[RISCV] Implement intrinsics for XAndesVBFHCVT (#145634)

This patch implements clang intrinsic support for XAndesVBFHCVT.

The clang intrinsicis for XAndesVBFHCVT is similar to Zvfbfmin, but it
doesn't have mask variants.

The document for the intrinsics can be found at:
https://github.com/andestech/andes-vector-intrinsic-doc/tree/ast-v5_4_0-release-v5/auto-generated/andes-v5

Co-authored-by: Tony Chuan-Yue Yuan <[email protected]>

Author: tclin914

clang/include/clang/Basic/riscv_andes_vector.td

@@ -18,6 +18,52 @@ include "riscv_vector_common.td"
 // Instruction definitions
 //===----------------------------------------------------------------------===//
 
+// Andes Vector BFLOAT16 Conversion Extension (XAndesVBFHCvt)
+
+let RequiredFeatures = ["xandesvbfhcvt"],
+    Log2LMUL = [-2, -1, 0, 1, 2],
+    HasMasked = false,
+    UnMaskedPolicyScheme = HasPassthruOperand in {
+  def nds_vfwcvt_s_bf16 : RVVConvBuiltin<"Fw", "Fwv", "y", "nds_vfwcvt_s">;
+
+  let ManualCodegen = [{
+    {
+      // LLVM intrinsic
+      // Unmasked: (passthru, op0, frm, vl)
+      SmallVector<llvm::Value*, 4> Operands;
+      bool HasMaskedOff = !(PolicyAttrs & RVV_VTA);
+      bool HasRoundModeOp = HasMaskedOff ? Ops.size() == 4 : Ops.size() == 3;
+
+      unsigned Offset = HasMaskedOff ? 1 : 0;
+
+      if (!HasMaskedOff)
+        Operands.push_back(llvm::PoisonValue::get(ResultType));
+      else
+        Operands.push_back(Ops[0]);
+
+      Operands.push_back(Ops[Offset]); // op0
+
+      if (HasRoundModeOp) {
+        Operands.push_back(Ops[Offset + 1]); // frm
+        Operands.push_back(Ops[Offset + 2]); // vl
+      } else {
+        Operands.push_back(ConstantInt::get(Ops[Offset + 1]->getType(), 7)); // frm
+        Operands.push_back(Ops[Offset + 1]); // vl
+      }
+
+      IntrinsicTypes = {ResultType, Ops[Offset]->getType(),
+                        Operands.back()->getType()};
+      llvm::Function *F = CGM.getIntrinsic(ID, IntrinsicTypes);
+      return Builder.CreateCall(F, Operands, "");
+    }
+  }] in {
+    let HasFRMRoundModeOp = 1, Name = "nds_vfncvt_bf16_s", IRName = "nds_vfncvt_bf16_s" in
+      def nds_vfncvt_bf16_s_rm : RVVConvBuiltin<"v", "vFwu", "y", "nds_vfncvt_bf16">;
+
+    def nds_vfncvt_bf16_s : RVVConvBuiltin<"v", "vFw", "y", "nds_vfncvt_bf16">;
+  }
+}
+
 // Andes Vector Packed FP16 Extension (XAndesVPackFPH)
 
 multiclass RVVFPMAD {

clang/lib/Sema/SemaRISCV.cpp

@@ -1438,8 +1438,15 @@ void SemaRISCV::checkRVVTypeSupport(QualType Ty, SourceLocation Loc, Decl *D,
            !FeatureMap.lookup("zvfhmin"))
     Diag(Loc, diag::err_riscv_type_requires_extension, D)
         << Ty << "zvfh or zvfhmin";
-  else if (Info.ElementType->isBFloat16Type() && !FeatureMap.lookup("zvfbfmin"))
-    Diag(Loc, diag::err_riscv_type_requires_extension, D) << Ty << "zvfbfmin";
+  else if (Info.ElementType->isBFloat16Type() &&
+           !FeatureMap.lookup("zvfbfmin") &&
+           !FeatureMap.lookup("xandesvbfhcvt"))
+    if (DeclareAndesVectorBuiltins) {
+      Diag(Loc, diag::err_riscv_type_requires_extension, D)
+          << Ty << "zvfbfmin or xandesvbfhcvt";
+    } else {
+      Diag(Loc, diag::err_riscv_type_requires_extension, D) << Ty << "zvfbfmin";
+    }
   else if (Info.ElementType->isSpecificBuiltinType(BuiltinType::Float) &&
            !FeatureMap.lookup("zve32f"))
     Diag(Loc, diag::err_riscv_type_requires_extension, D) << Ty << "zve32f";

clang/test/CodeGen/RISCV/rvv-intrinsics-autogenerated/non-policy/non-overloaded/nds_vfncvtbf16s.c

