Summary:
We don't use MinLatency any more since r184032.
Reviewers: atrick, hfinkel, mcrosier
Differential Revision: http://reviews.llvm.org/D19474
llvm-svn: 267502
Pass all of the state we need around as arguments, so that these
functions are easier to reuse. There is one part of this that is
unusual: we pass around a functor to look up a DomTree for a function.
This will be a necessary abstraction when we try to use this code in
both the legacy and the new pass manager.
llvm-svn: 267498
Kill-flags, which computeRegisterLiveness uses, are not reliable.
LivePhysRegs is.
Differential Revision: http://reviews.llvm.org/D19472
llvm-svn: 267495
The SparcV8 fneg and fabs instructions interestingly come only in a
single-float variant. Since the sign bit is always the topmost bit no
matter what size float it is, you simply operate on the high
subregister, as if it were a single float.
However, the layout of double-floats in the float registers is reversed
on little-endian CPUs, so that the high bits are in the second
subregister, rather than the first.
Thus, this expansion must check the endianness to use the correct
subregister.
llvm-svn: 267489
This patch is what was the "instcombine" portion of D14185, with an additional
test added (see julia_pseudovec in test/Transforms/InstCombine/insert-val-extract-elem.ll).
The patch causes instcombine to replace sequences of extractelement-insertvalue-store
that act essentially like a bitcast followed by a store.
Differential review: http://reviews.llvm.org/D14260
llvm-svn: 267482
Add a typedef for the std::map<GlobalValue::GUID, GlobalValueSummary *>
map that is passed around to identify summaries for values defined in a
particular module. This shortens up declarations in a variety of places.
llvm-svn: 267471
log2(Mask) is smaller than 32, we must use the 32-bit variant because the 64-bit
variant cannot encode it. Therefore, set the subreg part accordingly.
[AArch64] Fix optimizeCondBranch logic.
The opcode for the optimized branch does not depend on the size
of the activate bits in the AND masks, but the AND opcode itself.
Indeed, we need to use a X or W variant based on the AND variant
not based on whether the mask fits into the related variant.
Otherwise, we may end up using the W variant of the optimized branch
for 64-bit register inputs!
This fixes the last make check verifier issues for AArch64: PR27479.
llvm-svn: 267465
This replaces use of std::error_code and ErrorOr in the ORC RPC support library
with Error and Expected. This required updating the OrcRemoteTarget API, Client,
and server code, as well as updating the Orc C API.
This patch also fixes several instances where Errors were dropped.
llvm-svn: 267457
Use the operand for how long to wait. This is somewhat
distasteful, since it would be better to just emit s_nop
with the right argument in the first place. This would require
changing TII::insertNoop to emit N operands, which would be easy.
Slightly more problematic is the post-RA scheduler and hazard recognizer
represent nops as a single null node, and would require inventing
another way of representing N nops.
llvm-svn: 267456
There has been much recent confusion about the partition in the lattice between constant and non-constant values. Hopefully, documenting this will prevent confusion going forward.
llvm-svn: 267440
This function handled both unary and binary operators. Cloning and specializing leads to much easier to follow code with minimal duplicatation.
llvm-svn: 267438
Previously findClosestSuitableAluInstr was only considering the base register when checking the current instruction for suitability. Expand check to consider the offset if the offset is a register.
llvm-svn: 267424
visitAND, when folding and (load) forgets to check which output of
an indexed load is involved, happily folding the updated address
output on the following testcase:
target datalayout = "e-m:e-i64:64-n32:64"
target triple = "powerpc64le-unknown-linux-gnu"
%typ = type { i32, i32 }
define signext i32 @_Z8access_pP1Tc(%typ* %p, i8 zeroext %type) {
%b = getelementptr inbounds %typ, %typ* %p, i64 0, i32 1
%1 = load i32, i32* %b, align 4
%2 = ptrtoint i32* %b to i64
%3 = and i64 %2, -35184372088833
%4 = inttoptr i64 %3 to i32*
%_msld = load i32, i32* %4, align 4
%zzz = add i32 %1, %_msld
ret i32 %zzz
}
Fix this by checking ResNo.
I've found a few more places that currently neglect to check for
indexed load, and tightened them up as well, but I don't have test
cases for them. In fact, they might not be triggerable at all,
at least with current targets. Still, better safe than sorry.
Differential Revision: http://reviews.llvm.org/D19202
llvm-svn: 267420
Commit r266977 was reason for failing LLVM test suite with error message: fatal error: error in backend: Cannot select: t17: i32 = rotr t2, t11 ...
llvm-svn: 267418
Summary:
The expression is detected as a redundant expression.
Turn out, this is probably a bug.
```
/home/etienneb/llvm/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp:306:26: warning: both side of operator are equivalent [misc-redundant-expression]
if (isSMRD(*FirstLdSt) && isSMRD(*FirstLdSt)) {
```
Reviewers: rnk, tstellarAMD
Subscribers: arsenm, cfe-commits
Differential Revision: http://reviews.llvm.org/D19460
llvm-svn: 267415
We didn't have logic to correctly handle CFGs where there was more than
one EH-pad successor (these are novel with WinEH).
