The discriminators pass relied on the presence of llvm.dbg.cu to decide
whether to add discriminators, but this fails in the case where debug
info is only enabled partially when -fprofile-sample-use is active.
The reason llvm.dbg.cu is not present in these cases is to prevent
codegen from emitting debug info (as it is only used for the sample
profile pass).
This changes the discriminators pass to also emit discriminators even
when debug info is not being emitted.
llvm-svn: 252763
MIPS32 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any MIPS32
implementation.
The net result of allowing this speculation for the regression tests in this patch is
that we get this code:
ctlz:
jr $ra
clz $2, $4
cttz:
addiu $1, $4, -1
not $2, $4
and $1, $2, $1
clz $1, $1
addiu $2, $zero, 32
jr $ra
subu $2, $2, $1
Instead of:
ctlz:
beqz $4, $BB0_2
addiu $2, $zero, 32
clz $2, $4
$BB0_2:
jr $ra
nop
cttz:
beqz $4, $BB1_2
addiu $2, $zero, 32
addiu $1, $4, -1
not $2, $4
and $1, $2, $1
clz $1, $1
addiu $2, $zero, 32
subu $2, $2, $1
$BB1_2:
jr $ra
nop
See D14469 for the larger motivation.
Differential Revision: http://reviews.llvm.org/D14500
llvm-svn: 252755
I missed the side-effects of ParseBFI in my previous attempt (r252748).
Thanks dblaikie for the suggestion of adding a void use of the unused
variable instead.
llvm-svn: 252751
If we have a chain of BFIs, we may be able to combine several together into one merged BFI. We can do this if the "from" bits from one BFI OR'd with the "from" bits from the other BFI form a contiguous range, and the same with the "to" bits.
llvm-svn: 252740
Measurements primarily on AArch64 have shown this feature does not
significantly effect compile-time. The are no significant perf changes in LNT,
but for AArch64 at least, there are wins in third party benchmarks.
As discussed on llvm-dev, we're going to try turning this on by default and see
how other targets react to the change.
llvm-svn: 252733
If possible and profitable, replace lea %reg, 1(%reg) and lea %reg, -1(%reg) with inc %reg and dec %reg respectively.
Patch by: anton.nadolsky@intel.com
Differential Revision: http://reviews.llvm.org/D14059
llvm-svn: 252722
Right now isTruePredicate is only ever called with Pred == ICMP_SLE or
ICMP_ULE, and the ICMP_SLT and ICMP_ULT cases are dead. This change
removes the untested dead code so that the function is not misleading.
llvm-svn: 252676
Summary:
This change teaches isImpliedCondition to prove things like
(A | 15) < L ==> (A | 14) < L
if the low 4 bits of A are known to be zero.
Depends on D14391
Reviewers: majnemer, reames, hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14392
llvm-svn: 252673
This change would add functionality if isImpliedCondition worked on
vector types; but since it bail out on vector predicates this change is
an NFC.
llvm-svn: 252672
This patch adds a pass for doing PowerPC peephole optimizations at the
MI level while the code is still in SSA form. This allows for easy
modifications to the instructions while depending on a subsequent pass
of DCE. Both passes are very fast due to the characteristics of SSA.
At this time, the only peepholes added are for cleaning up various
redundancies involving the XXPERMDI instruction. However, I would
expect this will be a useful place to add more peepholes for
inefficiencies generated during instruction selection. The pass is
placed after VSX swap optimization, as it is best to let that pass
remove unnecessary swaps before performing any remaining clean-ups.
The utility of these clean-ups are demonstrated by changes to four
existing test cases, all of which now have tighter expected code
generation. I've also added Eric Schweiz's bugpoint-reduced test from
PR25157, for which we now generate tight code. One other test started
failing for me, and I've fixed it
(test/Transforms/PlaceSafepoints/finite-loops.ll) as well; this is not
related to my changes, and I'm not sure why it works before and not
after. The problem is that the CHECK-NOT: of "statepoint" from test1
fails because of the "statepoint" in test2, and so forth. Adding a
CHECK-LABEL in between keeps the different occurrences of that string
properly scoped.
llvm-svn: 252651
Summary:
The module linker lazy links some "discardable if unused" global
values (e.g. linkonce), materializing and linking them only
if they are referenced in the module. If a comdat group contains a
linkonce member that is not referenced, however, it would not be
materialized and linked, leading to an incomplete comdat group.
If there are other object files not part of the same LTO link that also
define and use that comdat group, the linker may select the incomplete
group leading to link time unsats.
To solve this, whenever a global value body is linked, make sure we
materialize any other members of the same comdat group that are not yet
materialized. This ensures they are in the lazy link list and get linked
as well.
Added new test and adjusted old test to remove parts that didn't
make sense with fix.
Reviewers: rafael
Subscribers: dexonsmith, davidxl, llvm-commits
Differential Revision: http://reviews.llvm.org/D14516
llvm-svn: 252647
This was an omission in the patch that landed initial support for
operand bundles. So far we haven't hit this, but we will once the
inliner is able to inline calls to functions that contain calls with
operand bundles.
llvm-svn: 252645
ARM V6T2 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any ARM V6T2
implementation.
The net result of allowing this speculation for the regression tests in this patch is
that we get this code:
ctlz:
clz r0, r0
bx lr
cttz:
rbit r0, r0
clz r0, r0
bx lr
Instead of:
ctlz:
cmp r0, #0
moveq r0, #32
clzne r0, r0
bx lr
cttz:
cmp r0, #0
moveq r0, #32
rbitne r0, r0
clzne r0, r0
bx lr
This will help solve a general speculation/despeculation problem noted in PR24818:
https://llvm.org/bugs/show_bug.cgi?id=24818
Differential Revision: http://reviews.llvm.org/D14469
llvm-svn: 252639
This allows avoiding the default Expand behavior which
introduces stack usage. Bitcast the scalar and replace
the missing elements with undef.
This is covered by existing tests and used by a future
commit which makes 64-bit vectors legal types on AMDGPU.
llvm-svn: 252632
This is for AMDGPU to implement v2i64 extract as extract of
half of a v4i32.
