ARM: Stop setting sincos_stret calling convention

[arsenm]

This was going out of its way to explicitly mark these as
ARM_AAPCS_VFP. This has been explicitly set since 8b40366,
where the commit message states that "sincos" (not sincos_stret)
has a special calling convention. However, that commit also sets
the calling convention for all libcalls to ARM_AAPCS_VFP, and
getEffectiveCallingConv returns the same for CCC anyway in tests
using isWatchABI triples.

The net result of this appears to be a change in behavior when
using -float-abi=soft with isWatchABI, which have no tests so
I assume this is a theoretical combination.

If I assert

  if (getTargetMachine().getTargetTriple().isWatchABI()) {
    assert(!useSoftFloat());
    assert(getEffectiveCallingConv(CallingConv::C, false) == CallingConv::ARM_AAPCS_VFP);
  }

Only 2 tests fail the second condition, which look like copy paste accidents
using v7k triples with linux and only needed a filler triple. This is a consequence
of strangely using the target architecture in place of the OS ABI check,
as was done in 042a6c1

[arsenm]

• ARM: Stop setting sincos_stret calling convention #147457 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-llvm-ir

@llvm/pr-subscribers-backend-arm

Author: Matt Arsenault (arsenm)

Changes

This was going out of its way to explicitly mark these as
ARM_AAPCS_VFP. This has been explicitly set since 8b40366,
where the commit message states that "sincos" (not sincos_stret)
has a special calling convention. However, that commit also sets
the calling convention for all libcalls to ARM_AAPCS_VFP, and
getEffectiveCallingConv returns the same for CCC anyway in tests
using isWatchABI triples.

The net result of this appears to be a change in behavior when
using -float-abi=soft with isWatchABI, which have no tests so
I assume this is a theoretical combination.

If I assert
if (getTargetMachine().getTargetTriple().isWatchABI()) {
assert(!useSoftFloat());
assert(getEffectiveCallingConv(CallingConv::C, false) == CallingConv::ARM_AAPCS_VFP);
}

Only 2 tests fail the second condition, which look like copy paste accidents
using v7k triples with linux and only needed a filler triple. This is a consequence
of strangely using the target architecture in place of the OS ABI check,
as was done in 042a6c1

---

Full diff: https://github.com/llvm/llvm-project/pull/147457.diff

1 Files Affected:

• (modified) llvm/lib/IR/RuntimeLibcalls.cpp (-6)

diff --git a/llvm/lib/IR/RuntimeLibcalls.cpp b/llvm/lib/IR/RuntimeLibcalls.cpp
index 2cbc2cdb79685..45b778d61d8e0 100644
--- a/llvm/lib/IR/RuntimeLibcalls.cpp
+++ b/llvm/lib/IR/RuntimeLibcalls.cpp
@@ -365,12 +365,6 @@ void RuntimeLibcallsInfo::initLibcalls(const Triple &TT,
     if (darwinHasSinCosStret(TT)) {
       setLibcallImpl(RTLIB::SINCOS_STRET_F32, RTLIB::__sincosf_stret);
       setLibcallImpl(RTLIB::SINCOS_STRET_F64, RTLIB::__sincos_stret);
-      if (TT.isWatchABI()) {
-        setLibcallImplCallingConv(RTLIB::__sincosf_stret,
-                                  CallingConv::ARM_AAPCS_VFP);
-        setLibcallImplCallingConv(RTLIB::__sincos_stret,
-                                  CallingConv::ARM_AAPCS_VFP);
-      }
     }
 
     if (darwinHasExp10(TT)) {

[arsenm]

ping

[efriedma-quic]

LGTM

Ideally, someone from Apple would look at this... but it's unlikely anyone cares; 32-bit WatchOS is dead.