[llvm][SveEmitter] SVE ACLE for quadword permute intrinsics.

[fpetrogalli]

Summary

The following intrinsics have been added, guarded by the macro
__ARM_FEATURE_SVE_MATMUL_FP64:

• svtrn1q[_*]
• svtrn2q[_*]
• svuzp1q[_*]
• svuzp2q[_*]
• svzip1q[_*]
• svzip2q[_*]

Supported types:

• svint[8|16|32|64]_t
• svuint[8|16|32|64]_t
• svfloat[16|32|64]_t

TODO: add support for svbfloat16_t

[sdesmalen]

LGTM!

clang/include/clang/Basic/arm_sve.td

 }
 
 let ArchGuard = "defined(__ARM_FEATURE_SVE_MATMUL_FP32)" in {
 def SVMLLA_F32 : SInst<"svmmla[_f32]", "dddd","f", MergeNone, "aarch64_sve_fmmla">;
 }
 
 let ArchGuard = "defined(__ARM_FEATURE_SVE_MATMUL_FP64)" in {
 def SVMLLA_F64 : SInst<"svmmla[_f64]", "dddd","d", MergeNone, "aarch64_sve_fmmla">;
+def SVTRN1Q      : SInst<"svtrn1q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn1q">;
+def SVTRN2Q      : SInst<"svtrn2q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn2q">;
+def SVUZP1Q      : SInst<"svuzp1q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp1q">;
+def SVUZP2Q      : SInst<"svuzp2q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp2q">;
+def SVZIP1Q      : SInst<"svzip1q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip1q">;
+def SVZIP2Q      : SInst<"svzip2q[_{d}]",     "ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip2q">;
 }
 
 
 ////////////////////////////////////////////////////////////////////////////////
 // SVE2 WhileGE/GT
 let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
 def SVWHILEGE_S32 : SInst<"svwhilege_{d}[_{1}]", "Pkk", "PcPsPiPl",     MergeNone, "aarch64_sve_whilege", [IsOverloadWhile]>;
 def SVWHILEGE_S64 : SInst<"svwhilege_{d}[_{1}]", "Pll", "PcPsPiPl",     MergeNone, "aarch64_sve_whilege", [IsOverloadWhile]>;

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_trn1-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svtrn1_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svtrn1_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.trn1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svtrn1_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svtrn1_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.trn1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svtrn1_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svtrn1_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.trn1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svtrn1_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svtrn1_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.trn1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svtrn1_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svtrn1_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.trn1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svtrn1_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svtrn1_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.trn1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svtrn1_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svtrn1_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.trn1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svtrn1_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svtrn1_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.trn1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svtrn1_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svtrn1_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.trn1q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svtrn1_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svtrn1_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.trn1q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svtrn1_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svtrn1_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.trn1q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn1q, _f64, , )(op1, op2);
+}

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_trn2-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svtrn2_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svtrn2_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.trn2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svtrn2_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svtrn2_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.trn2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svtrn2_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svtrn2_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.trn2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svtrn2_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svtrn2_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.trn2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svtrn2_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svtrn2_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.trn2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svtrn2_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svtrn2_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.trn2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svtrn2_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svtrn2_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.trn2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svtrn2_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svtrn2_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.trn2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svtrn2_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svtrn2_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.trn2q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svtrn2_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svtrn2_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.trn2q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svtrn2_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svtrn2_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.trn2q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svtrn2q, _f64, , )(op1, op2);
+}

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_uzp1-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svuzp1_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svuzp1_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svuzp1_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svuzp1_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svuzp1_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svuzp1_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svuzp1_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svuzp1_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uzp1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svuzp1_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svuzp1_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svuzp1_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svuzp1_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svuzp1_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svuzp1_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svuzp1_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svuzp1_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uzp1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svuzp1_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svuzp1_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.uzp1q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svuzp1_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svuzp1_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.uzp1q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svuzp1_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svuzp1_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.uzp1q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp1q, _f64, , )(op1, op2);
+}

