Lang Hames [Sun, 8 Jan 2017 20:09:35 +0000 (20:09 +0000)]
[Orc][RPC] Lock the pending results data structure when installing new result
handlers, make abandonPendingResults public API.
This should make installing asynchronous result handlers thread safe.
The abandonPendingResults method is made public so that clients can disconnect
from a remote even if they have asynchronous handlers awaing results from that
remote. The asynchronous handlers will all receive "abandoned result" errors as
their argument.
Lang Hames [Sun, 8 Jan 2017 01:13:47 +0000 (01:13 +0000)]
[Orc][RPC] Add an APICalls utility for grouping RPC funtions for registration.
APICalls allows groups of functions to be composed into an API that can be
registered as a unit with an RPC endpoint. Doing registration on a-whole API
basis (rather than per-function) allows missing API functions to be detected
early.
APICalls also allows Function membership to be tested at compile-time. This
allows clients to write static assertions that functions to be called are
members of registered APIs.
Mehdi Amini [Sun, 8 Jan 2017 00:44:45 +0000 (00:44 +0000)]
[ThinLTO] Fix lazy-loading of Metadata attachment, which left some Fwd ref behind
The change in r291362 was too agressive. We still need to flush at the
end of the block because function local metadata can introduce fwd
ref as well.
(Bootstrap with ThinLTO was broken)
Mehdi Amini [Sat, 7 Jan 2017 20:24:23 +0000 (20:24 +0000)]
[ThinLTO] Fix assertions on lazy-loading of Metadata TBAA attachments
Summary:
The issue happens with:
%0 = ....., !tbaa !0
%1 = ....., !tbaa !1
With !0 that references !1.
In this case when loading !0 we generates a temporary for the
operand !1. We now flush it immediately and trigger the load of
!1 before moving on. If we don't we get the temporary when
attaching to %1. This is usually not an issue except that we
eagerly try to update TBAA MDNodes, which is obviously not possible
if we only have a temporary.
Matt Arsenault [Sat, 7 Jan 2017 19:55:12 +0000 (19:55 +0000)]
SimplifyLibCalls: Remove incorrect optimization of fabs
fabs(x * x) is not generally safe to assume x is positive if x is a NaN.
This is also less general than it could be, so this will be replaced
with a transformation on the intrinsic.
Daniel Berlin [Sat, 7 Jan 2017 19:04:59 +0000 (19:04 +0000)]
Update update_test_checks to work properly with phi nodes and other fun things.
Summary:
Prior to this change, phi nodes were never considered defs, and so we ended up with undefined variables for any loop. Now, instead of trying to find just defs, we iterate over each actual IR value in the line, and replace them one by one with either a definition or a use.
We also don't try to match anything in the comment portions of the line.
I've tested it even on things like function pointer calls, etc, and against existing test cases uses update_test_checks
With this change, we are able to use update_tests on the cyclic cases in newgvn.
The only case i'm aware of that will misfire is if you have a string with which contains a valid token.
However, this is the same as it is now, with a slightly larger set of strings that may misfire.
Prior to this change, a test with the string " %a =" would be replaced.
Rui Ueyama [Sat, 7 Jan 2017 08:28:56 +0000 (08:28 +0000)]
TarWriter: Use Ustar header's "prefix" field to store long filenames.
Tar's Ustar header has the "prefix" field to store a directory
part of a filename. It is not as flexible as the PAX-extended
filename because there's still a limitation on the maximum filename
size, but it mitigates the situation.
This patch should unbreak some Windows buildbots that uses very
old tar command.
Craig Topper [Sat, 7 Jan 2017 06:56:54 +0000 (06:56 +0000)]
[X86] Disable load unfolding for 128-bit MOVDDUP instructions since the load size is smaller than the register size so unfolding would increase the load size.
Dan Gohman [Sat, 7 Jan 2017 01:50:01 +0000 (01:50 +0000)]
[WebAssembly] Don't abort on code with UB.
Gracefully leave code that performs function-pointer bitcasts implying
non-trivial pointer conversions alone, rather than aborting, since it's
just undefined behavior.
[MachineBasicBlock] Add a non-assert live-in accessor for debug mode.
