Summary:
Fixes a FIXME in MachineSinking. Instead of using the simple heuristics in
isPostDominatedBy, use the real MachinePostDominatorTree and MachineLoopInfo.
The old heuristics caused instructions to sink unnecessarily, and might create
register pressure.
This is the second try of the fix. The first one (D4814) caused a performance
regression due to failing to sink instructions out of loops (PR21115). This
patch fixes PR21115 by sinking an instruction from a deeper loop to a shallower
one regardless of whether the target block post-dominates the source.
Thanks Alexey Volkov for reporting PR21115!
Test Plan:
Added a NVPTX codegen test to verify that our change prevents the backend from
over-sinking. It also shows the unnecessary register pressure caused by
over-sinking.
Added an X86 test to verify we can sink instructions out of loops regardless of
the dominance relationship. This test is reduced from Alexey's test in PR21115.
Updated an affected test in X86.
Also ran SPEC CINT2006 and llvm-test-suite for compilation time and runtime
performance. Results are attached separately in the review thread.
Reviewers: Jiangning, resistor, hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, bruno, volkalexey, llvm-commits, meheff, eliben, jholewinski
Differential Revision: http://reviews.llvm.org/D5633
llvm-svn: 219773
in GetFullUnwindPlanForFrame() - the code was mostly checking
that we had an active Process and ABI but not always.
clang static analyzer fixit.
llvm-svn: 219772
Early attempts to support AAPCS bare metal MachO targets based the decision on
the CPU being compiled for. This was not a particularly great idea and we've
got a better option now, but this check remained.
No functional change for any target we care about.
llvm-svn: 219767
The various ways to create an ASTUnit all take a refcounted pointer to
a diagnostics engine as an argument, and if it isn't pointing at
anything they initialize it. This is a pretty confusing API, and it
really makes more sense for the caller to initialize the thing since
they control the lifetime anyway.
This fixes the one caller that didn't bother initializing the pointer
and asserts that the argument is initialized.
llvm-svn: 219752
This is a follow up to commit r219742. It removes the CCInMI variable
and accesses the CC in CSCINC directly. In the case of a conditional
branch accessing the CC with CCInMI was wrong.
llvm-svn: 219748
Summary:
AddressSanitizer currently doesn't support this configuration, and binaries
built with it will just get into an infinite loop during startup.
Test Plan: Includes an automated test.
Reviewers: samsonov
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D5764
llvm-svn: 219744
Peephole optimization that generates a single conditional branch
for csinc-branch sequences like in the examples below. This is
possible when the csinc sets or clears a register based on a condition
code and the branch checks that register. Also the condition
code may not be modified between the csinc and the original branch.
Examples:
1. Convert csinc w9, wzr, wzr, <CC>;tbnz w9, #0, 0x44
to b.<invCC>
2. Convert csinc w9, wzr, wzr, <CC>; tbz w9, #0, 0x44
to b.<CC>
rdar://problem/18506500
llvm-svn: 219742
A few minor changes to prevent @llvm.assume from interfering with loop
vectorization. First, treat @llvm.assume like the lifetime intrinsics, which
are scalarized (but don't otherwise interfere with the legality checking).
Second, ignore the cost of ephemeral instructions in the loop (these will go
away anyway during CodeGen).
Alignment assumptions and other uses of @llvm.assume can often end up inside of
loops that should be vectorized (this is not uncommon for assumptions generated
by __attribute__((align_value(n))), for example).
llvm-svn: 219741
Patch to provide shuffle decodes and asm comments for the sse pslldq/psrldq SSE2/AVX2 byte shift instructions.
Differential Revision: http://reviews.llvm.org/D5598
llvm-svn: 219738
When lazily constructing static member variable declarations (when
the vtable optimization fires and the definition of the type is omitted
(or built later, lazily), but the out of line definition of the static
member is provided and must be described in debug info) ensure we use
the canonical declaration when computing the file, line, etc for that
declaration (rather than the definition, which is also a declaration,
but not the canonical one).
llvm-svn: 219736
Thumb1 has legitimate reasons for preferring 32-bit alignment of types
i1/i8/i16, since the 16-bit encoding of "add rD, sp, #imm" requires #imm to be
a multiple of 4. However, this is a trade-off betweem code size and RAM usage;
the DataLayout string is not the best place to represent it even if desired.
So this patch removes the extra Thumb requirements, hopefully making ARM and
Thumb completely compatible in this respect.
llvm-svn: 219735
Thumb1 has legitimate reasons for preferring 32-bit alignment of types
i1/i8/i16, since the 16-bit encoding of "add rD, sp, #imm" requires #imm to be
a multiple of 4. However, this is a trade-off betweem code size and RAM usage;
the DataLayout string is not the best place to represent it even if desired.
