[clang] [clang-tools-extra] [llvm] [SVE2.1][Clang][LLVM]Add BFloat16 builtin in Clang and LLVM intrinisc (PR #70362)

[via cfe-commits]

================
@@ -0,0 +1,61 @@
+// 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 -target-feature +sve -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 -target-feature +sve -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 -target-feature +sve -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 -target-feature +sve -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 -target-feature +sve -target-feature -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_svmul_lane_bf16_idx1(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 1)
+// CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z25test_svmul_lane_bf16_idx1u14__SVBFloat16_tu14__SVBFloat16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 1)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svmul_lane_bf16_idx1(svbfloat16_t op1, svbfloat16_t op2)
+{
+  return SVE_ACLE_FUNC(svmul_lane, _bf16, )(op1, op2, 1);
+}
+
+// CHECK-LABEL: @test_svmul_lane_bf16_idx3(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 3)
+// CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z25test_svmul_lane_bf16_idx3u14__SVBFloat16_tu14__SVBFloat16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 3)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svmul_lane_bf16_idx3(svbfloat16_t op1, svbfloat16_t op2)
+{
+  return SVE_ACLE_FUNC(svmul_lane, _bf16, )(op1, op2, 3);
+}
+
+// CHECK-LABEL: @test_svmul_lane_bf16_idx7(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 7)
+// CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z25test_svmul_lane_bf16_idx7u14__SVBFloat16_tu14__SVBFloat16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmul.lane.nxv8bf16(<vscale x 8 x bfloat> [[OP1:%.*]], <vscale x 8 x bfloat> [[OP2:%.*]], i32 7)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svmul_lane_bf16_idx7(svbfloat16_t op1, svbfloat16_t op2)
+{
+  return SVE_ACLE_FUNC(svmul_lane, _bf16, )(op1, op2, 7);
----------------
CarolineConcatto wrote:

This is needs to be a constant. If you check the arm_sve_sme.td you will see that this is the definitions of these intrinsics:
ddddi - {d} stands for default and {i} for  constant uint64_t.
The index lane always needs to be a constant.


https://github.com/llvm/llvm-project/pull/70362