a memcpy to reflect that '0' has a different meaning when applied to
a load or store. Now we correctly use underaligned loads and stores for
the test case added.
llvm-svn: 165101
necessary during rewriting. As part of this, fix a real think-o here
where we might have left off an alignment specification when the address
is in fact underaligned. I haven't come up with any way to trigger this,
as there is always some other factor that reduces the alignment, but it
certainly might have been an observable bug in some way I can't think
of. This also slightly changes the strategy for placing explicit
alignments on loads and stores to only do so when the alignment does not
match that required by the ABI. This causes a few redundant alignments
to go away from test cases.
I've also added a couple of tests that really push on the alignment that
we end up with on loads and stores. More to come here as I try to fix an
underlying bug I have conjectured and produced test cases for, although
it's not clear if this bug is the one currently hitting dragonegg's
gcc47 bootstrap.
llvm-svn: 165100
scheduled for processing on the worklist eventually gets deleted while
we are processing another alloca, fixing the original test case in
PR13990.
To facilitate this, add a remove_if helper to the SetVector abstraction.
It's not easy to use the standard abstractions for this because of the
specifics of SetVectors types and implementation.
Finally, a nice small test case is included. Thanks to Benjamin for the
fantastic reduced test case here! All I had to do was delete some empty
basic blocks!
llvm-svn: 165065
alignment requirements of the new alloca. As one consequence which was
reported as a bug by Duncan, we overaligned memcpy calls to ranges of
allocas after they were rewritten to types with lower alignment
requirements. Other consquences are possible, but I don't have any test
cases for them.
llvm-svn: 164937
a pair of instructions, one for the used pointer and the second for the
user. This simplifies the representation and also makes it more dense.
This was noticed because of the miscompile in PR13926. In that case, we
were running up against a fundamental "bad idea" in the speculation of
PHI and select instructions: the speculation and rewriting are
interleaved, which requires phi speculation to also perform load
rewriting! This is bad, and causes us to miss opportunities to do (for
example) vector rewriting only exposed after PHI speculation, etc etc.
It also, in the old system, required us to insert *new* load uses into
the current partition's use list, which would then be ignored during
rewriting because we had already extracted an end iterator for the use
list. The appending behavior (and much of the other oddities) stem from
the strange de-duplication strategy in the PartitionUse builder.
Amusingly, all this went without notice for so long because it could
only be triggered by having *different* GEPs into the same partition of
the same alloca, where both different GEPs were operands of a single
PHI, and where the GEP which was not encountered first also had multiple
uses within that same PHI node... Hence the insane steps required to
reproduce.
So, step one in fixing this fundamental bad idea is to make the
PartitionUse actually contain a Use*, and to make the builder do proper
deduplication instead of funky de-duplication. This is enough to remove
the appending behavior, and fix the miscompile in PR13926, but there is
more work to be done here. Subsequent commits will lift the speculation
into its own visitor. It'll be a useful step toward potentially
extracting all of the speculation logic into a generic utility
transform.
The existing PHI test case for repeated operands has been made more
extreme to catch even these issues. This test case, run through the old
pass, will exactly reproduce the miscompile from PR13926. ;] We were so
close here!
llvm-svn: 164925
alignment could lose it due to the alloca type moving down to a much
smaller alignment guarantee.
Now SROA will actively compute a proper alignment, factoring the target
data, any explicit alignment, and the offset within the struct. This
will in some cases lower the alignment requirements, but when we lower
them below those of the type, we drop the alignment entirely to give
freedom to the code generator to align it however is convenient.
Thanks to Duncan for the lovely test case that pinned this down. =]
llvm-svn: 164891
rewriter in SROA to carry a proper alignment. This involves
interrogating various sources of alignment, etc. This is a more complete
and principled fix to PR13920 as well as related bugs pointed out by Eli
in review and by inspection in the area.
Also by inspection fix the integer and vector promotion paths to create
aligned loads and stores. I still need to work up test cases for
these... Sorry for the delay, they were found purely by inspection.
llvm-svn: 164689
This should really, really fix PR13916. For real this time. The
underlying bug is... a bit more subtle than I had imagined.
The setup is a code pattern that leads to an @llvm.memcpy call with two
equal pointers to an alloca in the source and dest. Now, not any pattern
will do. The alloca needs to be formed just so, and both pointers should
be wrapped in different bitcasts etc. When this precise pattern hits,
a funny sequence of events transpires. First, we correctly detect the
potential for overlap, and correctly optimize the memcpy. The first
time. However, we do simplify the set of users of the alloca, and that
causes us to run the alloca back through the SROA pass in case there are
knock-on simplifications. At this point, a curious thing has happened.
If we happen to have an i8 alloca, we have direct i8 pointer values. So
we don't bother creating a cast, we rewrite the arguments to the memcpy
to dircetly refer to the alloca.