So this patch removes the extra Thumb requirements, hopefully making ARM and
Thumb completely compatible in this respect.
llvm-svn: 219734
There's no hard requirement on LLVM to align local variable to 32-bits, so the
Thumb1 frame handling needs to be able to deal with variables that are only
naturally aligned without falling over.
llvm-svn: 219733
Specifically, avoid typo-correcting the variable name into a type before
typo-correcting the actual type name in the declaration. Doing so
results in a very unpleasant cascade of errors, with the typo correction
of the actual type name being buried in the middle.
llvm-svn: 219732
This implements Host::LaunchProcess for windows, and in doing so
does some minor refactor to move towards a more modular process
launching design.
The original motivation for this is that launching processes on
windows needs some very windows specific code, which would live
most appropriately in source/Host/windows somewhere. However,
there is already some common code that all platforms use when
launching a process before delegating to the platform specific
stuff, which lives in source/Host/common/Host.cpp which would
be nice to reuse without duplicating.
This commonality has been abstracted into MonitoringProcessLauncher,
a class which abstracts out the notion of launching a process using
an arbitrary algorithm, and then monitoring it for state changes.
The windows specific launching code lives in ProcessLauncherWindows,
and the posix specific launching code lives in ProcessLauncherPosix.
When launching a process MonitoringProcessLauncher is created, and
then an appropriate delegate launcher is created and given to the
MonitoringProcessLauncher.
Reviewed by: Greg Clayton
Differential Revision: http://reviews.llvm.org/D5781
llvm-svn: 219731
a FileSpec, UUID, and ArchSpec, and it's opening all the kexts
with the same bundle id to see if they're a match, don't set
the Arch in the ModuleSpec. If Module::GetObjectFile() sees
that the architecture of the kext is a mismatch for the arch
we're looking for, it'll spew a warning message to the dev (r217251).
Rely on the UUID match to get the correct file if we have
a UUID -- we'll get no warning if it's a mismatch.
<rdar://problem/18641477>
llvm-svn: 219728
This is mostly a copy of the existing FastISel GEP code, but on AArch64 we bail
out even for simple cases, because the standard fastEmit functions don't cover
MUL and ADD is lowered inefficientily.
llvm-svn: 219726
Because we use cast<> at the beginning of this function, it will
abort there if a given atom is not a DefinedAtom.
In the switch statement, we checked if a given atom is a DefinedAtom
again by evaluating definition() == Atom::definitionRegular.
This was always true. So we can remove the outer switch statement.
llvm-svn: 219724
We have to #define _WANT_RTENTRY to get struct rtentry, and add headers
that it depends on.
SIOCADDRT and SIOCDELRT also do not exist in FreeBSD 10.1 and were removed
in FreeBSD rev. 263203:
Garbage collect long time obsoleted (or never used) stuff from routing
API.
Differential Revision: http://reviews.llvm.org/D5703
llvm-svn: 219722
When compiling with -mfpu=vfpv3, those tests began to pass, like the others
with "Illegal Instruction" error, so removing the XFAIL from them should
get the bot green (and have more tests!).
llvm-svn: 219721
Before, ARM and Thumb mode code had different preferred alignments, which could
lead to some rather unexpected results. There's justification for reducing it
from the default 64-bits (wasted space), but I don't think there is for going
below 32-bits.
There's no actual ABI change here, just to reassure people.
llvm-svn: 219720
Before, ARM and Thumb mode code had different preferred alignments, which could
lead to some rather unexpected results. There's justification for reducing it
from the default 64-bits (wasted space), but I don't think there is for going
below 32-bits.
There's no actual ABI change here, just to reassure people.
llvm-svn: 219719
The CFL-AA implementation was missing a visit* routine for va_arg instructions,
causing it to assert when run on a function that had one. For now, handle these
in a conservative way.
Fixes PR20954.
llvm-svn: 219718
Eliminate library calls and intrinsic calls to fabs when the input
is a squared value.
Note that no unsafe-math / fast-math assumptions are needed for
this optimization.
Differential Revision: http://reviews.llvm.org/D5777
llvm-svn: 219717
Sign-/zero-extend folding depended on the load and the integer extend to be
both selected by FastISel. This cannot always be garantueed and SelectionDAG
might interfer. This commit adds additonal checks to load and integer extend
lowering to catch this.
Related to rdar://problem/18495928.
llvm-svn: 219716
Clang r219557 introduces libcalls to complex multiply/divide functions.
Since these functions are not available in iOS for arm64 devices, add them to
the static libraries.
llvm-svn: 219715
This moves some code from SemaType.cpp's hasVisibleDefinition() into
DeclCXX.cpp so that it can be used elsewhere. I found one other instance
of code trying to do the same thing, there are probably more. Search for
getInstantiatedFrom() to try to find more.
No functionality change.
Reviewed By: rsmith
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D5783
llvm-svn: 219714
Matt Arsenault