[Clang][LoongArch] Define more LoongArch specific built-in macros

[SixWeining]

Summary

Define below macros according to LoongArch toolchain conventions [1].

• __loongarch_grlen
• __loongarch_frlen
• __loongarch_lp64
• __loongarch_hard_float
• __loongarch_soft_float
• __loongarch_single_float
• __loongarch_double_float

And __loongarch64 for compatibility with existing code.

Note:

1. __loongarch__ is already covered.
2. __loongarch_arch is not defined because I don't know how to access the arch name specified by -march in TargetInfo.
3. __loongarch_tune will be defined in the future.

[1]: https://loongson.github.io/LoongArch-Documentation/LoongArch-toolchain-conventions-EN.html

Depends on D136146

[xry111]

Do we support --target=loongarch64 -mabi=ilp32d or -mfpu=64 -mabi=lp64s combinations now? If true I think we'll need test cases for such combinations.

[SixWeining]

In D136413#3873216, @xry111 wrote:

Do we support --target=loongarch64 -mabi=ilp32d or -mfpu=64 -mabi=lp64s combinations now? If true I think we'll need test cases for such combinations.

For --target=loongarch64 -mabi=ilp32d, I'm not sure about this. The ABI document doesn't clearly describe this. But in current llvm implementation, it is NOT supported. In clang/lib/Basic/Targets/LoongArch.h:LoongArch64TargetInfo:setABI:

bool setABI(const std::string &Name) override {
  if (Name == "lp64d" || Name == "lp64f" || Name == "lp64s") {
    ABI = Name;
    return true;
  }
  return false;
}

For -mfpu=64 -mabi=lp64s, I think it is valid and I will add some tests. Similarly, -mfpu=32 -mabi=lp64s.

How about -mfpu=64 -mabi=lp64f?

[tangyouling]

Do we need to add the definition of the __loongarch64 macro? gcc has that definition (but it is not recommended to use it, just for compatibility with previous code).
It has been used in some projects before, such as linux kernel, libunwind etc.

[SixWeining]

Add tests for more -mfpu and -mabi combinations.

[xen0n]

I've formatted the patch summary for you. Please familiarize yourself with Markdown syntax so next time your rendering would look better.

In D136413#3873317, @tangyouling wrote:

Do we need to add the definition of the __loongarch64 macro? gcc has that definition (but it is not recommended to use it, just for compatibility with previous code).
It has been used in some projects before, such as linux kernel, libunwind etc.

For minimum churn and hassle, I'm afraid we have to add it for now. I didn't realize [[ https://github.com/loongson/LoongArch-Documentation/pull/28#discussion_r917367794 | it's the equivalent of __mips64 ]] until too much software has been ported with it...

[SixWeining]

I've formatted the patch summary for you. Please familiarize yourself with Markdown syntax so next time your rendering would look better.

Thanks. Got it.

For minimum churn and hassle, I'm afraid we have to add it for now. I didn't realize [[ https://github.com/loongson/LoongArch-Documentation/pull/28#discussion_r917367794 | it's the equivalent of __mips64 ]] until too much software has been ported with it...

OK.

[SixWeining]

Define __loongarch64 which is equal to __loongarch_grlen == 64.

[xen0n]

Amended the patch summary a little for grammatical fixes and mentioning __loongarch64 too.

[SixWeining]

In D136413#3895369, @xen0n wrote:

Amended the patch summary a little for grammatical fixes and mentioning __loongarch64 too.

Thanks.

BTW, this patch depends on D136146. Could you help to have a look? :)

clang/lib/Basic/Targets/LoongArch.h

 #include "llvm/Support/Compiler.h"
 
 namespace clang {
 namespace targets {
 
 class LLVM_LIBRARY_VISIBILITY LoongArchTargetInfo : public TargetInfo {
 protected:
   std::string ABI;
+  bool HasFeatureD;
+  bool HasFeatureF;
 
 public:
   LoongArchTargetInfo(const llvm::Triple &Triple, const TargetOptions &)
       : TargetInfo(Triple) {
+    HasFeatureD = false;
+    HasFeatureF = false;
     LongDoubleWidth = 128;
     LongDoubleAlign = 128;
     LongDoubleFormat = &llvm::APFloat::IEEEquad();
     SuitableAlign = 128;
     WCharType = SignedInt;
     WIntType = UnsignedInt;
   }
 
public:
 
