You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

500 lines
12 KiB

(*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This source code is licensed under the MIT license found in the
* LICENSE file in the root directory of this source tree.
*)
let%test_module _ =
( module struct
open Equality
let () = Trace.init ~margin:68 ()
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
(* let () =
* Trace.init ~margin:160
* ~config:(Result.ok_exn (Trace.parse "+Equality"))
* ()
*
* [@@@warning "-32"] *)
let printf pp = Format.printf "@\n%a@." pp
let pp = printf pp
let pp_classes = Format.printf "@\n@[<hv> %a@]@." pp_classes
let ( ! ) i = Term.integer (Z.of_int i)
let ( + ) = Term.add
let ( - ) = Term.sub
let ( * ) = Term.mul
[sledge] Simplify type conversions Summary: The treatment of type conversions is too complicated, non-uniform, etc. This diff attempts to simplify things by separating integer to integer conversions, which are interpreted, from others, which are essentially just uninterpreted functions. Integer conversions are now handled using two expression and term forms: Signed and Unsigned. These each interpret their argument as either a signed or unsigned number of a given bitwidth: ``` | Signed of {bits: int} (** [Ap1 (Signed {bits= n}, dst, arg)] is [arg] interpreted as an [n]-bit signed integer and injected into the [dst] type. That is, it two's-complement--decodes the low [n] bits of the infinite two's-complement encoding of [arg]. The injection into [dst] is a no-op, so [dst] must be an integer type with bitwidth at least [n]. *) | Unsigned of {bits: int} (** [Ap1 (Unsigned {bits= n}, dst, arg)] is [arg] interpreted as an [n]-bit unsigned integer and injected into the [dst] type. That is, it unsigned-binary--decodes the low [n] bits of the infinite two's-complement encoding of [arg]. The injection into [dst] is a no-op, so [dst] must be an integer type with bitwidth greater than [n]. *) | Convert of {src: Typ.t} (** [Ap1 (Convert {src}, dst, arg)] is [arg] converted from type [src] to type [dst], possibly with loss of information. The [src] and [dst] types must be [Typ.convertible] and must not both be [Integer] types. *) ``` Reviewed By: ngorogiannis Differential Revision: D18298140 fbshipit-source-id: 690f065b4
5 years ago
let f = Term.unsigned 8
let g = Term.rem
let wrt = Var.Set.empty
let t_, wrt = Var.fresh "t" ~wrt
let u_, wrt = Var.fresh "u" ~wrt
let v_, wrt = Var.fresh "v" ~wrt
let w_, wrt = Var.fresh "w" ~wrt
let x_, wrt = Var.fresh "x" ~wrt
let y_, wrt = Var.fresh "y" ~wrt
let z_, wrt = Var.fresh "z" ~wrt
let t = Term.var t_
let u = Term.var u_
let v = Term.var v_
let w = Term.var w_
let x = Term.var x_
let y = Term.var y_
let z = Term.var z_
let of_eqs l =
List.fold ~init:(wrt, true_)
~f:(fun (us, r) (a, b) -> and_eq us a b r)
l
|> snd
let and_eq a b r = and_eq wrt a b r |> snd
let and_ r s = and_ wrt r s |> snd
let or_ r s = or_ wrt r s |> snd
(* tests *)
let f1 = of_eqs [(!0, !1)]
let%test _ = is_false f1
let%expect_test _ =