With r291169, it is now not possible to access the live-in information
when the liveness is not properly tracked. Although this is want we want
in general, for debugging purpose we may want to still be able to
traverse this information even if it may not be accurate.
Dan Gohman [Sat, 7 Jan 2017 00:34:54 +0000 (00:34 +0000)]
[WebAssembly] Add a pass to create wrappers for function bitcasts.
WebAssembly requires caller and callee signatures to match exactly. In LLVM,
there are a variety of circumstances where signatures may be mismatched in
practice, and one can bitcast a function address to another type to call it
as that type. This patch adds a pass which replaces bitcasted function
addresses with wrappers to replace the bitcasts.
This doesn't catch everything, but it does match many common cases.
Daniel Berlin [Sat, 7 Jan 2017 00:01:42 +0000 (00:01 +0000)]
NewGVN: Fix PR 31501.
Summary: LLVM's non-standard notion of phi nodes means we can't both try to substitute for undef in phi nodes *and* use phi nodes as leaders all the time. This changes NewGVN to use the same semantics as SimplifyPHINode to decide which phi nodes are equivalent.
Teresa Johnson [Fri, 6 Jan 2017 23:38:41 +0000 (23:38 +0000)]
[ThinLTO] Handle conflicting local names gracefully
Summary:
r285871 introduced an assert that was overly aggressive in the case
of a same-named local in different same-named files (in different
directories), where the source name and therefore the GUID ended up
the same because the files were compiled in their own directory without
any leading path. Change the handling in the promotion logic to get
the summary for the version in that module.
This also exposed an issue where we are not always importing the
right copy, which is a performance not correctness issue (because
the renaming is based on the module hash which must be different,
see the bug report for details). I will fix that as a follow-on.
Teresa Johnson [Fri, 6 Jan 2017 23:37:17 +0000 (23:37 +0000)]
[ThinLTO] Optionally ignore empty index file
Summary:
In order to simplify distributed build system integration, where actions
may be scheduled before the Thin Link which determines the list of
objects selected by the linker. The gold plugin currently will emit
0-sized index files for objects not selected by the link, to enable
checking for expected output files by the build system. If the build
system then schedules a backend action for these bitcode files, we want
to be able to fall back to normal compilation instead of failing.
This is the LLVM side support for optionally enabling fallback
instead of issuing an error. Return a null CombinedIndex from
llvm::getModuleSummaryIndexForFile under the option when the file
is empty. Clang can then ignore the index when it is null.
[gtest] Detect warning flags using the positive spelling.
Some GCC versions will accept any warning flag name after a '-Wno-',
which would cause us to try to disable warnings with names GCC didn't
understand. This will silently succeed unless there is some other output
from GCC in which case we get weird cc1plus warnings about the warning
name being bogus.
There is still the issue that gtest sets warning flags for building
gtest-all.cc using weird 'add_definitions' and the fact that there is
a GCC version which warns on the variadic macro usage in gtest under
-pedantic, but has no flag analogous to Clang's
-Wgnu-zero-variadic-macro-argumnets to suppress this warning. I haven't
been able to come up with any good solution here. The closest is to turn
off -pedantic for those versions of GCC, but that seems really nasty.
For now, those versinos of GCC aren't warning clean. If anyone is broken
by this, I'll work on CMake logic to detect and disable -pedantic in
these cases.
Michal Gorny [Fri, 6 Jan 2017 21:33:54 +0000 (21:33 +0000)]
[llvm-config] Print --system-libs only when static linking
Modify the --system-libs option in llvm-config to print system libs only
when using static linking. The system libraries are irrelevant when
linking to a shared library since the library has appropriate library
dependencies embedded.
Modify the --system-libs test appropriately to force static linking, and
disable it if static libs are not available (i.e. BUILD_SHARED_LIBS is
enabled).
Michal Gorny [Fri, 6 Jan 2017 21:33:48 +0000 (21:33 +0000)]
[cmake] Canonicalize CMake booleans to 0/1 for lit interop
Canonicalize all CMake booleans to 0/1 before passing them to lit, to
ensure that the Python side handles all of them consistently
and correctly. 0/1 is a safe choice of values that trigger the same
boolean interpretation in CMake, Python and C++.
