Summary:
When there is a call to an alias in the same module, we were not
adding a call edge. So we could incorrectly think that the alias
was dead if it was inlined in that function, despite having a
reference imported elsewhere. This resulted in unsats at link time.
Add a call edge when the call is to an alias.
Reviewers: davide, mehdi_amini
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D25384
llvm-svn: 283664
Summary:
While walking defs of pointer operands we were assuming that the pointer
size would remain constant. This is not true, because addresspacecast
instructions may cast the pointer to an address space with a different
pointer width.
This partial reverts r282612, which was a more conservative solution
to this problem.
Reviewers: reames, sanjoy, apilipenko
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D24772
llvm-svn: 283557
With the ROPI and RWPI relocation models we can't always have pointers
to global data or functions in constant data, so don't try to convert switches
into lookup tables if any value in the lookup table would require a relocation.
We can still safely emit lookup tables of other values, such as simple
constants.
Differential Revision: https://reviews.llvm.org/D24462
llvm-svn: 283530
Summary:
The computeKnownBits and ComputeNumSignBits functions in ValueTracking can now do a simple look-through of ExtractElement.
Reviewers: majnemer, spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24955
llvm-svn: 283434
The purpose of the YAML diagnostic output file is to collect information on
optimizations performed, or not performed, for later processing by tools that
help users (and compiler developers) understand how code was optimized. As
such, the diagnostics that appear in the file should not be coupled to what a
user might want to see summarized for them as the compiler runs, and in fact,
because the user likely does not know what optimization diagnostics their tools
might want to use, the user cannot provide a useful filter regardless. As such,
we shouldn't filter the diagnostics going to the output file.
Differential Revision: https://reviews.llvm.org/D25224
llvm-svn: 283236
Slightly improves the precision of GlobalsAA in certain situations, and
makes the behavior of optimization passes more predictable.
Differential Revision: https://reviews.llvm.org/D24104
llvm-svn: 283165
Summary: Added 6 new target hooks for the vectorizer in order to filter types, handle size constraints and decide how to split chains.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm, mzolotukhin, wdng, llvm-commits, nhaehnle
Differential Revision: https://reviews.llvm.org/D24727
llvm-svn: 283099
This was first landed in rL283058 and subsequenlty reverted since a
change this depends on (rL283057) was buggy and had to be reverted.
llvm-svn: 283079
They've broken the sanitizer-bootstrap bots. Reverting while I investigate.
Original commit messages:
r283057: "[ConstantRange] Make getEquivalentICmp smarter"
r283058: "[SCEV] Rely on ConstantRange instead of custom logic; NFCI"
llvm-svn: 283062
(Recommit after making sure IsVerbose gets properly initialized in
DiagnosticInfoOptimizationBase. See previous commit that takes care of
this.)
OptimizationRemarkAnalysis directly takes the role of the report that is
generated by LAA.
Then we need the magic to be able to turn an LAA remark into an LV
remark. This is done via a new OptimizationRemark ctor.
llvm-svn: 282813
OptimizationRemarkAnalysis directly takes the role of the report that is
generated by LAA.
Then we need the magic to be able to turn an LAA remark into an LV
remark. This is done via a new OptimizationRemark ctor.
llvm-svn: 282758
Summary:
When using llc with -compile-twice, module is generated twice, but getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI will still get the old PSI with the original (invalidated) Module. This patch checks if the module has changed when calling getPSI, if yes, update the module and invalidate the Summary.
The bug does not show up in the current llc because PSI is not used in CodeGen yet. But with https://reviews.llvm.org/D24989, the bug will be exposed by test/CodeGen/PowerPC/pr26378.ll
Reviewers: eraman, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24993
llvm-svn: 282616
Pointers in different addrspaces can have different sizes, so it's not valid to look through addrspace cast calculating base and offset for a value.
This is similar to D13008.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D24729
llvm-svn: 282612
Summary:
Instead of creating and destroying SCEVUnionPredicate instances (which
internally creates and destroys a DenseMap), use temporary SmallPtrSet
instances of remember the set of predicates that will get reified into a
SCEVUnionPredicate.
Reviewers: silviu.baranga, sbaranga
Subscribers: sanjoy, mcrosier, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D25000
llvm-svn: 282606
Ever since LAA was split out into an analysis on its own, this function
stopped emitting the report directly. Instead it stores it to be
retrieved by the client which can then emit it as its own report
(e.g. -Rpass-analysis=loop-vectorize).
llvm-svn: 282561
(Re-committed after moving the template specialization under the yaml
namespace. GCC was complaining about this.)
