[RISCV][test] Precommitted test for optimization for vmerge and unmasked intrinsics.

Precommitted test cases for D130442.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D130753

2 files changed, 573 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-peephole-vmerge-vops.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-peephole-vmerge-vops.ll
new file mode 100644
index 0000000..36e31aa
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-peephole-vmerge-vops.ll
@@ -0,0 +1,215 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-min=256 | FileCheck %s
+
+declare <8 x i16> @llvm.vp.merge.nxv2i16(<8 x i1>, <8 x i16>, <8 x i16>, i32)
+declare <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1>, <8 x i32>, <8 x i32>, i32)
+declare <8 x float> @llvm.vp.merge.nxv2f32(<8 x i1>, <8 x float>, <8 x float>, i32)
+declare <8 x double> @llvm.vp.merge.nxv2f64(<8 x i1>, <8 x double>, <8 x double>, i32)
+
+; Test binary operator with vp.merge and vp.smax.
+declare <8 x i32> @llvm.vp.add.nxv2i32(<8 x i32>, <8 x i32>, <8 x i1>, i32)
+define <8 x i32> @vpmerge_vpadd(<8 x i32> %passthru, <8 x i32> %x, <8 x i32> %y, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.add.nxv2i32(<8 x i32> %x, <8 x i32> %y, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test glued node of merge should not be deleted.
+declare <8 x i1> @llvm.vp.icmp.nxv2i32(<8 x i32>, <8 x i32>, metadata, <8 x i1>, i32)
+define <8 x i32> @vpmerge_vpadd2(<8 x i32> %passthru, <8 x i32> %x, <8 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v11, v9, v10
+; CHECK-NEXT:    vmseq.vv v0, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v11, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.add.nxv2i32(<8 x i32> %x, <8 x i32> %y, <8 x i1> %mask, i32 %vl)
+  %m = call <8 x i1> @llvm.vp.icmp.nxv2i32(<8 x i32> %x, <8 x i32> %y, metadata !"eq", <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test vp.merge have all-ones mask.
+define <8 x i32> @vpmerge_vpadd3(<8 x i32> %passthru, <8 x i32> %x, <8 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e8, mf4, ta, mu
+; CHECK-NEXT:    vmset.m v0
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.add.nxv2i32(<8 x i32> %x, <8 x i32> %y, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %mask, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test float binary operator with vp.merge and vp.fadd.
+declare <8 x float> @llvm.vp.fadd.nxv2f32(<8 x float>, <8 x float>, <8 x i1>, i32)
+define <8 x float> @vpmerge_vpfadd(<8 x float> %passthru, <8 x float> %x, <8 x float> %y, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfadd:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x float> @llvm.vp.fadd.nxv2f32(<8 x float> %x, <8 x float> %y, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x float> @llvm.vp.merge.nxv2f32(<8 x i1> %m, <8 x float> %a, <8 x float> %passthru, i32 %vl)
+  ret <8 x float> %b
+}
+
+; Test conversion by fptosi.
+declare <8 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<8 x float>, <8 x i1>, i32)
+define <8 x i16> @vpmerge_vpfptosi(<8 x i16> %passthru, <8 x float> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfptosi:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v9
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<8 x float> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i16> @llvm.vp.merge.nxv2i16(<8 x i1> %m, <8 x i16> %a, <8 x i16> %passthru, i32 %vl)
+  ret <8 x i16> %b
+}
+
+; Test conversion by sitofp.
+declare <8 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<8 x i64>, <8 x i1>, i32)
+define <8 x float> @vpmerge_vpsitofp(<8 x float> %passthru, <8 x i64> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpsitofp:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<8 x i64> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x float> @llvm.vp.merge.nxv2f32(<8 x i1> %m, <8 x float> %a, <8 x float> %passthru, i32 %vl)
+  ret <8 x float> %b
+}
+
+; Test integer extension by vp.zext.
+declare <8 x i32> @llvm.vp.zext.nxv2i32.nxv2i8(<8 x i8>, <8 x i1>, i32)
+define <8 x i32> @vpmerge_vpzext(<8 x i32> %passthru, <8 x i8> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpzext:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vzext.vf4 v10, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.zext.nxv2i32.nxv2i8(<8 x i8> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test integer truncation by vp.trunc.
