When multiple loop transformation are defined in a loop's metadata, their order of execution is defined by the order of their respective passes in the pass pipeline. For instance, e.g.
#pragma clang loop unroll_and_jam(enable)
#pragma clang loop distribute(enable)
is the same as
#pragma clang loop distribute(enable)
#pragma clang loop unroll_and_jam(enable)
and will try to loop-distribute before Unroll-And-Jam because the LoopDistribute pass is scheduled after UnrollAndJam pass. UnrollAndJamPass only supports one inner loop, i.e. it will necessarily fail after loop distribution. It is not possible to specify another execution order. Also,t the order of passes in the pipeline is subject to change between versions of LLVM, optimization options and which pass manager is used.
This patch adds 'followup' attributes to various loop transformation passes. These attributes define which attributes the resulting loop of a transformation should have. For instance,
!0 = !{!0, !1, !2}
!1 = !{!"llvm.loop.unroll_and_jam.enable"}
!2 = !{!"llvm.loop.unroll_and_jam.followup_inner", !3}
!3 = !{!"llvm.loop.distribute.enable"}
defines a loop ID (!0) to be unrolled-and-jammed (!1) and then the attribute !3 to be added to the jammed inner loop, which contains the instruction to distribute the inner loop.
Currently, in both pass managers, pass execution is in a fixed order and UnrollAndJamPass will not execute again after LoopDistribute. We hope to fix this in the future by allowing pass managers to run passes until a fixpoint is reached, use Polly to perform these transformations, or add a loop transformation pass which takes the order issue into account.
For mandatory/forced transformations (e.g. by having been declared by #pragma omp simd), the user must be notified when a transformation could not be performed. It is not possible that the responsible pass emits such a warning because the transformation might be 'hidden' in a followup attribute when it is executed, or it is not present in the pipeline at all. For this reason, this patche introduces a WarnMissedTransformations pass, to warn about orphaned transformations.
Since this changes the user-visible diagnostic message when a transformation is applied, two test cases in the clang repository need to be updated.
To ensure that no other transformation is executed before the intended one, the attribute `llvm.loop.disable_nonforced` can be added which should disable transformation heuristics before the intended transformation is applied. E.g. it would be surprising if a loop is distributed before a #pragma unroll_and_jam is applied.
With more supported code transformations (loop fusion, interchange, stripmining, offloading, etc.), transformations can be used as building blocks for more complex transformations (e.g. stripmining+stripmining+interchange -> tiling).
Reviewed By: hfinkel, dmgreen
Differential Revision: https://reviews.llvm.org/D49281
Differential Revision: https://reviews.llvm.org/D55288
llvm-svn: 348944
Summary:
It turns out that we need an OptimizerLast PassBuilder extension point
after all. I missed the relevance of this EP the first time. By legacy PM magic,
function passes added at this EP get added to the last _Function_ PM, which is a
feature we lost when dropping this EP for the new PM.
A key difference between this and the legacy PassManager's OptimizerLast
callback is that this extension point is not triggered at O0. Extensions
to the O0 pipeline should append their passes to the end of the overall
pipeline.
Differential Revision: https://reviews.llvm.org/D54374
llvm-svn: 346645
This reverts commit r342387 as it's showing significant performance
regressions in a number of benchmarks. Followed up with the
committer and original thread with an example and will get performance
numbers before recommitting.
llvm-svn: 343522
Pass Execution Instrumentation interface enables customizable instrumentation
of pass execution, as per "RFC: Pass Execution Instrumentation interface"
posted 06/07/2018 on llvm-dev@
The intent is to provide a common machinery to implement all
the pass-execution-debugging features like print-before/after,
opt-bisect, time-passes etc.
Here we get a basic implementation consisting of:
* PassInstrumentationCallbacks class that handles registration of callbacks
and access to them.
* PassInstrumentation class that handles instrumentation-point interfaces
that call into PassInstrumentationCallbacks.
* Callbacks accept StringRef which is just a name of the Pass right now.
There were some ideas to pass an opaque wrapper for the pointer to pass instance,
however it appears that pointer does not actually identify the instance
(adaptors and managers might have the same address with the pass they govern).
