POD-like and we can just splat the empty key across memory.
Sadly we can't optimize the normal loop well enough because we can't
turn the conditional store into an unconditional store according to the
memory model.
This loop actually showed up in a profile of code that was calling clear
as a serious source of time. =[
llvm-svn: 310189
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
Add memory synchronization semantics to LSE Atomics.
The memory semantics feature will be added in a subsequent patch.
In this patch, several corrections were added to the existing LSE Atomics
implementation, based on the ARM Errata D11904 from 05/12/2017.
Patch by: steleman
Differential Revision: https://reviews.llvm.org/D35319
llvm-svn: 310167
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
This extends the native reader to enable llvm-pdbutil to list the enums in a
PDB and it includes a simple test. It does not yet list the values in the
enumerations, which requires an actual implementation of
NativeEnumSymbol::FindChildren.
To exercise this code, use a command like:
llvm-pdbutil pretty -native -enums foo.pdb
Differential Revision: https://reviews.llvm.org/D35738
llvm-svn: 310144
Its sole purpose was to avoid spreading around ifdefs related to
building global-isel. Since r309990, GlobalISel is not optional anymore,
thus, we can get rid of this mechanism all together.
NFC.
llvm-svn: 310115
Summary:
This intrinsic lets us set inactive lanes to an identity value when
implementing wavefront reductions. In combination with Whole Wavefront
Mode, it lets inactive lanes be skipped over as required by GLSL/Vulkan.
Lowering the intrinsic needs to happen post-RA so that RA knows that the
destination isn't completely overwritten due to the EXEC shenanigans, so
we need another pseudo-instruction to represent the un-lowered
intrinsic.
Reviewers: tstellar, arsenm
Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Differential Revision: https://reviews.llvm.org/D34719
llvm-svn: 310088
Summary:
Whole Wavefront Wode (WWM) is similar to WQM, except that all of the
lanes are always enabled, regardless of control flow. This is required
for implementing wavefront reductions in non-uniform control flow, where
we need to use the inactive lanes to propagate intermediate results, so
they need to be enabled. We need to propagate WWM to uses (unless
they're explicitly marked as exact) so that they also propagate
intermediate results correctly. We do the analysis and exec mask munging
during the WQM pass, since there are interactions with WQM for things
that require both WQM and WWM. For simplicity, WWM is entirely
block-local -- blocks are never WWM on entry or exit of a block, and WWM
is not propagated to the block level. This means that computations
involving WWM cannot involve control flow, but we only ever plan to use
WWM for a few limited purposes (none of which involve control flow)
anyways.
Shaders can ask for WWM using the @llvm.amdgcn.wwm intrinsic. There
isn't yet a way to turn WWM off -- that will be added in a future
change.
Finally, it turns out that turning on inactive lanes causes a number of
problems with register allocation. While the best long-term solution
seems like teaching LLVM's register allocator about predication, for now
we need to add some hacks to prevent ourselves from getting into trouble
due to constraints that aren't currently expressed in LLVM. For the gory
details, see the comments at the top of SIFixWWMLiveness.cpp.
Reviewers: arsenm, nhaehnle, tpr
Subscribers: kzhuravl, wdng, mgorny, yaxunl, dstuttard, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D35524
llvm-svn: 310087
Summary:
Previously, we assumed that certain types of instructions needed WQM in
pixel shaders, particularly DS instructions and image sampling
instructions. This was ok because with OpenGL, the assumption was
correct. But we want to start using DPP instructions for derivatives as
well as other things, so the assumption that we can infer whether to use
WQM based on the instruction won't continue to hold. This intrinsic lets
frontends like Mesa indicate what things need WQM based on their
knowledge of the API, rather than second-guessing them in the backend.
We need to keep around the old method of enabling WQM, but eventually we
should remove it once Mesa catches up. For now, this will let us use DPP
instructions for computing derivatives correctly.
Reviewers: arsenm, tpr, nhaehnle
Subscribers: kzhuravl, wdng, yaxunl, dstuttard, llvm-commits, t-tye
Differential Revision: https://reviews.llvm.org/D35167
llvm-svn: 310085
Summary:
It was added to support clang warnings about includes with case
mismatches, but it ended up not being necessary.