+declare <8 x i32> @llvm.vp.trunc.nxv2i32.nxv2i64(<8 x i64>, <8 x i1>, i32)
+define <8 x i32> @vpmerge_vptrunc(<8 x i32> %passthru, <8 x i64> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vptrunc:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.trunc.nxv2i32.nxv2i64(<8 x i64> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test integer extension by vp.fpext.
+declare <8 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<8 x float>, <8 x i1>, i32)
+define <8 x double> @vpmerge_vpfpext(<8 x double> %passthru, <8 x float> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfpext:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v12, v10
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v12, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<8 x float> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x double> @llvm.vp.merge.nxv2f64(<8 x i1> %m, <8 x double> %a, <8 x double> %passthru, i32 %vl)
+  ret <8 x double> %b
+}
+
+; Test integer truncation by vp.trunc.
+declare <8 x float> @llvm.vp.fptrunc.nxv2f32.nxv2f64(<8 x double>, <8 x i1>, i32)
+define <8 x float> @vpmerge_vpfptrunc(<8 x float> %passthru, <8 x double> %x, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfptrunc:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x float> @llvm.vp.fptrunc.nxv2f32.nxv2f64(<8 x double> %x, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x float> @llvm.vp.merge.nxv2f32(<8 x i1> %m, <8 x float> %a, <8 x float> %passthru, i32 %vl)
+  ret <8 x float> %b
+}
+
+; Test load operation by vp.load.
+declare <8 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<8 x i32> *, <8 x i1>, i32)
+define <8 x i32> @vpmerge_vpload(<8 x i32> %passthru, <8 x i32> * %p, <8 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpload:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vle32.v v9, (a0)
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<8 x i32> * %p, <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
+
+; Test result have chain and glued node.
+define <8 x i32> @vpmerge_vpload2(<8 x i32> %passthru, <8 x i32> * %p, <8 x i32> %x, <8 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpload2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vle32.v v11, (a0)
+; CHECK-NEXT:    vmseq.vv v0, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v11, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <8 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <8 x i1> %splat, <8 x i1> poison, <8 x i32> zeroinitializer
+  %a = call <8 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<8 x i32> * %p, <8 x i1> %mask, i32 %vl)
+  %m = call <8 x i1> @llvm.vp.icmp.nxv2i32(<8 x i32> %x, <8 x i32> %y, metadata !"eq", <8 x i1> %mask, i32 %vl)
+  %b = call <8 x i32> @llvm.vp.merge.nxv2i32(<8 x i1> %m, <8 x i32> %a, <8 x i32> %passthru, i32 %vl)
+  ret <8 x i32> %b
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-vops.ll b/llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-vops.ll
new file mode 100644
index 0000000..4268b1b
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-vops.ll
@@ -0,0 +1,358 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=riscv64 -mattr=+v | FileCheck %s
+
+declare <vscale x 2 x i16> @llvm.vp.merge.nxv2i16(<vscale x 2 x i1>, <vscale x 2 x i16>, <vscale x 2 x i16>, i32)
+declare <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1>, <vscale x 2 x i32>, <vscale x 2 x i32>, i32)
+declare <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1>, <vscale x 2 x float>, <vscale x 2 x float>, i32)
+declare <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+; Test binary operator with vp.merge and vp.smax.
+declare <vscale x 2 x i32> @llvm.vp.add.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i32> @vpmerge_vpadd(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.add.nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test glued node of merge should not be deleted.
+declare <vscale x 2 x i1> @llvm.vp.icmp.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, metadata, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i32> @vpmerge_vpadd2(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v11, v9, v10
+; CHECK-NEXT:    vmseq.vv v0, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v11, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.add.nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %mask, i32 %vl)
+  %m = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y, metadata !"eq", <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test vp.merge have all-ones mask.
+define <vscale x 2 x i32> @vpmerge_vpadd3(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpadd3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vmset.m v0
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.add.nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %mask, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test float binary operator with vp.merge and vp.fadd.
+declare <vscale x 2 x float> @llvm.vp.fadd.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x float> @vpmerge_vpfadd(<vscale x 2 x float> %passthru, <vscale x 2 x float> %x, <vscale x 2 x float> %y, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfadd:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfadd.vv v9, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x float> @llvm.vp.fadd.nxv2f32(<vscale x 2 x float> %x, <vscale x 2 x float> %y, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
+  ret <vscale x 2 x float> %b
+}
+
+; Test for binary operator with specific EEW by riscv.vrgatherei16.
