aboutsummaryrefslogtreecommitdiffstats
path: root/lib/compiler/src/beam_dead.erl
diff options
context:
space:
mode:
Diffstat (limited to 'lib/compiler/src/beam_dead.erl')
-rw-r--r--lib/compiler/src/beam_dead.erl729
1 files changed, 492 insertions, 237 deletions
diff --git a/lib/compiler/src/beam_dead.erl b/lib/compiler/src/beam_dead.erl
index b15adfa889..bbe607cf19 100644
--- a/lib/compiler/src/beam_dead.erl
+++ b/lib/compiler/src/beam_dead.erl
@@ -21,112 +21,10 @@
-export([module/2]).
-%%% The following optimisations are done:
-%%%
-%%% (1) In this code
-%%%
-%%% move DeadValue {x,0}
-%%% jump L2
-%%% .
-%%% .
-%%% .
-%%% L2: move Anything {x,0}
-%%% .
-%%% .
-%%% .
-%%%
-%%% the first assignment to {x,0} has no effect (is dead),
-%%% so it can be removed. Besides removing a move instruction,
-%%% if the move was preceeded by a label, the resulting code
-%%% will look this
-%%%
-%%% L1: jump L2
-%%% .
-%%% .
-%%% .
-%%% L2: move Anything {x,0}
-%%% .
-%%% .
-%%% .
-%%%
-%%% which can be further optimized by the jump optimizer (beam_jump).
-%%%
-%%% (2) In this code
-%%%
-%%% L1: move AtomLiteral {x,0}
-%%% jump L2
-%%% .
-%%% .
-%%% .
-%%% L2: test is_atom FailLabel {x,0}
-%%% select_val {x,0}, FailLabel [... AtomLiteral => L3...]
-%%% .
-%%% .
-%%% .
-%%% L3: ...
-%%%
-%%% FailLabel: ...
-%%%
-%%% the first code fragment can be changed to
-%%%
-%%% L1: move AtomLiteral {x,0}
-%%% jump L3
-%%%
-%%% If the literal is not included in the table of literals in the
-%%% select_val instruction, the first code fragment will instead be
-%%% rewritten as:
-%%%
-%%% L1: move AtomLiteral {x,0}
-%%% jump FailLabel
-%%%
-%%% The move instruction will be removed by optimization (1) above,
-%%% if the code following the L3 label overwrites {x,0}.
-%%%
-%%% The code following the L2 label will be kept, but it will be removed later
-%%% by the jump optimizer.
-%%%
-%%% (3) In this code
-%%%
-%%% test is_eq_exact ALabel Src Dst
-%%% move Src Dst
-%%%
-%%% the move instruction can be removed.
-%%% Same thing for
-%%%
-%%% test is_nil ALabel Dst
-%%% move [] Dst
-%%%
-%%%
-%%% (4) In this code
-%%%
-%%% select_val {x,Reg}, ALabel [... Literal => L1...]
-%%% .
-%%% .
-%%% .
-%%% L1: move Literal {x,Reg}
-%%%
-%%% we can remove the move instruction.
-%%%
-%%% (5) In the following code
-%%%
-%%% bif '=:=' Fail Src1 Src2 {x,0}
-%%% jump L1
-%%% .
-%%% .
-%%% .
-%%% L1: select_val {x,0}, ALabel [... true => L2..., ...false => L3...]
-%%% .
-%%% .
-%%% .
-%%% L2: .... L3: ....
-%%%
-%%% the first two instructions can be replaced with
-%%%
-%%% test is_eq_exact L3 Src1 Src2
-%%% jump L2
-%%%
-%%% provided that {x,0} is killed at both L2 and L3.
-%%%
+%%% Dead code is code that is executed but has no effect. This
+%%% optimization pass either removes dead code or jumps around it,
+%%% potentially making it unreachable and a target for the
+%%% the beam_jump pass.
-import(lists, [mapfoldl/3,reverse/1]).
@@ -173,12 +71,39 @@ move_move_into_block([I|Is], Acc) ->
move_move_into_block([], Acc) -> reverse(Acc).
%%%
-%%% Scan instructions in execution order and remove dead code.
+%%% Scan instructions in execution order and remove redundant 'move'
+%%% instructions. 'move' instructions are redundant if we know that
+%%% the register already contains the value being assigned, as in the
+%%% following code:
+%%%
+%%% test is_eq_exact SomeLabel Src Dst
+%%% move Src Dst
+%%%
+%%% or in:
+%%%
+%%% test is_nil SomeLabel Dst
+%%% move nil Dst
+%%%
+%%% or in:
+%%%
+%%% select_val Register FailLabel [... Literal => L1...]
+%%% .
+%%% .
+%%% .
+%%% L1: move Literal Register
+%%%
+%%% Also add extra labels to help the second backward pass.