Hence it was decided to go simple for now and then later decide on what the proper
mental model of identifying a "pass in a phase of pipeline" is.
* Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies
on different IRUnits (e.g. Analyses).
* PassInstrumentationAnalysis analysis is explicitly requested from PassManager through
usual AnalysisManager::getResult. All pass managers were updated to run that
to get PassInstrumentation object for instrumentation calls.
* Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra
args out of a generic PassManager's extra args. This is the only way I was able to explicitly
run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or
RepeatedPass::run.
TODO: Upon lengthy discussions we agreed to accept this as an initial implementation
and then get rid of getAnalysisResult by improving RepeatedPass implementation.
* PassBuilder takes PassInstrumentationCallbacks object to pass it further into
PassInstrumentationAnalysis. Callbacks registration should be performed directly
through PassInstrumentationCallbacks.
* new-pm tests updated to account for PassInstrumentationAnalysis being run
* Added PassInstrumentation tests to PassBuilderCallbacks unit tests.
Other unit tests updated with registration of the now-required PassInstrumentationAnalysis.
Made getName helper to return std::string (instead of StringRef initially) to fix
asan builtbot failures on CGSCC tests.
Reviewers: chandlerc, philip.pfaffe
Differential Revision: https://reviews.llvm.org/D47858
llvm-svn: 342664
Pass Execution Instrumentation interface enables customizable instrumentation
of pass execution, as per "RFC: Pass Execution Instrumentation interface"
posted 06/07/2018 on llvm-dev@
The intent is to provide a common machinery to implement all
the pass-execution-debugging features like print-before/after,
opt-bisect, time-passes etc.
Here we get a basic implementation consisting of:
* PassInstrumentationCallbacks class that handles registration of callbacks
and access to them.
* PassInstrumentation class that handles instrumentation-point interfaces
that call into PassInstrumentationCallbacks.
* Callbacks accept StringRef which is just a name of the Pass right now.
There were some ideas to pass an opaque wrapper for the pointer to pass instance,
however it appears that pointer does not actually identify the instance
(adaptors and managers might have the same address with the pass they govern).
Hence it was decided to go simple for now and then later decide on what the proper
mental model of identifying a "pass in a phase of pipeline" is.
* Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies
on different IRUnits (e.g. Analyses).
* PassInstrumentationAnalysis analysis is explicitly requested from PassManager through
usual AnalysisManager::getResult. All pass managers were updated to run that
to get PassInstrumentation object for instrumentation calls.
* Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra
args out of a generic PassManager's extra args. This is the only way I was able to explicitly
run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or
RepeatedPass::run.
TODO: Upon lengthy discussions we agreed to accept this as an initial implementation
and then get rid of getAnalysisResult by improving RepeatedPass implementation.
* PassBuilder takes PassInstrumentationCallbacks object to pass it further into
PassInstrumentationAnalysis. Callbacks registration should be performed directly
through PassInstrumentationCallbacks.
* new-pm tests updated to account for PassInstrumentationAnalysis being run
* Added PassInstrumentation tests to PassBuilderCallbacks unit tests.
Other unit tests updated with registration of the now-required PassInstrumentationAnalysis.
Reviewers: chandlerc, philip.pfaffe
Differential Revision: https://reviews.llvm.org/D47858
llvm-svn: 342597
Summary:
Pass Execution Instrumentation interface enables customizable instrumentation
of pass execution, as per "RFC: Pass Execution Instrumentation interface"
posted 06/07/2018 on llvm-dev@
The intent is to provide a common machinery to implement all
the pass-execution-debugging features like print-before/after,
opt-bisect, time-passes etc.
Here we get a basic implementation consisting of:
* PassInstrumentationCallbacks class that handles registration of callbacks
and access to them.
* PassInstrumentation class that handles instrumentation-point interfaces
that call into PassInstrumentationCallbacks.
* Callbacks accept StringRef which is just a name of the Pass right now.
There were some ideas to pass an opaque wrapper for the pointer to pass instance,
however it appears that pointer does not actually identify the instance
(adaptors and managers might have the same address with the pass they govern).
Hence it was decided to go simple for now and then later decide on what the proper
mental model of identifying a "pass in a phase of pipeline" is.
* Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies
on different IRUnits (e.g. Analyses).
* PassInstrumentationAnalysis analysis is explicitly requested from PassManager through
usual AnalysisManager::getResult. All pass managers were updated to run that
to get PassInstrumentation object for instrumentation calls.
* Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra
args out of a generic PassManager's extra args. This is the only way I was able to explicitly
run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or
RepeatedPass::run.
TODO: Upon lengthy discussions we agreed to accept this as an initial implementation
and then get rid of getAnalysisResult by improving RepeatedPass implementation.
* PassBuilder takes PassInstrumentationCallbacks object to pass it further into
PassInstrumentationAnalysis. Callbacks registration should be performed directly
through PassInstrumentationCallbacks.
* new-pm tests updated to account for PassInstrumentationAnalysis being run
* Added PassInstrumentation tests to PassBuilderCallbacks unit tests.
Other unit tests updated with registration of the now-required PassInstrumentationAnalysis.
Reviewers: chandlerc, philip.pfaffe
Differential Revision: https://reviews.llvm.org/D47858
llvm-svn: 342544
Rebase rL341954 since https://bugs.llvm.org/show_bug.cgi?id=38912
has been fixed by rL342055.
Precommit testing performed:
* Overnight runs of csmith comparing the output between programs
compiled with gvn-hoist enabled/disabled.
* Bootstrap builds of clang with UbSan/ASan configurations.
llvm-svn: 342387
This reverts rL341954.
The builder `sanitizer-x86_64-linux-bootstrap-ubsan` has been
failing with timeouts at stage2 clang/ubsan:
[3065/3073] Linking CXX executable bin/lld
command timed out: 1200 seconds without output running python
../sanitizer_buildbot/sanitizers/buildbot_selector.py,
attempting to kill
llvm-svn: 342001
Rebase rL338240 since the excessive memory usage observed when using
GVNHoist with UBSan has been fixed by rL340818.
Differential Revision: https://reviews.llvm.org/D49858
llvm-svn: 340922
This is being done in order to make GVN able to better optimize certain inputs.
MemDep doesn't use PhiValues directly, but does need to notifiy it when things
get invalidated.
Differential Revision: https://reviews.llvm.org/D48489
llvm-svn: 338384
and diretory.
Also cleans up all the associated naming to be consistent and removes
the public access to the pass ID which was unused in LLVM.
Also runs clang-format over parts that changed, which generally cleans
up a bunch of formatting.
This is in preparation for doing some internal cleanups to the pass.
Differential Revision: https://reviews.llvm.org/D47352
llvm-svn: 336028
=== Generating the CG Profile ===
The CGProfile module pass simply gets the block profile count for each BB and scans for call instructions. For each call instruction it adds an edge from the current function to the called function with the current BB block profile count as the weight.
After scanning all the functions, it generates an appending module flag containing the data. The format looks like:
```
!llvm.module.flags = !{!0}
!0 = !{i32 5, !"CG Profile", !1}
!1 = !{!2, !3, !4} ; List of edges
!2 = !{void ()* @a, void ()* @b, i64 32} ; Edge from a to b with a weight of 32
!3 = !{void (i1)* @freq, void ()* @a, i64 11}
!4 = !{void (i1)* @freq, void ()* @b, i64 20}
```
Differential Revision: https://reviews.llvm.org/D48105
llvm-svn: 335794
loop-cleanup passes at the beginning of the loop pass pipeline, and
re-enqueue loops after even trivial unswitching.
This will allow us to much more consistently avoid simplifying code
while doing trivial unswitching. I've also added a test case that
specifically shows effective iteration using this technique.
I've unconditionally updated the new PM as that is always using the
SimpleLoopUnswitch pass, and I've made the pipeline changes for the old
PM conditional on using this new unswitch pass. I added a bunch of
comments to the loop pass pipeline in the old PM to make it more clear
what is going on when reviewing.
Hopefully this will unblock doing *partial* unswitching instead of just
full unswitching.
Differential Revision: https://reviews.llvm.org/D47408
llvm-svn: 333493
Summary: Makes this consistent with the old PM.
Reviewers: eraman
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D46526
llvm-svn: 331709
Combine expression patterns to form expressions with fewer, simple instructions.
