[RISCV][VP] Add RVV codegen for vp.select

[victor-eds]

Summary

Lower vp.select instrinsic to VSELECT_VL.

[victor-eds]

Test

[rogfer01]

Thanks for the patch @victor-eds!

LGTM.

After this change lands, the only lowering missing for vp.select will be for i1 vectors. AFAICT rvv doesn't have specific instructions for that but a sequence of vm<op>.mm and something like this (or a better sequence) https://graphics.stanford.edu/~seander/bithacks.html#MaskedMerge should do. We can't fully honour the vl in that case, but I think this is OK as VP doesn't demand us to do that.

(@craig.topper reading the spec I realised that vmerge (perhaps others?) doesn't seem to care about ma/mu only ta/tu so maybe this can be exploited when inserting vsetvl)

llvm/include/llvm/IR/VPIntrinsics.def

 
 #undef HELPER_REGISTER_REDUCTION_SEQ_VP
 
 ///// } Reduction
 
 ///// Shuffles {
 
 // llvm.vp.select(mask,on_true,on_false,vlen)
-BEGIN_REGISTER_VP_INTRINSIC(vp_select, 0, 3)
-// BEGIN_REGISTER_VP_SDNODE(VP_SELECT, -1, vp_select, 0, 4)
-// END_REGISTER_CASES(vp_select, VP_SELECT)
-END_REGISTER_VP_INTRINSIC(vp_select)
+BEGIN_REGISTER_VP(vp_select, 0, 3, VP_SELECT, -1)
+VP_PROPERTY_FUNCTIONAL_OPC(Select)
+END_REGISTER_VP(vp_select, VP_SELECT)
 
 BEGIN_REGISTER_VP(experimental_vp_splice, 3, 5, EXPERIMENTAL_VP_SPLICE, -1)
 END_REGISTER_VP(experimental_vp_splice, EXPERIMENTAL_VP_SPLICE)
 
 ///// } Shuffles
 
 #undef BEGIN_REGISTER_VP
 #undef BEGIN_REGISTER_VP_INTRINSIC

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

 //===-- RISCVISelLowering.cpp - RISCV DAG Lowering Implementation  --------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
if (Subtarget.hasVInstructions()) {
 
     static const unsigned IntegerVPOps[] = {
         ISD::VP_ADD,         ISD::VP_SUB,         ISD::VP_MUL,
         ISD::VP_SDIV,        ISD::VP_UDIV,        ISD::VP_SREM,
         ISD::VP_UREM,        ISD::VP_AND,         ISD::VP_OR,
         ISD::VP_XOR,         ISD::VP_ASHR,        ISD::VP_LSHR,
         ISD::VP_SHL,         ISD::VP_REDUCE_ADD,  ISD::VP_REDUCE_AND,
         ISD::VP_REDUCE_OR,   ISD::VP_REDUCE_XOR,  ISD::VP_REDUCE_SMAX,
-        ISD::VP_REDUCE_SMIN, ISD::VP_REDUCE_UMAX, ISD::VP_REDUCE_UMIN};
+        ISD::VP_REDUCE_SMIN, ISD::VP_REDUCE_UMAX, ISD::VP_REDUCE_UMIN,
+        ISD::VP_SELECT};
 
     static const unsigned FloatingPointVPOps[] = {
         ISD::VP_FADD,        ISD::VP_FSUB,        ISD::VP_FMUL,
         ISD::VP_FDIV,        ISD::VP_REDUCE_FADD, ISD::VP_REDUCE_SEQ_FADD,
-        ISD::VP_REDUCE_FMIN, ISD::VP_REDUCE_FMAX};
+        ISD::VP_REDUCE_FMIN, ISD::VP_REDUCE_FMAX, ISD::VP_SELECT};
 
     if (!Subtarget.is64Bit()) {
       // We must custom-lower certain vXi64 operations on RV32 due to the vector
       // element type being illegal.
       setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::i64, Custom);
       setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::i64, Custom);
 