%%%
forward(Is, Lc) ->
- forward(Is, gb_trees:empty(), Lc, []).
-
+ forward(Is, #{}, Lc, []).
+
+forward([{move,_,_}=Move|[{label,L}|_]=Is], D, Lc, Acc) ->
+ %% move/2 followed by jump/1 is optimized by backward/3.
+ forward([Move,{jump,{f,L}}|Is], D, Lc, Acc);
+forward([{bif,_,_,_,_}=Bif|[{label,L}|_]=Is], D, Lc, Acc) ->
+ %% bif/4 followed by jump/1 is optimized by backward/3.
+ forward([Bif,{jump,{f,L}}|Is], D, Lc, Acc);
forward([{block,[]}|Is], D, Lc, Acc) ->
%% Empty blocks can prevent optimizations.
forward(Is, D, Lc, Acc);
@@ -190,21 +115,24 @@ forward([{label,Lbl}=LblI,{block,[{set,[Dst],[Lit],move}|BlkIs]}=Blk|Is], D, Lc,
%% cannot be reached in any other way than through the select_val/3
%% instruction (i.e. there can be no fallthrough to such label and
%% it cannot be referenced by, for example, a jump/1 instruction).
- Block = case gb_trees:lookup({Lbl,Dst}, D) of
- {value,Lit} -> {block,BlkIs}; %Safe to remove move instruction.
- _ -> Blk %Must keep move instruction.
- end,
+ Key = {Lbl,Dst},
+ Block = case D of
+ #{Key := Lit} -> {block,BlkIs}; %Safe to remove move instruction.
+ _ -> Blk %Must keep move instruction.
+ end,
forward([Block|Is], D, Lc, [LblI|Acc]);
forward([{label,Lbl}=LblI|[{move,Lit,Dst}|Is1]=Is0], D, Lc, Acc) ->
%% Assumption: The target labels in a select_val/3 instruction
%% cannot be reached in any other way than through the select_val/3
%% instruction (i.e. there can be no fallthrough to such label and
%% it cannot be referenced by, for example, a jump/1 instruction).
- Is = case gb_trees:lookup({Lbl,Dst}, D) of
- {value,Lit} -> Is1; %Safe to remove move instruction.
- _ -> Is0 %Keep move instruction.
+ Is = case maps:find({Lbl,Dst}, D) of
+ {ok,Lit} -> Is1; %Safe to remove move instruction.
+ _ -> Is0 %Keep move instruction.
end,
forward(Is, D, Lc, [LblI|Acc]);
+forward([{test,is_eq_exact,_,[Same,Same]}|Is], D, Lc, Acc) ->
+ forward(Is, D, Lc, Acc);
forward([{test,is_eq_exact,_,[Dst,Src]}=I,
{block,[{set,[Dst],[Src],move}|Bl]}|Is], D, Lc, Acc) ->
forward([I,{block,Bl}|Is], D, Lc, Acc);
@@ -215,15 +143,13 @@ forward([{test,is_eq_exact,_,[Dst,Src]}=I,{move,Src,Dst}|Is], D, Lc, Acc) ->
forward([I|Is], D, Lc, Acc);
forward([{test,is_nil,_,[Dst]}=I,{move,nil,Dst}|Is], D, Lc, Acc) ->
forward([I|Is], D, Lc, Acc);
-forward([{test,is_eq_exact,_,_}=I|Is], D, Lc, Acc) ->
- case Is of
- [{label,_}|_] -> forward(Is, D, Lc, [I|Acc]);
- _ -> forward(Is, D, Lc+1, [{label,Lc},I|Acc])
- end;
-forward([{test,is_ne_exact,_,_}=I|Is], D, Lc, Acc) ->
- case Is of
- [{label,_}|_] -> forward(Is, D, Lc, [I|Acc]);
- _ -> forward(Is, D, Lc+1, [{label,Lc},I|Acc])
+forward([{test,_,_,_}=I|Is]=Is0, D, Lc, Acc) ->
+ %% Help the second, backward pass to by inserting labels after
+ %% relational operators so that they can be skipped if they are
+ %% known to be true.
+ case useful_to_insert_label(Is0) of
+ false -> forward(Is, D, Lc, [I|Acc]);
+ true -> forward(Is, D, Lc+1, [{label,Lc},I|Acc])
end;
forward([I|Is], D, Lc, Acc) ->
forward(Is, D, Lc, [I|Acc]);
@@ -231,17 +157,57 @@ forward([], _, Lc, Acc) -> {Acc,Lc}.
update_value_dict([Lit,{f,Lbl}|T], Reg, D0) ->
Key = {Lbl,Reg},
- D = case gb_trees:lookup(Key, D0) of
- none -> gb_trees:insert(Key, Lit, D0); %New.
- {value,inconsistent} -> D0; %Inconsistent.
