[ARM,MVE] Add the vmovlbq,vmovltq intrinsic family.

Summary:
These intrinsics take a vector of 2n elements, and return a vector of
n wider elements obtained by sign- or zero-extending every other
element of the input vector. They're represented in IR as a
shufflevector that extracts the odd or even elements of the input,
followed by a sext or zext.

Existing LLVM codegen already matches this pattern and generates the
VMOVLB instruction (which widens the even-index input lanes). But no
existing isel rule was generating VMOVLT, so I've added some. However,
the new rules currently only work in little-endian MVE, because the
pattern they expect from isel lowering includes a bitconvert which
doesn't have the right semantics in big-endian.

The output of one existing codegen test is improved by those new
rules.

This commit adds the unpredicated forms only.

Reviewers: dmgreen, miyuki, MarkMurrayARM, ostannard

Reviewed By: dmgreen

Subscribers: kristof.beyls, hiraditya, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74336
diff --git a/clang/include/clang/Basic/arm_mve.td b/clang/include/clang/Basic/arm_mve.td
index 21801b4..55ddfc2 100644
--- a/clang/include/clang/Basic/arm_mve.td
+++ b/clang/include/clang/Basic/arm_mve.td
@@ -420,6 +420,13 @@
 defm : float_int_conversions<f32, s32, fptosi, sitofp>;
 defm : float_int_conversions<f16, s16, fptosi, sitofp>;
 
+let params = [s8, u8, s16, u16] in {
+  def vmovlbq: Intrinsic<DblVector, (args Vector:$a),
+    (extend (unzip $a, 0), DblVector, (unsignedflag Scalar))>;
+  def vmovltq: Intrinsic<DblVector, (args Vector:$a),
+    (extend (unzip $a, 1), DblVector, (unsignedflag Scalar))>;
+}
+
 let params = T.Float in {
   def vrndq: Intrinsic<Vector, (args Vector:$a),
       (IRIntBase<"trunc", [Vector]> $a)>;
diff --git a/clang/include/clang/Basic/arm_mve_defs.td b/clang/include/clang/Basic/arm_mve_defs.td
index 9e5b7b3..f6816cd 100644
--- a/clang/include/clang/Basic/arm_mve_defs.td
+++ b/clang/include/clang/Basic/arm_mve_defs.td
@@ -128,6 +128,9 @@
 def vrev: CGHelperFn<"ARMMVEVectorElementReverse"> {
   let special_params = [IRBuilderIntParam<1, "unsigned">];
 }
+def unzip: CGHelperFn<"VectorUnzip"> {
+  let special_params = [IRBuilderIntParam<1, "bool">];
+}
 
 // Helper for making boolean flags in IR
 def i1: IRBuilderBase {
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 788f14b..b30a79a0 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -7056,6 +7056,17 @@
   }
 }
 
