[GISel] Funnel shift combiner port from SelectionDAG ISel to GlobalISel (llvm#135132)

The funnel shift combiner rule from
llvm@4a3708c
is currently missing from GlobalISel. The following is a port of that
combiner to GlobalISel.

Author: axelcool1234

llvm/include/llvm/Target/GlobalISel/Combine.td

@@ -1040,6 +1040,34 @@ def funnel_shift_overshift: GICombineRule<
   (apply [{ Helper.applyFunnelShiftConstantModulo(*${root}); }])
 >;
 
+// Transform: fshl x, z, y | shl x, y -> fshl x, z, y
+// Transform: shl x, y | fshl x, z, y  -> fshl x, z, y
+// FIXME: TableGen didn't handle G_OR commutativity on its own, 
+//        necessitating the use of !foreach to handle it manually.
+def funnel_shift_or_shift_to_funnel_shift_left_frags : GICombinePatFrag<
+  (outs root: $dst, $out1, $out2), (ins),
+  !foreach(inst, [(G_OR $dst, $out1, $out2), (G_OR $dst, $out2, $out1)], 
+           (pattern (G_FSHL $out1, $x, $z, $y), (G_SHL $out2, $x, $y), inst))>;
+def funnel_shift_or_shift_to_funnel_shift_left: GICombineRule<
+  (defs root:$root), 
+  (match (funnel_shift_or_shift_to_funnel_shift_left_frags $root, $out1, $out2)),
+  (apply (GIReplaceReg $root, $out1))
+>;
+
+// Transform: fshr z, x, y | srl x, y -> fshr z, x, y
+// Transform: srl x, y | fshr z, x, y -> fshr z, x, y
+// FIXME: TableGen didn't handle G_OR commutativity on its own, 
+//        necessitating the use of !foreach to handle it manually.
+def funnel_shift_or_shift_to_funnel_shift_right_frags : GICombinePatFrag<
+  (outs root: $dst, $out1, $out2), (ins),
+  !foreach(inst, [(G_OR $dst, $out1, $out2), (G_OR $dst, $out2, $out1)], 
+           (pattern (G_FSHR $out1, $z, $x, $y), (G_LSHR $out2, $x, $y), inst))>;
+def funnel_shift_or_shift_to_funnel_shift_right: GICombineRule<
+  (defs root:$root), 
+  (match (funnel_shift_or_shift_to_funnel_shift_right_frags $root, $out1, $out2)), 
+  (apply (GIReplaceReg $root, $out1))
+>;
+
 def rotate_out_of_range : GICombineRule<
   (defs root:$root),
   (match (wip_match_opcode G_ROTR, G_ROTL):$root,
@@ -1112,7 +1140,9 @@ def funnel_shift_combines : GICombineGroup<[funnel_shift_from_or_shift,
                                             funnel_shift_to_rotate,
                                             funnel_shift_right_zero,
                                             funnel_shift_left_zero,
-                                            funnel_shift_overshift]>;
+                                            funnel_shift_overshift,
+                                            funnel_shift_or_shift_to_funnel_shift_left,
+                                            funnel_shift_or_shift_to_funnel_shift_right]>;
 
 def bitfield_extract_from_sext_inreg : GICombineRule<
   (defs root:$root, build_fn_matchinfo:$info),