- {value,_} -> gb_trees:update(Key, inconsistent, D0)
- end,
+ D = case D0 of
+ #{Key := inconsistent} -> D0;
+ #{Key := _} -> D0#{Key := inconsistent};
+ _ -> D0#{Key => Lit}
+ end,
update_value_dict(T, Reg, D);
update_value_dict([], _, D) -> D.
+useful_to_insert_label([_,{label,_}|_]) ->
+ false;
+useful_to_insert_label([{test,Op,_,_}|_]) ->
+ case Op of
+ is_lt -> true;
+ is_ge -> true;
+ is_eq_exact -> true;
+ is_ne_exact -> true;
+ _ -> false
+ end.
+
+%%%
+%%% Scan instructions in reverse execution order and try to
+%%% shortcut branch instructions.
+%%%
+%%% For example, in this code:
%%%
-%%% Scan instructions in reverse execution order and remove dead code.
+%%% move Literal Register
+%%% jump L1
+%%% .
+%%% .
+%%% .
+%%% L1: test is_{integer,atom} FailLabel Register
+%%% select_val {x,0} FailLabel [... Literal => L2...]
+%%% .
+%%% .
+%%% .
+%%% L2: ...
%%%
+%%% the 'selectval' instruction will always transfer control to L2,
+%%% so we can just as well jump to L2 directly by rewriting the
+%%% first part of the sequence like this:
+%%%
+%%% move Literal Register
+%%% jump L2
+%%%
+%%% If register Register is killed at label L2, we can remove the
+%%% 'move' instruction, leaving just the 'jump' instruction:
+%%%
+%%% jump L2
+%%%
+%%% These transformations may leave parts of the code unreachable.
+%%% The beam_jump pass will remove the unreachable code.
backward(Is, D) ->
backward(Is, D, []).
@@ -277,15 +243,8 @@ backward([{select,select_val,Reg,{f,Fail0},List0}|Is], D, Acc) ->
Fail = shortcut_bs_test(Fail1, Is, D),
Sel = {select,select_val,Reg,{f,Fail},List},
backward(Is, D, [Sel|Acc]);
-backward([{jump,{f,To0}},{move,Src,Reg}=Move0|Is], D, Acc) ->
- {To,Move} = case Src of
- {atom,Val0} ->
- To1 = shortcut_select_label(To0, Reg, Val0, D),
- {To2,Val} = shortcut_boolean_label(To1, Reg, Val0, D),
- {To2,{move,{atom,Val},Reg}};
- _ ->
- {shortcut_label(To0, D),Move0}
- end,
+backward([{jump,{f,To0}},{move,Src,Reg}=Move|Is], D, Acc) ->
+ To = shortcut_select_label(To0, Reg, Src, D),
Jump = {jump,{f,To}},
case beam_utils:is_killed_at(Reg, To, D) of
false -> backward([Move|Is], D, [Jump|Acc]);
@@ -297,32 +256,39 @@ backward([{jump,{f,To}}=J|[{bif,Op,_,Ops,Reg}|Is]=Is0], D, Acc) ->
catch
throw:not_possible -> backward(Is0, D, [J|Acc])
end;
+backward([{test,bs_start_match2,F,_,[R,_],Ctxt}=I|Is], D,
+ [{test,bs_match_string,F,[Ctxt,Bs]},
+ {test,bs_test_tail2,F,[Ctxt,0]}|Acc0]=Acc) ->
+ case beam_utils:is_killed(Ctxt, Acc0, D) of
+ true ->
+ Eq = {test,is_eq_exact,F,[R,{literal,Bs}]},
+ backward(Is, D, [Eq|Acc0]);
+ false ->
+ backward(Is, D, [I|Acc])
+ end;
backward([{test,bs_start_match2,{f,To0},Live,[Src|_]=Info,Dst}|Is], D, Acc) ->
To = shortcut_bs_start_match(To0, Src, D),
I = {test,bs_start_match2,{f,To},Live,Info,Dst},
backward(Is, D, [I|Acc]);
-backward([{test,is_eq_exact,{f,To0},[Reg,{atom,Val}]=Ops}|Is], D, Acc) ->
- To1 = shortcut_bs_test(To0, Is, D),
- To = shortcut_fail_label(To1, Reg, Val, D),
- I = combine_eqs(To, Ops, D, Acc),
- backward(Is, D, [I|Acc]);
backward([{test,Op,{f,To0},Ops0}|Is], D, Acc) ->
To1 = shortcut_bs_test(To0, Is, D),
To2 = shortcut_label(To1, D),
+ To3 = shortcut_rel_op(To2, Op, Ops0, D),
+
%% Try to shortcut a repeated test:
%%
%% test Op {f,Fail1} Operands test Op {f,Fail2} Operands
%% . . . ==> ...
