For instructions that unlikely generate machine instructions, they should also have 0 latency.
Differential Revision: https://reviews.llvm.org/D37833
llvm-svn: 313288
invalidated SCCs even when we do not have an updated SCC to redirect
towards.
This comes up in a fairly subtle and surprising circumstance: we need to
have a connected but internal node in the call graph which later becomes
a disconnected island, and then gets deleted. All of this needs to
happen mid-CGSCC walk. Because it is disconnected, we have no way of
computing a new "current" SCC when it gets deleted. Instead, we need to
explicitly check for a deleted "current" SCC and bail out of the current
CGSCC step. This will bubble all the way up to the post-order walk and
then resume correctly.
I've included minimal tests for this bug. The specific behavior
matches something we've seen in the wild with the new PM combined with
ThinLTO and sample PGO, but I've not yet confirmed whether this is the
only issue there.
llvm-svn: 313242
This patch fixes pr34283, which exposed that the computation of
maximum legal width for vectorization was wrong, because it relied
on MaxInterleaveFactor to obtain the maximum stride used in the loop,
however not all strided accesses in the loop have an interleave-group
associated with them.
Instead of recording the maximum stride in the loop, which can be over
conservative (e.g. if the access with the maximum stride is not involved
in the dependence limitation), this patch tracks the actual maximum legal
width imposed by accesses that are involved in dependencies.
Differential Revision: https://reviews.llvm.org/D37507
llvm-svn: 313237
Summary:
Print profile counts as the third value in addition to the existing 'float' and
the 'int' values in the textual block frequency dump, if available.
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37835
llvm-svn: 313220
Summary:
Full inline cost is computed when -inline-cost-full is true or ORE is
non-null. This patch adds another way to compute full inline cost by
adding a field to InlineParams. This will be used by SampleProfileLoader
to check legality of inlining a callee that it wants to inline.
Reviewers: danielcdh, haicheng
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37819
llvm-svn: 313185
Summary:
Added text options to -pgo-view-counts and -pgo-view-raw-counts that dump block frequency and branch probability info in text.
This is useful when the graph is very large and complex (the dot command crashes, lines/edges too close to tell apart, hard to navigate without textual search) or simply when text is preferred.
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37776
llvm-svn: 313159
It now knows the tricks of both functions.
Also, fix a bug that considered allocas of non-zero address space to be always non null
Differential Revision: https://reviews.llvm.org/D37628
llvm-svn: 312869
This is intended to be a superset of the functionality from D31037 (EarlyCSE) but implemented
as an independent pass, so there's no stretching of scope and feature creep for an existing pass.
I also proposed a weaker version of this for SimplifyCFG in D30910. And I initially had almost
this same functionality as an addition to CGP in the motivating example of PR31028:
https://bugs.llvm.org/show_bug.cgi?id=31028
The advantage of positioning this ahead of SimplifyCFG in the pass pipeline is that it can allow
more flattening. But it needs to be after passes (InstCombine) that could sink a div/rem and
undo the hoisting that is done here.
Decomposing remainder may allow removing some code from the backend (PPC and possibly others).
Differential Revision: https://reviews.llvm.org/D37121
llvm-svn: 312862
In function TargetTransformInfo::getInstructionCost, all enum values in the switch statement has been covered, so the default is unnecessary, and may cause error with option -Werror,-Wcovered-switch-default, so remove it.
llvm-svn: 312834
Current TargetTransformInfo can support throughput cost model and code size model, but sometimes we also need instruction latency cost model in different optimizations. Hal suggested we need a single public interface to query the different cost of an instruction. So I proposed following interface:
enum TargetCostKind {
TCK_RecipThroughput, ///< Reciprocal throughput.
TCK_Latency, ///< The latency of instruction.
TCK_CodeSize ///< Instruction code size.
};
int getInstructionCost(const Instruction *I, enum TargetCostKind kind) const;
All clients should mainly use this function to query the cost of an instruction, parameter <kind> specifies the desired cost model.
This patch also provides a simple default implementation of getInstructionLatency.
The default getInstructionLatency provides latency numbers for only small number of instruction classes, those latency numbers are only reasonable for modern OOO processors. It can be extended in following ways:
Add more detail into this function.
Add getXXXLatency function and call it from here.
Implement target specific getInstructionLatency function.
Differential Revision: https://reviews.llvm.org/D37170
llvm-svn: 312832
SLP vectorizer supports horizontal reductions for Add/FAdd binary
operations. Patch adds support for horizontal min/max reductions.
Function getReductionCost() is split to getArithmeticReductionCost() for
binary operation reductions and getMinMaxReductionCost() for min/max
reductions.
Patch fixes PR26956.
Differential revision: https://reviews.llvm.org/D27846
llvm-svn: 312791
This is a preliminary step towards solving the remaining part of PR27145 - IR for isfinite():
https://bugs.llvm.org/show_bug.cgi?id=27145
In order to solve that one more generally, we need to add matching for and/or of fcmp ord/uno
with a constant operand.
But while looking at those patterns, I realized we were missing a canonicalization for nonzero
constants. Rather than limiting to just folds for constants, we're adding a general value
tracking method for this based on an existing DAG helper.
By transforming everything to 0.0, we can simplify the existing code in foldLogicOfFCmps()
and pick up missing vector folds.
Differential Revision: https://reviews.llvm.org/D37427
llvm-svn: 312591
This patch teaches decomposeBitTestICmp to look through truncate instructions on the input to the compare. If a truncate is found it will now return the pre-truncated Value and appropriately extend the APInt mask.
This allows some code to be removed from InstSimplify that was doing this functionality.
This allows InstCombine's bit test combining code to match a pre-truncate Value with the same Value appear with an 'and' on another icmp. Or it allows us to combine a truncate to i16 and a truncate to i8. This also required removing the type check from the beginning of getMaskedTypeForICmpPair, but I believe that's ok because we still have to find two values from the input to each icmp that are equal before we'll do any transformation. So the type check was really just serving as an early out.
