Hans Wennborg [Thu, 24 Aug 2017 16:16:07 +0000 (16:16 +0000)]
Merging r311623:
------------------------------------------------------------------------
r311623 | hans | 2017-08-23 18:08:27 -0700 (Wed, 23 Aug 2017) | 11 lines
[DAG] Fix Node Replacement in PromoteIntBinOp
When one operand is a user of another in a promoted binary operation
we may replace and delete the returned value before returning
triggering an assertion. Reorder node replacements to prevent this.
Hans Wennborg [Wed, 23 Aug 2017 21:30:37 +0000 (21:30 +0000)]
Merging r311565:
------------------------------------------------------------------------
r311565 | hans | 2017-08-23 08:43:28 -0700 (Wed, 23 Aug 2017) | 8 lines
LowerAtomic: Don't skip optnone functions; atomic still need lowering (PR34020)
The lowering isn't really an optimization, so optnone shouldn't make a
difference. ARM relies on the pass running when using "-mthread-model
single", because in that mode, it doesn't run AtomicExpand. See bug for
more details.
[IR] AutoUpgrade ModuleFlagBehavior for PIC and PIE level
Summary:
From r303590, ModuleFlagBehavior for PIC and PIE level is changed from
Error to Max. This will cause bitcode compatibility issue when linking
against a bitcode static archive built with old compiler.
Add an auto-ugprade path to upgrade the the ModuleFlagBehavior in the
old bitcode to match the new one so IRLinker can link them.
Hans Wennborg [Tue, 22 Aug 2017 17:44:59 +0000 (17:44 +0000)]
Merging r311263:
------------------------------------------------------------------------
r311263 | ctopper | 2017-08-19 15:02:02 -0700 (Sat, 19 Aug 2017) | 1 line
[AVX512] Use alignedstore256 in a pattern that's emitting a 256-bit movaps from an extract subvector operation.
------------------------------------------------------------------------
[X86] Prevent several calls to ISD::isConstantSplatVector from returning a narrower APInt than the original scalar type
ISD::isConstantSplatVector can shrink to the smallest splat width. But we don't check the size of the resulting APInt at all. This can cause us to misinterpret the results.
This patch just adds a flag to prevent the APInt from changing width.
When lowering a VLA, we emit a __chstk call. However, this call can
internally clobber CPSR. We did not mark this register as an ImpDef,
which could potentially allow a comparison to be hoisted above the call
to `__chkstk`. In such a case, the CPSR could be clobbered, and the
check invalidated. When the support was initially added, it seemed that
the call would take care of preventing CPSR from being clobbered, but
this is not the case. Mark the register as clobbered to fix a possible
state corruption.
------------------------------------------------------------------------
[SelectionDAG] Teach the vector-types operand scalarizer about SETCC
When v1i1 is legal (e.g. AVX512) the legalizer can reach
a case where a v1i1 SETCC with an illgeal vector type operand
wasn't scalarized (since v1i1 is legal) but its operands does
have to be scalarized. This used to assert because SETCC was
missing from the vector operand scalarizer.
This patch attemps to teach the legalizer to handle these cases
by scalazring the operands, converting the node into a scalar
SETCC node.
[llvm-dlltool] Fix creating stdcall/fastcall import libraries for i386
Hook up the -k option (that in the original GNU dlltool removes the
@n suffix from the symbol that the final executable ends up linked to).
In llvm-dlltool, make sure that functions end up with the undecorate
name type if this option is set and they are decorated. In mingw, when
creating import libraries from def files instead of creating an import
library as a side effect of linking a DLL, the symbol names in the def
contain the stdcall/fastcall decoration (but no leading underscore).
By setting the undecorate name type, a linker linking to the import
library will omit the decoration from the DLL import entry.
With this in place, mingw-w64 for i386 built with llvm-dlltool/clang
produces import libraries that actually work.
[cmake] Expose the dependencies of ExecutionEngine as PUBLIC
Expose the dependencies of LLVMExecutionEngine library as PUBLIC rather
than PRIVATE when building a shared library. This is necessary because
the library is not contained but exposes API of other LLVM libraries via
its headers.