This allows various presentation of this data using an external tool.
This was first recommended here[1].
As an example, consider this module:
1 int foo();
2 int bar();
3
4 int baz() {
5 return foo() + bar();
6 }
The inliner generates these missed-optimization remarks today (the
hotness information is pulled from PGO):
remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30)
remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30)
Now with -pass-remarks-output=<yaml-file>, we generate this YAML file:
--- !Missed
Pass: inline
Name: NotInlined
DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 }
Function: baz
Hotness: 30
Args:
- Callee: foo
- String: will not be inlined into
- Caller: baz
...
--- !Missed
Pass: inline
Name: NotInlined
DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 }
Function: baz
Hotness: 30
Args:
- Callee: bar
- String: will not be inlined into
- Caller: baz
...
This is a summary of the high-level decisions:
* There is a new streaming interface to emit optimization remarks.
E.g. for the inliner remark above:
ORE.emit(DiagnosticInfoOptimizationRemarkMissed(
DEBUG_TYPE, "NotInlined", &I)
<< NV("Callee", Callee) << " will not be inlined into "
<< NV("Caller", CS.getCaller()) << setIsVerbose());
NV stands for named value and allows the YAML client to process a remark
using its name (NotInlined) and the named arguments (Callee and Caller)
without parsing the text of the message.
Subsequent patches will update ORE users to use the new streaming API.
* I am using YAML I/O for writing the YAML file. YAML I/O requires you
to specify reading and writing at once but reading is highly non-trivial
for some of the more complex LLVM types. Since it's not clear that we
(ever) want to use LLVM to parse this YAML file, the code supports and
asserts that we're writing only.
On the other hand, I did experiment that the class hierarchy starting at
DiagnosticInfoOptimizationBase can be mapped back from YAML generated
here (see D24479).
* The YAML stream is stored in the LLVM context.
* In the example, we can probably further specify the IR value used,
i.e. print "Function" rather than "Value".
* As before hotness is computed in the analysis pass instead of
DiganosticInfo. This avoids the layering problem since BFI is in
Analysis while DiagnosticInfo is in IR.
[1] https://reviews.llvm.org/D19678#419445
Differential Revision: https://reviews.llvm.org/D24587
llvm-svn: 282539
I don't expect `PendingLoopPredicates` to have very many
elements (e.g. when -O3'ing the sqlite3 amalgamation,
`PendingLoopPredicates` has at most 3 elements). So now we use a
`SmallPtrSet` for it instead of the more heavyweight `DenseSet`.
llvm-svn: 282511
This allows various presentation of this data using an external tool.
This was first recommended here[1].
As an example, consider this module:
1 int foo();
2 int bar();
3
4 int baz() {
5 return foo() + bar();
6 }
The inliner generates these missed-optimization remarks today (the
hotness information is pulled from PGO):
remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30)
remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30)
Now with -pass-remarks-output=<yaml-file>, we generate this YAML file:
--- !Missed
Pass: inline
Name: NotInlined
DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 }
Function: baz
Hotness: 30
Args:
- Callee: foo
- String: will not be inlined into
- Caller: baz
...
--- !Missed
Pass: inline
Name: NotInlined
DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 }
Function: baz
Hotness: 30
Args:
- Callee: bar
- String: will not be inlined into
- Caller: baz
...
This is a summary of the high-level decisions:
* There is a new streaming interface to emit optimization remarks.
E.g. for the inliner remark above:
ORE.emit(DiagnosticInfoOptimizationRemarkMissed(
DEBUG_TYPE, "NotInlined", &I)
<< NV("Callee", Callee) << " will not be inlined into "
<< NV("Caller", CS.getCaller()) << setIsVerbose());
NV stands for named value and allows the YAML client to process a remark
using its name (NotInlined) and the named arguments (Callee and Caller)
without parsing the text of the message.
Subsequent patches will update ORE users to use the new streaming API.
* I am using YAML I/O for writing the YAML file. YAML I/O requires you
to specify reading and writing at once but reading is highly non-trivial
for some of the more complex LLVM types. Since it's not clear that we
(ever) want to use LLVM to parse this YAML file, the code supports and
asserts that we're writing only.
On the other hand, I did experiment that the class hierarchy starting at
DiagnosticInfoOptimizationBase can be mapped back from YAML generated
here (see D24479).
