[NVPTX] Add GenericToNVVM IR converter to better handle idiomatic LLVM IR inputs

This converter currently only handles global variables in address space 0. For
these variables, they are promoted to address space 1 (global memory), and all
uses are updated to point to the result of a cvta.global instruction on the new
variable.

The motivation for this is address space 0 global variables are illegal since we
cannot declare variables in the generic address space.  Instead, we place the
variables in address space 1 and explicitly convert the pointer to address
space 0. This is primarily intended to help new users who expect to be able to
place global variables in the default address space.

llvm-svn: 182254

1 files changed, 8 insertions, 0 deletions

diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
index 5f35edf2191..507fcb5d6af 100644
--- a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
@@ -49,6 +49,7 @@ using namespace llvm;
 
 namespace llvm {
 void initializeNVVMReflectPass(PassRegistry&);
+void initializeGenericToNVVMPass(PassRegistry&);
 }
 
 extern "C" void LLVMInitializeNVPTXTarget() {
@@ -62,6 +63,7 @@ extern "C" void LLVMInitializeNVPTXTarget() {
   // FIXME: This pass is really intended to be invoked during IR optimization,
   // but it's very NVPTX-specific.
   initializeNVVMReflectPass(*PassRegistry::getPassRegistry());
+  initializeGenericToNVVMPass(*PassRegistry::getPassRegistry());
 }
 
 NVPTXTargetMachine::NVPTXTargetMachine(
@@ -102,6 +104,7 @@ public:
     return getTM<NVPTXTargetMachine>();
   }
 
+  virtual void addIRPasses();
   virtual bool addInstSelector();
   virtual bool addPreRegAlloc();
 };
@@ -112,6 +115,11 @@ TargetPassConfig *NVPTXTargetMachine::createPassConfig(PassManagerBase &PM) {
   return PassConfig;
 }
 
+void NVPTXPassConfig::addIRPasses() {
+  TargetPassConfig::addIRPasses();
+  addPass(createGenericToNVVMPass());
+}
+
 bool NVPTXPassConfig::addInstSelector() {
   addPass(createLowerAggrCopies());
   addPass(createSplitBBatBarPass());