       setOperationAction(ISD::VECREDUCE_ADD, MVT::i64, Custom);
case ISD::VP_GATHER:
     return lowerMaskedGather(Op, DAG);
   case ISD::MSCATTER:
   case ISD::VP_SCATTER:
     return lowerMaskedScatter(Op, DAG);
   case ISD::FLT_ROUNDS_:
     return lowerGET_ROUNDING(Op, DAG);
   case ISD::SET_ROUNDING:
     return lowerSET_ROUNDING(Op, DAG);
+  case ISD::VP_SELECT:
+    return lowerVPOp(Op, DAG, RISCVISD::VSELECT_VL);
   case ISD::VP_ADD:
     return lowerVPOp(Op, DAG, RISCVISD::ADD_VL);
   case ISD::VP_SUB:
     return lowerVPOp(Op, DAG, RISCVISD::SUB_VL);
   case ISD::VP_MUL:
     return lowerVPOp(Op, DAG, RISCVISD::MUL_VL);
   case ISD::VP_SDIV:
     return lowerVPOp(Op, DAG, RISCVISD::SDIV_VL);

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+experimental-zfh,+experimental-v -target-abi=ilp32d -riscv-v-vector-bits-min=128 \
+; RUN:   -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+experimental-zfh,+experimental-v -target-abi=lp64d -riscv-v-vector-bits-min=128 \
+; RUN:   -verify-machineinstrs < %s | FileCheck %s
+
+declare <2 x i8> @llvm.vp.select.v2i8(<2 x i1>, <2 x i8>, <2 x i8>, i32)
+
+define <2 x i8> @select_v2i8(<2 x i1> %a, <2 x i8> %b, <2 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x i8> @llvm.vp.select.v2i8(<2 x i1> %a, <2 x i8> %b, <2 x i8> %c, i32 %evl)
+  ret <2 x i8> %v
+}
+
+declare <4 x i8> @llvm.vp.select.v4i8(<4 x i1>, <4 x i8>, <4 x i8>, i32)
+
+define <4 x i8> @select_v4i8(<4 x i1> %a, <4 x i8> %b, <4 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.select.v4i8(<4 x i1> %a, <4 x i8> %b, <4 x i8> %c, i32 %evl)
+  ret <4 x i8> %v
+}
+
+declare <8 x i8> @llvm.vp.select.v8i8(<8 x i1>, <8 x i8>, <8 x i8>, i32)
+
+define <8 x i8> @select_v8i8(<8 x i1> %a, <8 x i8> %b, <8 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x i8> @llvm.vp.select.v8i8(<8 x i1> %a, <8 x i8> %b, <8 x i8> %c, i32 %evl)
+  ret <8 x i8> %v
+}
+
+declare <16 x i8> @llvm.vp.select.v16i8(<16 x i1>, <16 x i8>, <16 x i8>, i32)
+
+define <16 x i8> @select_v16i8(<16 x i1> %a, <16 x i8> %b, <16 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x i8> @llvm.vp.select.v16i8(<16 x i1> %a, <16 x i8> %b, <16 x i8> %c, i32 %evl)
+  ret <16 x i8> %v
+}
+
+declare <2 x i16> @llvm.vp.select.v2i16(<2 x i1>, <2 x i16>, <2 x i16>, i32)
+
+define <2 x i16> @select_v2i16(<2 x i1> %a, <2 x i16> %b, <2 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x i16> @llvm.vp.select.v2i16(<2 x i1> %a, <2 x i16> %b, <2 x i16> %c, i32 %evl)
+  ret <2 x i16> %v
+}
+
+declare <4 x i16> @llvm.vp.select.v4i16(<4 x i1>, <4 x i16>, <4 x i16>, i32)
+
+define <4 x i16> @select_v4i16(<4 x i1> %a, <4 x i16> %b, <4 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.select.v4i16(<4 x i1> %a, <4 x i16> %b, <4 x i16> %c, i32 %evl)
+  ret <4 x i16> %v
+}
+
+declare <8 x i16> @llvm.vp.select.v8i16(<8 x i1>, <8 x i16>, <8 x i16>, i32)
+
+define <8 x i16> @select_v8i16(<8 x i1> %a, <8 x i16> %b, <8 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x i16> @llvm.vp.select.v8i16(<8 x i1> %a, <8 x i16> %b, <8 x i16> %c, i32 %evl)
+  ret <8 x i16> %v
+}
+
+declare <16 x i16> @llvm.vp.select.v16i16(<16 x i1>, <16 x i16>, <16 x i16>, i32)
+
