After the previous series of patches, this is now trivial and deletes
a pretty astonishing amount of complexity. This has been a long time
coming, as the move toward a PO sequence of RefSCCs started eroding the
underlying use cases for this half of the data structure.
Among the biggest advantages here is that now there aren't two
independent data structures that need to stay in sync.
Some of my profiling has also indicated that updating the parent sets
was among the most expensive parts of the lazy call graph. Eliminating
it whole sale is likely to be a nice win in terms of compile time.
Last but not least, I had discussed with some folks previously keeping
it around for asserts and other correctness checking, but once the
fundamentals of the parent and child checking were implemented without
the parent sets their value in correctness checking was tiny and no
where near worth the cost of the complexity required to keep everything
up-to-date.
llvm-svn: 310171
isDescendantOf methods on RefSCCs in terms of the forward edges rather
than the parent sets.
This is technically slower, but probably not interestingly slower, and
all of these routines were already so expensive that they're guarded
behind both !NDEBUG and EXPENSIVE_CHECKS.
This removes another non-critical usage of parent sets.
I've also added some comments to try and help clarify to any potential
users the costs of these routines. They're mostly useful for debugging,
asserts, or other queries.
llvm-svn: 310170
walk over the parent set.
When removing a single function from the call graph, we previously would
walk the entire RefSCC's parent set and then walk every outgoing edge
just to find the ones to remove. In addition to this being quite high
complexity in theory, it is also the last fundamental use of the parent
sets.
With this change, when we remove a function we transform the node
containing it to be recognizably "dead" and then teach the edge
iterators to recognize edges to such nodes and skip them the same way
they skip null edges.
We can't move fully to using "dead" nodes -- when disconnecting two live
nodes we need to null out the edge. But the complexity this adds to the
edge sequence isn't too bad and the simplification of lazily handling
this seems like a significant win.
llvm-svn: 310169
The definition of 'false' here was already pretty vague and debatable,
and I'm about to add another potential 'false' that would actually make
much more sense in a bool operator. Especially given how rarely this is
used, a nicely named method seems better.
llvm-svn: 310165
structures, actually null out the graph pointers as well. We won't ever
update these, and we certainly shouldn't be calling any methods on them,
so it seems good to defensively nuke them.
llvm-svn: 310164
pointers in node objects, just walk the map from function to node.
It doesn't have stable ordering, but works just as well and is much
simpler. We don't need ordering when just updating internal pointers.
llvm-svn: 310163
merging RefSCCs.
The logic to directly use the reference edges is simpler and not
substantially slower (despite the comments to the contrary) because this
is not actually an especially hot part of LCG in practice.
llvm-svn: 310161
Pushes the sext onto the operands of a Sub if NSW is present.
Also adds support for propagating the nowrap flags of the
llvm.ssub.with.overflow intrinsic during analysis.
Differential Revision: https://reviews.llvm.org/D35256
llvm-svn: 310117
Summary: We originally set the hotness threshold as 99.9% to be consistent with gcc FDO. But because the inline heuristic is different between 2 compilers: llvm uses bottom-up algorithm while gcc uses priority based. The LLVM algorithm tends to inline too much early that prevents hot callsites from further inlined into its caller. Due to this restriction, we think it is reasonable to lower the hotness threshold to give priority to those that are really hot. Our experiments show that this change would improve performance on large applications. Note that the inline heuristic has great room for further tuning. Once the inline heuristics are refined, we could adjust this threshold to allow inlining for less hot callsites.
Reviewers: davidxl, tejohnson, eraman
Reviewed By: tejohnson
Subscribers: sanjoy, llvm-commits
Differential Revision: https://reviews.llvm.org/D36317
llvm-svn: 310065
Adds function attributes to index: ReadNone, ReadOnly, NoRecurse, NoAlias. This attributes will be used for future ThinLTO optimizations that will propagate function attributes across modules.
llvm-svn: 310061
Summary:
This commit allows matchSelectPattern to recognize clamp of float
arguments in the presence of FMF the same way as already done for
integers.
