devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Introduce gpu_assert for assertion in device-code, and disable them with clang-cuda.

Browse Source
This commit is contained in:
Gael Guennebaud
2018-07-13 16:04:27 +02:00
parent 5fd03ddbfb
commit 06eb24cf4d
6 changed files with 44 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
unsupported/Eigen/CXX11/src/Tensor/TensorConvolutionSycl.h
View File

@@ -352,7 +352,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
size_t range_x, GRange_x, tileSize_x, range_y, GRange_y, tileSize_y;
m_device.parallel_for_setup(numX, numP, tileSize_x,tileSize_y,range_x,range_y, GRange_x, GRange_y );
const size_t shared_mem =(tileSize_x +kernel_size -1)*(tileSize_y);
assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
gpu_assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
auto global_range=cl::sycl::range<2>(GRange_x, GRange_y); // global range
auto local_range=cl::sycl::range<2>(tileSize_x, tileSize_y); // local range
InputLocalAcc local_acc(cl::sycl::range<1>(shared_mem), cgh);
@@ -377,7 +377,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
size_t range_x, GRange_x, tileSize_x, range_y, GRange_y, tileSize_y, range_z, GRange_z, tileSize_z;
m_device.parallel_for_setup(numX, numY, numP, tileSize_x, tileSize_y, tileSize_z, range_x, range_y, range_z, GRange_x, GRange_y, GRange_z );
const size_t shared_mem =(tileSize_x +kernel_size_x -1)*(tileSize_y +kernel_size_y -1) * tileSize_z;
assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
gpu_assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
auto global_range=cl::sycl::range<3>(GRange_x, GRange_y, GRange_z); // global range
auto local_range=cl::sycl::range<3>(tileSize_x, tileSize_y, tileSize_z); // local range
InputLocalAcc local_acc(cl::sycl::range<1>(shared_mem), cgh);
@@ -408,7 +408,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
size_t range_x, GRange_x, tileSize_x, range_y, GRange_y, tileSize_y, range_z, GRange_z, tileSize_z;
m_device.parallel_for_setup(numX, numY, numZ, tileSize_x, tileSize_y, tileSize_z, range_x, range_y, range_z, GRange_x, GRange_y, GRange_z );
const size_t shared_mem =(tileSize_x +kernel_size_x -1)*(tileSize_y +kernel_size_y -1) * (tileSize_z +kernel_size_y -1);
assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
gpu_assert(static_cast<unsigned long>(shared_mem) <= m_device.sharedMemPerBlock());
auto global_range=cl::sycl::range<3>(GRange_x, GRange_y, GRange_z); // global range
auto local_range=cl::sycl::range<3>(tileSize_x, tileSize_y, tileSize_z); // local range
InputLocalAcc local_acc(cl::sycl::range<1>(shared_mem), cgh);