   ArrayRef<TargetInfo::GCCRegAlias> getGCCRegAliases() const override;
 
   bool validateAsmConstraint(const char *&Name,
                              TargetInfo::ConstraintInfo &Info) const override;
   std::string convertConstraint(const char *&Constraint) const override;
 
   bool hasBitIntType() const override { return true; }
+
+  bool handleTargetFeatures(std::vector<std::string> &Features,
+                            DiagnosticsEngine &Diags) override;
 };
 
 class LLVM_LIBRARY_VISIBILITY LoongArch32TargetInfo
     : public LoongArchTargetInfo {
 public:
   LoongArch32TargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts)
       : LoongArchTargetInfo(Triple, Opts) {
     IntPtrType = SignedInt;

clang/lib/Basic/Targets/LoongArch.cpp

default:
     break;
   }
   return R;
 }
 
 void LoongArchTargetInfo::getTargetDefines(const LangOptions &Opts,
                                            MacroBuilder &Builder) const {
   Builder.defineMacro("__loongarch__");
-  // TODO: Define more macros.
+  unsigned GRLen = getRegisterWidth();
+  Builder.defineMacro("__loongarch_grlen", Twine(GRLen));
+  if (GRLen == 64)
+    Builder.defineMacro("__loongarch64");
+
+  if (HasFeatureD)
+    Builder.defineMacro("__loongarch_frlen", "64");
+  else if (HasFeatureF)
+    Builder.defineMacro("__loongarch_frlen", "32");
+  else
+    Builder.defineMacro("__loongarch_frlen", "0");
+
+  // TODO: define __loongarch_arch and __loongarch_tune.
+
+  StringRef ABI = getABI();
+  if (ABI == "lp64d" || ABI == "lp64f" || ABI == "lp64s")
+    Builder.defineMacro("__loongarch_lp64");
+
+  if (ABI == "lp64d" || ABI == "ilp32d") {
+    Builder.defineMacro("__loongarch_hard_float");
+    Builder.defineMacro("__loongarch_double_float");
+  } else if (ABI == "lp64f" || ABI == "ilp32f") {
+    Builder.defineMacro("__loongarch_hard_float");
+    Builder.defineMacro("__loongarch_single_float");
+  } else if (ABI == "lp64s" || ABI == "ilp32s") {
+    Builder.defineMacro("__loongarch_soft_float");
+  }
 }
 
 ArrayRef<Builtin::Info> LoongArchTargetInfo::getTargetBuiltins() const {
   // TODO: To be implemented in future.
   return {};
 }
+
+bool LoongArchTargetInfo::handleTargetFeatures(
+    std::vector<std::string> &Features, DiagnosticsEngine &Diags) {
+  for (const auto &Feature : Features) {
+    if (Feature == "+d" || Feature == "+f") {
+      // "d" implies "f".
+      HasFeatureF = true;
+      if (Feature == "+d") {
+        HasFeatureD = true;
+      }
+    }
+  }
+  return true;
+}

clang/test/Preprocessor/init-loongarch.c

 // LA32: #define __WCHAR_WIDTH__ 32
 // LA32: #define __WINT_MAX__ 4294967295U
 // LA32: #define __WINT_TYPE__ unsigned int
 // LA32: #define __WINT_UNSIGNED__ 1
 // LA32: #define __WINT_WIDTH__ 32
 // LA32-LINUX: #define __gnu_linux__ 1
 // LA32-LINUX: #define __linux 1
 // LA32-LINUX: #define __linux__ 1
+// LA32-NOT: #define __loongarch64 1
 // LA32: #define __loongarch__ 1
 // LA32-LINUX: #define __unix 1
 // LA32-LINUX: #define __unix__ 1
 // LA32-LINUX: #define linux 1
 // LA32-LINUX: #define unix 1
 
