Revert "[flang] Integrate the option -flang-experimental-integer-overflow into -fno-wrapv" (#113901)

Reverts #110063 due to the performance regression on
503.bwaves_r in SPEC2017.

Author: yus3710-fj

clang/include/clang/Driver/Options.td

@@ -6791,6 +6791,10 @@ def flang_deprecated_no_hlfir : Flag<["-"], "flang-deprecated-no-hlfir">,
   Flags<[HelpHidden]>, Visibility<[FlangOption, FC1Option]>,
   HelpText<"Do not use HLFIR lowering (deprecated)">;
 
+def flang_experimental_integer_overflow : Flag<["-"], "flang-experimental-integer-overflow">,
+  Flags<[HelpHidden]>, Visibility<[FlangOption, FC1Option]>,
+  HelpText<"Add nsw flag to internal operations such as do-variable increment (experimental)">;
+
 //===----------------------------------------------------------------------===//
 // FLangOption + CoreOption + NoXarchOption
 //===----------------------------------------------------------------------===//

clang/lib/Driver/ToolChains/Flang.cpp

@@ -148,6 +148,7 @@ void Flang::addCodegenOptions(const ArgList &Args,
 
   Args.addAllArgs(CmdArgs, {options::OPT_flang_experimental_hlfir,
                             options::OPT_flang_deprecated_no_hlfir,
+                            options::OPT_flang_experimental_integer_overflow,
                             options::OPT_fno_ppc_native_vec_elem_order,
                             options::OPT_fppc_native_vec_elem_order});
 }

flang/include/flang/Lower/LoweringOptions.def

@@ -38,5 +38,10 @@ ENUM_LOWERINGOPT(Underscoring, unsigned, 1, 1)
 /// (i.e. wraps around as two's complement). Off by default.
 ENUM_LOWERINGOPT(IntegerWrapAround, unsigned, 1, 0)
 
+/// If true, add nsw flags to loop variable increments.
+/// Off by default.
+/// TODO: integrate this option with the above
+ENUM_LOWERINGOPT(NSWOnLoopVarInc, unsigned, 1, 0)
+
 #undef LOWERINGOPT
 #undef ENUM_LOWERINGOPT

flang/include/flang/Optimizer/Transforms/Passes.h

@@ -66,7 +66,7 @@ namespace fir {
 std::unique_ptr<mlir::Pass> createAffineDemotionPass();
 std::unique_ptr<mlir::Pass>
 createArrayValueCopyPass(fir::ArrayValueCopyOptions options = {});
-std::unique_ptr<mlir::Pass> createCFGConversionPassWithoutNSW();
+std::unique_ptr<mlir::Pass> createCFGConversionPassWithNSW();
 std::unique_ptr<mlir::Pass> createMemDataFlowOptPass();
 std::unique_ptr<mlir::Pass> createPromoteToAffinePass();
 std::unique_ptr<mlir::Pass>
@@ -83,7 +83,7 @@ createVScaleAttrPass(std::pair<unsigned, unsigned> vscaleAttr);
 
 void populateCfgConversionRewrites(mlir::RewritePatternSet &patterns,
                                    bool forceLoopToExecuteOnce = false,
-                                   bool setNSW = true);
+                                   bool setNSW = false);
 
 // declarative passes
 #define GEN_PASS_REGISTRATION

flang/include/flang/Optimizer/Transforms/Passes.td

@@ -153,7 +153,7 @@ def CFGConversion : Pass<"cfg-conversion"> {
            /*default=*/"false",
            "force the body of a loop to execute at least once">,
     Option<"setNSW", "set-nsw", "bool",
-           /*default=*/"true",
+           /*default=*/"false",
            "set nsw on loop variable increment">
   ];
 }

flang/include/flang/Tools/CrossToolHelpers.h

@@ -122,7 +122,7 @@ struct MLIRToLLVMPassPipelineConfig : public FlangEPCallBacks {
   bool NoSignedZerosFPMath =
       false; ///< Set no-signed-zeros-fp-math attribute for functions.
   bool UnsafeFPMath = false; ///< Set unsafe-fp-math attribute for functions.
-  bool NSWOnLoopVarInc = true; ///< Add nsw flag to loop variable increments.
+  bool NSWOnLoopVarInc = false; ///< Add nsw flag to loop variable increments.
 };
 
 struct OffloadModuleOpts {

flang/lib/Frontend/CompilerInvocation.cpp

@@ -1350,6 +1350,12 @@ bool CompilerInvocation::createFromArgs(
     invoc.loweringOpts.setNoPPCNativeVecElemOrder(true);
   }
 
+  // -flang-experimental-integer-overflow
+  if (args.hasArg(
+          clang::driver::options::OPT_flang_experimental_integer_overflow)) {
+    invoc.loweringOpts.setNSWOnLoopVarInc(true);
+  }
+
   // Preserve all the remark options requested, i.e. -Rpass, -Rpass-missed or
   // -Rpass-analysis. This will be used later when processing and outputting the
   // remarks generated by LLVM in ExecuteCompilerInvocation.cpp.

flang/lib/Frontend/FrontendActions.cpp

@@ -828,8 +828,8 @@ void CodeGenAction::generateLLVMIR() {
     config.VScaleMax = vsr->second;
   }
 
-  if (ci.getInvocation().getLoweringOpts().getIntegerWrapAround())
-    config.NSWOnLoopVarInc = false;
+  if (ci.getInvocation().getLoweringOpts().getNSWOnLoopVarInc())
+    config.NSWOnLoopVarInc = true;
 
   // Create the pass pipeline
   fir::createMLIRToLLVMPassPipeline(pm, config, getCurrentFile());

flang/lib/Lower/Bridge.cpp

@@ -2271,7 +2271,7 @@ class FirConverter : public Fortran::lower::AbstractConverter {
     assert(!incrementLoopNestInfo.empty() && "empty loop nest");
     mlir::Location loc = toLocation();
     mlir::arith::IntegerOverflowFlags flags{};
-    if (!getLoweringOptions().getIntegerWrapAround())
+    if (getLoweringOptions().getNSWOnLoopVarInc())
       flags = bitEnumSet(flags, mlir::arith::IntegerOverflowFlags::nsw);
     auto iofAttr = mlir::arith::IntegerOverflowFlagsAttr::get(
         builder->getContext(), flags);

flang/lib/Lower/IO.cpp

@@ -929,7 +929,7 @@ static void genIoLoop(Fortran::lower::AbstractConverter &converter,
   fir::FirOpBuilder &builder = converter.getFirOpBuilder();
   mlir::Location loc = converter.getCurrentLocation();
   mlir::arith::IntegerOverflowFlags flags{};
-  if (!converter.getLoweringOptions().getIntegerWrapAround())
+  if (converter.getLoweringOptions().getNSWOnLoopVarInc())
     flags = bitEnumSet(flags, mlir::arith::IntegerOverflowFlags::nsw);
   auto iofAttr =
       mlir::arith::IntegerOverflowFlagsAttr::get(builder.getContext(), flags);