
};
void Compute(const paddle::framework::ExecutionContext& ctx) const override {
PADDLE_ENFORCE(paddle::platform::is_cpu_place(ctx.GetPlace()),
......@@ -85,7 +114,6 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
"dilation in convolution is not implemented yet");
const T* input_data = input->data<T>();
const float* filter_data = filter->data<float>();
std::vector<int> src_tz = paddle::framework::vectorize2int(input->dims());
std::vector<int> weights_tz =
......@@ -127,13 +155,13 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
need_s8_to_u8 = true;
}
std::shared_ptr<mkldnn::convolution_forward> conv_p;
std::shared_ptr<mkldnn::memory> src_memory_p;
std::shared_ptr<mkldnn::memory> user_src_memory_p;
std::shared_ptr<mkldnn::memory> dst_memory_p;
std::shared_ptr<mkldnn::convolution_forward> conv_p = nullptr;
std::shared_ptr<mkldnn::memory> src_memory_p = nullptr;
std::shared_ptr<mkldnn::memory> user_src_memory_p = nullptr;
std::shared_ptr<mkldnn::memory> dst_memory_p = nullptr;
std::vector<primitive> pipeline;
std::shared_ptr<mkldnn::convolution_forward::primitive_desc> conv_pd;
std::shared_ptr<platform::ConvMKLDNNHandler> handler;
std::shared_ptr<mkldnn::convolution_forward::primitive_desc> conv_pd = nullptr;
std::shared_ptr<platform::ConvMKLDNNHandler> handler = nullptr;
auto prim_key = key + "@conv_p";
auto dst_key = key + "@dst_mem_p";
......@@ -142,42 +170,38 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
auto src_reorder_key = key + "@src_mem_preorder_p";
conv_p = std::static_pointer_cast<mkldnn::convolution_forward>(dev_ctx.GetBlob(prim_key));
if(conv_p == nullptr){
struct ConvInfo convinfo;
struct MkldnnInfo mkldnninfo;
convinfo.strides = &strides;
convinfo.paddings = &paddings;
convinfo.dilations = &dilations;
convinfo.src_tz = &src_tz;
convinfo.weights_tz = &weights_tz;
convinfo.dst_tz = &dst_tz;
convinfo.g = g;
mkldnninfo.fuse_relu = fuse_relu;
mkldnninfo.fuse_residual_conn = fuse_residual_conn;
mkldnninfo.force_fp32_output = force_fp32_output;
mkldnninfo.is_test = is_test;
mkldnninfo.mkldnn_engine = &mkldnn_engine;
mkldnninfo.handler = handler;
mkldnninfo.pipeline = &pipeline;
mkldnninfo.key_conv_pd = &key_conv_pd;
mkldnninfo.key = &key;
mkldnninfo.src_memory_p = src_memory_p;
mkldnninfo.user_src_memory_p = user_src_memory_p;
mkldnninfo.dst_memory_p = dst_memory_p;
mkldnninfo.conv_p = conv_p;
mkldnninfo.conv_pd = conv_pd;
if(is_INT8){
CreateINT8Primitive(ctx, is_test, dev_ctx, mkldnn_engine, input, //filter,
bias, output,
strides, paddings,
dilations, fuse_relu,
fuse_residual_conn, input_data,
filter_data, src_tz,
weights_tz, g,
dst_tz, key,
dst_memory_p,
pipeline,
key_conv_pd,
src_memory_p,
user_src_memory_p,
conv_p,
conv_pd,
handler,
force_fp32_output);
CreateINT8Primitive(ctx, &dev_ctx, input, filter, bias, output, &convinfo, &mkldnninfo);
}else{
CreateFP32Primitive(ctx, is_test, dev_ctx, mkldnn_engine, input, //filter,
bias, output,
strides, paddings,
dilations, fuse_relu,
fuse_residual_conn, input_data,
filter_data, src_tz,
weights_tz, g,
dst_tz, key,
dst_memory_p,
pipeline,
key_conv_pd,
src_memory_p,
user_src_memory_p,
conv_p,
conv_pd,
handler);