pp f1 ; [%expect {| {sat= false; rep= [[-1 ]; [0 ]]} |}]
let%test _ = is_false (and_eq !1 !1 f1)
let f2 = of_eqs [(x, x + !1)]
let%test _ = is_false f2
let%expect_test _ =
pp f2 ;
[%expect {| {sat= false; rep= [[%x_5 ]; [-1 ]; [0 ]]} |}]
let f3 = of_eqs [(x + !0, x + !1)]
let%test _ = is_false f3
let%expect_test _ =
pp f3 ;
[%expect {| {sat= false; rep= [[%x_5 ]; [-1 ]; [0 ]]} |}]
let f4 = of_eqs [(x, y); (x + !0, y + !1)]
let%test _ = is_false f4
let%expect_test _ =
pp f4 ;
[%expect
{|
{sat= false; rep= [[%x_5 ]; [%y_6 %x_5]; [-1 ]; [0 ]]} |}]
let t1 = of_eqs [(!1, !1)]
let%test _ = is_true t1
let t2 = of_eqs [(x, x)]
let%test _ = is_true t2
let%test _ = is_false (and_ f3 t2)
let%test _ = is_false (and_ t2 f3)
let r0 = true_
let%expect_test _ =
pp r0 ; [%expect {| {sat= true; rep= [[-1 ]; [0 ]]} |}]
let%expect_test _ = pp_classes r0 ; [%expect {||}]
let%test _ = difference r0 (f x) (f x) |> Poly.equal (Some (Z.of_int 0))
let%test _ = difference r0 !4 !3 |> Poly.equal (Some (Z.of_int 1))
let r1 = of_eqs [(x, y)]
let%expect_test _ =
pp_classes r1 ;
pp r1 ;
[%expect
{|
%x_5 = %y_6
{sat= true; rep= [[%x_5 ]; [%y_6 %x_5]; [-1 ]; [0 ]]} |}]
let%test _ = entails_eq r1 x y
let r2 = of_eqs [(x, y); (f x, y); (f y, z)]
let%expect_test _ =
pp_classes r2 ;
pp r2 ;
[%expect
{|
%x_5 = %y_6 = %z_7 = ((u8) %x_5)
{sat= true;
rep= [[%x_5 ];
[%y_6 %x_5];
[%z_7 %x_5];
[((u8) %x_5) %x_5];
[-1 ];
[0 ]]} |}]
let%test _ = entails_eq r2 x z
let%test _ = entails_eq (or_ r1 r2) x y
let%test _ = not (entails_eq (or_ r1 r2) x z)
let%test _ = entails_eq (or_ f1 r2) x z
let%test _ = entails_eq (or_ r2 f3) x z
let%test _ = entails_eq r2 (f y) y
let%test _ = entails_eq r2 (f x) (f z)
let%test _ = entails_eq r2 (g x y) (g z y)
let%test _ = difference (or_ r1 r2) x z |> Poly.equal None
let%expect_test _ =
let r = of_eqs [(w, y); (y, z)] in
let s = of_eqs [(x, y); (y, z)] in
let rs = or_ r s in
pp r ;
pp s ;
pp rs ;
[%expect
{|
{sat= true;
rep= [[%w_4 ]; [%y_6 %w_4]; [%z_7 %w_4]; [-1 ]; [0 ]]}
{sat= true;
rep= [[%x_5 ]; [%y_6 %x_5]; [%z_7 %x_5]; [-1 ]; [0 ]]}
{sat= true; rep= [[%y_6 ]; [%z_7 %y_6]; [-1 ]; [0 ]]} |}]
let%test _ =
let r = of_eqs [(w, y); (y, z)] in
let s = of_eqs [(x, y); (y, z)] in
let rs = or_ r s in
entails_eq rs y z
let r3 = of_eqs [(g y z, w); (v, w); (g y w, t); (x, v); (x, u); (u, z)]
let%expect_test _ =
pp_classes r3 ;
pp r3 ;
[%expect
{|
%t_1 = %u_2 = %v_3 = %w_4 = %x_5 = %z_7 = (%y_6 rem %t_1)
= (%y_6 rem %t_1)
{sat= true;
rep= [[%t_1 ];
[%u_2 %t_1];
[%v_3 %t_1];
[%w_4 %t_1];
[%x_5 %t_1];
[%y_6 ];
[%z_7 %t_1];
[(%y_6 rem %v_3) %t_1];
[(%y_6 rem %z_7) %t_1];
[-1 ];
[0 ]]} |}]
let%test _ = entails_eq r3 t z
let%test _ = entails_eq r3 x z
let%test _ = entails_eq (and_ r2 r3) x z
let r4 = of_eqs [(w + !2, x - !3); (x - !5, y + !7); (y, z - !4)]
let%expect_test _ =
pp_classes r4 ;
pp r4 ;
[%expect
{|
(%z_7 + -4) = %y_6 (%z_7 + 3) = %w_4 (%z_7 + 8) = %x_5
{sat= true;
rep= [[%w_4 (%z_7 + 3)];
[%x_5 (%z_7 + 8)];
[%y_6 (%z_7 + -4)];
[%z_7 ];
[-1 ];
[0 ]]} |}]
let%test _ = entails_eq r4 x (w + !5)