There was one user of decomposeBitTestICmp that didn't want to look through truncates, so I've added a flag to prevent that behavior when necessary.
Differential Revision: https://reviews.llvm.org/D37158
llvm-svn: 312382
Previously this would sporadically crash as TargetType
was never initialized. We special-case the single-operand
case returning earlier and trying to mimic the behaviour of
isLegalAddressingMode as closely as possible.
Differential Revision: https://reviews.llvm.org/D37277
llvm-svn: 312357
We expect the pointer to be initialized by the above loop, but
if that's not executed, the contents are garbage.
A fix for the crash will be committed immediately after.
llvm-svn: 312353
In LLVM IR the following code:
%r = urem <ty> %t, %b
is equivalent to
%q = udiv <ty> %t, %b
%s = mul <ty> nuw %q, %b
%r = sub <ty> nuw %t, %q ; (t / b) * b + (t % b) = t
As UDiv, Mul and Sub are already supported by SCEV, URem can be implemented
with minimal effort using that relation:
%r --> (-%b * (%t /u %b)) + %t
We implement two special cases:
- if %b is 1, the result is always 0
- if %b is a power-of-two, we produce a zext/trunc based expression instead
That is, the following code:
%r = urem i32 %t, 65536
Produces:
%r --> (zext i16 (trunc i32 %a to i16) to i32)
Note that while this helps get a tighter bound on the range analysis and the
known-bits analysis, this exposes some normalization shortcoming of SCEVs:
%div = udim i32 %a, 65536
%mul = mul i32 %div, 65536
%rem = urem i32 %a, 65536
%add = add i32 %mul, %rem
Will usually not be reduced.
llvm-svn: 312329
Summary:
We add the precise cache sizes and associativity for the following Intel
architectures:
- Penry
- Nehalem
- Westmere
- Sandy Bridge
- Ivy Bridge
- Haswell
- Broadwell
- Skylake
- Kabylake
Polly uses since several months a performance model for BLAS computations that
derives optimal cache and register tile sizes from cache and latency
information (based on ideas from "Analytical Modeling Is Enough for High-Performance BLIS", by Tze Meng Low published at TOMS 2016).
While bootstrapping this model, these target values have been kept in Polly.
However, as our implementation is now rather mature, it seems time to teach
LLVM itself about cache sizes.
Interestingly, L1 and L2 cache sizes are pretty constant across
micro-architectures, hence a set of architecture specific default values
seems like a good start. They can be expanded to more target specific values,
in case certain newer architectures require different values. For now a set
of Intel architectures are provided.
Just as a little teaser, for a simple gemm kernel this model allows us to
improve performance from 1.2s to 0.27s. For gemm kernels with less optimal
memory layouts even larger speedups can be reported.
Reviewers: Meinersbur, bollu, singam-sanjay, hfinkel, gareevroman, fhahn, sebpop, efriedma, asb
Reviewed By: fhahn, asb
Subscribers: lsaba, asb, pollydev, llvm-commits
Differential Revision: https://reviews.llvm.org/D37051
llvm-svn: 311647
Currently, the inline cost model will bail once the inline cost exceeds the
inline threshold in order to avoid unnecessary compile-time. However, when
debugging it is useful to compute the full cost, so this command line option
is added to override the default behavior.
I took over this work from Chad Rosier (mcrosier@codeaurora.org).
Differential Revision: https://reviews.llvm.org/D35850
llvm-svn: 311371
Summary:
This updates the Inliner to only add a single Optimization
Remark when Inlining, rather than an Analysis Remark and an
Optimization Remark.
Fixes https://bugs.llvm.org/show_bug.cgi?id=33786
Reviewers: anemet, davidxl, chandlerc
Reviewed By: anemet
Subscribers: haicheng, fhahn, mehdi_amini, dblaikie, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D36054
llvm-svn: 311349
Summary:
This updates the Inliner to only add a single Optimization
Remark when Inlining, rather than an Analysis Remark and an
Optimization Remark.
Fixes https://bugs.llvm.org/show_bug.cgi?id=33786
Reviewers: anemet, davidxl, chandlerc
Reviewed By: anemet
Subscribers: haicheng, fhahn, mehdi_amini, dblaikie, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D36054
llvm-svn: 311273
This recommits r310869, with the moved files and no extra changes.
Original commit message:
This addresses a fixme in InstSimplify about using decomposeBitTest. This also fixes InstSimplify to handle ugt and ult compares too.
I've modified the interface a little to return only the APInt version of the mask that InstSimplify needs. InstCombine now has a small wrapper routine to create a Constant out of it. I've also dropped the returning of 0 since InstSimplify doesn't need that. So InstCombine creates a zero constant itself.
I also had to make decomposeBitTest support vectors since InstSimplify needs that.
As InstSimplify can't use something from the Transforms library, I've moved the CmpInstAnalysis code to the Analysis library.
Differential Revision: https://reviews.llvm.org/D36593
llvm-svn: 310889
On some targets, the penalty of executing runtime unrolling checks
and then not the unrolled loop can be significantly detrimental to
performance. This results in the need to be more conservative with
the unroll count, keeping a trip count of 2 reduces the overhead as
well as increasing the chance of the unrolled body being executed. But
being conservative leaves performance gains on the table.
This patch enables the unrolling of the remainder loop introduced by
runtime unrolling. This can help reduce the overhead of misunrolled
loops because the cost of non-taken branches is much less than the
cost of the backedge that would normally be executed in the remainder
loop. This allows larger unroll factors to be used without suffering
performance loses with smaller iteration counts.
Differential Revision: https://reviews.llvm.org/D36309
llvm-svn: 310824
printing techniques with a DEBUG_TYPE controlling them.