 // LA64: #define _LP64 1
 // LA64: #define __ATOMIC_ACQUIRE 2

 // LA64: #define __WCHAR_WIDTH__ 32
 // LA64: #define __WINT_MAX__ 4294967295U
 // LA64: #define __WINT_TYPE__ unsigned int
 // LA64: #define __WINT_UNSIGNED__ 1
 // LA64: #define __WINT_WIDTH__ 32
 // LA64-LINUX: #define __gnu_linux__ 1
 // LA64-LINUX: #define __linux 1
 // LA64-LINUX: #define __linux__ 1
+// LA64: #define __loongarch64 1
 // LA64: #define __loongarch__ 1
 // LA64-LINUX: #define __unix 1
 // LA64-LINUX: #define __unix__ 1
 // LA64-LINUX: #define linux 1
 // LA64-LINUX: #define unix 1
+
+
+/// Check various macros prefixed with "__loongarch_" in different cases.
+/// "__loongarch__"" is not listed here as it has been checked above.
+
+// RUN: %clang --target=loongarch32 -mfpu=64 -mabi=ilp32d -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU64-ILP32D %s
+// RUN: %clang --target=loongarch32 -mdouble-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU64-ILP32D %s
+// LA32-FPU64-ILP32D: __loongarch_double_float 1
+// LA32-FPU64-ILP32D-NEXT: __loongarch_frlen 64
+// LA32-FPU64-ILP32D-NEXT: __loongarch_grlen 32
+// LA32-FPU64-ILP32D-NEXT: __loongarch_hard_float 1
+// LA32-FPU64-ILP32D-NOT: __loongarch_lp64
+// LA32-FPU64-ILP32D-NOT: __loongarch_single_float
+// LA32-FPU64-ILP32D-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch32 -mfpu=64 -mabi=ilp32f -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU64-ILP32F %s
+// LA32-FPU64-ILP32F-NOT: __loongarch_double_float
+// LA32-FPU64-ILP32F: __loongarch_frlen 64
+// LA32-FPU64-ILP32F-NEXT: __loongarch_grlen 32
+// LA32-FPU64-ILP32F-NEXT: __loongarch_hard_float 1
+// LA32-FPU64-ILP32F-NOT: __loongarch_lp64
+// LA32-FPU64-ILP32F-NEXT: __loongarch_single_float 1
+// LA32-FPU64-ILP32F-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch32 -mfpu=64 -mabi=ilp32s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU64-ILP32S %s
+// LA32-FPU64-ILP32S-NOT: __loongarch_double_float
+// LA32-FPU64-ILP32S: __loongarch_frlen 64
+// LA32-FPU64-ILP32S-NEXT: __loongarch_grlen 32
+// LA32-FPU64-ILP32S-NOT: __loongarch_hard_float
+// LA32-FPU64-ILP32S-NOT: __loongarch_lp64
+// LA32-FPU64-ILP32S-NOT: __loongarch_single_float
+// LA32-FPU64-ILP32S-NEXT: __loongarch_soft_float 1
+
+// RUN: %clang --target=loongarch32 -mfpu=32 -mabi=ilp32f -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU32-ILP32F %s
+// RUN: %clang --target=loongarch32 -msingle-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU32-ILP32F %s
+// LA32-FPU32-ILP32F-NOT: __loongarch_double_float
+// LA32-FPU32-ILP32F: __loongarch_frlen 32
+// LA32-FPU32-ILP32F-NEXT: __loongarch_grlen 32
+// LA32-FPU32-ILP32F-NEXT: __loongarch_hard_float 1
+// LA32-FPU32-ILP32F-NOT: __loongarch_lp64
+// LA32-FPU32-ILP32F-NEXT: __loongarch_single_float 1
+// LA32-FPU32-ILP32F-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch32 -mfpu=32 -mabi=ilp32s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU32-ILP32S %s
+// LA32-FPU32-ILP32S-NOT: __loongarch_double_float
+// LA32-FPU32-ILP32S: __loongarch_frlen 32
+// LA32-FPU32-ILP32S-NEXT: __loongarch_grlen 32
+// LA32-FPU32-ILP32S-NOT: __loongarch_hard_float
+// LA32-FPU32-ILP32S-NOT: __loongarch_lp64
+// LA32-FPU32-ILP32S-NOT: __loongarch_single_float
+// LA32-FPU32-ILP32S-NEXT: __loongarch_soft_float 1
+
+// RUN: %clang --target=loongarch32 -mfpu=0 -mabi=ilp32s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU0-ILP32S %s
+// RUN: %clang --target=loongarch32 -mfpu=none -mabi=ilp32s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU0-ILP32S %s
+// RUN: %clang --target=loongarch32 -msoft-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA32-FPU0-ILP32S %s