CreateFP32Primitive(ctx, &dev_ctx, input, filter, bias, output, &convinfo, &mkldnninfo);
}
//src_memory_p = mkldnninfo.src_memory_p;
//user_src_memory_p = mkldnninfo.user_src_memory_p;
dst_memory_p = mkldnninfo.dst_memory_p;
//conv_p = mkldnninfo.conv_p;
} else {
auto src_memory_reorder_p = std::static_pointer_cast<mkldnn::memory>(dev_ctx.GetBlob(src_reorder_key));
src_memory_p = std::static_pointer_cast<mkldnn::memory>(dev_ctx.GetBlob(src_key));
......@@ -267,33 +291,18 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
private:
void CreateFP32Primitive(
paddle::framework::ExecutionContext ctx, bool is_test,
const paddle::platform::MKLDNNDeviceContext& dev_ctx,
const mkldnn::engine& mkldnn_engine,
const paddle::framework::Tensor* input,// const paddle::framework::Tensor* filter,
const paddle::framework::ExecutionContext& ctx,
const paddle::platform::MKLDNNDeviceContext* dev_ctx,
const paddle::framework::Tensor* input, const paddle::framework::Tensor* filter,
const paddle::framework::Tensor* bias, paddle::framework::Tensor* output,
std::vector<int> strides, std::vector<int> paddings,
std::vector<int> dilations, bool fuse_relu,
bool fuse_residual_conn, const T* input_data,
const float* filter_data, std::vector<int> src_tz,
std::vector<int> weights_tz, int g,
std::vector<int> dst_tz, const std::string key,
std::shared_ptr<mkldnn::memory> &dst_memory_p,
std::vector<primitive>& pipeline,
const std::string &key_conv_pd,
std::shared_ptr<mkldnn::memory> src_memory_p,
std::shared_ptr<mkldnn::memory> user_src_memory_p,
std::shared_ptr<mkldnn::convolution_forward> conv_p,
std::shared_ptr<mkldnn::convolution_forward::primitive_desc> conv_pd,
std::shared_ptr<platform::ConvMKLDNNHandler> handler) const{
//const T* input_data = input->data<T>();
ConvInfo* convinfo, MkldnnInfo* mkldnninfo) const{
const T* input_data = input->data<T>();
const float* filter_data = filter->data<float>();
auto user_src_md = platform::MKLDNNMemDesc(
{src_tz}, platform::MKLDNNGetDataType<T>(), input->format());
{*(convinfo->src_tz)}, platform::MKLDNNGetDataType<T>(), input->format());
auto user_weights_md = platform::MKLDNNMemDesc(
{weights_tz}, platform::MKLDNNGetDataType<T>(),
(g == 1) ? mkldnn::memory::format::oihw : mkldnn::memory::format::goihw);
{*(convinfo->weights_tz)}, platform::MKLDNNGetDataType<T>(),
(convinfo->g == 1) ? mkldnn::memory::format::oihw : mkldnn::memory::format::goihw);
/* create memory descriptor for convolution without specified format
* ('any') which lets a primitive (convolution in this case) choose
......@@ -304,46 +313,51 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
platform::data_format_to_memory_format(data_format);
auto src_md = platform::MKLDNNMemDesc(
src_tz, platform::MKLDNNGetDataType<T>(), chosen_memory_format);
*(convinfo->src_tz), platform::MKLDNNGetDataType<T>(), chosen_memory_format);
auto weights_md = platform::MKLDNNMemDesc(
weights_tz, platform::MKLDNNGetDataType<T>(), chosen_memory_format);
*(convinfo->weights_tz), platform::MKLDNNGetDataType<T>(), chosen_memory_format);
std::vector<int> bias_tz; // TODO(mgallus): avoid empty vector creation.
// Currently used whenever bias is != nullptr.