It was a mistake to start re-purposing the pass manager `DebugLogging`
variable for generic debug printing -- those logs are intended to be
very minimal and primarily used for testing. More detailed and
comprehensive logging doesn't make sense there (it would only make for
brittle tests).
Moreover, we kept forgetting to propagate the `DebugLogging` variable to
various places making it also ineffective and/or unavailable. Switching
to `DEBUG_TYPE` makes this a non-issue.
llvm-svn: 310695
must-alias(p, sz_p, p, sz_q) irrespective of access sizes sz_p, sz_q
As discussed a couple of weeks ago on the ML.
This makes the behavior consistent with that of BasicAA.
AA clients already check the obj size themselves and may not require the
obj size to match exactly the access size (e.g., in case of store forwarding)
llvm-svn: 310495
isLegalAddressingMode() has recently gained the extra optional Instruction*
parameter, and therefore it can now do the job that previously only
isFoldableMemAccess() could do.
The SystemZ implementation of isLegalAddressingMode() has gained the
functionality of checking for offsets, which used to be done with
isFoldableMemAccess().
The isFoldableMemAccess() hook has been removed everywhere.
Review: Quentin Colombet, Ulrich Weigand
https://reviews.llvm.org/D35933
llvm-svn: 310463
limited batch updates.
Specifically, allow removing multiple reference edges starting from
a common source node. There are a few constraints that play into
supporting this form of batching:
1) The way updates occur during the CGSCC walk, about the most we can
functionally batch together are those with a common source node. This
also makes the batching simpler to implement, so it seems
a worthwhile restriction.
2) The far and away hottest function for large C++ files I measured
(generated code for protocol buffers) showed a huge amount of time
was spent removing ref edges specifically, so it seems worth focusing
there.
3) The algorithm for removing ref edges is very amenable to this
restricted batching. There are just both API and implementation
special casing for the non-batch case that gets in the way. Once
removed, supporting batches is nearly trivial.
This does modify the API in an interesting way -- now, we only preserve
the target RefSCC when the RefSCC structure is unchanged. In the face of
any splits, we create brand new RefSCC objects. However, all of the
users were OK with it that I could find. Only the unittest needed
interesting updates here.
How much does batching these updates help? I instrumented the compiler
when run over a very large generated source file for a protocol buffer
and found that the majority of updates are intrinsically updating one
function at a time. However, nearly 40% of the total ref edges removed
are removed as part of a batch of removals greater than one, so these
are the cases batching can help with.
When compiling the IR for this file with 'opt' and 'O3', this patch
reduces the total time by 8-9%.
Differential Revision: https://reviews.llvm.org/D36352
llvm-svn: 310450
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
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
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
infinite-inlining across multiple runs of the inliner by keeping a tiny
history of internal-to-SCC inlining decisions.
This is still a bit gross, but I don't yet have any fundamentally better
ideas and numerous people are blocked on this to use new PM and ThinLTO
together.
The core of the idea is to detect when we are about to do an inline that
has a chance of re-splitting an SCC which we have split before with
a similar inlining step. That is a critical component in the inlining
forming a cycle and so far detects all of the various cyclic patterns
I can come up with as well as the original real-world test case (which
comes from a ThinLTO build of libunwind).
I've added some tests that I think really demonstrate what is going on
here. They are essentially state machines that march the inliner through
various steps of a cycle and check that we stop when the cycle is closed
and that we actually did do inlining to form that cycle.
A lot of thanks go to Eric Christopher and Sanjoy Das for the help
understanding this issue and improving the test cases.
The biggest "yuck" here is the layering issue -- the CGSCC pass manager
is providing somewhat magical state to the inliner for it to use to make
itself converge. This isn't great, but I don't honestly have a lot of
better ideas yet and at least seems nicely isolated.
I have tested this patch, and it doesn't block *any* inlining on the
entire LLVM test suite and SPEC, so it seems sufficiently narrowly
targeted to the issue at hand.
We have come up with hypothetical issues that this patch doesn't cover,
but so far none of them are practical and we don't have a viable
solution yet that covers the hypothetical stuff, so proceeding here in
the interim. Definitely an area that we will be back and revisiting in
the future.
Differential Revision: https://reviews.llvm.org/D36188
llvm-svn: 309784
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
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 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
Summary: We can use the template parameter `IsPostDom` to pick an appropriate SmallVector size to store DomTree roots for dominators and postdominators. Before, the code would always allocate memory with `std::vector`.
Reviewers: dberlin, davide, sanjoy, grosser
Reviewed By: grosser
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D35636
llvm-svn: 309148
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
Propagates the GraphT template argument to the default value of
the AnalysisGraphTraitsT template argument. This allows to specialize
the DefaultAnalysisGraphTraits<AnalysisT,GraphT> for analysis with a
graph type different from the analysis type and it will automatically
get picked-up.
Note: This was probably the intended purpose and should not result in any
functional change.
llvm-svn: 308878
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
functions.
In the prior commit, we provide ordering to the LCG between functions
and library function definitions that they might begin to call through
transformations. But we still would delete these library functions from
the call graph if they became dead during inlining.
While this immediately crashed, it also exposed a loss of information.
We shouldn't remove definitions of library functions that can still
usefully participate in the LCG-powered CGSCC optimization process. If
new call edges are formed, we want to have definitions to be called.
We can still remove these functions if truly dead using global-dce, etc,
but removing them during the CGSCC walk is premature.
This fixes a crash in the new PM when optimizing some unusual libraries
that end up with "internal" lib functions such as the code in the "R"
language's libraries.
llvm-svn: 308417
using runtime checks
Extend the SCEVPredicateRewriter to work a bit harder when it encounters an
UnknownSCEV for a Phi node; Try to build an AddRecurrence also for Phi nodes
whose update chain involves casts that can be ignored under the proper runtime
overflow test. This is one step towards addressing PR30654.