+// LA32-FPU0-ILP32S-NOT: __loongarch_double_float
+// LA32-FPU0-ILP32S: __loongarch_frlen 0
+// LA32-FPU0-ILP32S-NEXT: __loongarch_grlen 32
+// LA32-FPU0-ILP32S-NOT: __loongarch_hard_float
+// LA32-FPU0-ILP32S-NOT: __loongarch_lp64
+// LA32-FPU0-ILP32S-NOT: __loongarch_single_float
+// LA32-FPU0-ILP32S-NEXT: __loongarch_soft_float 1
+
+// RUN: %clang --target=loongarch64 -mfpu=64 -mabi=lp64d -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU64-LP64D %s
+// RUN: %clang --target=loongarch64 -mdouble-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU64-LP64D %s
+// LA64-FPU64-LP64D: __loongarch_double_float 1
+// LA64-FPU64-LP64D-NEXT: __loongarch_frlen 64
+// LA64-FPU64-LP64D-NEXT: __loongarch_grlen 64
+// LA64-FPU64-LP64D-NEXT: __loongarch_hard_float 1
+// LA64-FPU64-LP64D-NEXT: __loongarch_lp64 1
+// LA64-FPU64-LP64D-NOT: __loongarch_single_float
+// LA64-FPU64-LP64D-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch64 -mfpu=64 -mabi=lp64f -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU64-LP64F %s
+// LA64-FPU64-LP64F-NOT: __loongarch_double_float
+// LA64-FPU64-LP64F: __loongarch_frlen 64
+// LA64-FPU64-LP64F-NEXT: __loongarch_grlen 64
+// LA64-FPU64-LP64F-NEXT: __loongarch_hard_float 1
+// LA64-FPU64-LP64F-NEXT: __loongarch_lp64 1
+// LA64-FPU64-LP64F-NEXT: __loongarch_single_float 1
+// LA64-FPU64-LP64F-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch64 -mfpu=64 -mabi=lp64s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU64-LP64S %s
+// LA64-FPU64-LP64S-NOT: __loongarch_double_float
+// LA64-FPU64-LP64S: __loongarch_frlen 64
+// LA64-FPU64-LP64S-NEXT: __loongarch_grlen 64
+// LA64-FPU64-LP64S-NOT: __loongarch_hard_float
+// LA64-FPU64-LP64S-NEXT: __loongarch_lp64 1
+// LA64-FPU64-LP64S-NOT: __loongarch_single_float
+// LA64-FPU64-LP64S-NEXT: __loongarch_soft_float 1
+
+// RUN: %clang --target=loongarch64 -mfpu=32 -mabi=lp64f -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU32-LP64F %s
+// RUN: %clang --target=loongarch64 -msingle-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU32-LP64F %s
+// LA64-FPU32-LP64F-NOT: __loongarch_double_float
+// LA64-FPU32-LP64F: __loongarch_frlen 32
+// LA64-FPU32-LP64F-NEXT: __loongarch_grlen 64
+// LA64-FPU32-LP64F-NEXT: __loongarch_hard_float 1
+// LA64-FPU32-LP64F-NEXT: __loongarch_lp64 1
+// LA64-FPU32-LP64F-NEXT: __loongarch_single_float 1
+// LA64-FPU32-LP64F-NOT: __loongarch_soft_float
+
+// RUN: %clang --target=loongarch64 -mfpu=32 -mabi=lp64s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU32-LP64S %s
+// LA64-FPU32-LP64S-NOT: __loongarch_double_float
+// LA64-FPU32-LP64S: __loongarch_frlen 32
+// LA64-FPU32-LP64S-NEXT: __loongarch_grlen 64
+// LA64-FPU32-LP64S-NOT: __loongarch_hard_float
+// LA64-FPU32-LP64S-NEXT: __loongarch_lp64 1
+// LA64-FPU32-LP64S-NOT: __loongarch_single_float
+// LA64-FPU32-LP64S-NEXT: __loongarch_soft_float 1
+
+// RUN: %clang --target=loongarch64 -mfpu=0 -mabi=lp64s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU0-LP64S %s
+// RUN: %clang --target=loongarch64 -mfpu=none -mabi=lp64s -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU0-LP64S %s
+// RUN: %clang --target=loongarch64 -msoft-float -x c -E -dM %s -o - \
+// RUN:   | grep __loongarch_ | FileCheck --check-prefix=LA64-FPU0-LP64S %s
+// LA64-FPU0-LP64S-NOT: __loongarch_double_float
+// LA64-FPU0-LP64S: __loongarch_frlen 0
+// LA64-FPU0-LP64S-NEXT: __loongarch_grlen 64
+// LA64-FPU0-LP64S-NOT: __loongarch_hard_float
+// LA64-FPU0-LP64S-NEXT: __loongarch_lp64 1
+// LA64-FPU0-LP64S-NOT: __loongarch_single_float
+// LA64-FPU0-LP64S-NEXT: __loongarch_soft_float 1