* The YAML stream is stored in the LLVM context.
* In the example, we can probably further specify the IR value used,
i.e. print "Function" rather than "Value".
* As before hotness is computed in the analysis pass instead of
DiganosticInfo. This avoids the layering problem since BFI is in
Analysis while DiagnosticInfo is in IR.
[1] https://reviews.llvm.org/D19678#419445
Differential Revision: https://reviews.llvm.org/D24587
llvm-svn: 282499
Summary:
This patch improves thinlto importer
by importing 3x larger functions that are called from hot block.
I compared performance with the trunk on spec, and there
were about 2% on povray and 3.33% on milc. These results seems
to be consistant and match the results Teresa got with her simple
heuristic. Some benchmarks got slower but I think they are just
noisy (mcf, xalancbmki, omnetpp)- running the benchmarks again with
more iterations to confirm. Geomean of all benchmarks including the noisy ones
were about +0.02%.
I see much better improvement on google branch with Easwaran patch
for pgo callsite inlining (the inliner actually inline those big functions)
Over all I see +0.5% improvement, and I get +8.65% on povray.
So I guess we will see much bigger change when Easwaran patch will land
(it depends on new pass manager), but it is still worth putting this to trunk
before it.
Implementation details changes:
- Removed CallsiteCount.
- ProfileCount got replaced by Hotness
- hot-import-multiplier is set to 3.0 for now,
didn't have time to tune it up, but I see that we get most of the interesting
functions with 3, so there is no much performance difference with higher, and
binary size doesn't grow as much as with 10.0.
Reviewers: eraman, mehdi_amini, tejohnson
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D24638
llvm-svn: 282437
In a previous change I collapsed two different caches into one. When
doing that I noticed that ScalarEvolution's move constructor was not
moving those caches.
To keep the previous change simple, I've moved that bugfix into this
separate change.
llvm-svn: 282376
Both `loopHasNoSideEffects` and `loopHasNoAbnormalExits` involve walking
the loop and maintaining similar sorts of caches. This commit changes
SCEV to compute both the predicates via a single walk, and maintain a
single cache instead of two.
llvm-svn: 282375
This change simplifies a data structure optimization in the
`BackedgeTakenInfo` class for loops with exactly one computable exit.
I've sanity checked that this does not regress compile time performance,
using sqlite3's amalgamated build.
llvm-svn: 282365
There is no benefit in looking through assumptions on UndefValue to
guess known bits. Return early to avoid walking their use-lists, and
assert that all instances of ConstantData are handled here for similar
reasons (UndefValue was the only integer/pointer holdout).
llvm-svn: 282337
Check and return early for ConstantPointerNull and UndefValue
specifically in isKnownNonNullAt, and assert that ConstantData never
make it to isKnownNonNullFromDominatingCondition.
This confirms that isKnownNonNullFromDominatingCondition never walks
through the use-list of an instance of ConstantData. Given that such
use-lists cross module boundaries, it never really made sense to do so,
and was potentially very expensive.
llvm-svn: 282333
Summary: When identifying cold blocks, consider only the edge to the normal destination if the terminator is InvokeInst and let calcInvokeHeuristics() decide edge weights for the InvokeInst.
Reviewers: mcrosier, hfinkel, davidxl
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D24868
llvm-svn: 282262
computeKnownBits() already works for integer vectors, so allow vector types when calling that from InstCombine.
I don't think the change to use m_APInt in computeKnownBits is strictly necessary because we do check for
ConstantVector later, but it's more efficient to handle the splat case without needing to loop on vector elements.
This should work with InstSimplify, but doesn't yet, so I made that a FIXME comment on the test for PR24942:
https://llvm.org/bugs/show_bug.cgi?id=24942
Differential Revision: https://reviews.llvm.org/D24677
llvm-svn: 281777
Enhance SCEV to compute the trip count for some loops with unknown stride.
Patch by Pankaj Chawla
Differential Revision: https://reviews.llvm.org/D22377
llvm-svn: 281732
LazyCallGraph to support repeated, stable iterations, even in the face
of graph updates.
This is particularly important to allow the CGSCC pass manager to walk
the RefSCCs (and thus everything else) in a module more than once. Lots
of unittests and other tests were hard or impossible to write because
repeated CGSCC pass managers which didn't invalidate the LazyCallGraph
would conclude the module was empty after the first one. =[ Really,
really bad.
