Skip to content

GPU and multi-core accelerated neural networks in TMVA. #231

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

Closed
simonpf wants to merge 3 commits into from
Closed

GPU and multi-core accelerated neural networks in TMVA. #231

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8287099
GPU and multi-core accelerated DNNs in TMVA.
Aug 30, 2016
fac5cd7
Removed __restrict__ keyword for mac compatibility.
Aug 31, 2016
48d0417
Fixed linking to TBB library.
Sep 1, 2016
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
Prev Previous commit
Next Next commit
Removed __restrict__ keyword for mac compatibility.
  • Loading branch information
Simon Pfreundschuh committed Aug 31, 2016
commit fac5cd7c76fb9334c0ec32c0fa2d32294df19e68
6 changes: 3 additions & 3 deletions tmva/tmva/inc/TMVA/DNN/Architectures/Cpu/CpuMatrix.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ template<typename AFloat>
template<typename Function_t>
inline void TCpuMatrix<AFloat>::Map(Function_t &f)
{
AFloat __restrict__ *data = GetRawDataPointer();
AFloat *data = GetRawDataPointer();

auto fRange = [data, &f](const tbb::blocked_range<size_t> & range)
{
Expand All @@ -134,8 +134,8 @@ template<typename AFloat>
template<typename Function_t>
inline void TCpuMatrix<AFloat>::MapFrom(Function_t &f, const TCpuMatrix &A)
{
AFloat __restrict__ *dataB = GetRawDataPointer();
const AFloat __restrict__ *dataA = A.GetRawDataPointer();
AFloat *dataB = GetRawDataPointer();
const AFloat *dataA = A.GetRawDataPointer();

auto fRange = [&dataB, &dataA, &f](const tbb::blocked_range<size_t> & range)
{
Expand Down
4 changes: 2 additions & 2 deletions tmva/tmva/src/DNN/Architectures/Cpu/Arithmetic.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,8 +76,8 @@ template<typename Real_t>
void TCpu<Real_t>::Hadamard(TCpuMatrix<Real_t> &B,
const TCpuMatrix<Real_t> &A)
{
const Real_t __restrict__ *dataA = A.GetRawDataPointer();
Real_t __restrict__ *dataB = B.GetRawDataPointer();
const Real_t *dataA = A.GetRawDataPointer();
Real_t *dataB = B.GetRawDataPointer();

auto f = [&dataA, &dataB](const tbb::blocked_range<size_t> & range)
{
Expand Down
2 changes: 1 addition & 1 deletion tmva/tmva/src/DNN/Architectures/Cpu/Dropout.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ template<typename AFloat>
void TCpu<AFloat>::Dropout(TCpuMatrix<AFloat> &A,
AFloat dropoutProbability)
{
AFloat __restrict__ *data = A.GetRawDataPointer();
AFloat *data = A.GetRawDataPointer();

auto fRange = [&data, dropoutProbability](const tbb::blocked_range<size_t> & range)
{
Expand Down
20 changes: 10 additions & 10 deletions tmva/tmva/src/DNN/Architectures/Cpu/LossFunctions.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,8 @@ template<typename AFloat>
AFloat TCpu<AFloat>::MeanSquaredError(const TCpuMatrix<AFloat> &Y,
const TCpuMatrix<AFloat> &output)
{
const AFloat __restrict__ *dataY = Y.GetRawDataPointer();
const AFloat __restrict__ *dataOutput = output.GetRawDataPointer();
const AFloat *dataY = Y.GetRawDataPointer();
const AFloat *dataOutput = output.GetRawDataPointer();

auto f = [&dataY, &dataOutput](const tbb::blocked_range<size_t> & range,
AFloat partialSum)
Expand Down Expand Up @@ -63,9 +63,9 @@ void TCpu<AFloat>::MeanSquaredErrorGradients(
const TCpuMatrix<AFloat> & output)
{

