clang: Add __builtin_elementwise_rint and nearbyint

[arsenm]

Summary

These are basically the same thing and only differ for strictfp,
so add both for future proofing. Note all the elementwise functions are
currently broken for strictfp, and use non-constrained ops. Add a test
that demonstrates this, but doesn't attempt to fix it.

[sepavloff]

clang/docs/LanguageExtensions.rst

656

"halfway'" is superfluous here.

657

nearbyint should not rise only inexact exception. Other exceptions, like invalid are allowed.

662

"halfway'" is superfluous here.

clang/test/CodeGen/strictfp-elementwise-bulitins.cpp

2

For X86 -fexperimental-strict-floating-point is not needed.

13–14

Why vector type is bitcasted to scalar? The function must return <2 x float>, no?

[arsenm]

clang/test/CodeGen/strictfp-elementwise-bulitins.cpp

13–14

I assume this is some ABI thing. It doesn't happen for other targets and the actual code is correct. I've avoided it by using float4 instead

[bob80905]

clang/docs/LanguageExtensions.rst

656

typo with the word current.

clang/docs/ReleaseNotes.rst

234–239

Don't think this for is necessary or intended, and it appears in the nearbyint description too.

[arsenm]

Keep fixing documentation

[bob80905]

clang/docs/LanguageExtensions.rst

656

Nit: I don't think a "current rounding direction" exists, and personally don't see the significance of the second half of this sentence.

clang/lib/Sema/SemaChecking.cpp

18583

I don't believe you intended to remove all this code in your latest update, did you?

[arsenm]

clang/docs/LanguageExtensions.rst

656

It does exist, we just happen to ignore floating point environment most of the time. The current state of things is we have 3 LLVM intrinsics with the same behavior (nearbyint, rint and roundeven). The distinction only exists with fenv access / strictfp

clang/lib/Sema/SemaChecking.cpp

18583

I didn't delete any code?

[arsenm]

clang/lib/Sema/SemaChecking.cpp

18583

Are you looking at a history diff? I see nothing deleted here

[bob80905]

clang/lib/Sema/SemaChecking.cpp

18583

I am unsure what I was seeing prior, but I no longer see any deleted code, you can count this comment as resolved.

[bob80905]

clang/test/CodeGen/strictfp-elementwise-bulitins.cpp

191

I noticed a missing pattern here. Did you mean to write canonicalize instead of TRUNC?

[bob80905]

I'd like to ask, is there a reason why there isn't a test for these new builtins under SemaCXX/builtins_elementwise_math.cpp?

[arsenm]

clang/test/CodeGen/strictfp-elementwise-bulitins.cpp

191

This is just the generated check from the IR names. It doesn't matter, but there's a pre-existing "bug" where the debug value name is trunc for canonicalize.

[arsenm]

Rebase on fix for wrong debug name

[arsenm]

SemaCXX test

[bob80905]

This comment was removed by bob80905.

[bob80905]

Should there be tests added for usage of scalable vector types for RISC-V / AArch64?
I typically have added such tests in the past, as shown here:
https://reviews.llvm.org/D135011

Also, would you be able to add something at the top of the strictfp file that states the purpose? Like "testing that we don't use the constrained intrinsics"?

[arsenm]

In D153233#4443000, @bob80905 wrote:

Should there be tests added for usage of scalable vector types for RISC-V / AArch64?
I typically have added such tests in the past, as shown here:
https://reviews.llvm.org/D135011

Test coverage seems to be spotty for them. I think a separate patch to get consistent SVE testing of this should be separate

Also, would you be able to add something at the top of the strictfp file that states the purpose? Like "testing that we don't use the constrained intrinsics"?

