Author: Sander de Smalen Date: 2020-04-28T14:06:51+01:00 New Revision: 476ba8127bfa4553bf5ce1654cd844803e8d6dea
URL: https://github.com/llvm/llvm-project/commit/476ba8127bfa4553bf5ce1654cd844803e8d6dea DIFF: https://github.com/llvm/llvm-project/commit/476ba8127bfa4553bf5ce1654cd844803e8d6dea.diff LOG: [SveEmitter] Add builtins for zero/sign extension and bit/byte reversal. This patch adds builtins for predicated unary builtins svext[bhw] and svrev[bhw] and svrbit. Added: clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extb.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_exth.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extw.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_rbit.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revb.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revh.c clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revw.c Modified: clang/include/clang/Basic/arm_sve.td Removed: ################################################################################ diff --git a/clang/include/clang/Basic/arm_sve.td b/clang/include/clang/Basic/arm_sve.td index 03aef14cc2d7..e2c01023e55d 100644 --- a/clang/include/clang/Basic/arm_sve.td +++ b/clang/include/clang/Basic/arm_sve.td @@ -715,6 +715,24 @@ defm SVCLS : SInstCLS<"svcls", "csil", "aarch64_sve_cls">; defm SVCLZ : SInstCLS<"svclz", "csilUcUsUiUl", "aarch64_sve_clz">; defm SVCNT : SInstCLS<"svcnt", "csilUcUsUiUlhfd", "aarch64_sve_cnt">; +//////////////////////////////////////////////////////////////////////////////// +// Conversion + +defm SVEXTB_S : SInstZPZ<"svextb", "sil", "aarch64_sve_sxtb">; +defm SVEXTB_U : SInstZPZ<"svextb", "UsUiUl", "aarch64_sve_uxtb">; +defm SVEXTH_S : SInstZPZ<"svexth", "il", "aarch64_sve_sxth">; +defm SVEXTH_U : SInstZPZ<"svexth", "UiUl", "aarch64_sve_uxth">; +defm SVEXTW_S : SInstZPZ<"svextw", "l", "aarch64_sve_sxtw">; +defm SVEXTW_U : SInstZPZ<"svextw", "Ul", "aarch64_sve_uxtw">; + +//////////////////////////////////////////////////////////////////////////////// +// Reversal + +defm SVRBIT : SInstZPZ<"svrbit", "csilUcUsUiUl", "aarch64_sve_rbit">; +defm SVREVB : SInstZPZ<"svrevb", "silUsUiUl", "aarch64_sve_revb">; +defm SVREVH : SInstZPZ<"svrevh", "ilUiUl", "aarch64_sve_revh">; +defm SVREVW : SInstZPZ<"svrevw", "lUl", "aarch64_sve_revw">; + //////////////////////////////////////////////////////////////////////////////// // Floating-point arithmetic diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extb.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extb.c new file mode 100644 index 000000000000..720b81a28a18 --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extb.c @@ -0,0 +1,173 @@ +// RUN: %clang_cc1 -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 -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 + +svint16_t test_svextb_s16_z(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svextb_s16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sxtb.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s16,_z,)(pg, op); +} + +svint32_t test_svextb_s32_z(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svextb_s32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxtb.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s32,_z,)(pg, op); +} + +svint64_t test_svextb_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextb_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtb.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s64,_z,)(pg, op); +} + +svuint16_t test_svextb_u16_z(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svextb_u16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uxtb.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u16,_z,)(pg, op); +} + +svuint32_t test_svextb_u32_z(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svextb_u32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxtb.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u32,_z,)(pg, op); +} + +svuint64_t test_svextb_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextb_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtb.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u64,_z,)(pg, op); +} + +svint16_t test_svextb_s16_m(svint16_t inactive, svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svextb_s16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sxtb.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s16,_m,)(inactive, pg, op); +} + +svint32_t test_svextb_s32_m(svint32_t inactive, svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svextb_s32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxtb.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s32,_m,)(inactive, pg, op); +} + +svint64_t test_svextb_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextb_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtb.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s64,_m,)(inactive, pg, op); +} + +svuint16_t test_svextb_u16_m(svuint16_t inactive, svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svextb_u16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uxtb.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u16,_m,)(inactive, pg, op); +} + +svuint32_t test_svextb_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svextb_u32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxtb.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u32,_m,)(inactive, pg, op); +} + +svuint64_t test_svextb_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextb_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtb.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u64,_m,)(inactive, pg, op); +} + +svint16_t test_svextb_s16_x(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svextb_s16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sxtb.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s16,_x,)(pg, op); +} + +svint32_t test_svextb_s32_x(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svextb_s32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxtb.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s32,_x,)(pg, op); +} + +svint64_t test_svextb_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextb_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtb.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_s64,_x,)(pg, op); +} + +svuint16_t test_svextb_u16_x(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svextb_u16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.uxtb.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u16,_x,)(pg, op); +} + +svuint32_t test_svextb_u32_x(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svextb_u32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxtb.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u32,_x,)(pg, op); +} + +svuint64_t test_svextb_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextb_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtb.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextb,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_exth.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_exth.c new file mode 100644 index 000000000000..eb279566a82e --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_exth.c @@ -0,0 +1,119 @@ +// RUN: %clang_cc1 -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 -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 + +svint32_t test_svexth_s32_z(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svexth_s32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxth.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s32,_z,)(pg, op); +} + +svint64_t test_svexth_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svexth_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxth.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s64,_z,)(pg, op); +} + +svuint32_t test_svexth_u32_z(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svexth_u32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxth.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u32,_z,)(pg, op); +} + +svuint64_t test_svexth_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svexth_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxth.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u64,_z,)(pg, op); +} + +svint32_t test_svexth_s32_m(svint32_t inactive, svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svexth_s32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxth.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s32,_m,)(inactive, pg, op); +} + +svint64_t test_svexth_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svexth_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxth.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s64,_m,)(inactive, pg, op); +} + +svuint32_t test_svexth_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svexth_u32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxth.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u32,_m,)(inactive, pg, op); +} + +svuint64_t test_svexth_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svexth_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxth.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u64,_m,)(inactive, pg, op); +} + +svint32_t test_svexth_s32_x(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svexth_s32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sxth.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s32,_x,)(pg, op); +} + +svint64_t test_svexth_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svexth_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxth.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_s64,_x,)(pg, op); +} + +svuint32_t test_svexth_u32_x(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svexth_u32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.uxth.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u32,_x,)(pg, op); +} + +svuint64_t test_svexth_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svexth_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxth.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svexth,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extw.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extw.c new file mode 100644 index 000000000000..76626475c89a --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_extw.c @@ -0,0 +1,65 @@ +// RUN: %clang_cc1 -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 -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 + +svint64_t test_svextw_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextw_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtw.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_s64,_z,)(pg, op); +} + +svuint64_t test_svextw_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextw_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtw.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_u64,_z,)(pg, op); +} + +svint64_t test_svextw_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextw_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtw.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_s64,_m,)(inactive, pg, op); +} + +svuint64_t test_svextw_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextw_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtw.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_u64,_m,)(inactive, pg, op); +} + +svint64_t test_svextw_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svextw_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sxtw.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_s64,_x,)(pg, op); +} + +svuint64_t test_svextw_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svextw_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.uxtw.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svextw,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_rbit.