auto dst_md = platform::MKLDNNMemDesc(
dst_tz, platform::MKLDNNGetDataType<T>(), chosen_memory_format);
*(convinfo->dst_tz), platform::MKLDNNGetDataType<T>(), chosen_memory_format);
// create a conv primitive descriptor and save it for usage in backward
if (bias) {
bias_tz = paddle::framework::vectorize2int(bias->dims());
auto bias_md = platform::MKLDNNMemDesc(
bias_tz, platform::MKLDNNGetDataType<T>(), memory::format::x);
conv_pd = ConvFwdPrimitiveDesc(src_md, weights_md, bias_md, dst_md,
strides, paddings, mkldnn_engine,
fuse_relu, fuse_residual_conn, is_test);
mkldnninfo->conv_pd = ConvFwdPrimitiveDesc(src_md, weights_md, bias_md, dst_md,
*convinfo->strides, *convinfo->paddings,
*mkldnninfo->mkldnn_engine,
mkldnninfo->fuse_relu, mkldnninfo->fuse_residual_conn,
mkldnninfo->is_test);
} else {
conv_pd =
ConvFwdPrimitiveDesc(src_md, weights_md, dst_md, strides, paddings,
mkldnn_engine, fuse_relu, fuse_residual_conn, is_test);
mkldnninfo->conv_pd =
ConvFwdPrimitiveDesc(src_md, weights_md, dst_md,
*convinfo->strides, *convinfo->paddings,
*mkldnninfo->mkldnn_engine,
mkldnninfo->fuse_relu, mkldnninfo->fuse_residual_conn,
mkldnninfo->is_test);
}
// Save conv_pd/src_memory/weights_memory for backward pass
dev_ctx.SetBlob(key_conv_pd, conv_pd);
dev_ctx->SetBlob(*mkldnninfo->key_conv_pd, mkldnninfo->conv_pd);
handler.reset(new platform::ConvMKLDNNHandler(conv_pd, dev_ctx, mkldnn_engine, key));
mkldnninfo->handler.reset(new platform::ConvMKLDNNHandler(mkldnninfo->conv_pd, *dev_ctx, *mkldnninfo->mkldnn_engine, *mkldnninfo->key));
// create mkldnn memory from input tensors (data/weights)
user_src_memory_p =
handler->AcquireSrcMemory(user_src_md, to_void_cast<T>(input_data));
auto user_weights_memory_p = handler->AcquireWeightsMemory(
mkldnninfo->user_src_memory_p =
mkldnninfo->handler->AcquireSrcMemory(user_src_md, to_void_cast<T>(input_data));
auto user_weights_memory_p = mkldnninfo->handler->AcquireWeightsMemory(
user_weights_md, to_void_cast<float>(filter_data));
// create reorder primitive if the input format is not the preferred one
src_memory_p =
handler->AcquireSrcMemoryFromPrimitive(user_src_memory_p, pipeline);
auto weights_memory_p = handler->AcquireWeightsMemoryFromPrimitive(
user_weights_memory_p, pipeline, is_test);
mkldnninfo->src_memory_p =
mkldnninfo->handler->AcquireSrcMemoryFromPrimitive(mkldnninfo->user_src_memory_p, *mkldnninfo->pipeline);
auto weights_memory_p = mkldnninfo->handler->AcquireWeightsMemoryFromPrimitive(
user_weights_memory_p, *mkldnninfo->pipeline, mkldnninfo->is_test);
if (fuse_residual_conn) {
if (mkldnninfo->fuse_residual_conn) {
auto residual_param = ctx.Input<Tensor>("ResidualData");
auto residual_param_data = residual_param->data<T>();
......@@ -354,9 +368,9 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
"Output and elementwise parameter need to have the "
"same dimension sizes");
if (residual_param->format() != handler->GetDstFormat()) {
if (residual_param->format() != mkldnninfo->handler->GetDstFormat()) {
auto output_data =
output->mutable_data<T>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, handler->GetDstMemorySize());
output->mutable_data<T>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, mkldnninfo->handler->GetDstMemorySize());
auto residual_data_tz =
paddle::framework::vectorize2int(residual_param->dims());
auto residual_data_type =
......@@ -364,21 +378,21 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
auto user_residual_md = platform::MKLDNNMemDesc(