Done

[arsenm]

Add fixme to test

[bob80905]

LGTM

[arsenm]

9d84f8dc948b3188fdec9a1a080eb6d845c2082d

clang/docs/LanguageExtensions.rst

T __builtin_elementwise_roundeven(T x)      round x to the nearest integer value in floating point format,   floating point types
                                              that is an even integer), regardless of the current rounding
                                              direction.
  T __builtin_elementwise_round(T x)          round x to the nearest  integer value in floating point format,      floating point types
                                              rounding halfway cases away from zero, regardless of the
                                              current rounding direction. May raise floating-point
                                              exceptions.
  T __builtin_elementwise_trunc(T x)          return the integral value nearest to but no larger in            floating point types
                                              magnitude than x
+
+  T __builtin_elementwise_nearbyint(T x)     round x to the nearest  integer value in floating point format,      floating point types
+                                             rounding halfway according to the current current rounding
+                                             direction. May not raise floating-point exceptions. This is
+                                             treated the same as ``__builtin_elementwise_rint`` unless
+                                             :ref:`FENV_ACCESS is enabled. <floating-point-environment>`.
+
+ T __builtin_elementwise_rint(T x)           round x to the nearest  integer value in floating point format,      floating point types
+                                             rounding halfway cases according to the current rounding
+                                             direction. May raise floating-point exceptions. This is treated
+                                             the same as ``__builtin_elementwise_nearbyint`` unless
+                                             :ref:`FENV_ACCESS is enabled. <floating-point-environment>`.
+
  T __builtin_elementwise_canonicalize(T x)   return the platform specific canonical encoding                  floating point types
                                              of a floating-point number
  T __builtin_elementwise_copysign(T x, T y)  return the magnitude of x with the sign of y.                    floating point types
  T __builtin_elementwise_max(T x, T y)       return x or y, whichever is larger                               integer and floating point types
  T __builtin_elementwise_min(T x, T y)       return x or y, whichever is smaller                              integer and floating point types
  T __builtin_elementwise_add_sat(T x, T y)   return the sum of x and y, clamped to the range of               integer types
                                              representable values for the signed/unsigned integer type.
  T __builtin_elementwise_sub_sat(T x, T y)   return the difference of x and y, clamped to the range of        integer types

clang/docs/ReleaseNotes.rst

- Clang now support ``__builtin_FUNCSIG()`` which returns the same information
   This fixes (`#58951 <https://github.com/llvm/llvm-project/issues/58951>`_).
 - Clang now supports the `NO_COLOR <https://no-color.org/>`_ environment
   variable as a way to disable color diagnostics.
 - Clang now supports ``__builtin_isfpclass``, which checks if the specified
   floating-point value falls into any of the specified data classes.
 - Added ``__builtin_elementwise_round`` for  builtin for floating
   point types. This allows access to ``llvm.round`` for
   arbitrary floating-point and vector of floating-point types.
-
+- Added ``__builtin_elementwise_rint`` for  builtin for floating
+  point types. This allows access to ``llvm.rint`` for
+  arbitrary floating-point and vector of floating-point types.
+- Added ``__builtin_elementwise_nearbyint`` for  builtin for floating
+  point types. This allows access to ``llvm.nearbyint`` for
+  arbitrary floating-point and vector of floating-point types.
 
 New Compiler Flags
 ------------------
 - The flag ``-std=c++23`` has been added. This behaves the same as the existing
   flag ``-std=c++2b``.
 - ``-dumpdir`` has been implemented to specify auxiliary and dump output
   filenames for features like ``-gsplit-dwarf``.
 - ``-fcaret-diagnostics-max-lines=`` has been added as a driver options, which

clang/include/clang/Basic/Builtins.def

 BUILTIN(__builtin_elementwise_exp, "v.", "nct")
 BUILTIN(__builtin_elementwise_exp2, "v.", "nct")
 BUILTIN(__builtin_elementwise_floor, "v.", "nct")
 BUILTIN(__builtin_elementwise_log, "v.", "nct")
 BUILTIN(__builtin_elementwise_log2, "v.", "nct")
 BUILTIN(__builtin_elementwise_log10, "v.", "nct")
 BUILTIN(__builtin_elementwise_roundeven, "v.", "nct")
 BUILTIN(__builtin_elementwise_round, "v.", "nct")
+BUILTIN(__builtin_elementwise_rint, "v.", "nct")
+BUILTIN(__builtin_elementwise_nearbyint, "v.", "nct")
 BUILTIN(__builtin_elementwise_sin, "v.", "nct")
 BUILTIN(__builtin_elementwise_trunc, "v.", "nct")
 BUILTIN(__builtin_elementwise_canonicalize, "v.", "nct")
 BUILTIN(__builtin_elementwise_copysign, "v.", "nct")
 BUILTIN(__builtin_elementwise_fma, "v.", "nct")
 BUILTIN(__builtin_elementwise_add_sat, "v.", "nct")
 BUILTIN(__builtin_elementwise_sub_sat, "v.", "nct")
 BUILTIN(__builtin_reduce_max, "v.", "nct")