There were situations where a register was live in one exceptional
successor but not another but the code as written would only consider
the first exceptional successor it found.
This resulted in split points which were insufficiently early if an
invoke was present.
This fixes PR27501.
N.B. This removes getLandingPadSuccessor.
llvm-svn: 267412
Summary:
This patch adds support for the X asm constraint.
To do this, we lower the constraint to either a "w" or "r" constraint
depending on the operand type (both constraints are supported on ARM).
Fixes PR26493
Reviewers: t.p.northover, echristo, rengolin
Subscribers: joker.eph, jgreenhalgh, aemerson, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D19061
llvm-svn: 267411
The current logic assumes that any constant global will never be SRA'd. I presume this is because normally constant globals can be pushed into their uses and deleted. However, that sometimes can't happen (which is where you really want SRA, so the elements that can be eliminated, are!).
There seems to be no reason why we can't SRA constants too, so let's do it.
llvm-svn: 267393
If several regions cover the same area of code, we have to restore
the combined value for that area when return from a nested region.
This patch achieves that by combining regions before calling buildSegments.
Differential Revision: http://reviews.llvm.org/D18610
llvm-svn: 267390
Summary:
This implements a new method of run-time checking the NoWrap
SCEV predicates, which should be easier to optimize and nicer
for targets that don't correctly handle multiplication/addition
of large integer types (like i128).
If the AddRec is {a,+,b} and the backedge taken count is c,
the idea is to check that |b| * c doesn't have unsigned overflow,
and depending on the sign of b, that:
a + |b| * c >= a (b >= 0) or
a - |b| * c <= a (b <= 0)
where the comparisons above are signed or unsigned, depending on
the flag that we're checking.
The advantage of doing this is that we avoid extending to a larger
type and we avoid the multiplication of large types (multiplying
i128 can be expensive).
Reviewers: sanjoy
Subscribers: llvm-commits, mzolotukhin
Differential Revision: http://reviews.llvm.org/D19266
llvm-svn: 267389
ADD8TLS, a variant of add instruction used for initial-exec TLS,
currently accepts r0 as a source register. While add itself supports
r0 just fine, linker can relax it to a local-exec sequence, converting
it to addi - which doesn't support r0.
Differential Revision: http://reviews.llvm.org/D19193
llvm-svn: 267388
in a debug-info-bearing function has a debug location attached to it. Failure to
do so causes an "!dbg attachment points at wrong subprogram for function"
assertion failure when the inliner sets up inline scope info.
rdar://problem/25878916
This reaplies r267320 without changes after fixing an issue in the OpenMP IR
generator in clang.
llvm-svn: 267370
This corrects the MI annotations for the stack adjustment following the __chkstk
invocation. We were marking the original SP usage as a Def rather than Kill.
The (new) assigned value is the definition, the original reference is killed.
Adjust the ISelLowering to mark Kills and FrameSetup as well.
This partially resolves PR27480.
llvm-svn: 267361
As discussed on D19318, if we only demand the first element of a DIVSS/DIVSD intrinsic, then reduce to a FDIV call. This matches the existing FADD/FSUB/FMUL patterns.
llvm-svn: 267359
Split from D17490. This patch improves support for determining the demanded vector elements through SSE scalar intrinsics:
1 - demanded vector element support for unary and some extra binary scalar intrinsics (RCP/RSQRT/SQRT/FRCZ and ADD/CMP/DIV/ROUND).
2 - addss/addsd get simplified to a fadd call if we aren't interested in the pass through elements
3 - if we don't need the lowest element of a scalar operation then just use the first argument (the pass through elements) directly
We can add support for propagating demanded elements through any equivalent packed SSE intrinsics in a future patch (these wouldn't use the pass through patterns).
Differential Revision: http://reviews.llvm.org/D19318
llvm-svn: 267357
This patch improves support for determining the demanded vector elements through SSE scalar intrinsics:
1 - recognise that we only need the lowest element of the second input for binary scalar operations (and all the elements of the first input)
2 - recognise that the roundss/roundsd intrinsics use the lowest element of the second input and the remaining elements from the first input
Differential Revision: http://reviews.llvm.org/D17490
llvm-svn: 267356
We aren't currently making use of this in any successful mask decode and its actually incorrect as it inserts the wrong number of SM_SentinelUndef mask elements.
llvm-svn: 267350
There's hardly any functionality change here. Instead of calling
materializeMetadata on the first call to materialize(GlobalValue*), wait
until the first one that's actually going to do something. Noticed by
inspection; I don't have a concrete case where this makes a difference.
Added an assertion in materializeMetadata to be sure this (or a future
change) doesn't delay materializeMetadata after function-level metadata.
llvm-svn: 267345
Summary:
Remove the GlobalValueInfo and change the ModuleSummaryIndex to directly
reference summary objects. The info structure was there to support lazy
parsing of the combined index summary objects, which is no longer
needed and not supported.