Furthermore, using them without quotes improves the chance Python will
explicitly fail when an incorrect value (such as ON/OFF, TRUE/FALSE,
YES/NO) is accidentally passed, rather than silently misinterpreting
the value.
This replaces a lot of different logics spread around lit site files,
attempting to partially reproduce the boolean logic used in CMake
and usually silently failing when an uncommon value was used instead.
In fact, some of them were never working correctly since different
values were assigned in CMake and checked in Python.
The alternative solution could be to create a common parser for CMake
booleans in lit and use it consistently throughout the site files.
However, it does not seem like the best idea to create redundant
implementation of the same logic and have to follow upstream if it ever
is extended to handle more values.
Mehdi Amini [Fri, 6 Jan 2017 21:06:51 +0000 (21:06 +0000)]
Fix LoopLoadElimination to keep original alignment on the inital hoisted store
This is fixing a bug where Loop Vectorization is widening a load but
with a lower alignment. Hoisting the load without propagating the alignment
will allow inst-combine to later deduce a higher alignment that what the pointer
actually is.
Matthias Braun [Fri, 6 Jan 2017 19:22:01 +0000 (19:22 +0000)]
AArch64CollectLOH: Rewrite as block-local analysis.
Re-apply r288561: This time with a fix where the ADDs that are part of a
3 instruction LOH would not invalidate the "LastAdrp" state. This fixes
http://llvm.org/PR31361
Previously this pass was using up to 5% compile time in some cases which
is a bit much for what it is doing. The pass featured a full blown
data-flow analysis which in the default configuration was restricted to a
single block.
This rewrites the pass under the assumption that we only ever work on a
single block. This is done in a single pass maintaining a state machine
per general purpose register to catch LOH patterns.
Wolfgang Pieb [Fri, 6 Jan 2017 18:38:57 +0000 (18:38 +0000)]
[DWARF] Null out the debug locs of (loop invariant) instructions hoisted by LICM in
order to avoid jumpy line tables. Calls are left alone because they may be inlined.
Reid Kleckner [Fri, 6 Jan 2017 18:22:18 +0000 (18:22 +0000)]
Use %z for size_t and avoid deprecated string functions
This usage of strcpy and snprintf was certainly safe, but using them
sets off various deprecation and lint warnings. Easier to just write the
belt and suspenders version.
Daniel Sanders [Fri, 6 Jan 2017 14:29:34 +0000 (14:29 +0000)]
[globalisel] Stop requiring -debug/-debug-only=registerbankinfo for assertions.
Summary:
I've noticed that these assertions don't trigger when the condition is false.
The problem is that the DEBUG(x) macro only executes x when the pass is
emitting debug output via the -debug and -debug-only=registerbankinfo command
line arguments.
Debug builds should always execute the assertions so use '#ifndef NDEBUG' instead.
Also removed an assertion that is only true the first time it's tested. <Target>RegisterBankInfo's constructor will re-use register banks causing them to be valid on subsequent tests. That
assertion will fail on the first test too in the near future.
Michal Gorny [Fri, 6 Jan 2017 08:23:33 +0000 (08:23 +0000)]
[llvm-config] Add --cmakedir to obtain CMake module location
Add a --cmakedir option to llvm-config that returns the correct path to
built/installed CMake modules (i.e. lib/cmake/llvm). This is mostly
intended as a convenience option for stand-alone builds of other LLVM
projects that frequently reconstruct LLVM_CMAKE_PATH after querying
llvm-config.
Lang Hames [Fri, 6 Jan 2017 06:22:31 +0000 (06:22 +0000)]
[Orc][RPC] Fix an obvious locking-order bug in RawByteChannel::startSendMessage.
The lock needs to be acquired before the data is sent, not afterwards. This
think-o slipped in during the refactor in r286620, but went unnoticed as the
resulting bug only manifests in multi-threaded clients (of which there are none
in-tree).
No unit test as the bug depends on thread scheduling.
Rui Ueyama [Fri, 6 Jan 2017 05:33:45 +0000 (05:33 +0000)]
TarWriter: Emit PAX headers only when needed.
We use PAX headers to store long filenames (>= 100 bytes).
It is not needed to emit PAX headers if filenames fit in the
Ustar header. This patch implements that optimization.
projects / llvm / log