clang/lib/CodeGen/CGBuiltin.cpp

case Builtin::BI__builtin_elementwise_floor:
     return RValue::get(
         emitUnaryBuiltin(*this, E, llvm::Intrinsic::floor, "elt.floor"));
   case Builtin::BI__builtin_elementwise_roundeven:
     return RValue::get(emitUnaryBuiltin(*this, E, llvm::Intrinsic::roundeven,
                                         "elt.roundeven"));
   case Builtin::BI__builtin_elementwise_round:
     return RValue::get(emitUnaryBuiltin(*this, E, llvm::Intrinsic::round,
                                         "elt.round"));
+  case Builtin::BI__builtin_elementwise_rint:
+    return RValue::get(emitUnaryBuiltin(*this, E, llvm::Intrinsic::rint,
+                                        "elt.rint"));
+  case Builtin::BI__builtin_elementwise_nearbyint:
+    return RValue::get(emitUnaryBuiltin(*this, E, llvm::Intrinsic::nearbyint,
+                                        "elt.nearbyint"));
   case Builtin::BI__builtin_elementwise_sin:
     return RValue::get(
         emitUnaryBuiltin(*this, E, llvm::Intrinsic::sin, "elt.sin"));
 
   case Builtin::BI__builtin_elementwise_trunc:
     return RValue::get(
         emitUnaryBuiltin(*this, E, llvm::Intrinsic::trunc, "elt.trunc"));
   case Builtin::BI__builtin_elementwise_canonicalize:

clang/lib/Sema/SemaChecking.cpp

#include "clang/Basic/Builtins.def"
   case Builtin::BI__builtin_elementwise_exp:
   case Builtin::BI__builtin_elementwise_exp2:
   case Builtin::BI__builtin_elementwise_floor:
   case Builtin::BI__builtin_elementwise_log:
   case Builtin::BI__builtin_elementwise_log2:
   case Builtin::BI__builtin_elementwise_log10:
   case Builtin::BI__builtin_elementwise_roundeven:
   case Builtin::BI__builtin_elementwise_round:
+  case Builtin::BI__builtin_elementwise_rint:
+  case Builtin::BI__builtin_elementwise_nearbyint:
   case Builtin::BI__builtin_elementwise_sin:
   case Builtin::BI__builtin_elementwise_trunc:
   case Builtin::BI__builtin_elementwise_canonicalize: {
     if (PrepareBuiltinElementwiseMathOneArgCall(TheCall))
       return ExprError();
 
     QualType ArgTy = TheCall->getArg(0)->getType();
     if (checkFPMathBuiltinElementType(*this, TheCall->getArg(0)->getBeginLoc(),
void Sema::CheckTCBEnforcement(const SourceLocation CallExprLoc,
 
   // Go through the TCBs the caller is a part of and emit warnings if Caller
   // is in a TCB that the Callee is not.
   for (const auto *A : Caller->specific_attrs<EnforceTCBAttr>()) {
     StringRef CallerTCB = A->getTCBName();
     if (CalleeTCBs.count(CallerTCB) == 0) {
       this->Diag(CallExprLoc, diag::warn_tcb_enforcement_violation)
           << Callee << CallerTCB;
     }
   }
 }

clang/test/CodeGen/builtins-elementwise-math.c

void test_builtin_elementwise_round(float f1, float f2, double d1, double d2,
   // CHECK-NEXT: call double @llvm.round.f64(double [[D1]])
   d2 = __builtin_elementwise_round(d1);
 
   // CHECK:      [[VF1:%.+]] = load <4 x float>, ptr %vf1.addr, align 16
   // CHECK-NEXT: call <4 x float> @llvm.round.v4f32(<4 x float> [[VF1]])
   vf2 = __builtin_elementwise_round(vf1);
 }
 
