Summary:
...loop after the last iteration.
This is really hard to do correctly. The core problem is that we need to
model liveness through the induction PHIs from iteration to iteration in
order to get the correct results, and we need to correctly de-duplicate
the common subgraphs of instructions feeding some subset of the
induction PHIs. All of this can be driven either from a side effect at
some iteration or from the loop values used after the loop finishes.
This patch implements this by storing the forward-propagating analysis
of each instruction in a cache to recall whether it was free and whether
it has become live and thus counted toward the total unroll cost. Then,
at each sink for a value in the loop, we recursively walk back through
every value that feeds the sink, including looping back through the
iterations as needed, until we have marked the entire input graph as
live. Because we cache this, we never visit instructions more than twice
-- once when we analyze them and put them into the cache, and once when
we count their cost towards the unrolled loop. Also, because the cache
is only two bits and because we are dealing with relatively small
iteration counts, we can store all of this very densely in memory to
avoid this from becoming an excessively slow analysis.
The code here is still pretty gross. I would appreciate suggestions
about better ways to factor or split this up, I've stared too long at
the algorithmic side to really have a good sense of what the design
should probably look at.
Also, it might seem like we should do all of this bottom-up, but I think
that is a red herring. Specifically, the simplification power is *much*
greater working top-down. We can forward propagate very effectively,
even across strange and interesting recurrances around the backedge.
Because we use data to propagate, this doesn't cause a state space
explosion. Doing this level of constant folding, etc, would be very
expensive to do bottom-up because it wouldn't be until the last moment
that you could collapse everything. The current solution is essentially
a top-down simplification with a bottom-up cost accounting which seems
to get the best of both worlds. It makes the simplification incremental
and powerful while leaving everything dead until we *know* it is needed.
Finally, a core property of this approach is its *monotonicity*. At all
times, the current UnrolledCost is a conservatively low estimate. This
ensures that we will never early-exit from the analysis due to exceeding
a threshold when if we had continued, the cost would have gone back
below the threshold. These kinds of bugs can cause incredibly hard to
track down random changes to behavior.
We could use a techinque similar (but much simpler) within the inliner
as well to avoid considering speculated code in the inline cost.
Reviewers: chandlerc
Subscribers: sanjoy, mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D11758
llvm-svn: 269388
Summary:
Currently we consider such instructions as simplified, which is incorrect,
because if their user isn't simplified, we can't actually simplify them too.
This biases our estimates of profitability: for instance the analyzer expects
much more gains from unrolling memcpy loops than there actually are.
Reviewers: hfinkel, chandlerc
Subscribers: mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D17365
llvm-svn: 269387
Ported DA to the new PM by splitting the former DependenceAnalysis Pass
into a DependenceInfo result type and DependenceAnalysisWrapperPass type
and adding a new PM-style DependenceAnalysis analysis pass returning the
DependenceInfo.
Patch by Philip Pfaffe, most of the review by Justin.
Differential Revision: http://reviews.llvm.org/D18834
llvm-svn: 269370
SCEVExpander::replaceCongruentIVs assumes the backedge value of an
SCEV-analysable PHI to always be an instruction, when this is not
necessarily true. For now address this by bailing out of the
optimization if the backedge value of the PHI is a non-Instruction.
llvm-svn: 269213
`SCEVExpander::replaceCongruentIVs` bypasses `hoistIVInc` if both the
original and the isomorphic increments are PHI nodes. Doing this can
break SSA if the isomorphic increment is not dominated by the original
increment. Get rid of the bypass, and let `hoistIVInc` do the right
thing.
Fixes PR27232 (compile time crash/hang).
llvm-svn: 269212
... for AddRec's in loops for which SCEV is unable to compute a max
tripcount. This is not a problem for "normal" loops[0] that don't have
guards or assumes, but helps in cases where we have guards or assumes in
the loop that can be used to constrain incoming values over the backedge.
This partially fixes PR27691 (we still don't handle the NUW case).
[0]: for "normal" loops, in the cases where we'd be able to prove
no-wrap via isKnownPredicate, we'd also be able to compute a max
tripcount.
llvm-svn: 269211
Equivalent GEP indices with different types are treated as different
indices altogether, leading to an incorrect AA result. Fix the issue
by comparing indices based on their values.
Thanks to Mikael Holmén for reporting the issue!
Differential Revision: http://reviews.llvm.org/D19935
llvm-svn: 269197
Extract a part of isDereferenceableAndAlignedPointer functionality to Value:
Reviewed By: hfinkel, sanjoy
Differential Revision: http://reviews.llvm.org/D17611
llvm-svn: 269190
Do simplifications common to all shift instructions based on the amount shifted:
1. If the shift amount is known larger than the bitwidth, the result is undefined.
2. If the valid bits of the shift amount are all known to be 0, it's a shift by zero, so the shift operand is the result.
Note that we could generalize the shift-by-zero transform into a shift-by-constant if all of the valid bits in the shift
amount are known, but that would have to be done in InstCombine rather than here because it would mean we need to create
a new shift instruction.