Reviewers: joker.eph
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19462
llvm-svn: 267344
Reused the ability to split constants of a type wider than the shuffle mask to work with masks generated from scalar constants transfered to xmm.
This fixes an issue preventing PSHUFB target shuffle masks decoding rematerialized scalar constants and also exposes the XOP VPPERM bug described in PR27472.
llvm-svn: 267343
This fixes PR22248 on s390x. The previous attempt at this was D19101,
which was before LOAD_STACK_GUARD existed. Compared to the previous
version, this always emits a rather ugly block of 4 instructions, involving
a thread pointer load that can't be shared with other potential users.
However, this is necessary for SSP - spilling the guard value (or thread
pointer used to load it) is counter to the goal, since it could be
overwritten along with the frame it protects.
Differential Revision: http://reviews.llvm.org/D19363
llvm-svn: 267340
Add tests for some missing cases to bitcode upgrade in r267296.
- DICompositeType with an 'elements:' field, which will cause it to be
involved in a cycle after the upgrade.
- A DIDerivedType that references a class in 'extraData:'.
I updated test/Bitcode/dityperefs-3.8.ll with the missing cases and
regenerated test/Bitcode/dityperefs-3.8.ll.bc.
llvm-svn: 267332
The original patch caused crashes because it could derefence a null pointer
for SelectionDAGTargetInfo for targets that do not define it.
Evaluates fmul+fadd -> fmadd combines and similar code sequences in the
machine combiner. It adds support for float and double similar to the existing
integer implementation. The key features are:
- DAGCombiner checks whether it should combine greedily or let the machine
combiner do the evaluation. This is only supported on ARM64.
- It gives preference to throughput over latency: the heuristic used is
to combine always in loops. The targets decides whether the machine
combiner should optimize for throughput or latency.
- Supports for fmadd, f(n)msub, fmla, fmls patterns
- On by default at O3 ffast-math
llvm-svn: 267328
in a debug-info-bearing function has a debug location attached to it. Failure to
do so causes an "!dbg attachment points at wrong subprogram for function"
assertion failure when the inliner sets up inline scope info.
rdar://problem/25878916
llvm-svn: 267320
Right now it only contains the LinkageType, but will be extended
with "hasSection", "isOptSize", "hasInlineAssembly", etc.
Differential Revision: http://reviews.llvm.org/D19404
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267319
Keeping as much as possible internal/private is
known to help the optimizer. Let's try to benefit from
this in ThinLTO.
Note: this is early work, but is enough to build clang (and
all the LLVM tools). I still need to write some lit-tests...
Differential Revision: http://reviews.llvm.org/D19103
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267317
The option to control the emission of the new relocations
is -relax-relocations (blatantly copied from GNU as).
It can't be enabled by default because it breaks relatively
recent versions of ld.bfd/ld.gold (late 2015).
llvm-svn: 267307
Summary:
As discussed in D18298, some local globals can't
be renamed/promoted (because they have a section, or because
they are referenced from inline assembly).
To be able to detect naming collision, we need to keep around
the "GUID" using their original name without taking the linkage
into account.
Reviewers: tejohnson
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19454
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267304
Summary:
We are always importing the initializer for a GlobalVariable.
So if a GlobalVariable is in the export-list, we pull in any
refs as well.
Reviewers: tejohnson
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19102
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267303
A long overdue change to make DIGlobalVariable distinct. Much like
DISubprogram definitions (changed in r246098), it isn't logical to
unique DIGlobalVariable definitions from two different compile units.
(Longer-term, we should also find a way to reverse the link between
GlobalVariable and DIGlobalVariable, and between DIGlobalVariable and
DICompileUnit, so that debug info to do with optimized-out globals
disappears. Admittedly it's harder than with Function/DISubprogram,
since global variables may be constant-folded and the debug info should
still describe that somehow.)
llvm-svn: 267301
Eliminate DITypeIdentifierMap and make DITypeRef a thin wrapper around
DIType*. It is no longer legal to refer to a DICompositeType by its
'identifier:', and DIBuilder no longer retains all types with an
'identifier:' automatically.
Aside from the bitcode upgrade, this is mainly removing logic to resolve
an MDString-based reference to an actualy DIType. The commits leading
up to this have made the implicit type map in DICompileUnit's
'retainedTypes:' field superfluous.
This does not remove DITypeRef, DIScopeRef, DINodeRef, and
DITypeRefArray, or stop using them in DI-related metadata. Although as
of this commit they aren't serving a useful purpose, there are patchces
under review to reuse them for CodeView support.
The tests in LLVM were updated with deref-typerefs.sh, which is attached
to the thread "[RFC] Lazy-loading of debug info metadata":
http://lists.llvm.org/pipermail/llvm-dev/2016-April/098318.html
llvm-svn: 267296
Since forward references for uniqued node operands are expensive (and
those for distinct node operands are cheap due to
DistinctMDOperandPlaceholder), minimize forward references in uniqued
node operands.