This case is a little different though. With integers, given the
min/max pattern is recognized, DAGBuilder starts selecting MIN/MAX
"automatically". That is not the case for float, because for them only
full FMINNAN/FMINNUM/FMAXNAN/FMAXNUM ISD nodes exist and they do care
about NaNs. On the other hand, some backends (e.g. X86) have only
FMIN/FMAX nodes that do not care about NaNS and the former NAN/NUM
nodes are illegal thus selection is not happening. So I decided to do
such kind of transformation in IR (InstCombiner) instead of
complicating the logic in the backend.
Reviewers: spatel, jmolloy, majnemer, efriedma, craig.topper
Reviewed By: efriedma
Subscribers: hiraditya, javed.absar, n.bozhenov, llvm-commits
Patch by Andrei Elovikov <andrei.elovikov@intel.com>
Differential Revision: https://reviews.llvm.org/D33186
llvm-svn: 310054
Summary:
Detect when the working set size of a profiled application is huge,
by comparing the number of counts required to reach the hot percentile
in the profile summary to a large threshold*.
When the working set size is determined to be huge, disable peeling
to avoid bloating the working set further.
*Note that the selected threshold (15K) is significantly larger than the
largest working set value in SPEC cpu2006 (which is gcc at around 11K).
Reviewers: davidxl
Subscribers: mehdi_amini, mzolotukhin, eraman, llvm-commits
Differential Revision: https://reviews.llvm.org/D36288
llvm-svn: 310005
Summary:
This increases the inlining threshold for hot callsites. Hotness is
defined in terms of block frequency of the callsite relative to the
caller's entry block's frequency. Since this requires BFI in the
inliner, this only affects the new PM pipeline. This is enabled by
default at -O3.
This improves the performance of some internal benchmarks. Notably, an
internal benchmark for Gipfeli compression
(https://github.com/google/gipfeli) improves by ~7%. Povray in SPEC2006
improves by ~2.5%. I am running more experiments and will update the
thread if other benchmarks show improvement/regression.
In terms of text size, LLVM test-suite shows an 1.22% text size
increase. Diving into the results, 13 of the benchmarks in the
test-suite increases by > 10%. Most of these are small, but
Adobe-C++/loop_unroll (17.6% increases) and tramp3d(20.7% size increase)
have >250K text size. On a large application, the text size increases by
2%
Reviewers: chandlerc, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36199
llvm-svn: 309994
Summary:
(This is a second attempt as https://reviews.llvm.org/D34822 was reverted.)
LazyValueInfo currently computes the constant value of the switch condition through case edges, which allows the constant value to be propagated through the case edges.
But we have seen a case where a zero-extended value of the switch condition is used past case edges for which the constant propagation doesn't occur.
This patch adds a small logic to handle such a case in getEdgeValueLocal().
This is motivated by the Python 2.7 eval loop in PyEval_EvalFrameEx() where the lack of the constant propagation causes longer live ranges and more spill code than necessary.
With this patch, we see that the code size of PyEval_EvalFrameEx() decreases by ~5.4% and a performance test improves by ~4.6%.
Reviewers: sanjoy
Reviewed By: sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36247
llvm-svn: 309986
Summary: For SamplePGO, we already record the callsite count in the call instruction itself. So we do not want to use BFI to get profile count as it is less accurate.
Reviewers: tejohnson, davidxl, eraman
Reviewed By: eraman
Subscribers: sanjoy, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D36025
llvm-svn: 309964
The patch rL309080 was reverted because it did not clean up the cache on "forgetValue"
method call. This patch re-enables this change, adds the missing check and introduces
two new unit tests that make sure that the cache is cleaned properly.
Differential Revision: https://reviews.llvm.org/D36087
llvm-svn: 309925
This patch is update after the first patch (https://reviews.llvm.org/rL309651) based on the post-commit comments.
Stack coloring pass need to maintain AliasAnalysis information when merging stack slots of different types.
Actually, there is a FIXME comment in StackColoring.cpp
// FIXME: In order to enable the use of TBAA when using AA in CodeGen,
// we'll also need to update the TBAA nodes in MMOs with values
// derived from the merged allocas.
But, TBAA has been already enabled in CodeGen without fixing this pass.
The incorrect TBAA metadata results in recent failures in bootstrap test on ppc64le (PR33928) by allowing unsafe instruction scheduling.
Although we observed the problem on ppc64le, this is a platform neutral issue.
This patch makes the stack coloring pass maintains AliasAnalysis information when merging multiple stack slots.