+declare <vscale x 2 x i32> @llvm.riscv.vrgatherei16.vv.nxv2i32.i64(<vscale x 2 x i32>, <vscale x 2 x i32>, <vscale x 2 x i16>, i64)
+define <vscale x 2 x i32> @vpmerge_vrgatherei16(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i16> %y, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vrgatherei16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vrgatherei16.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %2 = tail call <vscale x 2 x i32> @llvm.riscv.vrgatherei16.vv.nxv2i32.i64(<vscale x 2 x i32> undef, <vscale x 2 x i32> %x, <vscale x 2 x i16> %y, i64 %1)
+  %3 = tail call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %2, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %2
+}
+
+; Test conversion by fptosi.
+declare <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i16> @vpmerge_vpfptosi(<vscale x 2 x i16> %passthru, <vscale x 2 x float> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfptosi:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v9
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<vscale x 2 x float> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i16> @llvm.vp.merge.nxv2i16(<vscale x 2 x i1> %m, <vscale x 2 x i16> %a, <vscale x 2 x i16> %passthru, i32 %vl)
+  ret <vscale x 2 x i16> %b
+}
+
+; Test conversion by sitofp.
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x float> @vpmerge_vpsitofp(<vscale x 2 x float> %passthru, <vscale x 2 x i64> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpsitofp:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<vscale x 2 x i64> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
+  ret <vscale x 2 x float> %b
+}
+
+; Test integer extension by vp.zext.
+declare <vscale x 2 x i32> @llvm.vp.zext.nxv2i32.nxv2i8(<vscale x 2 x i8>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i32> @vpmerge_vpzext(<vscale x 2 x i32> %passthru, <vscale x 2 x i8> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpzext:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vzext.vf4 v10, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.zext.nxv2i32.nxv2i8(<vscale x 2 x i8> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test integer truncation by vp.trunc.
+declare <vscale x 2 x i32> @llvm.vp.trunc.nxv2i32.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i32> @vpmerge_vptrunc(<vscale x 2 x i32> %passthru, <vscale x 2 x i64> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vptrunc:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.trunc.nxv2i32.nxv2i64(<vscale x 2 x i64> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test integer extension by vp.fpext.
+declare <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x double> @vpmerge_vpfpext(<vscale x 2 x double> %passthru, <vscale x 2 x float> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfpext:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v12, v10
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v12, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x double> @llvm.vp.merge.nxv2f64(<vscale x 2 x i1> %m, <vscale x 2 x double> %a, <vscale x 2 x double> %passthru, i32 %vl)
+  ret <vscale x 2 x double> %b
+}
+
+; Test integer truncation by vp.trunc.
+declare <vscale x 2 x float> @llvm.vp.fptrunc.nxv2f32.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+define <vscale x 2 x float> @vpmerge_vpfptrunc(<vscale x 2 x float> %passthru, <vscale x 2 x double> %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpfptrunc:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x float> @llvm.vp.fptrunc.nxv2f32.nxv2f64(<vscale x 2 x double> %x, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
+  ret <vscale x 2 x float> %b
+}
+
+; Test load operation by vp.load.
+declare <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<vscale x 2 x i32> *, <vscale x 2 x i1>, i32)
+define <vscale x 2 x i32> @vpmerge_vpload(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> * %p, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpload:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vle32.v v9, (a0)
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<vscale x 2 x i32> * %p, <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test result have chain and glued node.