This pass does not modify the CFG.
For example, this pass reduce width of expressions post-dominated by TruncInst
into smaller width when applicable.
It differs from instcombine pass in that it contains pattern optimization that
requires higher complexity than the O(1), thus, it should run fewer times than
instcombine pass.
Differential Revision: https://reviews.llvm.org/D38313
llvm-svn: 323321
This applies to most pipelines except the LTO and ThinLTO backend
actions - it is for use at the beginning of the overall pipeline.
This extension point will be used to add the GCOV pass when enabled in
Clang.
llvm-svn: 323166
Summary:
New pass manager driver passes DebugPM (-debug-pass-manager) flag into
individual PassManager constructors in order to enable debug logging.
FunctionToLoopPassAdaptor has its own internal LoopCanonicalizationPM
which never gets its debug logging enabled and that means canonicalization
passes like LoopSimplify are never present in -debug-pass-manager output.
Extending FunctionToLoopPassAdaptor's constructor and
createFunctionToLoopPassAdaptor wrapper with an optional
boolean DebugLogging argument.
Passing debug-logging flags there as appropriate.
Reviewers: chandlerc, davide
Reviewed By: davide
Subscribers: mehdi_amini, eraman, llvm-commits, JDevlieghere
Differential Revision: https://reviews.llvm.org/D41586
llvm-svn: 321548
This should solve:
https://bugs.llvm.org/show_bug.cgi?id=34603
...by preventing SimplifyCFG from altering redundant instructions before early-cse has a chance to run.
It changes the default (canonical-forming) behavior of SimplifyCFG, so we're only doing the
sinking transform later in the optimization pipeline.
Differential Revision: https://reviews.llvm.org/D38566
llvm-svn: 320749
Summary:
Passing AliasAnalysis results instead of nullptr appears to work just fine.
A couple new-pass-manager tests updated to align with new order of analyses.
Reviewers: chandlerc, spatel, craig.topper
Reviewed By: chandlerc
Subscribers: mehdi_amini, eraman, llvm-commits
Differential Revision: https://reviews.llvm.org/D41203
llvm-svn: 320687
The core idea is to (re-)introduce some redundancies where their cost is
hidden by the cost of materializing immediates for constant operands of
PHI nodes. When the cost of the redundancies is covered by this,
avoiding materializing the immediate has numerous benefits:
1) Less register pressure
2) Potential for further folding / combining
3) Potential for more efficient instructions due to immediate operand
As a motivating example, consider the remarkably different cost on x86
of a SHL instruction with an immediate operand versus a register
operand.
This pattern turns up surprisingly frequently, but is somewhat rarely
obvious as a significant performance problem.
The pass is entirely target independent, but it does rely on the target
cost model in TTI to decide when to speculate things around the PHI
node. I've included x86-focused tests, but any target that sets up its
immediate cost model should benefit from this pass.
There is probably more that can be done in this space, but the pass
as-is is enough to get some important performance on our internal
benchmarks, and should be generally performance neutral, but help with
more extensive benchmarking is always welcome.
One awkward part is that this pass has to be scheduled after
*everything* that can eliminate these kinds of redundancies. This
includes SimplifyCFG, GVN, etc. I'm open to suggestions about better
places to put this. We could in theory make it part of the codegen pass
pipeline, but there doesn't really seem to be a good reason for that --
it isn't "lowering" in any sense and only relies on pretty standard cost
model based TTI queries, so it seems to fit well with the "optimization"
pipeline model. Still, further thoughts on the pipeline position are
welcome.
I've also only implemented this in the new pass manager. If folks are
very interested, I can try to add it to the old PM as well, but I didn't
really see much point (my use case is already switched over to the new
PM).
I've tested this pretty heavily without issue. A wide range of
benchmarks internally show no change outside the noise, and I don't see
any significant changes in SPEC either. However, the size class
computation in tcmalloc is substantially improved by this, which turns
into a 2% to 4% win on the hottest path through tcmalloc for us, so
there are definitely important cases where this is going to make
a substantial difference.
Differential revision: https://reviews.llvm.org/D37467
llvm-svn: 319164
This recommit r317351 after fixing a buildbot failure.