This causes other libraries to fail to link if the linker verifies for
correctness of -l flags (i.e. fails on indirect dependencies). This e.g.
happens when building LLDB against shared LLVM:
lib64/liblldbExpression.a(IRExecutionUnit.cpp.o):(.data.rel.ro._ZTIN4llvm18MCJITMemoryManagerE[_ZTIN4llvm18MCJITMemoryManagerE]+0x10): undefined reference to `typeinfo for llvm::RuntimeDyld::MemoryManager'
lib64/liblldbExpression.a(IRExecutionUnit.cpp.o):(.data.rel.ro._ZTVN4llvm18MCJITMemoryManagerE[_ZTVN4llvm18MCJITMemoryManagerE]+0x60): undefined reference to `llvm::RuntimeDyld::MemoryManager::anchor()'
lib64/liblldbExpression.a(IRExecutionUnit.cpp.o):(.data.rel.ro._ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE[_ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE]+0x48): undefined reference to `llvm::RTDyldMemoryManager::deregisterEHFrames()'
lib64/liblldbExpression.a(IRExecutionUnit.cpp.o):(.data.rel.ro._ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE[_ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE]+0x60): undefined reference to `llvm::RuntimeDyld::MemoryManager::anchor()'
lib64/liblldbExpression.a(IRExecutionUnit.cpp.o):(.data.rel.ro._ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE[_ZTVN12lldb_private15IRExecutionUnit13MemoryManagerE]+0xd0): undefined reference to `llvm::JITSymbolResolver::anchor()'
collect2: error: ld returned 1 exit status
Declaring the dependencies as PUBLIC guarantees that any package using
the ExecutionEngine library will also get explicit -l flags for
the dependent libraries guaranteeing that the symbols exposed in headers
could be resolved.
[Inliner] Fix a nasty bug when inlining a non-recursive trace of
a function into itself.
We tried to fix this before in r306495 but that got reverted as the
assert was actually hit.
This fixes the original bug (which we seem to have lost track of with
the revert) by blocking a second remapping when the function being
inlined is also the caller and the remapping could succeed but
erroneously.
The included test case would actually load from an inlined copy of the
alloca before this change, failing to load the stored value and
miscompiling.
Many thanks to Richard Smith for diagnosing a user miscompile to this
bug, and to Kyle for the first attempt and initial analysis and David Li
for remembering the issue and how to fix it and suggesting the patch.
I'm just stitching it together and landing it. =]
------------------------------------------------------------------------
[ARM] Check the right order for halves of VZIP/VUZP if both parts are used
This is the exact same fix as in SVN r247254. In that commit, the fix was
applied only for isVTRNMask and isVTRN_v_undef_Mask, but the same issue
is present for VZIP/VUZP as well.
[dfsan] Add explicit zero extensions for shadow parameters in function wrappers.
In the case where dfsan provides a custom wrapper for a function,
shadow parameters are added for each parameter of the function.
These parameters are i16s. For targets which do not consider this
a legal type, the lack of sign extension information would cause
LLVM to generate anyexts around their usage with phi variables
and calling convention logic.
Address this by introducing zero exts for each shadow parameter.
[X86][AVX512] Choose correct registers in vpbroadcastb/w
Fixes the vpbroadcastb/w instructions which use GPRs as source operands, to use the correct registers.
The full GPR should be used, and not the subregister, as it happens before the patch.
[VirtRegRewriter] Properly model the register liveness on undef subreg definition
Undef subreg definition means that the content of the super register
doesn't matter at this point. While that's true for virtual registers,
this may not hold when replacing them with actual physical registers.
Indeed, some part of the physical register may be coalesced with the
related virtual register and thus, the values for those parts matter and
must be live.
The fix consists in checking whether or not subregs of the physical register
being assigned to an undef subreg definition are live through that def and
insert an implicit use if they are. Doing so, will keep them alive until
that point like they should be.