@@ -0,0 +1,108 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// REQUIRES: riscv-registered-target
+// RUN: %clang_cc1 -triple riscv64 -target-feature +zve64x \
+// RUN:   -target-feature +xandesvbfhcvt -disable-O0-optnone \
+// RUN:   -emit-llvm %s -o - | opt -S -passes=mem2reg | \
+// RUN:   FileCheck --check-prefix=CHECK-RV64 %s
+
+#include <andes_vector.h>
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 1 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf4
+// CHECK-RV64-SAME: (<vscale x 1 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv1bf16.nxv1f32.i64(<vscale x 1 x bfloat> poison, <vscale x 1 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 1 x bfloat> [[TMP0]]
+//
+vbfloat16mf4_t test_nds_vfncvt_bf16_s_bf16mf4(vfloat32mf2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16mf4(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 2 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf2
+// CHECK-RV64-SAME: (<vscale x 2 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv2bf16.nxv2f32.i64(<vscale x 2 x bfloat> poison, <vscale x 2 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 2 x bfloat> [[TMP0]]
+//
+vbfloat16mf2_t test_nds_vfncvt_bf16_s_bf16mf2(vfloat32m1_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16mf2(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x bfloat> @test_nds_vfncvt_bf16_s_bf16m1
+// CHECK-RV64-SAME: (<vscale x 4 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv4bf16.nxv4f32.i64(<vscale x 4 x bfloat> poison, <vscale x 4 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 4 x bfloat> [[TMP0]]
+//
+vbfloat16m1_t test_nds_vfncvt_bf16_s_bf16m1(vfloat32m2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m1(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 8 x bfloat> @test_nds_vfncvt_bf16_s_bf16m2
+// CHECK-RV64-SAME: (<vscale x 8 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv8bf16.nxv8f32.i64(<vscale x 8 x bfloat> poison, <vscale x 8 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+vbfloat16m2_t test_nds_vfncvt_bf16_s_bf16m2(vfloat32m4_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m2(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 16 x bfloat> @test_nds_vfncvt_bf16_s_bf16m4
+// CHECK-RV64-SAME: (<vscale x 16 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv16bf16.nxv16f32.i64(<vscale x 16 x bfloat> poison, <vscale x 16 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+//
+vbfloat16m4_t test_nds_vfncvt_bf16_s_bf16m4(vfloat32m8_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m4(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 1 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf4_rm
+// CHECK-RV64-SAME: (<vscale x 1 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv1bf16.nxv1f32.i64(<vscale x 1 x bfloat> poison, <vscale x 1 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 1 x bfloat> [[TMP0]]
+//
+vbfloat16mf4_t test_nds_vfncvt_bf16_s_bf16mf4_rm(vfloat32mf2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16mf4_rm(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 2 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf2_rm
+// CHECK-RV64-SAME: (<vscale x 2 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv2bf16.nxv2f32.i64(<vscale x 2 x bfloat> poison, <vscale x 2 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 2 x bfloat> [[TMP0]]
+//
+vbfloat16mf2_t test_nds_vfncvt_bf16_s_bf16mf2_rm(vfloat32m1_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16mf2_rm(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x bfloat> @test_nds_vfncvt_bf16_s_bf16m1_rm
+// CHECK-RV64-SAME: (<vscale x 4 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv4bf16.nxv4f32.i64(<vscale x 4 x bfloat> poison, <vscale x 4 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 4 x bfloat> [[TMP0]]
+//
+vbfloat16m1_t test_nds_vfncvt_bf16_s_bf16m1_rm(vfloat32m2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m1_rm(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 8 x bfloat> @test_nds_vfncvt_bf16_s_bf16m2_rm
+// CHECK-RV64-SAME: (<vscale x 8 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv8bf16.nxv8f32.i64(<vscale x 8 x bfloat> poison, <vscale x 8 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+vbfloat16m2_t test_nds_vfncvt_bf16_s_bf16m2_rm(vfloat32m4_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m2_rm(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 16 x bfloat> @test_nds_vfncvt_bf16_s_bf16m4_rm
+// CHECK-RV64-SAME: (<vscale x 16 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv16bf16.nxv16f32.i64(<vscale x 16 x bfloat> poison, <vscale x 16 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+//
+vbfloat16m4_t test_nds_vfncvt_bf16_s_bf16m4_rm(vfloat32m8_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16_s_bf16m4_rm(vs2, __RISCV_FRM_RNE, vl);
+}

