[clang] constexpr built-in reduce or and xor function.

[c8ef]

Part of #51787.
Follow up of #116822.

This patch adds constexpr support for the built-in reduce or and xor functions.

[llvmbot]

@llvm/pr-subscribers-clang

Author: None (c8ef)

Changes

Part of #51787.
Follow up of #116822.

This patch adds constexpr support for the built-in reduce or and xor functions.

---

Full diff: https://github.com/llvm/llvm-project/pull/116976.diff

4 Files Affected:

• (modified) clang/docs/ReleaseNotes.rst (+1)
• (modified) clang/include/clang/Basic/Builtins.td (+2-2)
• (modified) clang/lib/AST/ExprConstant.cpp (+11-1)
• (modified) clang/test/Sema/constant_builtins_vector.cpp (+20)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 999c88455b64a5..98ed21d09305b2 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -358,6 +358,7 @@ Non-comprehensive list of changes in this release
 - ``__builtin_reduce_add`` function can now be used in constant expressions.
 - ``__builtin_reduce_mul`` function can now be used in constant expressions.
 - ``__builtin_reduce_and`` function can now be used in constant expressions.
+- ``__builtin_reduce_or`` and ``__builtin_reduce_xor`` functions can now be used in constant expressions.
 
 New Compiler Flags
 ------------------
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index aa65f94e68f9c3..daf90b9570160e 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -1486,13 +1486,13 @@ def ReduceMinimum : Builtin {
 
 def ReduceXor : Builtin {
   let Spellings = ["__builtin_reduce_xor"];
-  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Attributes = [NoThrow, Const, CustomTypeChecking, Constexpr];
   let Prototype = "void(...)";
 }
 
 def ReduceOr : Builtin {
   let Spellings = ["__builtin_reduce_or"];
-  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Attributes = [NoThrow, Const, CustomTypeChecking, Constexpr];
   let Prototype = "void(...)";
 }
 
diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
index 33206f5cda2021..261de141637020 100644
--- a/clang/lib/AST/ExprConstant.cpp
+++ b/clang/lib/AST/ExprConstant.cpp
@@ -13529,7 +13529,9 @@ bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
 
   case Builtin::BI__builtin_reduce_add:
   case Builtin::BI__builtin_reduce_mul:
-  case Builtin::BI__builtin_reduce_and: {
+  case Builtin::BI__builtin_reduce_and:
+  case Builtin::BI__builtin_reduce_or:
+  case Builtin::BI__builtin_reduce_xor: {
     APValue Source;
     if (!EvaluateAsRValue(Info, E->getArg(0), Source))
       return false;
@@ -13558,6 +13560,14 @@ bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
         Reduced &= Source.getVectorElt(EltNum).getInt();
         break;
       }
+      case Builtin::BI__builtin_reduce_or: {
+        Reduced |= Source.getVectorElt(EltNum).getInt();
+        break;
+      }
+      case Builtin::BI__builtin_reduce_xor: {
+        Reduced ^= Source.getVectorElt(EltNum).getInt();
+        break;
+      }
       }
     }
 
