[clang] c577201 - [SveEmitter] Add builtins for bitcount operations
Sander de Smalen via cfe-commits
cfe-commits at lists.llvm.org
Tue Apr 28 05:55:35 PDT 2020
Author: Sander de Smalen
Date: 2020-04-28T13:53:54+01:00
New Revision: c57720125fa7596be2403e7810957655d04dfece
URL: https://github.com/llvm/llvm-project/commit/c57720125fa7596be2403e7810957655d04dfece
DIFF: https://github.com/llvm/llvm-project/commit/c57720125fa7596be2403e7810957655d04dfece.diff
LOG: [SveEmitter] Add builtins for bitcount operations
This patch adds builtins for svcls, svclz and svcnt.
For merging (_m), zeroing (_z) and don't-care (_x) predication.
Added:
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cls.c
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clz.c
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cnt.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 eaee860bd3c0..03aef14cc2d7 100644
--- a/clang/include/clang/Basic/arm_sve.td
+++ b/clang/include/clang/Basic/arm_sve.td
@@ -702,6 +702,19 @@ def SVWHILELS_U64 : SInst<"svwhilele_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNon
def SVWHILELT_S32 : SInst<"svwhilelt_{d}[_{1}]", "Pkk", "PcPsPiPl", MergeNone, "aarch64_sve_whilelt", [IsOverloadWhile]>;
def SVWHILELT_S64 : SInst<"svwhilelt_{d}[_{1}]", "Pll", "PcPsPiPl", MergeNone, "aarch64_sve_whilelt", [IsOverloadWhile]>;
+////////////////////////////////////////////////////////////////////////////////
+// Counting bit
+
+multiclass SInstCLS<string name, string types, string intrinsic, list<FlagType> flags=[]> {
+ def _M : SInst<name # "[_{d}]", "uuPd", types, MergeOp1, intrinsic, flags>;
+ def _X : SInst<name # "[_{d}]", "uPd", types, MergeAnyExp, intrinsic, flags>;
+ def _Z : SInst<name # "[_{d}]", "uPd", types, MergeZeroExp, intrinsic, flags>;
+}
+
+defm SVCLS : SInstCLS<"svcls", "csil", "aarch64_sve_cls">;
+defm SVCLZ : SInstCLS<"svclz", "csilUcUsUiUl", "aarch64_sve_clz">;
+defm SVCNT : SInstCLS<"svcnt", "csilUcUsUiUlhfd", "aarch64_sve_cnt">;
+
////////////////////////////////////////////////////////////////////////////////
// Floating-point arithmetic
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cls.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cls.c
new file mode 100644
index 000000000000..e464ac5a2ff6
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cls.c
@@ -0,0 +1,116 @@
+// 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
+
+svuint8_t test_svcls_s8_z(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcls_s8_z
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cls.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(svcls,_s8,_z,)(pg, op);
+}
+
+svuint16_t test_svcls_s16_z(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s16,_z,)(pg, op);
+}
+
+svuint32_t test_svcls_s32_z(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s32,_z,)(pg, op);
+}
+
+svuint64_t test_svcls_s64_z(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s64,_z,)(pg, op);
+}
+
+svuint8_t test_svcls_s8_m(svuint8_t inactive, svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcls_s8_m
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cls.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(svcls,_s8,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svcls_s16_m(svuint16_t inactive, svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcls_s32_m(svuint32_t inactive, svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcls_s64_m(svuint64_t inactive, svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s64,_m,)(inactive, pg, op);
+}
+
+svuint8_t test_svcls_s8_x(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcls_s8_x
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cls.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(svcls,_s8,_x,)(pg, op);
+}
+
+svuint16_t test_svcls_s16_x(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s16,_x,)(pg, op);
+}
+
+svuint32_t test_svcls_s32_x(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s32,_x,)(pg, op);
+}
+
+svuint64_t test_svcls_s64_x(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcls_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.cls.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(svcls,_s64,_x,)(pg, op);
+}
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clz.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clz.c
new file mode 100644
index 000000000000..fe46e4a655ac
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clz.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
+
+svuint8_t test_svclz_s8_z(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svclz_s8_z
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_s8,_z,)(pg, op);
+}
+
+svuint16_t test_svclz_s16_z(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s16,_z,)(pg, op);
+}
+
+svuint32_t test_svclz_s32_z(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s32,_z,)(pg, op);
+}
+
+svuint64_t test_svclz_s64_z(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s64,_z,)(pg, op);
+}
+
+svuint8_t test_svclz_u8_z(svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svclz_u8_z
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_u8,_z,)(pg, op);
+}
+
+svuint16_t test_svclz_u16_z(svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u16,_z,)(pg, op);
+}
+
+svuint32_t test_svclz_u32_z(svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u32,_z,)(pg, op);
+}
+
+svuint64_t test_svclz_u64_z(svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u64,_z,)(pg, op);
+}
+
+svuint8_t test_svclz_s8_m(svuint8_t inactive, svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svclz_s8_m
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_s8,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svclz_s16_m(svuint16_t inactive, svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svclz_s32_m(svuint32_t inactive, svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svclz_s64_m(svuint64_t inactive, svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s64,_m,)(inactive, pg, op);
+}
+
+svuint8_t test_svclz_u8_m(svuint8_t inactive, svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svclz_u8_m
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_u8,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svclz_u16_m(svuint16_t inactive, svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svclz_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svclz_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u64,_m,)(inactive, pg, op);
+}
+
+svuint8_t test_svclz_s8_x(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svclz_s8_x
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_s8,_x,)(pg, op);
+}
+
+svuint16_t test_svclz_s16_x(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s16,_x,)(pg, op);