E.g., let vreg14 being assigned to R0_R1 then
%vreg14:gsub_0<def,read-undef> = COPY %R0 ; <-- R1 is still live here
%vreg14:gsub_1<def> = COPY %R1
Before this changes, the rewriter would change the code into:
%R0<def> = KILL %R0, %R0_R1<imp-def> ; <-- this tells R1 is redefined
%R1<def> = KILL %R1, %R0_R1<imp-def>, %R0_R1<imp-use> ; this value of this R1
; is believed to come
; from the previous
; instruction
Because of this invalid liveness, later pass could make wrong choices and in
particular clobber live register as it happened with the register scavenger in
llvm.org/PR34107
Now we would generate:
%R0<def> = KILL %R0, %R0_R1<imp-def>, %R0_R1<imp-use> ; This tells R1 needs to
; reach this point
%R1<def> = KILL %R1, %R0_R1<imp-def>, %R0_R1<imp-use>
The bug has been here forever, it got exposed recently because the register
scavenger got smarter.
When creating an import library from lld, the cases with
Name != ExtName shouldn't end up as a weak alias, but as a real
export of the new name, which is what actually is exported from
the DLL.
This restores the behaviour of renamed exports to what it was in
4.0.
The other half of this commit, including test, goes into lld.
test-release.sh: Move test-suite setup to beginning of the script
Summary:
We want to catch failures early before do the full 3 stage build.
The goal here is to avoid running through the whole build process and have
it fail at the end (and not create the binary packages), just because
some prerequisites failed to install.
Remove RISCV from LLVM_ALL_TARGETS in CMakeLists.txt
It was mistakenly added to that list in D23560 (committed in rL285712). RISCV
is an experimental backend and should never have been in that list, I
mistakenly interpreted LLVM_ALL_TARGETS as a list of all targets rather than
targets to build by default. Unfortunately, because of this the RISCV backend
has been building by default when it shouldn't be.
This commet adds a description comment, which should help to avoid such
mistakes in the future.
See my message to llvm-dev for more information and analysis
<http://lists.llvm.org/pipermail/llvm-dev/2017-August/116347.html>.
[X86] When handling addcarry intrinsic, create the flag result with the correct type so we don't crash if we use a memory instruction
Summary:
Previously we were creating the flag result with MVT::Other which is interpretted as a Chain node. If we used a memory form of the instruction we would end up with a copyToReg that consumed the chain result of the adcx instruction instead of the flag result.
Pretty sure we should be using MVT::i32 here, that's what we do other places we create these node types.
[BDCE] clear poison generators after turning a value into zero (PR33695, PR34037)
nsw, nuw, and exact carry implicit assumptions about their operands, so we need
to clear those after trivializing a value. We decided there was no danger for
llvm.assume or metadata, so there's just a comment about that.
This fixes miscompiles as shown in:
https://bugs.llvm.org/show_bug.cgi?id=33695
https://bugs.llvm.org/show_bug.cgi?id=34037
The assert was added with r310779 and is usually correct,
but as the test shows, not always. The 'volatile' on the
load is needed to expose the faulty path because without
it, DemandedBits would return that the load is just dead
rather than not demanded, and so we wouldn't hit the
bogus assert.
Also, since the lambda is just a single-line now, get rid
of it and inline the DB.isAllOnesValue() calls.
This should fix (prevent execution of a faulty assert):
https://bugs.llvm.org/show_bug.cgi?id=34179
[lit] Fix UnboundLocalError for invalid shtest redirects
Replace the incorrect variable reference when invalid redirect is used.
This fixes the following issue:
File "/usr/src/llvm/utils/lit/lit/TestRunner.py", line 316, in processRedirects
raise InternalShellError(cmd, "Unsupported redirect: %r" % (r,))
UnboundLocalError: local variable 'r' referenced before assignment
which in turn broke shtest-shell.py and max-failures.py lit tests.
The breakage was introduced during refactoring in rL307310.
Revert "[lit] Remove dead code not referenced in the LLVM SVN repo."
Summary:
This reverts rL306623, which removed `FileBasedTest`, an abstract base class,
but did not also remove the usages of that class in the lit unit tests.
The revert fixes four test failures in the lit unit test suite.