llvm/test/CodeGen/AArch64/funnel-shift.ll

@@ -674,14 +674,12 @@ define i32 @or_shl_fshl_simplify(i32 %x, i32 %y, i32 %s) {
 ; CHECK-GI-LABEL: or_shl_fshl_simplify:
 ; CHECK-GI:       // %bb.0:
 ; CHECK-GI-NEXT:    mov w8, #31 // =0x1f
-; CHECK-GI-NEXT:    and w9, w2, #0x1f
-; CHECK-GI-NEXT:    lsr w10, w0, #1
-; CHECK-GI-NEXT:    lsl w11, w1, w2
+; CHECK-GI-NEXT:    lsr w9, w0, #1
+; CHECK-GI-NEXT:    and w10, w2, #0x1f
 ; CHECK-GI-NEXT:    bic w8, w8, w2
-; CHECK-GI-NEXT:    lsl w9, w1, w9
-; CHECK-GI-NEXT:    lsr w8, w10, w8
-; CHECK-GI-NEXT:    orr w9, w9, w11
-; CHECK-GI-NEXT:    orr w0, w9, w8
+; CHECK-GI-NEXT:    lsl w10, w1, w10
+; CHECK-GI-NEXT:    lsr w8, w9, w8
+; CHECK-GI-NEXT:    orr w0, w10, w8
 ; CHECK-GI-NEXT:    ret
   %shy = shl i32 %y, %s
   %fun = call i32 @llvm.fshl.i32(i32 %y, i32 %x, i32 %s)
@@ -702,14 +700,12 @@ define i32 @or_lshr_fshr_simplify(i32 %x, i32 %y, i32 %s) {
 ; CHECK-GI-LABEL: or_lshr_fshr_simplify:
 ; CHECK-GI:       // %bb.0:
 ; CHECK-GI-NEXT:    mov w8, #31 // =0x1f
-; CHECK-GI-NEXT:    and w9, w2, #0x1f
-; CHECK-GI-NEXT:    lsl w10, w0, #1
-; CHECK-GI-NEXT:    lsr w11, w1, w2
+; CHECK-GI-NEXT:    lsl w9, w0, #1
+; CHECK-GI-NEXT:    and w10, w2, #0x1f
 ; CHECK-GI-NEXT:    bic w8, w8, w2
-; CHECK-GI-NEXT:    lsr w9, w1, w9
-; CHECK-GI-NEXT:    lsl w8, w10, w8
-; CHECK-GI-NEXT:    orr w9, w11, w9
-; CHECK-GI-NEXT:    orr w0, w9, w8
+; CHECK-GI-NEXT:    lsl w8, w9, w8
+; CHECK-GI-NEXT:    lsr w9, w1, w10
+; CHECK-GI-NEXT:    orr w0, w8, w9
 ; CHECK-GI-NEXT:    ret
   %shy = lshr i32 %y, %s
   %fun = call i32 @llvm.fshr.i32(i32 %x, i32 %y, i32 %s)

llvm/test/CodeGen/RISCV/GlobalISel/shift.ll

@@ -105,3 +105,51 @@ define i16 @test_shl_i48_2(i48 %x, i48 %y) {
   %trunc = trunc i48 %shl to i16
   ret i16 %trunc
 }
+
+define i32 @test_fshl_i32(i32 %x, i32 %_, i32 %y) {
+; RV32-LABEL: test_fshl_i32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    not a3, a2
+; RV32-NEXT:    sll a0, a0, a2
+; RV32-NEXT:    srli a1, a1, 1
+; RV32-NEXT:    srl a1, a1, a3
+; RV32-NEXT:    or a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: test_fshl_i32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    not a3, a2
+; RV64-NEXT:    sllw a0, a0, a2
+; RV64-NEXT:    srliw a1, a1, 1
+; RV64-NEXT:    srlw a1, a1, a3
+; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    ret
+  %fshl = call i32 @llvm.fshl.i32(i32 %x, i32 %_, i32 %y)
+  %shl = shl i32 %x, %y
+  %or = or i32 %fshl, %shl
+  ret i32 %or
+}
+
+define i32 @test_fshr_i32(i32 %_, i32 %x, i32 %y) {
+; RV32-LABEL: test_fshr_i32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    not a3, a2
+; RV32-NEXT:    slli a0, a0, 1
+; RV32-NEXT:    sll a0, a0, a3
+; RV32-NEXT:    srl a1, a1, a2
+; RV32-NEXT:    or a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: test_fshr_i32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    not a3, a2
+; RV64-NEXT:    slli a0, a0, 1
+; RV64-NEXT:    sllw a0, a0, a3
+; RV64-NEXT:    srlw a1, a1, a2
+; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    ret
+  %fshr = call i32 @llvm.fshr.i32(i32 %_, i32 %x, i32 %y)
+  %lshr = lshr i32 %x, %y
+  %or = or i32 %lshr, %fshr
+  ret i32 %or
+}