Original commit message:
Summary:
This change add a pass which tries to split a call-site to pass
more constrained arguments if its argument is predicated in the control flow
so that we can expose better context to the later passes (e.g, inliner, jump
threading, or IPA-CP based function cloning, etc.).
As of now we support two cases :
1) If a call site is dominated by an OR condition and if any of its arguments
are predicated on this OR condition, try to split the condition with more
constrained arguments. For example, in the code below, we try to split the
call site since we can predicate the argument (ptr) based on the OR condition.
Split from :
if (!ptr || c)
callee(ptr);
to :
if (!ptr)
callee(null ptr) // set the known constant value
else if (c)
callee(nonnull ptr) // set non-null attribute in the argument
2) We can also split a call-site based on constant incoming values of a PHI
For example,
from :
BB0:
%c = icmp eq i32 %i1, %i2
br i1 %c, label %BB2, label %BB1
BB1:
br label %BB2
BB2:
%p = phi i32 [ 0, %BB0 ], [ 1, %BB1 ]
call void @bar(i32 %p)
to
BB0:
%c = icmp eq i32 %i1, %i2
br i1 %c, label %BB2-split0, label %BB1
BB1:
br label %BB2-split1
BB2-split0:
call void @bar(i32 0)
br label %BB2
BB2-split1:
call void @bar(i32 1)
br label %BB2
BB2:
%p = phi i32 [ 0, %BB2-split0 ], [ 1, %BB2-split1 ]
llvm-svn: 317362
Summary:
This change add a pass which tries to split a call-site to pass
more constrained arguments if its argument is predicated in the control flow
so that we can expose better context to the later passes (e.g, inliner, jump
threading, or IPA-CP based function cloning, etc.).
As of now we support two cases :
1) If a call site is dominated by an OR condition and if any of its arguments
are predicated on this OR condition, try to split the condition with more
constrained arguments. For example, in the code below, we try to split the
call site since we can predicate the argument (ptr) based on the OR condition.
Split from :
if (!ptr || c)
callee(ptr);
to :
if (!ptr)
callee(null ptr) // set the known constant value
else if (c)
callee(nonnull ptr) // set non-null attribute in the argument
2) We can also split a call-site based on constant incoming values of a PHI
For example,
from :
BB0:
%c = icmp eq i32 %i1, %i2
br i1 %c, label %BB2, label %BB1
BB1:
br label %BB2
BB2:
%p = phi i32 [ 0, %BB0 ], [ 1, %BB1 ]
call void @bar(i32 %p)
to
BB0:
%c = icmp eq i32 %i1, %i2
br i1 %c, label %BB2-split0, label %BB1
BB1:
br label %BB2-split1
BB2-split0:
call void @bar(i32 0)
br label %BB2
BB2-split1:
call void @bar(i32 1)
br label %BB2
BB2:
%p = phi i32 [ 0, %BB2-split0 ], [ 1, %BB2-split1 ]
Reviewers: davidxl, huntergr, chandlerc, mcrosier, eraman, davide
Reviewed By: davidxl
Subscribers: sdesmalen, ashutosh.nema, fhahn, mssimpso, aemerson, mgorny, mehdi_amini, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D39137
llvm-svn: 317351
This patch adds a new pass for attaching !callees metadata to indirect call
sites. The pass propagates values to call sites by performing an IPSCCP-like
analysis using the generic sparse propagation solver. For indirect call sites
having a small set of possible callees, the attached metadata indicates what
those callees are. The metadata can be used to facilitate optimizations like
intersecting the function attributes of the possible callees, refining the call
graph, performing indirect call promotion, etc.
Differential Revision: https://reviews.llvm.org/D37355
llvm-svn: 316576
This is the same exact change we did for the current pass manager
in rL314997, but the new pass manager pipeline already happened
to run GlobalOpt after the inliner, so we just insert a run of
GDCE here.
llvm-svn: 315003
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
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 is largely NFC*, in preparation for utilizing ProfileSummaryInfo
and BranchFrequencyInfo analyses. In this patch I am only doing the
splitting for the New PM, but I can do the same for the legacy PM as
a follow-on if this looks good.
