[GPGPU] Use separate basic block for GPU initialization code

This increases the readability of the IR and also clarifies that the GPU
inititialization is executed _after_ the scalar initialization which needs
to before the code of the transformed scop is executed.

Besides increased readability, the IR should not change. Specifically, I
do not expect any changes in program semantics due to this patch.

llvm-svn: 278125

polly/lib/CodeGen/PPCGCodeGeneration.cpp

@@ -35,6 +35,7 @@
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Transforms/IPO/PassManagerBuilder.h"
+#include "llvm/Transforms/Utils/BasicBlockUtils.h"
 
 #include "isl/union_map.h"
 
@@ -436,6 +437,11 @@ private:
 };
 
 void GPUNodeBuilder::initializeAfterRTH() {
+  BasicBlock *NewBB = SplitBlock(Builder.GetInsertBlock(),
+                                 &*Builder.GetInsertPoint(), &DT, &LI);
+  NewBB->setName("polly.acc.initialize");
+  Builder.SetInsertPoint(&NewBB->front());
+
   GPUContext = createCallInitContext();
   allocateDeviceArrays();
 }

polly/test/GPGPU/double-parallel-loop.ll

@@ -92,6 +92,9 @@
 ; IR-NEXT:    br i1 true, label %polly.start, label %bb2
 
 ; IR: polly.start:
+; IR-NEXT: br label %polly.acc.initialize
+
+; IR: polly.acc.initialize:
 ; IR-NEXT:    [[GPUContext:%.*]] = call i8* @polly_initContext()
 ; IR-NEXT:    %p_dev_array_MemRef_A = call i8* @polly_allocateMemoryForDevice(i64 4194304)
 ; IR-NEXT:    [[HostPtr:%.*]] = bitcast [1024 x float]* %A to i8*