+}
+
+svuint32_t test_svclz_s32_x(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s32,_x,)(pg, op);
+}
+
+svuint64_t test_svclz_s64_x(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_s64,_x,)(pg, op);
+}
+
+svuint8_t test_svclz_u8_x(svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svclz_u8_x
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.clz.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(svclz,_u8,_x,)(pg, op);
+}
+
+svuint16_t test_svclz_u16_x(svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u16,_x,)(pg, op);
+}
+
+svuint32_t test_svclz_u32_x(svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u32,_x,)(pg, op);
+}
+
+svuint64_t test_svclz_u64_x(svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svclz_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.clz.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(svclz,_u64,_x,)(pg, op);
+}
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cnt.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cnt.c
new file mode 100644
index 000000000000..be5b3c30e74c
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cnt.c
@@ -0,0 +1,302 @@
+// 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
+
+svuint8_t test_svcnt_s8_z(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_s8_z
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_s8,_z,)(pg, op);
+}
+
+svuint16_t test_svcnt_s16_z(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s16,_z,)(pg, op);
+}
+
+svuint32_t test_svcnt_s32_z(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s32,_z,)(pg, op);
+}
+
+svuint64_t test_svcnt_s64_z(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s64,_z,)(pg, op);
+}
+
+svuint8_t test_svcnt_u8_z(svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_u8_z
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_u8,_z,)(pg, op);
+}
+
+svuint16_t test_svcnt_u16_z(svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u16,_z,)(pg, op);
+}
+
+svuint32_t test_svcnt_u32_z(svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u32,_z,)(pg, op);
+}
+
+svuint64_t test_svcnt_u64_z(svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u64,_z,)(pg, op);
+}
+
+svuint16_t test_svcnt_f16_z(svbool_t pg, svfloat16_t op)
+{
+ // CHECK-LABEL: test_svcnt_f16_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.cnt.nxv8f16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f16,_z,)(pg, op);
+}
+
+svuint32_t test_svcnt_f32_z(svbool_t pg, svfloat32_t op)
+{
+ // CHECK-LABEL: test_svcnt_f32_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.cnt.nxv4f32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f32,_z,)(pg, op);
+}
+
+svuint64_t test_svcnt_f64_z(svbool_t pg, svfloat64_t op)
+{
+ // CHECK-LABEL: test_svcnt_f64_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.cnt.nxv2f64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f64,_z,)(pg, op);
+}
+
+svuint8_t test_svcnt_s8_m(svuint8_t inactive, svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_s8_m
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_s8,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svcnt_s16_m(svuint16_t inactive, svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcnt_s32_m(svuint32_t inactive, svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcnt_s64_m(svuint64_t inactive, svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s64,_m,)(inactive, pg, op);
+}
+
+svuint8_t test_svcnt_u8_m(svuint8_t inactive, svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_u8_m
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_u8,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svcnt_u16_m(svuint16_t inactive, svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcnt_u32_m(svuint32_t inactive, svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcnt_u64_m(svuint64_t inactive, svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u64,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svcnt_f16_m(svuint16_t inactive, svbool_t pg, svfloat16_t op)
+{
+ // CHECK-LABEL: test_svcnt_f16_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.cnt.nxv8f16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcnt_f32_m(svuint32_t inactive, svbool_t pg, svfloat32_t op)
+{
+ // CHECK-LABEL: test_svcnt_f32_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.cnt.nxv4f32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcnt_f64_m(svuint64_t inactive, svbool_t pg, svfloat64_t op)
+{
+ // CHECK-LABEL: test_svcnt_f64_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.cnt.nxv2f64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f64,_m,)(inactive, pg, op);
+}
+
+svuint8_t test_svcnt_s8_x(svbool_t pg, svint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_s8_x
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_s8,_x,)(pg, op);
+}
+
+svuint16_t test_svcnt_s16_x(svbool_t pg, svint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s16,_x,)(pg, op);
+}
+
+svuint32_t test_svcnt_s32_x(svbool_t pg, svint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s32,_x,)(pg, op);
+}
+
+svuint64_t test_svcnt_s64_x(svbool_t pg, svint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_s64,_x,)(pg, op);
+}
+
+svuint8_t test_svcnt_u8_x(svbool_t pg, svuint8_t op)
+{
+ // CHECK-LABEL: test_svcnt_u8_x
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.cnt.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(svcnt,_u8,_x,)(pg, op);
+}
+
+svuint16_t test_svcnt_u16_x(svbool_t pg, svuint16_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u16,_x,)(pg, op);
+}
+
+svuint32_t test_svcnt_u32_x(svbool_t pg, svuint32_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u32,_x,)(pg, op);
+}
+
+svuint64_t test_svcnt_u64_x(svbool_t pg, svuint64_t op)
+{
+ // CHECK-LABEL: test_svcnt_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.cnt.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(svcnt,_u64,_x,)(pg, op);
+}
+
+svuint16_t test_svcnt_f16_x(svbool_t pg, svfloat16_t op)
+{
+ // CHECK-LABEL: test_svcnt_f16_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.cnt.nxv8f16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f16,_x,)(pg, op);
+}
+
+svuint32_t test_svcnt_f32_x(svbool_t pg, svfloat32_t op)
+{
+ // CHECK-LABEL: test_svcnt_f32_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.cnt.nxv4f32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f32,_x,)(pg, op);
+}
+
+svuint64_t test_svcnt_f64_x(svbool_t pg, svfloat64_t op)
+{
+ // CHECK-LABEL: test_svcnt_f64_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.cnt.nxv2f64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ return SVE_ACLE_FUNC(svcnt,_f64,_x,)(pg, op);
+}
More information about the cfe-commits
mailing list