The interesting thing is that in many ways this simplifies the code. We
can now re-use the same code for handling reference edge insertion
updates of the RefSCC graph as we use for handling call edge insertion
updates of the SCC graph. Outside of adapting to the shared logic for
this (which isn't trivial, but is *much* simpler than the DFS it
replaces!), the new code involves putting newly created RefSCCs when
deleting a reference edge into the cached list in the correct way, and
to re-formulate the iterator to be stable and effective even in the face
of these kinds of updates.
I've updated the unittests for the LazyCallGraph to re-iterate the
postorder sequence and verify that this all works. We even check for
using alternating iterators to trigger the lazy formation of RefSCCs
after mutation has occured.
It's worth noting that there are a reasonable number of likely
simplifications we can make past this. It isn't clear that we need to
keep the "LeafRefSCCs" around any more. But I've not removed that mostly
because I want this to be a more isolated change.
Differential Revision: https://reviews.llvm.org/D24219
llvm-svn: 281716
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
value is a pointer.
This patch is to fix PR30213. When expanding an expr based on ValueOffsetPair,
if the value is of pointer type, we can only create a getelementptr instead
of sub expr.
Differential Revision: https://reviews.llvm.org/D24088
llvm-svn: 281439
The constant folder didn't know how to always fold bitcasts of constant integer
vectors. In particular, it was unable to handle the case where a constant vector
had some undef elements, and the resulting (i.e. bitcasted) vector type had more
elements than the original vector type.
Example:
%cast = bitcast <2 x i64><i64 undef, i64 2> to <4 x i32>
On a little endian target, %cast could have been folded to:
<4 x i32><i32 undef, i32 undef, i32 2, i32 0>
This patch improves the folding logic by teaching how to correctly propagate
undef elements in the folded vector.
Differential Revision: https://reviews.llvm.org/D24301
llvm-svn: 281343
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
Summary:
This will let e.g. the load/store vectorizer propagate this metadata
appropriately.
Reviewers: arsenm
Subscribers: tra, jholewinski, hfinkel, mzolotukhin
Differential Revision: https://reviews.llvm.org/D23479
llvm-svn: 281153
make_scope_exit now that we have that utility.
This makes the code much more clear and readable by isolating the check.
It also makes it easy to go through and make sure all the interesting
update routines have a start and end verify so we don't slowly let the
graph drift into an invalid state.
llvm-svn: 280619
a postorder-sequence based update after edge insertion into a generic
helper function.
This separates the SCC-specific logic into two fairly simple lambdas and
extracts the rest into a generic helper template function. I think this
is a net win on its own merits because it disentangles different pieces
of the algorithm. Now there is one place that does the two-step
partition to identify a set of newly connected components and at the
same time update the postorder sequence.
However, I'm also hoping to re-use this an upcoming patch to update
a cached post-order sequence of RefSCCs when doing the analogous update
to the RefSCC graph, and I don't want to have two copies.
The diff is quite messy but this really is just moving things around and
making types generic rather than specific.
llvm-svn: 280618
We don't need to call `GetCompareTy(LHS)' every single time true or false is
returned from function SimplifyFCmpInst as suggested by Sanjay in review D24142.
llvm-svn: 280491
This patch fixes a crash caused by an incorrect folding of an ordered comparison
between a packed floating point vector and a splat vector of NaN.
An ordered comparison between a vector and a constant vector of NaN, should
always be folded into a constant vector where each element is i1 false.
Since revision 266175, SimplifyFCmpInst folds the ordered fcmp into a scalar
'false'. Later on, this would cause an assertion failure, since the value type
of the folded value doesn't match the expected value type of the uses of the
original instruction: "Assertion failed: New->getType() == getType() &&
"replaceAllUses of value with new value of different type!".
This patch fixes the issue and adds a test case to the already existing test
InstSimplify/floating-point-compares.ll.
Differential Revision: https://reviews.llvm.org/D24143
llvm-svn: 280488
Summary:
Current implementation of LI verifier isn't ideal and fails to detect
some cases when LI is incorrect. For instance, it checks that all
recorded loops are in a correct form, but it has no way to check if
there are no more other (unrecorded in LI) loops in the function. This
patch adds a way to detect such bugs.
Reviewers: chandlerc, sanjoy, hfinkel
Subscribers: llvm-commits, silvas, mzolotukhin
Differential Revision: https://reviews.llvm.org/D23437
llvm-svn: 280280
There were paths where we wouldn't populate the visited set, causing us
to recurse forever if an SSA variable was defined in terms of itself.