Moreover, guarantee that when a cycle is broken by a distinct node, none
of the uniqued nodes have any forward references. In
ValueEnumerator::EnumerateMetadata, enumerate uniqued node subgraphs
first, delaying distinct nodes until all uniqued nodes have been
handled. This guarantees that uniqued nodes only have forward
references when there is a uniquing cycle (since r267276 changed
ValueEnumerator::organizeMetadata to partition distinct nodes in front
of uniqued nodes as a post-pass).
Note that a single uniqued subgraph can hit multiple distinct nodes at
its leaves. Ideally these would themselves be emitted in post-order,
but this commit doesn't attempt that; I think it requires an extra pass
through the edges, which I'm not convinced is worth it (since
DistinctMDOperandPlaceholder makes forward references quite cheap
between distinct nodes).
I've added two testcases:
- test/Bitcode/mdnodes-distinct-in-post-order.ll is just like
test/Bitcode/mdnodes-in-post-order.ll, except with distinct nodes
instead of uniqued ones. This confirms that, in the absence of
uniqued nodes, distinct nodes are still emitted in post-order.
- test/Bitcode/mdnodes-distinct-nodes-break-cycles.ll is the minimal
example where a naive post-order traversal would cause one uniqued
node to forward-reference another. IOW, it's the motivating test.
llvm-svn: 267278
When an operand of a distinct node hasn't been read yet, the reader can
use a DistinctMDOperandPlaceholder. This is much cheaper than forward
referencing from a uniqued node. Change
ValueEnumerator::organizeMetadata to partition distinct nodes and
uniqued nodes to reduce the overhead of cycles broken by distinct nodes.
Mehdi measured this for me; this removes most of the RAUW from the
importing step of -flto=thin, even after a WIP patch that removes
string-based DITypeRefs (introducing many more cycles to the metadata
graph).
llvm-svn: 267276
Summary:
As discussed in on the mailing list yesterday, I have refactored
BitcodeWriter.cpp to use classes to manage the bitcode writing process,
instead of passing around long lists of parameters between static
functions. See:
http://lists.llvm.org/pipermail/llvm-dev/2016-April/098610.html
I created a parent BitcodeWriter class to own the BitstreamWriter,
write the header, and contain the main entry point into the writing
process. There are two derived classes, one for writing a module and one
for writing a combined index file (for ThinLTO), which manage the
writing process specific to those bitcode file types.
I also changed the functions to conform to LLVM coding standards
(lowercase function name first letter). The only two routines that still
start with an uppercase letter are the two external interfaces, which
can be fixed as a follow-on (I wanted to keep this round just within
BitcodeWriter.cpp).
Reviewers: dexonsmith, joker.eph
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19447
llvm-svn: 267273
Mehdi's pattern recognition pulled this one out. This is cleaner with
std::find_if than with the strange helper function that took an iterator
by reference and updated it.
llvm-svn: 267271
Each reference to an unresolved MDNode is expensive, since the RAUW
support in MDNode uses a separate allocation and side map. Since
a distinct MDNode doesn't require its operands on creation (unlike
uniuqed nodes, there's no need to check for structural equivalence),
use nullptr for any of its unresolved operands. Besides reducing the
burden on MDNode maps, this can avoid allocating temporary MDNodes in
the first place.
We need some way to track operands. Invent DistinctMDOperandPlaceholder
for this purpose, which is a Metadata subclass that holds an ID and
points at its single user. DistinctMDOperandPlaceholder::replaceUseWith
is just like RAUW, but its name highlights that there is only ever
exactly one use.
There is no support for moving (or, obviously, copying) these. Move
support would be possible but expensive; leaving it unimplemented
prevents user error. In the BitcodeReader I originally considered
allocating on a BumpPtrAllocator and keeping a vector of pointers to
them, and then I realized that std::deque implements exactly this.
A couple of obvious follow-ups:
- Change ValueEnumerator to emit distinct nodes first to take more
advantage of this optimization. (How convenient... I think I might
have a couple of patches for this.)
- Change DIBuilder and its consumers (like CGDebugInfo in clang) to
use something like this when constructing debug info in the first
place.
llvm-svn: 267270
Consistently use the IsDistinct variable and start relying on it in
GET_OR_DISTINCT. This change has NFC, but prepares for using IsDistinct
to optimize the behaviour of the getMD() and getMDOrNull() helpers.
llvm-svn: 267268
The only functionality change was removing an error check from the
BitcodeReader (and an assertion from DILocation::getImpl) that is
already caught by Verifier::visitDILocation. The Verifier is a better
place for this anyway, and being inconsistent with other subclasses of
MDNode isn't serving anyone.
llvm-svn: 267267
Only one consumer (llvm-objdump) actually cared about the fact that there were
two triples. Others were actively working around the fact that the Triple
returned by getArch might have been invalid. As for llvm-objdump, it needs to
be acutely aware of both Triples anyway, so being generic in the exposed API is
no benefit.
Also rename the version of getArch returning a Triple. Users were having to
pass an unwanted nullptr to disambiguate the two, which was nasty.