Differential revision: http://reviews.llvm.org/D30041
llvm-svn: 308299
function to every defined function known to LLVM as a library function.
LLVM can introduce calls to these functions either by replacing other
library calls or by recognizing patterns (such as memset_pattern or
vector math patterns) and replacing those with calls. When these library
functions are actually defined in the module, we need to have reference
edges to them initially so that we visit them during the CGSCC walk in
the right order and can effectively rebuild the call graph afterward.
This was discovered when building code with Fortify enabled as that is
a common case of both inline definitions of library calls and
simplifications of code into calling them.
This can in extreme cases of LTO-ing with libc introduce *many* more
reference edges. I discussed a bunch of different options with folks but
all of them are unsatisfying. They either make the graph operations
substantially more complex even when there are *no* defined libfuncs, or
they introduce some other complexity into the callgraph. So this patch
goes with the simplest possible solution of actual synthetic reference
edges. If this proves to be a memory problem, I'm happy to implement one
of the clever techniques to save memory here.
llvm-svn: 308088
Now, getUserCost() only checks the src and dst types of EXT to decide it is free
or not. This change first checks the types, then calls isExtFreeImpl(), and
check if EXT can form ExtLoad at last. Currently, only AArch64 has customized
implementation of isExtFreeImpl() to check if EXT can be folded into its use.
Differential Revision: https://reviews.llvm.org/D34458
llvm-svn: 308076
Summary:
DominatorTreeBase used to have IsPostDominators (bool) member to indicate if the tree is a dominator or a postdominator tree. This made it possible to switch between the two 'modes' at runtime, but it isn't used in practice anywhere.
This patch makes IsPostDominator a template argument. This way, it is easier to switch between different algorithms at compile-time based on this argument and design external utilities around it. It also makes it impossible to incidentally assign a postdominator tree to a dominator tree (and vice versa), and to further simplify template code in GenericDominatorTreeConstruction.
Reviewers: dberlin, sanjoy, davide, grosser
Reviewed By: dberlin
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D35315
llvm-svn: 308040
Where is is needed (at the end of headers that define it), be
consistent about its use.
Also fix a few header guards that I found in the process.
Differential Revision: https://reviews.llvm.org/D34916
llvm-svn: 307840
Summary:
Solves PR33689.
If the pointer size is less than the size of the type used for the array
size in an alloca (the <ty> type below) then we could trigger the assert in
the PR. In that example we have pointer size i16 and <ty> is i32.
<result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>]
Handle the situation by allowing truncation as well as zero extension in
ObjectSizeOffsetVisitor::visitAllocaInst().
Also, we now detect overflow in visitAllocaInst(), similar to how it was
already done in visitCallSite().
Reviewers: craig.topper, rnk, george.burgess.iv
Reviewed By: george.burgess.iv
Subscribers: davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D35003
llvm-svn: 307754
invalidation of analyses when merging SCCs.
While I've added a bunch of testing of this, it takes something much
more like the inliner to really trigger this as you need to have
partially-analyzed SCCs with updates at just the right time. So I've
added a direct test for this using the inliner and verifying the
domtree. Without the changes here, this test ends up finding a stale
dominator tree.
However, to handle this properly, we need to invalidate analyses
*before* merging the SCCs. After talking to Philip and Sanjoy about this
they convinced me this was the right approach. To do this, we need
a callback mechanism when merging SCCs so we can observe the cycle that
will be merged before the merge happens. This API update ended up being
surprisingly easy.
With this commit, the new PM passes the test-suite again. It hadn't
since MemorySSA was enabled for EarlyCSE as that also will find this bug
very quickly.
llvm-svn: 307498
dependencies between analyses.
This uncovers even more issues with the proxies and the splitting apart
of SCCs which are fixed in this patch. I discovered this while trying to
add more rigorous testing for a change I'm making to the call graph
update invalidation logic.
llvm-svn: 307497
Adds loop expansions for known-size and unknown-sized memcpy calls, allowing the
target to provide the operand types through TTI callbacks. The default values
for the TTI callbacks use int8 operand types and matches the existing behaviour
if they aren't overridden by the target.
Differential revision: https://reviews.llvm.org/D32536
llvm-svn: 307346
This patch adds support for handling some forms of ands and ors in
ValueTracking's isImpliedCondition API.
PR33611
https://reviews.llvm.org/D34901
llvm-svn: 307304
This reverts commit r306907 and reapplies the patches in the title.
The patches used to make one of the
CodeGen/ARM/2011-02-07-AntidepClobber.ll test to fail because of a
missing null check.
llvm-svn: 306919
This reverts commit r306894.
Revert "[Dominators] Add NearestCommonDominator verification"
This reverts commit r306893.
Revert "[Dominators] Keep tree level in DomTreeNode and use it to find NCD and answer dominance queries"
This reverts commit r306892.
llvm-svn: 306907
Summary: This patch teaches IteratedDominanceFrontier to use the level information stored in DomTreeNodes instead of calculating it manually.
Reviewers: dberlin, sanjoy, davide
Reviewed By: davide
Subscribers: davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D34703
llvm-svn: 306894
Summary:
To enable profile hotness information in diagnostics output, Clang takes
the option `-fdiagnostics-show-hotness` -- that's "diagnostics", with an
"s" at the end. Clang also defines `CodeGenOptions::DiagnosticsWithHotness`.
LLVM, on the other hand, defines
`LLVMContext::getDiagnosticHotnessRequested` -- that's "diagnostic", not
"diagnostics". It's a small difference, but it's confusing, typo-inducing, and
frustrating.
Add a new method with the spelling "diagnostics", and "deprecate" the
old spelling.