%% Fail1: test Op {f,Fail2} Operands Fail1: test Op {f,Fail2} Operands
%%
- To = case beam_utils:code_at(To2, D) of
- [{test,Op,{f,To3},Ops}|_] ->
+ To = case beam_utils:code_at(To3, D) of
+ [{test,Op,{f,To4},Ops}|_] ->
case equal_ops(Ops0, Ops) of
- true -> To3;
- false -> To2
+ true -> To4;
+ false -> To3
end;
_Code ->
- To2
+ To3
end,
I = case Op of
is_eq_exact -> combine_eqs(To, Ops0, D, Acc);
@@ -367,8 +333,8 @@ equal_ops([Op|T0], [Op|T1]) ->
equal_ops([], []) -> true;
equal_ops(_, _) -> false.
-shortcut_select_list([{_,Val}=Lit,{f,To0}|T], Reg, D, Acc) ->
- To = shortcut_select_label(To0, Reg, Val, D),
+shortcut_select_list([Lit,{f,To0}|T], Reg, D, Acc) ->
+ To = shortcut_select_label(To0, Reg, Lit, D),
shortcut_select_list(T, Reg, D, [{f,To},Lit|Acc]);
shortcut_select_list([], _, _, Acc) -> reverse(Acc).
@@ -378,58 +344,29 @@ shortcut_label(To0, D) ->
_ -> To0
end.
-shortcut_select_label(To0, Reg, Val, D) ->
- case beam_utils:code_at(To0, D) of
- [{jump,{f,To}}|_] ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_atom,_,[Reg]},{select,select_val,Reg,{f,Fail},Map}|_] ->
- To = find_select_val(Map, Val, Fail),
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_eq_exact,{f,_},[Reg,{atom,Val}]},{label,To}|_] when is_atom(Val) ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_eq_exact,{f,_},[Reg,{atom,Val}]},{jump,{f,To}}|_] when is_atom(Val) ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_eq_exact,{f,To},[Reg,{atom,AnotherVal}]}|_]
- when is_atom(Val), Val =/= AnotherVal ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_ne_exact,{f,To},[Reg,{atom,Val}]}|_] when is_atom(Val) ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_ne_exact,{f,_},[Reg,{atom,_}]},{label,To}|_] when is_atom(Val) ->
- shortcut_select_label(To, Reg, Val, D);
- [{test,is_tuple,{f,To},[Reg]}|_] when is_atom(Val) ->
- shortcut_select_label(To, Reg, Val, D);
- _ ->
- To0
- end.
-
-shortcut_fail_label(To0, Reg, Val, D) ->
- case beam_utils:code_at(To0, D) of
- [{jump,{f,To}}|_] ->
- shortcut_fail_label(To, Reg, Val, D);
- [{test,is_eq_exact,{f,To},[Reg,{atom,Val}]}|_] when is_atom(Val) ->
- shortcut_fail_label(To, Reg, Val, D);
- _ ->
- To0
- end.
+shortcut_select_label(To, Reg, Lit, D) ->
+ shortcut_rel_op(To, is_ne_exact, [Reg,Lit], D).
-shortcut_boolean_label(To0, Reg, Bool0, D) when is_boolean(Bool0) ->
- case beam_utils:code_at(To0, D) of
- [{line,_},{bif,'not',_,[Reg],Reg},{jump,{f,To}}|_] ->
- Bool = not Bool0,
- {shortcut_select_label(To, Reg, Bool, D),Bool};
- _ ->
- {To0,Bool0}
- end;
-shortcut_boolean_label(To, _, Bool, _) -> {To,Bool}.
-
-find_select_val([{_,Val},{f,To}|_], Val, _) -> To;
-find_select_val([{_,_}, {f,_}|T], Val, Fail) ->
- find_select_val(T, Val, Fail);
-find_select_val([], _, Fail) -> Fail.
+%% Replace a comparison operator with a test instruction and a jump.
+%% For example, if we have this code:
+%%
+%% bif '=:=' Fail Src1 Src2 {x,0}
+%% jump L1
+%% .
+%% .
+%% .
+%% L1: select_val {x,0} FailLabel [... true => L2..., ...false => L3...]
+%%
+%% the first two instructions can be replaced with
+%%
+%% test is_eq_exact L3 Src1 Src2
+%% jump L2
+%%
+%% provided that {x,0} is killed at both L2 and L3.
replace_comp_op(To, Reg, Op, Ops, D) ->
- False = comp_op_find_shortcut(To, Reg, false, D),
- True = comp_op_find_shortcut(To, Reg, true, D),
+ False = comp_op_find_shortcut(To, Reg, {atom,false}, D),
+ True = comp_op_find_shortcut(To, Reg, {atom,true}, D),
[bif_to_test(Op, Ops, False),{jump,{f,True}}].
comp_op_find_shortcut(To0, Reg, Val, D) ->
@@ -461,9 +398,9 @@ not_possible() -> throw(not_possible).