clang/test/CodeGen/RISCV/rvv-intrinsics-autogenerated/non-policy/non-overloaded/nds_vfwcvtsbf16.c

@@ -0,0 +1,58 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// REQUIRES: riscv-registered-target
+// RUN: %clang_cc1 -triple riscv64 -target-feature +v \
+// RUN:   -target-feature +xandesvbfhcvt -disable-O0-optnone \
+// RUN:   -emit-llvm %s -o - | opt -S -passes=mem2reg | \
+// RUN:   FileCheck --check-prefix=CHECK-RV64 %s
+
+#include <andes_vector.h>
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_nds_vfwcvt_s_bf16_f32mf2
+// CHECK-RV64-SAME: (<vscale x 1 x bfloat> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.nds.vfwcvt.s.bf16.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> poison, <vscale x 1 x bfloat> [[VS2]], i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
+//
+vfloat32mf2_t test_nds_vfwcvt_s_bf16_f32mf2(vbfloat16mf4_t vs2, size_t vl) {
+  return __riscv_nds_vfwcvt_s_bf16_f32mf2(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_nds_vfwcvt_s_bf16_f32m1
+// CHECK-RV64-SAME: (<vscale x 2 x bfloat> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.nds.vfwcvt.s.bf16.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> poison, <vscale x 2 x bfloat> [[VS2]], i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
+//
+vfloat32m1_t test_nds_vfwcvt_s_bf16_f32m1(vbfloat16mf2_t vs2, size_t vl) {
+  return __riscv_nds_vfwcvt_s_bf16_f32m1(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_nds_vfwcvt_s_bf16_f32m2
+// CHECK-RV64-SAME: (<vscale x 4 x bfloat> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.nds.vfwcvt.s.bf16.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> poison, <vscale x 4 x bfloat> [[VS2]], i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+vfloat32m2_t test_nds_vfwcvt_s_bf16_f32m2(vbfloat16m1_t vs2, size_t vl) {
+  return __riscv_nds_vfwcvt_s_bf16_f32m2(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_nds_vfwcvt_s_bf16_f32m4
+// CHECK-RV64-SAME: (<vscale x 8 x bfloat> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.nds.vfwcvt.s.bf16.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> poison, <vscale x 8 x bfloat> [[VS2]], i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
+//
+vfloat32m4_t test_nds_vfwcvt_s_bf16_f32m4(vbfloat16m2_t vs2, size_t vl) {
+  return __riscv_nds_vfwcvt_s_bf16_f32m4(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_nds_vfwcvt_s_bf16_f32m8
+// CHECK-RV64-SAME: (<vscale x 16 x bfloat> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.nds.vfwcvt.s.bf16.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> poison, <vscale x 16 x bfloat> [[VS2]], i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
+//
+vfloat32m8_t test_nds_vfwcvt_s_bf16_f32m8(vbfloat16m4_t vs2, size_t vl) {
+  return __riscv_nds_vfwcvt_s_bf16_f32m8(vs2, vl);
+}

clang/test/CodeGen/RISCV/rvv-intrinsics-autogenerated/non-policy/overloaded/nds_vfncvtbf16s.c