The only functional change here is that armv7m and armv7em object files no
longer crash llvm-objdump.
llvm-svn: 267249
The dwo_name was added to dwo files to improve diagnostics in dwp, but
it confuses tools that attempt to load any dwo named by a dwo_name, even
ones inside dwos. Avoid this by keeping track of whether a unit is
already a dwo unit, and if so, not loading further dwos.
llvm-svn: 267241
The original commit was reverted because of a buildbot problem with LazyCallGraph::SCC handling (not related to the OptBisect handling).
Differential Revision: http://reviews.llvm.org/D19172
llvm-svn: 267231
Rather than relying on the gmlt-like data emitted into the .o/executable
which only contains the simple name of any inlined functions, use the
.dwo file if present.
Test symbolication with/without a .dwo, and the old test that was
testing behavior when no gmlt-like data was present. (I haven't included
a test of non-gmlt-like data + no .dwo (that would be akin to
symbolication with no debug info) but we could add one for completeness)
The test was simplified a bit to be a little clearer (unoptimized, force
inline, using a function call as the inlined entity) and regenerated
with ToT clang. For the no-gmlt-like-data case, I modified Clang back to
its old behavior temporarily & the .dwo file is identical so it is
shared between the two executables.
llvm-svn: 267227
Summary: The clang assembler assumes that the discriminator remains the same when there is source line change. The correct behavior is that when there is line change, discriminator will automatically reset to 0.
Reviewers: dnovillo, davidxl, echristo
Subscribers: echristo, llvm-commits
Differential Revision: http://reviews.llvm.org/D19436
llvm-svn: 267226
This intrinsic takes two arguments, ``%ptr`` and ``%offset``. It loads
a 32-bit value from the address ``%ptr + %offset``, adds ``%ptr`` to that
value and returns it. The constant folder specifically recognizes the form of
this intrinsic and the constant initializers it may load from; if a loaded
constant initializer is known to have the form ``i32 trunc(x - %ptr)``,
the intrinsic call is folded to ``x``.
LLVM provides that the calculation of such a constant initializer will
not overflow at link time under the medium code model if ``x`` is an
``unnamed_addr`` function. However, it does not provide this guarantee for
a constant initializer folded into a function body. This intrinsic can be
used to avoid the possibility of overflows when loading from such a constant.
Differential Revision: http://reviews.llvm.org/D18367
llvm-svn: 267223
This patch changes the interface for createPGOFuncNameMetadata() where we add
another PGOFuncName argument.
Differential Revision: http://reviews.llvm.org/D19433
llvm-svn: 267216
Summary:
We can fold compares to false when two distinct allocations within a
function are compared for equality.
Patch by Anna Thomas!
Reviewers: majnemer, reames, sanjoy
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19390
llvm-svn: 267214
The relative vtable ABI (PR26723) needs PLT relocations to refer to virtual
functions defined in other DSOs. The unnamed_addr attribute means that the
function's address is not significant, so we're allowed to substitute it
with the address of a PLT entry.
Also includes a bonus feature: addends for COFF image-relative references.
Differential Revision: http://reviews.llvm.org/D17938
llvm-svn: 267211
Extend the type canonicalization logic to work for unordered atomic loads and stores. Note that while this change itself is fairly simple and low risk, there's a reasonable chance this will expose problems in the backends by suddenly generating IR they wouldn't have seen before. Anything of this nature will be an existing bug in the backend (you could write an atomic float load), but this will definitely change the frequency with which such cases are encountered. If you see problems, feel free to revert this change, but please make sure you collect a test case.
llvm-svn: 267210
The opcode for the optimized branch does not depend on the size
of the activate bits in the AND masks, but the AND opcode itself.
Indeed, we need to use a X or W variant based on the AND variant
not based on whether the mask fits into the related variant.
Otherwise, we may end up using the W variant of the optimized branch
for 64-bit register inputs!
This fixes the last make check verifier issues for AArch64: PR27479.
llvm-svn: 267206
Summary: This change will shorten memset if the beginning of memset is overwritten by later stores.
Reviewers: hfinkel, eeckstein, dberlin, mcrosier
Subscribers: mgrang, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D18906
llvm-svn: 267197
E.g. for:
!1 = {"llvm.distribute", i32 1}
it now returns the MDOperand for 1.
I will use this in LoopDistribution to check the value of the metadata.
Note that the change is backward-compatible with its current use in
LoopVersioningLICM. An Optional implicitly converts to a bool depending
whether it contains a value or not.
llvm-svn: 267190
Avoid quadratic complexity in unusually large basic blocks by limiting
the size of the ready lists.
Differential Revision: http://reviews.llvm.org/D19349
llvm-svn: 267189
We used to simply set the kill flags to true when transforming a scalar
instruction to a vector one.
SrcScalar1 = copy SrcVector1
... = opScalar SrcScalar1
=>
SrcScalar1 = copy SrcVector1
... = opVector SrcVector1<kill>
This is obviously wrong. The proper update consists in:
1. Propagate the kill status from the copy to the new opVector
2. Reset the kill status on the copy, since the live-range of
SrcVector1 got extended.