Reviewers: anemet, davidxl
Reviewed By: anemet
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D34864
llvm-svn: 306848
In rL300494 there was an attempt to deal with excessive compile time on
invocations of getSign/ZeroExtExpr using local caching. This approach only
helps if we request the same SCEV multiple times throughout recursion. But
in the bug PR33431 we see a case where we request different values all the time,
so caching does not help and the size of the cache grows enormously.
In this patch we remove the local cache for this methods and add the recursion
depth limit instead, as we do for arithmetics. This gives us a guarantee that the
invocation sequence is limited and reasonably short.
Differential Revision: https://reviews.llvm.org/D34273
llvm-svn: 306785
The style guide states that the explicit `inline`
should not be used with inline methods. classof is
very common inline method with a fair amount on
inconsistency:
$ git grep classof ./include | grep inline | wc -l
230
$ git grep classof ./include | grep -v inline | wc -l
257
I chose to target this method rather the larger change
since this method is easily cargo-culted (I did it at
least once). I considered doing the larger change and
removing all occurrences but that would be a much larger
change.
Differential Revision: https://reviews.llvm.org/D33906
llvm-svn: 306731
Summary:
When we have patterns like
loop:
%la = load %ptr, !tbaa
%lba = load %ptr, !tbaa !noalias
AliasSetTracker would previously think that the two types of annotation for
the pointer conflict, dropping both for the purpose of determining alias sets.
That is clearly way too conservative, as the tbaa is still valid whether or
not one of the memory accesses has additional AA metadata. We could go
one step further and attempt to properly merge the AA metadata,
but it's not clear that that would be worth it since that may introduce
additional MD nodes, which may be undesirable since this is merely an
Analysis.
Reviewers: hfinkel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32139
llvm-svn: 306727
In LLVM IR the following code:
%r = urem <ty> %t, %b
is equivalent to:
%q = udiv <ty> %t, %b
%s = mul <ty> nuw %q, %b
%r = sub <ty> nuw %t, %q ; (t / b) * b + (t % b) = t
As UDiv, Mul and Sub are already supported by SCEV, URem can be
implemented with minimal effort this way.
Note: While SRem and SDiv are also related this way, SCEV does not
provides SDiv yet.
llvm-svn: 306695
The changes are a result of discussion of https://reviews.llvm.org/D33685.
It solves the following problem:
1. We can inform getGEPCost about simplified indices to help it with
calculating the cost. But getGEPCost does not take into account the
context which GEPs are used in.
2. We have getUserCost which can take the context into account but we cannot
inform about simplified indices.
With the changes getUserCost will have access to additional information
as getGEPCost has.
The one parameter getUserCost is also provided.
Differential Revision: https://reviews.llvm.org/D34057
llvm-svn: 306674
Summary: Make OrderedInstructions and OrderedBasicBlock use AssertingVH to try and catch mistakes
Reviewers: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34780
llvm-svn: 306605
Summary:
This patch changes getRange to getRangeRef and returns a reference to the ConstantRange object stored inside the DenseMap caches. We then take advantage of that to add new helper methods that can return min/max value of a signed or unsigned ConstantRange using that reference without first copying the ConstantRange.
getRangeRef calls itself recursively and I believe the reference return is fine for those calls.
I've left getSignedRange and getUnsignedRange returning a ConstantRange object so they will make a copy now. This is to ensure safety since the reference will be invalidated if the DenseMap changes.
I'm sure there are still more places that can take advantage of the reference and I'll submit future patches as I find them.
Reviewers: sanjoy, davide
Reviewed By: sanjoy
Subscribers: zzheng, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D32978
llvm-svn: 306229
Currently JumpThreading can use LazyValueInfo to analyze an 'and' or 'or' of compare if the compare is fed by a livein of a basic block. This can be used to to prove the condition can't be met for some predecessor and the jump from that predecessor can be moved to the false path of the condition.
But if the compare is something that InstCombine turns into an add and a single compare, it can't be analyzed because the livein is now an input to the add and not the compare.
This patch adds a new method to LVI to get a ConstantRange on an edge. Then we teach jump threading to detect the add livein feeding a compare and to get the ConstantRange and propagate it.
Differential Revision: https://reviews.llvm.org/D33262
llvm-svn: 306085
This is a fix for PR33292 that shows a case of extremely long compilation
of a single .c file with clang, with most time spent within SCEV.
We have a mechanism of limiting recursion depth for getAddExpr to avoid
long analysis in SCEV. However, there are calls from getAddExpr to getMulExpr
and back that do not propagate the info about depth. As result of this, a chain
getAddExpr -> ... .> getAddExpr -> getMulExpr -> getAddExpr -> ... -> getAddExpr
can be extremely long, with every segment of getAddExpr's being up to max depth long.
This leads either to long compilation or crash by stack overflow. We face this situation while
analyzing big SCEVs in the test of PR33292.
This patch applies the same limit on max expression depth for getAddExpr and getMulExpr.
Differential Revision: https://reviews.llvm.org/D33984
llvm-svn: 305463
User has 3 signatures for operator new today. They take a single size, a size and a number of users, and a size, number of users, and descriptor size.
Historically there used to only be one signature that took size and a number of uses. Long ago derived classes implemented their own versions that took just a size and would call the size and use count version. Then they left an unimplemented signature for the size and use count signature from User. As we moved to C++11 this unimplemented signature because = delete.
Since then operator new has picked up two new signatures for operator new. But when the 3 argument version was added it was never added to the delete list in all of the derived classes where the 2 argument version is deleted. This makes things inconsistent.
I believe once one version of operator new is created in a derived class name hiding will take care of making all of the base class signatures unavailable. So I don't think the deleted lines are needed at all.
This patch removes all of the deletes in cases where there is an override or there is already a delete of another signature (that should trigger name hiding too).