Test plan:
As per the instructions in `utils/lit/README.txt`, run the lit unit
test suite:
Summary:
Whereas rL299560 and rL309071 call `parallelism_groups.items()`, under the
assumption that `parallelism_groups` is a `dict` type, the default
parameter for that attribute is a `list`. Change the default to a
`dict` for type correctness.
This regression in the unit tests would have been caught if the
unit tests were being run continously. It also would have been caught
if the lit project used a Python type checker such as `mypy`.
Test Plan:
As per the instructions in `utils/lit/README.txt`, run the lit unit
test suite:
[lit] Fix race between shtest-shell and max-failures tests
Previously these tests would use the same Output directory leading to
flaky non-deterministic failures.
------------------------------------------------------------------------
[SelectionDAG] When scalarizing vselect, don't assert on
a legal cond operand.
When scalarizing the result of a vselect, the legalizer currently expects
to already have scalarized the operands. While this is true for the true/false
operands (which have the same type as the result), it is not case for the
condition operand. On X86 AVX512, v1i1 is legal - this leads to operations such
as '< N x type> vselect < N x i1> < N x type> < N x type>' where < N x type > is
illegal to hit an assertion during the scalarization.
The handling is similar to r205625.
This also exposes the fact that (v1i1 extract_subvector) should be legal
and selectable on AVX512 - We do this by custom lowering to vector_extract_elt.
This still leaves us in some cases with redundant dag nodes which will be
combined in a separate soon to come patch.
Hans Wennborg [Thu, 10 Aug 2017 17:33:55 +0000 (17:33 +0000)]
Merging r310481, r310492 and r310510:
------------------------------------------------------------------------
r310481 | davide | 2017-08-09 08:13:50 -0700 (Wed, 09 Aug 2017) | 8 lines
[ValueTracking] Honour recursion limit.
The recently improved support for `icmp` in ValueTracking
(r307304) exposes the fact that `isImplied` condition doesn't
really bail out if we hit the recursion limit (and calls
`computeKnownBits` which increases the depth and asserts).
[SCEV] Limit max size of AddRecExpr during evolving
When SCEV calculates product of two SCEVAddRecs from the same loop, it
tries to combine them into one big AddRecExpr. If the sizes of the initial
SCEVs were `S1` and `S2`, the size of their product is `S1 + S2 - 1`, and every
operand of the resulting SCEV is combined from operands of initial SCEV and
has much higher complexity than they have.
As result, if we try to calculate something like:
%x1 = {a,+,b}
%x2 = mul i32 %x1, %x1
%x3 = mul i32 %x2, %x1
%x4 = mul i32 %x3, %x2
...
The size of such SCEVs grows as `2^N`, and the arguments
become more and more complex as we go forth. This leads
to long compilation and huge memory consumption.
This patch sets a limit after which we don't try to combine two
`SCEVAddRecExpr`s into one. By default, max allowed size of the
resulting AddRecExpr is set to 16.
Clean up after my misguided attempt in r304267 to "fix" CMP_SWAP
returning an uninitialized status value.
- I was always using tMOVi8 to zero the status register which cannot
encode higher register numbers and llvm would silently miscompile)
- Nobody was ever looking at that status value outside the expansion.
ARMDAGToDAGISel::SelectCMP_SWAP() the only place creating CMP_SWAP
instructions was not mapping anything to it. (The cmpxchg status value
from llvm IR is lowered to a manual comparison after the CMP_SWAP)
So this:
- Renames the register from "status" to "temp" it make it obvious that
it isn't used outside the expansion.
- Remove the zeroing status/temp register.
- Keep the live-in list improvements from r304267
[X86] Enable isel to use the PAUSE instruction even when SSE2 is disabled
Summary:
On older processors this instruction encoding is treated as a NOP.
MSVC doesn't disable intrinsics based on features the way clang/gcc does. Because the PAUSE instruction encoding doesn't crash older processors, some software out there uses these intrinsics without checking for SSE2.
This change also seems to also be consistent with gcc behavior.
