summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--Modules/CMakeCUDAInformation.cmake6
-rw-r--r--Modules/CMakeDetermineCUDACompiler.cmake7
-rw-r--r--Modules/Compiler/NVIDIA-CUDA.cmake2
-rw-r--r--Modules/Platform/Windows-NVIDIA-CUDA.cmake2
-rw-r--r--Source/CMakeVersion.cmake2
-rw-r--r--Source/cmVisualStudioGeneratorOptions.cxx10
-rw-r--r--Source/kwsys/SystemInformation.cxx246
-rw-r--r--Source/kwsys/SystemInformation.hxx.in2
-rw-r--r--Tests/Cuda/CMakeLists.txt1
-rw-r--r--Tests/Cuda/WithC/CMakeLists.txt12
-rw-r--r--Tests/Cuda/WithC/cuda.cu16
-rw-r--r--Tests/Cuda/WithC/main.c14
12 files changed, 124 insertions, 196 deletions
diff --git a/Modules/CMakeCUDAInformation.cmake b/Modules/CMakeCUDAInformation.cmake
index 13b1789..1c48159 100644
--- a/Modules/CMakeCUDAInformation.cmake
+++ b/Modules/CMakeCUDAInformation.cmake
@@ -93,6 +93,12 @@ if(NOT CMAKE_NOT_USING_CONFIG_FLAGS)
endif()
+if(CMAKE_CUDA_STANDARD_LIBRARIES_INIT)
+ set(CMAKE_CUDA_STANDARD_LIBRARIES "${CMAKE_CUDA_STANDARD_LIBRARIES_INIT}"
+ CACHE STRING "Libraries linked by default with all CUDA applications.")
+ mark_as_advanced(CMAKE_CUDA_STANDARD_LIBRARIES)
+endif()
+
include(CMakeCommonLanguageInclude)
# now define the following rules:
diff --git a/Modules/CMakeDetermineCUDACompiler.cmake b/Modules/CMakeDetermineCUDACompiler.cmake
index 7b6d17b..375e230 100644
--- a/Modules/CMakeDetermineCUDACompiler.cmake
+++ b/Modules/CMakeDetermineCUDACompiler.cmake
@@ -75,13 +75,6 @@ if(MSVC_CUDA_ARCHITECTURE_ID)
"set(MSVC_CUDA_ARCHITECTURE_ID ${MSVC_CUDA_ARCHITECTURE_ID})")
endif()
-#if this compiler vendor is matches NVIDIA we can determine
-#what the host compiler is. This only needs to be done if the CMAKE_CUDA_HOST_COMPILER
-#has NOT been explicitly set
-#
-#Find the line from compiler ID that contains a.out ( or last line )
-#We also need to find the implicit link lines. Which can be done by replacing
-#the compiler with cuda-fake-ld and pass too CMAKE_PARSE_IMPLICIT_LINK_INFO
if(CMAKE_CUDA_COMPILER_ID STREQUAL NVIDIA)
set(_nvcc_log "")
string(REPLACE "\r" "" _nvcc_output_orig "${CMAKE_CUDA_COMPILER_PRODUCED_OUTPUT}")
diff --git a/Modules/Compiler/NVIDIA-CUDA.cmake b/Modules/Compiler/NVIDIA-CUDA.cmake
index ae35132..7903313 100644
--- a/Modules/Compiler/NVIDIA-CUDA.cmake
+++ b/Modules/Compiler/NVIDIA-CUDA.cmake
@@ -13,7 +13,7 @@ set(CMAKE_INCLUDE_SYSTEM_FLAG_CUDA -isystem=)
string(APPEND CMAKE_CUDA_FLAGS_INIT " ")
string(APPEND CMAKE_CUDA_FLAGS_DEBUG_INIT " -g")
-string(APPEND CMAKE_CUDA_FLAGS_MINSIZEREL_INIT " -Os -DNDEBUG")
+string(APPEND CMAKE_CUDA_FLAGS_MINSIZEREL_INIT " -O1 -DNDEBUG")
string(APPEND CMAKE_CUDA_FLAGS_RELEASE_INIT " -O3 -DNDEBUG")
string(APPEND CMAKE_CUDA_FLAGS_RELWITHDEBINFO_INIT " -O2 -g -DNDEBUG")
diff --git a/Modules/Platform/Windows-NVIDIA-CUDA.cmake b/Modules/Platform/Windows-NVIDIA-CUDA.cmake
index 809ee06..eda41e0 100644
--- a/Modules/Platform/Windows-NVIDIA-CUDA.cmake
+++ b/Modules/Platform/Windows-NVIDIA-CUDA.cmake
@@ -40,3 +40,5 @@ string(APPEND CMAKE_CUDA_FLAGS_DEBUG_INIT " -Xcompiler=-MDd,-Zi,-RTC1")
string(APPEND CMAKE_CUDA_FLAGS_RELEASE_INIT " -Xcompiler=-MD")
string(APPEND CMAKE_CUDA_FLAGS_RELWITHDEBINFO_INIT " -Xcompiler=-MD")
string(APPEND CMAKE_CUDA_FLAGS_MINSIZEREL_INIT " -Xcompiler=-MD")
+
+set(CMAKE_CUDA_STANDARD_LIBRARIES_INIT "${CMAKE_C_STANDARD_LIBRARIES_INIT}")