Now, in an unrelated area of the pass, we have clever logic which
ensures that when visiting each User of a particular pointer derived
from an alloca, we only visit that User once, and directly inspect all
of its operands which refer to that particular pointer value. However,
the mechanism used to detect memcpy's with the potential to overlap
relied upon getting visited once per *Use*, not once per *User*. This is
always true *unless* the same exact value is both source and dest. It
turns out that almost nothing actually produces that pattern though.
We can hand craft test cases that more directly test this behavior of
course, and those are included. Also, note that there is a significant
missed optimization here -- we prove in many cases that there is
a non-volatile memcpy call with identical source and dest addresses. We
shouldn't prevent splitting the alloca in that case, and in fact we
should just remove such memcpy calls eagerly. I'll address that in
a subsequent commit.
llvm-svn: 164669
only a missed optimization opportunity if the store is over-aligned, but a
miscompile if the store's new type has a higher natural alignment than the
memcpy did. Fixes PR13920!
llvm-svn: 164641
Chandler, it's not obvious that it's okay that this alloca gets into the list
twice to begin with. Please review and see whether this is the fix you really
want, but I wanted to get a fix checked in quickly.
llvm-svn: 164634
to chains or cycles between PHIs and/or selects. Also add a couple of
really nice test cases reduced from Kostya's reports in PR13905 and
PR13906. Both are fixed by this patch.
llvm-svn: 164596
integer promotion analogous to vector promotion. When there is an
integer alloca being accessed both as its integer type and as a narrower
integer type, promote the narrower access to "insert" and "extract" the
smaller integer from the larger one, and make the integer alloca
a candidate for promotion.
In the new formulation, we don't care about target legal integer or use
thresholds to control things. Instead, we only perform this promotion to
an integer type which the frontend has already emitted a load or store
for. This bounds the scope and prevents optimization passes from
coalescing larger and larger entities into a single integer.
llvm-svn: 164479
across the uses of the alloca. It's entirely possible for negative
numbers to come up here, and in some rare cases simply doing the 2's
complement arithmetic isn't the correct decision. Notably, we can't zext
the index of the GEP. The definition of GEP is that these offsets are
sign extended or truncated to the size of the pointer, and then wrapping
2's complement arithmetic used.
This patch fixes an issue that comes up with *no* input from the
buildbots or bootstrap afaict. The only place where it manifested,
disturbingly, is Clang's own regression test suite. A reduced and
targeted collection of tests are added to cope with this. Note that I've
tried to pin down the potential cases of overflow, but may have missed
some cases. I've tried to add a few cases to test this, but its hard
because LLVM has quite limited support for >64bit constructs.
llvm-svn: 164475
selects with a constant condition. This resulted in the operands
remaining live through the SROA rewriter. Most of the time, this just
caused some dead allocas to persist and get zapped by later passes, but
in one case found by Joerg, it caused a crash when we tried to *promote*
the alloca despite it having this dead use. We already have the
mechanisms in place to handle this, just wire select up to them.
llvm-svn: 164427
from the dragonegg build bots when we turned on the full version of the
pass. Included a much reduced test case for this pesky bug, despite
bugpoint's uncooperative behavior.
Also, I audited all the similar code I could find and didn't spot any
other cases where this mistake cropped up.
llvm-svn: 164178
working on FCA splitting. Instead of refusing to form a common type when
there are uses of a subsection of the alloca as well as a use of the
entire alloca, just skip the subsection uses and continue looking for
a whole-alloca use with a type that we can use.
This produces slightly prettier IR I think, and also fixes the other
failure in the test.
llvm-svn: 164146
FCAs. This is essential in order to promote allocas that are used in
struct returns by frontends like Clang. The FCA load would block the
rest of the pass from firing, resulting is significant regressions with
the bullet benchmark in the nightly test suite.
Thanks to Duncan for repeated discussions about how best to do this, and
to both him and Benjamin for review.
This appears to have blocked many places where the pass tries to fire,
and so I'm expect somewhat different results with this fix added.
As with the last big patch, I'm including a change to enable the SROA by
default *temporarily*. Ben is going to remove this as soon as the LNT
bots pick up the patch. I'm just trying to get a round of LNT numbers
from the stable machines in the lab.
NOTE: Four clang tests are expected to fail in the brief window where
this is enabled. Sorry for the noise!
llvm-svn: 164119
new one, and add support for running the new pass in that mode and in
that slot of the pass manager. With this the new pass can completely
replace the old one within the pipeline.
The strategy for enabling or disabling the SSAUpdater logic is to do it
by making the requirement of the domtree analysis optional. By default,
it is required and we get the standard mem2reg approach. This is usually
the desired strategy when run in stand-alone situations. Within the
CGSCC pass manager, we disable requiring of the domtree analysis and
consequentially trigger fallback to the SSAUpdater promotion.