let%test _ = difference r4 x w |> Poly.equal (Some (Z.of_int 5))
let r5 = of_eqs [(x, y); (g w x, y); (g w y, f z)]
let%test _ = Var.Set.equal (fv r5) (Var.Set.of_list [w_; x_; y_; z_])
let r6 = of_eqs [(x, !1); (!1, y)]
let%expect_test _ =
pp_classes r6 ;
pp r6 ;
[%expect
{|
1 = %x_5 = %y_6
{sat= true; rep= [[%x_5 1]; [%y_6 1]; [-1 ]; [0 ]]} |}]
let%test _ = entails_eq r6 x y
let r7 = of_eqs [(v, x); (w, z); (y, z)]
let%expect_test _ =
pp_classes r7 ;
pp r7 ;
pp (and_eq x z r7) ;
pp_classes (and_eq x z r7) ;
[%expect
{|
%v_3 = %x_5 %w_4 = %y_6 = %z_7
{sat= true;
rep= [[%v_3 ];
[%w_4 ];
[%x_5 %v_3];
[%y_6 %w_4];
[%z_7 %w_4];
[-1 ];
[0 ]]}
{sat= true;
rep= [[%v_3 ];
[%w_4 %v_3];
[%x_5 %v_3];
[%y_6 %v_3];
[%z_7 %v_3];
[-1 ];
[0 ]]}
%v_3 = %w_4 = %x_5 = %y_6 = %z_7 |}]
let%expect_test _ =
printf (List.pp " , " Term.pp) (Equality.class_of r7 t) ;
printf (List.pp " , " Term.pp) (Equality.class_of r7 x) ;
printf (List.pp " , " Term.pp) (Equality.class_of r7 z) ;
[%expect
{|
%t_1
%v_3 , %x_5
%w_4 , %z_7 , %y_6 |}]
let r7' = and_eq x z r7
let%expect_test _ =
pp_classes r7' ;
pp r7' ;
[%expect
{|
%v_3 = %w_4 = %x_5 = %y_6 = %z_7
{sat= true;
rep= [[%v_3 ];
[%w_4 %v_3];
[%x_5 %v_3];
[%y_6 %v_3];
[%z_7 %v_3];
[-1 ];
[0 ]]} |}]
let%test _ = normalize r7' w |> Term.equal v
let%test _ =
entails_eq (of_eqs [(g w x, g y z); (x, z)]) (g w x) (g w z)
let%test _ =
entails_eq (of_eqs [(g w x, g y w); (x, z)]) (g w x) (g w z)
let r8 = of_eqs [(x + !42, (!3 * y) + (!13 * z)); (!13 * z, x)]
let%expect_test _ =
pp_classes r8 ;
pp r8 ;
[%expect
{|
(13 × %z_7) = %x_5 14 = %y_6
{sat= true;
rep= [[%x_5 (13 × %z_7)]; [%y_6 14]; [%z_7 ]; [-1 ]; [0 ]]} |}]
let%test _ = entails_eq r8 y !14
let r9 = of_eqs [(x, z - !16)]
let%expect_test _ =
pp_classes r9 ;
pp r9 ;
[%expect
{|
(%z_7 + -16) = %x_5
{sat= true;
rep= [[%x_5 (%z_7 + -16)]; [%z_7 ]; [-1 ]; [0 ]]} |}]
let%test _ = difference r9 z (x + !8) |> Poly.equal (Some (Z.of_int 8))
let r10 = of_eqs [(!16, z - x)]
let%expect_test _ =
pp_classes r10 ;
pp r10 ;
Format.printf "@.%a@." Term.pp (z - (x + !8)) ;
Format.printf "@.%a@." Term.pp (normalize r10 (z - (x + !8))) ;
Format.printf "@.%a@." Term.pp (x + !8 - z) ;
Format.printf "@.%a@." Term.pp (normalize r10 (x + !8 - z)) ;
[%expect
{|
(%z_7 + -16) = %x_5
{sat= true;
rep= [[%x_5 (%z_7 + -16)]; [%z_7 ]; [-1 ]; [0 ]]}
(-1 × %x_5 + %z_7 + -8)
8
(%x_5 + -1 × %z_7 + 8)
-8 |}]
let%test _ = difference r10 z (x + !8) |> Poly.equal (Some (Z.of_int 8))
let%test _ =
difference r10 (x + !8) z |> Poly.equal (Some (Z.of_int (-8)))
let r11 = of_eqs [(!16, z - x); (x + !8 - z, z - !16 + !8 - z)]
let%expect_test _ = pp_classes r11 ; [%expect {| (%z_7 + -16) = %x_5 |}]
let r12 = of_eqs [(!16, z - x); (x + !8 - z, z + !16 + !8 - z)]
let%expect_test _ = pp_classes r12 ; [%expect {| (%z_7 + -16) = %x_5 |}]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let r13 = of_eqs [(Term.eq x !2, y); (Term.dq x !2, z); (y, z)]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let%expect_test _ =
pp r13 ;
