[CIR][X86] Implement lowering for sqrt builtins

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -4677,6 +4677,27 @@ class CIR_UnaryFPToFPBuiltinOp<string mnemonic, string llvmOpName>
   let llvmOp = llvmOpName;
 }
 
+def CIR_SqrtOp : CIR_UnaryFPToFPBuiltinOp<"sqrt", "SqrtOp"> {
+  let summary = "Floating-point square root operation";
+
+  let description = [{
+    Computes the square root of a floating-point value or vector.
+
+    The input must be either:
+      • a floating-point scalar type, or
+      • a vector whose element type is floating-point.
+
+    The result type must match the input type exactly.
+
+    Examples:
+      // scalar
+      %r = cir.sqrt %x : !cir.fp64
+
+      // vector
+      %v = cir.sqrt %vec : !cir.vector<!cir.fp32 x 4>
+  }];
+}
+
 def CIR_ACosOp : CIR_UnaryFPToFPBuiltinOp<"acos", "ACosOp"> {
   let summary = "Computes the arcus cosine of the specified value";
   let description = [{

clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp

@@ -1347,13 +1347,15 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_sqrtsh_round_mask:
   case X86::BI__builtin_ia32_sqrtsd_round_mask:
   case X86::BI__builtin_ia32_sqrtss_round_mask:
+    errorNYI("Unimplemented builtin");
+    return {};
   case X86::BI__builtin_ia32_sqrtph512:
   case X86::BI__builtin_ia32_sqrtps512:
-  case X86::BI__builtin_ia32_sqrtpd512:
-    cgm.errorNYI(expr->getSourceRange(),
-                 std::string("unimplemented X86 builtin call: ") +
-                     getContext().BuiltinInfo.getName(builtinID));
-    return {};
+  case X86::BI__builtin_ia32_sqrtpd512: {
+    mlir::Location loc = getLoc(expr->getExprLoc());
+    mlir::Value arg = ops[0];
+    return cir::SqrtOp::create(builder, loc, arg.getType(), arg).getResult();
+  }
   case X86::BI__builtin_ia32_pmuludq128:
   case X86::BI__builtin_ia32_pmuludq256:
   case X86::BI__builtin_ia32_pmuludq512: {

clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp

@@ -1,4 +1,4 @@
-//====- LowerToLLVM.cpp - Lowering from CIR to LLVMIR ---------------------===//
+//====- LowerToLLVM.cpp - Lowering from CIR to LLVMIR ---------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -186,6 +186,14 @@ mlir::LogicalResult CIRToLLVMCopyOpLowering::matchAndRewrite(
   return mlir::success();
 }
 
+mlir::LogicalResult SqrtOpLowering::matchAndRewrite(
+    cir::SqrtOp op, OpAdaptor adaptor, 
+    mlir::ConversionPatternRewriter &rewriter) const {
+  mlir::Type resTy = typeConverter->convertType(op.getType());
+  rewriter.replaceOpWithNewOp<mlir::LLVM::SqrtOp>(op, resTy, adaptor.getSrc());
+  return mlir::success();
+}
+
 mlir::LogicalResult CIRToLLVMCosOpLowering::matchAndRewrite(
     cir::CosOp op, OpAdaptor adaptor,
     mlir::ConversionPatternRewriter &rewriter) const {

clang/test/CIR/CodeGen/X86/cir-sqrt-builtins.c

@@ -0,0 +1,45 @@
+#include <immintrin.h>
+// Test X86-specific sqrt builtins
+
+// 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=OGCG --input-file=%t.ll %s
+
+// Test __builtin_ia32_sqrtph512
+__m512h test_sqrtph512(__m512h a) {
+  return __builtin_ia32_sqrtph512(a);
+}
+// CIR: cir.func @test_sqrtph512
+// CIR: [[RES:%.*]] = cir.sqrt {{%.*}} : !cir.vector<!cir.fp16 x 32>
+// CIR: cir.return [[RES]]
+// LLVM: define {{.*}} @test_sqrtph512
+// LLVM: call <32 x half> @llvm.sqrt.v32f16
+// OGCG: define {{.*}} @test_sqrtph512
+// OGCG: call <32 x half> @llvm.sqrt.v32f16
+
+// Test __builtin_ia32_sqrtps512
+__m512 test_sqrtps512(__m512 a) {
+  return __builtin_ia32_sqrtps512(a);
+}
+// CIR: cir.func @test_sqrtps512
+// CIR: [[RES:%.*]] = cir.sqrt {{%.*}} : !cir.vector<!cir.float x 16>
+// CIR: cir.return [[RES]]
+// LLVM: define {{.*}} @test_sqrtps512
+// LLVM: call <16 x float> @llvm.sqrt.v16f32
+// OGCG: define {{.*}} @test_sqrtps512
+// OGCG: call <16 x float> @llvm.sqrt.v16f32
+
+// Test __builtin_ia32_sqrtpd512
+__m512d test_sqrtpd512(__m512d a) {
+  return __builtin_ia32_sqrtpd512(a);
+}
+// CIR: cir.func @test_sqrtpd512
+// CIR: [[RES:%.*]] = cir.sqrt {{%.*}} : !cir.vector<!cir.double x 8>
+// CIR: cir.return [[RES]]
+// LLVM: define {{.*}} @test_sqrtpd512
+// LLVM: call <8 x double> @llvm.sqrt.v8f64
+// OGCG: define {{.*}} @test_sqrtpd512
+// OGCG: call <8 x double> @llvm.sqrt.v8f64