Reviewers: twoh, rafael
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D36328
llvm-svn: 310078
Adds function attributes to index: ReadNone, ReadOnly, NoRecurse, NoAlias. This attributes will be used for future ThinLTO optimizations that will propagate function attributes across modules.
llvm-svn: 310061
Summary:
This commit allows matchSelectPattern to recognize clamp of float
arguments in the presence of FMF the same way as already done for
integers.
This case is a little different though. With integers, given the
min/max pattern is recognized, DAGBuilder starts selecting MIN/MAX
"automatically". That is not the case for float, because for them only
full FMINNAN/FMINNUM/FMAXNAN/FMAXNUM ISD nodes exist and they do care
about NaNs. On the other hand, some backends (e.g. X86) have only
FMIN/FMAX nodes that do not care about NaNS and the former NAN/NUM
nodes are illegal thus selection is not happening. So I decided to do
such kind of transformation in IR (InstCombiner) instead of
complicating the logic in the backend.
Reviewers: spatel, jmolloy, majnemer, efriedma, craig.topper
Reviewed By: efriedma
Subscribers: hiraditya, javed.absar, n.bozhenov, llvm-commits
Patch by Andrei Elovikov <andrei.elovikov@intel.com>
Differential Revision: https://reviews.llvm.org/D33186
llvm-svn: 310054
The full story is in the comments:
// Do not attempt to close stdout or stderr. We used to try to maintain the
// property that tools that support writing file to stdout should not also
// write informational output to stdout, but in practice we were never able to
// maintain this invariant. Many features have been added to LLVM and clang
// (-fdump-record-layouts, optimization remarks, etc) that print to stdout, so
// users must simply be aware that mixed output and remarks is a possibility.
NFC, I am just updating comments to reflect reality.
llvm-svn: 310016
This is similar to what we are doing in "regular" SROA and creates
DW_OP_LLVM_fragment operations to describe the resulting variables.
rdar://problem/33654891
llvm-svn: 310014
Summary:
Detect when the working set size of a profiled application is huge,
by comparing the number of counts required to reach the hot percentile
in the profile summary to a large threshold*.
When the working set size is determined to be huge, disable peeling
to avoid bloating the working set further.
*Note that the selected threshold (15K) is significantly larger than the
largest working set value in SPEC cpu2006 (which is gcc at around 11K).
Reviewers: davidxl
Subscribers: mehdi_amini, mzolotukhin, eraman, llvm-commits
Differential Revision: https://reviews.llvm.org/D36288
llvm-svn: 310005
Summary:
This increases the inlining threshold for hot callsites. Hotness is
defined in terms of block frequency of the callsite relative to the
caller's entry block's frequency. Since this requires BFI in the
inliner, this only affects the new PM pipeline. This is enabled by
default at -O3.
This improves the performance of some internal benchmarks. Notably, an
internal benchmark for Gipfeli compression
(https://github.com/google/gipfeli) improves by ~7%. Povray in SPEC2006
improves by ~2.5%. I am running more experiments and will update the
thread if other benchmarks show improvement/regression.
In terms of text size, LLVM test-suite shows an 1.22% text size
increase. Diving into the results, 13 of the benchmarks in the
test-suite increases by > 10%. Most of these are small, but
Adobe-C++/loop_unroll (17.6% increases) and tramp3d(20.7% size increase)
have >250K text size. On a large application, the text size increases by
2%
Reviewers: chandlerc, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36199
llvm-svn: 309994
Summary:
PDB section contributions are supposed to use output section indices and
offsets, not input section indices and offsets.
This allows the debugger to look up the index of the module that it
should look up in the modules stream for symbol information. With this
change, windbg can now find line tables, but it still cannot print local
variables.