%%
%% is_eq_exact F1 Reg Lit1 select_val Reg F2 [ Lit1 L1
%% L1: . Lit2 L2 ]
-%% .
-%% . ==>
-%% .
+%% .
+%% . ==>
+%% .
%% F1: is_eq_exact F2 Reg Lit2 F1: is_eq_exact F2 Reg Lit2
%% L2: .... L2:
%%
@@ -488,31 +425,26 @@ remove_from_list(Lit, [Val,{f,_}=Fail|T]) ->
[Val,Fail|remove_from_list(Lit, T)];
remove_from_list(_, []) -> [].
-%% shortcut_bs_test(TargetLabel, [Instruction], D) -> TargetLabel'
-%% Try to shortcut the failure label for a bit syntax matching.
-%% We know that the binary contains at least Bits bits after
-%% the latest save point.
+%% shortcut_bs_test(TargetLabel, ReversedInstructions, D) -> TargetLabel'
+%% Try to shortcut the failure label for bit syntax matching.
shortcut_bs_test(To, Is, D) ->
shortcut_bs_test_1(beam_utils:code_at(To, D), Is, To, D).
-shortcut_bs_test_1([{bs_restore2,Reg,SavePoint}|Is], PrevIs, To, D) ->
- shortcut_bs_test_2(Is, {Reg,SavePoint}, PrevIs, To, D);
-shortcut_bs_test_1([_|_], _, To, _) -> To.
-
-shortcut_bs_test_2([{label,_}|Is], Save, PrevIs, To, D) ->
- shortcut_bs_test_2(Is, Save, PrevIs, To, D);
-shortcut_bs_test_2([{test,bs_test_tail2,{f,To},[_,TailBits]}|_],
- {Reg,_Point} = RP, PrevIs, To0, D) ->
- case count_bits_matched(PrevIs, RP, 0) of
+shortcut_bs_test_1([{bs_restore2,Reg,SavePoint},
+ {label,_},
+ {test,bs_test_tail2,{f,To},[_,TailBits]}|_],
+ PrevIs, To0, D) ->
+ case count_bits_matched(PrevIs, {Reg,SavePoint}, 0) of
Bits when Bits > TailBits ->
%% This instruction will fail. We know because a restore has been
- %% done from the previous point SavePoint in the binary, and we also know
- %% that the binary contains at least Bits bits from SavePoint.
+ %% done from the previous point SavePoint in the binary, and we
+ %% also know that the binary contains at least Bits bits from
+ %% SavePoint.
%%
%% Since we will skip a bs_restore2 if we shortcut to label To,
- %% we must now make sure that code at To does not depend on the position
- %% in the context in any way.
+ %% we must now make sure that code at To does not depend on
+ %% the position in the context in any way.
case shortcut_bs_pos_used(To, Reg, D) of
false -> To;
true -> To0
@@ -520,15 +452,26 @@ shortcut_bs_test_2([{test,bs_test_tail2,{f,To},[_,TailBits]}|_],
_Bits ->
To0
end;
-shortcut_bs_test_2([_|_], _, _, To, _) -> To.
+shortcut_bs_test_1([_|_], _, To, _) -> To.
+%% counts_bits_matched(ReversedInstructions, SavePoint, Bits) -> Bits'
+%% Given a reversed instruction stream, determine the minimum number
+%% of bits that will be matched by bit syntax instructions up to the
+%% given save point.
+
+count_bits_matched([{test,bs_get_utf8,{f,_},_,_,_}|Is], SavePoint, Bits) ->
+ count_bits_matched(Is, SavePoint, Bits+8);
+count_bits_matched([{test,bs_get_utf16,{f,_},_,_,_}|Is], SavePoint, Bits) ->
+ count_bits_matched(Is, SavePoint, Bits+16);
+count_bits_matched([{test,bs_get_utf32,{f,_},_,_,_}|Is], SavePoint, Bits) ->
+ count_bits_matched(Is, SavePoint, Bits+32);
count_bits_matched([{test,_,_,_,[_,Sz,U,{field_flags,_}],_}|Is], SavePoint, Bits) ->
case Sz of
{integer,N} -> count_bits_matched(Is, SavePoint, Bits+N*U);
_ -> count_bits_matched(Is, SavePoint, Bits)
end;
-count_bits_matched([{test,bs_match_string,_,[_,Bits,_]}|Is], SavePoint, Bits0) ->
- count_bits_matched(Is, SavePoint, Bits0+Bits);
+count_bits_matched([{test,bs_match_string,_,[_,Bs]}|Is], SavePoint, Bits) ->
+ count_bits_matched(Is, SavePoint, Bits+bit_size(Bs));
count_bits_matched([{test,_,_,_}|Is], SavePoint, Bits) ->
count_bits_matched(Is, SavePoint, Bits);
count_bits_matched([{bs_save2,Reg,SavePoint}|_], {Reg,SavePoint}, Bits) ->
@@ -545,20 +488,332 @@ shortcut_bs_pos_used_1(Is, Reg, D) ->
not beam_utils:is_killed(Reg, Is, D).
