Skip to content

Commit b32d14c

Browse files
sdesmalen-armsdesmalen-arm
committed
[SveEmitter] Add builtins for SVE2 uniform DSP operations
This patch adds builtins for: - svqadd, svhadd, svrhadd - svqsub, svhsub, svqusbr, svhsubr - svqabs - svqneg - svrecpe - svrsqrte
1 parent 35de496 commit b32d14c

12 files changed

Lines changed: 4382 additions & 0 deletions

File tree

‎clang/include/clang/Basic/arm_sve.td‎

Lines changed: 26 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1248,6 +1248,32 @@ def SVWHILEHS_U32 : SInst<"svwhilege_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNon
12481248
def SVWHILEHS_U64 : SInst<"svwhilege_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehs", [IsOverloadWhile]>;
12491249
}
12501250

1251+
////////////////////////////////////////////////////////////////////////////////
1252+
// SVE2 - Uniform DSP operations
1253+
1254+
let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
1255+
defm SVQADD_S : SInstZPZZ<"svqadd", "csli", "aarch64_sve_sqadd">;
1256+
defm SVQADD_U : SInstZPZZ<"svqadd", "UcUsUiUl", "aarch64_sve_uqadd">;
1257+
defm SVHADD_S : SInstZPZZ<"svhadd", "csli", "aarch64_sve_shadd">;
1258+
defm SVHADD_U : SInstZPZZ<"svhadd", "UcUsUiUl", "aarch64_sve_uhadd">;
1259+
defm SVRHADD_S : SInstZPZZ<"svrhadd", "csli", "aarch64_sve_srhadd">;
1260+
defm SVRHADD_U : SInstZPZZ<"svrhadd", "UcUsUiUl", "aarch64_sve_urhadd">;
1261+
1262+
defm SVQSUB_S : SInstZPZZ<"svqsub", "csli", "aarch64_sve_sqsub">;
1263+
defm SVQSUB_U : SInstZPZZ<"svqsub", "UcUsUiUl", "aarch64_sve_uqsub">;
1264+
defm SVQSUBR_S : SInstZPZZ<"svqsubr", "csli", "aarch64_sve_sqsubr">;
1265+
defm SVQSUBR_U : SInstZPZZ<"svqsubr", "UcUsUiUl", "aarch64_sve_uqsubr">;
1266+
defm SVHSUB_S : SInstZPZZ<"svhsub", "csli", "aarch64_sve_shsub">;
1267+
defm SVHSUB_U : SInstZPZZ<"svhsub", "UcUsUiUl", "aarch64_sve_uhsub">;
1268+
defm SVHSUBR_S : SInstZPZZ<"svhsubr", "csli", "aarch64_sve_shsubr">;
1269+
defm SVHSUBR_U : SInstZPZZ<"svhsubr", "UcUsUiUl", "aarch64_sve_uhsubr">;
1270+
1271+
defm SVQABS : SInstZPZ<"svqabs", "csil", "aarch64_sve_sqabs">;
1272+
defm SVQNEG : SInstZPZ<"svqneg", "csil", "aarch64_sve_sqneg">;
1273+
defm SVRECPE : SInstZPZ<"svrecpe", "Ui", "aarch64_sve_urecpe">;
1274+
defm SVRSQRTE : SInstZPZ<"svrsqrte", "Ui", "aarch64_sve_ursqrte">;
1275+
}
1276+
12511277
////////////////////////////////////////////////////////////////////////////////
12521278
// SVE2 - Non-temporal gather/scatter
12531279
let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {

‎clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hadd.c‎

Lines changed: 569 additions & 0 deletions
Large diffs are not rendered by default.

‎clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hsub.c‎

Lines changed: 569 additions & 0 deletions
Large diffs are not rendered by default.

‎clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hsubr.c‎

Lines changed: 568 additions & 0 deletions
Large diffs are not rendered by default.