Fixes PR34048
Reviewers: zturner
Subscribers: hiraditya, ruiu, llvm-commits
Differential Revision: https://reviews.llvm.org/D36285
llvm-svn: 309987
Summary:
Peeling should not occur during the full unrolling invocation early
in the pipeline, but rather later with partial and runtime loop
unrolling. The later loop unrolling invocation will also eventually
utilize profile summary and branch frequency information, which
we would like to use to control peeling. And for ThinLTO we want
to delay peeling until the backend (post thin link) phase, just as
we do for most types of unrolling.
Ensure peeling doesn't occur during the full unrolling invocation
by adding a parameter to the shared implementation function, similar
to the way partial and runtime loop unrolling are disabled.
Performance results for ThinLTO suggest this has a neutral to positive
effect on some internal benchmarks.
Reviewers: chandlerc, davidxl
Subscribers: mzolotukhin, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D36258
llvm-svn: 309966
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
IMHO it is an antipattern to have a enum value that is Default.
At any given piece of code it is not clear if we have to handle
Default or if has already been mapped to a concrete value. In this
case in particular, only the target can do the mapping and it is nice
to make sure it is always done.
This deletes the two default enum values of CodeModel and uses an
explicit Optional<CodeModel> when it is possible that it is
unspecified.
llvm-svn: 309911
The CoverageMapping::getInstantiations() API retrieved all function
records corresponding to functions with more than one instantiation (e.g
template functions with multiple specializations). However, there was no
simple way to determine *which* function a given record was an
instantiation of. This was an oversight, since it's useful to aggregate
coverage information over all instantiations of a function.
llvm-cov works around this by building a mapping of source locations to
instantiation sets, but this duplicates logic that libCoverage already
has (see FunctionInstantiationSetCollector).
This change adds a new API, CoverageMapping::getInstantiationGroups(),
which returns a list of InstantiationGroups. A group contains records
for each instantiation of some particular function, and also provides
utilities to get the total execution count within the group, the source
location of the common definition, etc.
This lets removes some hacky logic in llvm-cov by reusing
FunctionInstantiationSetCollector and makes the CoverageMapping API
friendlier for other clients.
llvm-svn: 309904
The PDB reserves certain blocks for the FPM that describe which
blocks in the file are allocated and which are free. We weren't
filling that out at all, and in some cases we were even stomping
it with incorrect data. This patch writes a correct FPM.
Differential Revision: https://reviews.llvm.org/D36235
llvm-svn: 309896
Recently problems have been discovered in the way we write the FPM
(free page map). In order to fix this, we first need to establish
a baseline about what a correct FPM looks like using an MSVC
generated PDB, so that we can then make our own generated PDBs
match. And in order to do this, the dumper needs a mode where it
can dump an FPM so that we can write tests for it.
This patch adds a command to dump the FPM, as well as a test against
a known-good PDB.
llvm-svn: 309894
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
I was surprised to see the code model being passed to MC. After all,
it assembles code, it doesn't create it.
The one place it is used is in the expansion of .cfi directives to
handle .eh_frame being more that 2gb away from the code.
As far as I can tell, gnu assembler doesn't even have an option to
enable this. Compiling a c file with gcc -mcmodel=large produces a
regular looking .eh_frame. This is probably because in practice linker
parse and recreate .eh_frames.
In llvm this is used because the JIT can place the code and .eh_frame
very far apart. Ideally we would fix the jit and delete this
option. This is hard.
Apart from confusion another problem with the current interface is
that most callers pass CodeModel::Default, which is bad since MC has
no way to map it to the target default if it actually needed to.
This patch then replaces the argument with a boolean with a default
value. The vast majority of users don't ever need to look at it. In
fact, only CodeGen and llvm-mc use it and llvm-mc just to enable more
testing.
llvm-svn: 309884
Followup to r309570, fixing it slightly differently (ranges_base and
addr_base should never be read from a DWO file - so there shouldn't be
any issue with 'overriding' the values - conditionalize the code and
assert that the values aren't being overriden).
llvm-svn: 309879
Summary:
This patch makes LoopDeletion use the incremental DominatorTree API.
We modify LoopDeletion to perform the deletion in 5 steps:
1. Create a new dummy edge from the preheader to the exit, by adding a conditional branch.
2. Inform the DomTree about the new edge.
3. Remove the conditional branch and replace it with an unconditional edge to the exit. This removes the edge to the loop header, making it unreachable.