diff --git a/Source/CMakeVersion.cmake b/Source/CMakeVersion.cmake
index 8b1882f..2efc5bf 100644
--- a/Source/CMakeVersion.cmake
+++ b/Source/CMakeVersion.cmake
@@ -1,5 +1,5 @@
# CMake version number components.
set(CMake_VERSION_MAJOR 3)
set(CMake_VERSION_MINOR 8)
-set(CMake_VERSION_PATCH 20170214)
+set(CMake_VERSION_PATCH 20170215)
#set(CMake_VERSION_RC 1)
diff --git a/Source/cmVisualStudioGeneratorOptions.cxx b/Source/cmVisualStudioGeneratorOptions.cxx
index c0913e6..6bacfa1 100644
--- a/Source/cmVisualStudioGeneratorOptions.cxx
+++ b/Source/cmVisualStudioGeneratorOptions.cxx
@@ -252,19 +252,19 @@ void cmVisualStudioGeneratorOptions::OutputPreprocessorDefinitions(
if (this->Defines.empty()) {
return;
}
+ const char* tag = "PreprocessorDefinitions";
if (this->Version >= cmGlobalVisualStudioGenerator::VS10) {
// if there are configuration specific flags, then
// use the configuration specific tag for PreprocessorDefinitions
if (!this->Configuration.empty()) {
fout << prefix;
this->TargetGenerator->WritePlatformConfigTag(
- "PreprocessorDefinitions", this->Configuration.c_str(), 0, 0, 0,
- &fout);
+ tag, this->Configuration.c_str(), 0, 0, 0, &fout);
} else {
- fout << prefix << "<PreprocessorDefinitions>";
+ fout << prefix << "<" << tag << ">";
}
} else {
- fout << prefix << "PreprocessorDefinitions=\"";
+ fout << prefix << tag << "=\"";
}
const char* sep = "";
for (std::vector<std::string>::const_iterator di = this->Defines.begin();
@@ -291,7 +291,7 @@ void cmVisualStudioGeneratorOptions::OutputPreprocessorDefinitions(
sep = ";";
}
if (this->Version >= cmGlobalVisualStudioGenerator::VS10) {
- fout << ";%(PreprocessorDefinitions)</PreprocessorDefinitions>" << suffix;
+ fout << ";%(" << tag << ")</" << tag << ">" << suffix;
} else {
fout << "\"" << suffix;
}
diff --git a/Source/kwsys/SystemInformation.cxx b/Source/kwsys/SystemInformation.cxx
index e01dcd7..86f7552 100644
--- a/Source/kwsys/SystemInformation.cxx
+++ b/Source/kwsys/SystemInformation.cxx
@@ -35,8 +35,13 @@
#include "SystemInformation.hxx.in"
#endif
+#include <algorithm>
+#include <bitset>
+#include <cassert>
#include <fstream>
#include <iostream>
+#include <limits>
+#include <set>
#include <sstream>
#include <string>
#include <vector>
@@ -394,7 +399,6 @@ public:
bool SupportsMP;
bool HasMMXPlus;
bool HasSSEMMX;
- bool SupportsHyperthreading;
unsigned int LogicalProcessorsPerPhysical;
int APIC_ID;
CPUPowerManagement PowerManagement;
@@ -463,10 +467,9 @@ protected:
unsigned int NumberOfLogicalCPU;
unsigned int NumberOfPhysicalCPU;
- int CPUCount(); // For windows
- unsigned char LogicalCPUPerPhysicalCPU();
+ void CPUCountWindows(); // For windows
unsigned char GetAPICId(); // For windows
- bool IsHyperThreadingSupported();
+ bool IsSMTSupported();
static LongLong GetCyclesDifference(DELAY_FUNC, unsigned int); // For windows
// For Linux and Cygwin, /proc/cpuinfo formats are slightly different
@@ -1542,7 +1545,7 @@ void SystemInformationImplementation::RunCPUCheck()
RetrieveProcessorSerialNumber();
}
- this->CPUCount();
+ this->CPUCountWindows();
#elif defined(__APPLE__)
this->ParseSysCtl();
@@ -2090,16 +2093,10 @@ bool SystemInformationImplementation::RetrieveCPUFeatures()
// Retrieve Intel specific extended features.
