blender/intern/cycles/util/util_cuda.cpp
Brecht Van Lommel 16204bd647 Cycles: prepare to make CUDA 5.0 the official version we use
* Add CUDA compiler version detection to cmake/scons/runtime
* Remove noinline in kernel_shader.h and reenable --use_fast_math if CUDA 5.x
  is used, these were workarounds for CUDA 4.2 bugs
* Change max number of registers to 32 for sm 2.x (based on performance tests
  from Martijn Berger and confirmed here), and also for NVidia OpenCL.

Overall it seems that with these changes and the latest CUDA 5.0 download, that
performance is as good as or better than the 2.67b release with the scenes and
graphics cards I tested.
2013-06-19 17:54:23 +00:00

498 lines
15 KiB
C++

/*
* Copyright 2011, Blender Foundation.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*/
#include <iostream>
#include <stdlib.h>
#include <stdio.h>
#include "util_cuda.h"
#include "util_debug.h"
#include "util_dynlib.h"
#include "util_path.h"
#include "util_string.h"
#ifdef _WIN32
#define popen _popen
#define pclose _pclose
#endif
/* function defininitions */
tcuInit *cuInit;
tcuDriverGetVersion *cuDriverGetVersion;
tcuDeviceGet *cuDeviceGet;
tcuDeviceGetCount *cuDeviceGetCount;
tcuDeviceGetName *cuDeviceGetName;
tcuDeviceComputeCapability *cuDeviceComputeCapability;
tcuDeviceTotalMem *cuDeviceTotalMem;
tcuDeviceGetProperties *cuDeviceGetProperties;
tcuDeviceGetAttribute *cuDeviceGetAttribute;
tcuCtxCreate *cuCtxCreate;
tcuCtxDestroy *cuCtxDestroy;
tcuCtxAttach *cuCtxAttach;
tcuCtxDetach *cuCtxDetach;
tcuCtxPushCurrent *cuCtxPushCurrent;
tcuCtxPopCurrent *cuCtxPopCurrent;
tcuCtxGetDevice *cuCtxGetDevice;
tcuCtxSynchronize *cuCtxSynchronize;
tcuModuleLoad *cuModuleLoad;
tcuModuleLoadData *cuModuleLoadData;
tcuModuleLoadDataEx *cuModuleLoadDataEx;
tcuModuleLoadFatBinary *cuModuleLoadFatBinary;
tcuModuleUnload *cuModuleUnload;
tcuModuleGetFunction *cuModuleGetFunction;
tcuModuleGetGlobal *cuModuleGetGlobal;
tcuModuleGetTexRef *cuModuleGetTexRef;
tcuModuleGetSurfRef *cuModuleGetSurfRef;
tcuMemGetInfo *cuMemGetInfo;
tcuMemAlloc *cuMemAlloc;
tcuMemAllocPitch *cuMemAllocPitch;
tcuMemFree *cuMemFree;
tcuMemGetAddressRange *cuMemGetAddressRange;
tcuMemAllocHost *cuMemAllocHost;
tcuMemFreeHost *cuMemFreeHost;
tcuMemHostAlloc *cuMemHostAlloc;
tcuMemHostGetDevicePointer *cuMemHostGetDevicePointer;
tcuMemHostGetFlags *cuMemHostGetFlags;
tcuMemcpyHtoD *cuMemcpyHtoD;
tcuMemcpyDtoH *cuMemcpyDtoH;
tcuMemcpyDtoD *cuMemcpyDtoD;
tcuMemcpyDtoA *cuMemcpyDtoA;
tcuMemcpyAtoD *cuMemcpyAtoD;
