Squashed feature branch:

* Added CUDA support * Performance improvements * API improvements * Added "depth" parameter to GEMM * Android cross-compilation
2015-04-29 15:50:57 -04:00
parent 5ff16bfcb6
commit cf5028d55b
3819 changed files with 7080 additions and 2916 deletions
--- a/lib/driver/device.cpp
+++ b/lib/driver/device.cpp
@@ -0,0 +1,190 @@
+#include "isaac/driver/device.h"
+#include <algorithm>
+
+namespace isaac
+{
+
+namespace driver
+{
+
+
+#ifdef ISAAC_WITH_CUDA
+template<CUdevice_attribute attr>
+int Device::cuGetInfo() const
+{
+  int res;
+  cuda::check(cuDeviceGetAttribute(&res, attr, *h_.cu));
+  return res;
+}
+
+Device::Device(int ordinal): backend_(CUDA), h_(backend_)
+{ cuda::check(cuDeviceGet(h_.cu.get(), ordinal)); }
+
+#endif
+
+
+Device::Device(cl::Device const & device) : backend_(OPENCL), h_(backend_)
+{ *h_.cl = device; }
+
+backend_type Device::backend() const
+{ return backend_; }
+
+unsigned int Device::address_bits() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA: return sizeof(long long)*8;
+#endif
+    case OPENCL: return h_.cl->getInfo<CL_DEVICE_ADDRESS_BITS>();
+    default: throw;
+  }
+
+  return backend_;
+}
+
+driver::Platform Device::platform() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA: return Platform(CUDA);
+#endif
+    case OPENCL: return Platform(h_.cl->getInfo<CL_DEVICE_PLATFORM>());
+    default: throw;
+  }
+}
+
+std::string Device::name() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA:
+      char tmp[128];
+      cuda::check(cuDeviceGetName(tmp, 128, *h_.cu));
+      return std::string(tmp);
+#endif
+    case OPENCL: return h_.cl->getInfo<CL_DEVICE_NAME>();
+    default: throw;
+  }
+}
+
+std::string Device::vendor_str() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA: return "NVidia";
+#endif
+    case OPENCL: return h_.cl->getInfo<CL_DEVICE_VENDOR>();
+    default: throw;
+  }
+}
+
+Device::VENDOR Device::vendor() const
+{
+    std::string vname = vendor_str();
+    std::transform(vname.begin(), vname.end(), vname.begin(), ::tolower);
+    if(vname.find("nvidia")!=std::string::npos)
+        return NVIDIA;
+    else if(vname.find("intel")!=std::string::npos)
+        return INTEL;
+    else if(vname.find("amd")!=std::string::npos || vname.find("advanced micro devices")!=std::string::npos)
+        return AMD;
+    else
+        return UNKNOWN;
+}
+
+std::vector<size_t> Device::max_work_item_sizes() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA:
+    {
+      std::vector<size_t> result(3);
+      result[0] = cuGetInfo<CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X>();
+      result[1] = cuGetInfo<CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y>();
+      result[2] = cuGetInfo<CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z>();
+      return result;
+    }
+#endif
+    case OPENCL:
+      return h_.cl->getInfo<CL_DEVICE_MAX_WORK_ITEM_SIZES>();
+    default:
+      throw;
+  }
+}
+
+device_type Device::type() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA: return DEVICE_TYPE_GPU;
+#endif
+    case OPENCL: return static_cast<device_type>(h_.cl->getInfo<CL_DEVICE_TYPE>());
+    default: throw;
+  }
+}
+
+std::string Device::extensions() const
+{
+  switch(backend_)
+  {
+#ifdef ISAAC_WITH_CUDA
+    case CUDA: return "";
+#endif
+    case OPENCL: return h_.cl->getInfo<CL_DEVICE_EXTENSIONS>();
+    default: throw;
+  }
+}
+
+#ifdef ISAAC_WITH_CUDA
+    #define CUDACASE(CUNAME) case CUDA: return cuGetInfo<CUNAME>();
+#else
+    #define CUDACASE(CUNAME)
+#endif\
+
+#define WRAP_ATTRIBUTE(ret, fname, CUNAME, CLNAME) \
+  ret Device::fname() const\
+  {\
+    switch(backend_)\
+    {\
+      CUDACASE(CUNAME)\
+      case OPENCL: return h_.cl->getInfo<CLNAME>();\
+      default: throw;\
+    }\
+  }\
+
+
+WRAP_ATTRIBUTE(size_t, max_work_group_size, CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK, CL_DEVICE_MAX_WORK_GROUP_SIZE)
+WRAP_ATTRIBUTE(size_t, local_mem_size, CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK, CL_DEVICE_LOCAL_MEM_SIZE)
+WRAP_ATTRIBUTE(size_t, warp_wavefront_size, CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK, CL_DEVICE_WAVEFRONT_WIDTH_AMD)
+WRAP_ATTRIBUTE(size_t, clock_rate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, CL_DEVICE_MAX_CLOCK_FREQUENCY)
+
+
+std::pair<unsigned int, unsigned int> Device::nv_compute_capability() const
+{
+  switch(backend_)
+  {
+      case OPENCL:
+          return std::pair<unsigned int, unsigned int>( h_.cl->getInfo<CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV>(), h_.cl->getInfo<CL_DEVICE_COMPUTE_CAPABILITY_MINOR_NV> ());
+#ifdef ISAAC_WITH_CUDA
+      case CUDA:
+          return std::pair<unsigned int, unsigned int>(cuGetInfo<CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR>(), cuGetInfo<CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR>());
+#endif
+      default:
+          throw;
+  }
+
+
+}
+
+
+
+}
+
+}
+