From c60ade5fa04dd0730d7cd52980680f47cd685331 Mon Sep 17 00:00:00 2001
From: Tim Northover <tnorthover@apple.com>
Date: Wed, 19 Feb 2014 10:37:13 +0000
Subject: [PATCH] ARM NEON: add vcvtX (with rounding mode) intrinsics to v8
ARM.
These instructions (well, the f32 ones) are supported on 32-bit ARMv8, not just
AArch64. Now that the arm_neon.td refactoring is complete, adding them is
surprisingly simple.
rdar://problem/16035743
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@201661 91177308-0d34-0410-b5e6-96231b3b80d8
---
include/clang/Basic/arm_neon.td | 41 ++++---
lib/CodeGen/CGBuiltin.cpp | 206 +++++++++++++-------------------
test/CodeGen/arm-neon-vcvtX.c | 99 +++++++++++++++
3 files changed, 206 insertions(+), 140 deletions(-)
create mode 100644 test/CodeGen/arm-neon-vcvtX.c
diff --git a/include/clang/Basic/arm_neon.td b/include/clang/Basic/arm_neon.td
index 4b24634297c..eab3b50e384 100644
--- a/include/clang/Basic/arm_neon.td
+++ b/include/clang/Basic/arm_neon.td
@@ -690,22 +690,6 @@ def FRINTZ : SInst<"vrnd", "dd", "fdQfQd">;
def FRINTI : SInst<"vrndi", "dd", "fdQfQd">;
def VCVT_S64 : SInst<"vcvt_s64", "xd", "dQd">;
def VCVT_U64 : SInst<"vcvt_u64", "ud", "dQd">;
-def FCVTNS_S32 : SInst<"vcvtn_s32", "xd", "fQf">;
-def FCVTNS_S64 : SInst<"vcvtn_s64", "xd", "dQd">;
-def FCVTNU_S32 : SInst<"vcvtn_u32", "ud", "fQf">;
-def FCVTNU_S64 : SInst<"vcvtn_u64", "ud", "dQd">;
-def FCVTPS_S32 : SInst<"vcvtp_s32", "xd", "fQf">;
-def FCVTPS_S64 : SInst<"vcvtp_s64", "xd", "dQd">;
-def FCVTPU_S32 : SInst<"vcvtp_u32", "ud", "fQf">;
-def FCVTPU_S64 : SInst<"vcvtp_u64", "ud", "dQd">;
-def FCVTMS_S32 : SInst<"vcvtm_s32", "xd", "fQf">;
-def FCVTMS_S64 : SInst<"vcvtm_s64", "xd", "dQd">;
-def FCVTMU_S32 : SInst<"vcvtm_u32", "ud", "fQf">;
-def FCVTMU_S64 : SInst<"vcvtm_u64", "ud", "dQd">;
-def FCVTAS_S32 : SInst<"vcvta_s32", "xd", "fQf">;
-def FCVTAS_S64 : SInst<"vcvta_s64", "xd", "dQd">;
-def FCVTAU_S32 : SInst<"vcvta_u32", "ud", "fQf">;
-def FCVTAU_S64 : SInst<"vcvta_u64", "ud", "dQd">;
def FRECPE : SInst<"vrecpe", "dd", "dQd">;
def FRSQRTE : SInst<"vrsqrte", "dd", "dQd">;
def FSQRT : SInst<"vsqrt", "dd", "fdQfQd">;
@@ -948,6 +932,31 @@ def SHA256H2 : SInst<"vsha256h2", "dddd", "QUi">;
def SHA256SU1 : SInst<"vsha256su1", "dddd", "QUi">;
}
+////////////////////////////////////////////////////////////////////////////////
+// Float -> Int conversions with explicit rounding mode
+
+let ArchGuard = "__ARM_ARCH >= 8" in {
+def FCVTNS_S32 : SInst<"vcvtn_s32", "xd", "fQf">;
+def FCVTNU_S32 : SInst<"vcvtn_u32", "ud", "fQf">;
+def FCVTPS_S32 : SInst<"vcvtp_s32", "xd", "fQf">;
+def FCVTPU_S32 : SInst<"vcvtp_u32", "ud", "fQf">;
+def FCVTMS_S32 : SInst<"vcvtm_s32", "xd", "fQf">;
+def FCVTMU_S32 : SInst<"vcvtm_u32", "ud", "fQf">;
+def FCVTAS_S32 : SInst<"vcvta_s32", "xd", "fQf">;
+def FCVTAU_S32 : SInst<"vcvta_u32", "ud", "fQf">;
+}
+
+let ArchGuard = "__ARM_ARCH >= 8 && defined(__aarch64__)" in {
+def FCVTNS_S64 : SInst<"vcvtn_s64", "xd", "dQd">;
+def FCVTNU_S64 : SInst<"vcvtn_u64", "ud", "dQd">;
+def FCVTPS_S64 : SInst<"vcvtp_s64", "xd", "dQd">;
