[RISCV] Teach doPeepholeMaskedRVV to handle vslide{up,down}

[luke]

Summary

We already handle vslide1{up,down}, so this extends it to vslide{up,down}.

This was unintentionally added in https://reviews.llvm.org/D150463 and
then removed in 37cfcfcef76bb615b941d7077ca81168bd7ad080, but unless I'm
missing something this should still be ok as the mask only controls what
destination elements are written to.

[craig.topper]

LGTM

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

 
 multiclass VPseudoVSLDVWithPolicy<VReg RetClass,
                                   RegisterClass Op1Class,
                                   DAGOperand Op2Class,
                                   LMULInfo MInfo,
                                   string Constraint = ""> {
   let VLMul = MInfo.value in {
     def "_" # MInfo.MX : VPseudoTernaryNoMaskWithPolicy<RetClass, Op1Class, Op2Class, Constraint>;
-    def "_" # MInfo.MX # "_MASK" : VPseudoBinaryMaskPolicy<RetClass, Op1Class, Op2Class, Constraint>;
+    def "_" # MInfo.MX # "_MASK" : VPseudoBinaryMaskPolicy<RetClass, Op1Class, Op2Class, Constraint>,
+                                   RISCVMaskedPseudo</*MaskOpIdx*/ 3, /*HasTU*/ false, /*IsTernary*/ true>;
   }
 }
 
 multiclass VPseudoVSLDV_VX<LMULInfo m, string Constraint = ""> {
   defm _VX : VPseudoVSLDVWithPolicy<m.vrclass, m.vrclass, GPR, m, Constraint>;
 }
 
 multiclass VPseudoVSLDV_VI<Operand ImmType = simm5, LMULInfo m, string Constraint = ""> {

llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-masked-vops.ll

 }
 declare <vscale x 2 x i32> @llvm.riscv.vle.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>*, <vscale x 2 x i1>, i64, i64)
 
 declare <vscale x 2 x i32> @llvm.riscv.vslideup.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, <vscale x 2 x i1>, i64, i64)
 define <vscale x 2 x i32> @vpmerge_vslideup(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl) {
 ; CHECK-LABEL: vpmerge_vslideup:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a1, e32, m1, tu, mu
-; CHECK-NEXT:    vmv1r.v v10, v8
-; CHECK-NEXT:    vslideup.vx v10, v9, a0, v0.t
-; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    vslideup.vx v8, v9, a0, v0.t
 ; CHECK-NEXT:    ret
   %a = call <vscale x 2 x i32> @llvm.riscv.vslideup.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
   %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
   %b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
   ret <vscale x 2 x i32> %b
 }
 
 declare <vscale x 2 x i32> @llvm.riscv.vslidedown.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, <vscale x 2 x i1>, i64, i64)
 define <vscale x 2 x i32> @vpmerge_vslidedown(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl) {
 ; CHECK-LABEL: vpmerge_vslidedown:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a1, e32, m1, tu, mu
-; CHECK-NEXT:    vmv1r.v v10, v8
-; CHECK-NEXT:    vslidedown.vx v10, v9, a0, v0.t
-; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    vslidedown.vx v8, v9, a0, v0.t
 ; CHECK-NEXT:    ret
   %a = call <vscale x 2 x i32> @llvm.riscv.vslidedown.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
   %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
   %b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
   ret <vscale x 2 x i32> %b
 }
 

llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-vops.ll

; CHECK-NEXT:    ret
   ret <vscale x 2 x float> %b
 }
 
 ; Test slides
 declare <vscale x 2 x i32> @llvm.riscv.vslideup.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, i64, i64)
 define <vscale x 2 x i32> @vpselect_vslideup(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
 ; CHECK-LABEL: vpselect_vslideup:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
-; CHECK-NEXT:    vslideup.vx v10, v9, a0
-; CHECK-NEXT:    vmerge.vvm v8, v8, v10, v0
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vslideup.vx v8, v9, a0, v0.t
 ; CHECK-NEXT:    ret
   %1 = zext i32 %vl to i64
   %a = call <vscale x 2 x i32> @llvm.riscv.vslideup.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i64 %x, i64 %1, i64 0)
   %b = call <vscale x 2 x i32> @llvm.vp.select.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
 }
 
 declare <vscale x 2 x i32> @llvm.riscv.vslidedown.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, i64, i64)
 define <vscale x 2 x i32> @vpselect_vslidedown(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
 ; CHECK-LABEL: vpselect_vslidedown:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
-; CHECK-NEXT:    vslidedown.vx v9, v9, a0
-; CHECK-NEXT:    vmerge.vvm v8, v8, v9, v0
+; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, mu
+; CHECK-NEXT:    vslidedown.vx v8, v9, a0, v0.t
 ; CHECK-NEXT:    ret
   %1 = zext i32 %vl to i64
   %a = call <vscale x 2 x i32> @llvm.riscv.vslidedown.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i64 %x, i64 %1, i64 0)
   %b = call <vscale x 2 x i32> @llvm.vp.select.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
 }
 
 declare <vscale x 2 x i32> @llvm.riscv.vslide1up.nxv2i32.i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i32, i64)