Skip to content

Revert "Revert "[SVE2.1][Clang][LLVM]Add BFloat16 builtin in Clang an… #73110

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 22, 2023
Merged

Revert "Revert "[SVE2.1][Clang][LLVM]Add BFloat16 builtin in Clang an… #73110

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 17 additions & 1 deletion clang/include/clang/Basic/arm_sve.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2039,7 +2039,23 @@ def SVCNTP_COUNT : SInst<"svcntp_{d}", "n}i", "QcQsQiQl", MergeNone, "aarch64_sv
defm SVREVD : SInstZPZ<"svrevd", "csilUcUsUiUl", "aarch64_sve_revd">;
}

////////////////////////////////////////////////////////////////////////////////

let TargetGuard = "sve2p1,b16b16" in {
defm SVMUL_BF : SInstZPZZ<"svmul", "b", "aarch64_sve_fmul", "aarch64_sve_fmul_u">;
defm SVADD_BF : SInstZPZZ<"svadd", "b", "aarch64_sve_fadd", "aarch64_sve_fadd_u">;
defm SVSUB_BF : SInstZPZZ<"svsub", "b", "aarch64_sve_fsub", "aarch64_sve_fsub_u">;
defm SVMAXNM_BF : SInstZPZZ<"svmaxnm","b", "aarch64_sve_fmaxnm", "aarch64_sve_fmaxnm_u">;
defm SVMINNM_BF : SInstZPZZ<"svminnm","b", "aarch64_sve_fminnm", "aarch64_sve_fminnm_u">;
defm SVMAX_BF : SInstZPZZ<"svmax", "b", "aarch64_sve_fmax", "aarch64_sve_fmax_u">;
defm SVMIN_BF : SInstZPZZ<"svmin", "b", "aarch64_sve_fmin", "aarch64_sve_fmin_u">;
defm SVMLA_BF : SInstZPZZZ<"svmla", "b", "aarch64_sve_fmla", "aarch64_sve_fmla_u", []>;
defm SVMLS_BF : SInstZPZZZ<"svmls", "b", "aarch64_sve_fmls", "aarch64_sve_fmls_u", []>;
def SVMLA_LANE_BF : SInst<"svmla_lane[_{d}]", "ddddi", "b", MergeNone, "aarch64_sve_fmla_lane", [], [ImmCheck<3, ImmCheckLaneIndex, 2>]>;
def SVMLS_LANE_BF : SInst<"svmls_lane[_{d}]", "ddddi", "b", MergeNone, "aarch64_sve_fmls_lane", [], [ImmCheck<3, ImmCheckLaneIndex, 2>]>;
def SVMUL_LANE_BF : SInst<"svmul_lane[_{d}]", "dddi", "b", MergeNone, "aarch64_sve_fmul_lane", [], [ImmCheck<2, ImmCheckLaneIndex, 1>]>;
def SVFCLAMP_BF : SInst<"svclamp[_{d}]", "dddd", "b", MergeNone, "aarch64_sve_fclamp", [], []>;
} //sve2p1,b16b16

// SME2

// SME intrinsics which operate only on vectors and do not require ZA should be added here,
Expand Down
133 changes: 133 additions & 0 deletions clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_bfadd.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,133 @@
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// REQUIRES: aarch64-registered-target
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -o /dev/null %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) A1##A3
#else
#define SVE_ACLE_FUNC(A1,A2,A3) A1##A2##A3
#endif

// CHECK-LABEL: @test_svadd_bf16_m(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z17test_svadd_bf16_mu10__SVBool_tu14__SVBfloat16_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
svbfloat16_t test_svadd_bf16_m(svbool_t pg, svbfloat16_t op1, svbfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _bf16, _m)(pg, op1, op2);
}

// CHECK-LABEL: @test_svadd_bf16_z(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = select <vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> zeroinitializer
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z17test_svadd_bf16_zu10__SVBool_tu14__SVBfloat16_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = select <vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> zeroinitializer
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
svbfloat16_t test_svadd_bf16_z(svbool_t pg, svbfloat16_t op1, svbfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _bf16, _z)(pg, op1, op2);
}

// CHECK-LABEL: @test_svadd_bf16_x(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.u.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z17test_svadd_bf16_xu10__SVBool_tu14__SVBfloat16_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.u.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
svbfloat16_t test_svadd_bf16_x(svbool_t pg, svbfloat16_t op1, svbfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _bf16, _x)(pg, op1, op2);
}

// CHECK-LABEL: @test_svadd_bf16_n_m(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z19test_svadd_bf16_n_mu10__SVBool_tu14__SVBfloat16_tu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CPP-CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
svbfloat16_t test_svadd_bf16_n_m(svbool_t pg, svbfloat16_t op1, bfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _n_bf16, _m)(pg, op1, op2);
}

// CHECK-LABEL: @test_svadd_bf16_n_z(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CHECK-NEXT: [[TMP1:%.*]] = select <vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> zeroinitializer
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svadd_bf16_n_zu10__SVBool_tu14__SVBfloat16_tu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CPP-CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CPP-CHECK-NEXT: [[TMP1:%.*]] = select <vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> zeroinitializer
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
svbfloat16_t test_svadd_bf16_n_z(svbool_t pg, svbfloat16_t op1, bfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _n_bf16, _z)(pg, op1, op2);
}

// CHECK-LABEL: @test_svadd_bf16_n_x(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.u.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z19test_svadd_bf16_n_xu10__SVBool_tu14__SVBfloat16_tu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x bfloat> poison, bfloat [[OP2:%.*]], i64 0
// CPP-CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x bfloat> [[DOTSPLATINSERT]], <vscale x 8 x bfloat> poison, <vscale x 8 x i32> zeroinitializer
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fadd.u.nxv8bf16(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[DOTSPLAT]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP1]]
//
svbfloat16_t test_svadd_bf16_n_x(svbool_t pg, svbfloat16_t op1, bfloat16_t op2)
{
return SVE_ACLE_FUNC(svadd, _n_bf16, _x)(pg, op1, op2);
}
31 changes: 31 additions & 0 deletions clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_bfclamp.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,31 @@
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// REQUIRES: aarch64-registered-target
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -target-feature +b16b16 -disable-O0-optnone -Werror -Wall -o /dev/null %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) A1##A3
#else
#define SVE_ACLE_FUNC(A1,A2,A3) A1##A2##A3
#endif

// CHECK-LABEL: @test_svclamp_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fclamp.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], <vscale x 8 x bfloat> [[OP3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z17test_svclamp_bf16u14__SVBfloat16_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fclamp.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], <vscale x 8 x bfloat> [[OP3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
svbfloat16_t test_svclamp_bf16(svbfloat16_t op1, svbfloat16_t op2, svbfloat16_t op3)
{
return SVE_ACLE_FUNC(svclamp, _bf16,)(op1, op2, op3);
}

Loading