tcuMemcpyHtoA *cuMemcpyHtoA;
tcuMemcpyAtoH *cuMemcpyAtoH;
tcuMemcpyAtoA *cuMemcpyAtoA;
tcuMemcpy2D *cuMemcpy2D;
tcuMemcpy2DUnaligned *cuMemcpy2DUnaligned;
tcuMemcpy3D *cuMemcpy3D;
tcuMemcpyHtoDAsync *cuMemcpyHtoDAsync;
tcuMemcpyDtoHAsync *cuMemcpyDtoHAsync;
tcuMemcpyDtoDAsync *cuMemcpyDtoDAsync;
tcuMemcpyHtoAAsync *cuMemcpyHtoAAsync;
tcuMemcpyAtoHAsync *cuMemcpyAtoHAsync;
tcuMemcpy2DAsync *cuMemcpy2DAsync;
tcuMemcpy3DAsync *cuMemcpy3DAsync;
tcuMemsetD8 *cuMemsetD8;
tcuMemsetD16 *cuMemsetD16;
tcuMemsetD32 *cuMemsetD32;
tcuMemsetD2D8 *cuMemsetD2D8;
tcuMemsetD2D16 *cuMemsetD2D16;
tcuMemsetD2D32 *cuMemsetD2D32;
tcuFuncSetBlockShape *cuFuncSetBlockShape;
tcuFuncSetSharedSize *cuFuncSetSharedSize;
tcuFuncGetAttribute *cuFuncGetAttribute;
tcuFuncSetCacheConfig *cuFuncSetCacheConfig;
tcuArrayCreate *cuArrayCreate;
tcuArrayGetDescriptor *cuArrayGetDescriptor;
tcuArrayDestroy *cuArrayDestroy;
tcuArray3DCreate *cuArray3DCreate;
tcuArray3DGetDescriptor *cuArray3DGetDescriptor;
tcuTexRefCreate *cuTexRefCreate;
tcuTexRefDestroy *cuTexRefDestroy;
tcuTexRefSetArray *cuTexRefSetArray;
tcuTexRefSetAddress *cuTexRefSetAddress;
tcuTexRefSetAddress2D *cuTexRefSetAddress2D;
tcuTexRefSetFormat *cuTexRefSetFormat;
tcuTexRefSetAddressMode *cuTexRefSetAddressMode;
tcuTexRefSetFilterMode *cuTexRefSetFilterMode;
tcuTexRefSetFlags *cuTexRefSetFlags;
tcuTexRefGetAddress *cuTexRefGetAddress;
tcuTexRefGetArray *cuTexRefGetArray;
tcuTexRefGetAddressMode *cuTexRefGetAddressMode;
tcuTexRefGetFilterMode *cuTexRefGetFilterMode;
tcuTexRefGetFormat *cuTexRefGetFormat;
tcuTexRefGetFlags *cuTexRefGetFlags;
tcuSurfRefSetArray *cuSurfRefSetArray;
tcuSurfRefGetArray *cuSurfRefGetArray;
tcuParamSetSize *cuParamSetSize;
tcuParamSeti *cuParamSeti;
tcuParamSetf *cuParamSetf;
tcuParamSetv *cuParamSetv;
tcuParamSetTexRef *cuParamSetTexRef;
tcuLaunch *cuLaunch;
tcuLaunchGrid *cuLaunchGrid;
tcuLaunchGridAsync *cuLaunchGridAsync;
tcuEventCreate *cuEventCreate;
tcuEventRecord *cuEventRecord;
tcuEventQuery *cuEventQuery;
tcuEventSynchronize *cuEventSynchronize;
tcuEventDestroy *cuEventDestroy;
tcuEventElapsedTime *cuEventElapsedTime;
tcuStreamCreate *cuStreamCreate;
tcuStreamQuery *cuStreamQuery;
tcuStreamSynchronize *cuStreamSynchronize;
tcuStreamDestroy *cuStreamDestroy;
tcuGraphicsUnregisterResource *cuGraphicsUnregisterResource;
tcuGraphicsSubResourceGetMappedArray *cuGraphicsSubResourceGetMappedArray;
tcuGraphicsResourceGetMappedPointer *cuGraphicsResourceGetMappedPointer;
tcuGraphicsResourceSetMapFlags *cuGraphicsResourceSetMapFlags;