[%expect
{|
{sat= true;
rep= [[%x_5 ];
[%y_6 ];
[%z_7 %y_6];
[(%x_5 = 2) %y_6];
[(%x_5 2) %y_6];
[-1 ];
[0 ]]} |}]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let%test _ = not (is_false r13) (* incomplete *)
let a = Term.dq x !0
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let r14 = of_eqs [(a, a); (x, !1)]
let%expect_test _ =
pp r14 ;
[%expect
{|
{sat= true; rep= [[%x_5 1]; [(%x_5 0) -1]; [-1 ]; [0 ]]} |}]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let%test _ = entails_eq r14 a Term.true_
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let b = Term.dq y !0
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let r14 = of_eqs [(a, b); (x, !1)]
let%expect_test _ =
pp r14 ;
[%expect
{|
{sat= true;
rep= [[%x_5 1];
[%y_6 ];
[(%x_5 0) -1];
[(%y_6 0) -1];
[-1 ];
[0 ]]} |}]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let%test _ = entails_eq r14 a Term.true_
let%test _ = entails_eq r14 b Term.true_
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let b = Term.dq x !0
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
let r15 = of_eqs [(b, b); (x, !1)]
let%expect_test _ =
pp r15 ;
[%expect
{|
{sat= true; rep= [[%x_5 1]; [(%x_5 0) -1]; [-1 ]; [0 ]]} |}]
[sledge] Classify fully-interpreted and simplified exps differently Summary: For some Exp forms, Exp.solve is not complete, and this is necessary since the result of solve is a substitution that needs to encode the input equation as a conjunction of equations each between a variable and an exp. This is tantamount to, stronger even than, the theory being convex. So Exp.solve is not complete for some exps, and some of those have constructors that perform some simplification. For example, `(1 ≠ 0)` simplifies to `-1` (i.e. true). To enable deductions such as `((x ≠ 0) = b) && x = 1 |- b = -1` needs the equality solver to substitute through subexps of simplifiable exps like = and ≠, as it does for interpreted exps like + and ×. At the same time, since Exp.solve for non-interpreted exps cannot be complete, to enable deductions such as `((x ≠ 0) = (y ≠ 0)) && x = 1 |- y ≠ 0` needs the equality solver to congruence-close over subexps of simplifiable exps such as = and ≠, as it does for uninterpreted exps. To strengthen the equality solver in these sorts of cases, this diff adds a new class of exps for = and ≠, and revises the equality solver to handle them in a hybrid fashion between interpreted and uninterpreted. I am not currently sure whether or not this breaks the termination proof, but I have also not managed to adapt usual divergent cases to break this. One notable point is that simplifying = and ≠ exps always produces genuinely simpler and smaller exps, in contrast to e.g. polynomial simplification and gaussian elimination. Note that the real solution to this problem is likely to be to eliminate the i1 type in favor or a genuine boolean type, and translate all integer operations on i1 to boolean/logical ones. Then the disjunction implicit in e.g. equations between disequations would appear as actual disjunction, and could be dealt with as such. Reviewed By: jvillard Differential Revision: D15424823 fbshipit-source-id: 67d62df1f