[ArgPromotion] Preserve alignment of byval argument in new alloca
The frontend may have requested a higher alignment for any reason, and
downstream optimizations may already have taken advantage of it. We
should keep the same alignment when moving the allocation from the
parameter area to the local variable area.
[Cloning] Move distinct GlobalVariable debug info metadata in CloneModule
Duplicating the distinct Subprogram and CU metadata nodes seems like the incorrect thing to do in CloneModule for GlobalVariable debug info. As it results in the scope of the GlobalVariable DI no longer being consistent with the rest of the module, and the new CU is absent from llvm.dbg.cu.
Fixed by adding RF_MoveDistinctMDs to MapMetadata flags for GlobalVariables.
Current unit test IR after clone:
```
@gv = global i32 1, comdat($comdat), !dbg !0, !type !5
[StackColoring] Update AliasAnalysis information in stack coloring pass
Stack coloring pass need to maintain AliasAnalysis information when merging stack slots of different types.
Actually, there is a FIXME comment in StackColoring.cpp
// FIXME: In order to enable the use of TBAA when using AA in CodeGen,
// we'll also need to update the TBAA nodes in MMOs with values
// derived from the merged allocas.
But, TBAA has been already enabled in CodeGen without fixing this pass.
The incorrect TBAA metadata results in recent failures in bootstrap test on ppc64le (PR33928) by allowing unsafe instruction scheduling.
Although we observed the problem on ppc64le, this is a platform neutral issue.
This patch makes the stack coloring pass maintains AliasAnalysis information when merging multiple stack slots.
------------------------------------------------------------------------
[StackColoring] Update AliasAnalysis information in stack coloring pass (part 2)
This patch is update after the first patch (https://reviews.llvm.org/rL309651) based on the post-commit comments.
Stack coloring pass need to maintain AliasAnalysis information when merging stack slots of different types.
Actually, there is a FIXME comment in StackColoring.cpp
// FIXME: In order to enable the use of TBAA when using AA in CodeGen,
// we'll also need to update the TBAA nodes in MMOs with values
// derived from the merged allocas.
But, TBAA has been already enabled in CodeGen without fixing this pass.
The incorrect TBAA metadata results in recent failures in bootstrap test on ppc64le (PR33928) by allowing unsafe instruction scheduling.
Although we observed the problem on ppc64le, this is a platform neutral issue.
This patch makes the stack coloring pass maintains AliasAnalysis information when merging multiple stack slots.
This patch fixes PR33928.
------------------------------------------------------------------------
rL306209 taught SelectionDAG how to add the dereferenceable flag when
expanding memcpy and memmove. The fix however contained a nit where
the offset + size was constructed as an APInt of PointerSize rather
than PointerSizeInBits.
This lead to isDereferenceableAndAlignedPointer() get truncated values or
values which would be sign extended within that function leading to
incorrect results.
[AArch64] Rewrite stack frame handling for win64 vararg functions
The previous attempt, which made do with a single offset in
computeCalleeSaveRegisterPairs, wasn't quite enough. The previous
attempt only worked as long as CombineSPBump == true (since the
offset would be adjusted later in fixupCalleeSaveRestoreStackOffset).
Instead include the size for the fixed stack area used for win64
varargs in calculations in emitPrologue/emitEpilogue. The stack
consists of mainly three parts;
- AFI->getLocalStackSize()
- AFI->getCalleeSavedStackSize()
- FixedObject
Most of the places in the code which previously used the CSStackSize
now use PrologueSaveSize instead, which is the sum of the latter
two, while some cases which need exactly the middle one use
AFI->getCalleeSavedStackSize() explicitly instead of a local variable.
In addition to moving the offsetting into emitPrologue/emitEpilogue
(which fixes functions with CombineSPBump == false), also set the
frame pointer to point to the right location, where the frame pointer
and link register actually are stored. In addition to the prologue/epilogue,
this also requires changes to resolveFrameIndexReference.
Add tests for a function that keeps a frame pointer and another one
that uses a VLA.
Previously, the created object files for the import library were broken.
Write the symbol table before the string table. Simplify the code by
using a separate variable Prefix instead of duplicating a few lines.