+define <vscale x 2 x i32> @vpmerge_vpload2(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> * %p, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vpload2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vle32.v v11, (a0)
+; CHECK-NEXT:    vmseq.vv v0, v9, v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v11, v0
+; CHECK-NEXT:    ret
+  %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
+  %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
+  %a = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0nxv2i32(<vscale x 2 x i32> * %p, <vscale x 2 x i1> %mask, i32 %vl)
+  %m = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y, metadata !"eq", <vscale x 2 x i1> %mask, i32 %vl)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; FIXME: Merge vmerge.vvm and vleffN.v
+declare { <vscale x 2 x i32>, i64 } @llvm.riscv.vleff.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>*, i64)
+define <vscale x 2 x i32> @vpmerge_vleff(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> * %p, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vleff:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vle32ff.v v9, (a0)
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call { <vscale x 2 x i32>, i64 } @llvm.riscv.vleff.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32>* %p, i64 %1)
+  %b = extractvalue { <vscale x 2 x i32>, i64 } %a, 0
+  %c = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %b, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %c
+}
+
+; Test strided load by riscv.vlse
+declare <vscale x 2 x i32> @llvm.riscv.vlse.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>*, i64, i64)
+define <vscale x 2 x i32> @vpmerge_vlse(<vscale x 2 x i32> %passthru,  <vscale x 2 x i32> * %p, <vscale x 2 x i1> %m, i64 %s, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vlse:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a2, e32, m1, ta, mu
+; CHECK-NEXT:    vlse32.v v9, (a0), a1
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x i32> @llvm.riscv.vlse.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32>* %p, i64 %s, i64 %1)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test indexed load by riscv.vluxei
+declare <vscale x 2 x i32> @llvm.riscv.vluxei.nxv2i32.nxv2i64(<vscale x 2 x i32>, <vscale x 2 x i32>*, <vscale x 2 x i64>, i64)
+define <vscale x 2 x i32> @vpmerge_vluxei(<vscale x 2 x i32> %passthru,  <vscale x 2 x i32> * %p, <vscale x 2 x i64> %idx, <vscale x 2 x i1> %m, i64 %s, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vluxei:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a2, e32, m1, ta, mu
+; CHECK-NEXT:    vluxei64.v v9, (a0), v10
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x i32> @llvm.riscv.vluxei.nxv2i32.nxv2i64(<vscale x 2 x i32> undef, <vscale x 2 x i32>* %p, <vscale x 2 x i64> %idx, i64 %1)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test vector index by riscv.vid
+declare <vscale x 2 x i32> @llvm.riscv.vid.nxv2i32(<vscale x 2 x i32>, i64)
+define <vscale x 2 x i32> @vpmerge_vid(<vscale x 2 x i32> %passthru, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vid:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vid.v v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x i32> @llvm.riscv.vid.nxv2i32(<vscale x 2 x i32> undef, i64 %1)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test riscv.viota
+declare <vscale x 2 x i32> @llvm.riscv.viota.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i1>, i64)
+define <vscale x 2 x i32> @vpmerge_viota(<vscale x 2 x i32> %passthru, <vscale x 2 x i1> %m, <vscale x 2 x i1> %vm, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_viota:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    viota.m v10, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x i32> @llvm.riscv.viota.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i1> %vm, i64 %1)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test riscv.vfclass
+declare <vscale x 2 x i32> @llvm.riscv.vfclass.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x float>, i64)
+define <vscale x 2 x i32> @vpmerge_vflcass(<vscale x 2 x i32> %passthru, <vscale x 2 x float> %vf, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vflcass:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfclass.v v9, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x i32> @llvm.riscv.vfclass.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x float> %vf, i64 %1)
+  %b = call <vscale x 2 x i32> @llvm.vp.merge.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
+  ret <vscale x 2 x i32> %b
+}
+
+; Test riscv.vfsqrt
+declare <vscale x 2 x float> @llvm.riscv.vfsqrt.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, i64)
+define <vscale x 2 x float> @vpmerge_vfsqrt(<vscale x 2 x float> %passthru, <vscale x 2 x float> %vf, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vfsqrt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfsqrt.v v9, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x float> @llvm.riscv.vfsqrt.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> %vf, i64 %1)
+  %b = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
+  ret <vscale x 2 x float> %b
+}
+
+; Test reciprocal operation by riscv.vfrec7
+declare <vscale x 2 x float> @llvm.riscv.vfrec7.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>, i64)
+define <vscale x 2 x float> @vpmerge_vfrec7(<vscale x 2 x float> %passthru, <vscale x 2 x float> %vf, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vpmerge_vfrec7:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfrec7.v v9, v9
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, tu, mu
+; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    ret
+  %1 = zext i32 %vl to i64
+  %a = call <vscale x 2 x float> @llvm.riscv.vfrec7.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> %vf, i64 %1)
+  %b = call <vscale x 2 x float> @llvm.vp.merge.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
+  ret <vscale x 2 x float> %b
+}