This patch fixes PR33928.
llvm-svn: 309849
If SCEV can prove that the backedge taken count for a loop is zero, it does not
need to "understand" a recursive PHI to compute its exiting value.
This should fix PR33885.
llvm-svn: 309758
This causes assertion failures in (a somewhat old version of) SpiderMonkey.
I have already forwarded reproduction instructions to the patch author.
llvm-svn: 309659
Stack coloring pass need to maintain AliasAnalysis information when merging stack slots of different types.
Actually, there is a FIXME comment in StackColoring.cpp
// FIXME: In order to enable the use of TBAA when using AA in CodeGen,
// we'll also need to update the TBAA nodes in MMOs with values
// derived from the merged allocas.
But, TBAA has been already enabled in CodeGen without fixing this pass.
The incorrect TBAA metadata results in recent failures in bootstrap test on ppc64le (PR33928) by allowing unsafe instruction scheduling.
Although we observed the problem on ppc64le, this is a platform neutral issue.
This patch makes the stack coloring pass maintains AliasAnalysis information when merging multiple stack slots.
llvm-svn: 309651
Summary:
Adding part of the changes in D30369 (needed to make progress):
Current patch updates AliasAnalysis and MemoryLocation, but does _not_ clean up MemorySSA.
Original summary from D30369, by dberlin:
Currently, we have instructions which affect memory but have no memory
location. If you call, for example, MemoryLocation::get on a fence,
it asserts. This means things specifically have to avoid that. It
also means we end up with a copy of each API, one taking a memory
location, one not.
This starts to fix that.
We add MemoryLocation::getOrNone as a new call, and reimplement the
old asserting version in terms of it.
We make MemoryLocation optional in the (Instruction, MemoryLocation)
version of getModRefInfo, and kill the old one argument version in
favor of passing None (it had one caller). Now both can handle fences
because you can just use MemoryLocation::getOrNone on an instruction
and it will return a correct answer.
We use all this to clean up part of MemorySSA that had to handle this difference.
Note that literally every actual getModRefInfo interface we have could be made private and replaced with:
getModRefInfo(Instruction, Optional<MemoryLocation>)
and
getModRefInfo(Instruction, Optional<MemoryLocation>, Instruction, Optional<MemoryLocation>)
and delegating to the right ones, if we wanted to.
I have not attempted to do this yet.
Reviewers: dberlin, davide, dblaikie
Subscribers: sanjoy, hfinkel, chandlerc, llvm-commits
Differential Revision: https://reviews.llvm.org/D35441
llvm-svn: 309641
Summary:
Inlining threshold is increased by application of bonuses when the
callee has a single reachable basic block or is rich in vector
instructions. Similarly, inlining cost is reduced by applying a large
bonus when the last call to a static function is considered for
inlining. This patch disables the application of these bonuses when the
callsite or the callee is cold. The intention here is to prevent a large
cold callsite from being inlined to a non-cold caller that could prevent
the caller from being inlined. This is especially important when the
cold callsite is a last call to a static since the associated bonus is
very high.
Reviewers: chandlerc, davidxl
Subscribers: danielcdh, llvm-commits
Differential Revision: https://reviews.llvm.org/D35823
llvm-svn: 309441
Summary:
LazyValueInfo currently computes the constant value of the switch condition through case edges, which allows the constant value to be propagated through the case edges.
But we have seen a case where a zero-extended value of the switch condition is used past case edges for which the constant propagation doesn't occur.
This patch adds a small logic to handle such a case in getEdgeValueLocal().
This is motivated by the Python 2.7 eval loop in PyEval_EvalFrameEx() where the lack of the constant propagation causes longer live ranges and more spill code than necessary.
With this patch, we see that the code size of PyEval_EvalFrameEx() decreases by ~5.4% and a performance test improves by ~4.6%.
Reviewers: wmi, dberlin, sanjoy
Reviewed By: sanjoy
Subscribers: davide, davidxl, llvm-commits
Differential Revision: https://reviews.llvm.org/D34822
llvm-svn: 309415
This patch reworks the function that searches constants in Add and Mul SCEV expression
chains so that now it does not visit a node more than once, and also renames this function
for better correspondence between its implementation and semantics.
Differential Revision: https://reviews.llvm.org/D35931
llvm-svn: 309367
This reverts commit r309080. The patch needs to clear out the
ScalarEvolution::ExitLimits cache in forgetMemoizedResults.