@@ -0,0 +1,108 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// REQUIRES: riscv-registered-target
+// RUN: %clang_cc1 -triple riscv64 -target-feature +v \
+// RUN:   -target-feature +xandesvbfhcvt -disable-O0-optnone \
+// RUN:   -emit-llvm %s -o - | opt -S -passes=mem2reg | \
+// RUN:   FileCheck --check-prefix=CHECK-RV64 %s
+
+#include <andes_vector.h>
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 1 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf4
+// CHECK-RV64-SAME: (<vscale x 1 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv1bf16.nxv1f32.i64(<vscale x 1 x bfloat> poison, <vscale x 1 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 1 x bfloat> [[TMP0]]
+//
+vbfloat16mf4_t test_nds_vfncvt_bf16_s_bf16mf4(vfloat32mf2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 2 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf2
+// CHECK-RV64-SAME: (<vscale x 2 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv2bf16.nxv2f32.i64(<vscale x 2 x bfloat> poison, <vscale x 2 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 2 x bfloat> [[TMP0]]
+//
+vbfloat16mf2_t test_nds_vfncvt_bf16_s_bf16mf2(vfloat32m1_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x bfloat> @test_nds_vfncvt_bf16_s_bf16m1
+// CHECK-RV64-SAME: (<vscale x 4 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv4bf16.nxv4f32.i64(<vscale x 4 x bfloat> poison, <vscale x 4 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 4 x bfloat> [[TMP0]]
+//
+vbfloat16m1_t test_nds_vfncvt_bf16_s_bf16m1(vfloat32m2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 8 x bfloat> @test_nds_vfncvt_bf16_s_bf16m2
+// CHECK-RV64-SAME: (<vscale x 8 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv8bf16.nxv8f32.i64(<vscale x 8 x bfloat> poison, <vscale x 8 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+vbfloat16m2_t test_nds_vfncvt_bf16_s_bf16m2(vfloat32m4_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 16 x bfloat> @test_nds_vfncvt_bf16_s_bf16m4
+// CHECK-RV64-SAME: (<vscale x 16 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv16bf16.nxv16f32.i64(<vscale x 16 x bfloat> poison, <vscale x 16 x float> [[VS2]], i64 7, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+//
+vbfloat16m4_t test_nds_vfncvt_bf16_s_bf16m4(vfloat32m8_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 1 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf4_rm
+// CHECK-RV64-SAME: (<vscale x 1 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv1bf16.nxv1f32.i64(<vscale x 1 x bfloat> poison, <vscale x 1 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 1 x bfloat> [[TMP0]]
+//
+vbfloat16mf4_t test_nds_vfncvt_bf16_s_bf16mf4_rm(vfloat32mf2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 2 x bfloat> @test_nds_vfncvt_bf16_s_bf16mf2_rm
+// CHECK-RV64-SAME: (<vscale x 2 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv2bf16.nxv2f32.i64(<vscale x 2 x bfloat> poison, <vscale x 2 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 2 x bfloat> [[TMP0]]
+//
+vbfloat16mf2_t test_nds_vfncvt_bf16_s_bf16mf2_rm(vfloat32m1_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x bfloat> @test_nds_vfncvt_bf16_s_bf16m1_rm
+// CHECK-RV64-SAME: (<vscale x 4 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv4bf16.nxv4f32.i64(<vscale x 4 x bfloat> poison, <vscale x 4 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 4 x bfloat> [[TMP0]]
+//
+vbfloat16m1_t test_nds_vfncvt_bf16_s_bf16m1_rm(vfloat32m2_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 8 x bfloat> @test_nds_vfncvt_bf16_s_bf16m2_rm
+// CHECK-RV64-SAME: (<vscale x 8 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv8bf16.nxv8f32.i64(<vscale x 8 x bfloat> poison, <vscale x 8 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+vbfloat16m2_t test_nds_vfncvt_bf16_s_bf16m2_rm(vfloat32m4_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, __RISCV_FRM_RNE, vl);
+}
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 16 x bfloat> @test_nds_vfncvt_bf16_s_bf16m4_rm
+// CHECK-RV64-SAME: (<vscale x 16 x float> [[VS2:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x bfloat> @llvm.riscv.nds.vfncvt.bf16.s.nxv16bf16.nxv16f32.i64(<vscale x 16 x bfloat> poison, <vscale x 16 x float> [[VS2]], i64 0, i64 [[VL]])
+// CHECK-RV64-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+//
+vbfloat16m4_t test_nds_vfncvt_bf16_s_bf16m4_rm(vfloat32m8_t vs2, size_t vl) {
+  return __riscv_nds_vfncvt_bf16(vs2, __RISCV_FRM_RNE, vl);
+}