This is an archive of the discontinued LLVM Phabricator instance.

Differential D11466

Correct x86_64 Android fp128 mangled name
ClosedPublic

Authored by chh on Jul 23 2015, 10:21 AM.

Details

Reviewers
majnemer
echristo
rnk
Commits
rG0997fb43b33c: Merging r243133: --------------------------------------------------------------…
rG0b0eeaaaf6a5: Correct x86_64 Android fp128 mangled name
rC243133: Correct x86_64 Android fp128 mangled name
rL243133: Correct x86_64 Android fp128 mangled name
Summary

These changes are for Android x86_64 targets to be compatible with current Android g++.




https://llvm.org/bugs/show_bug.cgi?id=23897






    

  • Use 'g' and 'Cg' for "long double" and "long double _Complex" mangled type names.
    • 

Diff Detail
Event Timeline
chh updated this revision to Diff 30501.Jul 23 2015, 10:21 AM
chh retitled this revision from  to Correct x86_64 Android fp128 mangled name.
chh updated this object.
chh added reviewers: majnemer, echristo.
chh added subscribers: cfe-commits, davidxl, rnk and 4 others.
rnk added inline comments.Jul 23 2015, 10:26 AM
test/CodeGen/x86-long-double.cpp


1–12 ↗(On Diff #30501)
Don't use %clang in CodeGen tests.
majnemer added inline comments.Jul 23 2015, 10:35 AM
test/CodeGen/x86-long-double.cpp


5–6 ↗(On Diff #30501)
This test is called x86-long-double.cpp but you have powerpc tests here.
chh updated this revision to Diff 30502.Jul 23 2015, 10:42 AM
Use %clang_cc1 in unit test.
chh updated this revision to Diff 30504.Jul 23 2015, 10:49 AM
Change unit test name to long_double_fp128.
rnk added a comment.Jul 23 2015, 10:56 AM
Thanks, this seems almost done. Sorry for missing the -O on the first round.
test/CodeGen/long_double_fp128.cpp


1
I don't think these tests need optimizations (-O).
echristo added inline comments.Jul 23 2015, 10:58 AM
test/CodeGen/long_double_fp128.cpp


1
If they do there's a different problem :)
chh updated this revision to Diff 30505.Jul 23 2015, 11:09 AM
Remove -O from unit tests.
rnk accepted this revision.Jul 23 2015, 1:42 PM
rnk added a reviewer: rnk.
lgtm
This revision is now accepted and ready to land.Jul 23 2015, 1:42 PM
chh added a comment.Jul 23 2015, 3:08 PM
Thanks for the review and approval.



I will commit this patch when I get my commit access request approved.

If anyone can help me commit this patch, please go ahead.

Thank you.
Closed by commit rL243133: Correct x86_64 Android fp128 mangled name (authored by chh).  ·  Explain WhyJul 24 2015, 11:13 AM
This revision was automatically updated to reflect the committed changes.
Revision Contents
PathSize
lib/
Basic/
4 lines
test/
CodeGen/
22 lines
Diff 30504
lib/Basic/Targets.cpp
Show First 20 Lines • Show All 7,008 Lines • ▼ Show 20 Lines
namespace {
namespace {

// x86_64 Android target
// x86_64 Android target

class AndroidX86_64TargetInfo : public LinuxTargetInfo<X86_64TargetInfo> {
class AndroidX86_64TargetInfo : public LinuxTargetInfo<X86_64TargetInfo> {

public:
public:

  AndroidX86_64TargetInfo(const llvm::Triple &Triple)
  AndroidX86_64TargetInfo(const llvm::Triple &Triple)

      : LinuxTargetInfo<X86_64TargetInfo>(Triple) {
      : LinuxTargetInfo<X86_64TargetInfo>(Triple) {

    LongDoubleFormat = &llvm::APFloat::IEEEquad;
    LongDoubleFormat = &llvm::APFloat::IEEEquad;

  }
  }



  bool useFloat128ManglingForLongDouble() const override {

    return true;

  }

};
};

} // end anonymous namespace
} // end anonymous namespace







//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//

// Driver code
// Driver code

//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//




▲ Show 20 Lines • Show All 468 Lines • Show Last 20 Lines
test/CodeGen/long_double_fp128.cpp
  • This file was added.
// RUN: %clang_cc1 -triple x86_64-linux-android -emit-llvm -O -o - %s \

rnkUnsubmitted
Not Done
ReplyInline Actions
I don't think these tests need optimizations (-O).
rnk: I don't think these tests need optimizations (-O).
echristoUnsubmitted
Not Done
ReplyInline Actions
If they do there's a different problem :)
echristo: If they do there's a different problem :)
// RUN:    | FileCheck %s --check-prefix=A64

// RUN: %clang_cc1 -triple x86_64-linux-gnu -emit-llvm -O -o - %s \

// RUN:    | FileCheck %s --check-prefix=G64

// RUN: %clang_cc1 -triple powerpc64-linux-gnu -emit-llvm -O -o - %s \

// RUN:    | FileCheck %s --check-prefix=P64

// RUN: %clang_cc1 -triple i686-linux-android -emit-llvm -O -o - %s \

// RUN:    | FileCheck %s --check-prefix=A32

// RUN: %clang_cc1 -triple i686-linux-gnu -emit-llvm -O -o - %s \

// RUN:    | FileCheck %s --check-prefix=G32

// RUN: %clang_cc1 -triple powerpc-linux-gnu -emit-llvm -O -o - %s \

// RUN:    | FileCheck %s --check-prefix=P32



// Check mangled name of long double.

// Android's gcc and llvm use fp128 for long double.

void test(long, float, double, long double, long double _Complex) { }

// A64:  define void @_Z4testlfdgCg(i64, float, double, fp128, { fp128, fp128 }*

// G64:  define void @_Z4testlfdeCe(i64, float, double, x86_fp80, { x86_fp80, x86_fp80 }*

// P64:  define void @_Z4testlfdgCg(i64, float, double, ppc_fp128, ppc_fp128 {{.*}}, ppc_fp128

// A32:  define void @_Z4testlfdeCe(i32, float, double, double, { double, double }*

// G32:  define void @_Z4testlfdeCe(i32, float, double, x86_fp80, { x86_fp80, x86_fp80 }*

// P32:  define void @_Z4testlfdgCg(i32, float, double, ppc_fp128, { ppc_fp128, ppc_fp128 }*