4. Inform the DomTree about the deleted edge.
5. Remove the unreachable block from the function.
Creating the dummy conditional branch is necessary to perform incremental DomTree update.
We should consider using the batch updater when it's ready.
Reviewers: dberlin, davide, grosser, sanjoy
Reviewed By: dberlin, grosser
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D35391
llvm-svn: 309850
This should enable us to test the generation of target-specific constant
pools, e.g. for ARM:
constants:
- id: 0
value: 'g(GOT_PREL)-(LPC0+8-.)'
alignment: 4
isTargetSpecific: true
I intend to use this to test PIC support in GlobalISel for ARM.
This is difficult to test outside of that context, since the existing
MIR tests usually rely on parser support as well, and that seems a bit
trickier to add. We could try to add a unit test, but the setup for that
seems rather convoluted and overkill.
We do test however that the parser reports a nice error when
encountering a target-specific constant pool.
Differential Revision: https://reviews.llvm.org/D36092
llvm-svn: 309806
Summary:
Fix a bug discovered in an out-of-tree target where memoperands from
pseudo-instructions that weren't part of the match were being merged into the
result instructions as part of GIR_MergeMemOperands.
This bug was caused by a change to the handling of State.MIs between rules when
the state machine tables were fused into a single table. Previously, each rule
would reset State.MIs using State.MIs.resize(1) but this is no longer done, as a
result stale data is occasionally left in some elements of State.MIs. Most
opcodes aren't affected by this but GIR_MergeMemOperands merges all memoperands
from the intructions recorded in State.MIs into the result instruction.
Suppose for example, we processed but rejected the following pattern:
(signextend (load x))
at this point, State.MIs contains the signextend and the load. Now suppose we
process and accept this pattern:
(add x, y)
at this point, State.MIs contains the add as well as the (now irrelevant) load.
When GIR_MergeMemOperands is processed, the memoperands from that irrelevant
load will be merged into the result instruction even though it was not part of
the match.
Bringing back the State.MIs.resize(1) would fix the problem but it would limit
our ability to optimize the table in the future. Instead, this patch fixes the
problem by explicitly stating which instructions should be merged into the result.
There's no direct test case in this commit because a test case would be very brittle.
However, at the time of writing this should fix the failures in
http://green.lab.llvm.org/green/job/Compiler_Verifiers_GlobalISEL/ as well as a
failure in test/CodeGen/ARM/GlobalISel/arm-isel.ll when expensive checks are enabled.
Reviewers: ab, t.p.northover, qcolombet, rovka, aditya_nandakumar
Subscribers: fhahn, kristof.beyls, igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D36094
llvm-svn: 309804
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
In the last half-dozen commits to LLVM I removed code that became dead
after removing the offset parameter from llvm.dbg.value gradually
proceeding from IR towards the backend. Before I can move on to
DwarfDebug and friends there is one last side-called offset I need to
remove: This patch modifies PrologEpilogInserter's use of the
DBG_VALUE's offset argument to use a DIExpression instead. Because the
PrologEpilogInserter runs at the Machine level I had to play a little
trick with a named llvm.dbg.mir node to get the DIExpressions to print
in MIR dumps (which print the llvm::Module followed by the
MachineFunction dump).
I also had to add rudimentary DwarfExpression support to CodeView and
as a side-effect also fixed a bug (CodeViewDebug::collectVariableInfo
was supposed to give up on variables with complex DIExpressions, but
would fail to do so for fragments, which are also modeled as
DIExpressions).
With this last holdover removed we will have only one canonical way of
representing offsets to debug locations which will simplify the code
in DwarfDebug (and future versions of CodeViewDebug once it starts
handling more complex expressions) and make it easier to reason about.
This patch is NFC-ish: All test case changes are for assembler
comments and the binary output does not change.
rdar://problem/33580047
Differential Revision: https://reviews.llvm.org/D36125
llvm-svn: 309751
Chandler Carruth