+define <16 x i16> @select_v16i16(<16 x i1> %a, <16 x i16> %b, <16 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x i16> @llvm.vp.select.v16i16(<16 x i1> %a, <16 x i16> %b, <16 x i16> %c, i32 %evl)
+  ret <16 x i16> %v
+}
+
+declare <2 x i32> @llvm.vp.select.v2i32(<2 x i1>, <2 x i32>, <2 x i32>, i32)
+
+define <2 x i32> @select_v2i32(<2 x i1> %a, <2 x i32> %b, <2 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x i32> @llvm.vp.select.v2i32(<2 x i1> %a, <2 x i32> %b, <2 x i32> %c, i32 %evl)
+  ret <2 x i32> %v
+}
+
+declare <4 x i32> @llvm.vp.select.v4i32(<4 x i1>, <4 x i32>, <4 x i32>, i32)
+
+define <4 x i32> @select_v4i32(<4 x i1> %a, <4 x i32> %b, <4 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.select.v4i32(<4 x i1> %a, <4 x i32> %b, <4 x i32> %c, i32 %evl)
+  ret <4 x i32> %v
+}
+
+declare <8 x i32> @llvm.vp.select.v8i32(<8 x i1>, <8 x i32>, <8 x i32>, i32)
+
+define <8 x i32> @select_v8i32(<8 x i1> %a, <8 x i32> %b, <8 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x i32> @llvm.vp.select.v8i32(<8 x i1> %a, <8 x i32> %b, <8 x i32> %c, i32 %evl)
+  ret <8 x i32> %v
+}
+
+declare <16 x i32> @llvm.vp.select.v16i32(<16 x i1>, <16 x i32>, <16 x i32>, i32)
+
+define <16 x i32> @select_v16i32(<16 x i1> %a, <16 x i32> %b, <16 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x i32> @llvm.vp.select.v16i32(<16 x i1> %a, <16 x i32> %b, <16 x i32> %c, i32 %evl)
+  ret <16 x i32> %v
+}
+
+declare <2 x i64> @llvm.vp.select.v2i64(<2 x i1>, <2 x i64>, <2 x i64>, i32)
+
+define <2 x i64> @select_v2i64(<2 x i1> %a, <2 x i64> %b, <2 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x i64> @llvm.vp.select.v2i64(<2 x i1> %a, <2 x i64> %b, <2 x i64> %c, i32 %evl)
+  ret <2 x i64> %v
+}
+
+declare <4 x i64> @llvm.vp.select.v4i64(<4 x i1>, <4 x i64>, <4 x i64>, i32)
+
+define <4 x i64> @select_v4i64(<4 x i1> %a, <4 x i64> %b, <4 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.select.v4i64(<4 x i1> %a, <4 x i64> %b, <4 x i64> %c, i32 %evl)
+  ret <4 x i64> %v
+}
+
+declare <8 x i64> @llvm.vp.select.v8i64(<8 x i1>, <8 x i64>, <8 x i64>, i32)
+
+define <8 x i64> @select_v8i64(<8 x i1> %a, <8 x i64> %b, <8 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x i64> @llvm.vp.select.v8i64(<8 x i1> %a, <8 x i64> %b, <8 x i64> %c, i32 %evl)
+  ret <8 x i64> %v
+}
+
+declare <16 x i64> @llvm.vp.select.v16i64(<16 x i1>, <16 x i64>, <16 x i64>, i32)
+
+define <16 x i64> @select_v16i64(<16 x i1> %a, <16 x i64> %b, <16 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x i64> @llvm.vp.select.v16i64(<16 x i1> %a, <16 x i64> %b, <16 x i64> %c, i32 %evl)
+  ret <16 x i64> %v
+}
+
+declare <2 x half> @llvm.vp.select.v2f16(<2 x i1>, <2 x half>, <2 x half>, i32)
+
+define <2 x half> @select_v2f16(<2 x i1> %a, <2 x half> %b, <2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x half> @llvm.vp.select.v2f16(<2 x i1> %a, <2 x half> %b, <2 x half> %c, i32 %evl)
+  ret <2 x half> %v
+}
+
+declare <4 x half> @llvm.vp.select.v4f16(<4 x i1>, <4 x half>, <4 x half>, i32)
+
+define <4 x half> @select_v4f16(<4 x i1> %a, <4 x half> %b, <4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.select.v4f16(<4 x i1> %a, <4 x half> %b, <4 x half> %c, i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <8 x half> @llvm.vp.select.v8f16(<8 x i1>, <8 x half>, <8 x half>, i32)