This is covered by existing tests and used by a future
commit which makes 64-bit vectors legal types on AMDGPU.
llvm-svn: 252630
This is a cleaned up version of a patch by John Regehr with permission. Originally found via the souper tool.
If we add an odd number to x, then bitwise-and the result with x, we know that the low bit of the result must be zero. Either it was zero in x originally, or the add cleared it in the temporary value. As a result, one of the two values anded together must have the bit cleared.
Differential Revision: http://reviews.llvm.org/D14315
llvm-svn: 252629
Ensure WeakAny variables are imported as ExternalWeak declarations. To
handle WeakAny more consistently and fix this issue:
1) Update helper doImportAsDefinition to properly flag WeakAny variables
and aliases as not importing defintions.
Update callers of doImportAsDefinition to remove now redundant checks for
WeakAny aliases, or ignore aliases, as appropriate.
2) Add any !doImportAsDefinition GVs to DoNotLinkFromSource set during
linking of the GV prototype, where we usually add GVs to the
DoNotLinkFromSource set for other reasons.
Remove now unnecessary adding of WeakAny aliases to
DoNotLinkFromSource set from copyGlobalAliasProto.
Remove now unnecessary guard against linking non-imported function
bodies from ModuleLinker::run.
llvm-svn: 252626
AArch64 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any AArch64
implementation.
The net result of allowing this speculation for the regression tests in this
patch is that we get this code:
ctlz:
clz w0, w0
ret
cttz:
rbit w8, w0
clz w0, w8
ret
Instead of:
ctlz:
cbz w0, .LBB0_2
clz w0, w0
ret
.LBB0_2:
orr w0, wzr, #0x20
ret
cttz:
cbz w0, .LBB1_2
rbit w8, w0
clz w0, w8
ret
.LBB1_2:
orr w0, wzr, #0x20
ret
See D14469 for the larger motivation.
Differential Revision: http://reviews.llvm.org/D14505
llvm-svn: 252625
This is one of the problems noted in PR25016:
https://llvm.org/bugs/show_bug.cgi?id=25016
and:
http://lists.llvm.org/pipermail/llvm-dev/2015-October/090998.html
The spilling problem is independent and not addressed by this patch.
The MachineCombiner was doing reassociations that don't improve or even worsen the critical path.
This is caused by inclusion of the "slack" factor when calculating the critical path of the original
code sequence. If we don't add that, then we have a more conservative cost comparison of the old code
sequence vs. a new sequence. The more liberal calculation must be preserved, however, for the AArch64
MULADD patterns because benchmark regressions were observed without that.
The two failing test cases now have identical asm that does what we want:
a + b + c + d ---> (a + b) + (c + d)
Differential Revision: http://reviews.llvm.org/D13417
llvm-svn: 252616
Added fixes for stage2 failures: CMOV is not commutable; commuting the operands results in the condition being flipped! d'oh!
Original commit message:
If we have a CMOV, OR and AND combination such as:
if (x & CN)
y |= CM;
And:
* CN is a single bit;
* All bits covered by CM are known zero in y;
Then we can convert this to a sequence of BFI instructions. This will always be a win if CM is a single bit, will always be no worse than the TST & OR sequence if CM is two bits, and for thumb will be no worse if CM is three bits (due to the extra IT instruction).
llvm-svn: 252606
This is fix for PR24059.
When we are hoisting instruction above some condition it may turn out
that metadata on this instruction was control dependant on the condition.
This metadata becomes invalid and we need to drop it.
This patch should cover most obvious places of speculative execution (which
I have found by greping isSafeToSpeculativelyExecute). I think there are more
cases but at least this change covers the severe ones.
Differential Revision: http://reviews.llvm.org/D14398
llvm-svn: 252604
For big-endian targets, when we merge two halfword loads into a word load, the
order of the halfwords in the loaded value is reversed compared to
little-endian, so the load-store optimiser needs to swap the destination
registers.
This does not affect merging of two word loads, as we use ldp, which treats the
memory as two separate 32-bit words.
llvm-svn: 252597
For CoreCLR on Windows, stack probes must be emitted as inline sequences that probe successive stack pages
between the current stack limit and the desired new stack pointer location. This implements support for
the inline expansion on x64.
For in-body alloca probes, expansion is done during instruction lowering. For prolog probes, a stub call
is initially emitted during prolog creation, and expanded after epilog generation, to avoid complications
that arise when introducing new machine basic blocks during prolog and epilog creation.
Added a new test case, modified an existing one to exclude non-x64 coreclr (for now).
Add test case
Fix tests
llvm-svn: 252578
AArch64 has the ability to use the top 8-bits of an "address" for extra
information, with the memory subsystem automatically masking them off for loads
and stores. When that's happening, we can sometimes skip masks on memory
operations in the compiler.
However, this requires the host OS and support stack to preserve those bits so
it can't be enabled everywhere. In principle iOS 8.0 and above do take the
required precautions and but we'll put it under a flag for now.
llvm-svn: 252573
Lower LLVM's 'unreachable' terminator to ISD::TRAP, and lower ISD::TRAP to
wasm's 'unreachable' expression.
WebAssembly type-checks expressions, but a noreturn function with a
return type that doesn't match the context will cause a check
failure. So we lower LLVM 'unreachable' to ISD::TRAP and then lower that
to WebAssembly's 'unreachable' expression, which typechecks in any
context and causes a trap if executed.
Differential Revision: http://reviews.llvm.org/D14515
llvm-svn: 252566
I'm not sure what the point of this was. I'm not sure why
you would ever define an instruction that produces an unallocatable
register class. No tests fail with this removed, and it seems like
it should be a verifier error to define such an instruction.
This was problematic for AMDGPU because it would make bad decisions
by arbitrarily changing the register class when unsetting isAllocatable
for VS_32/VS_64, which is currently set as a workaround to this problem.