tcuGraphicsMapResources *cuGraphicsMapResources;
tcuGraphicsUnmapResources *cuGraphicsUnmapResources;
tcuGetExportTable *cuGetExportTable;
tcuCtxSetLimit *cuCtxSetLimit;
tcuCtxGetLimit *cuCtxGetLimit;
tcuGLCtxCreate *cuGLCtxCreate;
tcuGraphicsGLRegisterBuffer *cuGraphicsGLRegisterBuffer;
tcuGraphicsGLRegisterImage *cuGraphicsGLRegisterImage;
tcuCtxSetCurrent *cuCtxSetCurrent;
CCL_NAMESPACE_BEGIN
/* utility macros */
#define CUDA_LIBRARY_FIND_CHECKED(name) \
name = (t##name*)dynamic_library_find(lib, #name);
#define CUDA_LIBRARY_FIND(name) \
name = (t##name*)dynamic_library_find(lib, #name); \
assert(name);
#define CUDA_LIBRARY_FIND_V2(name) \
name = (t##name*)dynamic_library_find(lib, #name "_v2"); \
assert(name);
/* initialization function */
bool cuLibraryInit()
{
static bool initialized = false;
static bool result = false;
if(initialized)
return result;
initialized = true;
/* library paths */
#ifdef _WIN32
/* expected in c:/windows/system or similar, no path needed */
const char *path = "nvcuda.dll";
#elif defined(__APPLE__)
/* default installation path */
const char *path = "/usr/local/cuda/lib/libcuda.dylib";
#else
const char *path = "libcuda.so";
#endif
/* load library */
DynamicLibrary *lib = dynamic_library_open(path);
if(lib == NULL)
return false;
/* detect driver version */
int driver_version = 1000;
CUDA_LIBRARY_FIND_CHECKED(cuDriverGetVersion);
if(cuDriverGetVersion)
cuDriverGetVersion(&driver_version);
/* we require version 4.0 */
if(driver_version < 4000)
return false;
/* fetch all function pointers */
CUDA_LIBRARY_FIND(cuInit);
CUDA_LIBRARY_FIND(cuDeviceGet);
CUDA_LIBRARY_FIND(cuDeviceGetCount);
CUDA_LIBRARY_FIND(cuDeviceGetName);
CUDA_LIBRARY_FIND(cuDeviceComputeCapability);
CUDA_LIBRARY_FIND(cuDeviceTotalMem);
CUDA_LIBRARY_FIND(cuDeviceGetProperties);
CUDA_LIBRARY_FIND(cuDeviceGetAttribute);
CUDA_LIBRARY_FIND(cuCtxCreate);
CUDA_LIBRARY_FIND(cuCtxDestroy);
CUDA_LIBRARY_FIND(cuCtxAttach);
CUDA_LIBRARY_FIND(cuCtxDetach);
CUDA_LIBRARY_FIND(cuCtxPushCurrent);
CUDA_LIBRARY_FIND(cuCtxPopCurrent);
CUDA_LIBRARY_FIND(cuCtxGetDevice);
CUDA_LIBRARY_FIND(cuCtxSynchronize);
CUDA_LIBRARY_FIND(cuModuleLoad);
CUDA_LIBRARY_FIND(cuModuleLoadData);
CUDA_LIBRARY_FIND(cuModuleUnload);
CUDA_LIBRARY_FIND(cuModuleGetFunction);
CUDA_LIBRARY_FIND(cuModuleGetGlobal);
CUDA_LIBRARY_FIND(cuModuleGetTexRef);
CUDA_LIBRARY_FIND(cuMemGetInfo);
CUDA_LIBRARY_FIND(cuMemAlloc);
CUDA_LIBRARY_FIND(cuMemAllocPitch);
CUDA_LIBRARY_FIND(cuMemFree);
CUDA_LIBRARY_FIND(cuMemGetAddressRange);