residual_data_tz, residual_data_type, residual_param->format());
auto user_residual_memory_p = handler->AcquireResidualDataMemory(
auto user_residual_memory_p = mkldnninfo->handler->AcquireResidualDataMemory(
user_residual_md, to_void_cast<T>(residual_param_data));
dst_memory_p = handler->AcquireDstMemoryFromResidualDataMemory(
user_residual_memory_p, to_void_cast<T>(output_data), pipeline);
mkldnninfo->dst_memory_p = mkldnninfo->handler->AcquireDstMemoryFromResidualDataMemory(
user_residual_memory_p, to_void_cast<T>(output_data), *mkldnninfo->pipeline);
} else {
output->ShareDataWith(*residual_param);
auto output_data = output->mutable_data<T>(ctx.GetPlace());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<T>(output_data));
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<T>(output_data));
}
} else {
auto output_data =
output->mutable_data<T>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, handler->GetDstMemorySize());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<T>(output_data));
output->mutable_data<T>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, mkldnninfo->handler->GetDstMemorySize());
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<T>(output_data));
}
// create convolution op primitive
......@@ -387,72 +401,41 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
auto user_bias_md = platform::MKLDNNMemDesc(
{bias_tz}, platform::MKLDNNGetDataType<T>(), memory::format::x);
auto user_bias_memory_p =
handler->AcquireBiasMemory(user_bias_md, to_void_cast<T>(bias_data));
mkldnninfo->handler->AcquireBiasMemory(user_bias_md, to_void_cast<T>(bias_data));
auto bias_memory_p =
handler->AcquireBiasMemoryFromPrimitive(user_bias_memory_p, pipeline, is_test);
conv_p = handler->AcquireConvolution(src_memory_p, weights_memory_p,
bias_memory_p, dst_memory_p);
mkldnninfo->handler->AcquireBiasMemoryFromPrimitive(user_bias_memory_p, *mkldnninfo->pipeline, mkldnninfo->is_test);
mkldnninfo->conv_p = mkldnninfo->handler->AcquireConvolution(
mkldnninfo->src_memory_p, weights_memory_p,
bias_memory_p, mkldnninfo->dst_memory_p);
} else {
conv_p = handler->AcquireConvolution(src_memory_p, weights_memory_p,
dst_memory_p);
mkldnninfo->conv_p = mkldnninfo->handler->AcquireConvolution(
mkldnninfo->src_memory_p, weights_memory_p,
mkldnninfo->dst_memory_p);
}
// push primitive to stream and wait until it's executed
pipeline.push_back(*conv_p);
mkldnninfo->pipeline->push_back(*mkldnninfo->conv_p);
};
void CreateINT8Primitive(
const paddle::framework::ExecutionContext& ctx, bool is_test,
const paddle::platform::MKLDNNDeviceContext & dev_ctx,
const mkldnn::engine & mkldnn_engine,
const paddle::framework::Tensor* input, //const paddle::framework::Tensor* filter,
const paddle::framework::ExecutionContext& ctx,
const paddle::platform::MKLDNNDeviceContext* dev_ctx,
const paddle::framework::Tensor* input, const paddle::framework::Tensor* filter,
const paddle::framework::Tensor* bias, paddle::framework::Tensor* output,
std::vector<int> strides, std::vector<int> paddings,
std::vector<int> dilations, bool fuse_relu,
bool fuse_residual_conn, const T* input_data,
const float* filter_data, std::vector<int> src_tz,
std::vector<int> weights_tz, int g,
std::vector<int> dst_tz, const std::string key,
std::shared_ptr<mkldnn::memory>& dst_memory_p,
std::vector<primitive>& pipeline,
const std::string &key_conv_pd,
std::shared_ptr<mkldnn::memory> src_memory_p,
std::shared_ptr<mkldnn::memory> user_src_memory_p,
std::shared_ptr<mkldnn::convolution_forward> conv_p,