+
+define <8 x half> @select_v8f16(<8 x i1> %a, <8 x half> %b, <8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x half> @llvm.vp.select.v8f16(<8 x i1> %a, <8 x half> %b, <8 x half> %c, i32 %evl)
+  ret <8 x half> %v
+}
+
+declare <16 x half> @llvm.vp.select.v16f16(<16 x i1>, <16 x half>, <16 x half>, i32)
+
+define <16 x half> @select_v16f16(<16 x i1> %a, <16 x half> %b, <16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x half> @llvm.vp.select.v16f16(<16 x i1> %a, <16 x half> %b, <16 x half> %c, i32 %evl)
+  ret <16 x half> %v
+}
+
+declare <2 x float> @llvm.vp.select.v2f32(<2 x i1>, <2 x float>, <2 x float>, i32)
+
+define <2 x float> @select_v2f32(<2 x i1> %a, <2 x float> %b, <2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x float> @llvm.vp.select.v2f32(<2 x i1> %a, <2 x float> %b, <2 x float> %c, i32 %evl)
+  ret <2 x float> %v
+}
+
+declare <4 x float> @llvm.vp.select.v4f32(<4 x i1>, <4 x float>, <4 x float>, i32)
+
+define <4 x float> @select_v4f32(<4 x i1> %a, <4 x float> %b, <4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.select.v4f32(<4 x i1> %a, <4 x float> %b, <4 x float> %c, i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <8 x float> @llvm.vp.select.v8f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+
+define <8 x float> @select_v8f32(<8 x i1> %a, <8 x float> %b, <8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x float> @llvm.vp.select.v8f32(<8 x i1> %a, <8 x float> %b, <8 x float> %c, i32 %evl)
+  ret <8 x float> %v
+}
+
+declare <16 x float> @llvm.vp.select.v16f32(<16 x i1>, <16 x float>, <16 x float>, i32)
+
+define <16 x float> @select_v16f32(<16 x i1> %a, <16 x float> %b, <16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x float> @llvm.vp.select.v16f32(<16 x i1> %a, <16 x float> %b, <16 x float> %c, i32 %evl)
+  ret <16 x float> %v
+}
+
+declare <2 x double> @llvm.vp.select.v2f64(<2 x i1>, <2 x double>, <2 x double>, i32)
+
+define <2 x double> @select_v2f64(<2 x i1> %a, <2 x double> %b, <2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <2 x double> @llvm.vp.select.v2f64(<2 x i1> %a, <2 x double> %b, <2 x double> %c, i32 %evl)
+  ret <2 x double> %v
+}
+
+declare <4 x double> @llvm.vp.select.v4f64(<4 x i1>, <4 x double>, <4 x double>, i32)
+
+define <4 x double> @select_v4f64(<4 x i1> %a, <4 x double> %b, <4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.select.v4f64(<4 x i1> %a, <4 x double> %b, <4 x double> %c, i32 %evl)
+  ret <4 x double> %v
+}
+
+declare <8 x double> @llvm.vp.select.v8f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+
+define <8 x double> @select_v8f64(<8 x i1> %a, <8 x double> %b, <8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <8 x double> @llvm.vp.select.v8f64(<8 x i1> %a, <8 x double> %b, <8 x double> %c, i32 %evl)
+  ret <8 x double> %v
+}
+
+declare <16 x double> @llvm.vp.select.v16f64(<16 x i1>, <16 x double>, <16 x double>, i32)
+
+define <16 x double> @select_v16f64(<16 x i1> %a, <16 x double> %b, <16 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_v16f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <16 x double> @llvm.vp.select.v16f64(<16 x i1> %a, <16 x double> %b, <16 x double> %c, i32 %evl)
+  ret <16 x double> %v
+}

llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+experimental-zfh,+experimental-v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+experimental-zfh,+experimental-v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x i8> @llvm.vp.select.nxv1i8(<vscale x 1 x i1>, <vscale x 1 x i8>, <vscale x 1 x i8>, i32)
+
+define <vscale x 1 x i8> @select_nxv1i8(<vscale x 1 x i1> %a, <vscale x 1 x i8> %b, <vscale x 1 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x i8> @llvm.vp.select.nxv1i8(<vscale x 1 x i1> %a, <vscale x 1 x i8> %b, <vscale x 1 x i8> %c, i32 %evl)
+  ret <vscale x 1 x i8> %v
+}
+
+declare <vscale x 2 x i8> @llvm.vp.select.nxv2i8(<vscale x 2 x i1>, <vscale x 2 x i8>, <vscale x 2 x i8>, i32)
+
+define <vscale x 2 x i8> @select_nxv2i8(<vscale x 2 x i1> %a, <vscale x 2 x i8> %b, <vscale x 2 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.select.nxv2i8(<vscale x 2 x i1> %a, <vscale x 2 x i8> %b, <vscale x 2 x i8> %c, i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+declare <vscale x 4 x i8> @llvm.vp.select.nxv4i8(<vscale x 4 x i1>, <vscale x 4 x i8>, <vscale x 4 x i8>, i32)
+
+define <vscale x 4 x i8> @select_nxv4i8(<vscale x 4 x i1> %a, <vscale x 4 x i8> %b, <vscale x 4 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x i8> @llvm.vp.select.nxv4i8(<vscale x 4 x i1> %a, <vscale x 4 x i8> %b, <vscale x 4 x i8> %c, i32 %evl)
+  ret <vscale x 4 x i8> %v
+}
+
+declare <vscale x 8 x i8> @llvm.vp.select.nxv8i8(<vscale x 8 x i1>, <vscale x 8 x i8>, <vscale x 8 x i8>, i32)
+
+define <vscale x 8 x i8> @select_nxv8i8(<vscale x 8 x i1> %a, <vscale x 8 x i8> %b, <vscale x 8 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x i8> @llvm.vp.select.nxv8i8(<vscale x 8 x i1> %a, <vscale x 8 x i8> %b, <vscale x 8 x i8> %c, i32 %evl)
+  ret <vscale x 8 x i8> %v
+}
+
+declare <vscale x 16 x i8> @llvm.vp.select.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
+
+define <vscale x 16 x i8> @select_nxv16i8(<vscale x 16 x i1> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 16 x i8> @llvm.vp.select.nxv16i8(<vscale x 16 x i1> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c, i32 %evl)
+  ret <vscale x 16 x i8> %v
+}
+
+declare <vscale x 32 x i8> @llvm.vp.select.nxv32i8(<vscale x 32 x i1>, <vscale x 32 x i8>, <vscale x 32 x i8>, i32)
+
+define <vscale x 32 x i8> @select_nxv32i8(<vscale x 32 x i1> %a, <vscale x 32 x i8> %b, <vscale x 32 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 32 x i8> @llvm.vp.select.nxv32i8(<vscale x 32 x i1> %a, <vscale x 32 x i8> %b, <vscale x 32 x i8> %c, i32 %evl)
+  ret <vscale x 32 x i8> %v
+}
+
+declare <vscale x 64 x i8> @llvm.vp.select.nxv64i8(<vscale x 64 x i1>, <vscale x 64 x i8>, <vscale x 64 x i8>, i32)
+
+define <vscale x 64 x i8> @select_nxv64i8(<vscale x 64 x i1> %a, <vscale x 64 x i8> %b, <vscale x 64 x i8> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv64i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 64 x i8> @llvm.vp.select.nxv64i8(<vscale x 64 x i1> %a, <vscale x 64 x i8> %b, <vscale x 64 x i8> %c, i32 %evl)
+  ret <vscale x 64 x i8> %v
+}
+
+declare <vscale x 1 x i16> @llvm.vp.select.nxv1i16(<vscale x 1 x i1>, <vscale x 1 x i16>, <vscale x 1 x i16>, i32)
+
+define <vscale x 1 x i16> @select_nxv1i16(<vscale x 1 x i1> %a, <vscale x 1 x i16> %b, <vscale x 1 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x i16> @llvm.vp.select.nxv1i16(<vscale x 1 x i1> %a, <vscale x 1 x i16> %b, <vscale x 1 x i16> %c, i32 %evl)
+  ret <vscale x 1 x i16> %v
+}
+
+declare <vscale x 2 x i16> @llvm.vp.select.nxv2i16(<vscale x 2 x i1>, <vscale x 2 x i16>, <vscale x 2 x i16>, i32)
+