This fixes PR30210.
llvm-svn: 280191
Or they were not instantiated as expected;
llvm::InnerAnalysisManagerProxy<llvm::AnalysisManager<llvm::Function>, llvm::LazyCallGraph::SCC>::PassID
llvm::InnerAnalysisManagerProxy<llvm::AnalysisManager<llvm::Function>, llvm::LazyCallGraph::SCC>::PassID
llvm-svn: 280105
Summary:
Changed this code because it was not very readable.
The one question that I got after changing it is, should we
count calls to intrinsics? We don't add them to caller summary,
so maybe we shouldn't also count them?
Reviewers: tejohnson, eraman, mehdi_amini
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23949
llvm-svn: 280036
Fixed a bug in run-time checks for possible memory conflicts inside loop.
The bug is in Low <-> High boundaries calculation. The High boundary should be calculated as "last memory access pointer + element size".
Differential revision: https://reviews.llvm.org/D23176
llvm-svn: 279930
Summary:
This is obviously an interesting case because it may motivate code
restructuring or LTO.
Reporting this requires instantiation of ORE in the loop where the call
sites are first gathered. I've checked compile-time
overhead *with* -Rpass-with-hotness and the worst slow-down was 6% in
mcf and quickly tailing off. As before without -Rpass-with-hotness
there is no overhead.
Because this could be a pretty noisy diagnostics, it is currently
qualified as 'verbose'. As of this patch, 'verbose' diagnostics are
only emitted with -Rpass-with-hotness, i.e. when the output is expected
to be filtered.
Reviewers: eraman, chandlerc, davidxl, hfinkel
Subscribers: tejohnson, Prazek, davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D23415
llvm-svn: 279860
Summary: Dead store elimination gets very expensive when large numbers of instructions need to be analyzed. This patch limits the number of instructions analyzed per store to the value of the memdep-block-scan-limit parameter (which defaults to 100). This resulted in no observed difference in performance of the generated code, and no change in the statistics for the dead store elimination pass, but improved compilation time on some files by more than an order of magnitude.
Reviewers: dexonsmith, bruno, george.burgess.iv, dberlin, reames, davidxl
Subscribers: davide, chandlerc, dberlin, davidxl, eraman, tejohnson, mbodart, llvm-commits
Differential Revision: https://reviews.llvm.org/D15537
llvm-svn: 279833
This patch changes LLVM_CONSTEXPR variable declarations to const
variable declarations, since LLVM_CONSTEXPR expands to nothing if the
current compiler doesn't support constexpr. In all of the changed
cases, it looks like the code intended the variable to be const instead
of sometimes-constexpr sometimes-not.
llvm-svn: 279696
manager, including both plumbing and logic to handle function pass
updates.
There are three fundamentally tied changes here:
1) Plumbing *some* mechanism for updating the CGSCC pass manager as the
CG changes while passes are running.
2) Changing the CGSCC pass manager infrastructure to have support for
the underlying graph to mutate mid-pass run.
3) Actually updating the CG after function passes run.
I can separate them if necessary, but I think its really useful to have
them together as the needs of #3 drove #2, and that in turn drove #1.
The plumbing technique is to extend the "run" method signature with
extra arguments. We provide the call graph that intrinsically is
available as it is the basis of the pass manager's IR units, and an
output parameter that records the results of updating the call graph
during an SCC passes's run. Note that "...UpdateResult" isn't a *great*
name here... suggestions very welcome.
I tried a pretty frustrating number of different data structures and such
for the innards of the update result. Every other one failed for one
reason or another. Sometimes I just couldn't keep the layers of
complexity right in my head. The thing that really worked was to just
directly provide access to the underlying structures used to walk the
call graph so that their updates could be informed by the *particular*
nature of the change to the graph.
The technique for how to make the pass management infrastructure cope
with mutating graphs was also something that took a really, really large
number of iterations to get to a place where I was happy. Here are some
of the considerations that drove the design:
- We operate at three levels within the infrastructure: RefSCC, SCC, and
Node. In each case, we are working bottom up and so we want to
continue to iterate on the "lowest" node as the graph changes. Look at
how we iterate over nodes in an SCC running function passes as those
function passes mutate the CG. We continue to iterate on the "lowest"
SCC, which is the one that continues to contain the function just
processed.