‎clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_qabs.c‎

Lines changed: 142 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,142 @@
1+
// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
2+
// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
3+
// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
4+
// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
5+
6+
#include <arm_sve.h>
7+
8+
#ifdef SVE_OVERLOADED_FORMS
9+
// A simple used,unused... macro, long enough to represent any SVE builtin.
10+
#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
11+
#else
12+
#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
13+
#endif
14+
15+
svint8_t test_svqabs_s8_z(svbool_t pg, svint8_t op)
16+
{
17+
// CHECK-LABEL: test_svqabs_s8_z
18+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.sqabs.nxv16i8(<vscale x 16 x i8> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op)
19+
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
20+
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
21+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_z'}}
22+
return SVE_ACLE_FUNC(svqabs,_s8,_z,)(pg, op);
23+
}
24+
25+
svint16_t test_svqabs_s16_z(svbool_t pg, svint16_t op)
26+
{
27+
// CHECK-LABEL: test_svqabs_s16_z
28+
// CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
29+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sqabs.nxv8i16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op)
30+
// CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
31+
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
32+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_z'}}
33+
return SVE_ACLE_FUNC(svqabs,_s16,_z,)(pg, op);
34+
}
35+
36+
svint32_t test_svqabs_s32_z(svbool_t pg, svint32_t op)
37+
{
38+
// CHECK-LABEL: test_svqabs_s32_z
39+
// CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
40+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sqabs.nxv4i32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
41+
// CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
42+
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
43+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_z'}}
44+
return SVE_ACLE_FUNC(svqabs,_s32,_z,)(pg, op);
45+
}
46+
47+
svint64_t test_svqabs_s64_z(svbool_t pg, svint64_t op)
48+
{
49+
// CHECK-LABEL: test_svqabs_s64_z
50+
// CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
51+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sqabs.nxv2i64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
52+
// CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
53+
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
54+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_z'}}
55+
return SVE_ACLE_FUNC(svqabs,_s64,_z,)(pg, op);
56+
}
57+
58+
svint8_t test_svqabs_s8_m(svint8_t inactive, svbool_t pg, svint8_t op)
59+
{
60+
// CHECK-LABEL: test_svqabs_s8_m
61+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.sqabs.nxv16i8(<vscale x 16 x i8> %inactive, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op)
62+
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
63+
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
64+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_m'}}
65+
return SVE_ACLE_FUNC(svqabs,_s8,_m,)(inactive, pg, op);
66+
}
67+
68+
svint16_t test_svqabs_s16_m(svint16_t inactive, svbool_t pg, svint16_t op)
69+
{
70+
// CHECK-LABEL: test_svqabs_s16_m
71+
// CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
72+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sqabs.nxv8i16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op)
73+
// CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
74+
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
75+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_m'}}
76+
return SVE_ACLE_FUNC(svqabs,_s16,_m,)(inactive, pg, op);
77+
}
78+
79+
svint32_t test_svqabs_s32_m(svint32_t inactive, svbool_t pg, svint32_t op)
80+
{
81+
// CHECK-LABEL: test_svqabs_s32_m
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 i32> @llvm.aarch64.sve.sqabs.nxv4i32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
84+
// CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
85+
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
86+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_m'}}
87+
return SVE_ACLE_FUNC(svqabs,_s32,_m,)(inactive, pg, op);
88+
}
89+
90+
svint64_t test_svqabs_s64_m(svint64_t inactive, svbool_t pg, svint64_t op)
91+
{
92+
// CHECK-LABEL: test_svqabs_s64_m
93+
// CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
94+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sqabs.nxv2i64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
95+
// CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
96+
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
97+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_m'}}
98+
return SVE_ACLE_FUNC(svqabs,_s64,_m,)(inactive, pg, op);
99+
}
100+
101+
svint8_t test_svqabs_s8_x(svbool_t pg, svint8_t op)
102+
{
103+
// CHECK-LABEL: test_svqabs_s8_x
104+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.sqabs.nxv16i8(<vscale x 16 x i8> undef, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %op)
105+
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
106+
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
107+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_x'}}
108+
return SVE_ACLE_FUNC(svqabs,_s8,_x,)(pg, op);
109+
}
110+
111+
svint16_t test_svqabs_s16_x(svbool_t pg, svint16_t op)
112+
{
113+
// CHECK-LABEL: test_svqabs_s16_x
114+
// CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
115+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.sqabs.nxv8i16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x i16> %op)
116+
// CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
117+
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
118+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_x'}}
119+
return SVE_ACLE_FUNC(svqabs,_s16,_x,)(pg, op);
120+
}
121+
122+
svint32_t test_svqabs_s32_x(svbool_t pg, svint32_t op)
123+
{
124+
// CHECK-LABEL: test_svqabs_s32_x
125+
// CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
126+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.sqabs.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
127+
// CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
128+
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
129+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_x'}}
130+
return SVE_ACLE_FUNC(svqabs,_s32,_x,)(pg, op);
131+
}
132+
133+
svint64_t test_svqabs_s64_x(svbool_t pg, svint64_t op)
134+
{
135+
// CHECK-LABEL: test_svqabs_s64_x
136+
// CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
137+
// CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.sqabs.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
138+
// CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
139+
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
140+
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_x'}}
141+
return SVE_ACLE_FUNC(svqabs,_s64,_x,)(pg, op);
142+
}

0 commit comments

Comments
 (0)