diff --git a/clang/test/Sema/constant_builtins_vector.cpp b/clang/test/Sema/constant_builtins_vector.cpp
index 7063c290479f6c..e84d09b24672b4 100644
--- a/clang/test/Sema/constant_builtins_vector.cpp
+++ b/clang/test/Sema/constant_builtins_vector.cpp
@@ -777,3 +777,23 @@ static_assert(__builtin_reduce_and((vector4int){(int)~0x11111111, (int)~0x222222
 static_assert(__builtin_reduce_and((vector4long){(long long)~0x1111111111111111L, (long long)~0x2222222222222222L, (long long)~0x4444444444444444L, (long long)-1}) == 0x8888888888888888L);
 static_assert(__builtin_reduce_and((vector4uint){0x11111111U, 0x22222222U, 0x44444444U, 0x88888888U}) == 0U);
 static_assert(__builtin_reduce_and((vector4ulong){0x1111111111111111UL, 0x2222222222222222UL, 0x4444444444444444UL, 0x8888888888888888UL}) == 0L);
+
+static_assert(__builtin_reduce_or((vector4char){}) == 0);
+static_assert(__builtin_reduce_or((vector4char){(char)0x11, (char)0x22, (char)0x44, (char)0x88}) == (char)0xFF);
+static_assert(__builtin_reduce_or((vector4short){(short)0x1111, (short)0x2222, (short)0x4444, (short)0x8888}) == (short)0xFFFF);
+static_assert(__builtin_reduce_or((vector4int){(int)0x11111111, (int)0x22222222, (int)0x44444444, (int)0x88888888}) == (int)0xFFFFFFFF);
+static_assert(__builtin_reduce_or((vector4long){(long long)0x1111111111111111L, (long long)0x2222222222222222L, (long long)0x4444444444444444L, (long long)0x8888888888888888L}) == (long long)0xFFFFFFFFFFFFFFFFL);
+static_assert(__builtin_reduce_or((vector4char){(char)0, (char)0x22, (char)0x44, (char)0x88}) == ~0x11);
+static_assert(__builtin_reduce_or((vector4short){(short)0x1111, (short)0, (short)0x4444, (short)0x8888}) == ~0x2222);
+static_assert(__builtin_reduce_or((vector4int){(int)0x11111111, (int)0x22222222, (int)0, (int)0x88888888}) == ~0x44444444);
+static_assert(__builtin_reduce_or((vector4long){(long long)0x1111111111111111L, (long long)0x2222222222222222L, (long long)0x4444444444444444L, (long long)0}) == ~0x8888888888888888L);
+static_assert(__builtin_reduce_or((vector4uint){0x11111111U, 0x22222222U, 0x44444444U, 0x88888888U}) == 0xFFFFFFFFU);
+static_assert(__builtin_reduce_or((vector4ulong){0x1111111111111111UL, 0x2222222222222222UL, 0x4444444444444444UL, 0x8888888888888888UL}) == 0xFFFFFFFFFFFFFFFFL);
+
+static_assert(__builtin_reduce_xor((vector4char){}) == 0);
+static_assert(__builtin_reduce_xor((vector4char){(char)0x11, (char)0x22, (char)0x44, (char)0x88}) == (char)0xFF);
+static_assert(__builtin_reduce_xor((vector4short){(short)0x1111, (short)0x2222, (short)0x4444, (short)0x8888}) == (short)0xFFFF);
+static_assert(__builtin_reduce_xor((vector4int){(int)0x11111111, (int)0x22222222, (int)0x44444444, (int)0x88888888}) == (int)0xFFFFFFFF);
+static_assert(__builtin_reduce_xor((vector4long){(long long)0x1111111111111111L, (long long)0x2222222222222222L, (long long)0x4444444444444444L, (long long)0x8888888888888888L}) == (long long)0xFFFFFFFFFFFFFFFFL);
+static_assert(__builtin_reduce_xor((vector4uint){0x11111111U, 0x22222222U, 0x44444444U, 0x88888888U}) == 0xFFFFFFFFU);
+static_assert(__builtin_reduce_xor((vector4ulong){0x1111111111111111UL, 0x2222222222222222UL, 0x4444444444444444UL, 0x8888888888888888UL}) == 0xFFFFFFFFFFFFFFFFUL);

[RKSimon]

Please can you update the LanguageExtensions doc to mention that these are now constexpr

llvm-project/clang/docs/LanguageExtensions.rst

Lines 723 to 765 in 934140a

	*Reduction Builtins*
	Each builtin returns a scalar equivalent to applying the specified
	operation(x, y) as recursive even-odd pairwise reduction to all vector
	elements. ``operation(x, y)`` is repeatedly applied to each non-overlapping
	even-odd element pair with indices ``i * 2`` and ``i * 2 + 1`` with
	``i in [0, Number of elements / 2)``. If the numbers of elements is not a
	power of 2, the vector is widened with neutral elements for the reduction
	at the end to the next power of 2.
	These reductions support both fixed-sized and scalable vector types.
	Example:
	.. code-block:: c++
	__builtin_reduce_add([e3, e2, e1, e0]) = __builtin_reduced_add([e3 + e2, e1 + e0])
	= (e3 + e2) + (e1 + e0)
	Let ``VT`` be a vector type and ``ET`` the element type of ``VT``.
	======================================= ====================================================================== ==================================
	Name Operation Supported element types
	======================================= ====================================================================== ==================================
	ET __builtin_reduce_max(VT a) return the largest element of the vector. The floating point result integer and floating point types
	will always be a number unless all elements of the vector are NaN.
	ET __builtin_reduce_min(VT a) return the smallest element of the vector. The floating point result integer and floating point types
	will always be a number unless all elements of the vector are NaN.
	ET __builtin_reduce_add(VT a) \+ integer types
	ET __builtin_reduce_mul(VT a) \* integer types
	ET __builtin_reduce_and(VT a) & integer types
	ET __builtin_reduce_or(VT a) \| integer types
	ET __builtin_reduce_xor(VT a) ^ integer types
	ET __builtin_reduce_maximum(VT a) return the largest element of the vector. Follows IEEE 754-2019 floating point types
	semantics, see `LangRef
	<http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation>`_
	for the comparison.
	ET __builtin_reduce_minimum(VT a) return the smallest element of the vector. Follows IEEE 754-2019 floating point types
	semantics, see `LangRef
	<http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation>`_
	for the comparison.
	======================================= ====================================================================== ==================================

[Fznamznon]

Thanks!