if (this->ChipManufacturer == Intel) {
- this->Features.ExtendedFeatures.SupportsHyperthreading =
- ((cpuinfo[3] & 0x10000000) !=
- 0); // Intel specific: Hyperthreading --> Bit 28
- this->Features.ExtendedFeatures.LogicalProcessorsPerPhysical =
- (this->Features.ExtendedFeatures.SupportsHyperthreading)
- ? ((cpuinfo[1] & 0x00FF0000) >> 16)
- : 1;
-
- if ((this->Features.ExtendedFeatures.SupportsHyperthreading) &&
- (this->Features.HasAPIC)) {
+ bool SupportsSMT =
+ ((cpuinfo[3] & 0x10000000) != 0); // Intel specific: SMT --> Bit 28
+
+ if ((SupportsSMT) && (this->Features.HasAPIC)) {
// Retrieve APIC information if there is one present.
this->Features.ExtendedFeatures.APIC_ID =
((cpuinfo[1] & 0xFF000000) >> 24);
@@ -3401,7 +3398,7 @@ bool SystemInformationImplementation::RetreiveInformationFromCpuInfoFile()
fclose(fd);
buffer.resize(fileSize - 2);
// Number of logical CPUs (combination of multiple processors, multi-core
- // and hyperthreading)
+ // and SMT)
size_t pos = buffer.find("processor\t");
while (pos != buffer.npos) {
this->NumberOfLogicalCPU++;
@@ -3409,30 +3406,25 @@ bool SystemInformationImplementation::RetreiveInformationFromCpuInfoFile()
}
#ifdef __linux
- // Find the largest physical id.
- int maxId = -1;
+ // Count sockets.
+ std::set<int> PhysicalIDs;
std::string idc = this->ExtractValueFromCpuInfoFile(buffer, "physical id");
while (this->CurrentPositionInFile != buffer.npos) {
int id = atoi(idc.c_str());
- if (id > maxId) {
- maxId = id;
- }
+ PhysicalIDs.insert(id);
idc = this->ExtractValueFromCpuInfoFile(buffer, "physical id",
this->CurrentPositionInFile + 1);
}
+ uint64_t NumberOfSockets = PhysicalIDs.size();
+ NumberOfSockets = std::max(NumberOfSockets, (uint64_t)1);
// Physical ids returned by Linux don't distinguish cores.
// We want to record the total number of cores in this->NumberOfPhysicalCPU
// (checking only the first proc)
- std::string cores = this->ExtractValueFromCpuInfoFile(buffer, "cpu cores");
- int numberOfCoresPerCPU = atoi(cores.c_str());
- if (maxId > 0) {
- this->NumberOfPhysicalCPU =
- static_cast<unsigned int>(numberOfCoresPerCPU * (maxId + 1));
- } else {
- // Linux Sparc: get cpu count
- this->NumberOfPhysicalCPU =
- atoi(this->ExtractValueFromCpuInfoFile(buffer, "ncpus active").c_str());
- }
+ std::string Cores = this->ExtractValueFromCpuInfoFile(buffer, "cpu cores");
+ unsigned int NumberOfCoresPerSocket = (unsigned int)atoi(Cores.c_str());
+ NumberOfCoresPerSocket = std::max(NumberOfCoresPerSocket, 1u);
+ this->NumberOfPhysicalCPU =
+ NumberOfCoresPerSocket * (unsigned int)NumberOfSockets;
#else // __CYGWIN__
// does not have "physical id" entries, neither "cpu cores"
@@ -3447,7 +3439,7 @@ bool SystemInformationImplementation::RetreiveInformationFromCpuInfoFile()
if (this->NumberOfPhysicalCPU <= 0) {
this->NumberOfPhysicalCPU = 1;
}
- // LogicalProcessorsPerPhysical>1 => hyperthreading.
