[polly] r275957 - GPGPU: Emit in-kernel synchronization statements

[Tobias Grosser via llvm-commits]

Author: grosser
Date: Tue Jul 19 02:33:16 2016
New Revision: 275957

URL: http://llvm.org/viewvc/llvm-project?rev=275957&view=rev
Log:
GPGPU: Emit in-kernel synchronization statements

We use this opportunity to further classify the different user statements that
can arise and add TODOs for the ones not yet implemented.

Modified:
    polly/trunk/lib/CodeGen/PPCGCodeGeneration.cpp
    polly/trunk/test/GPGPU/host-statement.ll

Modified: polly/trunk/lib/CodeGen/PPCGCodeGeneration.cpp
URL: http://llvm.org/viewvc/llvm-project/polly/trunk/lib/CodeGen/PPCGCodeGeneration.cpp?rev=275957&r1=275956&r2=275957&view=diff
==============================================================================
--- polly/trunk/lib/CodeGen/PPCGCodeGeneration.cpp (original)
+++ polly/trunk/lib/CodeGen/PPCGCodeGeneration.cpp Tue Jul 19 02:33:16 2016
@@ -115,6 +115,8 @@ private:
   ///   - ScopStmt:      A computational statement (TODO)
   ///   - Kernel:        A GPU kernel call (TODO)
   ///   - Data-Transfer: A GPU <-> CPU data-transfer (TODO)
+  ///   - In-kernel synchronization
+  ///   - In-kernel memory copy statement
   ///
   /// @param UserStmt The ast node to generate code for.
   virtual void createUser(__isl_take isl_ast_node *UserStmt);
@@ -153,6 +155,9 @@ private:
   /// @param The kernel to generate the intrinsic functions for.
   void insertKernelIntrinsics(ppcg_kernel *Kernel);
 
+  /// Create an in-kernel synchronization call.
+  void createKernelSync();
+
   /// Finalize the generation of the kernel function.
   ///
   /// Free the LLVM-IR module corresponding to the kernel and -- if requested --
@@ -160,6 +165,14 @@ private:
   void finalizeKernelFunction();
 };
 
+/// Check if one string is a prefix of another.
+///
+/// @param String The string in which to look for the prefix.
+/// @param Prefix The prefix to look for.
+static bool isPrefix(std::string String, std::string Prefix) {
+  return String.find(Prefix) == 0;
+}
+
 void GPUNodeBuilder::createUser(__isl_take isl_ast_node *UserStmt) {
   isl_ast_expr *Expr = isl_ast_node_user_get_expr(UserStmt);
   isl_ast_expr *StmtExpr = isl_ast_expr_get_op_arg(Expr, 0);
@@ -174,11 +187,47 @@ void GPUNodeBuilder::createUser(__isl_ta
     return;
   }
 
+  if (isPrefix(Str, "to_device") || isPrefix(Str, "from_device")) {
+    // TODO: Insert memory copies
+    isl_ast_expr_free(Expr);
+    isl_ast_node_free(UserStmt);
+    return;
+  }
+
+  isl_id *Anno = isl_ast_node_get_annotation(UserStmt);
+  struct ppcg_kernel_stmt *KernelStmt =
+      (struct ppcg_kernel_stmt *)isl_id_get_user(Anno);
+  isl_id_free(Anno);
+
+  switch (KernelStmt->type) {
+  case ppcg_kernel_domain:
+    // TODO Create kernel user stmt
+    isl_ast_expr_free(Expr);
+    isl_ast_node_free(UserStmt);
+    return;
+  case ppcg_kernel_copy:
+    // TODO: Create kernel copy stmt
+    isl_ast_expr_free(Expr);
+    isl_ast_node_free(UserStmt);
+    return;
+  case ppcg_kernel_sync:
+    createKernelSync();
+    isl_ast_expr_free(Expr);
+    isl_ast_node_free(UserStmt);
+    return;
+  }
+
   isl_ast_expr_free(Expr);
   isl_ast_node_free(UserStmt);
   return;
 }
 
+void GPUNodeBuilder::createKernelSync() {
+  Module *M = Builder.GetInsertBlock()->getParent()->getParent();
+  auto *Sync = Intrinsic::getDeclaration(M, Intrinsic::nvvm_barrier0);
+  Builder.CreateCall(Sync, {});
+}
+
 void GPUNodeBuilder::createKernel(__isl_take isl_ast_node *KernelStmt) {
   isl_id *Id = isl_ast_node_get_annotation(KernelStmt);
   ppcg_kernel *Kernel = (ppcg_kernel *)isl_id_get_user(Id);

Modified: polly/trunk/test/GPGPU/host-statement.ll
URL: http://llvm.org/viewvc/llvm-project/polly/trunk/test/GPGPU/host-statement.ll?rev=275957&r1=275956&r2=275957&view=diff
==============================================================================
--- polly/trunk/test/GPGPU/host-statement.ll (original)
+++ polly/trunk/test/GPGPU/host-statement.ll Tue Jul 19 02:33:16 2016
@@ -2,6 +2,10 @@
 ; RUN: -disable-output < %s | \
 ; RUN: FileCheck -check-prefix=CODE %s
 
+; RUN: opt %loadPolly -polly-codegen-ppcg -polly-acc-dump-kernel-ir \
+; RUN: -disable-output < %s | \
+; RUN: FileCheck -check-prefix=KERNEL-IR %s
+
 ; REQUIRES: pollyacc
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -68,6 +72,7 @@ declare void @llvm.lifetime.start(i64, i
 ; CODE-NEXT:     for (int c3 = 0; c3 <= 1; c3 += 1)
 ; CODE-NEXT:       Stmt_for_body62(32 * b0 + t0 + 8192 * c0, 32 * b1 + t1 + 16 * c3);
 
+; KERNEL-IR: call void @llvm.nvvm.barrier0()
 
 ; Function Attrs: nounwind uwtable
 define internal void @kernel_gramschmidt(i32 %ni, i32 %nj, [512 x double]* %A, [512 x double]* %R, [512 x double]* %Q) #1 {