Summary:
LLVM.global_initializer casues a cast assertion failure if the value
passed to it is not a GlobalVariable. So we first check if it is a
GlobalVariable and only then ask for an initiliazer.
This is hidden if LLVM is built without assertions.
Reviewed By: jberdine
Differential Revision: D15601632
fbshipit-source-id: e9db23a12
Summary:
Sledge does not terminate on programs with recursion, because
functions get "infinitely inlined" and therefore recursion is not
treated as retreating edge.
This patch bounds the number of times the same function can "inlined"
to respect the bound (`-b` option). On each call we check the number of
occurances of the called function in the call stack. If that is higher
than the bound, we skip it.
Reviewed By: jvillard
Differential Revision: D15577134
fbshipit-source-id: 4cd3b62c6
Summary:
This diff changes the translation of global variables to translate the
initializer whenever it exists in LLVM, rather than relying on
linkage. Previously code such as
```
char *mutable_string = "hahaha";
```
would lead to LLVM code
```
nutritious_string = global i8* getelementptr inbounds ([7 x i8], [7 x i8]* @.str, i32 0, i32 0), align 8, !dbg !0 ; [#uses=2 type=i8**]
```
in which `mutable_string` had `External` linkage according to
`Llvm.linkage` even though it has an initializer. This could cause
sledge to drop the initializer.
Reviewed By: kren1
Differential Revision: D15577755
fbshipit-source-id: 50aa06c5e
Summary:
Global variables have pointer type. The size needed by the backend is
of the element type, not of the pointer itself. This diff corrects
this.
Reviewed By: kren1
Differential Revision: D15577756
fbshipit-source-id: 948ecf3cd
Summary:
The root cause is not clear, but it seems that not calling
Llvm.dispose_context avoids segfaults in the GC.
Reviewed By: kren1
Differential Revision: D15535434
fbshipit-source-id: 280e44d0b
Summary:
The name and loc tables are added-to almost exactly in sync, so
combine them to amortize the overhead.
Reviewed By: kren1
Differential Revision: D15535435
fbshipit-source-id: 801da75bb
Summary:
Format is slow. Especially Format.sprintf, which has to allocate and
initialize a buffer every time.
Reviewed By: kren1
Differential Revision: D15535437
fbshipit-source-id: ea43f44e1
Summary:
This isn't free and is expected to hold of bitcode produced by
clang/llvm. There are tests that fail verification, so keep it in
debug mode.
Reviewed By: kren1
Differential Revision: D15535438
fbshipit-source-id: 9390a8363
Summary:
If the input file has a .bc or .ll suffix, treat it as a pre-linked
bitcode file. Otherwise, treat it as before, as a file containing a
list of bitcode files to be linked. Also, perform global dead-code
elimination only when linking multiple files.
Reviewed By: kren1
Differential Revision: D15513345
fbshipit-source-id: 4c80ff9c3
Summary:
Change the sledge input format from a bitcode file to a newline
separated list of paths to LLVM bitcode files.
Reviewed By: jberdine
Differential Revision: D15470082
fbshipit-source-id: 8860f947c
Summary:
Llvm.string_of_llvalue, which just calls llvm::Value::print, is
extremely slow when called on instructions or functions. In these
cases, it initializes metadata slots for everything in the *parent
module* of the instruction or function being printed, on every
call. This is ridiculously slow, don't do it.
Reviewed By: kren1
Differential Revision: D15511376
fbshipit-source-id: 658eeccab
Summary:
Add shortcut code paths to return early in some cases guaranteed to be
the identity function.
Reviewed By: ngorogiannis
Differential Revision: D15468704
fbshipit-source-id: f137049c6
Summary:
It is pointless to track membership of atomic exps in the congruence
relation, as they cannot have any subexps which might later become
equal to something else.
Reviewed By: jvillard
Differential Revision: D15424820
fbshipit-source-id: 048dbc9e1
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
Summary:
* Adds compilation of cxxabi.bc
* Includes the cxxabi.bc into the sledge executable via opam-crunch
* Links cxxabi.bc in sledge frontend
Reviewed By: jberdine
Differential Revision: D15415190
fbshipit-source-id: cc42f09fb
Summary:
There are two motivations for this:
1. Distinguish between `Unreachable`, which silently terminates
execution a la `assume false`; and `Abort`, which vocally
terminates execution a la `assert false`.
