1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
2 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2 -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
3 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2 -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
4 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
5 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
7 // REQUIRES: aarch64-registered-target
11 #ifdef SVE_OVERLOADED_FORMS
12 // A simple used,unused... macro, long enough to represent any SVE builtin.
13 #define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
15 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
18 // CHECK-LABEL: @test_svcdot_s32(
20 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 0)
21 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
23 // CPP-CHECK-LABEL: @_Z15test_svcdot_s32u11__SVInt32_tu10__SVInt8_tS0_(
24 // CPP-CHECK-NEXT: entry:
25 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 0)
26 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
28 svint32_t
test_svcdot_s32(svint32_t op1
, svint8_t op2
, svint8_t op3
)
30 return SVE_ACLE_FUNC(svcdot
,_s32
,,)(op1
, op2
, op3
, 0);
33 // CHECK-LABEL: @test_svcdot_s32_1(
35 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 90)
36 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
38 // CPP-CHECK-LABEL: @_Z17test_svcdot_s32_1u11__SVInt32_tu10__SVInt8_tS0_(
39 // CPP-CHECK-NEXT: entry:
40 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 90)
41 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
43 svint32_t
test_svcdot_s32_1(svint32_t op1
, svint8_t op2
, svint8_t op3
)
45 return SVE_ACLE_FUNC(svcdot
,_s32
,,)(op1
, op2
, op3
, 90);
48 // CHECK-LABEL: @test_svcdot_s32_2(
50 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 180)
51 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
53 // CPP-CHECK-LABEL: @_Z17test_svcdot_s32_2u11__SVInt32_tu10__SVInt8_tS0_(
54 // CPP-CHECK-NEXT: entry:
55 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 180)
56 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
58 svint32_t
test_svcdot_s32_2(svint32_t op1
, svint8_t op2
, svint8_t op3
)
60 return SVE_ACLE_FUNC(svcdot
,_s32
,,)(op1
, op2
, op3
, 180);
63 // CHECK-LABEL: @test_svcdot_s32_3(
65 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 270)
66 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
68 // CPP-CHECK-LABEL: @_Z17test_svcdot_s32_3u11__SVInt32_tu10__SVInt8_tS0_(
69 // CPP-CHECK-NEXT: entry:
70 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 270)
71 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
73 svint32_t
test_svcdot_s32_3(svint32_t op1
, svint8_t op2
, svint8_t op3
)
75 return SVE_ACLE_FUNC(svcdot
,_s32
,,)(op1
, op2
, op3
, 270);
78 // CHECK-LABEL: @test_svcdot_s64(
80 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 0)
81 // CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
83 // CPP-CHECK-LABEL: @_Z15test_svcdot_s64u11__SVInt64_tu11__SVInt16_tS0_(
84 // CPP-CHECK-NEXT: entry:
85 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 0)
86 // CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
88 svint64_t
test_svcdot_s64(svint64_t op1
, svint16_t op2
, svint16_t op3
)
90 return SVE_ACLE_FUNC(svcdot
,_s64
,,)(op1
, op2
, op3
, 0);
93 // CHECK-LABEL: @test_svcdot_s64_1(
95 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 90)
96 // CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
98 // CPP-CHECK-LABEL: @_Z17test_svcdot_s64_1u11__SVInt64_tu11__SVInt16_tS0_(
99 // CPP-CHECK-NEXT: entry:
100 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 90)
101 // CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
103 svint64_t
test_svcdot_s64_1(svint64_t op1
, svint16_t op2
, svint16_t op3
)
105 return SVE_ACLE_FUNC(svcdot
,_s64
,,)(op1
, op2
, op3
, 90);
108 // CHECK-LABEL: @test_svcdot_s64_2(
109 // CHECK-NEXT: entry:
110 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 180)
111 // CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
113 // CPP-CHECK-LABEL: @_Z17test_svcdot_s64_2u11__SVInt64_tu11__SVInt16_tS0_(
114 // CPP-CHECK-NEXT: entry:
115 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 180)
116 // CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
118 svint64_t
test_svcdot_s64_2(svint64_t op1
, svint16_t op2
, svint16_t op3
)
120 return SVE_ACLE_FUNC(svcdot
,_s64
,,)(op1
, op2
, op3
, 180);
123 // CHECK-LABEL: @test_svcdot_s64_3(
124 // CHECK-NEXT: entry:
125 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 270)
126 // CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
128 // CPP-CHECK-LABEL: @_Z17test_svcdot_s64_3u11__SVInt64_tu11__SVInt16_tS0_(
129 // CPP-CHECK-NEXT: entry:
130 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 270)
131 // CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
133 svint64_t
test_svcdot_s64_3(svint64_t op1
, svint16_t op2
, svint16_t op3
)
135 return SVE_ACLE_FUNC(svcdot
,_s64
,,)(op1
, op2
, op3
, 270);
138 // CHECK-LABEL: @test_svcdot_lane_s32(
139 // CHECK-NEXT: entry:
140 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.lane.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 0, i32 0)
141 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
143 // CPP-CHECK-LABEL: @_Z20test_svcdot_lane_s32u11__SVInt32_tu10__SVInt8_tS0_(
144 // CPP-CHECK-NEXT: entry:
145 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.lane.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 0, i32 0)
146 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
148 svint32_t
test_svcdot_lane_s32(svint32_t op1
, svint8_t op2
, svint8_t op3
)
150 return SVE_ACLE_FUNC(svcdot_lane
,_s32
,,)(op1
, op2
, op3
, 0, 0);
153 // CHECK-LABEL: @test_svcdot_lane_s32_1(
154 // CHECK-NEXT: entry:
155 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.lane.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 2, i32 90)
156 // CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
158 // CPP-CHECK-LABEL: @_Z22test_svcdot_lane_s32_1u11__SVInt32_tu10__SVInt8_tS0_(
159 // CPP-CHECK-NEXT: entry:
160 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.cdot.lane.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 16 x i8> [[OP2:%.*]], <vscale x 16 x i8> [[OP3:%.*]], i32 2, i32 90)
161 // CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
163 svint32_t
test_svcdot_lane_s32_1(svint32_t op1
, svint8_t op2
, svint8_t op3
)
165 return SVE_ACLE_FUNC(svcdot_lane
,_s32
,,)(op1
, op2
, op3
, 2, 90);
168 // CHECK-LABEL: @test_svcdot_lane_s64(
169 // CHECK-NEXT: entry:
170 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.lane.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 0, i32 180)
171 // CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
173 // CPP-CHECK-LABEL: @_Z20test_svcdot_lane_s64u11__SVInt64_tu11__SVInt16_tS0_(
174 // CPP-CHECK-NEXT: entry:
175 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.cdot.lane.nxv2i64(<vscale x 2 x i64> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 0, i32 180)
176 // CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
178 svint64_t
test_svcdot_lane_s64(svint64_t op1
, svint16_t op2
, svint16_t op3
)
180 return SVE_ACLE_FUNC(svcdot_lane
,_s64
,,)(op1
, op2
, op3
, 0, 180);