- The call graph structure re-uses SCCs (and RefSCCs) during mutation
events for the *highest* entry in the resulting new subgraph, not the
lowest. This means that it is necessary to continually update the
current SCC or RefSCC as it shifts. This is really surprising and
subtle, and took a long time for me to work out. I actually tried
changing the call graph to provide the opposite behavior, and it
breaks *EVERYTHING*. The graph update algorithms are really deeply
tied to this particualr pattern.
- When SCCs or RefSCCs are split apart and refined and we continually
re-pin our processing to the bottom one in the subgraph, we need to
enqueue the newly formed SCCs and RefSCCs for subsequent processing.
Queuing them presents a few challenges:
1) SCCs and RefSCCs use wildly different iteration strategies at
a high level. We end up needing to converge them on worklist
approaches that can be extended in order to be able to handle the
mutations.
2) The order of the enqueuing need to remain bottom-up post-order so
that we don't get surprising order of visitation for things like
the inliner.
3) We need the worklists to have set semantics so we don't duplicate
things endlessly. We don't need a *persistent* set though because
we always keep processing the bottom node!!!! This is super, super
surprising to me and took a long time to convince myself this is
correct, but I'm pretty sure it is... Once we sink down to the
bottom node, we can't re-split out the same node in any way, and
the postorder of the current queue is fixed and unchanging.
4) We need to make sure that the "current" SCC or RefSCC actually gets
enqueued here such that we re-visit it because we continue
processing a *new*, *bottom* SCC/RefSCC.
- We also need the ability to *skip* SCCs and RefSCCs that get merged
into a larger component. We even need the ability to skip *nodes* from
an SCC that are no longer part of that SCC.
This led to the design you see in the patch which uses SetVector-based
worklists. The RefSCC worklist is always empty until an update occurs
and is just used to handle those RefSCCs created by updates as the
others don't even exist yet and are formed on-demand during the
bottom-up walk. The SCC worklist is pre-populated from the RefSCC, and
we push new SCCs onto it and blacklist existing SCCs on it to get the
desired processing.
We then *directly* update these when updating the call graph as I was
never able to find a satisfactory abstraction around the update
strategy.
Finally, we need to compute the updates for function passes. This is
mostly used as an initial customer of all the update mechanisms to drive
their design to at least cover some real set of use cases. There are
a bunch of interesting things that came out of doing this:
- It is really nice to do this a function at a time because that
function is likely hot in the cache. This means we want even the
function pass adaptor to support online updates to the call graph!
- To update the call graph after arbitrary function pass mutations is
quite hard. We have to build a fairly comprehensive set of
data structures and then process them. Fortunately, some of this code
is related to the code for building the cal graph in the first place.
Unfortunately, very little of it makes any sense to share because the
nature of what we're doing is so very different. I've factored out the
one part that made sense at least.
- We need to transfer these updates into the various structures for the
CGSCC pass manager. Once those were more sanely worked out, this
became relatively easier. But some of those needs necessitated changes
to the LazyCallGraph interface to make it significantly easier to
extract the changed SCCs from an update operation.
- We also need to update the CGSCC analysis manager as the shape of the
graph changes. When an SCC is merged away we need to clear analyses
associated with it from the analysis manager which we didn't have
support for in the analysis manager infrsatructure. New SCCs are easy!
But then we have the case that the original SCC has its shape changed
but remains in the call graph. There we need to *invalidate* the
analyses associated with it.
- We also need to invalidate analyses after we *finish* processing an
SCC. But the analyses we need to invalidate here are *only those for
the newly updated SCC*!!! Because we only continue processing the
bottom SCC, if we split SCCs apart the original one gets invalidated
once when its shape changes and is not processed farther so its
analyses will be correct. It is the bottom SCC which continues being
processed and needs to have the "normal" invalidation done based on
the preserved analyses set.
All of this is mostly background and context for the changes here.
Many thanks to all the reviewers who helped here. Especially Sanjoy who
caught several interesting bugs in the graph algorithms, David, Sean,
and others who all helped with feedback.
Differential Revision: http://reviews.llvm.org/D21464
llvm-svn: 279618
And add a FIXME because the helper excludes folds for vectors. It's
not clear yet how many of these are actually testable (and therefore
necessary?) because later analysis uses computeKnownBits and other
methods to catch many of these cases.
llvm-svn: 279492
This change cause performance regression on MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt from LNT and some other bechmarks.
See https://reviews.llvm.org/D18777 for details.
llvm-svn: 279433
Teresa Johnson