chore: fix acNf bug in bv_decide by switching over to grind #9309
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Investigated. The issue is that the use of import Std.Tactic.BVDecide
set_option trace.Meta.Tactic.simp true
set_option trace.Meta.Tactic.simp.all true
set_option trace.Meta.AC true
-- set_option diagnostics true
theorem simp_timeout (x y : BitVec 32) : (s * (~~~ (0#32) + s)) = (s * (~~~ (0#32) + s)) := by
-- simp only [BitVec.not_zero, BitVec.reduceAllOnes]
ac_nf
theorem grind_success (x y : BitVec 32) : (s * (~~~ (0#32) + s)) = (s * (~~~ (0#32) + s)) := by
grind
theorem grind_success_original_problem (x y : BitVec 32) :
(s * (4294967295#32 + t) == s * (t + 4294967295#32)) = (s * (4294967295#32 + t) == s * (t + 4294967295#32)) := by
grind -- grind justifies this rewrite without needing to unfold.
/-
(declare-fun s () (_ BitVec 32))
(declare-fun t () (_ BitVec 32))
-- SAT (exists s t, not P s t ) => USAT (forall s t, P s t)
(assert (not (= (bvmul s (bvadd (bvnot #b00000000000000000000000000000000) t)) (bvmul s (bvnot (bvneg t))))))
-/
theorem original_report (s t : BitVec 32) : (s * ((~~~(0#32) + t)) = (s * (~~~(-t)))) := by
bv_normalize
-- ac_nf
bv_decide (config := { acNf := true }) |
This reverts commit 537b08a.
|
Oh, that's a bit unfortunate, it seems that import Std.Tactic.BVDecide
open Lean.Grind.CommRing
#synth Lean.Grind.CommRing (BitVec 256)
example (x y : BitVec 10) : x * y = y * x := by grind (config := {}) -- succeeds
/--
error: `grind` failed
case grind
x y : BitVec 256
h : ¬x * y = y * x
⊢ False
[grind] Goal diagnostics
[facts] Asserted facts
[grind] Issues
[issue] maximum recursion depth has been reached
use `set_option maxRecDepth <num>` to increase limit
use `set_option diagnostics true` to get diagnostic information
-/
#guard_msgs in example (x y : BitVec 256) : x * y = y * x := by grind (config := {}) -- fails |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The bug seems to be because
acNfUnnormalizedcreates asimpset that loops. I'll isolate an explanation of what's going on later.For now, we choose instead to justify our rewrites using
grind. We will improve this to usegrind's internalCommRingsolver, rather than all of the grind infrastructure.We make this change as we wish to implement more of bitwuzla's associativity/commutatitivy normalization, which rely on canonicalizing non-linear combinations (such as$$x (y + x) + x y \mapsto 2xy + x^2$$ ), which requires a theory solver at least as powerful as
ringto justify.