returnValue VectorspaceElement::printToString( char** string,
const char* const name,
PrintScheme printScheme,
BooleanType allocateMemory
) const
{
char* startString = 0;
char* endString = 0;
uint width = 0;
uint precision = 0;
char* colSeparator = 0;
char* rowSeparator = 0;
returnValue returnvalue;
returnvalue = getGlobalStringDefinitions( printScheme,&startString,&endString,
width,precision,&colSeparator,&rowSeparator );
if ( returnvalue == SUCCESSFUL_RETURN )
{
returnvalue = printToString( string,name,startString,endString,width,precision,colSeparator,rowSeparator,allocateMemory );
}
if ( startString != 0 ) delete[] startString;
if ( endString != 0 ) delete[] endString;
if ( colSeparator != 0 ) delete[] colSeparator;
if ( rowSeparator != 0 ) delete[] rowSeparator;
return returnvalue;
}
void puts(char *s) {
char c;
// Loops through each character in string 's'
while (c = *s++) {
if(toSer)
sendByte(c);
else
printToString(c);
}
}
returnValue VectorspaceElement::print( const char* const name,
PrintScheme printScheme
) const
{
char* string = 0;
printToString( &string, name,printScheme );
acadoPrintf( "%s",string );
if ( string != 0 )
delete[] string;
return SUCCESSFUL_RETURN;
}
returnValue VectorspaceElement::print( const char* const name,
const char* const startString,
const char* const endString,
uint width,
uint precision,
const char* const colSeparator,
const char* const rowSeparator
) const
{
char* string = 0;
printToString( &string, name,startString,endString,width,precision,colSeparator,rowSeparator );
acadoPrintf( "%s",string );
if ( string != 0 )
delete[] string;
return SUCCESSFUL_RETURN;
}
std::vector<std::string> generateLibraryPtx(int devMajor, int devMinor) {
if (libdevicePtxCache.size() > 0 &&
intrinsicsPtxCache.size() > 0) {
return {libdevicePtxCache, intrinsicsPtxCache};
}
// Build libdevice (math libraries, etc.) module
//
// Reference:
// http://docs.nvidia.com/cuda/libdevice-users-guide/basic-usage.html
//
// The device to libdevice version mapping is weird (note 3.1-3.4=compute_20)
// 2.0 ≤ Arch < 3.0 libdevice.compute_20.XX.bc
// Arch = 3.0 libdevice.compute_30.XX.bc
// 3.1 ≤ Arch < 3.5 libdevice.compute_20.XX.bc
// Arch = 3.5 libdevice.compute_35.XX.bc
// Identify device by Compute API level
const char *libdevice;
int libdevice_length;
if ((devMajor == 3 && devMinor >= 5) ||
devMajor > 3) {
libdevice = reinterpret_cast<const char*>(simit_gpu_libdevice_compute_35);
libdevice_length = simit_gpu_libdevice_compute_35_length;
}
else if (devMajor == 3 && devMinor >= 0) {
libdevice = reinterpret_cast<const char*>(simit_gpu_libdevice_compute_30);
libdevice_length = simit_gpu_libdevice_compute_30_length;
} else {
libdevice = reinterpret_cast<const char*>(simit_gpu_libdevice_compute_20);
libdevice_length = simit_gpu_libdevice_compute_20_length;
}
llvm::SMDiagnostic errReport;
libdevice_length = alignBitreaderLength(libdevice_length);
llvm::MemoryBufferRef libdeviceBuf(
llvm::StringRef(libdevice, libdevice_length), "libdevice");
std::unique_ptr<llvm::Module> libdeviceModule =
llvm::parseIR(libdeviceBuf, errReport, LLVM_CTX);
iassert((bool)libdeviceModule)
<< "Failed to load libdevice: " << printToString(errReport);
setNVVMModuleProps(libdeviceModule.get());
logModule(libdeviceModule.get(), "simit-libdevice.ll");
std::string libdevicePtx = generatePtx(libdeviceModule.get(), devMajor, devMinor);
// Build intrinsics module
const char *intrinsics = reinterpret_cast<const char*>(simit_gpu_intrinsics);
int intrinsics_length = alignBitreaderLength(simit_gpu_intrinsics_length);
llvm::MemoryBufferRef intrinsicsBuf(
llvm::StringRef(intrinsics, intrinsics_length), "intrinsics");
std::unique_ptr<llvm::Module> intrinsicsModule =
llvm::parseIR(intrinsicsBuf, errReport, LLVM_CTX);
iassert((bool)intrinsicsModule)
<< "Failed to load intrinsics: " << printToString(errReport);
setNVVMModuleProps(intrinsicsModule.get());
logModule(intrinsicsModule.get(), "simit-intrinsics.ll");
std::string intrinsicsPtx = generatePtx(intrinsicsModule.get(), devMajor, devMinor);
// Cache the compiled libries
libdevicePtxCache = libdevicePtx;
intrinsicsPtxCache = intrinsicsPtx;
return {libdevicePtx, intrinsicsPtx};
}
/**
* puts() is used by printf() to display or send a character. This function
* determines where printf prints to. For this case it sends a character
* out over UART.
**/
void putc(unsigned b) {
if(toSer)
sendByte(b);
else
printToString(b);
}
