//===----------------------------------------------------------------------===//
#include "NVPTXSubtarget.h"
+#include "NVPTXTargetMachine.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "nvptx-subtarget"
+
#define GET_SUBTARGETINFO_ENUM
#define GET_SUBTARGETINFO_TARGET_DESC
#define GET_SUBTARGETINFO_CTOR
#include "NVPTXGenSubtargetInfo.inc"
-using namespace llvm;
-
-// Select Driver Interface
-#include "llvm/Support/CommandLine.h"
-namespace {
-cl::opt<NVPTX::DrvInterface>
-DriverInterface(cl::desc("Choose driver interface:"),
- cl::values(
- clEnumValN(NVPTX::NVCL, "drvnvcl", "Nvidia OpenCL driver"),
- clEnumValN(NVPTX::CUDA, "drvcuda", "Nvidia CUDA driver"),
- clEnumValN(NVPTX::TEST, "drvtest", "Plain Test"),
- clEnumValEnd),
- cl::init(NVPTX::NVCL));
-}
+// Pin the vtable to this file.
+void NVPTXSubtarget::anchor() {}
-NVPTXSubtarget::NVPTXSubtarget(const std::string &TT, const std::string &CPU,
- const std::string &FS, bool is64Bit)
-:NVPTXGenSubtargetInfo(TT, "", FS), // Don't pass CPU to subtarget,
- // because we don't register all
- // nvptx targets.
- Is64Bit(is64Bit) {
+NVPTXSubtarget &NVPTXSubtarget::initializeSubtargetDependencies(StringRef CPU,
+ StringRef FS) {
+ // Provide the default CPU if we don't have one.
+ if (CPU.empty() && FS.size())
+ llvm_unreachable("we are not using FeatureStr");
+ TargetName = CPU.empty() ? "sm_20" : CPU;
- drvInterface = DriverInterface;
+ ParseSubtargetFeatures(TargetName, FS);
- // Provide the default CPU if none
- std::string defCPU = "sm_10";
+ // Set default to PTX 3.2 (CUDA 5.5)
+ if (PTXVersion == 0) {
+ PTXVersion = 32;
+ }
- // Get the TargetName from the FS if available
- if (FS.empty() && CPU.empty())
- TargetName = defCPU;
- else if (!CPU.empty())
- TargetName = CPU;
- else
- llvm_unreachable("we are not using FeatureStr");
+ return *this;
+}
- // Set up the SmVersion
- SmVersion = atoi(TargetName.c_str()+3);
+NVPTXSubtarget::NVPTXSubtarget(const Triple &TT, const std::string &CPU,
+ const std::string &FS,
+ const NVPTXTargetMachine &TM)
+ : NVPTXGenSubtargetInfo(TT, CPU, FS), PTXVersion(0), SmVersion(20), TM(TM),
+ InstrInfo(), TLInfo(TM, initializeSubtargetDependencies(CPU, FS)),
+ FrameLowering() {}
+
+bool NVPTXSubtarget::hasImageHandles() const {
+ // Enable handles for Kepler+, where CUDA supports indirect surfaces and
+ // textures
+ if (TM.getDrvInterface() == NVPTX::CUDA)
+ return (SmVersion >= 30);
+
+ // Disabled, otherwise
+ return false;
}