*Not NFC since for partial unrolling we lose the updates done to the
loop traversal (adding new sibling and child loops) - according to
Chandler this is not very useful for partial unrolling, but it also
means that the debugging flag -unroll-revisit-child-loops no longer
works for partial unrolling.
Reviewers: chandlerc
Subscribers: mehdi_amini, mzolotukhin, eraman, llvm-commits
Differential Revision: https://reviews.llvm.org/D36157
llvm-svn: 309886
Summary:
This changes SimplifyLibCalls to use the new OptimizationRemarkEmitter
API.
In fact, as SimplifyLibCalls is only ever called via InstCombine,
(as far as I can tell) the OptimizationRemarkEmitter is added there,
and then passed through to SimplifyLibCalls later.
I have avoided changing any remark text.
This closes PR33787
Patch by Sam Elliott!
Reviewers: anemet, davide
Reviewed By: anemet
Subscribers: davide, mehdi_amini, eraman, fhahn, llvm-commits
Differential Revision: https://reviews.llvm.org/D35608
llvm-svn: 309158
Summary:
This patch adds a callback registration API to the PassBuilder,
enabling registering out-of-tree passes with it.
Through the Callback API, callers may register callbacks with the
various stages at which passes are added into pass managers, including
parsing of a pass pipeline as well as at extension points within the
default -O pipelines.
Registering utilities like `require<>` and `invalidate<>` needs to be
handled manually by the caller, but a helper is provided.
Additionally, adding passes at pipeline extension points is exposed
through the opt tool. This patch adds a `-passes-ep-X` commandline
option for every extension point X, which opt parses into pipelines
inserted into that extension point.
Reviewers: chandlerc
Reviewed By: chandlerc
Subscribers: lksbhm, grosser, davide, mehdi_amini, llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D33464
llvm-svn: 307532
Based on the original patch by Davide, but I've adjusted the API exposed
to just be different entry points rather than exposing more state
parameters. I've factored all the common logic out so that we don't have
any duplicate pipelines, we just stitch them together in different ways.
I think this makes the build easier to reason about and understand.
This adds a direct method for getting the module simplification pipeline
as well as a method to get the optimization pipeline. While not my
express goal, this seems nice and gives a good place comment about the
restrictions that are imposed on them.
I did make some minor changes to the way the pipelines are structured
here, but hopefully not ones that are significant or controversial:
1) I sunk the PGO indirect call promotion to only be run when we have
PGO enabled (or as part of the special ThinLTO pipeline).
2) I made the extra GlobalOpt run in ThinLTO just happen all the time
and at a slightly more powerful place (before we remove available
externaly functions). This seems like general goodness and not a big
compile time sink, so it didn't make sense to *only* use it in
ThinLTO. Fewer differences in the pipeline makes everything simpler
IMO.
3) I hoisted the ThinLTO stop point pre-link above the the RPO function
attr inference. The RPO inference won't infer anything terribly
meaningful pre-link (recursiveness?) so it didn't make a lot of
sense. But if the placement of RPO inference starts to matter, we
should move it to the canonicalization phase anyways which seems like
a better place for it (and there is a FIXME to this effect!). But
that seemed a bridge too far for this patch.
If we ever need to parameterize these pipelines more heavily, we can
always sink the logic to helper functions with parameters to keep those
parameters out of the public API. But the changes above seemed minor
that we could possible get away without the parameters entirely.
I added support for parsing 'thinlto' and 'thinlto-pre-link' names in
pass pipelines to make it easy to test these routines and play with them
in larger pipelines. I also added a really basic manifest of passes test
that will show exactly how the pipelines behave and work as well as
making updates to them clear.
Lastly, this factoring does introduce a nesting layer of module pass
managers in the default pipeline. I don't think this is a big deal and
the flexibility of decoupling the pipelines seems easily worth it.
Differential Revision: https://reviews.llvm.org/D33540
llvm-svn: 304407
also a discussion about exactly what we should do prior to re-enabling
it.
The current bug is http://llvm.org/PR32821 and the discussion about this
is in the review thread for r300200.
llvm-svn: 301505
Summary:
Otherwise we might end up with some empty basic blocks or
single-entry-single-exit basic blocks.