This fixes some of the machine verifier errors for AArch64 with make check.
llvm-svn: 267180
Summary: eq imply [u|s]ge and [u|s]le are true.
Remove redundant logic by implementing isImpliedFalseByMatchingCmp(Pred1, Pred2)
as isImpliedTrueByMatchingCmp(Pred1, getInversePredicate(Pred2)).
llvm-svn: 267177
Summary:
(... while still not using a PostDomTree)
The way we use isKnownNotFullPoison from SCEV today, the new CFG walking
logic will not trigger for any realistic cases -- it will kick in only
for situations where we could have merged the contiguous basic blocks
anyway[0], since the poison generating instruction dominates all of its
non-PHI uses (which are the only uses we consider right now).
However, having this change in place will allow a later bugfix to break
fewer llvm-lit tests.
[0]: i.e. cases where block A branches to block B and B is A's only
successor and A is B's only predecessor.
Reviewers: broune, bjarke.roune
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19212
llvm-svn: 267175
Summary: [u|s]gt and [u|s]lt imply [u|s]ge and [u|s]le are true, respectively.
I've simplified the existing tests and added additional tests to cover the new
cases mentioned above. I've also added tests for all the cases where the
first compare doesn't imply anything about the second compare.
llvm-svn: 267171
- Switch few loops to range-based for loops
- Fix nop insertion at the end of BB
- Fix formatting
- Check for endpgm
Differential Revision: http://reviews.llvm.org/D19380
llvm-svn: 267167
Summary:
CachingMemorySSAWalker::invalidateInfo was using IsCall to determine
which cache map needed to be cleared of entries referring to the invalidated
MemoryAccess, but there could also be entries referring to it in the
other cache map (value entries, not key entries). This change just
clears both tables to be conservatively correct.
Also add a verifyRemoved() function, called when expensive
checks (i.e. XDEBUG) are enabled to verify that the invalidated
MemoryAccess object is not referenced in any of the caches.
Reviewers: dberlin, george.burgess.iv
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19388
llvm-svn: 267157
Summary:
This new pass allows targets to use the hazard recognizer without having
to also run one of the schedulers. This is useful when compiling with
optimizations disabled for targets that still need noop hazards
to be handled correctly.
Reviewers: hfinkel, atrick
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18594
llvm-svn: 267156
We take the intersection of overflow flags while CSE'ing.
This permits us to consider two instructions with different overflow
behavior to be replaceable.
llvm-svn: 267153
Summary:
When generating assembly using -m16 we must explicitly mark it as
16-bit. Emit .code16 at beginning of file. Fixes wrong results when
using -fno-integrated-as.
Reviewers: dwmw2
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19392
llvm-svn: 267152
When targetting MIPS64R6 some of the patterns for select were guarded by a
broken predicate. The predicate was supposed to test if a constant value
could fit in a 16 bit zero-extended field. Instead the value was tested to
fit in a 16 bit sign-extended field. For negative constants of native word
width this resulted in wrong code generation.
Reviewers: vkalintiris, dsanders
Differential Review: http://reviews.llvm.org/D19378
llvm-svn: 267151
r267049 broke multiple buildbots (e.g. clang-cmake-mips, and clang-x86_64-linux-selfhost-modules) which the follow-ups have not yet resolved and this is preventing subsequent committers from being notified about additional failures on the affected buildbots.
llvm-svn: 267148
Summary:
When optimizing PHIs which have inputs floating point binary
operators, we preserve all IR flags except the fast math
flags.
This change removes the logic which tracked some of the IR flags
(no wrap, exact) and replaces it by doing an and on the IR flags of
all inputs to the PHI - which will also handle the fast math
flags.
Reviewers: majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19370
llvm-svn: 267139
Summary:
rL256194 transforms truncations between vectors of integers into PACKUS/PACKSS
operations during DAG combine. This generates better code for truncate, so cost
of truncate needs to be changed but looks like it got changed only in SSE2 table
Whereas this change is also applicable for SSE4.1, so the cost of truncate needs
to be changed for that as well. Cost of “TRUNCATE v16i32 to v16i8” & “TRUNCATE
v16i16 to v16i8” should be same in SSE4.1 & SSE2 table. Removing their cost from
SSE4.1, so it will fall back to SSE2.
Reviewers: Simon Pilgrim
llvm-svn: 267123
Specifically, itineraries for LEON processors has been added, along with several LEON processor Subtargets. Although currently all these targets are pretty much identical, support for features that will differ among these processors will be added in the very near future.
The different Instruction Itinerary Classes (IICs) added are sufficient to differentiate between the instruction timings used by LEON and, quite probably, by generic Sparc processors too, but the focus of the exercise has been for LEON processors, as the requirement of my project. If the IICs are not sufficient for other Sparc processor types and you want to add a new itinerary for one of those, it should be relatively trivial to adapt this.
As none of the LEON processors has Quad Floats, or is a Version 9 processor, none of those instructions have itinerary classes defined and revert to the default "NoItinerary" instruction itinerary.
