and to expose a handle to represent the actual case rather than having
the iterator return a reference to itself.
All of this allows the iterator to be used with common STL facilities,
standard algorithms, etc.
Doing this exposed some missing facilities in the iterator facade that
I've fixed and required some work to the actual iterator to fully
support the necessary API.
Differential Revision: https://reviews.llvm.org/D31548
llvm-svn: 300032
Using AssemblyAnnotationWriter for LVI printer prints
for instructions and basic blocks.
So, we explicitly need to print LVI info for the arguments of the function (these
are values and not instructions).
llvm-svn: 298640
Summary:
Adding a printer pass for printing the LVI cache values after transformations
that use LVI.
This will help us in identifying cases where LVI
invariants are violated, or transforms that leave LVI in an incorrect state.
Right now, I have added two test cases to show that the printer pass is working.
I will be adding more test cases in a later change, once this change is
checked in upstream.
Reviewers: reames, dberlin, sanjoy, apilipenko
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D30790
llvm-svn: 298542
Summary:
LVI is now depth first, which is optimal for iteration strategy in
terms of work per call. However, the way the results get cached means
it can still go very badly N^2 or worse right now. The overdefined
cache is per-block, because LVI wants to try to get different results
for the same name in different blocks (IE solve the problem
PredicateInfo solves). This means even if we discover a value is
overdefined after going very deep, it doesn't cache this information,
causing it to end up trying to rediscover it again and again. The
same is true for values along the way. In practice, overdefined
anywhere should mean overdefined everywhere (this is how, for example,
SCCP works).
Until we get around to reworking the overdefined cache, we need to
limit the worklist size we process. Note that permanently reverting
the DFS strategy exploration seems the wrong strategy (temporarily
seems fine if we really want). BFS is clearly the wrong approach, it
just gets luckier on some testcases. It's also very hard to design
an effective throttle for BFS. For DFS, the throttle is directly related
to the depth of the CFG. So really deep CFGs will get cutoff, smaller
ones will not. As the CFG simplifies, you get better results.
In BFS, the limit is it's related to the fan-out times average block size,
which is harder to reason about or make good choices for.
Bug being filed about the overdefined cache, but it will require major
surgery to fix it (plumbing predicateinfo through CVP or LVI).
Note: I did not make this number configurable because i'm not sure
anyone really needs to tweak this knob. We run CVP 3 times. On the
testcases i have the slow ones happen in the middle, where CVP is
doing cleanup work other things are effective at. Over the course of
3 runs, we don't see to have any real loss of performance.
I haven't gotten a minimized testcase yet, but just imagine in your
head a testcase where, going *up* the CFG, you have branches, one of
which leads 50000 blocks deep, and the other, to something where the
answer is overdefined immediately. BFS would discover the overdefined
faster than DFS, but do more work to do so. In practice, the right
answer is "once DFS discovers overdefined for a value, stop trying to
get more info about that value" (and so, DFS would normally cache the
overdefined results for every value it passed through in those 50k
blocks, and never do that work again. But it don't, because of the
naming problem)
Reviewers: chandlerc, djasper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D29715
llvm-svn: 294463
This patch changes the order in which LVI explores previously unexplored paths.
Previously, the code used an BFS strategy where each unexplored input was added to the search queue before any of them were explored. This has the effect of causing all inputs to be explored before returning to re-evaluate the merge point (non-local or phi node). This has the unfortunate property of doing redundant work if one of the inputs to the merge is found to be overdefined (i.e. unanalysable). If any input is overdefined, the result of the merge will be too; regardless of the values of other inputs.
The new code uses a DFS strategy where we re-evaluate the merge after evaluating each input. If we discover an overdefined input, we immediately return without exploring other inputs.
We have reports of large (4-10x) improvements of compile time with this patch and some reports of more precise analysis results as well. See the review discussion for details. The original motivating case was pr10584.
Differential Revision: https://reviews.llvm.org/D28190
llvm-svn: 294264
with it.
This code was dereferencing the PoisoningVH which isn't allowed once it
is poisoned. But the code itself really doesn't need to access the
pointer, it is just doing the safe stuff of clearing out data structures
keyed on the pointer value.
Change the code to use iterators to erase directly from a DenseMap. This
is also substantially more efficient as it avoids lots of hashing and
lookups to do the erasure. DenseMap supports iterating behind the
iteration which is fairly easy to implement.
Sadly, I don't have a test case here. I'm not even close and I don't
know that I ever will be. The issue is that several of the tricky
aspects of fixing this only show up when you cause the stack's
SmallVector to be in *EXACTLY* the right location. I only ever got
a reproduction for those with Clang, and only with *exactly* the right
command line flags. Any adjustment, even to seemingly unrelated flags,
would make partial and half-way solutions magically start to "work". In
good news, all of this was caught with the LLVM test suite. Also, there
is no *specific* code here that is untested, just that the old pattern
of code won't immediately fail on any test case I've managed to
contrive.
llvm-svn: 293160
a lazy-asserting PoisoningVH.
