[llvm-branch-commits] [clang] ddc49d0 - [clang] [MSVC] Implement __mulh and __umulh builtins for aarch64

[Tom Stellard via llvm-branch-commits]

Author: Martin Storsjö
Date: 2021-08-19T21:47:01-07:00
New Revision: ddc49d0f1f2bd80e9b23a8d93fa60129c0b90121

URL: https://github.com/llvm/llvm-project/commit/ddc49d0f1f2bd80e9b23a8d93fa60129c0b90121
DIFF: https://github.com/llvm/llvm-project/commit/ddc49d0f1f2bd80e9b23a8d93fa60129c0b90121.diff

LOG: [clang] [MSVC] Implement __mulh and __umulh builtins for aarch64

The code is based on the same __mulh and __umulh intrinsics for
x86.

This should fix PR51128.

Differential Revision: https://reviews.llvm.org/D106721

(cherry picked from commit cc3affd8b02091bf475f9dd083802e1b6a232be1)

Added: 
    

Modified: 
    clang/include/clang/Basic/BuiltinsAArch64.def
    clang/lib/CodeGen/CGBuiltin.cpp
    clang/lib/Headers/intrin.h
    clang/test/CodeGen/arm64-microsoft-intrinsics.c

Removed: 
    


################################################################################
diff  --git a/clang/include/clang/Basic/BuiltinsAArch64.def b/clang/include/clang/Basic/BuiltinsAArch64.def
index 1dac5d2371d48..634bcaed20a6f 100644
--- a/clang/include/clang/Basic/BuiltinsAArch64.def
+++ b/clang/include/clang/Basic/BuiltinsAArch64.def
@@ -243,6 +243,9 @@ TARGET_HEADER_BUILTIN(_ReadStatusReg,  "LLii",  "nh", "intrin.h", ALL_MS_LANGUAG
 TARGET_HEADER_BUILTIN(_WriteStatusReg, "viLLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_AddressOfReturnAddress, "v*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
+TARGET_HEADER_BUILTIN(__mulh,  "SLLiSLLiSLLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(__umulh, "ULLiULLiULLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+
 #undef BUILTIN
 #undef LANGBUILTIN
 #undef TARGET_HEADER_BUILTIN

diff  --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index d9b2a5fe16bec..1a02965b223ed 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -9732,6 +9732,29 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(F);
   }
 
+  if (BuiltinID == AArch64::BI__mulh || BuiltinID == AArch64::BI__umulh) {
+    llvm::Type *ResType = ConvertType(E->getType());
+    llvm::Type *Int128Ty = llvm::IntegerType::get(getLLVMContext(), 128);
+
+    bool IsSigned = BuiltinID == AArch64::BI__mulh;
+    Value *LHS =
+        Builder.CreateIntCast(EmitScalarExpr(E->getArg(0)), Int128Ty, IsSigned);
+    Value *RHS =
+        Builder.CreateIntCast(EmitScalarExpr(E->getArg(1)), Int128Ty, IsSigned);
+
+    Value *MulResult, *HigherBits;
+    if (IsSigned) {
+      MulResult = Builder.CreateNSWMul(LHS, RHS);
+      HigherBits = Builder.CreateAShr(MulResult, 64);
+    } else {
+      MulResult = Builder.CreateNUWMul(LHS, RHS);
+      HigherBits = Builder.CreateLShr(MulResult, 64);
+    }
+    HigherBits = Builder.CreateIntCast(HigherBits, ResType, IsSigned);
+
+    return HigherBits;
+  }
+
   // Handle MSVC intrinsics before argument evaluation to prevent double
   // evaluation.
   if (Optional<MSVCIntrin> MsvcIntId = translateAarch64ToMsvcIntrin(BuiltinID))

diff  --git a/clang/lib/Headers/intrin.h b/clang/lib/Headers/intrin.h
index ff8eb8fca2687..34ec79d6acbc6 100644
--- a/clang/lib/Headers/intrin.h
+++ b/clang/lib/Headers/intrin.h
@@ -574,6 +574,9 @@ void _WriteStatusReg(int, __int64);
 unsigned short __cdecl _byteswap_ushort(unsigned short val);
 unsigned long __cdecl _byteswap_ulong (unsigned long val);
 unsigned __int64 __cdecl _byteswap_uint64(unsigned __int64 val);
+
+__int64 __mulh(__int64 __a, __int64 __b);
+unsigned __int64 __umulh(unsigned __int64 __a, unsigned __int64 __b);
 #endif
 
 /*----------------------------------------------------------------------------*\

diff  --git a/clang/test/CodeGen/arm64-microsoft-intrinsics.c b/clang/test/CodeGen/arm64-microsoft-intrinsics.c
index ca8f270bd4f36..36f182433edd2 100644
--- a/clang/test/CodeGen/arm64-microsoft-intrinsics.c
+++ b/clang/test/CodeGen/arm64-microsoft-intrinsics.c
@@ -81,6 +81,28 @@ void check_ReadWriteBarrier() {
 // CHECK-MSVC: fence syncscope("singlethread")
 // CHECK-LINUX: error: implicit declaration of function '_ReadWriteBarrier'
 
+long long check_mulh(long long a, long long b) {
+  return __mulh(a, b);
+}
+
+// CHECK-MSVC: %[[ARG1:.*]] = sext i64 {{.*}} to i128
+// CHECK-MSVC: %[[ARG2:.*]] = sext i64 {{.*}} to i128
+// CHECK-MSVC: %[[PROD:.*]] = mul nsw i128 %[[ARG1]], %[[ARG2]]
+// CHECK-MSVC: %[[HIGH:.*]] = ashr i128 %[[PROD]], 64
+// CHECK-MSVC: %[[RES:.*]] = trunc i128 %[[HIGH]] to i64
+// CHECK-LINUX: error: implicit declaration of function '__mulh'
+
+unsigned long long check_umulh(unsigned long long a, unsigned long long b) {
+  return __umulh(a, b);
+}
+
+// CHECK-MSVC: %[[ARG1:.*]] = zext i64 {{.*}} to i128
+// CHECK-MSVC: %[[ARG2:.*]] = zext i64 {{.*}} to i128
+// CHECK-MSVC: %[[PROD:.*]] = mul nuw i128 %[[ARG1]], %[[ARG2]]
+// CHECK-MSVC: %[[HIGH:.*]] = lshr i128 %[[PROD]], 64
+// CHECK-MSVC: %[[RES:.*]] = trunc i128 %[[HIGH]] to i64
+// CHECK-LINUX: error: implicit declaration of function '__umulh'
+
 unsigned __int64 check__getReg() {
   unsigned volatile __int64 reg;
   reg = __getReg(18);