Phabricator Review: http://reviews.llvm.org/D19359
llvm-svn: 267121
EarlyCSE had inconsistent behavior with regards to flag'd instructions:
- In some cases, it would pessimize if the available instruction had
different flags by not performing CSE.
- In other cases, it would miscompile if it replaced an instruction
which had no flags with an instruction which has flags.
Fix this by being more consistent with our flag handling by utilizing
andIRFlags.
llvm-svn: 267111
Summary:
Also adds a small comment blurb on control flow + no-wrap flags, since
that question came up a few days back on llvm-dev.
Reviewers: bjarke.roune, broune
Subscribers: sanjoy, mcrosier, llvm-commits, mzolotukhin
Differential Revision: http://reviews.llvm.org/D19209
llvm-svn: 267110
Summary:
This intrinsic returns true if the current thread belongs to a live pixel
and false if it belongs to a pixel that we are executing only for derivative
computation. It will be used by Mesa to implement gl_HelperInvocation.
Note that for pixels that are killed during the shader, this implementation
also returns true, but it doesn't matter because those pixels are always
disabled in the EXEC mask.
This unearthed a corner case in the instruction verifier, which complained
about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but
correct code, so make the verifier accept it as such.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19191
llvm-svn: 267102
Re-layer the functions in the new (i.e., newly correct) post-order
traversals in ValueEnumerator (r266947) and ValueMapper (r266949).
Instead of adding a node to the worklist in a helper function and
returning a flag to say what happened, return the node itself. This
makes the code way cleaner: the worklist is local to the main function,
there is no flag for an early loop exit (since we can cleanly bury the
loop), and it's perfectly clear when pointers into the worklist might be
invalidated.
I'm fixing both algorithms in the same commit to avoid repeating the
commit message; if you take the time to understand one the other should
be easy. The diff itself isn't entirely obvious since the traversals
have some noise (i.e., things to do), but here's the high-level change:
auto helper = [&WL](T *Op) { auto helper = [](T **&I, T **E) {
=> while (I != E) {
if (shouldVisit(Op)) { T *Op = *I++;
WL.push(Op, Op->begin()); if (shouldVisit(Op)) {
return true; return Op;
} }
return false; return nullptr;
}; };
=>
WL.push(S, S->begin()); WL.push(S, S->begin());
while (!empty()) { while (!empty()) {
auto *N = WL.top().N; auto *N = WL.top().N;
auto *&I = WL.top().I; auto *&I = WL.top().I;
bool DidChange = false;
while (I != N->end())
if (helper(*I++)) { => if (T *Op = helper(I, N->end()) {
DidChange = true; WL.push(Op, Op->begin());
break; continue;
} }
if (DidChange)
continue;
POT.push(WL.pop()); => POT.push(WL.pop());
} }
Thanks to Mehdi for helping me find a better way to layer this.
llvm-svn: 267099
Evaluates fmul+fadd -> fmadd combines and similar code sequences in the
machine combiner. It adds support for float and double similar to the existing
integer implementation. The key features are:
- DAGCombiner checks whether it should combine greedily or let the machine
combiner do the evaluation. This is only supported on ARM64.
- It gives preference to throughput over latency: the heuristic used is
to combine always in loops. The targets decides whether the machine
combiner should optimize for throughput or latency.
- Supports for fmadd, f(n)msub, fmla, fmls patterns
- On by default at O3 ffast-math
llvm-svn: 267098
This removes the interfaces added (and not yet complete) to support
lazy reading of summaries. This support is not expected to be needed
since we are moving to a model where the full index is only being
traversed in the thin link step, instead of the back ends.
(The second part of this that I plan to do next is remove the
GlobalValueInfo from the ModuleSummaryIndex - it was mostly needed to
support lazy parsing of summaries. The index can instead reference the
summary structures directly.)
llvm-svn: 267097
WIN__DBZCHK will insert a CBZ instruction into the stream. This instruction
reserves 3 bits for the condition register (rn). As such, we must ensure that
we restrict the register to a low register. Use the tGPR class instead of GPR
to ensure that this is properly constrained. In debug builds, we would attempt
to use lr as a condition register which would silently get truncated with no
hint that the register selection was incorrect.
llvm-svn: 267080
We'd disabled them on x86 because back in the early days some host tools
couldn't handle the new load commands. This no longer holds: anyone capable of
deploying Clang should be able to deploy its copies of ar/ranlib/etc.
rdar://25254790
llvm-svn: 267075
When printing the properties required by a pass, only print the
properties that are set, and not those that are clear (only properties
that are set are verified, clear properties are "don't-care").
llvm-svn: 267070
Summary:
Adds an instrumentation pass for the new EfficiencySanitizer ("esan")
performance tuning family of tools. Multiple tools will be supported
within the same framework. Preliminary support for a cache fragmentation
tool is included here.
The shared instrumentation includes:
+ Turn mem{set,cpy,move} instrinsics into library calls.
+ Slowpath instrumentation of loads and stores via callouts to
the runtime library.
+ Fastpath instrumentation will be per-tool.
+ Which memory accesses to ignore will be per-tool.