+ // LogicalProcessorsPerPhysical>1 => SMT.
this->Features.ExtendedFeatures.LogicalProcessorsPerPhysical =
this->NumberOfLogicalCPU / this->NumberOfPhysicalCPU;
@@ -4322,68 +4314,10 @@ void SystemInformationImplementation::DelayOverhead(unsigned int uiMS)
(void)uiMS;
}
-/** Return the number of logical CPU per physical CPUs Works only for windows
- */
-unsigned char SystemInformationImplementation::LogicalCPUPerPhysicalCPU(void)
-{
-#ifdef __APPLE__
- size_t len = 4;
- int cores_per_package = 0;
- int err = sysctlbyname("machdep.cpu.cores_per_package", &cores_per_package,
- &len, NULL, 0);
- if (err != 0) {
- return 1; // That name was not found, default to 1
- }
- return static_cast<unsigned char>(cores_per_package);
-#else
- int Regs[4] = { 0, 0, 0, 0 };
-#if USE_CPUID
- if (!this->IsHyperThreadingSupported()) {
- return static_cast<unsigned char>(1); // HT not supported
- }
- call_cpuid(1, Regs);
-#endif
- return static_cast<unsigned char>((Regs[1] & NUM_LOGICAL_BITS) >> 16);
-#endif
-}
-
/** Works only for windows */
-bool SystemInformationImplementation::IsHyperThreadingSupported()
+bool SystemInformationImplementation::IsSMTSupported()
{
- if (this->Features.ExtendedFeatures.SupportsHyperthreading) {
- return true;
- }
-
-#if USE_CPUID
- int Regs[4] = { 0, 0, 0, 0 }, VendorId[4] = { 0, 0, 0, 0 };
- // Get vendor id string
- if (!call_cpuid(0, VendorId)) {
- return false;
- }
- // eax contains family processor type
- // edx has info about the availability of hyper-Threading
- if (!call_cpuid(1, Regs)) {
- return false;
- }
-
- if (((Regs[0] & FAMILY_ID) == PENTIUM4_ID) || (Regs[0] & EXT_FAMILY_ID)) {
- if (VendorId[1] == 0x756e6547) // 'uneG'
- {
- if (VendorId[3] == 0x49656e69) // 'Ieni'
- {
- if (VendorId[2] == 0x6c65746e) // 'letn'
- {
- // Genuine Intel with hyper-Threading technology
- this->Features.ExtendedFeatures.SupportsHyperthreading =
- ((Regs[3] & HT_BIT) != 0);
- return this->Features.ExtendedFeatures.SupportsHyperthreading;
- }
- }
- }
- }
-#endif
-
- return 0; // Not genuine Intel processor
+ return this->Features.ExtendedFeatures.LogicalProcessorsPerPhysical > 1;
}
/** Return the APIC Id. Works only for windows. */
@@ -4392,7 +4326,7 @@ unsigned char SystemInformationImplementation::GetAPICId()
int Regs[4] = { 0, 0, 0, 0 };
#if USE_CPUID
- if (!this->IsHyperThreadingSupported()) {
+ if (!this->IsSMTSupported()) {
return static_cast<unsigned char>(-1); // HT not supported
} // Logical processor = 1
call_cpuid(1, Regs);
@@ -4402,102 +4336,46 @@ unsigned char SystemInformationImplementation::GetAPICId()
}
/** Count the number of CPUs. Works only on windows. */
-int SystemInformationImplementation::CPUCount()
+void SystemInformationImplementation::CPUCountWindows()
{
#if defined(_WIN32)
- unsigned char StatusFlag = 0;
- SYSTEM_INFO info;
-
+ std::vector<SYSTEM_LOGICAL_PROCESSOR_INFORMATION> ProcInfo;
this->NumberOfPhysicalCPU = 0;
this->NumberOfLogicalCPU = 0;
- info.dwNumberOfProcessors = 0;
- GetSystemInfo(&info);
-
- // Number of physical processors in a non-Intel system
- // or in a 32-bit Intel system with Hyper-Threading technology disabled
- this->NumberOfPhysicalCPU = (unsigned char)info.dwNumberOfProcessors;
-
- if (this->IsHyperThreadingSupported()) {
- unsigned char HT_Enabled = 0;
- this->NumberOfLogicalCPU = this->LogicalCPUPerPhysicalCPU();
- if (this->NumberOfLogicalCPU >=
- 1) // >1 Doesn't mean HT is enabled in the BIOS
- {
- HANDLE hCurrentProcessHandle;
-#ifndef _WIN64
-#define DWORD_PTR DWORD
-#endif
- DWORD_PTR dwProcessAffinity;
- DWORD_PTR dwSystemAffinity;
- DWORD dwAffinityMask;
-
- // Calculate the appropriate shifts and mask based on the
- // number of logical processors.