+static llvm::Value *VectorUnzip(CGBuilderTy &Builder, llvm::Value *V, bool Odd) {
+  // Make a shufflevector that extracts every other element of a vector (evens
+  // or odds, as desired).
+  SmallVector<uint32_t, 16> Indices;
+  unsigned InputElements = V->getType()->getVectorNumElements();
+  for (unsigned i = 0; i < InputElements; i += 2)
+    Indices.push_back(i + Odd);
+  return Builder.CreateShuffleVector(V, llvm::UndefValue::get(V->getType()),
+                                     Indices);
+}
+
 template<unsigned HighBit, unsigned OtherBits>
 static llvm::Value *ARMMVEConstantSplat(CGBuilderTy &Builder, llvm::Type *VT) {
   // MVE-specific helper function to make a vector splat of a constant such as
diff --git a/clang/test/CodeGen/arm-mve-intrinsics/vmovl.c b/clang/test/CodeGen/arm-mve-intrinsics/vmovl.c
new file mode 100644
index 0000000..0b8ef59
--- /dev/null
+++ b/clang/test/CodeGen/arm-mve-intrinsics/vmovl.c
@@ -0,0 +1,126 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
+// RUN: %clang_cc1 -triple thumbv8.1m.main-arm-none-eabi -target-feature +mve -mfloat-abi hard -fallow-half-arguments-and-returns -O0 -disable-O0-optnone -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s
+// RUN: %clang_cc1 -triple thumbv8.1m.main-arm-none-eabi -target-feature +mve -mfloat-abi hard -fallow-half-arguments-and-returns -O0 -disable-O0-optnone -DPOLYMORPHIC -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s
+
+#include <arm_mve.h>
+
+// CHECK-LABEL: @test_vmovlbq_s8(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i8> [[A:%.*]], <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+// CHECK-NEXT:    [[TMP1:%.*]] = sext <8 x i8> [[TMP0]] to <8 x i16>
+// CHECK-NEXT:    ret <8 x i16> [[TMP1]]
+//
+int16x8_t test_vmovlbq_s8(int8x16_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovlbq(a);
+#else /* POLYMORPHIC */
+    return vmovlbq_s8(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovlbq_s16(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+// CHECK-NEXT:    [[TMP1:%.*]] = sext <4 x i16> [[TMP0]] to <4 x i32>
+// CHECK-NEXT:    ret <4 x i32> [[TMP1]]
+//
+int32x4_t test_vmovlbq_s16(int16x8_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovlbq(a);
+#else /* POLYMORPHIC */
+    return vmovlbq_s16(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovlbq_u8(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i8> [[A:%.*]], <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+// CHECK-NEXT:    [[TMP1:%.*]] = zext <8 x i8> [[TMP0]] to <8 x i16>
+// CHECK-NEXT:    ret <8 x i16> [[TMP1]]
+//
+uint16x8_t test_vmovlbq_u8(uint8x16_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovlbq(a);
+#else /* POLYMORPHIC */
+    return vmovlbq_u8(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovlbq_u16(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+// CHECK-NEXT:    [[TMP1:%.*]] = zext <4 x i16> [[TMP0]] to <4 x i32>
+// CHECK-NEXT:    ret <4 x i32> [[TMP1]]
+//
+uint32x4_t test_vmovlbq_u16(uint16x8_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovlbq(a);
+#else /* POLYMORPHIC */
+    return vmovlbq_u16(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovltq_s8(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i8> [[A:%.*]], <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+// CHECK-NEXT:    [[TMP1:%.*]] = sext <8 x i8> [[TMP0]] to <8 x i16>
+// CHECK-NEXT:    ret <8 x i16> [[TMP1]]
+//
+int16x8_t test_vmovltq_s8(int8x16_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovltq(a);
+#else /* POLYMORPHIC */
+    return vmovltq_s8(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovltq_s16(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+// CHECK-NEXT:    [[TMP1:%.*]] = sext <4 x i16> [[TMP0]] to <4 x i32>
+// CHECK-NEXT:    ret <4 x i32> [[TMP1]]
+//
+int32x4_t test_vmovltq_s16(int16x8_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovltq(a);
+#else /* POLYMORPHIC */
+    return vmovltq_s16(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovltq_u8(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i8> [[A:%.*]], <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+// CHECK-NEXT:    [[TMP1:%.*]] = zext <8 x i8> [[TMP0]] to <8 x i16>
+// CHECK-NEXT:    ret <8 x i16> [[TMP1]]
+//
+uint16x8_t test_vmovltq_u8(uint8x16_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovltq(a);
+#else /* POLYMORPHIC */
+    return vmovltq_u8(a);
+#endif /* POLYMORPHIC */
+}
+
+// CHECK-LABEL: @test_vmovltq_u16(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+// CHECK-NEXT:    [[TMP1:%.*]] = zext <4 x i16> [[TMP0]] to <4 x i32>
+// CHECK-NEXT:    ret <4 x i32> [[TMP1]]
+//
+uint32x4_t test_vmovltq_u16(uint16x8_t a)
+{
+#ifdef POLYMORPHIC
+    return vmovltq(a);
+#else /* POLYMORPHIC */
+    return vmovltq_u16(a);
+#endif /* POLYMORPHIC */
+}
+