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_rbit.c new file mode 100644 index 000000000000..907360a5e7c5 --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_rbit.c @@ -0,0 +1,221 @@ +// RUN: %clang_cc1 -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 -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_svrbit_s8_z(svbool_t pg, svint8_t op) +{ + // CHECK-LABEL: test_svrbit_s8_z + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s8,_z,)(pg, op); +} + +svint16_t test_svrbit_s16_z(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrbit_s16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s16,_z,)(pg, op); +} + +svint32_t test_svrbit_s32_z(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrbit_s32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s32,_z,)(pg, op); +} + +svint64_t test_svrbit_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrbit_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s64,_z,)(pg, op); +} + +svuint8_t test_svrbit_u8_z(svbool_t pg, svuint8_t op) +{ + // CHECK-LABEL: test_svrbit_u8_z + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u8,_z,)(pg, op); +} + +svuint16_t test_svrbit_u16_z(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrbit_u16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u16,_z,)(pg, op); +} + +svuint32_t test_svrbit_u32_z(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrbit_u32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u32,_z,)(pg, op); +} + +svuint64_t test_svrbit_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrbit_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u64,_z,)(pg, op); +} + +svint8_t test_svrbit_s8_m(svint8_t inactive, svbool_t pg, svint8_t op) +{ + // CHECK-LABEL: test_svrbit_s8_m + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> %inactive, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s8,_m,)(inactive, pg, op); +} + +svint16_t test_svrbit_s16_m(svint16_t inactive, svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrbit_s16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s16,_m,)(inactive, pg, op); +} + +svint32_t test_svrbit_s32_m(svint32_t inactive, svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrbit_s32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s32,_m,)(inactive, pg, op); +} + +svint64_t test_svrbit_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrbit_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s64,_m,)(inactive, pg, op); +} + +svuint8_t test_svrbit_u8_m(svuint8_t inactive, svbool_t pg, svuint8_t op) +{ + // CHECK-LABEL: test_svrbit_u8_m + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> %inactive, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u8,_m,)(inactive, pg, op); +} + +svuint16_t test_svrbit_u16_m(svuint16_t inactive, svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrbit_u16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u16,_m,)(inactive, pg, op); +} + +svuint32_t test_svrbit_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrbit_u32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u32,_m,)(inactive, pg, op); +} + +svuint64_t test_svrbit_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrbit_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u64,_m,)(inactive, pg, op); +} + +svint8_t test_svrbit_s8_x(svbool_t pg, svint8_t op) +{ + // CHECK-LABEL: test_svrbit_s8_x + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> undef, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s8,_x,)(pg, op); +} + +svint16_t test_svrbit_s16_x(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrbit_s16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s16,_x,)(pg, op); +} + +svint32_t test_svrbit_s32_x(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrbit_s32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s32,_x,)(pg, op); +} + +svint64_t test_svrbit_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrbit_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_s64,_x,)(pg, op); +} + +svuint8_t test_svrbit_u8_x(svbool_t pg, svuint8_t op) +{ + // CHECK-LABEL: test_svrbit_u8_x + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.rbit.nxv16i8(<vscale x 16 x i8> undef, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op) + // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u8,_x,)(pg, op); +} + +svuint16_t test_svrbit_u16_x(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrbit_u16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.rbit.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u16,_x,)(pg, op); +} + +svuint32_t test_svrbit_u32_x(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrbit_u32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.rbit.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u32,_x,)(pg, op); +} + +svuint64_t test_svrbit_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrbit_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.rbit.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrbit,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revb.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revb.c new file mode 100644 index 000000000000..2aaf2dc8644f --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revb.c @@ -0,0 +1,173 @@ +// RUN: %clang_cc1 -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 -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 + +svint16_t test_svrevb_s16_z(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrevb_s16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s16,_z,)(pg, op); +} + +svint32_t test_svrevb_s32_z(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevb_s32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s32,_z,)(pg, op); +} + +svint64_t test_svrevb_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevb_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s64,_z,)(pg, op); +} + +svuint16_t test_svrevb_u16_z(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrevb_u16_z + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u16,_z,)(pg, op); +} + +svuint32_t test_svrevb_u32_z(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevb_u32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u32,_z,)(pg, op); +} + +svuint64_t test_svrevb_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevb_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u64,_z,)(pg, op); +} + +svint16_t test_svrevb_s16_m(svint16_t inactive, svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrevb_s16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s16,_m,)(inactive, pg, op); +} + +svint32_t test_svrevb_s32_m(svint32_t inactive, svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevb_s32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s32,_m,)(inactive, pg, op); +} + +svint64_t