Also update the coff-weak-exports to actually check that the generated
weak symbols can be found as intended.
[IPSCCP] Guard a user of getInitializer with hasDefinitiveInitializer
We are not allowed to reason about an initializer value without first
consulting hasDefinitiveInitializer.
------------------------------------------------------------------------
[AArch64] Tie source and destination operands for AESMC/AESIMC.
Summary:
Most CPUs implementing AES fusion require instruction pairs of the form
AESE Vn, _
AESMC Vn, Vn
and
AESD Vn, _
AESIMC Vn, Vn
The constraint is added to AES(I)MC instructions which use the result of
an AES(E|D) instruction by using AES(I)MCTrr pseudo instructions, which
constraint source and destination registers to be the same.
A nice side effect of this change is that now all possible pairs are
scheduled back-to-back on the exynos-m1 for the misched-fusion-aes.ll
test case.
I had to update aes_load_store. The version I added initially was very
reduced and with the new constraint, AESE/AESMC could not be scheduled
back-to-back. I updated the test to be more realistic and still expose
the same scheduling problem as the initial test case.
[ConstantFolder] Don't try to fold gep when the idx is a vector.
The code in ConstantFoldGetElementPtr() assumes integers, and
therefore it crashes trying to get the integer bidwith of a vector
type (in this case <4 x i32>. I just changed the code to prevent
the folding in case of vectors and I didn't bother to generalize
as this doesn't seem to me something that really happens in
practice, but I'm willing to change the patch if you think
it's worth it.
This is hard to trigger from -instsimplify or -instcombine
only as the second instruction is dead, so the test uses loop-unroll.
[OCaml] Install dynamic libraries in 'stubdirs' directory
Install the OCaml dynamic libraries in the 'stubdirs' directory rather
than the llvm subdirectory in order to fix running executables created
by ocamlc. Otherwise, the executables fail to run being unable to locate
the libraries (unless the LLVM directory is explicitly added to
LD_LIBRARY_PATH).
The staging directories are not altered since they work for our
development setup anyway, and installing into two directories would
unnecessarily make the code more complex.
Detect [/-][DU]NDEBUG in CMAKE_C_FLAGS* and pass them through to ocamlc.
This is necessary because their value might affect visibility of dump
functions in LLVM and ocamlc uses its own compiler and flags by default.
[OCaml] Fix undefined reference to LLVMDumpType() with NDEBUG
Account for the possibility of LLVMDumpType() not being available with
NDEBUG in the OCaml bindings. If it is not built into LLVM, make
the dump function raise an exception.
Since rL293359, the dump functions are built only if either NDEBUG is
not defined, or LLVM_ENABLE_DUMP is defined. As a result, if the dump
functions are not built in LLVM, the dynamic OCaml libraries fail to
load due to undefined LLVMDumpType symbol.
[X86] Don't lie about legality to TLI's demanded bits.
Like r309323, X86 had a typo where it passed the wrong flags to TLO.
Found by inspection; I haven't been able to tickle this into having
observable behavior. I don't think it does, given that X86 doesn't have
custom demanded bits logic, and the generic logic doesn't have a lot of
exposure to illegal constructs.
------------------------------------------------------------------------
[AArch64] Fix legality info passed to demanded bits for TBI opt.
The (seldom-used) TBI-aware optimization had a typo lying dormant since
it was first introduced, in r252573: when asking for demanded bits, it
told TLI that it was running after legalize, where the opposite was
true.
This is an important piece of information, that the demanded bits
analysis uses to make assumptions about the node. r301019 added such an
assumption, which was broken by the TBI combine.
Instead, pass the correct flags to TLO.
------------------------------------------------------------------------
The X86 tail call eligibility logic was correct when it was written, but
the addition of inalloca and argument copy elision broke its
assumptions. It was assuming that fixed stack objects were immutable.
Currently, we aim to emit a tail call if no arguments have to be
re-arranged in memory. This code would trace the outgoing argument
values back to check if they are loads from an incoming stack object.
If the stack argument is immutable, then we won't need to store it back
to the stack when we tail call.