2. Distinguish between undefined functions, which have `Unreachable`
bodies, and bomb functions such as:
```
define void bomb() {
tail call void llvm.trap()
unreachable
}
```
Reviewed By: ngorogiannis
Differential Revision: D15408246
fbshipit-source-id: b64354cdb
Summary:
Require a final `()` argument to explicitly indicate the end of the
arguments to the printf-like functions. For `warn` this is not
any safer because the return type is `unit` anyhow, but for `fail` the
return type is polymorphic so the final `()` prevents unintentionally
forgetting an argument.
Reviewed By: ngorogiannis
Differential Revision: D15403367
fbshipit-source-id: ce3fe4035
Summary:
llvm.trap is noreturn nounwind so calls to it are always succeeded by
Unreachable, therefore unless an alarm is desired for reaching it,
translating it as nop suffices.
Reviewed By: ngorogiannis
Differential Revision: D15328302
fbshipit-source-id: 54efe6c21
Summary:
A previous rebase fumble sometimes led to the wrong arguments being
passed to retpolines of Invoke instructions.
Reviewed By: ngorogiannis
Differential Revision: D15323950
fbshipit-source-id: f6eb6fbe2
Summary:
It is (now?) the case that `[%sexp_of: type_name]` generates just
`sexp_of_type_name`, rather than expanding `type_name` to its
definition and generating a conversion function for that. Hence, when
such cases appear within `let rec sexp_of_type_name`, they get
captured, sometimes leading to divergence.
This diff fixes this by manually expanding such types into their
definitions.
Reviewed By: ngorogiannis
Differential Revision: D15314736
fbshipit-source-id: 716fff7cc
Summary: There was a missing case for singleton monomials of degree > 1.
Reviewed By: mbouaziz
Differential Revision: D15098810
fbshipit-source-id: e6d17d899
Summary:
Boolean Xor expressions are treated directly by the solver, so the 'no
new subexps' invariant for congruence closure is not needed.
Reviewed By: mbouaziz
Differential Revision: D15098817
fbshipit-source-id: d118c5881
Summary: Shifting by too many bits is undefined, so do not normalize them.
Reviewed By: mbouaziz
Differential Revision: D15098812
fbshipit-source-id: 96f4606d7
Summary:
When constructing division expressions from rational coefficients of
polynomials, the constant numerator and denominator should be clamped
to the number of bits in the result type.
Reviewed By: mbouaziz
Differential Revision: D15098825
fbshipit-source-id: fa448c39a
Summary:
Some constants are used during Exp normalization, to encode
subtraction using addition, units of addition and multiplication,
etc. Previously these were unconditionally integers, but need to be
floats for float arithmetic.
Reviewed By: mbouaziz
Differential Revision: D15098815
fbshipit-source-id: 13d1be142
Summary:
The frontend would implicitly assume there was (at least) one argument
to calls to operator new. If code declares operator new with the wrong
type, this can lead to crashing trying to access a missing arg.
Reviewed By: mbouaziz
Differential Revision: D15098820
fbshipit-source-id: 539281a83
Summary: These relaxations are needed for some of the llvm test suite.
Reviewed By: mbouaziz
Differential Revision: D15098813
fbshipit-source-id: 702d3ffd9
Summary:
Some globals have initializers which refer to the global
itself (e.g. for program counter relative offsets). Memoizing
translation of globals gives enough machinery to detect and handle
this situation.
Reviewed By: mbouaziz
Differential Revision: D15098819
fbshipit-source-id: ecc9dce92
Summary:
In some cases there were extra metadata operands (such as alignment of
an alloc) which would cause the debug locations to not be recorded.
Reviewed By: mbouaziz
Differential Revision: D15098816
fbshipit-source-id: a7e83f590
Summary:
When comparing stacks as part of a control-flow edge, treat each as a
code location in a hypothetical expansion of the program where all
non-recursive functions have been completely inlined. In particular,
this means to compare stacks as if all Locals frames or Return frames
for recursive calls had been removed. Additionally, the from_call info
in Return frames is ignored.
Reviewed By: jvillard
Differential Revision: D14657601
fbshipit-source-id: b8a42d3fa
Summary:
- Ensure that popping Throw or Return does not leave stale Throw or
Return frames
- Add a module for Control.stack
- Add invariant to enforce that a Throw frame can appear only
immediately above a Return frame
Reviewed By: jvillard
Differential Revision: D14547263
fbshipit-source-id: deb31b8af
Summary:
The backend does not generally interpret conditional expressions
specially. Therefore it is logically stronger to hoist them over other
boolean exps, for example normalizing
`e = (c ? t : f)` to `(c ? e = t : e =f)`
This enables the treatment of top-level conditional expressions in
terms of disjunction, conjunction and negation.