AFloat __restrict__ *dataDY = dY.GetRawDataPointer();
const AFloat __restrict__ *dataY = Y.GetRawDataPointer();
const AFloat __restrict__ *dataOutput = output.GetRawDataPointer();
AFloat *dataDY = dY.GetRawDataPointer();
const AFloat *dataY = Y.GetRawDataPointer();
const AFloat *dataOutput = output.GetRawDataPointer();
AFloat norm = 1.0 / ((AFloat) Y.GetNrows() * Y.GetNcols());

auto f = [&dataDY, &dataY, &dataOutput, norm](const tbb::blocked_range<size_t> & range)
Expand All @@ -87,8 +87,8 @@ template<typename AFloat>
AFloat TCpu<AFloat>::CrossEntropy(const TCpuMatrix<AFloat> &Y,
const TCpuMatrix<AFloat> &output)
{
const AFloat __restrict__ *dataY = Y.GetRawDataPointer();
const AFloat __restrict__ *dataOutput = output.GetRawDataPointer();
const AFloat *dataY = Y.GetRawDataPointer();
const AFloat *dataOutput = output.GetRawDataPointer();

auto f = [&dataY, &dataOutput](const tbb::blocked_range<size_t> & range,
AFloat partialSum)
Expand Down Expand Up @@ -122,9 +122,9 @@ void TCpu<AFloat>::CrossEntropyGradients(
const TCpuMatrix<AFloat> & Y,
const TCpuMatrix<AFloat> & output)
{
AFloat __restrict__ *dataDY = dY.GetRawDataPointer();
const AFloat __restrict__ *dataY = Y.GetRawDataPointer();
const AFloat __restrict__ *dataOutput = output.GetRawDataPointer();
AFloat *dataDY = dY.GetRawDataPointer();
const AFloat *dataY = Y.GetRawDataPointer();
const AFloat *dataOutput = output.GetRawDataPointer();
AFloat norm = 1.0 / ((AFloat) Y.GetNrows() * Y.GetNcols());

auto f = [&dataDY, &dataY, &dataOutput, norm](const tbb::blocked_range<size_t> & range)
Expand Down
12 changes: 6 additions & 6 deletions tmva/tmva/src/DNN/Architectures/Cpu/Regularization.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ namespace DNN
template<typename AFloat>
AFloat TCpu<AFloat>::L1Regularization(const TCpuMatrix<AFloat> &Weights)
{
const AFloat __restrict__ *data = Weights.GetRawDataPointer();
const AFloat *data = Weights.GetRawDataPointer();

auto f = [&data](const tbb::blocked_range<size_t> & range,
AFloat partialSum)
Expand Down Expand Up @@ -58,8 +58,8 @@ void TCpu<AFloat>::AddL1RegularizationGradients(
AFloat weightDecay)
{

AFloat __restrict__ *dataB = B.GetRawDataPointer();
const AFloat __restrict__ *dataA = A.GetRawDataPointer();
AFloat *dataB = B.GetRawDataPointer();
const AFloat *dataA = A.GetRawDataPointer();

auto f = [&dataA, &dataB, weightDecay](const tbb::blocked_range<size_t> & range)
{
Expand All @@ -80,7 +80,7 @@ void TCpu<AFloat>::AddL1RegularizationGradients(
template<typename AFloat>
AFloat TCpu<AFloat>::L2Regularization(const TCpuMatrix<AFloat> &Weights)
{
const AFloat __restrict__ *data = Weights.GetRawDataPointer();
const AFloat *data = Weights.GetRawDataPointer();

auto f = [&data](const tbb::blocked_range<size_t> & range,
AFloat partialSum)
Expand Down Expand Up @@ -112,8 +112,8 @@ void TCpu<AFloat>::AddL2RegularizationGradients(
AFloat weightDecay)
{

AFloat __restrict__ *dataB = B.GetRawDataPointer();
const AFloat __restrict__ *dataA = A.GetRawDataPointer();
AFloat *dataB = B.GetRawDataPointer();
const AFloat *dataA = A.GetRawDataPointer();

auto f = [&dataA, &dataB, weightDecay](const tbb::blocked_range<size_t> & range)
{
Expand Down