This fixes PR32085
Reviewers: chandlerc, danielcdh
Subscribers: mehdi_amini, RKSimon, llvm-commits
Differential Revision: https://reviews.llvm.org/D30468
llvm-svn: 301395
entire SCC before iterating on newly-introduced call edges resulting
from any inlined function bodies.
This more closely matches the behavior of the old PM's inliner. While it
wasn't really clear to me initially, this behavior is actually essential
to the inliner behaving reasonably in its current design.
Because the inliner is fundamentally a bottom-up inliner and all of its
cost modeling is designed around that it often runs into trouble within
an SCC where we don't have any meaningful bottom-up ordering to use. In
addition to potentially cyclic, infinite inlining that we block with the
inline history mechanism, it can also take seemingly simple call graph
patterns within an SCC and turn them into *insanely* large functions by
accidentally working top-down across the SCC without any of the
threshold limitations that traditional top-down inliners use.
Consider this diabolical monster.cpp file that Richard Smith came up
with to help demonstrate this issue:
```
template <int N> extern const char *str;
void g(const char *);
template <bool K, int N> void f(bool *B, bool *E) {
if (K)
g(str<N>);
if (B == E)
return;
if (*B)
f<true, N + 1>(B + 1, E);
else
f<false, N + 1>(B + 1, E);
}
template <> void f<false, MAX>(bool *B, bool *E) { return f<false, 0>(B, E); }
template <> void f<true, MAX>(bool *B, bool *E) { return f<true, 0>(B, E); }
extern bool *arr, *end;
void test() { f<false, 0>(arr, end); }
```
When compiled with '-DMAX=N' for various values of N, this will create an SCC
with a reasonably large number of functions. Previously, the inliner would try
to exhaust the inlining candidates in a single function before moving on. This,
unfortunately, turns it into a top-down inliner within the SCC. Because our
thresholds were never built for that, we will incrementally decide that it is
always worth inlining and proceed to flatten the entire SCC into that one
function.
What's worse, we'll then proceed to the next function, and do the exact same
thing except we'll skip the first function, and so on. And at each step, we'll
also make some of the constant factors larger, which is awesome.
The fix in this patch is the obvious one which makes the new PM's inliner use
the same technique used by the old PM: consider all the call edges across the
entire SCC before beginning to process call edges introduced by inlining. The
result of this is essentially to distribute the inlining across the SCC so that
every function incrementally grows toward the inline thresholds rather than
allowing the inliner to grow one of the functions vastly beyond the threshold.
The code for this is a bit awkward, but it works out OK.
We could consider in the future doing something more powerful here such as
prioritized order (via lowest cost and/or profile info) and/or a code-growth
budget per SCC. However, both of those would require really substantial work
both to design the system in a way that wouldn't break really useful
abstraction decomposition properties of the current inliner and to be tuned
across a reasonably diverse set of code and workloads. It also seems really
risky in many ways. I have only found a single real-world file that triggers
the bad behavior here and it is generated code that has a pretty pathological
pattern. I'm not worried about the inliner not doing an *awesome* job here as
long as it does *ok*. On the other hand, the cases that will be tricky to get
right in a prioritized scheme with a budget will be more common and idiomatic
for at least some frontends (C++ and Rust at least). So while these approaches
are still really interesting, I'm not in a huge rush to go after them. Staying
even closer to the existing PM's behavior, especially when this easy to do,
seems like the right short to medium term approach.
I don't really have a test case that makes sense yet... I'll try to find a
variant of the IR produced by the monster template metaprogram that is both
small enough to be sane and large enough to clearly show when we get this wrong
in the future. But I'm not confident this exists. And the behavior change here
*should* be unobservable without snooping on debug logging. So there isn't
really much to test.
The test case updates come from two incidental changes:
1) We now visit functions in an SCC in the opposite order. I don't think there
really is a "right" order here, so I just update the test cases.
2) We no longer compute some analyses when an SCC has no call instructions that
we consider for inlining.
llvm-svn: 297374
All the invalidation issues and bugs in this seem to be fixed, it has
survived a full build of the test suite plus SPEC with asserts and ASan
enabled on the Clang binary used.
Differential Revision: https://reviews.llvm.org/D29815
llvm-svn: 294887
Michael Kruse