AMDGPU uses the VS_32/VS_64 register classes to represent operands which
can use either VGPRs or SGPRs. When isAllocatable is unset for these,
this would need to pick either the SGPR or VGPR class and insert either
a copy we don't want, or an illegal copy we would need to deal with
later. A semi-arbitrary register class ordering decision is made in tablegen,
which resulted in always picking a VGPR class because it happens to have
more registers than the SGPR register class. We really just want to
use whatever register class the original register had.
llvm-svn: 252565
- Make indexed value profile data more compact by peeling out
the per-site value count field into its own smaller sized array.
- Introduced formal data structure definitions to specify value
profile data layout in indexed format. Previously the layout
of the data is only assumed in the client code (scattered in
three different places : size computation, EmitData, and ReadData
- The new data structure serves as a central place for layout documentation.
- Add interfaces to force BE output for value profile data (testing purpose)
- Add byte swap unit tests
Differential Revision: http://reviews.llvm.org/D14401
llvm-svn: 252563
This fixes a bug in ARMAsmPrinter::EmitUnwindingInstruction where
llvm_unreachable was reached because t2ADDri wasn't handled.
Test case provided by Tim Northover.
rdar://problem/23270609
http://reviews.llvm.org/D14518
llvm-svn: 252557
Simply perform additional report_context() calls after a report()
instead of adding more and more overloaded variations of report(). Also
improve several instances where information was output in an ad-hoc way
probably because no matching report() overload was available.
llvm-svn: 252552
This is a prerequisite for further optimisations of these functions,
which will be commited as a separate patch.
Differential Revision: http://reviews.llvm.org/D14219
llvm-svn: 252535
The motivation for this patch starts with the epic fail example in PR18007:
https://llvm.org/bugs/show_bug.cgi?id=18007
...unfortunately, this patch makes no difference for that case, but it solves some
simpler cases. We'll get there some day. :)
The current 'or' matching code was using computeKnownBits() via
isBaseWithConstantOffset() -> MaskedValueIsZero(), but that's an unnecessarily limited use.
We can do more by copying the logic in ValueTracking's haveNoCommonBitsSet(), so we can
treat the 'or' as if it was an 'add'.
There's a TODO comment here because we should lift the bit-checking logic into a helper
function, so it's not duplicated in DAGCombiner.
An example of the better LEA matching:
leal (%rdi,%rdi), %eax
andl $1, %esi
orl %esi, %eax
Becomes:
andl $1, %esi
leal (%rsi,%rdi,2), %eax
Differential Revision: http://reviews.llvm.org/D13956
llvm-svn: 252515
For some reason we'd never run MachineVerifier on WinEH code, and you
explicitly have to ask for it with llc. I added it to a few test cases
to get some coverage.
Fixes PR25461.
llvm-svn: 252512
This patch makes ASAN for aarch64 use the same shadow offset for all
currently supported VMAs (39 and 42 bits). The shadow offset is the
same for 39-bit (36). Similar to ppc64 port, aarch64 transformation
also requires to use an add instead of 'or' for 42-bit VMA.
llvm-svn: 252495
Summary: Call instructions that are from the same line and same basic block needs to have separate discriminators to distinguish between different callsites.
Reviewers: davidxl, dnovillo, dblaikie
Subscribers: dblaikie, probinson, llvm-commits
Differential Revision: http://reviews.llvm.org/D14464
llvm-svn: 252492
When GlobalOpt splits an internal, global variable with an aggregate type, it
should propagate the externally_initialized flag to the newly created globals.
This makes the pass safe for our downstream use of this flag, while still
allowing some useful optimisations (such as removing dead parts of the split
aggregate) to be performed.
Differential Revision: http://reviews.llvm.org/D13382
llvm-svn: 252490
Implemented as many of Michael's suggestions as were possible:
* clang-format the added code while it is still fresh.
* tried to change Value* to Instruction* in many places in computeMinimumValueSizes - unfortunately there are several places where Constants need to be handled so this wasn't possible.
* Reduce the pass list on loop-vectorization-factors.ll.
* Fix a bug where we were querying MinBWs for I->getOperand(0) but using MinBWs[I].
llvm-svn: 252469
Summary:
LAA currently generates a set of SCEV predicates that must be checked by users.
In the case of Loop Distribute/Loop Load Elimination, no such predicates could have
been emitted, since we don't allow stride versioning. However, in the future there
could be SCEV predicates that will need to be checked.
This change adds support for SCEV predicate versioning in the Loop Distribute, Loop
Load Eliminate and the loop versioning infrastructure.
Reviewers: anemet
Subscribers: mssimpso, sanjoy, llvm-commits
Differential Revision: http://reviews.llvm.org/D14240
llvm-svn: 252467
Summary:
This matches the sum-of-absdiff patterns emitted by the vectoriser using log2 shuffles.
Relies on D14207 to be able to match the `extract_subvector(..., 0)`
Reviewers: t.p.northover, jmolloy
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D14208
llvm-svn: 252465
Summary:
Lowering this pattern early to an `EXTRACT_SUBREG` was making it impossible to match larger patterns in tblgen that use `extract_subvector(..., 0)` as part of the their input pattern.
It seems like there will exist somewhere a better way of specifying this pattern over all relevant register value types, but I didn't manage to find it.
Reviewers: t.p.northover, jmolloy
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D14207
llvm-svn: 252464
"GCC requires the freestanding environment provide memcpy, memmove, memset
and memcmp": https://gcc.gnu.org/onlinedocs/gcc-5.2.0/gcc/Standards.html
Hence in GNUEABI targets LLVM should not convert 'memops' to their equivalent
'__aeabi_memops'. This convertion violates GCC contract.
The -meabi flag controls whether or not LLVM will modify 'memops' in GNUEABI
targets.
Without -meabi: use the triple default EABI.
With -meabi=default: use the triple default EABI.
With -meabi=gnu: use 'memops'.
With -meabi=4 or -meabi=5: use '__aeabi_memops'.
With -meabi set to an unknown value: same as -meabi=default.
Patch by Vinicius Tinti.
llvm-svn: 252462
We don't currently have any runtime library functions for operations on
f16 values (other than conversions to and from f32 and f64), so we
should always promote it to f32, even if that is not a legal type. In
that case, the f32 values would be softened to f32 library calls.
