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

Fixing LLVM error on TensorMorphingSycl.h on GPU; fixing int64_t crash for tensor_broadcast_sycl on GPU; adding get_sycl_supported_devices() on syclDevice.h.

Browse Source
This commit is contained in:
Mehdi Goli
2016-11-25 16:19:07 +00:00
parent b8cc5635d5
commit 7318daf887
18 changed files with 203 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
unsupported/test/cxx11_tensor_broadcast_sycl.cpp
View File

@@ -136,21 +136,14 @@ template<typename DataType> void sycl_broadcast_test_per_device(const cl::sycl::
test_broadcast_sycl<DataType, RowMajor, int>(sycl_device);
test_broadcast_sycl_fixed<DataType, ColMajor, int>(sycl_device);
test_broadcast_sycl<DataType, ColMajor, int>(sycl_device);
test_broadcast_sycl<DataType, RowMajor, int64_t>(sycl_device);
test_broadcast_sycl<DataType, ColMajor, int64_t>(sycl_device);
// the folowing two test breaks the intel gpu and amd gpu driver (cannot create opencl kernel)
// test_broadcast_sycl_fixed<DataType, RowMajor, int64_t>(sycl_device);
// test_broadcast_sycl_fixed<DataType, ColMajor, int64_t>(sycl_device);
test_broadcast_sycl_fixed<DataType, RowMajor, int64_t>(sycl_device);
test_broadcast_sycl_fixed<DataType, ColMajor, int64_t>(sycl_device);
}
void test_cxx11_tensor_broadcast_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos)
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_broadcast_test_per_device<float>(device));
}
}

15
unsupported/test/cxx11_tensor_builtins_sycl.cpp
View File

@@ -264,15 +264,10 @@ static void test_builtin_binary_sycl(const Eigen::SyclDevice &sycl_device) {
}
void test_cxx11_tensor_builtins_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos){
QueueInterface queueInterface(device);
Eigen::SyclDevice sycl_device(&queueInterface);
CALL_SUBTEST(test_builtin_unary_sycl(sycl_device));
CALL_SUBTEST(test_builtin_binary_sycl(sycl_device));
}
for (const auto& device :Eigen::get_sycl_supported_devices()) {
QueueInterface queueInterface(device);
Eigen::SyclDevice sycl_device(&queueInterface);
CALL_SUBTEST(test_builtin_unary_sycl(sycl_device));
CALL_SUBTEST(test_builtin_binary_sycl(sycl_device));
}
}

8
unsupported/test/cxx11_tensor_device_sycl.cpp
View File

@@ -71,11 +71,7 @@ template<typename DataType> void sycl_device_test_per_device(const cl::sycl::dev
}
void test_cxx11_tensor_device_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos)
CALL_SUBTEST(sycl_device_test_per_device<float>(device));
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_device_test_per_device<float>(device));
}
}

8
unsupported/test/cxx11_tensor_forced_eval_sycl.cpp
View File

@@ -70,11 +70,7 @@ template <typename DataType, typename Dev_selector> void tensorForced_evalperDev
test_forced_eval_sycl<DataType, ColMajor>(sycl_device);
}
void test_cxx11_tensor_forced_eval_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos)
CALL_SUBTEST(tensorForced_evalperDevice<float>(device));
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(tensorForced_evalperDevice<float>(device));
}
}

9
unsupported/test/cxx11_tensor_morphing_sycl.cpp
View File

@@ -82,12 +82,7 @@ template<typename DataType, typename dev_Selector> void sycl_slicing_test_per_de
}
void test_cxx11_tensor_morphing_sycl()
{
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
/// Currentlly it only works on cpu. Adding GPU cause LLVM ERROR in cunstructing OpenCL Kernel at runtime.
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(device.is_cpu() && s.find("amd")==std::string::npos)
CALL_SUBTEST(sycl_slicing_test_per_device<float>(device));
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_slicing_test_per_device<float>(device));
}
}

8
unsupported/test/cxx11_tensor_reduction_sycl.cpp
View File

@@ -141,11 +141,7 @@ template<typename DataType> void sycl_reduction_test_per_device(const cl::sycl::
test_last_dim_reductions_sycl<DataType, ColMajor>(sycl_device);
}
void test_cxx11_tensor_reduction_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos)
CALL_SUBTEST(sycl_reduction_test_per_device<float>(device));
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_reduction_test_per_device<float>(device));
}
}

9
unsupported/test/cxx11_tensor_sycl.cpp
View File

@@ -197,11 +197,8 @@ template<typename DataType, typename dev_Selector> void sycl_computing_test_per_
test_sycl_computations<DataType, ColMajor>(sycl_device);
}
void test_cxx11_tensor_sycl() {
for (const auto& device : cl::sycl::device::get_devices()) {
/// get_devices returns all the available opencl devices. Either use device_selector or exclude devices that computecpp does not support (AMD OpenCL for CPU )
auto s= device.template get_info<cl::sycl::info::device::vendor>();
std::transform(s.begin(), s.end(), s.begin(), ::tolower);
if(!device.is_cpu() || s.find("amd")==std::string::npos)
CALL_SUBTEST(sycl_computing_test_per_device<float>(device));
auto devices =Eigen::get_sycl_supported_devices();
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_computing_test_per_device<float>(device));
}
}