- unsigned int i = 1;
- unsigned char PHY_ID_MASK = 0xFF;
- // unsigned char PHY_ID_SHIFT = 0;
-
- while (i < this->NumberOfLogicalCPU) {
- i *= 2;
- PHY_ID_MASK <<= 1;
- // PHY_ID_SHIFT++;
- }
- hCurrentProcessHandle = GetCurrentProcess();
- GetProcessAffinityMask(hCurrentProcessHandle, &dwProcessAffinity,
- &dwSystemAffinity);
-
- // Check if available process affinity mask is equal to the
- // available system affinity mask
- if (dwProcessAffinity != dwSystemAffinity) {
- StatusFlag = HT_CANNOT_DETECT;
- this->NumberOfPhysicalCPU = (unsigned char)-1;
- return StatusFlag;
- }
-
- dwAffinityMask = 1;
- while (dwAffinityMask != 0 && dwAffinityMask <= dwProcessAffinity) {
- // Check if this CPU is available
- if (dwAffinityMask & dwProcessAffinity) {
- if (SetProcessAffinityMask(hCurrentProcessHandle, dwAffinityMask)) {
- unsigned char APIC_ID, LOG_ID;
- Sleep(0); // Give OS time to switch CPU
-
- APIC_ID = GetAPICId();
- LOG_ID = APIC_ID & ~PHY_ID_MASK;
-
- if (LOG_ID != 0) {
- HT_Enabled = 1;
- }
- }
- }
- dwAffinityMask = dwAffinityMask << 1;
- }
- // Reset the processor affinity
- SetProcessAffinityMask(hCurrentProcessHandle, dwProcessAffinity);
+ {
+ DWORD Length = 0;
+ DWORD rc = GetLogicalProcessorInformation(NULL, &Length);
+ assert(FALSE == rc);
+ (void)rc; // Silence unused variable warning in Borland C++ 5.81
+ assert(GetLastError() == ERROR_INSUFFICIENT_BUFFER);
+ ProcInfo.resize(Length / sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION));
+ rc = GetLogicalProcessorInformation(&ProcInfo[0], &Length);
+ assert(rc != FALSE);
+ (void)rc; // Silence unused variable warning in Borland C++ 5.81
+ }
+
+ typedef std::vector<SYSTEM_LOGICAL_PROCESSOR_INFORMATION>::iterator
+ pinfoIt_t;
+ for (pinfoIt_t it = ProcInfo.begin(); it != ProcInfo.end(); ++it) {
+ SYSTEM_LOGICAL_PROCESSOR_INFORMATION PInfo = *it;
+ if (PInfo.Relationship != RelationProcessorCore) {
+ continue;
+ }
- if (this->NumberOfLogicalCPU ==
- 1) // Normal P4 : HT is disabled in hardware
- {
- StatusFlag = HT_DISABLED;
- } else {
- if (HT_Enabled) {
- // Total physical processors in a Hyper-Threading enabled system.
- this->NumberOfPhysicalCPU /= (this->NumberOfLogicalCPU);
- StatusFlag = HT_ENABLED;
- } else {
- StatusFlag = HT_SUPPORTED_NOT_ENABLED;
- }
- }
+ std::bitset<std::numeric_limits<ULONG_PTR>::digits> ProcMask(
+ (unsigned long long)PInfo.ProcessorMask);
+ unsigned int count = (unsigned int)ProcMask.count();
+ if (count == 0) { // I think this should never happen, but just to be safe.