+def FCVTPU_S64 : SInst<"vcvtp_u64", "ud", "dQd">;
+def FCVTMS_S64 : SInst<"vcvtm_s64", "xd", "dQd">;
+def FCVTMU_S64 : SInst<"vcvtm_u64", "ud", "dQd">;
+def FCVTAS_S64 : SInst<"vcvta_s64", "xd", "dQd">;
+def FCVTAU_S64 : SInst<"vcvta_u64", "ud", "dQd">;
+}
+
////////////////////////////////////////////////////////////////////////////////
// Permutation
def VTRN1 : SOpInst<"vtrn1", "ddd",
diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp
index e25a1aae1b6..e4b5c999125 100644
--- a/lib/CodeGen/CGBuiltin.cpp
+++ b/lib/CodeGen/CGBuiltin.cpp
@@ -1914,14 +1914,94 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(unsigned BuiltinID,
}
case NEON::BI__builtin_neon_vcvt_s32_v:
case NEON::BI__builtin_neon_vcvt_u32_v:
+ case NEON::BI__builtin_neon_vcvt_s64_v:
+ case NEON::BI__builtin_neon_vcvt_u64_v:
case NEON::BI__builtin_neon_vcvtq_s32_v:
- case NEON::BI__builtin_neon_vcvtq_u32_v: {
+ case NEON::BI__builtin_neon_vcvtq_u32_v:
+ case NEON::BI__builtin_neon_vcvtq_s64_v:
+ case NEON::BI__builtin_neon_vcvtq_u64_v: {
+ bool Double =
+ (cast<llvm::IntegerType>(VTy->getElementType())->getBitWidth() == 64);
llvm::Type *FloatTy =
- GetNeonType(this, NeonTypeFlags(NeonTypeFlags::Float32, false, Quad));
+ GetNeonType(this, NeonTypeFlags(Double ? NeonTypeFlags::Float64
+ : NeonTypeFlags::Float32,
+ false, Quad));
Ops[0] = Builder.CreateBitCast(Ops[0], FloatTy);
return Usgn ? Builder.CreateFPToUI(Ops[0], Ty, "vcvt")
: Builder.CreateFPToSI(Ops[0], Ty, "vcvt");
}
+ case NEON::BI__builtin_neon_vcvta_s32_v:
+ case NEON::BI__builtin_neon_vcvta_s64_v:
+ case NEON::BI__builtin_neon_vcvta_u32_v:
+ case NEON::BI__builtin_neon_vcvta_u64_v:
+ case NEON::BI__builtin_neon_vcvtaq_s32_v:
+ case NEON::BI__builtin_neon_vcvtaq_s64_v:
+ case NEON::BI__builtin_neon_vcvtaq_u32_v:
+ case NEON::BI__builtin_neon_vcvtaq_u64_v: {
+ Int = Usgn ? Intrinsic::arm_neon_vcvtau : Intrinsic::arm_neon_vcvtas;
+ bool Double =
+ (cast<llvm::IntegerType>(VTy->getElementType())->getBitWidth() == 64);
+ llvm::Type *InTy =
+ GetNeonType(this,
+ NeonTypeFlags(Double ? NeonTypeFlags::Float64
+ : NeonTypeFlags::Float32, false, Quad));
+ llvm::Type *Tys[2] = { Ty, InTy };
+ return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvta");
+ }
+ case NEON::BI__builtin_neon_vcvtn_s32_v:
+ case NEON::BI__builtin_neon_vcvtn_s64_v:
+ case NEON::BI__builtin_neon_vcvtn_u32_v:
+ case NEON::BI__builtin_neon_vcvtn_u64_v:
+ case NEON::BI__builtin_neon_vcvtnq_s32_v:
+ case NEON::BI__builtin_neon_vcvtnq_s64_v:
+ case NEON::BI__builtin_neon_vcvtnq_u32_v:
+ case NEON::BI__builtin_neon_vcvtnq_u64_v: {
+ Int = Usgn ? Intrinsic::arm_neon_vcvtnu : Intrinsic::arm_neon_vcvtns;
+ bool Double =
+ (cast<llvm::IntegerType>(VTy->getElementType())->getBitWidth() == 64);
+ llvm::Type *InTy =
+ GetNeonType(this,
+ NeonTypeFlags(Double ? NeonTypeFlags::Float64
+ : NeonTypeFlags::Float32, false, Quad));
+ llvm::Type *Tys[2] = { Ty, InTy };
+ return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtn");
+ }
+ case NEON::BI__builtin_neon_vcvtp_s32_v:
+ case NEON::BI__builtin_neon_vcvtp_s64_v:
+ case NEON::BI__builtin_neon_vcvtp_u32_v:
+ case NEON::BI__builtin_neon_vcvtp_u64_v:
+ case NEON::BI__builtin_neon_vcvtpq_s32_v:
+ case NEON::BI__builtin_neon_vcvtpq_s64_v:
+ case NEON::BI__builtin_neon_vcvtpq_u32_v:
+ case NEON::BI__builtin_neon_vcvtpq_u64_v: {
+ Int = Usgn ? Intrinsic::arm_neon_vcvtpu : Intrinsic::arm_neon_vcvtps;
+ bool Double =
+ (cast<llvm::IntegerType>(VTy->getElementType())->getBitWidth() == 64);
+ llvm::Type *InTy =
+ GetNeonType(this,
+ NeonTypeFlags(Double ? NeonTypeFlags::Float64
+ : NeonTypeFlags::Float32, false, Quad));
+ llvm::Type *Tys[2] = { Ty, InTy };
+ return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtp");
+ }
+ case NEON::BI__builtin_neon_vcvtm_s32_v:
+ case NEON::BI__builtin_neon_vcvtm_s64_v:
+ case NEON::BI__builtin_neon_vcvtm_u32_v:
+ case NEON::BI__builtin_neon_vcvtm_u64_v:
+ case NEON::BI__builtin_neon_vcvtmq_s32_v:
+ case NEON::BI__builtin_neon_vcvtmq_s64_v:
+ case NEON::BI__builtin_neon_vcvtmq_u32_v:
+ case NEON::BI__builtin_neon_vcvtmq_u64_v: {
+ Int = Usgn ? Intrinsic::arm_neon_vcvtmu : Intrinsic::arm_neon_vcvtms;
+ bool Double =
+ (cast<llvm::IntegerType>(VTy->getElementType())->getBitWidth() == 64);
+ llvm::Type *InTy =
+ GetNeonType(this,
+ NeonTypeFlags(Double ? NeonTypeFlags::Float64
+ : NeonTypeFlags::Float32, false, Quad));
+ llvm::Type *Tys[2] = { Ty, InTy };
+ return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtm");
+ }
case NEON::BI__builtin_neon_vext_v:
case NEON::BI__builtin_neon_vextq_v: {
int CV = cast<ConstantInt>(Ops[2])->getSExtValue();
@@ -4059,128 +4139,6 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
Int = Intrinsic::nearbyint;
return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrndi");
}
- case NEON::BI__builtin_neon_vcvt_s64_v:
- case NEON::BI__builtin_neon_vcvt_u64_v:
- case NEON::BI__builtin_neon_vcvtq_s64_v:
- case NEON::BI__builtin_neon_vcvtq_u64_v: {
- llvm::Type *DoubleTy =
- GetNeonType(this, NeonTypeFlags(NeonTypeFlags::Float64, false, quad));
- Ops[0] = Builder.CreateBitCast(Ops[0], DoubleTy);
- return usgn ? Builder.CreateFPToUI(Ops[0], Ty, "vcvt")
- : Builder.CreateFPToSI(Ops[0], Ty, "vcvt");
- }
- case NEON::BI__builtin_neon_vcvtn_s32_v:
- case NEON::BI__builtin_neon_vcvtnq_s32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtns;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtns_f32");
- }
- case NEON::BI__builtin_neon_vcvtn_s64_v:
- case NEON::BI__builtin_neon_vcvtnq_s64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtns;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtns_f64");
- }
- case NEON::BI__builtin_neon_vcvtn_u32_v:
- case NEON::BI__builtin_neon_vcvtnq_u32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtnu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtnu_f32");
- }
- case NEON::BI__builtin_neon_vcvtn_u64_v:
- case NEON::BI__builtin_neon_vcvtnq_u64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtnu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtnu_f64");
- }
- case NEON::BI__builtin_neon_vcvtp_s32_v:
- case NEON::BI__builtin_neon_vcvtpq_s32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtps;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtps_f32");