Differential Revision: http://reviews.llvm.org/D19874
llvm-svn: 269114
The plan is to eventually make this logic simpler, however I expect it to
be a little tricky for the foreseeable future (at least until we're rid of
pointee types), so move it here so that it can be reused to build a summary
index for devirtualization.
Differential Revision: http://reviews.llvm.org/D20005
llvm-svn: 269081
This removes a redundant stride versioning step (we already
do it in getPtrStride, so it has no effect) and uses PSE to
get the SCEV expressions for the source and destination
(this might have changed when getPtrStride was called).
I discovered this through code inspection, and couldn't
produce a regression test for it.
llvm-svn: 269052
Summary:
The idea is very close to what we do for assume intrinsics: we mark the
guard intrinsics as writing to arbitrary memory to maintain control
dependence, but under the covers we teach AA that they do not mod any
particular memory location.
Reviewers: chandlerc, hfinkel, gbiv, reames
Subscribers: george.burgess.iv, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19575
llvm-svn: 269007
We can use calls to @llvm.experimental.guard to prove predicates,
relying on the fact that in all locations domianted by a call to
@llvm.experimental.guard the predicate it is guarding is known to be
true.
llvm-svn: 268997
When we encounter unsafe memory dependencies, loop distribution could
help.
Even though, the diagnostics is in LAA, it's only currently emitted in
the vectorizer.
llvm-svn: 268987
When deciding if a vector calculation can be done in a smaller bitwidth, use sign bit information from ValueTracking to add more information and allow more truncations.
llvm-svn: 268921
A number of libcalls don't exist in any particular lib but are, instead,
defined in math.h as inline functions (even in C mode!). Don't rely on
their existence when lowering @llvm.{cos,sin,floor,..}.f32, promote them
instead.
N.B. We had logic to handle FREM but were missing out on a number of
others. This change generalizes the FREM handling.
llvm-svn: 268875
This test was crashing, and currently it breaks bootstrapping clang with debuginfo
Differential Revision: http://reviews.llvm.org/D20008
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 268715
This message used to be correct, when all we cared about was whether the
dependence was safe (i.e. NoDep) or unsafe. With the current more
precise characterization, this is a forward dep.
llvm-svn: 268695
We assumed that ConstantVectors would be rather uninteresting from the
perspective of analysis. However, this is not the case due to a quirk
of how LLVM handles vectors of i1. Vectors of i1 are not
ConstantDataVectors like vectors of i8, i16, i32 or i64 because i1's
SizeInBits differs from it's StoreSizeInBytes. This leads to it being
categorized as a ConstantVector instead of a ConstantDataVector.
Instead, treat ConstantVector more uniformly.
This fixes PR27591.
llvm-svn: 268479
A loop pass that didn't preserve this entire set of passes wouldn't
play well with other loop passes, since these are generally a basic
requirement to do any interesting transformations to a loop.
Adds a helper to get the set of analyses a loop pass should preserve,
and checks that any loop pass we run satisfies the requirement.
llvm-svn: 268444
In the "LoopDispositions:" section:
- Instead of printing out a list, print out a "dictionary" to make it
obvious by inspection which disposition is for which loop. This is
just a cosmetic change.
- Print dispositions for parent _and_ sibling loops. I will use this
to write a test case.
llvm-svn: 268405
Summary
When a non-escaping pointer is compared to a global value, the
comparison can be folded even if the corresponding malloc/allocation
call cannot be elided.
We need to make sure the global value is not null, since comparisons to
null cannot be folded.
In future, we should also handle cases when the the comparison
instruction dominates the pointer escape.
Reviewers: sanjoy
Subscribers s.egerton, llvm-commits
Differential Revision: http://reviews.llvm.org/D19549
llvm-svn: 268390
We were overly cautious in our analysis of loops which have invokes
which unwind to EH pads. The loop unroll transform is safe because it
only clones blocks in the loop body, it does not try to split critical
edges involving EH pads. Instead, move the necessary safety check to
LoopUnswitch.
N.B. The safety check for loop unswitch is covered by an existing test
which fails without it.
llvm-svn: 268357
that it computes. Currently this is used for testing and precision
tuning, but it might be used by optimizations later.
Differential Revision: http://reviews.llvm.org/D19179
llvm-svn: 268291
As shown in the diff, we used to add to CFLAA's cache by doing
`Cache[Fn] = buildSetsFrom(Fn)`. `buildSetsFrom(Fn)` may cause `Cache`
to reallocate its underlying storage, if this happens and `Cache[Fn]`
was evaluated prior to `buildSetsFrom(Fn)`, then we'll store the result
to a bad address.
Patch by Jia Chen.
llvm-svn: 268269
There are currently some bugs in tree around SCEV caching an incorrect
loop disposition. Printing out loop dispositions will let us write
whitebox tests as those are fixed.
The dispositions are printed as a list in "inside out" order,
i.e. innermost loop first.
llvm-svn: 268177
matchSelectPattern attempts to see through casts which mask min/max
patterns from being more obvious. Under certain circumstances, it would
misidentify a sequence of instructions as a min/max because it assumed
that folding casts would preserve the result. This is not the case for
floating point <-> integer casts.
This fixes PR27575.
llvm-svn: 268086