Reviewed By: mbouaziz
Differential Revision: D14495815
fbshipit-source-id: 4f2e8053f
Summary:
Eq and Dq expressions are interpreted by Equality and Exp, rather than
being considered uninterpreted functions. Classifying them as
interpreted results in stronger Equality normalization.
Reviewed By: mbouaziz
Differential Revision: D14495817
fbshipit-source-id: 44bb376c0
Summary: Also fix the spec for posix_memalign, and minor cleanup.
Reviewed By: ngorogiannis
Differential Revision: D14403648
fbshipit-source-id: 6b1cb3e3a
Summary:
The SL solver is currently not always able to append segments which
have been split symbolically, that is, at an internal point expressed
using a variable, rather than merely a constant.
Also, existential instantiation, that is, the choice of witnesses
during proof search, is currently sensitive to the order of
subformulas. This can lead to fragile incompleteness.
Reviewed By: mbouaziz
Differential Revision: D14481991
fbshipit-source-id: 80fe2f0a8
Summary:
Strengthen the Sh.seg constructor and Sh.is_false test to account for
the axiom that `null -[_;_)-> ⟨_,_⟩` is inconsistent.
Reviewed By: ngorogiannis
Differential Revision: D14481986
fbshipit-source-id: 7016e7451
Summary:
The dnf implementation dates to before nested existentials were
added. Updating it was overlooked, and it is just wrong.
Reviewed By: ngorogiannis
Differential Revision: D14481988
fbshipit-source-id: 9bba570f0
Summary:
P ∨ (∃x. Q ∨ R) could be simplified to (∃x. P ∨ Q ∨ R) and capture
occurrences of x in P.
Reviewed By: ngorogiannis
Differential Revision: D14481990
fbshipit-source-id: 92b474d59
Summary:
Also minor change to make strlen implementation syntactically closer
to spec.
Reviewed By: ngorogiannis
Differential Revision: D14403647
fbshipit-source-id: 48c771329
Summary:
This diff adds support in symbolic execution for calls to intrinsic
functions, to be used in lieu of adding a separate Llair instruction
for each intrinsic. This involves:
- adding skeleton support in Exec for symbolically execution an
intrinsic function call;
- exposing this in Domain;
- allowing symbolic execution of block terminators (e.g. function
call) to possibly fail; and
- generalizing Report for failing terminators.
Reviewed By: ngorogiannis
Differential Revision: D14403652
fbshipit-source-id: d86d9d1b8
Summary:
Some functions are translated directly to an expression, rather than a
function call. The llvm instructions for calling functions that may
and may not raise exceptions differ (Invoke and Call, resp.). This
diff adds support for the Invoke case.
Reviewed By: jvillard
Differential Revision: D14385596
fbshipit-source-id: 24458d960
Summary:
These tell llvm that contents of memory will not change during
execution of some code, which the analyzer does not need.
Reviewed By: jvillard
Differential Revision: D14385597
fbshipit-source-id: 0ef566a6f
Summary:
The backend has an implicit assumption that the entry block's locals
are all the function locals, not just those that happen to appear in
the entry block.
An alternative would be to collect the inverse renaming substitutions
produced by symbolic execution of jumps, which rename variables to
avoid clashes with the destination block's locals. These inverse
substitutions could be applied upon function return, in inverse
order. The complication with this approach is that it would be
necessary to distinguish between a destination local that clashes
because it has the same name as a local of some function higher in the
call stack versus a clash due to being a previous incarnation of the
same local due to a control-flow cycle. Accumulating unrenamings for
the latter case is expensive and pointless, as well as incongrous with
true interprocedural extensions.
So it seems preferable to exploit the asymmetry between the entry
block and others a bit more.
Reviewed By: jvillard
Differential Revision: D14354530
fbshipit-source-id: 815cdb224
Summary:
Interpreted subexps were not handled correctly: they must not be in
the carrier, but their non-interpreted subexps should.