+ continue;
}
- } else {
- // Processors do not have Hyper-Threading technology
- StatusFlag = HT_NOT_CAPABLE;
- this->NumberOfLogicalCPU = 1;
+ this->NumberOfPhysicalCPU++;
+ this->NumberOfLogicalCPU += (unsigned int)count;
+ this->Features.ExtendedFeatures.LogicalProcessorsPerPhysical = count;
}
- return StatusFlag;
+ this->NumberOfPhysicalCPU = std::max(1u, this->NumberOfPhysicalCPU);
+ this->NumberOfLogicalCPU = std::max(1u, this->NumberOfLogicalCPU);
#else
- return 0;
#endif
}
@@ -4559,8 +4437,14 @@ bool SystemInformationImplementation::ParseSysCtl()
sysctlbyname("hw.physicalcpu", &this->NumberOfPhysicalCPU, &len, NULL, 0);
len = sizeof(this->NumberOfLogicalCPU);
sysctlbyname("hw.logicalcpu", &this->NumberOfLogicalCPU, &len, NULL, 0);
+
+ int cores_per_package = 0;
+ len = sizeof(cores_per_package);
+ err = sysctlbyname("machdep.cpu.cores_per_package", &cores_per_package, &len,
+ NULL, 0);
+ // That name was not found, default to 1
this->Features.ExtendedFeatures.LogicalProcessorsPerPhysical =
- this->LogicalCPUPerPhysicalCPU();
+ err != 0 ? 1 : static_cast<unsigned char>(cores_per_package);
len = sizeof(value);
sysctlbyname("hw.cpufrequency", &value, &len, NULL, 0);
diff --git a/Source/kwsys/SystemInformation.hxx.in b/Source/kwsys/SystemInformation.hxx.in
index 0fc1067..cc09393 100644
--- a/Source/kwsys/SystemInformation.hxx.in
+++ b/Source/kwsys/SystemInformation.hxx.in
@@ -67,7 +67,7 @@ public:
bool Is64Bits();
- unsigned int GetNumberOfLogicalCPU(); // per physical cpu
+ unsigned int GetNumberOfLogicalCPU();
unsigned int GetNumberOfPhysicalCPU();
bool DoesCPUSupportCPUID();
diff --git a/Tests/Cuda/CMakeLists.txt b/Tests/Cuda/CMakeLists.txt
index 42b00e1..de48501 100644
--- a/Tests/Cuda/CMakeLists.txt
+++ b/Tests/Cuda/CMakeLists.txt
@@ -4,3 +4,4 @@ ADD_TEST_MACRO(Cuda.ConsumeCompileFeatures CudaConsumeCompileFeatures)
ADD_TEST_MACRO(Cuda.ObjectLibrary CudaObjectLibrary)
ADD_TEST_MACRO(Cuda.ToolkitInclude CudaToolkitInclude)
ADD_TEST_MACRO(Cuda.ProperLinkFlags ProperLinkFlags)
+ADD_TEST_MACRO(Cuda.WithC CudaWithC)
diff --git a/Tests/Cuda/WithC/CMakeLists.txt b/Tests/Cuda/WithC/CMakeLists.txt
new file mode 100644
index 0000000..7596804
--- /dev/null
+++ b/Tests/Cuda/WithC/CMakeLists.txt
@@ -0,0 +1,12 @@
+cmake_minimum_required(VERSION 3.7)
+project(CudaComplex CUDA C)
+
+set(CMAKE_CUDA_FLAGS "-gencode arch=compute_30,code=compute_30")
+
+add_executable(CudaWithC main.c cuda.cu)
+
+if(APPLE)
+ # We need to add the default path to the driver (libcuda.dylib) as an rpath, so that
+ # the static cuda runtime can find it at runtime.
+ target_link_libraries(CudaWithC PRIVATE -Wl,-rpath,/usr/local/cuda/lib)
+endif()
diff --git a/Tests/Cuda/WithC/cuda.cu b/Tests/Cuda/WithC/cuda.cu
new file mode 100644
index 0000000..06bd7b9
--- /dev/null
+++ b/Tests/Cuda/WithC/cuda.cu
@@ -0,0 +1,16 @@
+#include <cuda.h>
+
+#include <iostream>
+
+extern "C" int use_cuda(void)
+{
+ int nDevices = 0;
+ cudaError_t err = cudaGetDeviceCount(&nDevices);
+ if (err != cudaSuccess) {
+ std::cerr << "Failed to retrieve the number of CUDA enabled devices"
+ << std::endl;
+ return 1;
+ }
+ std::cout << "Found " << nDevices << " CUDA enabled devices" << std::endl;
+ return 0;
+}
diff --git a/Tests/Cuda/WithC/main.c b/Tests/Cuda/WithC/main.c
new file mode 100644
index 0000000..cb5fddc
--- /dev/null
+++ b/Tests/Cuda/WithC/main.c
@@ -0,0 +1,14 @@
+extern int use_cuda(void);
+
+#ifdef _WIN32
+#include <windows.h>
+#endif
+
+int main()
+{
+#ifdef _WIN32
+ /* Use an API that requires CMake's "standard" C libraries. */
+ GetOpenFileName(NULL);
+#endif
+ return use_cuda();
+}
generated by cgit v0.12 at 2025-01-25 13:02:44 (GMT)