%% shortcut_bs_start_match(TargetLabel, Reg) -> TargetLabel
-%% A failing bs_start_match2 instruction means that the source
-%% cannot be a binary, so there is no need to jump bs_context_to_binary/1
-%% or another bs_start_match2 instruction.
+%% A failing bs_start_match2 instruction means that the source (Reg)
+%% cannot be a binary. That means that it is safe to skip
+%% bs_context_to_binary instructions operating on Reg, and
+%% bs_start_match2 instructions operating on Reg.
shortcut_bs_start_match(To, Reg, D) ->
- shortcut_bs_start_match_1(beam_utils:code_at(To, D), Reg, To).
+ shortcut_bs_start_match_1(beam_utils:code_at(To, D), Reg, To, D).
+
+shortcut_bs_start_match_1([{bs_context_to_binary,Reg}|Is], Reg, To, D) ->
+ shortcut_bs_start_match_1(Is, Reg, To, D);
+shortcut_bs_start_match_1([{jump,{f,To}}|_], Reg, _, D) ->
+ Code = beam_utils:code_at(To, D),
+ shortcut_bs_start_match_1(Code, Reg, To, D);
+shortcut_bs_start_match_1([{test,bs_start_match2,{f,To},_,[Reg|_],_}|_],
+ Reg, _, D) ->
+ Code = beam_utils:code_at(To, D),
+ shortcut_bs_start_match_1(Code, Reg, To, D);
+shortcut_bs_start_match_1(_, _, To, _) ->
+ To.
+
+%% shortcut_rel_op(FailLabel, Operator, [Operand], D) -> FailLabel'
+%% Try to shortcut the given test instruction. Example:
+%%
+%% is_ge L1 {x,0} 48
+%% .
+%% .
+%% .
+%% L1: is_ge L2 {x,0} 65
+%%
+%% The first test instruction can be rewritten to "is_ge L2 {x,0} 48"
+%% since the instruction at L1 will also fail.
+%%
+%% If there are instructions between L1 and the other test instruction
+%% it may still be possible to do the shortcut. For example:
+%%
+%% L1: is_eq_exact L3 {x,0} 92
+%% is_ge L2 {x,0} 65
+%%
+%% Since the first test instruction failed, we know that {x,0} must
+%% be less than 48; therefore, we know that {x,0} cannot be equal to
+%% 92 and the jump to L3 cannot happen.
+
+shortcut_rel_op(To, Op, Ops, D) ->
+ case normalize_op({test,Op,{f,To},Ops}) of
+ {{NormOp,A,B},_} ->
+ Normalized = {negate_op(NormOp),A,B},
+ shortcut_rel_op_fp(To, Normalized, D);
+ {_,_} ->
+ To;
+ error ->
+ To
+ end.
-shortcut_bs_start_match_1([{bs_context_to_binary,Reg}|Is], Reg, To) ->
- shortcut_bs_start_match_2(Is, Reg, To);
-shortcut_bs_start_match_1(_, _, To) -> To.
+shortcut_rel_op_fp(To0, Normalized, D) ->
+ Code = beam_utils:code_at(To0, D),
+ case shortcut_any_label(Code, Normalized) of
+ error ->
+ To0;
+ To ->
+ shortcut_rel_op_fp(To, Normalized, D)
+ end.
-shortcut_bs_start_match_2([{jump,{f,To}}|_], _, _) ->
- To;
-shortcut_bs_start_match_2([{test,bs_start_match2,{f,To},_,[Reg|_],_}|_], Reg, _) ->
- To;
-shortcut_bs_start_match_2(_Is, _Reg, To) ->
- To.
+%% shortcut_any_label([Instruction], PrevCondition) -> FailLabel | error
+%% Using PrevCondition (a previous condition known to be true),
+%% try to shortcut to another failure label.
+
+shortcut_any_label([{jump,{f,Lbl}}|_], _Prev) ->
+ Lbl;
+shortcut_any_label([{label,Lbl}|_], _Prev) ->
+ Lbl;
+shortcut_any_label([{select,select_val,R,{f,Fail},L}|_], Prev) ->
+ shortcut_selectval(L, R, Fail, Prev);
+shortcut_any_label([I|Is], Prev) ->
+ case normalize_op(I) of
+ error ->
+ error;
+ {Normalized,Fail} ->
+ %% We have a relational operator.
+ case will_succeed(Prev, Normalized) of
+ no ->
+ %% This test instruction will always branch
+ %% to Fail.
+ Fail;
+ yes ->
+ %% This test instruction will never branch,
+ %% so we will look at the next instruction.
