GPGPU: generate control flow within the kernel

llvm-svn: 275956

diff --git a/polly/lib/CodeGen/PPCGCodeGeneration.cpp b/polly/lib/CodeGen/PPCGCodeGeneration.cpp
index fcfc9ff..de261fc 100644 (file)
--- a/polly/lib/CodeGen/PPCGCodeGeneration.cpp
+++ b/polly/lib/CodeGen/PPCGCodeGeneration.cpp
@@ -192,6 +192,8 @@ void GPUNodeBuilder::createKernel(__isl_take isl_ast_node *KernelStmt) {
 
   createKernelFunction(Kernel);
 
+  create(isl_ast_node_copy(Kernel->tree));
+
   Builder.SetInsertPoint(&HostInsertPoint);
   IDToValue = HostIDs;
 
@@ -304,8 +306,12 @@ void GPUNodeBuilder::createKernelFunction(ppcg_kernel *Kernel) {
 
   Function *FN = createKernelFunctionDecl(Kernel);
 
+  BasicBlock *PrevBlock = Builder.GetInsertBlock();
   auto EntryBlock = BasicBlock::Create(Builder.getContext(), "entry", FN);
 
+  DominatorTree &DT = P->getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+  DT.addNewBlock(EntryBlock, PrevBlock);
+
   Builder.SetInsertPoint(EntryBlock);
   Builder.CreateRetVoid();
   Builder.SetInsertPoint(EntryBlock, EntryBlock->begin());

diff --git a/polly/test/GPGPU/double-parallel-loop.ll b/polly/test/GPGPU/double-parallel-loop.ll
index 5e80c3f..f9b3f1f 100644 (file)
--- a/polly/test/GPGPU/double-parallel-loop.ll
+++ b/polly/test/GPGPU/double-parallel-loop.ll
@@ -93,7 +93,7 @@
 ; IR: polly.exiting:
 ; IR-NEXT:    br label %polly.merge_new_and_old
 
-; KERNEL-IR: define ptx_kernel void @kernel_0(i8* %MemRef_A) {
+; KERNEL-IR-LABEL: define ptx_kernel void @kernel_0(i8* %MemRef_A) {
 ; KERNEL-IR-NEXT: entry:
 ; KERNEL-IR-NEXT:   %0 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x()
 ; KERNEL-IR-NEXT:   %b0 = zext i32 %0 to i64
@@ -103,7 +103,19 @@
 ; KERNEL-IR-NEXT:   %t0 = zext i32 %2 to i64
 ; KERNEL-IR-NEXT:   %3 = call i32 @llvm.nvvm.read.ptx.sreg.tid.y()
 ; KERNEL-IR-NEXT:   %t1 = zext i32 %3 to i64
+; KERNEL-IR-NEXT:   br label %polly.loop_preheader
+
+; KERNEL-IR-LABEL: polly.loop_exit:
 ; KERNEL-IR-NEXT:   ret void
+
+; KERNEL-IR-LABEL: polly.loop_header:
+; KERNEL-IR-NEXT:   %polly.indvar = phi i64 [ 0, %polly.loop_preheader ], [ %polly.indvar_next, %polly.loop_header ]
+; KERNEL-IR-NEXT:   %polly.indvar_next = add nsw i64 %polly.indvar, 1
+; KERNEL-IR-NEXT:   %polly.loop_cond = icmp sle i64 %polly.indvar, 0
+; KERNEL-IR-NEXT:   br i1 %polly.loop_cond, label %polly.loop_header, label %polly.loop_exit
+
+; KERNEL-IR-LABEL: polly.loop_preheader:
+; KERNEL-IR-NEXT:   br label %polly.loop_header
 ; KERNEL-IR-NEXT: }
 
 ;    void double_parallel_loop(float A[][1024]) {

diff --git a/polly/test/GPGPU/host-control-flow.ll b/polly/test/GPGPU/host-control-flow.ll
index 9125863..d3ec4bb 100644 (file)
--- a/polly/test/GPGPU/host-control-flow.ll
+++ b/polly/test/GPGPU/host-control-flow.ll
@@ -34,13 +34,28 @@
 ; IR-NEXT:   %polly.loop_cond = icmp sle i64 %polly.indvar, 98
 ; IR-NEXT:   br i1 %polly.loop_cond, label %polly.loop_header, label %polly.loop_exit
 
-; KERNEL-IR: define ptx_kernel void @kernel_0(i8* %MemRef_A, i64 %c0) {
+; KERNEL-IR-LABEL: define ptx_kernel void @kernel_0(i8* %MemRef_A, i64 %c0) {
 ; KERNEL-IR-NEXT: entry:
 ; KERNEL-IR-NEXT:   %0 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x()
 ; KERNEL-IR-NEXT:   %b0 = zext i32 %0 to i64
 ; KERNEL-IR-NEXT:   %1 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
 ; KERNEL-IR-NEXT:   %t0 = zext i32 %1 to i64
+; KERNEL-IR-NEXT:   br label %polly.cond
+
+; KERNEL-IR-LABEL: polly.cond:
+; KERNEL-IR-NEXT:   %2 = mul nsw i64 32, %b0
+; KERNEL-IR-NEXT:   %3 = add nsw i64 %2, %t0
+; KERNEL-IR-NEXT:   %4 = icmp sle i64 %3, 97
+; KERNEL-IR-NEXT:   br i1 %4, label %polly.then, label %polly.else
+
+; KERNEL-IR-LABEL: polly.merge:
 ; KERNEL-IR-NEXT:   ret void
+
+; KERNEL-IR-LABEL: polly.then:
+; KERNEL-IR-NEXT:   br label %polly.merge
+
+; KERNEL-IR-LABEL: polly.else:
+; KERNEL-IR-NEXT:   br label %polly.merge
 ; KERNEL-IR-NEXT: }
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"

diff --git a/polly/test/GPGPU/kernel-params-scop-parameter.ll b/polly/test/GPGPU/kernel-params-scop-parameter.ll
index d73af4d..b409816 100644 (file)
--- a/polly/test/GPGPU/kernel-params-scop-parameter.ll
+++ b/polly/test/GPGPU/kernel-params-scop-parameter.ll
@@ -8,13 +8,6 @@
 ;    }
 
 ; KERNEL-IR: define ptx_kernel void @kernel_0(i8* %MemRef_A, i64 %n) {
-; KERNEL-IR-NEXT: entry:
-; KERNEL-IR-NEXT:   %0 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x()
-; KERNEL-IR-NEXT:   %b0 = zext i32 %0 to i64
-; KERNEL-IR-NEXT:   %1 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
-; KERNEL-IR-NEXT:   %t0 = zext i32 %1 to i64
-; KERNEL-IR-NEXT:   ret void
-; KERNEL-IR-NEXT: }
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"