AssertVH is fundamentally incompatible with cache-invalidation of
analysis results. The invaliadtion happens after the AssertingVH has
already fired. Instead, use a PoisoningVH that will assert if the
dangling handle is ever used rather than merely be assigned or
destroyed.
This patch also removes all of the (numerous) doomed attempts to work
around this fundamental incompatibility. It is a pretty significant
simplification IMO.
The most interesting change is in the Inliner where we still do some
clearing because we don't want to rely on the coarse grained
invalidation strategy of the containing pass manager. However, I prefer
the approach that contains this logic to the cleanup phase of the
Inliner, and I think we could enhance the CGSCC analysis management
layer to make this even better in the future if desired.
The rest is straight cleanup.
I've also added a test for one of the harder cases to work around: when
a *module analysis* contains many AssertingVHes pointing at functions.
Differential Revision: https://reviews.llvm.org/D29006
llvm-svn: 292928
become unavailable.
The AssumptionCache is now immutable but it still needs to respond to
DomTree invalidation if it ended up caching one.
This lets us remove one of the explicit invalidates of LVI but the
other one continues to avoid hitting a latent bug.
llvm-svn: 292769
Here's my second try at making @llvm.assume processing more efficient. My
previous attempt, which leveraged operand bundles, r289755, didn't end up
working: it did make assume processing more efficient but eliminating the
assumption cache made ephemeral value computation too expensive. This is a
more-targeted change. We'll keep the assumption cache, but extend it to keep a
map of affected values (i.e. values about which an assumption might provide
some information) to the corresponding assumption intrinsics. This allows
ValueTracking and LVI to find assumptions relevant to the value being queried
without scanning all assumptions in the function. The fact that ValueTracking
started doing O(number of assumptions in the function) work, for every
known-bits query, has become prohibitively expensive in some cases.
As discussed during the review, this is a pragmatic fix that, longer term, will
likely be replaced by a more-principled solution (perhaps based on an extended
SSA form).
Differential Revision: https://reviews.llvm.org/D28459
llvm-svn: 291671
After r289755, the AssumptionCache is no longer needed. Variables affected by
assumptions are now found by using the new operand-bundle-based scheme. This
new scheme is more computationally efficient, and also we need much less
code...
llvm-svn: 289756
There was an efficiency problem with how we processed @llvm.assume in
ValueTracking (and other places). The AssumptionCache tracked all of the
assumptions in a given function. In order to find assumptions relevant to
computing known bits, etc. we searched every assumption in the function. For
ValueTracking, that means that we did O(#assumes * #values) work in InstCombine
and other passes (with a constant factor that can be quite large because we'd
repeat this search at every level of recursion of the analysis).
Several of us discussed this situation at the last developers' meeting, and
this implements the discussed solution: Make the values that an assume might
affect operands of the assume itself. To avoid exposing this detail to
frontends and passes that need not worry about it, I've used the new
operand-bundle feature to add these extra call "operands" in a way that does
not affect the intrinsic's signature. I think this solution is relatively
clean. InstCombine adds these extra operands based on what ValueTracking, LVI,
etc. will need and then those passes need only search the users of the values
under consideration. This should fix the computational-complexity problem.
At this point, no passes depend on the AssumptionCache, and so I'll remove
that as a follow-up change.
Differential Revision: https://reviews.llvm.org/D27259
llvm-svn: 289755
I believe this is the cause of the failure, but have not been able to confirm. Note that this is a speculative fix; I'm still waiting for a full build to finish as I synced and ended up doing a clean build which takes 20+ minutes on my machine.
llvm-svn: 288886
Integers are expressed in the lattice via constant ranges. They can never be represented by constants or not-constants; those are reserved for non-integer types. This code has been dead for literaly years.
llvm-svn: 288767
This completes a small series of patches to hide the stateful updates of LVILatticeVal from the consuming code. The only remaining stateful API is mergeIn.
llvm-svn: 288765
This just extracts out the transfer rules for constant ranges into a single shared point. As it happens, neither bit of code actually overlaps in terms of the handled operators, but with this change that could easily be tweaked in the future.
I also want to have this separated out to make experimenting with a eager value info implementation and possibly a ValueTracking-like fixed depth recursion peephole version. There's no reason all four of these can't share a common implementation which reduces the chances of bugs.
Differential Revision: https://reviews.llvm.org/D27294
llvm-svn: 288413
analyses to have a common type which is enforced rather than using
a char object and a `void *` type when used as an identifier.
This has a number of advantages. First, it at least helps some of the
confusion raised in Justin Lebar's code review of why `void *` was being
used everywhere by having a stronger type that connects to documentation
about this.
However, perhaps more importantly, it addresses a serious issue where
the alignment of these pointer-like identifiers was unknown. This made
it hard to use them in pointer-like data structures. We were already
dodging this in dangerous ways to create the "all analyses" entry. In
a subsequent patch I attempted to use these with TinyPtrVector and
things fell apart in a very bad way.
