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

Adding tensor contraction operation backend for Sycl; adding test for contractionOp sycl backend; adding temporary solution to prevent memory leak in buffer; cleaning up cxx11_tensor_buildins_sycl.h

Browse Source
This commit is contained in:
Mehdi Goli
2016-12-14 15:30:37 +00:00
parent 3d59a47720
commit 2d4a091beb
15 changed files with 648 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
unsupported/test/cxx11_tensor_builtins_sycl.cpp
View File

@@ -154,7 +154,6 @@ template <typename T> T cwiseMin(T x, T y) { return std::min(x, y); }
Tensor<SCALAR, 3> out(tensorRange); \
in_1 = in_1.random() + static_cast<SCALAR>(0.01); \
in_2 = in_2.random() + static_cast<SCALAR>(0.01); \
out = out.random() + static_cast<SCALAR>(0.01); \
Tensor<SCALAR, 3> reference(out); \
SCALAR *gpu_data_1 = static_cast<SCALAR *>( \
sycl_device.allocate(in_1.size() * sizeof(SCALAR))); \
@@ -169,8 +168,6 @@ template <typename T> T cwiseMin(T x, T y) { return std::min(x, y); }
(in_1.size()) * sizeof(SCALAR)); \
sycl_device.memcpyHostToDevice(gpu_data_2, in_2.data(), \
(in_2.size()) * sizeof(SCALAR)); \
sycl_device.memcpyHostToDevice(gpu_data_out, out.data(), \
(out.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) = gpu_1.FUNC(gpu_2); \
sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \
(out.size()) * sizeof(SCALAR)); \
@@ -192,7 +189,6 @@ template <typename T> T cwiseMin(T x, T y) { return std::min(x, y); }
Tensor<SCALAR, 3> out(tensorRange); \
in_1 = in_1.random() + static_cast<SCALAR>(0.01); \
in_2 = in_2.random() + static_cast<SCALAR>(0.01); \
out = out.random() + static_cast<SCALAR>(0.01); \
Tensor<SCALAR, 3> reference(out); \
SCALAR *gpu_data_1 = static_cast<SCALAR *>( \
sycl_device.allocate(in_1.size() * sizeof(SCALAR))); \
@@ -207,8 +203,6 @@ template <typename T> T cwiseMin(T x, T y) { return std::min(x, y); }
(in_1.size()) * sizeof(SCALAR)); \
sycl_device.memcpyHostToDevice(gpu_data_2, in_2.data(), \
(in_2.size()) * sizeof(SCALAR)); \
sycl_device.memcpyHostToDevice(gpu_data_out, out.data(), \
(out.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) = gpu_1 OPERATOR gpu_2; \
sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \
(out.size()) * sizeof(SCALAR)); \
@@ -235,8 +229,6 @@ template <typename T> T cwiseMin(T x, T y) { return std::min(x, y); }
TensorMap<Tensor<SCALAR, 3>> gpu_out(gpu_data_out, tensorRange); \
sycl_device.memcpyHostToDevice(gpu_data_1, in_1.data(), \
(in_1.size()) * sizeof(SCALAR)); \
sycl_device.memcpyHostToDevice(gpu_data_out, out.data(), \
(out.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) = gpu_1 OPERATOR 2; \
sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \
(out.size()) * sizeof(SCALAR)); \