Append options for floating-point commutivity when related diagnostics are produced.

With this patch clang appends the command line options that would allow vectorization when floating-point commutativity is required. Specifically those are enabling fast-math or specifying a loop hint. 

llvm-svn: 244492

1 files changed, 19 insertions, 0 deletions

diff --git a/clang/lib/CodeGen/CodeGenAction.cpp b/clang/lib/CodeGen/CodeGenAction.cpp
index d97a64e..d47997e 100644
--- a/clang/lib/CodeGen/CodeGenAction.cpp
+++ b/clang/lib/CodeGen/CodeGenAction.cpp
@@ -256,6 +256,8 @@ namespace clang {
         const llvm::DiagnosticInfoOptimizationRemarkMissed &D);
     void OptimizationRemarkHandler(
         const llvm::DiagnosticInfoOptimizationRemarkAnalysis &D);
+    void OptimizationRemarkHandler(
+        const llvm::DiagnosticInfoOptimizationRemarkAnalysisFPCommute &D);
     void OptimizationFailureHandler(
         const llvm::DiagnosticInfoOptimizationFailure &D);
   };
@@ -500,6 +502,17 @@ void BackendConsumer::OptimizationRemarkHandler(
         D, diag::remark_fe_backend_optimization_remark_analysis);
 }
 
+void BackendConsumer::OptimizationRemarkHandler(
+    const llvm::DiagnosticInfoOptimizationRemarkAnalysisFPCommute &D) {
+  // Optimization analysis remarks are active only if the -Rpass-analysis
+  // flag has a regular expression that matches the name of the pass
+  // name in \p D.
+  if (CodeGenOpts.OptimizationRemarkAnalysisPattern &&
+      CodeGenOpts.OptimizationRemarkAnalysisPattern->match(D.getPassName()))
+    EmitOptimizationMessage(
+        D, diag::remark_fe_backend_optimization_remark_analysis_fpcommute);
+}
+
 void BackendConsumer::OptimizationFailureHandler(
     const llvm::DiagnosticInfoOptimizationFailure &D) {
   EmitOptimizationMessage(D, diag::warn_fe_backend_optimization_failure);
@@ -553,6 +566,12 @@ void BackendConsumer::DiagnosticHandlerImpl(const DiagnosticInfo &DI) {
     OptimizationRemarkHandler(
         cast<DiagnosticInfoOptimizationRemarkAnalysis>(DI));
     return;
+  case llvm::DK_OptimizationRemarkAnalysisFPCommute:
+    // Optimization remarks are always handled completely by this
+    // handler. There is no generic way of emitting them.
+    OptimizationRemarkHandler(
+        cast<DiagnosticInfoOptimizationRemarkAnalysisFPCommute>(DI));
+    return;
   case llvm::DK_OptimizationFailure:
     // Optimization failures are always handled completely by this
     // handler.