Fortunately, stack objects track their mutability, so we can just make
the obvious check to fix the bug.
This was http://crbug.com/749826
------------------------------------------------------------------------
Fix conditional tail call branch folding when both edges are the same
The conditional tail call logic did the wrong thing when both
destinations of a conditional branch were the same:
BB#1: derived from LLVM BB %entry
Live Ins: %EFLAGS
Predecessors according to CFG: BB#0
JE_1 <BB#5>, %EFLAGS<imp-use,kill>
JMP_1 <BB#5>
BB#5: derived from LLVM BB %sw.epilog
Predecessors according to CFG: BB#1
TCRETURNdi64 <ga:@mergeable_conditional_tailcall>, 0, ...
We would fold the JE_1 to a TCRETURNdi64cc, and then remove our BB#5
successor. Then BB#5 would be deleted as it had no predecessors, leaving
a dangling "JMP_1 <BB#5>" reference behind to cause assertions later.
This patch checks that both conditional branch destinations are
different before doing the transform. The standard branch folding logic
is able to remove both the JMP_1 and the JE_1, and for my test case we
end up forming a better conditional tail call later.
JumpThreading claims to preserve LVI, but it doesn't preserve
the analyses which LVI holds a reference to (e.g. the Dominator).
In the current pass manager infrastructure, after JT runs, the
PM frees these analyses (including DominatorTree) but preserves
LVI.
CorrelatedValuePropagation runs immediately after and queries
a corrupted domtree, causing weird miscompiles.
This commit disables the preservation of LVI for the time being.
Eventually, we should either move LVI to a proper dependency
tracking mechanism (i.e. an analyses shouldn't hold references
to other analyses and compute them on demand if needed), or
we should teach all the passes preserving LVI to preserve the
analyses LVI depends on.
The new pass manager has a mechanism to invalidate LVI in case
one of the analyses it depends on becomes invalid, so this problem
shouldn't exist (at least not in this immediate form), but handling
of analyses holding references is still a very delicate subject.
[Bash-autocompletion] Show HelpText with possible flags
Summary:
`clang --autocomplete=-std` will show
```
-std: Language standard to compile for
-std= Language standard to compile for
-stdlib= C++ standard library to use
```
after this change.
However, showing HelpText with completion in bash seems super tricky, so
this feature will be used in other shells (fish, zsh...).
Improve DAGTypeLegalizer::convertMask's isSETCCorConvertedSETCC assertion to properly check for any mixture of SETCC or BUILD_VECTOR of constants, or a logical mask op of them.
------------------------------------------------------------------------
[Sparc] invalid adjustments in TLS_LE/TLS_LDO relocations removed
Summary:
Some SPARC TLS relocations were applying nontrivial adjustments
to zero value, leading to unexpected non-zero values in ELF and then
Solaris linker failures.
There's no need for these to be part of a class since
they are immediately replaced. New unreachable hit in
existing tests.'
------------------------------------------------------------------------
[AArch64] Reserve a 16 byte aligned amount of fixed stack for win64 varargs
Create a dummy 8 byte fixed object for the unused slot below the first
stored vararg.
Alternative ideas tested but skipped: One could try to align the whole
fixed object to 16, but I haven't found how to add an offset to the stack
frame used in LowerWin64_VASTART.
If only the size of the fixed stack object size is padded but not the offset, via
MFI.CreateFixedObject(alignTo(GPRSaveSize, 16), -(int)GPRSaveSize, false),
PrologEpilogInserter crashes due to "Attempted to reset backwards range!".
This fixes misconceptions about where registers are spilled, since
AArch64FrameLowering.cpp assumes the offset from fixed objects is
aligned to 16 bytes (and the Win64 case there already manually aligns
the offset to 16 bytes).
This fixes cases where local stack allocations could overwrite callee
saved registers on the stack.
[CodeGenPrepare] Cut off FindAllMemoryUses if there are too many uses.
This avoids excessive compile time. The case I'm looking at is
Function.cpp from an old version of LLVM that still had the giant memcmp
string matcher in it. Before r308322 this compiled in about 2 minutes,
after it, clang takes infinite* time to compile it. With this patch
we're at 5 min, which is still bad but this is a pathological case.
