[OPENMP] Fix mixture of omp and clang pragmas

[hishiguro]

Summary

Fixes PR45753

When a program that contains a loop to which both "omp parallel for" pragma and "clang loop" pragma are associated is compiled with the -fopenmp option, "clang loop" pragma will not take effect.
The example below should not be vectorized by the "clang loop` pragma but is it actually vectorized.
The cause is that "llvm.loop.vectorize.width" is not output to the IR when -fopenmp is specified.
The fix attaches attributes if they exist in the loop.

[example.c]

int a[100], b[100];
int foo() {
#pragma omp parallel for
#pragma clang loop vectorize(disable)
for (int i=0; i<100; i++)
  a[i]+=b[i]*i;
}

[compile]
clang -O2 -fopenmp a.c -c -Rpass=vect
a.c:3:1: remark: vectorized loop (vectorization width: 4, interleaved count: 2) [-Rpass=loop-vectorize] #pragma omp parallel for ^

[IR]

• -fopenmp

$ clang -O2 a.c -S -emit-llvm -mllvm -disable-llvm-optzns -o - -fopenmp |grep "vectorize\.width"
$

• -fno-openmp

$ clang -O2 a.c -S -emit-llvm -mllvm -disable-llvm-optzns -o - -fno-openmp |grep "vectorize\.width"
!7 = !{!"llvm.loop.vectorize.width", i32 1}

[hishiguro]

Fix clang-format error.

[ABataev]

Tests?

clang/lib/CodeGen/CGStmtOpenMP.cpp

1749

1. Cast it to OMPExecutableDirective here, it should be enough.
2. No need for conditional casting, use cast<OMPExecutableDirective>(...). Or you can change the function itself to accept const OMPExecutableDirective &S instead of const Stmt &S.

1750

Use getInnermostCapturedStmt() instead.

[hishiguro]

Thank you for your comments. I will check those comments.

[hishiguro]

I checked that your comments are correct. I modified the source, and added a test.

[ABataev]

clang/lib/CodeGen/CGStmtOpenMP.cpp

1735

const auto &OMPED = cast<OMPExecutableDirective>(S);

[hishiguro]

Thank you for your comment. I updated source.

[ABataev]

LG

clang/lib/CodeGen/CGStmtOpenMP.cpp

void CodeGenFunction::EmitOMPInnerLoop(
     const llvm::function_ref<void(CodeGenFunction &)> BodyGen,
     const llvm::function_ref<void(CodeGenFunction &)> PostIncGen) {
   auto LoopExit = getJumpDestInCurrentScope("omp.inner.for.end");
 
   // Start the loop with a block that tests the condition.
   auto CondBlock = createBasicBlock("omp.inner.for.cond");
   EmitBlock(CondBlock);
   const SourceRange R = S.getSourceRange();
+
+  // If attributes are attached, push to the basic block with them.
+  const auto &OMPED = cast<OMPExecutableDirective>(S);
+  const CapturedStmt *ICS = OMPED.getInnermostCapturedStmt();
+  const Stmt *SS = ICS->getCapturedStmt();
+  const AttributedStmt *AS = dyn_cast_or_null<AttributedStmt>(SS);
+  if (AS)
+    LoopStack.push(CondBlock, CGM.getContext(), CGM.getCodeGenOpts(),
+                   AS->getAttrs(), SourceLocToDebugLoc(R.getBegin()),
+                   SourceLocToDebugLoc(R.getEnd()));
+  else
     LoopStack.push(CondBlock, SourceLocToDebugLoc(R.getBegin()),
                    SourceLocToDebugLoc(R.getEnd()));
 
   // If there are any cleanups between here and the loop-exit scope,
   // create a block to stage a loop exit along.
   llvm::BasicBlock *ExitBlock = LoopExit.getBlock();
   if (RequiresCleanup)
     ExitBlock = createBasicBlock("omp.inner.for.cond.cleanup");
 
   llvm::BasicBlock *LoopBody = createBasicBlock("omp.inner.for.body");
 
   // Emit condition.
   EmitBranchOnBoolExpr(LoopCond, LoopBody, ExitBlock, getProfileCount(&S));
   if (ExitBlock != LoopExit.getBlock()) {
     EmitBlock(ExitBlock);

clang/test/OpenMP/omp_with_loop_pragma.c

+// RUN: %clang_cc1 -verify -fopenmp -x c -emit-llvm %s -triple x86_64-unknown-linux -o - -femit-all-decls -disable-llvm-passes | FileCheck %s
+// RUN: %clang_cc1 -verify -x c -emit-llvm %s -triple x86_64-unknown-linux -o - -femit-all-decls -disable-llvm-passes | FileCheck %s
+// expected-no-diagnostics
+
+// CHECK: !{{[0-9]+}} = !{!"llvm.loop.vectorize.width", i32 1}
+void sub(double *restrict a, double *restrict b, int n) {
+  int i;
+
+#pragma omp parallel for
+#pragma clang loop vectorize(disable)
+  for (i = 0; i < n; i++) {
+    a[i] = a[i] + b[i];
+  }
+}