SoftenFloatRes_FP_EXTEND now needs to check the promoted operand's type,
as it may ne a no-op or require a different library call.
getCopyFromParts and getCopyToParts now need to cope with a
floating-point value stored in a larger integer part, as is the case for
any target that needs to store an f16 value in a 32-bit integer
register.
Differential Revision: http://reviews.llvm.org/D12856
llvm-svn: 252459
Summary: Mimic parseTriple(); and exposes it to LTOModule.cpp
Reviewers: dexonsmith, rafael
Subscribers: llvm-commits
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 252442
llvm/lib/Support/Signals.cpp:66:13: warning: unused function 'printSymbolizedStackTrace' [-Wunused-function]
llvm/lib/Support/Signals.cpp:52:13: warning: function 'findModulesAndOffsets' has internal linkage but is not defined [-Wundefined-internal]
llvm-svn: 252418
Under most circumstances, if SCEV can simplify X-Y to a constant, then it can
also simplify Y-X to a constant. However, there is no guarantee that this is
always true, and concensus is not to consider that a correctness bug in SCEV
(although it is undesirable).
PPCLoopPreIncPrep gathers pointers used to access memory (via loads, stores and
prefetches) into buckets, where in each bucket the relative pointer offsets are
constant. We used to keep each bucket as a multimap, where SCEV's subtraction
operation was used to define the ordering predicate. Instead, use a fixed SCEV
base expression for each bucket, record the constant offsets from that base
expression, and adjust it later, if desirable, once all pointers have been
collected.
Doing it this way should be more compile-time efficient than the previous
scheme (in addition to making the implementation less sensitive to SCEV
simplification quirks).
Fixes PR25170.
llvm-svn: 252417
The TailDuplication machine pass ran across a malformed CFG: a PHI node
referred it's predecessor's predecessor instead of it's predecessor.
This occurred because we split the edge in X86ISelLowering when we
processed the CATCHRET but forgot to do something about the PHI nodes.
This fixes PR25444.
llvm-svn: 252413
This commit adds enums in LLVMBitCodes.h to improve readability and
maintainability. This is a follow-up to r252368 which was discussed
here:
http://reviews.llvm.org/D12923
llvm-svn: 252395
Summary:
Teach the FunctionAttrs to do the right thing for IR with operand
bundles.
Reviewers: reames, chandlerc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14408
llvm-svn: 252387
Summary:
This change fixes an iterator wraparound bug in
`determinePointerReadAttrs`.
Ideally, ++'ing off the `end()` of an iplist should result in a failed
assert, but currently iplist seems to silently wrap to the head of the
list on `end()++`. This is why the bad behavior is difficult to
demonstrate.
Reviewers: chandlerc, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14350
llvm-svn: 252386
Summary:
The CLR's personality routine passes these in rdx/edx, not rax/eax.
Make getExceptionPointerRegister a virtual method parameterized by
personality function to allow making this distinction.
Similarly make getExceptionSelectorRegister a virtual method parameterized
by personality function, for symmetry.
Reviewers: pgavlin, majnemer, rnk
Subscribers: jyknight, dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D14344
llvm-svn: 252383
FoldPHIArgZextsIntoPHI cannot insert an instruction after the PHI if
there is an EHPad in the BB. Doing so would result in an instruction
inserted after a terminator.
llvm-svn: 252377
Some implicit ilist iterator conversions have crept back into Analysis,
Transforms, Hexagon, and llvm-stress. This removes them.
I'll commit a patch immediately after this to disallow them (in a
separate patch so that it's easy to revert if necessary).
llvm-svn: 252371
We tried to insert a cast of a phi in a block whose terminator is an
EHPad. This is invalid. Do not attempt the transform in these
circumstances.
llvm-svn: 252370
This marker prevents optimization passes from adding 'tail' or
'musttail' markers to a call. Is is used to prevent tail call
optimization from being performed on the call.
rdar://problem/22667622
Differential Revision: http://reviews.llvm.org/D12923
llvm-svn: 252368
Summary:
In general GetTempDir follows the same logic as the replaced code: checks env variables TMP, TEMP, USERPROFILE in order. However, it also perform other checks like making separators native (\), making the path absolute, etc.
This change fixes FileSystemTest.CreateDir unittest that had been failing when run from Unix-like shell on Windows (Unix-like path separator (/) used in env variables).
Reviewers: chapuni, rafael, aaron.ballman
Subscribers: rafael, llvm-commits
Differential Revision: http://reviews.llvm.org/D14231
llvm-svn: 252366
We used to try to constant-fold them to i32 immediates.
Given that fast-isel doesn't otherwise support vNi1, when selecting
the result users, we'd fallback to SDAG anyway.
However, if the users were in another block, we'd insert broken
cross-class copies (GPR32 to FPR64).
Give up, let SDAG agree with itself on a vNi1 legalization strategy.
llvm-svn: 252364
When matching non-LSB-extracting truncating broadcasts, we now insert
the necessary SRL. If the scalar resulted from a load, the SRL will be
folded into it, creating a narrower, offset, load.
However, i16 loads aren't Desirable, so we get i16->i32 zextloads.
We already catch i16 aextloads; catch these as well.
llvm-svn: 252363
Now that we recognize this, we can support it instead of bailing out.
That is, we can fold:
(v8i16 (shufflevector
(v8i16 (bitcast (v4i32 (build_vector X, Y, ...)))),
<1,1,...,1>))
into:
(v8i16 (vbroadcast (i16 (trunc (srl Y, 16)))))
llvm-svn: 252362
We used to incorrectly assume that the offset we're extracting from
was a multiple of the element size. So, we'd fold:
(v8i16 (shufflevector
(v8i16 (bitcast (v4i32 (build_vector X, Y, ...)))),
<1,1,...,1>))
into:
(v8i16 (vbroadcast (i16 (trunc Y))))
whereas we should have extracted the higher bits from X.
Instead, bail out if the assumption doesn't hold.
llvm-svn: 252361
Summary:
Pass the VOPProfile object all the through to *_m multiclasses. This will
allow us to do more simplifications in the future.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D13437
llvm-svn: 252339
The SLPVectorizer had a very crude way of trying to benefit
from associativity: it tried to optimize for splat/broadcast
or in order to have the same operator on the same side.