+define <vscale x 2 x i16> @select_nxv2i16(<vscale x 2 x i1> %a, <vscale x 2 x i16> %b, <vscale x 2 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.select.nxv2i16(<vscale x 2 x i1> %a, <vscale x 2 x i16> %b, <vscale x 2 x i16> %c, i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+declare <vscale x 4 x i16> @llvm.vp.select.nxv4i16(<vscale x 4 x i1>, <vscale x 4 x i16>, <vscale x 4 x i16>, i32)
+
+define <vscale x 4 x i16> @select_nxv4i16(<vscale x 4 x i1> %a, <vscale x 4 x i16> %b, <vscale x 4 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x i16> @llvm.vp.select.nxv4i16(<vscale x 4 x i1> %a, <vscale x 4 x i16> %b, <vscale x 4 x i16> %c, i32 %evl)
+  ret <vscale x 4 x i16> %v
+}
+
+declare <vscale x 8 x i16> @llvm.vp.select.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+
+define <vscale x 8 x i16> @select_nxv8i16(<vscale x 8 x i1> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x i16> @llvm.vp.select.nxv8i16(<vscale x 8 x i1> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c, i32 %evl)
+  ret <vscale x 8 x i16> %v
+}
+
+declare <vscale x 16 x i16> @llvm.vp.select.nxv16i16(<vscale x 16 x i1>, <vscale x 16 x i16>, <vscale x 16 x i16>, i32)
+
+define <vscale x 16 x i16> @select_nxv16i16(<vscale x 16 x i1> %a, <vscale x 16 x i16> %b, <vscale x 16 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 16 x i16> @llvm.vp.select.nxv16i16(<vscale x 16 x i1> %a, <vscale x 16 x i16> %b, <vscale x 16 x i16> %c, i32 %evl)
+  ret <vscale x 16 x i16> %v
+}
+
+declare <vscale x 32 x i16> @llvm.vp.select.nxv32i16(<vscale x 32 x i1>, <vscale x 32 x i16>, <vscale x 32 x i16>, i32)
+
+define <vscale x 32 x i16> @select_nxv32i16(<vscale x 32 x i1> %a, <vscale x 32 x i16> %b, <vscale x 32 x i16> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv32i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 32 x i16> @llvm.vp.select.nxv32i16(<vscale x 32 x i1> %a, <vscale x 32 x i16> %b, <vscale x 32 x i16> %c, i32 %evl)
+  ret <vscale x 32 x i16> %v
+}
+
+declare <vscale x 1 x i32> @llvm.vp.select.nxv1i32(<vscale x 1 x i1>, <vscale x 1 x i32>, <vscale x 1 x i32>, i32)
+
+define <vscale x 1 x i32> @select_nxv1i32(<vscale x 1 x i1> %a, <vscale x 1 x i32> %b, <vscale x 1 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x i32> @llvm.vp.select.nxv1i32(<vscale x 1 x i1> %a, <vscale x 1 x i32> %b, <vscale x 1 x i32> %c, i32 %evl)
+  ret <vscale x 1 x i32> %v
+}
+
+declare <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1>, <vscale x 2 x i32>, <vscale x 2 x i32>, i32)
+
+define <vscale x 2 x i32> @select_nxv2i32(<vscale x 2 x i1> %a, <vscale x 2 x i32> %b, <vscale x 2 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1> %a, <vscale x 2 x i32> %b, <vscale x 2 x i32> %c, i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+declare <vscale x 4 x i32> @llvm.vp.select.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+
+define <vscale x 4 x i32> @select_nxv4i32(<vscale x 4 x i1> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x i32> @llvm.vp.select.nxv4i32(<vscale x 4 x i1> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c, i32 %evl)
+  ret <vscale x 4 x i32> %v
+}
+
+declare <vscale x 8 x i32> @llvm.vp.select.nxv8i32(<vscale x 8 x i1>, <vscale x 8 x i32>, <vscale x 8 x i32>, i32)
+
+define <vscale x 8 x i32> @select_nxv8i32(<vscale x 8 x i1> %a, <vscale x 8 x i32> %b, <vscale x 8 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x i32> @llvm.vp.select.nxv8i32(<vscale x 8 x i1> %a, <vscale x 8 x i32> %b, <vscale x 8 x i32> %c, i32 %evl)
+  ret <vscale x 8 x i32> %v
+}
+
+declare <vscale x 16 x i32> @llvm.vp.select.nxv16i32(<vscale x 16 x i1>, <vscale x 16 x i32>, <vscale x 16 x i32>, i32)
+
+define <vscale x 16 x i32> @select_nxv16i32(<vscale x 16 x i1> %a, <vscale x 16 x i32> %b, <vscale x 16 x i32> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 16 x i32> @llvm.vp.select.nxv16i32(<vscale x 16 x i1> %a, <vscale x 16 x i32> %b, <vscale x 16 x i32> %c, i32 %evl)