6 years ago
[sledge] Simplify type conversions Summary: The treatment of type conversions is too complicated, non-uniform, etc. This diff attempts to simplify things by separating integer to integer conversions, which are interpreted, from others, which are essentially just uninterpreted functions. Integer conversions are now handled using two expression and term forms: Signed and Unsigned. These each interpret their argument as either a signed or unsigned number of a given bitwidth: ``` | Signed of {bits: int} (** [Ap1 (Signed {bits= n}, dst, arg)] is [arg] interpreted as an [n]-bit signed integer and injected into the [dst] type. That is, it two's-complement--decodes the low [n] bits of the infinite two's-complement encoding of [arg]. The injection into [dst] is a no-op, so [dst] must be an integer type with bitwidth at least [n]. *) | Unsigned of {bits: int} (** [Ap1 (Unsigned {bits= n}, dst, arg)] is [arg] interpreted as an [n]-bit unsigned integer and injected into the [dst] type. That is, it unsigned-binary--decodes the low [n] bits of the infinite two's-complement encoding of [arg]. The injection into [dst] is a no-op, so [dst] must be an integer type with bitwidth greater than [n]. *) | Convert of {src: Typ.t} (** [Ap1 (Convert {src}, dst, arg)] is [arg] converted from type [src] to type [dst], possibly with loss of information. The [src] and [dst] types must be [Typ.convertible] and must not both be [Integer] types. *) ``` Reviewed By: ngorogiannis Differential Revision: D18298140 fbshipit-source-id: 690f065b4
5 years ago
let%test _ = entails_eq r15 b (Term.signed 1 !1)
let%test _ = entails_eq r15 (Term.unsigned 1 b) !1
(* f(x1)1=x+1, f(y)+1=y1, y+1=x ⊢ false *)
let r16 =
of_eqs [(f (x - !1) - !1, x + !1); (f y + !1, y - !1); (y + !1, x)]
let%expect_test _ =
pp r16 ;
[%expect
{|
{sat= false;
rep= [[%x_5 (%y_6 + 1)];
[%y_6 ];
[((u8) %y_6) (%y_6 + -2)];
[((u8) (%x_5 + -1)) (%y_6 + 3)];
[-1 ];
[0 ]]} |}]
let%test _ = is_false r16
(* f(x) = x, f(y) = y 1, y = x ⊢ false *)
let r17 = of_eqs [(f x, x); (f y, y - !1); (y, x)]
let%expect_test _ =
pp r17 ;
[%expect
{|
{sat= false;
rep= [[%x_5 ];
[%y_6 %x_5];
[((u8) %x_5) %x_5];
[((u8) %y_6) (%x_5 + -1)];
[-1 ];
[0 ]]} |}]
let%test _ = is_false r17
let%expect_test _ =
let r18 = of_eqs [(f x, x); (f y, y - !1)] in
pp r18 ;
pp_classes r18 ;
[%expect
{|
{sat= true;
rep= [[%x_5 ];
[%y_6 ];
[((u8) %x_5) %x_5];
[((u8) %y_6) (%y_6 + -1)];
[-1 ];
[0 ]]}
%x_5 = ((u8) %x_5) (%y_6 + -1) = ((u8) %y_6) |}]
let r19 = of_eqs [(x, y + z); (x, !0); (y, !0)]
let%expect_test _ =
pp r19 ;
[%expect
{|
{sat= true;
rep= [[%x_5 0]; [%y_6 0]; [%z_7 0]; [-1 ]; [0 ]]} |}]
let%test _ = entails_eq r19 z !0
let%expect_test _ =
Equality.replay
{|(And_eq () (Var (id 10) (name v))
(Mul (((Var (id 8) (name v)) 1) ((Var (id 9) (name v)) 1)))
((xs ()) (sat true) (rep ())))|} ;
[%expect {| |}]
end )