This is benefitial to the cost model and allows more vectorization
to occur.
This patch improve the logic and make the detection optimal (locally,
we don't look at the full tree but only at the immediate children).
Should fix https://llvm.org/bugs/show_bug.cgi?id=25247
Reviewers: mzolotukhin
Differential Revision: http://reviews.llvm.org/D13996
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 252337
All 3 operands of FMA3 instructions are commutable now.
Patch by Slava Klochkov
Reviewers: Quentin Colombet(qcolombet), Ahmed Bougacha(ab).
Differential Revision: http://reviews.llvm.org/D13269
llvm-svn: 252335
Modelling of the expression stack is evolving. This patch takes another
step by making pushes explicit.
Differential Revision: http://reviews.llvm.org/D14338
llvm-svn: 252334
Summary:
This change makes the `isImpliedCondition` interface similar to the rest
of the functions in ValueTracking (in that it takes a DataLayout,
AssumptionCache etc.). This is an NFC, intended to make a later diff
less noisy.
Depends on D14369
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14391
llvm-svn: 252333
Summary:
Currently `isImpliedCondition` will optimize "I +_nuw C < L ==> I < L"
only if C is positive. This is an unnecessary restriction -- the
implication holds even if `C` is negative.
Reviewers: reames, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14369
llvm-svn: 252332
Summary:
This change adds a framework for adding more smarts to
`isImpliedCondition` around inequalities. Informally,
`isImpliedCondition` will now try to prove "A < B ==> C < D" by proving
"C <= A && B <= D", since then it follows "C <= A < B <= D".
While this change is in principle NFC, I could not think of a way to not
handle cases like "i +_nsw 1 < L ==> i < L +_nsw 1" (that ValueTracking
did not handle before) while keeping the change understandable. I've
added tests for these cases.
Reviewers: reames, majnemer, hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14368
llvm-svn: 252331
Mark kernels that use certain features that require user
SGPRs to support with kernel attributes. We need to know
before instruction selection begins because it impacts
the kernel calling convention lowering.
For now this only detects the workitem intrinsics.
llvm-svn: 252323
For some reason VS_32 ends up factoring into the pressure heuristics
even though we should never see a virtual register with this class.
When SGPRs are reserved for register spilling, this for some reason
triggers reg-crit scheduling.
Setting isAllocatable = 0 may help with this since that seems to remove
it from the default implementation's generated table.
llvm-svn: 252321
Summary:
This reverts commit r251965.
Restore "Move metadata linking after lazy global materialization/linking."
This restores commit r251926, with fixes for the LTO bootstrapping bot
failure.
The bot failure was caused by references from debug metadata to
otherwise unreferenced globals. Previously, this caused the lazy linking
to link in their defs, which is unnecessary. With this patch, because
lazy linking is complete when we encounter the metadata reference, the
materializer created a declaration. For definitions such as aliases and
comdats, it is illegal to have a declaration. Furthermore, metadata
linking should not change code generation. Therefore, when linking of
global value bodies is complete, the materializer will simply return
nullptr as the new reference for the linked metadata.
This change required fixing a different test to ensure there was a
real reference to a linkonce global that was only being reference from
metadata.
Note that the new changes to the only-needed-named-metadata.ll test
illustrate an issue with llvm-link -only-needed handling of comdat
groups, whereby it may result in an incomplete comdat group. I note this
in the test comments, but the issue is orthogonal to this patch (it can
be reproduced without any metadata at head).
Reviewers: dexonsmith, rafael, tra
Subscribers: tobiasvk, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D14447
llvm-svn: 252320
Summary:
In this implementation, LiveIntervalAnalysis invents a few register
masks on basic block boundaries that preserve no registers. The nice
thing about this is that it prevents the prologue inserter from thinking
it needs to spill all XMM CSRs, because it doesn't see any explicit
physreg defs in the MI.
Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer
Subscribers: MatzeB, llvm-commits
Differential Revision: http://reviews.llvm.org/D14407
llvm-svn: 252318
The benefit from converting narrow loads into a wider load (r251438) could be
micro-architecturally dependent, as it assumes that a single load with two bitfield
extracts is cheaper than two narrow loads. Currently, this conversion is
enabled only in cortex-a57 on which performance benefits were verified.
llvm-svn: 252316
We now create the .eh_frame section early, just like every other special
section.
This means that the special flags are visible in code that explicitly
asks for ".eh_frame".
llvm-svn: 252313
Summary:
The bug was that the sldi instructions have immediate widths dependant on
their element size. So sldi.d has a 1-bit immediate and sldi.b has a 4-bit
immediate. All of these were using 4-bit immediates previously.
Reviewers: vkalintiris
Subscribers: llvm-commits, atanasyan, dsanders
Differential Revision: http://reviews.llvm.org/D14018
llvm-svn: 252297
Summary:
The bug was that the MIPS32R6/MIPS64R6/microMIPS32R6 versions of LSA and DLSA
(unlike the MSA version) failed to account for the off-by-one encoding of the
immediate. The range is actually 1..4 rather than 0..3.
Reviewers: vkalintiris
Subscribers: atanasyan, dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D14015
llvm-svn: 252295
Summary:
Without these patterns we would generate a complete LL/SC sequence.
This would be problematic for memory regions marked as WRITE-only or
READ-only, as the instructions LL/SC would read/write to the protected
memory regions correspondingly.
Reviewers: dsanders
Subscribers: llvm-commits, dsanders
Differential Revision: http://reviews.llvm.org/D14397
llvm-svn: 252293
This attribute allows the compiler to assume that the function never recurses into itself, either directly or indirectly (transitively). This can be used among other things to demote global variables to locals.
llvm-svn: 252282
This adds the EH_RESTORE x86 pseudo instr, which is responsible for
restoring the stack pointers: EBP and ESP, and ESI if stack realignment
is involved. We only need this on 32-bit x86, because on x64 the runtime
restores CSRs for us.
Previously we had to keep the CATCHRET instruction around during SEH so
that we could convince X86FrameLowering to restore our frame pointers.
Now we can split these instructions earlier.
