[RISCV][VLOPT] Add vmerge to isSupportedInstr

llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp

@@ -944,6 +944,10 @@ static bool isSupportedInstr(const MachineInstr &MI) {
   case RISCV::VMADD_VX:
   case RISCV::VNMSUB_VV:
   case RISCV::VNMSUB_VX:
+  // Vector Integer Merge Instructions
+  case RISCV::VMERGE_VIM:
+  case RISCV::VMERGE_VVM:
+  case RISCV::VMERGE_VXM:
   // Vector Widening Integer Multiply-Add Instructions
   case RISCV::VWMACCU_VV:
   case RISCV::VWMACCU_VX:

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-buildvec.ll

@@ -697,7 +697,7 @@ define void @buildvec_seq_v9i8(ptr %x) {
 ; CHECK-LABEL: buildvec_seq_v9i8:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    li a1, 73
-; CHECK-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; CHECK-NEXT:    vsetivli zero, 9, e8, m1, ta, ma
 ; CHECK-NEXT:    vmv.v.i v9, 3
 ; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
 ; CHECK-NEXT:    vmv.s.x v0, a1
@@ -706,7 +706,6 @@ define void @buildvec_seq_v9i8(ptr %x) {
 ; CHECK-NEXT:    vsetvli zero, zero, e8, m1, ta, ma
 ; CHECK-NEXT:    vmerge.vim v9, v9, 1, v0
 ; CHECK-NEXT:    vmv1r.v v0, v8
-; CHECK-NEXT:    vsetivli zero, 9, e8, m1, ta, ma
 ; CHECK-NEXT:    vmerge.vim v8, v9, 2, v0
 ; CHECK-NEXT:    vse8.v v8, (a0)
 ; CHECK-NEXT:    ret

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll

@@ -1290,10 +1290,8 @@ define void @mulhs_v6i16(ptr %x) {
 ; CHECK-NEXT:    vle16.v v8, (a0)
 ; CHECK-NEXT:    li a1, 22
 ; CHECK-NEXT:    vmv.s.x v0, a1
-; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; CHECK-NEXT:    vmv.v.i v9, -7
 ; CHECK-NEXT:    vmerge.vim v9, v9, 7, v0
-; CHECK-NEXT:    vsetivli zero, 6, e16, m1, ta, ma
 ; CHECK-NEXT:    vdiv.vv v8, v8, v9
 ; CHECK-NEXT:    vse16.v v8, (a0)
 ; CHECK-NEXT:    ret

llvm/test/CodeGen/RISCV/rvv/narrow-shift-extend.ll

@@ -142,10 +142,9 @@ entry:
 define <vscale x 4 x i32> @test_vloxei7(ptr %ptr, <vscale x 4 x i1> %offset, i64 %vl) {
 ; CHECK-LABEL: test_vloxei7:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetvli a2, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vsetvli zero, a1, e64, m4, ta, ma
 ; CHECK-NEXT:    vmv.v.i v8, 0
 ; CHECK-NEXT:    vmerge.vim v8, v8, 1, v0
-; CHECK-NEXT:    vsetvli zero, a1, e64, m4, ta, ma
 ; CHECK-NEXT:    vsll.vi v12, v8, 2
 ; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
 ; CHECK-NEXT:    vloxei64.v v8, (a0), v12

llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll

@@ -3535,3 +3535,63 @@ define <vscale x 4 x i1> @vmfgt_vv(<vscale x 4 x float> %a, <vscale x 4 x i1> %b
   %2 = call <vscale x 4 x i1> @llvm.riscv.vmand.nxv4i1(<vscale x 4 x i1> %1, <vscale x 4 x i1> %b, iXLen %vl)
   ret <vscale x 4 x i1> %2
 }
+
+define <vscale x 4 x i32> @vmerge_vvm(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i1> %c, iXLen %vl) {
+; NOVLOPT-LABEL: vmerge_vvm:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vmerge.vvm v8, v8, v10, v0
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v10
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vmerge_vvm:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vmerge.vvm v8, v8, v10, v0
+; VLOPT-NEXT:    vadd.vv v8, v8, v10
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vmerge.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i1> %c, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vmerge_vxm(<vscale x 4 x i32> %a, i32 %b, <vscale x 4 x i1> %c, iXLen %vl) {
+; NOVLOPT-LABEL: vmerge_vxm:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vmerge.vxm v8, v8, a0, v0
+; NOVLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vmerge_vxm:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; VLOPT-NEXT:    vmerge.vxm v8, v8, a0, v0
+; VLOPT-NEXT:    vadd.vv v8, v8, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vmerge.nxv4i32.i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, <vscale x 4 x i1> %c, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %1, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vmerge_vim(<vscale x 4 x i32> %a, <vscale x 4 x i1> %c, iXLen %vl) {
+; NOVLOPT-LABEL: vmerge_vim:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vmerge.vim v8, v8, 9, v0
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vmerge_vim:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vmerge.vim v8, v8, 9, v0
+; VLOPT-NEXT:    vadd.vv v8, v8, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vmerge.nxv4i32.i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 9, <vscale x 4 x i1> %c, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %1, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}