And it isn't just a compile time or type system issue. Worse than that,
the actual alignment of these pointer-like opaque identifiers wasn't
guaranteed to be a useful alignment as they were just characters.
This change introduces a type to use as the "key" object whose address
forms the opaque identifier. This both forces the objects to have proper
alignment, and provides type checking that we get it right everywhere.
It also makes the types somewhat less mysterious than `void *`.
We could go one step further and introduce a truly opaque pointer-like
type to return from the `ID()` static function rather than returning
`AnalysisKey *`, but that didn't seem to be a clear win so this is just
the initial change to get to a reliably typed and aligned object serving
is a key for all the analyses.
Thanks to Richard Smith and Justin Lebar for helping pick plausible
names and avoid making this refactoring many times. =] And thanks to
Sean for the super fast review!
While here, I've tried to move away from the "PassID" nomenclature
entirely as it wasn't really helping and is overloaded with old pass
manager constructs. Now we have IDs for analyses, and key objects whose
address can be used as IDs. Where possible and clear I've shortened this
to just "ID". In a few places I kept "AnalysisID" to make it clear what
was being identified.
Differential Revision: https://reviews.llvm.org/D27031
llvm-svn: 287783
The patch is to partially fix PR10584. Correlated Value Propagation queries LVI
to check non-null for pointer params of each callsite. If we know the def of
param is an alloca instruction, we know it is non-null and can return early from
LVI. Similarly, CVP queries LVI to check whether pointer for each mem access is
constant. If the def of the pointer is an alloca instruction, we know it is not
a constant pointer. These shortcuts can reduce the cost of CVP significantly.
Differential Revision: https://reviews.llvm.org/D18066
llvm-svn: 281586
Convert the previous introduced is-a relationship between the LVICache and LVIImple clases into a has-a relationship and hide all the implementation details of the cache from the lazy query layer.
The only slightly concerning change here is removing the addition of a queried block into the SeenBlock set in LVIImpl::getBlockValue. As far as I can tell, this was effectively dead code. I think it *used* to be the case that getCachedValueInfo wasn't const and might end up inserting elements in the cache during lookup. That's no longer true and hasn't been for a while. I did fixup the const usage to make that more obvious.
llvm-svn: 281272
Seperate the caching logic from the implementation of the lazy analysis. For the moment, the lazy analysis impl has a is-a relationship with the cache; this will change to a has-a relationship shortly. This was done as two steps merely to keep the changes simple and the diff understandable.
llvm-svn: 281266
Rewrite Visited[Cond] = getValueFromConditionImpl(..., Visited) statement which can lead to a memory corruption since getValueFromConditionImpl changes Visited map and invalidates the iterators.
llvm-svn: 278514
Take range metadata into account for conditions like this:
%length = load i32, i32* %length_ptr, !range !{i32 0, i32 2147483647}
%cmp = icmp ult i32 %a, %length
This is a common pattern for range checks where the length of the array is dynamically loaded.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D23267
llvm-svn: 278496
Currently LVI can only gather value constraints from comparisons like:
* icmp <pred> Val, ...
* icmp ult (add Val, Offset), ...
In fact we can handle any predicate in latter comparisons.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D23357
llvm-svn: 278493
Teach LVI how to gather information from conditions in the form of (cond1 && cond2). Our out-of-tree front-end emits range checks in this form.
Reviewed By: sanjoy
Differential Revision: http://reviews.llvm.org/D23200
llvm-svn: 278231
Instead of returning bool and setting LVILatticeValue reference argument return LVILattice value. Use overdefined value to denote the case when we didn't gather any information from the condition.
This change was separated from the review "[LVI] Handle conditions in the form of (cond1 && cond2)" (https://reviews.llvm.org/D23200#inline-199531). Once getValueFromCondition returns LVILatticeValue we can cache the result in Visited map.
llvm-svn: 278224
The problem was triggered by my recent change in CVP (D23059). Current code expected that integer constants are represented by constantrange LVILatticeVal and never represented as LVILatticeVal with constant tag. That is true for ConstantInt constants, although ConstantExpr integer type constants are legally represented as constant LVILatticeVal.
This code fails with CVP change in:
@b = global i32 0, align 4
define void @test6(i32 %a) {
bb:
%add = add i32 %a, ptrtoint (i32* @b to i32)
ret void
}
Currently getConstantRange code is not executed by any of the upstream passes. I'm going to add a test case to test/Transforms/CorrelatedValuePropagation/add.ll once I resubmit the CVP change.
Reviewed By: sanjoy
Differential Revision: http://reviews.llvm.org/D23194
llvm-svn: 278217
Gathering constantins from a condition on the false path ask makeAllowedICmpRegion about inverse predicate instead of inversing the resulting range.
This change was separated from the review "[LVI] Make LVI smarter about comparisons with non-constants" (https://reviews.llvm.org/D23205#inline-198361)
llvm-svn: 278009
Chandler Carruth