Differential Revision: https://reviews.llvm.org/D34120
llvm-svn: 305251
Summary: The method TargetTransformInfo::getRegisterBitWidth() is declared const, but the type erasing implementation classes (TargetTransformInfo::Concept & TargetTransformInfo::Model) that were introduced by Chandler in https://reviews.llvm.org/D7293 do not have the method declared const. This is an NFC to tidy up the const consistency between TTI and its implementation.
Reviewers: chandlerc, rnk, reames
Reviewed By: reames
Subscribers: reames, jfb, arsenm, dschuff, nemanjai, nhaehnle, javed.absar, sbc100, jgravelle-google, llvm-commits
Differential Revision: https://reviews.llvm.org/D33903
llvm-svn: 305189
The zero heuristic assumes that integers are more likely positive than negative,
but this also has the effect of assuming that strcmp return values are more
likely positive than negative. Given that for nonzero strcmp return values it's
the ordering of arguments that determines the sign of the result there's no
reason to assume that's true.
Fix this by inspecting the LHS of the compare and using TargetLibraryInfo to
decide if it's strcmp-like, and if so only assume that nonzero is more likely
than zero i.e. strings are more often different than the same. This causes a
slight code generation change in the spec2006 benchmark 403.gcc, but with no
noticeable performance impact. The intent of this patch is to allow better
optimisation of dhrystone on Cortex-M cpus, but currently it won't as there are
also some changes that need to be made to if-conversion.
Differential Revision: https://reviews.llvm.org/D33934
llvm-svn: 304970
Summary:
LVIPrinter pass was previously relying on the LVICache. We now directly call the
the LVI functions which solves the value if the LVI information is not already
available in the cache. This has 2 benefits over the printing of LVI cache:
1. higher coverage (i.e. catches errors) in LVI code when cache value is
invalidated.
2. relies on the core functions, and not dependent on the LVI cache (which may
be scrapped at some point).
It would still catch any cache invalidation errors, since we first go through
the cache.
Reviewers: reames, dberlin, sanjoy
Reviewed by: reames
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32135
llvm-svn: 304819
Summary:
Expanding the loop idiom test for memcpy to also recognize
unordered atomic memcpy. The only difference for recognizing
an unordered atomic memcpy and instead of a normal memcpy is
that the loads and/or stores involved are unordered atomic operations.
Background: http://lists.llvm.org/pipermail/llvm-dev/2017-May/112779.html
Patch by Daniel Neilson!
Reviewers: reames, anna, skatkov
Reviewed By: reames, anna
Subscribers: llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D33243
llvm-svn: 304806
I did this a long time ago with a janky python script, but now
clang-format has built-in support for this. I fed clang-format every
line with a #include and let it re-sort things according to the precise
LLVM rules for include ordering baked into clang-format these days.
I've reverted a number of files where the results of sorting includes
isn't healthy. Either places where we have legacy code relying on
particular include ordering (where possible, I'll fix these separately)
or where we have particular formatting around #include lines that
I didn't want to disturb in this patch.
This patch is *entirely* mechanical. If you get merge conflicts or
anything, just ignore the changes in this patch and run clang-format
over your #include lines in the files.
Sorry for any noise here, but it is important to keep these things
stable. I was seeing an increasing number of patches with irrelevant
re-ordering of #include lines because clang-format was used. This patch
at least isolates that churn, makes it easy to skip when resolving
conflicts, and gets us to a clean baseline (again).
llvm-svn: 304787
This patch does an inline expansion of memcmp.
It changes the memcmp library call into an inline expansion when the size is
known at compile time and is under a target specified threshold.
This expansion is implemented in CodeGenPrepare and expands into straight line
code. The target specifies a maximum load size and the expansion works by using
this size to load the two sources, compare, and exit early if a difference is
found. It also has a special case when the memcmp result is used in a compare
to zero equality.
Differential Revision: https://reviews.llvm.org/D28637
llvm-svn: 304313
Params DT and LI are redundant, because these values are contained in fields anyways.
Differential Revision: https://reviews.llvm.org/D33668
llvm-svn: 304204
The patch rL303730 was reverted because test lsr-expand-quadratic.ll failed on
many non-X86 configs with this patch. The reason of this is that the patch
makes a correctless fix that changes optimizer's behavior for this test.
Without the change, LSR was making an overconfident simplification basing on a
wrong SCEV. Apparently it did not need the IV analysis to do this. With the
change, it chose a different way to simplify (that wasn't so confident), and
this way required the IV analysis. Now, following the right execution path,
LSR tries to make a transformation relying on IV Users analysis. This analysis
is target-dependent due to this code:
// LSR is not APInt clean, do not touch integers bigger than 64-bits.
// Also avoid creating IVs of non-native types. For example, we don't want a
// 64-bit IV in 32-bit code just because the loop has one 64-bit cast.
uint64_t Width = SE->getTypeSizeInBits(I->getType());
if (Width > 64 || !DL.isLegalInteger(Width))
return false;
To make a proper transformation in this test case, the type i32 needs to be
legal for the specified data layout. When the test runs on some non-X86
configuration (e.g. pure ARM 64), opt gets confused by the specified target
and does not use it, rejecting the specified data layout as well. Instead,
it uses some default layout that does not treat i32 as a legal type
(currently the layout that is used when it is not specified does not have
legal types at all). As result, the transformation we expect to happen does
not happen for this test.
This re-enabling patch does not have any source code changes compared to the
original patch rL303730. The only difference is that the failing test is
moved to X86 directory and now has requirement of running on x86 only to comply
with the specified target triple and data layout.
Differential Revision: https://reviews.llvm.org/D33543
llvm-svn: 303971
having it internally allocate the loop.
This is a much more flexible API and necessary in the new loop unswitch
to reasonably support both new and old PMs in common code. It also just
seems like a cleaner separation of concerns.
NFC, this should just be a pure refactoring.