+void test_builtin_elementwise_rint(float f1, float f2, double d1, double d2,
+                                   float4 vf1, float4 vf2) {
+  // CHECK-LABEL: define void @test_builtin_elementwise_rint(
+  // CHECK:      [[F1:%.+]] = load float, ptr %f1.addr, align 4
+  // CHECK-NEXT:  call float @llvm.rint.f32(float [[F1]])
+  f2 = __builtin_elementwise_rint(f1);
+
+  // CHECK:      [[D1:%.+]] = load double, ptr %d1.addr, align 8
+  // CHECK-NEXT: call double @llvm.rint.f64(double [[D1]])
+  d2 = __builtin_elementwise_rint(d1);
+
+  // CHECK:      [[VF1:%.+]] = load <4 x float>, ptr %vf1.addr, align 16
+  // CHECK-NEXT: call <4 x float> @llvm.rint.v4f32(<4 x float> [[VF1]])
+  vf2 = __builtin_elementwise_rint(vf1);
+}
+
+void test_builtin_elementwise_nearbyint(float f1, float f2, double d1, double d2,
+                                        float4 vf1, float4 vf2) {
+  // CHECK-LABEL: define void @test_builtin_elementwise_nearbyint(
+  // CHECK:      [[F1:%.+]] = load float, ptr %f1.addr, align 4
+  // CHECK-NEXT:  call float @llvm.nearbyint.f32(float [[F1]])
+  f2 = __builtin_elementwise_nearbyint(f1);
+
+  // CHECK:      [[D1:%.+]] = load double, ptr %d1.addr, align 8
+  // CHECK-NEXT: call double @llvm.nearbyint.f64(double [[D1]])
+  d2 = __builtin_elementwise_nearbyint(d1);
+
+  // CHECK:      [[VF1:%.+]] = load <4 x float>, ptr %vf1.addr, align 16
+  // CHECK-NEXT: call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[VF1]])
+  vf2 = __builtin_elementwise_nearbyint(vf1);
+}
+
 void test_builtin_elementwise_sin(float f1, float f2, double d1, double d2,
                                   float4 vf1, float4 vf2) {
   // CHECK-LABEL: define void @test_builtin_elementwise_sin(
   // CHECK:      [[F1:%.+]] = load float, ptr %f1.addr, align 4
   // CHECK-NEXT:  call float @llvm.sin.f32(float [[F1]])
   f2 = __builtin_elementwise_sin(f1);
 