std::shared_ptr<mkldnn::convolution_forward::primitive_desc> conv_pd,
std::shared_ptr<platform::ConvMKLDNNHandler> handler,
bool force_fp32_output) const {
//const T* input_data = input->data<T>();
ConvInfo* convinfo, MkldnnInfo* mkldnninfo) const {
const T* input_data = input->data<T>();
const float* filter_data = filter->data<float>();
bool is_INT8 = true;
auto scale_in_data = ctx.Attr<float>("Scale_in");
auto scale_in_eltwise_data = ctx.Attr<float>("Scale_in_eltwise");
auto scale_weights_data = ctx.Attr<std::vector<float>>("Scale_weights");
auto scale_out_data = force_fp32_output? 1.0f : ctx.Attr<float>("Scale_out");
auto scale_out_data = mkldnninfo->force_fp32_output? 1.0f : ctx.Attr<float>("Scale_out");
bool is_multi_channel = scale_weights_data.size() > 1 ? true : false;
auto scale_in_key = key + "@scale_in";
auto scale_weights_key = key + "@scale_weights";
auto scale_out_key = key + "@scale_out";
auto output_shift_scale_key = key + "@output_shift_scale";
auto sum_scale_key = key + "@sum_scale";
auto scale_in_eltwise_key = key + "@scale_in_eltwise";
//std::vector<float> scale_in_data;
//std::vector<float> scale_out_data = {1.0f};
//std::vector<float> scale_weights_data;
//std::vector<float> scale_in_eltwise_data;
std::vector<float> output_shift_scale;
float sum_scale = 1.0f;
int count = is_multi_channel? (g>1? weights_tz[1]*weights_tz[0] : weights_tz[0]) : 1;
//scale_in_data = {scale_in};
//scale_weights_data.resize(count);
//#pragma omp parallel for if (count > 1)
//for(int i=0; i<count; i++){
//scale_weights_data[i] =*(scale_weights->data<float>() + i);
//}
//if(!force_fp32_output)
//scale_out_data = {*(scale_out->data<float>())};
int count = is_multi_channel? (convinfo->g>1? (*convinfo->weights_tz)[1] * (*convinfo->weights_tz)[0] : (*convinfo->weights_tz)[0]) : 1;
output_shift_scale.resize(count);
#pragma omp parallel for if (count > 1)
for(int i=0; i<count; i++){
......@@ -461,16 +444,15 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
else
output_shift_scale[i] = scale_out_data / (scale_in_data * scale_weights_data[i]);
}
if(fuse_residual_conn){
//scale_in_eltwise_data = {*(scale_in_eltwise->data<float>())};
if(mkldnninfo->fuse_residual_conn){
sum_scale = scale_out_data / scale_in_eltwise_data;
}
auto user_src_md = platform::MKLDNNMemDesc(
{src_tz}, paddle::framework::ToMKLDNNDataType(input->type()), input->format());
{*convinfo->src_tz}, paddle::framework::ToMKLDNNDataType(input->type()), input->format());
auto user_weights_md = platform::MKLDNNMemDesc(
{weights_tz}, platform::MKLDNNGetDataType<float>(),
(g == 1) ? mkldnn::memory::format::oihw : mkldnn::memory::format::goihw);
{*convinfo->weights_tz}, platform::MKLDNNGetDataType<float>(),
((convinfo->g) == 1) ? mkldnn::memory::format::oihw : mkldnn::memory::format::goihw);
/* create memory descriptor for convolution without specified format
* ('any') which lets a primitive (convolution in this case) choose
......@@ -483,123 +465,123 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
auto bias_tz = paddle::framework::vectorize2int(bias->dims());
auto src_md = platform::MKLDNNMemDesc(
src_tz, memory::data_type::u8, chosen_memory_format);
*convinfo->src_tz, memory::data_type::u8, chosen_memory_format);
auto weights_md = platform::MKLDNNMemDesc(
weights_tz, memory::data_type::s8, chosen_memory_format);
*convinfo->weights_tz, memory::data_type::s8, chosen_memory_format);
auto dst_dt = fuse_relu?
auto dst_dt = mkldnninfo->fuse_relu?