The cut off at 20 uses was chosen by looking at other cut-offs in LLVM
for user scanning. It's probably too high, but does the job and is very
unlikely to regress anything.
Fixes PR33900.
* I'm impatient and aborted after 15 minutes, on the bug report it was
killed after 2h.
------------------------------------------------------------------------
[X86][CGP] Reduce memcmp() expansion to 2 load pairs (PR33914)
D35067/rL308322 attempted to support up to 4 load pairs for memcmp inlining which resulted in regressions for some optimized libc memcmp implementations (PR33914).
Until we can match these more optimal cases, this patch reduces the memcmp expansion to a maximum of 2 load pairs (which matches what we do for -Os).
This patch should be considered for the 5.0.0 release branch as well
test-release.sh: Fix phase2 and phase3 binary comparision
Summary:
scudo_utils.cpp.o from compiler-rt has one of the host compiler's builtin
include paths stored in the .debug_line section. So we need to do
sed 's,Phase1,Phase2,g` on the Phase2 object file so it matches Phase3.
../llvm-project/llvm/lib/Support/ErrorHandling.cpp:172:51: warning:
ignoring return value of ‘ssize_t write(int, const void*, size_t)’,
declared with attribute warn_unused_result [-Wunused-result]
(void)::write(2, OOMMessage, strlen(OOMMessage));
Work around the warning by storing the return value in a variable and
casting that to void instead. We already did this for the other write()
call in this file.
------------------------------------------------------------------------
[LoopUtils] Add an extra parameter OpValue to propagateIRFlags function,
If OpValue is non-null, we only consider operations similar to OpValue
when intersecting.
[SimplifyCFG] Defer folding unconditional branches to LateSimplifyCFG if it can destroy canonical loop structure.
Summary:
When simplifying unconditional branches from empty blocks, we pre-test if the
BB belongs to a set of loop headers and keep the block to prevent passes from
destroying canonical loop structure. However, the current algorithm fails if
the destination of the branch is a loop header. Especially when such a loop's
latch block is folded into loop header it results in additional backedges and
LoopSimplify turns it into a nested loop which prevent later optimizations
from being applied (e.g., loop unrolling and loop interleaving).
This patch augments the existing algorithm by further checking if the
destination of the branch belongs to a set of loop headers and defer
eliminating it if yes to LateSimplifyCFG.
[LV] Test once if vector trip count is zero, instead of twice
Generate a single test to decide if there are enough iterations to jump to the
vectorized loop, or else go to the scalar remainder loop. This test compares the
Scalar Trip Count: if STC < VF * UF go to the scalar loop. If
requiresScalarEpilogue() holds, at-least one iteration must remain scalar; the
rest can be used to form vector iterations. So in this case the test checks
instead if (STC - 1) < VF * UF by comparing STC <= VF * UF, and going to the
scalar loop if so. Otherwise the vector loop is entered for at-least one vector
iteration.
This test covers the case where incrementing the backedge-taken count will
overflow leading to an incorrect trip count of zero. In this (rare) case we will
also avoid the vector loop and jump to the scalar loop.
This patch simplifies the existing tests and effectively removes the basic-block
originally named "min.iters.checked", leaving the single test in block
"vector.ph".
Original observation and initial patch by Evgeny Stupachenko.
[PM/LCG] Follow-up fix to r308088 to handle deletion of library
functions.
In the prior commit, we provide ordering to the LCG between functions
and library function definitions that they might begin to call through
transformations. But we still would delete these library functions from
the call graph if they became dead during inlining.
While this immediately crashed, it also exposed a loss of information.
We shouldn't remove definitions of library functions that can still
usefully participate in the LCG-powered CGSCC optimization process. If
new call edges are formed, we want to have definitions to be called.
We can still remove these functions if truly dead using global-dce, etc,
but removing them during the CGSCC walk is premature.
This fixes a crash in the new PM when optimizing some unusual libraries
that end up with "internal" lib functions such as the code in the "R"
language's libraries.
projects / llvm / log