   // CHECK:      [[D1:%.+]] = load double, ptr %d1.addr, align 8

clang/test/CodeGen/strictfp-elementwise-bulitins.cpp

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// RUN: %clang_cc1 -triple x86_64-linux-gnu -fexperimental-strict-floating-point -frounding-math -ffp-exception-behavior=strict -O2 -emit-llvm -o - %s | FileCheck %s
+
+typedef float float2 __attribute__((ext_vector_type(2)));
+
+// Sanity check we're getting constrained ops for a non-builtin.
+// CHECK-LABEL: define dso_local noundef double @_Z11strict_faddDv2_fS_
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]], double noundef [[B_COERCE:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double [[B_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ADD:%.*]] = tail call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> [[TMP0]], <2 x float> [[TMP1]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR4:[0-9]+]]
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[ADD]] to double
+// CHECK-NEXT:    ret double [[TMP2]]
+//
+float2 strict_fadd(float2 a, float2 b) {
+  return a + b;
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_absDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_ABS:%.*]] = tail call <2 x float> @llvm.fabs.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_ABS]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_abs(float2 a) {
+  return __builtin_elementwise_abs(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_maxDv2_fS_
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]], double noundef [[B_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double [[B_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_MAX:%.*]] = tail call <2 x float> @llvm.maxnum.v2f32(<2 x float> [[TMP0]], <2 x float> [[TMP1]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[ELT_MAX]] to double
+// CHECK-NEXT:    ret double [[TMP2]]
+//
+float2 strict_elementwise_max(float2 a, float2 b) {
+  return __builtin_elementwise_max(a, b);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_minDv2_fS_
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]], double noundef [[B_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double [[B_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_MIN:%.*]] = tail call <2 x float> @llvm.minnum.v2f32(<2 x float> [[TMP0]], <2 x float> [[TMP1]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[ELT_MIN]] to double
+// CHECK-NEXT:    ret double [[TMP2]]
+//
+float2 strict_elementwise_min(float2 a, float2 b) {
+  return __builtin_elementwise_min(a, b);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z23strict_elementwise_ceilDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_CEIL:%.*]] = tail call <2 x float> @llvm.ceil.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_CEIL]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_ceil(float2 a) {
+  return __builtin_elementwise_ceil(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_cosDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_COS:%.*]] = tail call <2 x float> @llvm.cos.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_COS]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_cos(float2 a) {
+  return __builtin_elementwise_cos(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_expDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_EXP:%.*]] = tail call <2 x float> @llvm.exp.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_EXP]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_exp(float2 a) {
+  return __builtin_elementwise_exp(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z23strict_elementwise_exp2Dv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_EXP2:%.*]] = tail call <2 x float> @llvm.exp2.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_EXP2]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_exp2(float2 a) {
+  return __builtin_elementwise_exp2(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z24strict_elementwise_floorDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_FLOOR:%.*]] = tail call <2 x float> @llvm.floor.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_FLOOR]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_floor(float2 a) {
+  return __builtin_elementwise_floor(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_logDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_LOG:%.*]] = tail call <2 x float> @llvm.log.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_LOG]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_log(float2 a) {
+  return __builtin_elementwise_log(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z23strict_elementwise_log2Dv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_LOG2:%.*]] = tail call <2 x float> @llvm.log2.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_LOG2]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_log2(float2 a) {
+  return __builtin_elementwise_log2(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z24strict_elementwise_log10Dv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_LOG2:%.*]] = tail call <2 x float> @llvm.log2.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_LOG2]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_log10(float2 a) {
+  return __builtin_elementwise_log2(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z28strict_elementwise_roundevenDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_ROUNDEVEN:%.*]] = tail call <2 x float> @llvm.roundeven.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_ROUNDEVEN]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_roundeven(float2 a) {
+  return __builtin_elementwise_roundeven(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z24strict_elementwise_roundDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_ROUND:%.*]] = tail call <2 x float> @llvm.round.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_ROUND]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_round(float2 a) {
+  return __builtin_elementwise_round(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z23strict_elementwise_rintDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_RINT:%.*]] = tail call <2 x float> @llvm.rint.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_RINT]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_rint(float2 a) {
+  return __builtin_elementwise_rint(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z28strict_elementwise_nearbyintDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_NEARBYINT:%.*]] = tail call <2 x float> @llvm.nearbyint.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_NEARBYINT]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_nearbyint(float2 a) {
+  return __builtin_elementwise_nearbyint(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_sinDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_SIN:%.*]] = tail call <2 x float> @llvm.sin.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_SIN]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_sin(float2 a) {
+  return __builtin_elementwise_sin(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z24strict_elementwise_truncDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_TRUNC:%.*]] = tail call <2 x float> @llvm.trunc.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_TRUNC]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_trunc(float2 a) {
+  return __builtin_elementwise_trunc(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z31strict_elementwise_canonicalizeDv2_f
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[ELT_TRUNC:%.*]] = tail call <2 x float> @llvm.canonicalize.v2f32(<2 x float> [[TMP0]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[ELT_TRUNC]] to double
+// CHECK-NEXT:    ret double [[TMP1]]
+//
+float2 strict_elementwise_canonicalize(float2 a) {
+  return __builtin_elementwise_canonicalize(a);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z27strict_elementwise_copysignDv2_fS_
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]], double noundef [[B_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double [[B_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x float> @llvm.copysign.v2f32(<2 x float> [[TMP0]], <2 x float> [[TMP1]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x float> [[TMP2]] to double
+// CHECK-NEXT:    ret double [[TMP3]]
+//
+float2 strict_elementwise_copysign(float2 a, float2 b) {
+  return __builtin_elementwise_copysign(a, b);
+}
+
+// CHECK-LABEL: define dso_local noundef double @_Z22strict_elementwise_fmaDv2_fS_S_
+// CHECK-SAME: (double noundef [[A_COERCE:%.*]], double noundef [[B_COERCE:%.*]], double noundef [[C_COERCE:%.*]]) local_unnamed_addr #[[ATTR2]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[A_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double [[B_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double [[C_COERCE]] to <2 x float>
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <2 x float> @llvm.fma.v2f32(<2 x float> [[TMP0]], <2 x float> [[TMP1]], <2 x float> [[TMP2]]) #[[ATTR4]]
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to double
+// CHECK-NEXT:    ret double [[TMP4]]
+//
+float2 strict_elementwise_fma(float2 a, float2 b, float2 c) {
+  return __builtin_elementwise_fma(a, b, c);
+}
+