paddle::framework::ToMKLDNNDataType(std::type_index(typeid(unsigned char)))
: paddle::framework::ToMKLDNNDataType(std::type_index(typeid(signed char)));
if(force_fp32_output){
if(mkldnninfo->force_fp32_output){
dst_dt = paddle::framework::ToMKLDNNDataType(std::type_index(typeid(float)));
}
if(fuse_residual_conn){
if(mkldnninfo->fuse_residual_conn){
auto residual = ctx.Input<Tensor>("ResidualData");
auto residual_dt = paddle::framework::ToMKLDNNDataType(residual->type());
if(dst_dt != residual_dt)
dst_dt = residual_dt;
}
auto dst_md = platform::MKLDNNMemDesc(dst_tz, dst_dt, chosen_memory_format);
auto dst_md = platform::MKLDNNMemDesc(*convinfo->dst_tz, dst_dt, chosen_memory_format);
// create a conv primitive descriptor and save it for usage in backward
if (bias) {
auto bias_md = platform::MKLDNNMemDesc(
bias_tz, memory::data_type::s32, memory::format::x);
conv_pd = ConvFwdPrimitiveDesc(src_md, weights_md, bias_md, dst_md,
strides, paddings, mkldnn_engine,
fuse_relu, fuse_residual_conn,
output_shift_scale, sum_scale, is_test);
mkldnninfo->conv_pd = ConvFwdPrimitiveDesc(src_md, weights_md, bias_md, dst_md,
*convinfo->strides, *convinfo->paddings, *mkldnninfo->mkldnn_engine,
mkldnninfo->fuse_relu, mkldnninfo->fuse_residual_conn,
output_shift_scale, sum_scale, mkldnninfo->is_test);
} else {
conv_pd =
ConvFwdPrimitiveDesc(src_md, weights_md, dst_md, strides, paddings,
mkldnn_engine, fuse_relu, fuse_residual_conn,
output_shift_scale, sum_scale, is_test);
mkldnninfo->conv_pd =
ConvFwdPrimitiveDesc(src_md, weights_md, dst_md, *convinfo->strides, *convinfo->paddings,
*mkldnninfo->mkldnn_engine, mkldnninfo->fuse_relu, mkldnninfo->fuse_residual_conn,
output_shift_scale, sum_scale, mkldnninfo->is_test);
}
// Save conv_pd/src_memory/weights_memory for backward pass
dev_ctx.SetBlob(key_conv_pd, conv_pd);
dev_ctx->SetBlob(*mkldnninfo->key_conv_pd, mkldnninfo->conv_pd);
handler.reset(new platform::ConvMKLDNNHandler(conv_pd, dev_ctx, mkldnn_engine, key));
mkldnninfo->handler.reset(new platform::ConvMKLDNNHandler(mkldnninfo->conv_pd, *dev_ctx, *mkldnninfo->mkldnn_engine, *mkldnninfo->key));
// create mkldnn memory from input tensors (data/weights)
user_src_memory_p =
handler->AcquireSrcMemory(user_src_md, to_void_cast<T>(input_data));
auto user_weights_memory_p = handler->AcquireWeightsMemory(
mkldnninfo->user_src_memory_p =
mkldnninfo->handler->AcquireSrcMemory(user_src_md, to_void_cast<T>(input_data));
auto user_weights_memory_p = mkldnninfo->handler->AcquireWeightsMemory(
user_weights_md, to_void_cast<float>(filter_data));
// create reorder primitive if the input format is not the preferred one
src_memory_p =
handler->AcquireSrcMemoryFromPrimitive(user_src_memory_p, pipeline);
mkldnninfo->src_memory_p =
mkldnninfo->handler->AcquireSrcMemoryFromPrimitive(mkldnninfo->user_src_memory_p, *mkldnninfo->pipeline);
std::shared_ptr<mkldnn::memory> weights_memory_p;
int mask_reorder = is_multi_channel? ((g!= 1) ? (1<<1)+(1<<0) : 1<<0) : 0;
weights_memory_p = handler->AcquireWeightsMemoryFromPrimitive(
user_weights_memory_p, pipeline, is_test, is_INT8, scale_weights_data, mask_reorder);
int mask_reorder = is_multi_channel? ((convinfo->g!= 1) ? (1<<1)+(1<<0) : 1<<0) : 0;
weights_memory_p = mkldnninfo->handler->AcquireWeightsMemoryFromPrimitive(
user_weights_memory_p, *mkldnninfo->pipeline, mkldnninfo->is_test, is_INT8, scale_weights_data, mask_reorder);
if(fuse_residual_conn) {
if(mkldnninfo->fuse_residual_conn) {
auto residual_param = ctx.Input<Tensor>("ResidualData");
PADDLE_ENFORCE_EQ(output->dims(), residual_param->dims(),
"Output and elementwise parameter need to have the "
"same dimension sizes");
auto residual_dt = paddle::framework::ToMKLDNNDataType(residual_param->type());
PADDLE_ENFORCE_EQ(residual_param->format(), handler->GetDstFormat(),
PADDLE_ENFORCE_EQ(residual_param->format(), mkldnninfo->handler->GetDstFormat(),
"Conv input dimension and filter dimension should be the same.");
output->ShareDataWith(*residual_param);
if(residual_dt == mkldnn::memory::data_type::u8){