Reviewed By: jvillard
Differential Revision: D14344291
fbshipit-source-id: 995896640
Summary:
In case the starting locations of two heap segments are
related (provably equal up to some offset), add equations between
their enclosing block to the goal. In these cases, the enclosing
blocks must be the same, so no completeness is lost. This has the
effect of instantiating existentials in the enclosing block prior to
others, which can avoid incomplete instantiation guesses.
Reviewed By: mbouaziz
Differential Revision: D14323550
fbshipit-source-id: 89a34a2c8
Summary: An initial set of basic sanity checks for frame inference.
Reviewed By: mbouaziz
Differential Revision: D14323549
fbshipit-source-id: d7cd4235f
Summary:
`('a, exn) result` is a more widely used type than with something
custom in place of `exn`, e.g. by the `Result` operations.
Reviewed By: mbouaziz
Differential Revision: D14322627
fbshipit-source-id: e2ed167ed
Summary:
- Change representation of Concat expressions from curried binary
operator to an nary one. This enables normalizing Concat expressions
with respect to associativity.
- Generalize Exp.solve to return a map rather than a pair of exps, to
enable expressing cases where solving an equation yields multiple
equations.
- Strengthen solver with implied equalities between sums of sizes of
concatenations of byte arrays.
Reviewed By: ngorogiannis
Differential Revision: D14297865
fbshipit-source-id: b40871559
Summary:
This patch adds an embarrassingly inefficient implementation of a
decision procedure for equality in the theories of uninterpreted
functions and linear arithmetic. A Shostak-style approach, where a
single congruence closure structure is shared by all theories, is
used. This is mostly based on the corrected variant of Shostak's
algorithm from:
Harald Rueβ and Natarajan Shankar. 2001. Deconstructing Shostak. In
Proceedings of the 16th Annual IEEE Symposium on Logic in Computer
Science (LICS '01).
Reviewed By: jvillard
Differential Revision: D14251655
fbshipit-source-id: 8a080145f
Summary:
Now with rational coefficients, dividing a polynomial by an integer
can be simplified.
Reviewed By: jvillard
Differential Revision: D14251656
fbshipit-source-id: 01094bc61
Summary:
Exp simplification must not generate fresh subexpressions in its
output that do not appear in its input, lest congruence closure gets
confused. Previously this check only considered immediate
subexpressions, which is overly restrictive.
Reviewed By: ngorogiannis
Differential Revision: D14251653
fbshipit-source-id: f8d5d9756
Summary:
Trace.report is essentially redundant with Trace.fail, and does not
behave as well wrt flushing when raising exceptions.
Reviewed By: ngorogiannis
Differential Revision: D14251657
fbshipit-source-id: 69a61c915
Summary:
Represent Add and Mul directly as associative and commutative nary
operations, rather than have Add and Mul operators and a separate AppN
nary application.
Reviewed By: ngorogiannis
Differential Revision: D14231544
fbshipit-source-id: 4fb7a06cf
Summary:
- Add nary expressions implemented using a form of multisets which
support any integer multiplicity
- Reimplement polynomials using new nary expressions
- Move the decomposition of exps into "base plus offset" form into
Exp, to enforce simplification invariants
- Revise expression simplification to cooperate with congruence
closure (mainly: simplification should not invent new
subexpressions)
- Reimplement congruence closure plus integer offsets to
+ cope with new representation of polynomials using nary expression forms
+ be diligent about maintaining which expressions are in the relation
+ add lots of invariant checking for the correlations between the
componnents of the congruence closure data structures
Reviewed By: jvillard
Differential Revision: D14075512
fbshipit-source-id: 2dbaf3d11
Summary:
Even pretty-printed congruence relations are verbose, and most
operations do not change much of the relation. So on return, print
only the symmetric difference between input and output relations.
Reviewed By: mbouaziz
Differential Revision: D14075513
fbshipit-source-id: b1f0ae6d0
Summary:
This is significant for operations involving the NULL pointer, which
is an integer of pointer type.
Reviewed By: mbouaziz
Differential Revision: D14075516
fbshipit-source-id: b727cb4c8
Summary:
Trace.parse is exposed across a module boundary, so handle its own
exceptions and return an `error`.
Reviewed By: mbouaziz
Differential Revision: D14081602
fbshipit-source-id: 4087c5d5b
Summary:
This allows exposing `Trace.parse`, which can be used to call
`Trace.init` in test code.