Differential Revision: https://reviews.llvm.org/D33528
llvm-svn: 303834
The loop vectorizer usually vectorizes any instruction it can and then
extracts the elements for a scalarized use. On SystemZ, all elements
containing addresses must be extracted into address registers (GRs). Since
this extraction is not free, it is better to have the address in a suitable
register to begin with. By forcing address arithmetic instructions and loads
of addresses to be scalar after vectorization, two benefits result:
* No need to extract the register
* LSR optimizations trigger (LSR isn't handling vector addresses currently)
Benchmarking show improvements on SystemZ with this new behaviour.
Any other target could try this by returning false in the new hook
prefersVectorizedAddressing().
Review: Renato Golin, Elena Demikhovsky, Ulrich Weigand
https://reviews.llvm.org/D32422
llvm-svn: 303744
When folding arguments of AddExpr or MulExpr with recurrences, we rely on the fact that
the loop of our base recurrency is the bottom-lost in terms of domination. This assumption
may be broken by an expression which is treated as invariant, and which depends on a complex
Phi for which SCEVUnknown was created. If such Phi is a loop Phi, and this loop is lower than
the chosen AddRecExpr's loop, it is invalid to fold our expression with the recurrence.
Another reason why it might be invalid to fold SCEVUnknown into Phi start value is that unlike
other SCEVs, SCEVUnknown are sometimes position-bound. For example, here:
for (...) { // loop
phi = {A,+,B}
}
X = load ...
Folding phi + X into {A+X,+,B}<loop> actually makes no sense, because X does not exist and cannot
exist while we are iterating in loop (this memory can be even not allocated and not filled by this moment).
It is only valid to make such folding if X is defined before the loop. In this case the recurrence {A+X,+,B}<loop>
may be existant.
This patch prohibits folding of SCEVUnknown (and those who use them) into the start value of an AddRecExpr,
if this instruction is dominated by the loop. Merging the dominating unknown values is still valid. Some tests that
relied on the fact that some SCEVUnknown should be folded into AddRec's are changed so that they no longer
expect such behavior.
llvm-svn: 303730
The forward declarations and the SimplifyQuery class at the beginning of the namespace weren't indented. But the closing brace for SimplifyQuery and everything after it were indented.
This commit makes the whole file consistent to no identation per coding standards. The signature of every function in this file changed a few weeks ago so this isn't a big disturbance to the revision history.
llvm-svn: 303588
This is a re-application of a r303497 that was reverted in r303498.
I thought it had broken a bot when it had not (the breakage did not
go away with the revert).
This change makes the split between the "exact" backedge taken count
and the "maximum" backedge taken count a bit more obvious. Both of
these are upper bounds on the number of times the loop header
executes (since SCEV does not account for most kinds of abnormal
control flow), but the latter is guaranteed to be a constant.
There were a few places where the max backedge taken count *was* a
non-constant; I've changed those to compute constants instead.
At this point, I'm not sure if the constant max backedge count can be
computed by calling `getUnsignedRange(Exact).getUnsignedMax()` without
losing precision. If it can, we can simplify even further by making
`getMaxBackedgeTakenCount` a thin wrapper around
`getBackedgeTakenCount` and `getUnsignedRange`.
llvm-svn: 303531
This change makes the split between the "exact" backedge taken count
and the "maximum" backedge taken count a bit more obvious. Both of
these are upper bounds on the number of times the loop header
executes (since SCEV does not account for most kinds of abnormal
control flow), but the latter is guaranteed to be a constant.
There were a few places where the max backedge taken count *was* a
non-constant; I've changed those to compute constants instead.
At this point, I'm not sure if the constant max backedge count can be
computed by calling `getUnsignedRange(Exact).getUnsignedMax()` without
losing precision. If it can, we can simplify even further by making
`getMaxBackedgeTakenCount` a thin wrapper around
`getBackedgeTakenCount` and `getUnsignedRange`.
llvm-svn: 303497
Summary: This allows pthread_self to be pulled out of a loop by LICM.
Reviewers: hfinkel, arsenm, davide
Reviewed By: davide
Subscribers: davide, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D32782
llvm-svn: 303495
Refactor the strlen optimization code to work for both strlen and wcslen.
This especially helps with programs in the wild where people pass
L"string"s to const std::wstring& function parameters and the wstring
constructor gets inlined.
This also fixes a lingerind API problem/bug in getConstantStringInfo()
where zeroinitializers would always give you an empty string (without a
length) back regardless of the actual length of the initializer which
did not work well in the TrimAtNul==false causing the PR mentioned
below.
Note that the fixed getConstantStringInfo() needed fixes to SelectionDAG
memcpy lowering and may lead to some cases for out-of-bounds
zeroinitializer accesses not getting optimized anymore. So some code
with UB may produce out of bound memory reads now instead of just
producing zeros.
The refactoring "accidentally" fixes http://llvm.org/PR32124
Differential Revision: https://reviews.llvm.org/D32839
llvm-svn: 303461
Summary:
Implements PR889
Removing the virtual table pointer from Value saves 1% of RSS when doing
LTO of llc on Linux. The impact on time was positive, but too noisy to
conclusively say that performance improved. Here is a link to the
spreadsheet with the original data:
https://docs.google.com/spreadsheets/d/1F4FHir0qYnV0MEp2sYYp_BuvnJgWlWPhWOwZ6LbW7W4/edit?usp=sharing
This change makes it invalid to directly delete a Value, User, or
Instruction pointer. Instead, such code can be rewritten to a null check
and a call Value::deleteValue(). Value objects tend to have their
lifetimes managed through iplist, so for the most part, this isn't a big
deal. However, there are some places where LLVM deletes values, and
those places had to be migrated to deleteValue. I have also created
llvm::unique_value, which has a custom deleter, so it can be used in
place of std::unique_ptr<Value>.
I had to add the "DerivedUser" Deleter escape hatch for MemorySSA, which
derives from User outside of lib/IR. Code in IR cannot include MemorySSA
headers or call the MemoryAccess object destructors without introducing
a circular dependency, so we need some level of indirection.