+  ret <vscale x 16 x i32> %v
+}
+
+declare <vscale x 1 x i64> @llvm.vp.select.nxv1i64(<vscale x 1 x i1>, <vscale x 1 x i64>, <vscale x 1 x i64>, i32)
+
+define <vscale x 1 x i64> @select_nxv1i64(<vscale x 1 x i1> %a, <vscale x 1 x i64> %b, <vscale x 1 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x i64> @llvm.vp.select.nxv1i64(<vscale x 1 x i1> %a, <vscale x 1 x i64> %b, <vscale x 1 x i64> %c, i32 %evl)
+  ret <vscale x 1 x i64> %v
+}
+
+declare <vscale x 2 x i64> @llvm.vp.select.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+
+define <vscale x 2 x i64> @select_nxv2i64(<vscale x 2 x i1> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.select.nxv2i64(<vscale x 2 x i1> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c, i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+declare <vscale x 4 x i64> @llvm.vp.select.nxv4i64(<vscale x 4 x i1>, <vscale x 4 x i64>, <vscale x 4 x i64>, i32)
+
+define <vscale x 4 x i64> @select_nxv4i64(<vscale x 4 x i1> %a, <vscale x 4 x i64> %b, <vscale x 4 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x i64> @llvm.vp.select.nxv4i64(<vscale x 4 x i1> %a, <vscale x 4 x i64> %b, <vscale x 4 x i64> %c, i32 %evl)
+  ret <vscale x 4 x i64> %v
+}
+
+declare <vscale x 8 x i64> @llvm.vp.select.nxv8i64(<vscale x 8 x i1>, <vscale x 8 x i64>, <vscale x 8 x i64>, i32)
+
+define <vscale x 8 x i64> @select_nxv8i64(<vscale x 8 x i1> %a, <vscale x 8 x i64> %b, <vscale x 8 x i64> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x i64> @llvm.vp.select.nxv8i64(<vscale x 8 x i1> %a, <vscale x 8 x i64> %b, <vscale x 8 x i64> %c, i32 %evl)
+  ret <vscale x 8 x i64> %v
+}
+
+declare <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1>, <vscale x 1 x half>, <vscale x 1 x half>, i32)
+
+define <vscale x 1 x half> @select_nxv1f16(<vscale x 1 x i1> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x half> @llvm.vp.select.nxv1f16(<vscale x 1 x i1> %a, <vscale x 1 x half> %b, <vscale x 1 x half> %c, i32 %evl)
+  ret <vscale x 1 x half> %v
+}
+
+declare <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1>, <vscale x 2 x half>, <vscale x 2 x half>, i32)
+
+define <vscale x 2 x half> @select_nxv2f16(<vscale x 2 x i1> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.select.nxv2f16(<vscale x 2 x i1> %a, <vscale x 2 x half> %b, <vscale x 2 x half> %c, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1>, <vscale x 4 x half>, <vscale x 4 x half>, i32)
+
+define <vscale x 4 x half> @select_nxv4f16(<vscale x 4 x i1> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x half> @llvm.vp.select.nxv4f16(<vscale x 4 x i1> %a, <vscale x 4 x half> %b, <vscale x 4 x half> %c, i32 %evl)
+  ret <vscale x 4 x half> %v
+}
+
+declare <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+
+define <vscale x 8 x half> @select_nxv8f16(<vscale x 8 x i1> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x half> @llvm.vp.select.nxv8f16(<vscale x 8 x i1> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c, i32 %evl)
+  ret <vscale x 8 x half> %v
+}
+
+declare <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1>, <vscale x 16 x half>, <vscale x 16 x half>, i32)
+
+define <vscale x 16 x half> @select_nxv16f16(<vscale x 16 x i1> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 16 x half> @llvm.vp.select.nxv16f16(<vscale x 16 x i1> %a, <vscale x 16 x half> %b, <vscale x 16 x half> %c, i32 %evl)
+  ret <vscale x 16 x half> %v
+}
+
+declare <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1>, <vscale x 32 x half>, <vscale x 32 x half>, i32)
+