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_uzp2-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svuzp2_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svuzp2_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svuzp2_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svuzp2_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svuzp2_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svuzp2_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svuzp2_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svuzp2_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uzp2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svuzp2_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svuzp2_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svuzp2_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svuzp2_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svuzp2_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svuzp2_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svuzp2_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svuzp2_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uzp2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svuzp2_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svuzp2_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.uzp2q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svuzp2_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svuzp2_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.uzp2q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svuzp2_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svuzp2_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.uzp2q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svuzp2q, _f64, , )(op1, op2);
+}

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_zip1-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svzip1_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svzip1_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.zip1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svzip1_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svzip1_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.zip1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svzip1_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svzip1_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.zip1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svzip1_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svzip1_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.zip1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svzip1_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svzip1_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.zip1q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svzip1_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svzip1_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.zip1q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svzip1_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svzip1_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.zip1q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svzip1_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svzip1_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.zip1q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svzip1_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svzip1_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.zip1q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svzip1_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svzip1_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.zip1q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svzip1_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svzip1_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.zip1q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip1q, _f64, , )(op1, op2);
+}

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_zip2-fp64.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE_MATMUL_FP64 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1, A2, A3, A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_svzip2_s8(svint8_t op1, svint8_t op2) {
+  // CHECK-LABEL: test_svzip2_s8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.zip2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _s8, , )(op1, op2);
+}
+
+svint16_t test_svzip2_s16(svint16_t op1, svint16_t op2) {
+  // CHECK-LABEL: test_svzip2_s16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.zip2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _s16, , )(op1, op2);
+}
+
+svint32_t test_svzip2_s32(svint32_t op1, svint32_t op2) {
+  // CHECK-LABEL: test_svzip2_s32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.zip2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _s32, , )(op1, op2);
+}
+
+svint64_t test_svzip2_s64(svint64_t op1, svint64_t op2) {
+  // CHECK-LABEL: test_svzip2_s64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.zip2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _s64, , )(op1, op2);
+}
+
+svuint8_t test_svzip2_u8(svuint8_t op1, svuint8_t op2) {
+  // CHECK-LABEL: test_svzip2_u8
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.zip2q.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+  // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _u8, , )(op1, op2);
+}
+
+svuint16_t test_svzip2_u16(svuint16_t op1, svuint16_t op2) {
+  // CHECK-LABEL: test_svzip2_u16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.zip2q.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _u16, , )(op1, op2);
+}
+
+svuint32_t test_svzip2_u32(svuint32_t op1, svuint32_t op2) {
+  // CHECK-LABEL: test_svzip2_u32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.zip2q.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _u32, , )(op1, op2);
+}
+
+svuint64_t test_svzip2_u64(svuint64_t op1, svuint64_t op2) {
+  // CHECK-LABEL: test_svzip2_u64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.zip2q.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _u64, , )(op1, op2);
+}
+
+svfloat16_t test_svzip2_f16(svfloat16_t op1, svfloat16_t op2) {
+  // CHECK-LABEL: test_svzip2_f16
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.zip2q.nxv8f16(<vscale x 8 x half> %op1, <vscale x 8 x half> %op2)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _f16, , )(op1, op2);
+}
+
+svfloat32_t test_svzip2_f32(svfloat32_t op1, svfloat32_t op2) {
+  // CHECK-LABEL: test_svzip2_f32
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.zip2q.nxv4f32(<vscale x 4 x float> %op1, <vscale x 4 x float> %op2)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _f32, , )(op1, op2);
+}
+
+svfloat64_t test_svzip2_f64(svfloat64_t op1, svfloat64_t op2) {
+  // CHECK-LABEL: test_svzip2_f64
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.zip2q.nxv2f64(<vscale x 2 x double> %op1, <vscale x 2 x double> %op2)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svzip2q, _f64, , )(op1, op2);
+}