[clang] [CIR] Lowering for __builtin_reduce_add/mul (PR #224204)

[Kunal Dubey via cfe-commits]

https://github.com/xakep8 updated https://github.com/llvm/llvm-project/pull/224204

>From 422867a172ef2fbf49ef601ae0c2435720eb303f Mon Sep 17 00:00:00 2001
From: Kunal Dubey <xakep8 at protonmail.com>
Date: Thu, 17 Sep 2026 12:17:03 +0530
Subject: [PATCH] [CIR] Lowering for __builtin_reduce_add/mul

Added lowering for __builtin_reduce_add and __builtin_reduce_mul by
mimicking the same shape as __builtin_reduce_xor which mirrors classic
Codegen shape

Added test for the same
---
 clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp       | 10 +++-
 .../builtin-reduce-arithmetic.c               | 49 +++++++++++++++++++
 .../CodeGenBuiltins/builtin-undef-rvalue.cpp  |  8 +--
 3 files changed, 62 insertions(+), 5 deletions(-)
 create mode 100644 clang/test/CIR/CodeGenBuiltins/builtin-reduce-arithmetic.c

diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp
index e2fe3adefd4af..b0e3f0ae00ebe 100644
--- a/clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp
@@ -2141,9 +2141,17 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl &gd, unsigned builtinID,
   case Builtin::BI__builtin_elementwise_minimumnum:
   case Builtin::BI__builtin_reduce_max:
   case Builtin::BI__builtin_reduce_min:
+    return errorBuiltinNYI(*this, e, builtinID);
   case Builtin::BI__builtin_reduce_add:
+    return emitBuiltinWithOneOverloadedType<1>(
+        e, "vector.reduce.add",
+        cast<cir::VectorType>(convertType(e->getArg(0)->getType()))
+            .getElementType());
   case Builtin::BI__builtin_reduce_mul:
-    return errorBuiltinNYI(*this, e, builtinID);
+    return emitBuiltinWithOneOverloadedType<1>(
+        e, "vector.reduce.mul",
+        cast<cir::VectorType>(convertType(e->getArg(0)->getType()))
+            .getElementType());
   case Builtin::BI__builtin_reduce_xor:
     return emitBuiltinWithOneOverloadedType<1>(
         e, "vector.reduce.xor",
diff --git a/clang/test/CIR/CodeGenBuiltins/builtin-reduce-arithmetic.c b/clang/test/CIR/CodeGenBuiltins/builtin-reduce-arithmetic.c
new file mode 100644
index 0000000000000..271aaecf5e0a8
--- /dev/null
+++ b/clang/test/CIR/CodeGenBuiltins/builtin-reduce-arithmetic.c
@@ -0,0 +1,49 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm %s -o %t-cir.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t-cir.ll %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -emit-llvm %s -o %t.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t.ll %s
+
+typedef int v4si __attribute__((vector_size(16)));
+typedef unsigned int v4su __attribute__((vector_size(16)));
+
+int test_reduce_add(v4si x) {
+  // CIR-LABEL: @test_reduce_add
+  // CIR: cir.call_llvm_intrinsic "vector.reduce.add"
+  // CIR: cir.return
+  // LLVM-LABEL: @test_reduce_add
+  // LLVM: call i32 @llvm.vector.reduce.add.v4i32(<4 x i32>
+  // LLVM: ret i32
+  return __builtin_reduce_add(x);
+}
+
+unsigned test_reduce_add_unsigned(v4su x) {
+  // CIR-LABEL: @test_reduce_add_unsigned
+  // CIR: cir.call_llvm_intrinsic "vector.reduce.add"
+  // CIR: cir.return
+  // LLVM-LABEL: @test_reduce_add_unsigned
+  // LLVM: call i32 @llvm.vector.reduce.add.v4i32(<4 x i32>
+  // LLVM: ret i32
+  return __builtin_reduce_add(x);
+}
+
+int test_reduce_mul(v4si x) {
+  // CIR-LABEL: @test_reduce_mul
+  // CIR: cir.call_llvm_intrinsic "vector.reduce.mul"
+  // CIR: cir.return
+  // LLVM-LABEL: @test_reduce_mul
+  // LLVM: call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32>
+  // LLVM: ret i32
+  return __builtin_reduce_mul(x);
+}
+
+unsigned test_reduce_mul_unsigned(v4su x) {
+  // CIR-LABEL: @test_reduce_mul_unsigned
+  // CIR: cir.call_llvm_intrinsic "vector.reduce.mul"
+  // CIR: cir.return
+  // LLVM-LABEL: @test_reduce_mul_unsigned
+  // LLVM: call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32>
+  // LLVM: ret i32
+  return __builtin_reduce_mul(x);
+}
diff --git a/clang/test/CIR/CodeGenBuiltins/builtin-undef-rvalue.cpp b/clang/test/CIR/CodeGenBuiltins/builtin-undef-rvalue.cpp
index c15375621f08a..372eb97c323f3 100644
--- a/clang/test/CIR/CodeGenBuiltins/builtin-undef-rvalue.cpp
+++ b/clang/test/CIR/CodeGenBuiltins/builtin-undef-rvalue.cpp
@@ -3,11 +3,11 @@
 
 typedef int v4si __attribute__((vector_size(16)));
 
-int test_builtin_reduce_add_undef_rvalue(v4si x) {
-  // expected-error at +1 {{unimplemented builtin call: __builtin_reduce_add}}
-  return __builtin_reduce_add(x);
+int test_builtin_reduce_max_undef_rvalue(v4si x) {
+  // expected-error at +1 {{unimplemented builtin call: __builtin_reduce_max}}
+  return __builtin_reduce_max(x);
 }
 
-// CIR-LABEL: @_Z36test_builtin_reduce_add_undef_rvalueDv4_i
+// CIR-LABEL: @_Z36test_builtin_reduce_max_undef_rvalueDv4_i
 // CIR:         cir.const #cir.undef : !s32i
 // CIR:         cir.return