I've replied on the commit thread for the patch with more details.
llvm-svn: 309357
Summary: In performance tuning, we see performance benefits when enlarge the maximum num promotion targets to 3. This is safe as soon as we have total percentage threshold properly setup (https://reviews.llvm.org/D35962)
Reviewers: davidxl, tejohnson
Reviewed By: tejohnson
Subscribers: llvm-commits, sanjoy
Differential Revision: https://reviews.llvm.org/D35966
llvm-svn: 309346
Summary: In the current implementation, isPromotionProfitable only checks if the call count to a direct target is no less than a certain percentage threshold of the remaining call counts that have not been promoted. This causes code size problems when the target count is small but greater than a large portion of remaining counts. E.g. target1 takes 99.9%, while target2 takes 0.1%. Both targets will be promoted and inlined, makes the function size too large, which potentially prevents it from further inlining into its callers. This patch adds another percentage threshold against the total indirect call count. If the target count needs to be no less than both thresholds in order to be promoted speculatively.
Reviewers: davidxl, tejohnson
Reviewed By: tejohnson
Subscribers: sanjoy, llvm-commits
Differential Revision: https://reviews.llvm.org/D35962
llvm-svn: 309345
Currently CallAnalyzer::isGEPFree uses TTI::getGEPCost to check if GEP is free.
TTI::getGEPCost cannot handle cases when GEPs participate in Def-Use dependencies
(see https://reviews.llvm.org/D31186 for example).
There is TTI::getUserCost which can calculate the cost more accurately by
taking dependencies into account.
Differential Revision: https://reviews.llvm.org/D33685
llvm-svn: 309268
This patch adds a cache for computeExitLimit to save compilation time. A lot of examples of
tests that take extensive time to compile are attached to the bug 33494.
Differential Revision: https://reviews.llvm.org/D35827
llvm-svn: 309080
`SCEVUnknown::allUsesReplacedWith` does not need to call `forgetMemoizedResults`
since RAUW does a value-equivalent replacement by assumption. If this
assumption was false then the later setValPtr(New) call would be incorrect too.
This is a non-trivial performance optimization for functions with a large number
of loops since `forgetMemoizedResults` walks all loop backedge taken counts to
see if any of them use the SCEVUnknown being RAUWed. However, this improvement
is difficult to demonstrate without checking in an excessively large IR file.
llvm-svn: 309072
When SCEV calculates product of two SCEVAddRecs from the same loop, it
tries to combine them into one big AddRecExpr. If the sizes of the initial
SCEVs were `S1` and `S2`, the size of their product is `S1 + S2 - 1`, and every
operand of the resulting SCEV is combined from operands of initial SCEV and
has much higher complexity than they have.
As result, if we try to calculate something like:
%x1 = {a,+,b}
%x2 = mul i32 %x1, %x1
%x3 = mul i32 %x2, %x1
%x4 = mul i32 %x3, %x2
...
The size of such SCEVs grows as `2^N`, and the arguments
become more and more complex as we go forth. This leads
to long compilation and huge memory consumption.
This patch sets a limit after which we don't try to combine two
`SCEVAddRecExpr`s into one. By default, max allowed size of the
resulting AddRecExpr is set to 16.
Differential Revision: https://reviews.llvm.org/D35664
llvm-svn: 308847
This patch makes LSR generate better code for SystemZ in the cases of memory
intrinsics, Load->Store pairs or comparison of immediate with memory.
In order to achieve this, the following common code changes were made:
* New TTI hook: LSRWithInstrQueries(), which defaults to false. Controls if
LSR should do instruction-based addressing evaluations by calling
isLegalAddressingMode() with the Instruction pointers.
* In LoopStrengthReduce: handle address operands of memset, memmove and memcpy
as address uses, and call isFoldableMemAccessOffset() for any LSRUse::Address,
not just loads or stores.
SystemZ changes:
* isLSRCostLess() implemented with Insns first, and without ImmCost.
* New function supportedAddressingMode() that is a helper for TTI methods
looking at Instructions passed via pointers.
Review: Ulrich Weigand, Quentin Colombet
https://reviews.llvm.org/D35262https://reviews.llvm.org/D35049
llvm-svn: 308729
Chandler Carruth