Reviewers: eugenis, vitalybuka, aizatsky, filcab
Subscribers: filcab, vkalintiris, pcc, silvas, llvm-commits, zhaoqin, kcc
Differential Revision: http://reviews.llvm.org/D19167
llvm-svn: 267058
Summary: As per title. This will help work on the C API.
Reviewers: Wallbraker, whitequark, joker.eph, echristo, rafael
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19173
llvm-svn: 267057
InstrProfSymtab::create can fail with instrprof_error::malformed, but
this error is silently dropped. Propagate the error up to the caller so
we fail early.
Eventually, I'd like to transition ProfileData over to the new Error
class so we can't ignore hard failures like this.
llvm-svn: 267055
splitting edges.
MachineBasicBlock::SplitCriticalEdges will crash if a nullptr would have
been passed for the Pass argument. Do not allow that by turning this
argument into a reference.
The alternative would have been to make the Pass a truly optional
argument, but although this is easy to do, I was afraid users using it
like this would not be aware the livness information, dominator tree and
such would silently be broken.
llvm-svn: 267052
PDB parsing code was hand-rolled into llvm-pdbdump. This patch moves the
parsing of this code into DebugInfoPDB and makes the dumper use this.
This is achieved by implementing the skeleton of RawPdbSession, the
non-DIA counterpart to the existing PDB read interface. None of the type /
source file / etc information is accessible yet, so this implementation is
not yet close to achieving parity with the DIA counterpart, but the
RawSession class simply holds a reference to a PDBFile class which handles
parsing the file format. Additionally a PDBStream class is introduced
which allows accessing the bytes of a particular stream in a PDB file.
Differential Revision: http://reviews.llvm.org/D19343
Reviewed By: majnemer
llvm-svn: 267049
Introduce canSplitCriticalEdge, so that clients can now query whether or
not a critical edge can be split without actually needing to split it.
This may be useful when gathering information for cost models for
instance.
llvm-svn: 267046
The previous allocation code was over-estimating the amount of memory required.
No test case: we don't currently have a good way to detect conervative
over-allocation.
llvm-svn: 267041
Summary:
If we know that the pointer allocated within a function does not escape,
we can fold away comparisons that are done with global pointers
Patch by Anna Thomas!
Reviewers: reames, majnemer, sanjoy
Subscribers: mgrang, mcrosier, majnemer, llvm-commits
Differential Revision: http://reviews.llvm.org/D19276
llvm-svn: 267035
When custom lowered, this is not called if the store is custom
lowered. Move it to be a utility function so targets can
easily expand unaligned accesses when custom lowering.
llvm-svn: 267029
Instead of holding a mask, hold two value: the start index and the
length of the mapping. This is a more compact representation, although
less powerful. That being said, arbitrary masks would not have worked
for the generic so do not allow them in the first place.
llvm-svn: 267025
This patch implements a optimization bisect feature, which will allow optimizations to be selectively disabled at compile time in order to track down test failures that are caused by incorrect optimizations.
The bisection is enabled using a new command line option (-opt-bisect-limit). Individual passes that may be skipped call the OptBisect object (via an LLVMContext) to see if they should be skipped based on the bisect limit. A finer level of control (disabling individual transformations) can be managed through an addition OptBisect method, but this is not yet used.
The skip checking in this implementation is based on (and replaces) the skipOptnoneFunction check. Where that check was being called, a new call has been inserted in its place which checks the bisect limit and the optnone attribute. A new function call has been added for module and SCC passes that behaves in a similar way.
Differential Revision: http://reviews.llvm.org/D19172
llvm-svn: 267022
Summary:
IntrReadWriteArgMem simply becomes IntrArgMemOnly.
So there are fewer intrinsic properties that express their orthogonality
better, and correspond more closely to the corresponding IR attributes.
Suggested by: Philip Reames
Reviewers: joker.eph, reames, tstellarAMD
Subscribers: jholewinski, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19291
llvm-svn: 267021
"Into" was misleading. I am also planning to use this helper to look
for loop metadata and return the argument, so find seems like a better
name.
llvm-svn: 267013
This builds on 266999 which made FindAvailableValue do the right thing. Tests included show the newly enabled transforms and those which disabled either due to conservatism or correctness requirements.
llvm-svn: 267006
Before this fix, DILexicalBlockFile entries were skipped only in some cases and were not in other cases.
Differential Revision: http://reviews.llvm.org/D18724
llvm-svn: 267004
This change adds a couple of test cases to make sure FindAvailableLoadedValue does the right thing. At the moment, the code added is dead, but separating it makes follow on changes far more obvious.
llvm-svn: 266999
AArch64InstrInfo::optimizeCompareInstr has bug PR27158 which causes generation of incorrect code.
A compare instruction is substituted with another instruction which does not
produce the same flags as the original compare instruction.
This patch contains:
1. Fix of the bug.
2. A regression test in MIR.
3. A new test to check that SUBS is replaced by SUB.
Differential Revision: http://reviews.llvm.org/D18838
llvm-svn: 266969