This was confusing, because we had a return instruction which wasn't
really a return and was ultimately going to be removed by
X86FrameLowering. This change also simplifies X86FrameLowering, which
really shouldn't be building new MBBs.
No observable functional change currently, but with the new register
mask stuff in D14407, CATCHRET will become a register allocator barrier,
and our existing tests rely on us having reasonable register allocation
around SEH.
llvm-svn: 252266
Windows EH funclets need to always return to a single parent funclet. However, it is possible for earlier optimizations to combine funclets (probably based on one funclet having an unreachable terminator) in such a way that this condition is violated.
These changes add code to the WinEHPrepare pass to detect situations where a funclet has multiple parents and clone such funclets, fixing up the unwind and catch return edges so that each copy of the funclet returns to the correct parent funclet.
Differential Revision: http://reviews.llvm.org/D13274?id=39098
llvm-svn: 252249
The bug: I missed adding break statements in the switch / case.
Original commit message:
[SCEV] Teach SCEV some axioms about non-wrapping arithmetic
Summary:
- A s< (A + C)<nsw> if C > 0
- A s<= (A + C)<nsw> if C >= 0
- (A + C)<nsw> s< A if C < 0
- (A + C)<nsw> s<= A if C <= 0
Right now `C` needs to be a constant, but we can later generalize it to
be a non-constant if needed.
Reviewers: atrick, hfinkel, reames, nlewycky
Subscribers: sanjoy, llvm-commits
Differential Revision: http://reviews.llvm.org/D13686
llvm-svn: 252236
Previously, subprograms contained a metadata reference to the function they
described. Because most clients need to get or set a subprogram for a given
function rather than the other way around, this created unneeded inefficiency.
For example, many passes needed to call the function llvm::makeSubprogramMap()
to build a mapping from functions to subprograms, and the IR linker needed to
fix up function references in a way that caused quadratic complexity in the IR
linking phase of LTO.
This change reverses the direction of the edge by storing the subprogram as
function-level metadata and removing DISubprogram's function field.
Since this is an IR change, a bitcode upgrade has been provided.
Fixes PR23367. An upgrade script for textual IR for out-of-tree clients is
attached to the PR.
Differential Revision: http://reviews.llvm.org/D14265
llvm-svn: 252219
inalloca variables were not treated as static allocas, therefore didn't
participate in regular stack instrumentation. We don't want them to
participate in dynamic alloca instrumentation as well.
llvm-svn: 252213
We already had a test for this for 32-bit SEH catchpads, but those don't
actually create funclets. We had a bug that only appeared in funclet
prologues, where we would establish EBP and ESI as our FP and BP, and
then downstream prologue code would overwrite them.
While I was at it, I fixed Win64+funclets+stackrealign. This issue
doesn't come up as often there due to the ABI requring 16 byte stack
alignment, but now we can rest easy that AVX and WinEH will work well
together =P.
llvm-svn: 252210
Mangling type information into MachineInstr opcode names was a temporary
measure, and it's starting to get hairy. At the same time, the MC instruction
printer wants to use AsmString strings for printing. This patch takes the
first step, starting the process of adding AsmStrings for instructions.
llvm-svn: 252203
Also, remove an enum hack where enum values were used as indexes into an array.
We may want to make this a real class to allow pattern-based queries/customization (D13417).
llvm-svn: 252196
The needed lld matching changes to be submitted immediately next,
but this revision will cause lld failures with this alone which is expected.
This removes the eating of the error in Archive::Child::getSize() when the characters
in the size field in the archive header for the member is not a number. To do this we
have all of the needed methods return ErrorOr to push them up until we get out of lib.
Then the tools and can handle the error in whatever way is appropriate for that tool.
So the solution is to plumb all the ErrorOr stuff through everything that touches archives.
This include its iterators as one can create an Archive object but the first or any other
Child object may fail to be created due to a bad size field in its header.
Thanks to Lang Hames on the changes making child_iterator contain an
ErrorOr<Child> instead of a Child and the needed changes to ErrorOr.h to add
operator overloading for * and -> .
We don’t want to use llvm_unreachable() as it calls abort() and is produces a “crash”
and using report_fatal_error() to move the error checking will cause the program to
stop, neither of which are really correct in library code. There are still some uses of
these that should be cleaned up in this library code for other than the size field.
The test cases use archives with text files so one can see the non-digit character,
in this case a ‘%’, in the size field.
These changes will require corresponding changes to the lld project. That will be
committed immediately after this change. But this revision will cause lld failures
with this alone which is expected.
llvm-svn: 252192
Summary:
This review is related to another review request http://reviews.llvm.org/D11268, does the same and merely fixes a couple of issues with it.
D11268 is quite old and has merge conflicts against the current trunk.
This request
- rebases D11268 onto the new trunk;
- resolves the merge conflicts;
- fixes the prologue_end tests, which do not pass due to the subprogram definitions not marked as distinct.
Reviewers: echristo, rengolin, kubabrecka
Subscribers: aemerson, rengolin, jyknight, dsanders, llvm-commits, asl
Differential Revision: http://reviews.llvm.org/D14338
llvm-svn: 252177
This fixes the issue of wrong CFA calculation in the following case:
0x08048400 <+0>: push %ebx
0x08048401 <+1>: sub $0x8,%esp
0x08048404 <+4>: **call 0x8048409 <test+9>**
0x08048409 <+9>: **pop %eax**
0x0804840a <+10>: add $0x1bf7,%eax
0x08048410 <+16>: mov %eax,%ebx
0x08048412 <+18>: call 0x80483f0 <bar>
0x08048417 <+23>: add $0x8,%esp
0x0804841a <+26>: pop %ebx
0x0804841b <+27>: ret
The highlighted instructions are a product of movpc instruction. The call
instruction changes the stack pointer, and pop instruction restores its
value. However, the rule for computing CFA is not updated and is wrong on
the pop instruction. So, e.g. backtrace in gdb does not work when on the pop
instruction. This adds cfi instructions for both call and pop instructions.
cfi_adjust_cfa_offset** instruction is used with the appropriate offset for
setting the rules to calculate CFA correctly.