- }
- case NEON::BI__builtin_neon_vcvtp_s64_v:
- case NEON::BI__builtin_neon_vcvtpq_s64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtps;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtps_f64");
- }
- case NEON::BI__builtin_neon_vcvtp_u32_v:
- case NEON::BI__builtin_neon_vcvtpq_u32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtpu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtpu_f32");
- }
- case NEON::BI__builtin_neon_vcvtp_u64_v:
- case NEON::BI__builtin_neon_vcvtpq_u64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtpu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtpu_f64");
- }
- case NEON::BI__builtin_neon_vcvtm_s32_v:
- case NEON::BI__builtin_neon_vcvtmq_s32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtms;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtms_f32");
- }
- case NEON::BI__builtin_neon_vcvtm_s64_v:
- case NEON::BI__builtin_neon_vcvtmq_s64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtms;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtms_f64");
- }
- case NEON::BI__builtin_neon_vcvtm_u32_v:
- case NEON::BI__builtin_neon_vcvtmq_u32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtmu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtmu_f32");
- }
- case NEON::BI__builtin_neon_vcvtm_u64_v:
- case NEON::BI__builtin_neon_vcvtmq_u64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtmu;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtmu_f64");
- }
- case NEON::BI__builtin_neon_vcvta_s32_v:
- case NEON::BI__builtin_neon_vcvtaq_s32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtas;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtas_f32");
- }
- case NEON::BI__builtin_neon_vcvta_s64_v:
- case NEON::BI__builtin_neon_vcvtaq_s64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtas;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtas_f64");
- }
- case NEON::BI__builtin_neon_vcvta_u32_v:
- case NEON::BI__builtin_neon_vcvtaq_u32_v: {
- llvm::Type *OpTy = llvm::VectorType::get(FloatTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtau;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtau_f32");
- }
- case NEON::BI__builtin_neon_vcvta_u64_v:
- case NEON::BI__builtin_neon_vcvtaq_u64_v: {
- llvm::Type *OpTy = llvm::VectorType::get(DoubleTy, VTy->getNumElements());
- llvm::Type *Tys[2] = { Ty, OpTy };
- Int = Intrinsic::arm_neon_vcvtau;
- return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtau_f64");
- }
case NEON::BI__builtin_neon_vsqrt_v:
case NEON::BI__builtin_neon_vsqrtq_v: {
Int = Intrinsic::sqrt;
diff --git a/test/CodeGen/arm-neon-vcvtX.c b/test/CodeGen/arm-neon-vcvtX.c
new file mode 100644
index 00000000000..ff8ce7ea3e2
--- /dev/null
+++ b/test/CodeGen/arm-neon-vcvtX.c
@@ -0,0 +1,99 @@
+// RUN: %clang_cc1 -triple thumbv8-linux-gnueabihf -target-cpu cortex-a57 -ffreestanding -O1 -emit-llvm %s -o - | FileCheck %s
+
+#include <arm_neon.h>
+
+int32x2_t test_vcvta_s32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvta_s32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtas.v2i32.v2f32(<2 x float> %a)