CUDA_LIBRARY_FIND(cuMemAllocHost);
CUDA_LIBRARY_FIND(cuMemFreeHost);
CUDA_LIBRARY_FIND(cuMemHostAlloc);
CUDA_LIBRARY_FIND(cuMemHostGetDevicePointer);
CUDA_LIBRARY_FIND(cuMemcpyHtoD);
CUDA_LIBRARY_FIND(cuMemcpyDtoH);
CUDA_LIBRARY_FIND(cuMemcpyDtoD);
CUDA_LIBRARY_FIND(cuMemcpyDtoA);
CUDA_LIBRARY_FIND(cuMemcpyAtoD);
CUDA_LIBRARY_FIND(cuMemcpyHtoA);
CUDA_LIBRARY_FIND(cuMemcpyAtoH);
CUDA_LIBRARY_FIND(cuMemcpyAtoA);
CUDA_LIBRARY_FIND(cuMemcpy2D);
CUDA_LIBRARY_FIND(cuMemcpy2DUnaligned);
CUDA_LIBRARY_FIND(cuMemcpy3D);
CUDA_LIBRARY_FIND(cuMemcpyHtoDAsync);
CUDA_LIBRARY_FIND(cuMemcpyDtoHAsync);
CUDA_LIBRARY_FIND(cuMemcpyHtoAAsync);
CUDA_LIBRARY_FIND(cuMemcpyAtoHAsync);
CUDA_LIBRARY_FIND(cuMemcpy2DAsync);
CUDA_LIBRARY_FIND(cuMemcpy3DAsync);
CUDA_LIBRARY_FIND(cuMemsetD8);
CUDA_LIBRARY_FIND(cuMemsetD16);
CUDA_LIBRARY_FIND(cuMemsetD32);
CUDA_LIBRARY_FIND(cuMemsetD2D8);
CUDA_LIBRARY_FIND(cuMemsetD2D16);
CUDA_LIBRARY_FIND(cuMemsetD2D32);
CUDA_LIBRARY_FIND(cuFuncSetBlockShape);
CUDA_LIBRARY_FIND(cuFuncSetSharedSize);
CUDA_LIBRARY_FIND(cuFuncGetAttribute);
CUDA_LIBRARY_FIND(cuArrayCreate);
CUDA_LIBRARY_FIND(cuArrayGetDescriptor);
CUDA_LIBRARY_FIND(cuArrayDestroy);
CUDA_LIBRARY_FIND(cuArray3DCreate);
CUDA_LIBRARY_FIND(cuArray3DGetDescriptor);
CUDA_LIBRARY_FIND(cuTexRefCreate);
CUDA_LIBRARY_FIND(cuTexRefDestroy);
CUDA_LIBRARY_FIND(cuTexRefSetArray);
CUDA_LIBRARY_FIND(cuTexRefSetAddress);
CUDA_LIBRARY_FIND(cuTexRefSetAddress2D);
CUDA_LIBRARY_FIND(cuTexRefSetFormat);
CUDA_LIBRARY_FIND(cuTexRefSetAddressMode);
CUDA_LIBRARY_FIND(cuTexRefSetFilterMode);
CUDA_LIBRARY_FIND(cuTexRefSetFlags);
CUDA_LIBRARY_FIND(cuTexRefGetAddress);
CUDA_LIBRARY_FIND(cuTexRefGetArray);
CUDA_LIBRARY_FIND(cuTexRefGetAddressMode);
CUDA_LIBRARY_FIND(cuTexRefGetFilterMode);
CUDA_LIBRARY_FIND(cuTexRefGetFormat);
CUDA_LIBRARY_FIND(cuTexRefGetFlags);
CUDA_LIBRARY_FIND(cuParamSetSize);
CUDA_LIBRARY_FIND(cuParamSeti);
CUDA_LIBRARY_FIND(cuParamSetf);
CUDA_LIBRARY_FIND(cuParamSetv);
CUDA_LIBRARY_FIND(cuParamSetTexRef);
CUDA_LIBRARY_FIND(cuLaunch);
CUDA_LIBRARY_FIND(cuLaunchGrid);
CUDA_LIBRARY_FIND(cuLaunchGridAsync);
CUDA_LIBRARY_FIND(cuEventCreate);
CUDA_LIBRARY_FIND(cuEventRecord);
CUDA_LIBRARY_FIND(cuEventQuery);
CUDA_LIBRARY_FIND(cuEventSynchronize);
CUDA_LIBRARY_FIND(cuEventDestroy);
CUDA_LIBRARY_FIND(cuEventElapsedTime);
CUDA_LIBRARY_FIND(cuStreamCreate);
CUDA_LIBRARY_FIND(cuStreamQuery);
CUDA_LIBRARY_FIND(cuStreamSynchronize);
CUDA_LIBRARY_FIND(cuStreamDestroy);
/* cuda 2.1 */
CUDA_LIBRARY_FIND(cuModuleLoadDataEx);