Patch by Violeta Vukobrat.
Differential Revision: http://reviews.llvm.org/D14021
llvm-svn: 252176
We can conservatively know that CMOV's known bits are the intersection of known bits for each of its operands. This helps PerformCMOVToBFICombine find more opportunities.
I tried hard to create a testcase for this and failed - we have to sufficiently confuse DAG.computeKnownBits which can see through all the cheap tricks I tried to narrow my larger testcase down :(
This code is actually exercised in CodeGen/ARM/bfi.ll, there's just no functional difference because DAG.computeKnownBits gets the right answer in that case.
llvm-svn: 252168
We were correctly skipping dbginfo intrinsics and terminators, but the initial bailout wasn't, causing it to bail out on almost any block.
llvm-svn: 252152
Summary:
GetUnderlyingObjects() can return "null" among its list of objects,
we don't want to deduce that two pointers can point to the same
memory in this case, so filter it out.
Reviewers: anemet
Subscribers: dexonsmith, llvm-commits
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 252149
Summary:
Remove the loop over the uses of the CallSite in ArgumentUsesTracker.
Since we have the `Use *` for actual argument operand, we can just use
pointer subtraction.
The time complexity remains the same though (except for a vararg
argument) -- `std::advance` is O(UseIndex) for the ArgumentList
iterator.
The real motivation is to make a later change adding support for operand
bundles simpler.
Reviewers: reames, chandlerc, nlewycky
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14363
llvm-svn: 252141
The operand layout is slightly different for the atomic
opcodes from the usual MUBUF loads and stores.
This should only fix it on SI/CI. VI is still broken
because it still emits the addr64 replacement.
llvm-svn: 252140
Summary:
The CLR's personality routine passes the pointer to the establisher frame
in RCX, not RDX.
Reviewers: pgavlin, majnemer, rnk
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14343
llvm-svn: 252135
Summary:
llvm-symbolizer understands both PDBs and DWARF, so it is more likely to
succeed at symbolization. If llvm-symbolizer is unavailable, we will
fall back to dbghelp. This also makes our crash traces more similar
between Windows and Linux.
Reviewers: Bigcheese, zturner, chapuni
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D12884
llvm-svn: 252118
With this change, instrumentation code and reader/write
code related to profile data structs are kept strictly
in-sync. THis will be extended to cfe and compile-rt
references as well.
Differential Revision: http://reviews.llvm.org/D13843
llvm-svn: 252113
Summary:
Earlier CaptureTracking would assume all "interesting" operands to a
call or invoke were its arguments. With operand bundles this is no
longer true.
Note: an earlier change got `doesNotCapture` working correctly with
operand bundles.
This change uses DSE to test the changes to CaptureTracking. DSE is a
vehicle for testing only, and is not directly involved in this change.
Reviewers: reames, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14306
llvm-svn: 252095
The generic infrastructure already did a lot of work to decide if the
fixup value is know or not. It doesn't make sense to reimplement a very
basic case: same fragment.
llvm-svn: 252090
Win64 has some strict requirements for the epilogue. As a result, we disable
shrink-wrapping for Win64 unless the block that gets the epilogue is already an
exit block.
Fixes PR24193.
llvm-svn: 252088
Splits PrintLoopPass into a new-style pass and a PrintLoopPassWrapper,
much like we already do for PrintFunctionPass and PrintModulePass.
llvm-svn: 252085
This is part-1 of the patch that replaces all edge weights in MBB by
probabilities, which only adds new interfaces. No functional changes.
Differential revision: http://reviews.llvm.org/D13908
llvm-svn: 252083
Summary:
Data operands of a call or invoke consist of the call arguments, and
the bundle operands associated with the `call` (or `invoke`)
instruction. The motivation for this change is that we'd like to be
able to query "argument attributes" like `readonly` and `nocapture`
for bundle operands naturally.
This change also provides a conservative "implementation" for these
attributes for any bundle operand, and an extension point for future
work.
Reviewers: chandlerc, majnemer, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14305
llvm-svn: 252077
This patch improves the memory folding of the inserted float element for the (V)INSERTPS instruction.
The existing implementation occurs in the DAGCombiner and relies on the narrowing of a whole vector load into a scalar load (and then converted into a vector) to (hopefully) allow folding to occur later on. Not only has this proven problematic for debug builds, it also prevents other memory folds (notably stack reloads) from happening.
This patch removes the old implementation and moves the folding code to the X86 foldMemoryOperand handler. A new private 'special case' function - foldMemoryOperandCustom - has been added to deal with memory folding of instructions that can't just use the lookup tables - (V)INSERTPS is the first of several that could be done.
It also tweaks the memory operand folding code with an additional pointer offset that allows existing memory addresses to be modified, in this case to convert the vector address to the explicit address of the scalar element that will be inserted.
Unlike the previous implementation we now set the insertion source index to zero, although this is ignored for the (V)INSERTPSrm version, anything that relied on shuffle decodes (such as unfolding of insertps loads) was incorrectly calculating the source address - I've added a test for this at insertps-unfold-load-bug.ll
Differential Revision: http://reviews.llvm.org/D13988
llvm-svn: 252074
Summary:
This is intended to make a later change simpler.
Note: adding this bounds checking required fixing `X86FastISel`. As
far I can tell I've preserved original behavior but a careful review
will be appreciated.
Reviewers: reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14304
llvm-svn: 252073
Patch by Slava Klochkov
The key difference between FMA* and FMA*_Int opcodes is that FMA*_Int opcodes are handled more conservatively. It is illegal to commute the 1st operand of FMA*_Int instructions as the upper bits of scalar FMA intrinsic result must be taken from the 1st operand, but such commute transformation would change those upper bits and invalidate the intrinsic's result.
