1 // REQUIRES: aarch64-registered-target
2 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
3 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s
4 // RUN: %clang_cc1 -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
5 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s
6 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -o - %s >/dev/null
7 #include <arm_sve.h>
8
9 #ifdef SVE_OVERLOADED_FORMS
10 // A simple used,unused... macro, long enough to represent any SVE builtin.
11 #define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
12 #else
13 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
14 #endif
15
test_svsqrt_f16_z(svbool_t pg,svfloat16_t op)16 svfloat16_t test_svsqrt_f16_z(svbool_t pg, svfloat16_t op)
17 {
18 // CHECK-LABEL: test_svsqrt_f16_z
19 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
20 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fsqrt.nxv8f16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
21 // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
22 return SVE_ACLE_FUNC(svsqrt,_f16,_z,)(pg, op);
23 }
24
test_svsqrt_f32_z(svbool_t pg,svfloat32_t op)25 svfloat32_t test_svsqrt_f32_z(svbool_t pg, svfloat32_t op)
26 {
27 // CHECK-LABEL: test_svsqrt_f32_z
28 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
29 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fsqrt.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
30 // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
31 return SVE_ACLE_FUNC(svsqrt,_f32,_z,)(pg, op);
32 }
33
test_svsqrt_f64_z(svbool_t pg,svfloat64_t op)34 svfloat64_t test_svsqrt_f64_z(svbool_t pg, svfloat64_t op)
35 {
36 // CHECK-LABEL: test_svsqrt_f64_z
37 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
38 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fsqrt.nxv2f64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
39 // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
40 return SVE_ACLE_FUNC(svsqrt,_f64,_z,)(pg, op);
41 }
42
test_svsqrt_f16_m(svfloat16_t inactive,svbool_t pg,svfloat16_t op)43 svfloat16_t test_svsqrt_f16_m(svfloat16_t inactive, svbool_t pg, svfloat16_t op)
44 {
45 // CHECK-LABEL: test_svsqrt_f16_m
46 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
47 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fsqrt.nxv8f16(<vscale x 8 x half> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
48 // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
49 return SVE_ACLE_FUNC(svsqrt,_f16,_m,)(inactive, pg, op);
50 }
51
test_svsqrt_f32_m(svfloat32_t inactive,svbool_t pg,svfloat32_t op)52 svfloat32_t test_svsqrt_f32_m(svfloat32_t inactive, svbool_t pg, svfloat32_t op)
53 {
54 // CHECK-LABEL: test_svsqrt_f32_m
55 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
56 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fsqrt.nxv4f32(<vscale x 4 x float> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
57 // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
58 return SVE_ACLE_FUNC(svsqrt,_f32,_m,)(inactive, pg, op);
59 }
60
test_svsqrt_f64_m(svfloat64_t inactive,svbool_t pg,svfloat64_t op)61 svfloat64_t test_svsqrt_f64_m(svfloat64_t inactive, svbool_t pg, svfloat64_t op)
62 {
63 // CHECK-LABEL: test_svsqrt_f64_m
64 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
65 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fsqrt.nxv2f64(<vscale x 2 x double> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
66 // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
67 return SVE_ACLE_FUNC(svsqrt,_f64,_m,)(inactive, pg, op);
68 }
69
test_svsqrt_f16_x(svbool_t pg,svfloat16_t op)70 svfloat16_t test_svsqrt_f16_x(svbool_t pg, svfloat16_t op)
71 {
72 // CHECK-LABEL: test_svsqrt_f16_x
73 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
74 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fsqrt.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
75 // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
76 return SVE_ACLE_FUNC(svsqrt,_f16,_x,)(pg, op);
77 }
78
test_svsqrt_f32_x(svbool_t pg,svfloat32_t op)79 svfloat32_t test_svsqrt_f32_x(svbool_t pg, svfloat32_t op)
80 {
81 // CHECK-LABEL: test_svsqrt_f32_x
82 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
83 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fsqrt.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
84 // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
85 return SVE_ACLE_FUNC(svsqrt,_f32,_x,)(pg, op);
86 }
87
test_svsqrt_f64_x(svbool_t pg,svfloat64_t op)88 svfloat64_t test_svsqrt_f64_x(svbool_t pg, svfloat64_t op)
89 {
90 // CHECK-LABEL: test_svsqrt_f64_x
91 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
92 // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fsqrt.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
93 // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
94 return SVE_ACLE_FUNC(svsqrt,_f64,_x,)(pg, op);
95 }
96