+ shortcut_any_label(Is, Prev);
+ maybe ->
+ %% May or may not branch. From now on, we can only
+ %% shortcut to the this specific failure label
+ %% Fail.
+ shortcut_specific_label(Is, Fail, Prev)
+ end
+ end.
+
+%% shortcut_specific_label([Instruction], FailLabel, PrevCondition) ->
+%% FailLabel | error
+%% We have previously encountered a test instruction that may or
+%% may not branch to FailLabel. Therefore we are only allowed
+%% to do the shortcut to the same fail label (FailLabel).
+
+shortcut_specific_label([{label,_}|Is], Fail, Prev) ->
+ shortcut_specific_label(Is, Fail, Prev);
+shortcut_specific_label([{select,select_val,R,{f,F},L}|_], Fail, Prev) ->
+ case shortcut_selectval(L, R, F, Prev) of
+ Fail -> Fail;
+ _ -> error
+ end;
+shortcut_specific_label([I|Is], Fail, Prev) ->
+ case normalize_op(I) of
+ error ->
+ error;
+ {Normalized,Fail} ->
+ case will_succeed(Prev, Normalized) of
+ no ->
+ %% Will branch to FailLabel.
+ Fail;
+ yes ->
+ %% Will definitely never branch.
+ shortcut_specific_label(Is, Fail, Prev);
+ maybe ->
+ %% May branch, but still OK since it will branch
+ %% to FailLabel.
+ shortcut_specific_label(Is, Fail, Prev)
+ end;
+ {Normalized,_} ->
+ %% This test instruction will branch to a different
+ %% fail label, if it branches at all.
+ case will_succeed(Prev, Normalized) of
+ yes ->
+ %% Still OK, since the branch will never be
+ %% taken.
+ shortcut_specific_label(Is, Fail, Prev);
+ no ->
+ %% Give up. The branch will definitely be taken
+ %% to a different fail label.
+ error;
+ maybe ->
+ %% Give up. If the branch is taken, it will be
+ %% to a different fail label.
+ error
+ end
+ end.
+
+
+%% shortcut_selectval(List, Reg, Fail, PrevCond) -> FailLabel | error
+%% Try to shortcut a selectval instruction. A selectval instruction
+%% is equivalent to the following instruction sequence:
+%%
+%% is_ne_exact L1 Reg Value1
+%% .
+%% .
+%% .
+%% is_ne_exact LN Reg ValueN
+%% jump DefaultFailLabel
+%%
+shortcut_selectval([Val,{f,Lbl}|T], R, Fail, Prev) ->
+ case will_succeed(Prev, {'=/=',R,get_literal(Val)}) of
+ yes -> shortcut_selectval(T, R, Fail, Prev);
+ no -> Lbl;
+ maybe -> error
+ end;
+shortcut_selectval([], _, Fail, _) -> Fail.
+
+%% will_succeed(PrevCondition, Condition) -> yes | no | maybe
+%% PrevCondition is a condition known to be true. This function
+%% will tell whether Condition will succeed.
+
+will_succeed({Op1,Reg,A}, {Op2,Reg,B}) ->
+ will_succeed_1(Op1, A, Op2, B);
+will_succeed({'=:=',Reg,{literal,A}}, {TypeTest,Reg}) ->
+ case erlang:TypeTest(A) of
+ false -> no;
+ true -> yes
+ end;
+will_succeed({_,_,_}, maybe) ->
+ maybe;
+will_succeed({_,_,_}, Test) when is_tuple(Test) ->
+ maybe.