clang/test/Sema/builtins-elementwise-math.c

void test_builtin_elementwise_roundeven(int i, float f, double d, float4 v, int3 iv, unsigned u, unsigned4 uv) {
   u = __builtin_elementwise_roundeven(u);
   // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned int')}}
 
   uv = __builtin_elementwise_roundeven(uv);
   // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned4' (vector of 4 'unsigned int' values))}}
 }
 
 void test_builtin_elementwise_round(int i, float f, double d, float4 v, int3 iv, unsigned u, unsigned4 uv) {
-
   struct Foo s = __builtin_elementwise_round(f);
   // expected-error@-1 {{initializing 'struct Foo' with an expression of incompatible type 'float'}}
 
   i = __builtin_elementwise_round();
   // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
 
   i = __builtin_elementwise_round(i);
   // expected-error@-1 {{1st argument must be a floating point type (was 'int')}}
 
   i = __builtin_elementwise_round(f, f);
   // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
 
   u = __builtin_elementwise_round(u);
   // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned int')}}
 
   uv = __builtin_elementwise_round(uv);
   // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned4' (vector of 4 'unsigned int' values))}}
 
   // FIXME: Error should not mention integer
   _Complex float c1, c2;
   c1 = __builtin_elementwise_round(c1);
   // expected-error@-1 {{1st argument must be a vector, integer or floating point type (was '_Complex float')}}
 }
 
+void test_builtin_elementwise_rint(int i, float f, double d, float4 v, int3 iv, unsigned u, unsigned4 uv) {
+  struct Foo s = __builtin_elementwise_rint(f);
+  // expected-error@-1 {{initializing 'struct Foo' with an expression of incompatible type 'float'}}
+
+  i = __builtin_elementwise_rint();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+
+  i = __builtin_elementwise_rint(i);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'int')}}
+
+  i = __builtin_elementwise_rint(f, f);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+
+  u = __builtin_elementwise_rint(u);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned int')}}
+
+  uv = __builtin_elementwise_rint(uv);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned4' (vector of 4 'unsigned int' values))}}
+
+  // FIXME: Error should not mention integer
+  _Complex float c1, c2;
+  c1 = __builtin_elementwise_rint(c1);
+  // expected-error@-1 {{1st argument must be a vector, integer or floating point type (was '_Complex float')}}
+}
+
+void test_builtin_elementwise_nearbyint(int i, float f, double d, float4 v, int3 iv, unsigned u, unsigned4 uv) {
+  struct Foo s = __builtin_elementwise_nearbyint(f);
+  // expected-error@-1 {{initializing 'struct Foo' with an expression of incompatible type 'float'}}
+
+  i = __builtin_elementwise_nearbyint();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+
+  i = __builtin_elementwise_nearbyint(i);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'int')}}
+
+  i = __builtin_elementwise_nearbyint(f, f);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+
+  u = __builtin_elementwise_nearbyint(u);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned int')}}
+
+  uv = __builtin_elementwise_nearbyint(uv);
+  // expected-error@-1 {{1st argument must be a floating point type (was 'unsigned4' (vector of 4 'unsigned int' values))}}
+
+  // FIXME: Error should not mention integer
+  _Complex float c1, c2;
+  c1 = __builtin_elementwise_nearbyint(c1);
+  // expected-error@-1 {{1st argument must be a vector, integer or floating point type (was '_Complex float')}}
+}
+
 void test_builtin_elementwise_sin(int i, float f, double d, float4 v, int3 iv, unsigned u, unsigned4 uv) {
 
   struct Foo s = __builtin_elementwise_sin(f);
   // expected-error@-1 {{initializing 'struct Foo' with an expression of incompatible type 'float'}}
 
   i = __builtin_elementwise_sin();
   // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}