CUDA_LIBRARY_FIND(cuModuleLoadFatBinary);
CUDA_LIBRARY_FIND(cuGLCtxCreate);
CUDA_LIBRARY_FIND(cuGraphicsGLRegisterBuffer);
CUDA_LIBRARY_FIND(cuGraphicsGLRegisterImage);
/* cuda 2.3 */
CUDA_LIBRARY_FIND(cuMemHostGetFlags);
CUDA_LIBRARY_FIND(cuGraphicsGLRegisterBuffer);
CUDA_LIBRARY_FIND(cuGraphicsGLRegisterImage);
/* cuda 3.0 */
CUDA_LIBRARY_FIND(cuMemcpyDtoDAsync);
CUDA_LIBRARY_FIND(cuFuncSetCacheConfig);
CUDA_LIBRARY_FIND(cuGraphicsUnregisterResource);
CUDA_LIBRARY_FIND(cuGraphicsSubResourceGetMappedArray);
CUDA_LIBRARY_FIND(cuGraphicsResourceGetMappedPointer);
CUDA_LIBRARY_FIND(cuGraphicsResourceSetMapFlags);
CUDA_LIBRARY_FIND(cuGraphicsMapResources);
CUDA_LIBRARY_FIND(cuGraphicsUnmapResources);
CUDA_LIBRARY_FIND(cuGetExportTable);
/* cuda 3.1 */
CUDA_LIBRARY_FIND(cuModuleGetSurfRef);
CUDA_LIBRARY_FIND(cuSurfRefSetArray);
CUDA_LIBRARY_FIND(cuSurfRefGetArray);
CUDA_LIBRARY_FIND(cuCtxSetLimit);
CUDA_LIBRARY_FIND(cuCtxGetLimit);
/* functions which changed 3.1 -> 3.2 for 64 bit stuff, the cuda library
* has both the old ones for compatibility and new ones with _v2 postfix,
* we load the _v2 ones here. */
CUDA_LIBRARY_FIND_V2(cuDeviceTotalMem);
CUDA_LIBRARY_FIND_V2(cuCtxCreate);
CUDA_LIBRARY_FIND_V2(cuModuleGetGlobal);
CUDA_LIBRARY_FIND_V2(cuMemGetInfo);
CUDA_LIBRARY_FIND_V2(cuMemAlloc);
CUDA_LIBRARY_FIND_V2(cuMemAllocPitch);
CUDA_LIBRARY_FIND_V2(cuMemFree);
CUDA_LIBRARY_FIND_V2(cuMemGetAddressRange);
CUDA_LIBRARY_FIND_V2(cuMemAllocHost);
CUDA_LIBRARY_FIND_V2(cuMemHostGetDevicePointer);
CUDA_LIBRARY_FIND_V2(cuMemcpyHtoD);
CUDA_LIBRARY_FIND_V2(cuMemcpyDtoH);
CUDA_LIBRARY_FIND_V2(cuMemcpyDtoD);
CUDA_LIBRARY_FIND_V2(cuMemcpyDtoA);
CUDA_LIBRARY_FIND_V2(cuMemcpyAtoD);
CUDA_LIBRARY_FIND_V2(cuMemcpyHtoA);
CUDA_LIBRARY_FIND_V2(cuMemcpyAtoH);
CUDA_LIBRARY_FIND_V2(cuMemcpyAtoA);
CUDA_LIBRARY_FIND_V2(cuMemcpyHtoAAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpyAtoHAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpy2D);
CUDA_LIBRARY_FIND_V2(cuMemcpy2DUnaligned);
CUDA_LIBRARY_FIND_V2(cuMemcpy3D);
CUDA_LIBRARY_FIND_V2(cuMemcpyHtoDAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpyDtoHAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpyDtoDAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpy2DAsync);
CUDA_LIBRARY_FIND_V2(cuMemcpy3DAsync);
CUDA_LIBRARY_FIND_V2(cuMemsetD8);
CUDA_LIBRARY_FIND_V2(cuMemsetD16);
CUDA_LIBRARY_FIND_V2(cuMemsetD32);
CUDA_LIBRARY_FIND_V2(cuMemsetD2D8);
CUDA_LIBRARY_FIND_V2(cuMemsetD2D16);
CUDA_LIBRARY_FIND_V2(cuMemsetD2D32);