Unfortunately, no class derived from User may have any virtual methods,
because adding a virtual method would break User::getHungOffOperands(),
which assumes that it can find the use list immediately prior to the
User object. I've added a static_assert to the appropriate OperandTraits
templates to help people avoid this trap.
Reviewers: chandlerc, mehdi_amini, pete, dberlin, george.burgess.iv
Reviewed By: chandlerc
Subscribers: krytarowski, eraman, george.burgess.iv, mzolotukhin, Prazek, nlewycky, hans, inglorion, pcc, tejohnson, dberlin, llvm-commits
Differential Revision: https://reviews.llvm.org/D31261
llvm-svn: 303362
ProfileSummaryInfo already checks whether the module has sample profile
in determining profile counts. This will also be useful in inliner to
clean up threshold updates.
llvm-svn: 303204
ARM Neon has native support for half-sized vector registers (64 bits). This
is beneficial for example for 2D and 3D graphics. This patch adds the option
to lower MinVecRegSize from 128 via a TTI in the SLP Vectorizer.
*** Performance Analysis
This change was motivated by some internal benchmarks but it is also
beneficial on SPEC and the LLVM testsuite.
The results are with -O3 and PGO. A negative percentage is an improvement.
The testsuite was run with a sample size of 4.
** SPEC
* CFP2006/482.sphinx3 -3.34%
A pretty hot loop is SLP vectorized resulting in nice instruction reduction.
This used to be a +22% regression before rL299482.
* CFP2000/177.mesa -3.34%
* CINT2000/256.bzip2 +6.97%
My current plan is to extend the fix in rL299482 to i16 which brings the
regression down to +2.5%. There are also other problems with the codegen in
this loop so there is further room for improvement.
** LLVM testsuite
* SingleSource/Benchmarks/Misc/ReedSolomon -10.75%
There are multiple small SLP vectorizations outside the hot code. It's a bit
surprising that it adds up to 10%. Some of this may be code-layout noise.
* MultiSource/Benchmarks/VersaBench/beamformer/beamformer -8.40%
The opt-viewer screenshot can be seen at F3218284. We start at a colder store
but the tree leads us into the hottest loop.
* MultiSource/Applications/lambda-0.1.3/lambda -2.68%
* MultiSource/Benchmarks/Bullet/bullet -2.18%
This is using 3D vectors.
* SingleSource/Benchmarks/Shootout-C++/Shootout-C++-lists +6.67%
Noise, binary is unchanged.
* MultiSource/Benchmarks/Ptrdist/anagram/anagram +4.90%
There is an additional SLP in the cold code. The test runs for ~1sec and
prints out over 2000 lines. This is most likely noise.
* MultiSource/Applications/aha/aha +1.63%
* MultiSource/Applications/JM/lencod/lencod +1.41%
* SingleSource/Benchmarks/Misc/richards_benchmark +1.15%
Differential Revision: https://reviews.llvm.org/D31965
llvm-svn: 303116
This pass uses a new target hook to decide whether or not to expand a particular
intrinsic to the shuffevector sequence.
Differential Revision: https://reviews.llvm.org/D32245
llvm-svn: 302631
This change is required because the notion of count is different for
sample profiling and getProfileCount will need to determine the
underlying profile type.
Differential revision: https://reviews.llvm.org/D33012
llvm-svn: 302597
- This change allows targets to opt-in to using them instead of the log2
shufflevector algorithm.
- The SLP and Loop vectorizers have the common code to do shuffle reductions
factored out into LoopUtils, and now have a unified interface for generating
reductions regardless of the preference of the target. LoopUtils now uses TTI
to determine what kind of reductions the target wants to handle.
- For CodeGen, basic legalization support is added.
Differential Revision: https://reviews.llvm.org/D30086
llvm-svn: 302514
This introduces a new interface for computeKnownBits that returns the KnownBits object instead of requiring it to be pre-constructed and passed in by reference.
This is a much more convenient interface as it doesn't require the caller to figure out the BitWidth to pre-construct the object. It's so convenient that I believe we can use this interface to remove the special ComputeSignBit flavor of computeKnownBits.
As a step towards that idea, this patch replaces all of the internal usages of ComputeSignBit with this new interface. As you can see from the patch there were a couple places where we called ComputeSignBit which really called computeKnownBits, and then called computeKnownBits again directly. I've reduced those places to only making one call to computeKnownBits. I bet there are probably external users that do it too.
A future patch will update the external users and remove the ComputeSignBit interface. I'll also working on moving more locations to the KnownBits returning interface for computeKnownBits.
Differential Revision: https://reviews.llvm.org/D32848
llvm-svn: 302437
Summary: This makes setRange take ConstantRange by rvalue reference since most callers were passing an unnamed temporary ConstantRange. We can then move that ConstantRange into the DenseMap caches. For the callers that weren't passing a temporary, I've added std::move to to the local variable being passed.
Reviewers: sanjoy, mzolotukhin, efriedma
Reviewed By: sanjoy
Subscribers: takuto.ikuta, llvm-commits
Differential Revision: https://reviews.llvm.org/D32943
llvm-svn: 302371
Summary:
When writing a loop pass I made a mistake and hit the assertion
"Unreachable block in loop". Later, I hit an assertion when I called
`BasicBlock::eraseFromParent()` incorrectly: "Use still stuck around
after Def is destroyed". This latter assertion, however, printed out
exactly which value is being deleted and what uses remain, which helped
me debug the issue.
To help people debugging their loop passes in the future, print out
exactly which basic block is unreachable in a loop.
Reviewers: sanjoy, hfinkel, mehdi_amini
Reviewed By: mehdi_amini
Subscribers: mzolotukhin
Differential Revision: https://reviews.llvm.org/D32878
llvm-svn: 302354