+
+will_succeed_1('=:=', A, '<', B) ->
+ if
+ B =< A -> no;
+ true -> yes
+ end;
+will_succeed_1('=:=', A, '=<', B) ->
+ if
+ B < A -> no;
+ true -> yes
+ end;
+will_succeed_1('=:=', A, '=:=', B) ->
+ if
+ A =:= B -> yes;
+ true -> no
+ end;
+will_succeed_1('=:=', A, '=/=', B) ->
+ if
+ A =:= B -> no;
+ true -> yes
+ end;
+will_succeed_1('=:=', A, '>=', B) ->
+ if
+ B > A -> no;
+ true -> yes
+ end;
+will_succeed_1('=:=', A, '>', B) ->
+ if
+ B >= A -> no;
+ true -> yes
+ end;
+
+will_succeed_1('=/=', A, '=/=', B) when A =:= B -> yes;
+will_succeed_1('=/=', A, '=:=', B) when A =:= B -> no;
+
+will_succeed_1('<', A, '=:=', B) when B >= A -> no;
+will_succeed_1('<', A, '=/=', B) when B >= A -> yes;
+will_succeed_1('<', A, '<', B) when B >= A -> yes;
+will_succeed_1('<', A, '=<', B) when B > A -> yes;
+will_succeed_1('<', A, '>=', B) when B > A -> no;
+will_succeed_1('<', A, '>', B) when B >= A -> no;
+
+will_succeed_1('=<', A, '=:=', B) when B > A -> no;
+will_succeed_1('=<', A, '=/=', B) when B > A -> yes;
+will_succeed_1('=<', A, '<', B) when B > A -> yes;
+will_succeed_1('=<', A, '=<', B) when B >= A -> yes;
+will_succeed_1('=<', A, '>=', B) when B > A -> no;
+will_succeed_1('=<', A, '>', B) when B >= A -> no;
+
+will_succeed_1('>=', A, '=:=', B) when B < A -> no;
+will_succeed_1('>=', A, '=/=', B) when B < A -> yes;
+will_succeed_1('>=', A, '<', B) when B =< A -> no;
+will_succeed_1('>=', A, '=<', B) when B < A -> no;
+will_succeed_1('>=', A, '>=', B) when B =< A -> yes;
+will_succeed_1('>=', A, '>', B) when B < A -> yes;
+
+will_succeed_1('>', A, '=:=', B) when B =< A -> no;
+will_succeed_1('>', A, '=/=', B) when B =< A -> yes;
+will_succeed_1('>', A, '<', B) when B =< A -> no;
+will_succeed_1('>', A, '=<', B) when B < A -> no;
+will_succeed_1('>', A, '>=', B) when B =< A -> yes;
+will_succeed_1('>', A, '>', B) when B < A -> yes;
+
+will_succeed_1(_, _, _, _) -> maybe.
+
+%% normalize_op(Instruction) -> {Normalized,FailLabel} | error
+%% Normalized = {Operator,Register,Literal} |
+%% {TypeTest,Register} |
+%% maybe
+%% Operation = '<' | '=<' | '=:=' | '=/=' | '>=' | '>'
+%% TypeTest = is_atom | is_integer ...
+%% Literal = {literal,Term}
+%%
+%% Normalize a relational operator to facilitate further
+%% comparisons between operators. Always make the register
+%% operand the first operand. Thus the following instruction:
+%%
+%% {test,is_ge,{f,99},{integer,13},{x,0}}
+%%
+%% will be normalized to:
+%%
+%% {'=<',{x,0},{literal,13}}
+%%
+%% NOTE: Bit syntax test instructions are scary. They may change the
+%% state of match contexts and update registers, so we don't dare
+%% mess with them.
+
+normalize_op({test,is_ge,{f,Fail},Ops}) ->
+ normalize_op_1('>=', Ops, Fail);
+normalize_op({test,is_lt,{f,Fail},Ops}) ->
+ normalize_op_1('<', Ops, Fail);
+normalize_op({test,is_eq_exact,{f,Fail},Ops}) ->
+ normalize_op_1('=:=', Ops, Fail);
+normalize_op({test,is_ne_exact,{f,Fail},Ops}) ->
+ normalize_op_1('=/=', Ops, Fail);
+normalize_op({test,is_nil,{f,Fail},[R]}) ->
+ normalize_op_1('=:=', [R,nil], Fail);
+normalize_op({test,Op,{f,Fail},[R]}) ->
+ case erl_internal:new_type_test(Op, 1) of
+ true -> {{Op,R},Fail};
+ false -> {maybe,Fail}
+ end;
+normalize_op({test,_,{f,Fail},_}=I) ->
+ case beam_utils:is_pure_test(I) of
+ true -> {maybe,Fail};
+ false -> error
+ end;
+normalize_op(_) ->
+ error.
+
+normalize_op_1(Op, [Op1,Op2], Fail) ->
+ case {get_literal(Op1),get_literal(Op2)} of
+ {error,error} ->
+ %% Both operands are registers.
+ {maybe,Fail};
+ {error,Lit} ->
+ {{Op,Op1,Lit},Fail};
+ {Lit,error} ->
+ {{turn_op(Op),Op2,Lit},Fail};
+ {_,_} ->
+ %% Both operands are literals. Can probably only
+ %% happen if the Core Erlang optimizations passes were
+ %% turned off, so don't bother trying to do something
+ %% smart here.
+ {maybe,Fail}
+ end.
+
+turn_op('<') -> '>';
+turn_op('>=') -> '=<';
+turn_op('=:='=Op) -> Op;
+turn_op('=/='=Op) -> Op.
+
+negate_op('>=') -> '<';
+negate_op('<') -> '>=';
+negate_op('=<') -> '>';
+negate_op('>') -> '=<';
+negate_op('=:=') -> '=/=';
+negate_op('=/=') -> '=:='.
+
+get_literal({atom,Val}) ->
+ {literal,Val};
+get_literal({integer,Val}) ->
+ {literal,Val};
+get_literal({float,Val}) ->
+ {literal,Val};
+get_literal(nil) ->
+ {literal,[]};
+get_literal({literal,_}=Lit) ->
+ Lit;
+get_literal({_,_}) -> error.