Reviewers: Quentin Colombet, Elena Demikhovsky
Differential Revision: http://reviews.llvm.org/D13710
llvm-svn: 252060
If we have a CMOV, OR and AND combination such as:
if (x & CN)
y |= CM;
And:
* CN is a single bit;
* All bits covered by CM are known zero in y;
Then we can convert this to a sequence of BFI instructions. This will always be a win if CM is a single bit, will always be no worse than the TST & OR sequence if CM is two bits, and for thumb will be no worse if CM is three bits (due to the extra IT instruction).
llvm-svn: 252057
When converting an alias to a non-alias when the aliasee is not
imported, ensure that the linkage type is set to external so that it is
a valid linkage type. Added a test case that exposed this issue.
llvm-svn: 252054
We can often end up with conditional stores that cannot be speculated. They can come from fairly simple, idiomatic code:
if (c & flag1)
*a = x;
if (c & flag2)
*a = y;
...
There is no dominating or post-dominating store to a, so it is not legal to move the store unconditionally to the end of the sequence and cache the intermediate result in a register, as we would like to.
It is, however, legal to merge the stores together and do the store once:
tmp = undef;
if (c & flag1)
tmp = x;
if (c & flag2)
tmp = y;
if (c & flag1 || c & flag2)
*a = tmp;
The real power in this optimization is that it allows arbitrary length ladders such as these to be completely and trivially if-converted. The typical code I'd expect this to trigger on often uses binary-AND with constants as the condition (as in the above example), which means the ending condition can simply be truncated into a single binary-AND too: 'if (c & (flag1|flag2))'. As in the general case there are bitwise operators here, the ladder can often be optimized further too.
This optimization involves potentially increasing register pressure. Even in the simplest case, the lifetime of the first predicate is extended. This can be elided in some cases such as using binary-AND on constants, but not in the general case. Threading 'tmp' through all branches can also increase register pressure.
The optimization as in this patch is enabled by default but kept in a very conservative mode. It will only optimize if it thinks the resultant code should be if-convertable, and additionally if it can thread 'tmp' through at least one existing PHI, so it will only ever in the worst case create one more PHI and extend the lifetime of a predicate.
This doesn't trigger much in LNT, unfortunately, but it does trigger in a big way in a third party test suite.
llvm-svn: 252051
The x86 "sitofp i64 to double" dag combine, in 32-bit mode, lowers sitofp
directly to X86ISD::FILD (or FILD_FLAG). This should not be done in soft-float mode.
llvm-svn: 252042
In my previous change to CVP (251606), I made CVP much more aggressive about trying to constant fold comparisons. This patch is a reversal in direction. Rather than being agressive about every compare, we restore the non-block local restriction for most, and then try hard for compares feeding returns.
The motivation for this is two fold:
* The more I thought about it, the less comfortable I got with the possible compile time impact of the other approach. There have been no reported issues, but after talking to a couple of folks, I've come to the conclusion the time probably isn't justified.
* It turns out we need to know the context to leverage the full power of LVI. In particular, asking about something at the end of it's block (the use of a compare in a return) will frequently get more precise results than something in the middle of a block. This is an implementation detail, but it's also hard to get around since mid-block queries have to reason about possible throwing instructions and don't get to use most of LVI's block focused infrastructure. This will become particular important when combined with http://reviews.llvm.org/D14263.
Differential Revision: http://reviews.llvm.org/D14271
llvm-svn: 252032
There is no point in having invoke safepoints handled differently than the
call safepoints. All relevant decisions could be made by looking at whether
or not gc.result and gc.relocate lay in a same basic block. This change will
allow to lower call safepoints with relocates and results in a different
basic blocks. See test case for example.
Differential Revision: http://reviews.llvm.org/D14158
llvm-svn: 252028
Summary:
The goal of this pass is to perform store-to-load forwarding across the
backedge of a loop. E.g.:
for (i)
A[i + 1] = A[i] + B[i]
=>
T = A[0]
for (i)
T = T + B[i]
A[i + 1] = T
The pass relies on loop dependence analysis via LoopAccessAnalisys to
find opportunities of loop-carried dependences with a distance of one
between a store and a load. Since it's using LoopAccessAnalysis, it was
easy to also add support for versioning away may-aliasing intervening
stores that would otherwise prevent this transformation.
This optimization is also performed by Load-PRE in GVN without the
option of multi-versioning. As was discussed with Daniel Berlin in
http://reviews.llvm.org/D9548, this is inferior to a more loop-aware
solution applied here. Hopefully, we will be able to remove some
complexity from GVN/MemorySSA as a consequence.
In the long run, we may want to extend this pass (or create a new one if
there is little overlap) to also eliminate loop-indepedent redundant
loads and store that *require* versioning due to may-aliasing
intervening stores/loads. I have some motivating cases for store
elimination. My plan right now is to wait for MemorySSA to come online
first rather than using memdep for this.
The main motiviation for this pass is the 456.hmmer loop in SPECint2006
where after distributing the original loop and vectorizing the top part,
we are left with the critical path exposed in the bottom loop. Being
able to promote the memory dependence into a register depedence (even
though the HW does perform store-to-load fowarding as well) results in a
major gain (~20%). This gain also transfers over to x86: it's
around 8-10%.
Right now the pass is off by default and can be enabled
with -enable-loop-load-elim. On the LNT testsuite, there are two
performance changes (negative number -> improvement):
1. -28% in Polybench/linear-algebra/solvers/dynprog: the length of the
critical paths is reduced
2. +2% in Polybench/stencils/adi: Unfortunately, I couldn't reproduce this
outside of LNT
The pass is scheduled after the loop vectorizer (which is after loop
distribution). The rational is to try to reuse LAA state, rather than
recomputing it. The order between LV and LLE is not critical because
normally LV does not touch scalar st->ld forwarding cases where
vectorizing would inhibit the CPU's st->ld forwarding to kick in.
LoopLoadElimination requires LAA to provide the full set of dependences
(including forward dependences). LAA is known to omit loop-independent
dependences in certain situations. The big comment before
removeDependencesFromMultipleStores explains why this should not occur
for the cases that we're interested in.
Reviewers: dberlin, hfinkel
Subscribers: junbuml, dberlin, mssimpso, rengolin, sanjoy, llvm-commits
Differential Revision: http://reviews.llvm.org/D13259
llvm-svn: 252017
Summary: Will be used by the LoopLoadElimination pass.
Reviewers: hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D13258
llvm-svn: 252016
Diego Novillo