In theory this would allow the pass to re-use a domtree if one happened
to be available even when run in a mode that doesn't require it. In
practice, it lets us have a single pass rather than two which was
simpler for me to wrap my head around.
There is a hidden flag to force the use of the SSAUpdater code path for
the purpose of testing. The primary testing strategy is just to run the
existing tests through that path. One notable difference is that it has
custom code to handle lifetime markers, and one of the tests has been
enhanced to exercise that code.
This has survived a bootstrap and the test suite without serious
correctness issues, however my run of the test suite produced *very*
alarming performance numbers. I don't entirely understand or trust them
though, so more investigation is on-going.
To aid my understanding of the performance impact of the new SROA now
that it runs throughout the optimization pipeline, I'm enabling it by
default in this commit, and will disable it again once the LNT bots have
picked up one iteration with it. I want to get those bots (which are
much more stable) to evaluate the impact of the change before I jump to
any conclusions.
NOTE: Several Clang tests will fail because they run -O3 and check the
result's order of output. They'll go back to passing once I disable it
again.
llvm-svn: 163965
This is essentially a ground up re-think of the SROA pass in LLVM. It
was initially inspired by a few problems with the existing pass:
- It is subject to the bane of my existence in optimizations: arbitrary
thresholds.
- It is overly conservative about which constructs can be split and
promoted.
- The vector value replacement aspect is separated from the splitting
logic, missing many opportunities where splitting and vector value
formation can work together.
- The splitting is entirely based around the underlying type of the
alloca, despite this type often having little to do with the reality
of how that memory is used. This is especially prevelant with unions
and base classes where we tail-pack derived members.
- When splitting fails (often due to the thresholds), the vector value
replacement (again because it is separate) can kick in for
preposterous cases where we simply should have split the value. This
results in forming i1024 and i2048 integer "bit vectors" that
tremendously slow down subsequnet IR optimizations (due to large
APInts) and impede the backend's lowering.
The new design takes an approach that fundamentally is not susceptible
to many of these problems. It is the result of a discusison between
myself and Duncan Sands over IRC about how to premptively avoid these
types of problems and how to do SROA in a more principled way. Since
then, it has evolved and grown, but this remains an important aspect: it
fixes real world problems with the SROA process today.
First, the transform of SROA actually has little to do with replacement.
It has more to do with splitting. The goal is to take an aggregate
alloca and form a composition of scalar allocas which can replace it and
will be most suitable to the eventual replacement by scalar SSA values.
The actual replacement is performed by mem2reg (and in the future
SSAUpdater).
The splitting is divided into four phases. The first phase is an
analysis of the uses of the alloca. This phase recursively walks uses,
building up a dense datastructure representing the ranges of the
alloca's memory actually used and checking for uses which inhibit any
aspects of the transform such as the escape of a pointer.
Once we have a mapping of the ranges of the alloca used by individual
operations, we compute a partitioning of the used ranges. Some uses are
inherently splittable (such as memcpy and memset), while scalar uses are
not splittable. The goal is to build a partitioning that has the minimum
number of splits while placing each unsplittable use in its own
partition. Overlapping unsplittable uses belong to the same partition.
This is the target split of the aggregate alloca, and it maximizes the
number of scalar accesses which become accesses to their own alloca and
candidates for promotion.
Third, we re-walk the uses of the alloca and assign each specific memory
access to all the partitions touched so that we have dense use-lists for
each partition.
Finally, we build a new, smaller alloca for each partition and rewrite
each use of that partition to use the new alloca. During this phase the
pass will also work very hard to transform uses of an alloca into a form
suitable for promotion, including forming vector operations, speculating
loads throguh PHI nodes and selects, etc.
After splitting is complete, each newly refined alloca that is
a candidate for promotion to a scalar SSA value is run through mem2reg.
There are lots of reasonably detailed comments in the source code about
the design and algorithms, and I'm going to be trying to improve them in
subsequent commits to ensure this is well documented, as the new pass is
in many ways more complex than the old one.
Some of this is still a WIP, but the current state is reasonbly stable.
It has passed bootstrap, the nightly test suite, and Duncan has run it
successfully through the ACATS and DragonEgg test suites. That said, it
remains behind a default-off flag until the last few pieces are in
place, and full testing can be done.
Specific areas I'm looking at next:
- Improved comments and some code cleanup from reviews.
- SSAUpdater and enabling this pass inside the CGSCC pass manager.
- Some datastructure tuning and compile-time measurements.
- More aggressive FCA splitting and vector formation.
Many thanks to Duncan Sands for the thorough final review, as well as
Benjamin Kramer for lots of review during the process of writing this
pass, and Daniel Berlin for reviewing the data structures and algorithms
and general theory of the pass. Also, several other people on IRC, over
lunch tables, etc for lots of feedback and advice.
llvm-svn: 163883
Chandler Carruth