+ return vcvta_s32_f32(a);
+}
+
+uint32x2_t test_vcvta_u32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvta_u32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtau.v2i32.v2f32(<2 x float> %a)
+ return vcvta_u32_f32(a);
+}
+
+int32x4_t test_vcvtaq_s32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtaq_s32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtas.v4i32.v4f32(<4 x float> %a)
+ return vcvtaq_s32_f32(a);
+}
+
+uint32x4_t test_vcvtaq_u32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtaq_u32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtau.v4i32.v4f32(<4 x float> %a)
+ return vcvtaq_u32_f32(a);
+}
+
+int32x2_t test_vcvtn_s32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtn_s32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtns.v2i32.v2f32(<2 x float> %a)
+ return vcvtn_s32_f32(a);
+}
+
+uint32x2_t test_vcvtn_u32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtn_u32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtnu.v2i32.v2f32(<2 x float> %a)
+ return vcvtn_u32_f32(a);
+}
+
+int32x4_t test_vcvtnq_s32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtnq_s32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtns.v4i32.v4f32(<4 x float> %a)
+ return vcvtnq_s32_f32(a);
+}
+
+uint32x4_t test_vcvtnq_u32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtnq_u32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtnu.v4i32.v4f32(<4 x float> %a)
+ return vcvtnq_u32_f32(a);
+}
+
+int32x2_t test_vcvtp_s32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtp_s32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtps.v2i32.v2f32(<2 x float> %a)
+ return vcvtp_s32_f32(a);
+}
+
+uint32x2_t test_vcvtp_u32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtp_u32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtpu.v2i32.v2f32(<2 x float> %a)
+ return vcvtp_u32_f32(a);
+}
+
+int32x4_t test_vcvtpq_s32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtpq_s32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtps.v4i32.v4f32(<4 x float> %a)
+ return vcvtpq_s32_f32(a);
+}
+
+uint32x4_t test_vcvtpq_u32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtpq_u32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtpu.v4i32.v4f32(<4 x float> %a)
+ return vcvtpq_u32_f32(a);
+}
+
+int32x2_t test_vcvtm_s32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtm_s32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtms.v2i32.v2f32(<2 x float> %a)
+ return vcvtm_s32_f32(a);
+}
+
+uint32x2_t test_vcvtm_u32_f32(float32x2_t a) {
+ // CHECK-LABEL: test_vcvtm_u32_f32
+ // CHECK-LABEL: call <2 x i32> @llvm.arm.neon.vcvtmu.v2i32.v2f32(<2 x float> %a)
+ return vcvtm_u32_f32(a);
+}
+
+int32x4_t test_vcvtmq_s32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtmq_s32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtms.v4i32.v4f32(<4 x float> %a)
+ return vcvtmq_s32_f32(a);
+}
+
+uint32x4_t test_vcvtmq_u32_f32(float32x4_t a) {
+ // CHECK-LABEL: test_vcvtmq_u32_f32
+ // CHECK-LABEL: call <4 x i32> @llvm.arm.neon.vcvtmu.v4i32.v4f32(<4 x float> %a)
+ return vcvtmq_u32_f32(a);
+}
--
GitLab