Skip to content

Commit

Permalink
[SveEmitter] Add builtins for SVE2 uniform DSP operations
Browse files Browse the repository at this point in the history
This patch adds builtins for:
- svqadd, svhadd, svrhadd
- svqsub, svhsub, svqusbr, svhsubr
- svqabs
- svqneg
- svrecpe
- svrsqrte
  • Loading branch information
sdesmalen-arm committed May 7, 2020
1 parent 35de496 commit b32d14c
Show file tree
Hide file tree
Showing 12 changed files with 4,382 additions and 0 deletions.
26 changes: 26 additions & 0 deletions clang/include/clang/Basic/arm_sve.td
Expand Up @@ -1248,6 +1248,32 @@ def SVWHILEHS_U32 : SInst<"svwhilege_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNon
def SVWHILEHS_U64 : SInst<"svwhilege_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehs", [IsOverloadWhile]>;
}

////////////////////////////////////////////////////////////////////////////////
// SVE2 - Uniform DSP operations

let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
defm SVQADD_S : SInstZPZZ<"svqadd", "csli", "aarch64_sve_sqadd">;
defm SVQADD_U : SInstZPZZ<"svqadd", "UcUsUiUl", "aarch64_sve_uqadd">;
defm SVHADD_S : SInstZPZZ<"svhadd", "csli", "aarch64_sve_shadd">;
defm SVHADD_U : SInstZPZZ<"svhadd", "UcUsUiUl", "aarch64_sve_uhadd">;
defm SVRHADD_S : SInstZPZZ<"svrhadd", "csli", "aarch64_sve_srhadd">;
defm SVRHADD_U : SInstZPZZ<"svrhadd", "UcUsUiUl", "aarch64_sve_urhadd">;

defm SVQSUB_S : SInstZPZZ<"svqsub", "csli", "aarch64_sve_sqsub">;
defm SVQSUB_U : SInstZPZZ<"svqsub", "UcUsUiUl", "aarch64_sve_uqsub">;
defm SVQSUBR_S : SInstZPZZ<"svqsubr", "csli", "aarch64_sve_sqsubr">;
defm SVQSUBR_U : SInstZPZZ<"svqsubr", "UcUsUiUl", "aarch64_sve_uqsubr">;
defm SVHSUB_S : SInstZPZZ<"svhsub", "csli", "aarch64_sve_shsub">;
defm SVHSUB_U : SInstZPZZ<"svhsub", "UcUsUiUl", "aarch64_sve_uhsub">;
defm SVHSUBR_S : SInstZPZZ<"svhsubr", "csli", "aarch64_sve_shsubr">;
defm SVHSUBR_U : SInstZPZZ<"svhsubr", "UcUsUiUl", "aarch64_sve_uhsubr">;

defm SVQABS : SInstZPZ<"svqabs", "csil", "aarch64_sve_sqabs">;
defm SVQNEG : SInstZPZ<"svqneg", "csil", "aarch64_sve_sqneg">;
defm SVRECPE : SInstZPZ<"svrecpe", "Ui", "aarch64_sve_urecpe">;
defm SVRSQRTE : SInstZPZ<"svrsqrte", "Ui", "aarch64_sve_ursqrte">;
}

////////////////////////////////////////////////////////////////////////////////
// SVE2 - Non-temporal gather/scatter
let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
Expand Down
569 changes: 569 additions & 0 deletions clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hadd.c

Large diffs are not rendered by default.

569 changes: 569 additions & 0 deletions clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hsub.c

Large diffs are not rendered by default.

568 changes: 568 additions & 0 deletions clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_hsubr.c

Large diffs are not rendered by default.

142 changes: 142 additions & 0 deletions clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_qabs.c
@@ -0,0 +1,142 @@
// 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
// 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
// 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
// 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

#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_svqabs_s8_z(svbool_t pg, svint8_t op)
{
// CHECK-LABEL: test_svqabs_s8_z
// 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)
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_z'}}
return SVE_ACLE_FUNC(svqabs,_s8,_z,)(pg, op);
}

svint16_t test_svqabs_s16_z(svbool_t pg, svint16_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_z'}}
return SVE_ACLE_FUNC(svqabs,_s16,_z,)(pg, op);
}

svint32_t test_svqabs_s32_z(svbool_t pg, svint32_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_z'}}
return SVE_ACLE_FUNC(svqabs,_s32,_z,)(pg, op);
}

svint64_t test_svqabs_s64_z(svbool_t pg, svint64_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_z'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_z'}}
return SVE_ACLE_FUNC(svqabs,_s64,_z,)(pg, op);
}

svint8_t test_svqabs_s8_m(svint8_t inactive, svbool_t pg, svint8_t op)
{
// CHECK-LABEL: test_svqabs_s8_m
// 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)
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_m'}}
return SVE_ACLE_FUNC(svqabs,_s8,_m,)(inactive, pg, op);
}

svint16_t test_svqabs_s16_m(svint16_t inactive, svbool_t pg, svint16_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_m'}}
return SVE_ACLE_FUNC(svqabs,_s16,_m,)(inactive, pg, op);
}

svint32_t test_svqabs_s32_m(svint32_t inactive, svbool_t pg, svint32_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_m'}}
return SVE_ACLE_FUNC(svqabs,_s32,_m,)(inactive, pg, op);
}

svint64_t test_svqabs_s64_m(svint64_t inactive, svbool_t pg, svint64_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_m'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_m'}}
return SVE_ACLE_FUNC(svqabs,_s64,_m,)(inactive, pg, op);
}

svint8_t test_svqabs_s8_x(svbool_t pg, svint8_t op)
{
// CHECK-LABEL: test_svqabs_s8_x
// 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)
// CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s8_x'}}
return SVE_ACLE_FUNC(svqabs,_s8,_x,)(pg, op);
}

svint16_t test_svqabs_s16_x(svbool_t pg, svint16_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s16_x'}}
return SVE_ACLE_FUNC(svqabs,_s16,_x,)(pg, op);
}

svint32_t test_svqabs_s32_x(svbool_t pg, svint32_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s32_x'}}
return SVE_ACLE_FUNC(svqabs,_s32,_x,)(pg, op);
}

svint64_t test_svqabs_s64_x(svbool_t pg, svint64_t op)
{
// CHECK-LABEL: test_svqabs_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.sqabs.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]]
// overload-warning@+2 {{implicit declaration of function 'svqabs_x'}}
// expected-warning@+1 {{implicit declaration of function 'svqabs_s64_x'}}
return SVE_ACLE_FUNC(svqabs,_s64,_x,)(pg, op);
}

0 comments on commit b32d14c

Please sign in to comment.