+define <vscale x 32 x half> @select_nxv32f16(<vscale x 32 x i1> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 32 x half> @llvm.vp.select.nxv32f16(<vscale x 32 x i1> %a, <vscale x 32 x half> %b, <vscale x 32 x half> %c, i32 %evl)
+  ret <vscale x 32 x half> %v
+}
+
+declare <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1>, <vscale x 1 x float>, <vscale x 1 x float>, i32)
+
+define <vscale x 1 x float> @select_nxv1f32(<vscale x 1 x i1> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, mf2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x float> @llvm.vp.select.nxv1f32(<vscale x 1 x i1> %a, <vscale x 1 x float> %b, <vscale x 1 x float> %c, i32 %evl)
+  ret <vscale x 1 x float> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+
+define <vscale x 2 x float> @select_nxv2f32(<vscale x 2 x i1> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %a, <vscale x 2 x float> %b, <vscale x 2 x float> %c, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+
+define <vscale x 4 x float> @select_nxv4f32(<vscale x 4 x i1> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x float> @llvm.vp.select.nxv4f32(<vscale x 4 x i1> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c, i32 %evl)
+  ret <vscale x 4 x float> %v
+}
+
+declare <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1>, <vscale x 8 x float>, <vscale x 8 x float>, i32)
+
+define <vscale x 8 x float> @select_nxv8f32(<vscale x 8 x i1> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x float> @llvm.vp.select.nxv8f32(<vscale x 8 x i1> %a, <vscale x 8 x float> %b, <vscale x 8 x float> %c, i32 %evl)
+  ret <vscale x 8 x float> %v
+}
+
+declare <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1>, <vscale x 16 x float>, <vscale x 16 x float>, i32)
+
+define <vscale x 16 x float> @select_nxv16f32(<vscale x 16 x i1> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 16 x float> @llvm.vp.select.nxv16f32(<vscale x 16 x i1> %a, <vscale x 16 x float> %b, <vscale x 16 x float> %c, i32 %evl)
+  ret <vscale x 16 x float> %v
+}
+
+declare <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1>, <vscale x 1 x double>, <vscale x 1 x double>, i32)
+
+define <vscale x 1 x double> @select_nxv1f64(<vscale x 1 x i1> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v9, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 1 x double> @llvm.vp.select.nxv1f64(<vscale x 1 x i1> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, i32 %evl)
+  ret <vscale x 1 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+define <vscale x 2 x double> @select_nxv2f64(<vscale x 2 x i1> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v10, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.select.nxv2f64(<vscale x 2 x i1> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1>, <vscale x 4 x double>, <vscale x 4 x double>, i32)
+
+define <vscale x 4 x double> @select_nxv4f64(<vscale x 4 x i1> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m4, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v12, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 4 x double> @llvm.vp.select.nxv4f64(<vscale x 4 x i1> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, i32 %evl)
+  ret <vscale x 4 x double> %v
+}
+
+declare <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1>, <vscale x 8 x double>, <vscale x 8 x double>, i32)
+
+define <vscale x 8 x double> @select_nxv8f64(<vscale x 8 x i1> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, i32 zeroext %evl) {
+; CHECK-LABEL: select_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m8, ta, mu
+; CHECK-NEXT:    vmerge.vvm v8, v16, v8, v0
+; CHECK-NEXT:    ret
+  %v = call <vscale x 8 x double> @llvm.vp.select.nxv8f64(<vscale x 8 x i1> %a, <vscale x 8 x double> %b, <vscale x 8 x double> %c, i32 %evl)
+  ret <vscale x 8 x double> %v
+}