CUDA_LIBRARY_FIND_V2(cuArrayCreate);
CUDA_LIBRARY_FIND_V2(cuArrayGetDescriptor);
CUDA_LIBRARY_FIND_V2(cuArray3DCreate);
CUDA_LIBRARY_FIND_V2(cuArray3DGetDescriptor);
CUDA_LIBRARY_FIND_V2(cuTexRefSetAddress);
CUDA_LIBRARY_FIND_V2(cuTexRefSetAddress2D);
CUDA_LIBRARY_FIND_V2(cuTexRefGetAddress);
CUDA_LIBRARY_FIND_V2(cuGraphicsResourceGetMappedPointer);
CUDA_LIBRARY_FIND_V2(cuGLCtxCreate);
/* cuda 4.0 */
CUDA_LIBRARY_FIND(cuCtxSetCurrent);
if(cuHavePrecompiledKernels())
result = true;
#ifndef _WIN32
else if(cuCompilerPath() != "")
result = true;
#endif
return result;
}
bool cuHavePrecompiledKernels()
{
string cubins_path = path_get("lib");
return path_exists(cubins_path);
}
string cuCompilerPath()
{
#ifdef _WIN32
const char *defaultpaths[] = {"C:/CUDA/bin", NULL};
const char *executable = "nvcc.exe";
#else
const char *defaultpaths[] = {
"/Developer/NVIDIA/CUDA-5.0/bin",
"/usr/local/cuda-5.0/bin",
"/usr/local/cuda/bin",
"/Developer/NVIDIA/CUDA-4.2/bin",
"/usr/local/cuda-4.2/bin",
"/Developer/NVIDIA/CUDA-5.5/bin",
"/usr/local/cuda-5.5/bin",
NULL};
const char *executable = "nvcc";
#endif
const char *binpath = getenv("CUDA_BIN_PATH");
string nvcc;
if(binpath) {
nvcc = path_join(binpath, executable);
if(path_exists(nvcc))
return nvcc;
}
for(int i = 0; defaultpaths[i]; i++) {
nvcc = path_join(defaultpaths[i], executable);
if(path_exists(nvcc))
return nvcc;
}
#ifndef _WIN32
{
FILE *handle = popen("which nvcc", "r");
if(handle) {
char buffer[4096] = {0};
int len = fread(buffer, 1, sizeof(buffer) - 1, handle);
buffer[len] = '\0';
pclose(handle);
if(buffer[0])
return "nvcc";
}
}
#endif
return "";
}
int cuCompilerVersion()
{
string path = cuCompilerPath();
if(path == "")
return 0;
/* get --version output */
FILE *pipe = popen((path + " --version").c_str(), "r");
if(!pipe) {
fprintf(stderr, "CUDA: failed to run compiler to retrieve version");
return 0;
}
char buf[128];
string output = "";
while(!feof(pipe))
if(fgets(buf, 128, pipe) != NULL)
output += buf;
pclose(pipe);
/* parse version number */
string marker = "Cuda compilation tools, release ";
size_t offset = output.find(marker);
if(offset == string::npos) {
fprintf(stderr, "CUDA: failed to find version number in:\n\n%s\n", output.c_str());
return 0;
}
string versionstr = output.substr(offset + marker.size(), string::npos);
int major, minor;
if(sscanf(versionstr.c_str(), "%d.%d", &major, &minor) < 2) {
fprintf(stderr, "CUDA: failed to parse version number from:\n\n%s\n", output.c_str());
return 0;
}
return 10*major + minor;
}
CCL_NAMESPACE_END