uint8_t* output_data = output->mutable_data<uint8_t>(ctx.GetPlace());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<uint8_t>(output_data));
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<uint8_t>(output_data));
} else{
int8_t* output_data = output->mutable_data<int8_t>(ctx.GetPlace());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<int8_t>(output_data));
}
} else if(!force_fp32_output){
if(fuse_relu){
uint8_t* output_data = output->mutable_data<uint8_t>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, handler->GetDstMemorySize());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<uint8_t>(output_data));
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<int8_t>(output_data));
}
} else if(!mkldnninfo->force_fp32_output){
if(mkldnninfo->fuse_relu){
uint8_t* output_data = output->mutable_data<uint8_t>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, mkldnninfo->handler->GetDstMemorySize());
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<uint8_t>(output_data));
} else{
int8_t* output_data = output->mutable_data<int8_t>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, handler->GetDstMemorySize());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<int8_t>(output_data));
int8_t* output_data = output->mutable_data<int8_t>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, mkldnninfo->handler->GetDstMemorySize());
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<int8_t>(output_data));
}
} else {
float* output_data = output->mutable_data<float>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, handler->GetDstMemorySize());
dst_memory_p =
handler->AcquireDstMemoryFromPrimitive(to_void_cast<float>(output_data));
float* output_data = output->mutable_data<float>(ctx.GetPlace(), ::paddle::memory::Allocator::kDefault, mkldnninfo->handler->GetDstMemorySize());
mkldnninfo->dst_memory_p =
mkldnninfo->handler->AcquireDstMemoryFromPrimitive(to_void_cast<float>(output_data));
}
// create convolution op primitive
std::vector<float> scale_bias_data;
auto scale_bias_key = key + "@scale_bias";
auto scale_bias_key = *mkldnninfo->key + "@scale_bias";
if (bias) {
const float* bias_data = bias->data<float>();
auto user_bias_md = platform::MKLDNNMemDesc(
{bias_tz}, platform::MKLDNNGetDataType<float>(), memory::format::x);
auto user_bias_memory_p =
handler->AcquireBiasMemory(user_bias_md, to_void_cast<float>(bias_data));
mkldnninfo->handler->AcquireBiasMemory(user_bias_md, to_void_cast<float>(bias_data));
std::shared_ptr<mkldnn::memory> bias_memory_p;
int mask_reorder = is_multi_channel? 1<<0 : 1;
int count = is_multi_channel? (g>1? weights_tz[1]*weights_tz[0] : weights_tz[0]) : 1;
int count = is_multi_channel? (convinfo->g>1? (*convinfo->weights_tz)[1] * (*convinfo->weights_tz)[0] : (*convinfo->weights_tz)[0]) : 1;
scale_bias_data.resize(count);
#pragma omp parallel for if (count > 1)
for(int i=0; i<count; i++){
scale_bias_data[i] = scale_in_data * scale_weights_data[i];
}
bias_memory_p =
handler->AcquireBiasMemoryFromPrimitive(user_bias_memory_p, pipeline, is_test, is_INT8, scale_bias_data, mask_reorder);
conv_p = handler->AcquireConvolution(src_memory_p, weights_memory_p,
bias_memory_p, dst_memory_p);
mkldnninfo->handler->AcquireBiasMemoryFromPrimitive(user_bias_memory_p, *mkldnninfo->pipeline, mkldnninfo->is_test, is_INT8, scale_bias_data, mask_reorder);
mkldnninfo->conv_p = mkldnninfo->handler->AcquireConvolution(mkldnninfo->src_memory_p, weights_memory_p,
bias_memory_p, mkldnninfo->dst_memory_p);
} else {
conv_p = handler->AcquireConvolution(src_memory_p, weights_memory_p,
dst_memory_p);
mkldnninfo->conv_p = mkldnninfo->handler->AcquireConvolution(mkldnninfo->src_memory_p, weights_memory_p,
mkldnninfo->dst_memory_p);
}
// push primitive to stream and wait until it's executed
pipeline.push_back(*conv_p);
mkldnninfo->pipeline->push_back(*mkldnninfo->conv_p);
};
void AppendKey(std::string& key, mkldnn::memory::dims& input_dims, // NOLINT
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或