Reviewed By: mbouaziz
Differential Revision: D14075522
fbshipit-source-id: 4715c643f
Summary:
This makes it convenient to call
`Trace.init ~margin:70 ~config:all ()`
in test modules.
Reviewed By: mbouaziz
Differential Revision: D14075515
fbshipit-source-id: 3d3940eb1
Summary:
Comparison operations always return Typ.bool, and Add/Mul/Integer exps
are typed. Use these types to check that arguments are type correct up
to Typ.castable.
Reviewed By: mbouaziz
Differential Revision: D12854504
fbshipit-source-id: fff1426d5
Summary:
Require Exp clients to provide the type of the result of arithmetic
operations.
Reviewed By: mbouaziz
Differential Revision: D12854511
fbshipit-source-id: cc91a39ca
Summary:
Previously only the arguments of binary operations on integers were
clamped to the given bitwidth, but the results could overflow. This
was incorrect.
Also, use a clamped version of equality to simplify equality
expressions. This was previously overlooked.
Reviewed By: mbouaziz
Differential Revision: D12854502
fbshipit-source-id: 8786b4217
Summary:
Types of integer constants, in particular their bit-width, are
necessary for:
- correctly interpreting bitwise operations (e.g. `-1 xor 1` at type
`i1` is `0` while without the type the result is `-2`), and;
- distinguishing between integers and booleans, which are one-bit
integers, since booleans admit stronger algebraic simplification.
Note that code does genuinely treat 1-bit integers interchangeably as
booleans and integers, e.g. with expressions such as `e + (b != 42)`.
Therefore a lighter-weight early syntactic distinction between boolean
and bitwise operations is nontrivial/impossible to make robust.
This patch:
- adds the type to the representation of Exp.Integer;
- adds checks that Integer values fit within their specified bit-width
- factors out code handling 1-bit integers as booleans into `Z`, as it
is easy to make mistakes when forgetting that `-1`, not `1`, is the
representation of `true`;
- corrects the treatment of Exp.Convert in some cases involving
treating negative integers as unsigned;
- corrects and strengthens Exp simplification based on the bit-width
information;
- removes the `e - e ==> 0` simplification, due to not having the type
for `0`.
Reviewed By: mbouaziz
Differential Revision: D10488407
fbshipit-source-id: ff4320a29
Summary:
The main motivation is to factor out the code that determines the test
condition, e.g. to uniformly test `!= 0` vs `= 1`.
Also a convenience, and there are enough uses to make it worthwhile
and a readability improvement.
Reviewed By: mbouaziz
Differential Revision: D10488408
fbshipit-source-id: 520e843f3
Summary:
Just a convenience, but there are enough uses to make it worthwhile,
and a readability improvement.
Reviewed By: mbouaziz
Differential Revision: D10488404
fbshipit-source-id: 30cca06d5
Summary:
`sledge -c foo.bc` generates a binary file `foo.bc.llair` which can be
analyzed with `sledge foo.bc.llair`, and converted to textual form
with `sledge -c foo.bc.llair -o foo.llair`. Binary files are not
compatible between debug and release builds.
Reviewed By: mbouaziz
Differential Revision: D10389473
fbshipit-source-id: dfcabf33b
Summary:
Also for debugging support, note that analysis can be stopped after
reporting an attempt to call an unknown function with `sledge
-tReport.unknown_call` and likewise for invalid memory accesses with
`sledge -tReport.invalid_access`.
Reviewed By: mbouaziz
Differential Revision: D10389474
fbshipit-source-id: b006480d3
Summary:
Call __cxa_demangle from libstdc++ using Ctypes instead of the
wrapper in Llvm.
Reviewed By: mbouaziz
Differential Revision: D9939682
fbshipit-source-id: a9f02fff5
Summary:
Change the license of the source code from BSD + PATENTS to MIT.
Change `checkCopyright` to reflect the new license and learn some new file
types.
Generated with:
```
git grep BSD | xargs -n 1 ./scripts/checkCopyright -i
```
Reviewed By: jeremydubreil, mbouaziz, jberdine
Differential Revision: D8071249
fbshipit-source-id: 97ca23a
Summary:
Upgrade ocamlformat, and base which needs to be done in sync in order to build
ocamlformat, and the other deps can come for the ride.
Reviewed By: jvillard
Differential Revision: D7663537
fbshipit-source-id: 3e90970
Timotej Kapus
Josh Berdine
Mehdi Bouaziz
Jules Villard