[x86] Front-end part of MCU psABI support

[andreybokhanko]

Summary

This patch implements front-end parts of MCU psABI (https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnx4MzJhYml8Z3g6NzFhMGYzMjUzZDQzODllMw) support:

1. "long double type is the same as double."
2. "New predefined C/C++ pre-processor symbols: iamcu and iamcu__.

[mkuper]

LGTM

lib/Basic/Targets.cpp

2327 ↗

(On Diff #38814)

Do you know if we should also change this to be 0?

[andreybokhanko]

Michael,

Thanks for the review!

As for your comment, I will investigate and fix (if needed).

Andrey

lib/Basic/Targets.cpp

2327 ↗

(On Diff #38814)

Good catch.

Let me investigate and change (if needed) in another patch.

cfe/trunk/lib/Basic/Targets.cpp

enum FPMathKind {
     FP_Default,
     FP_SSE,
     FP_387
   } FPMath = FP_Default;
 
 public:
   X86TargetInfo(const llvm::Triple &Triple) : TargetInfo(Triple) {
     BigEndian = false;
-    LongDoubleFormat = &llvm::APFloat::x87DoubleExtended;
+    LongDoubleFormat = Triple.isOSIAMCU() ? &llvm::APFloat::IEEEdouble
+                                          : &llvm::APFloat::x87DoubleExtended;
   }
   unsigned getFloatEvalMethod() const override {
     // X87 evaluates with 80 bits "long double" precision.
     return SSELevel == NoSSE ? 2 : 0;
   }
   ArrayRef<Builtin::Info> getTargetBuiltins() const override {
     return llvm::makeArrayRef(BuiltinInfo,
                              clang::X86::LastTSBuiltin-Builtin::FirstTSBuiltin);
void X86TargetInfo::getTargetDefines(const LangOptions &Opts,
 
   if (CPU >= CK_i486) {
     Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_1");
     Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2");
     Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4");
   }
   if (CPU >= CK_i586)
     Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8");
+
+  if (getTriple().isOSIAMCU()) {
+    Builder.defineMacro("__iamcu");
+    Builder.defineMacro("__iamcu__");
+  }
 }
 
 bool X86TargetInfo::hasFeature(StringRef Feature) const {
   return llvm::StringSwitch<bool>(Feature)
       .Case("aes", HasAES)
       .Case("avx", SSELevel >= AVX)
       .Case("avx2", SSELevel >= AVX2)
       .Case("avx512f", SSELevel >= AVX512F)

cfe/trunk/test/CodeGen/x86_32-arguments-iamcu.c

 // CHECK-LABEL: define void @retLargeStruct(%struct.st12_t* inreg noalias sret %agg.result, i32 inreg %i1, i32 %r.0, i32 %r.1, i32 %r.2)
 st12_t retLargeStruct(int i1, st12_t r) { return r; }
 
 // FIXME: We really shouldn't be marking this inreg. Right now the
 // inreg gets ignored by the CG for varargs functions, but that's
 // insane.
 // CHECK-LABEL: define i32 @varArgs(i32 inreg %i1, ...)
 int varArgs(int i1, ...) { return i1; }
+
+// CHECK-LABEL: define double @longDoubleArg(double %ld1)
+long double longDoubleArg(long double ld1) { return ld1; }
+

cfe/trunk/test/Preprocessor/elfiamcu-predefines.c

+// RUN: %clang_cc1 -E -dM -triple i586-intel-elfiamcu | FileCheck %s
+
+// CHECK: #define __iamcu
+// CHECK: #define __iamcu__
+