test_svrevb_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevb_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s64,_m,)(inactive, pg, op); +} + +svuint16_t test_svrevb_u16_m(svuint16_t inactive, svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrevb_u16_m + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u16,_m,)(inactive, pg, op); +} + +svuint32_t test_svrevb_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevb_u32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u32,_m,)(inactive, pg, op); +} + +svuint64_t test_svrevb_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevb_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u64,_m,)(inactive, pg, op); +} + +svint16_t test_svrevb_s16_x(svbool_t pg, svint16_t op) +{ + // CHECK-LABEL: test_svrevb_s16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s16,_x,)(pg, op); +} + +svint32_t test_svrevb_s32_x(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevb_s32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s32,_x,)(pg, op); +} + +svint64_t test_svrevb_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevb_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_s64,_x,)(pg, op); +} + +svuint16_t test_svrevb_u16_x(svbool_t pg, svuint16_t op) +{ + // CHECK-LABEL: test_svrevb_u16_x + // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.revb.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op) + // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u16,_x,)(pg, op); +} + +svuint32_t test_svrevb_u32_x(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevb_u32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revb.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u32,_x,)(pg, op); +} + +svuint64_t test_svrevb_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevb_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revb.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevb,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revh.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revh.c new file mode 100644 index 000000000000..ac7d5c0b2545 --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revh.c @@ -0,0 +1,119 @@ +// RUN: %clang_cc1 -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 -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 + +svint32_t test_svrevh_s32_z(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevh_s32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s32,_z,)(pg, op); +} + +svint64_t test_svrevh_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevh_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s64,_z,)(pg, op); +} + +svuint32_t test_svrevh_u32_z(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevh_u32_z + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u32,_z,)(pg, op); +} + +svuint64_t test_svrevh_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevh_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u64,_z,)(pg, op); +} + +svint32_t test_svrevh_s32_m(svint32_t inactive, svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevh_s32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s32,_m,)(inactive, pg, op); +} + +svint64_t test_svrevh_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevh_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s64,_m,)(inactive, pg, op); +} + +svuint32_t test_svrevh_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevh_u32_m + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u32,_m,)(inactive, pg, op); +} + +svuint64_t test_svrevh_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevh_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u64,_m,)(inactive, pg, op); +} + +svint32_t test_svrevh_s32_x(svbool_t pg, svint32_t op) +{ + // CHECK-LABEL: test_svrevh_s32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s32,_x,)(pg, op); +} + +svint64_t test_svrevh_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevh_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_s64,_x,)(pg, op); +} + +svuint32_t test_svrevh_u32_x(svbool_t pg, svuint32_t op) +{ + // CHECK-LABEL: test_svrevh_u32_x + // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.revh.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op) + // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u32,_x,)(pg, op); +} + +svuint64_t test_svrevh_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevh_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revh.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevh,_u64,_x,)(pg, op); +} diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revw.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revw.c new file mode 100644 index 000000000000..1d6d414b85b6 --- /dev/null +++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_revw.c @@ -0,0 +1,65 @@ +// RUN: %clang_cc1 -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 -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 + +svint64_t test_svrevw_s64_z(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevw_s64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_s64,_z,)(pg, op); +} + +svuint64_t test_svrevw_u64_z(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevw_u64_z + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_u64,_z,)(pg, op); +} + +svint64_t test_svrevw_s64_m(svint64_t inactive, svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevw_s64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_s64,_m,)(inactive, pg, op); +} + +svuint64_t test_svrevw_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevw_u64_m + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_u64,_m,)(inactive, pg, op); +} + +svint64_t test_svrevw_s64_x(svbool_t pg, svint64_t op) +{ + // CHECK-LABEL: test_svrevw_s64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_s64,_x,)(pg, op); +} + +svuint64_t test_svrevw_u64_x(svbool_t pg, svuint64_t op) +{ + // CHECK-LABEL: test_svrevw_u64_x + // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg) + // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.revw.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op) + // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]] + return SVE_ACLE_FUNC(svrevw,_u64,_x,)(pg, op); +} _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
