JF Bastien [Mon, 18 Apr 2016 18:01:49 +0000 (18:01 +0000)]
NFC: unify clang / LLVM atomic ordering
This makes the C11 / C++11 *ABI* atomic ordering accessible from LLVM,
as discussed in http://reviews.llvm.org/D18200#inline-151433
This re-applies r266574 which I had reverted in r266575.
Depends on http://reviews.llvm.org/D18875
Original review: http://reviews.llvm.org/D18876
llvm-svn: 266641
JF Bastien [Mon, 18 Apr 2016 18:01:43 +0000 (18:01 +0000)]
NFC: unify clang / LLVM atomic ordering
This makes the C11 / C++11 *ABI* atomic ordering accessible from LLVM,
as discussed in http://reviews.llvm.org/D18200#inline-151433
This re-applies r266573 which I had reverted in r266576.
Original review: http://reviews.llvm.org/D18875
llvm-svn: 266640
Xinliang David Li [Mon, 18 Apr 2016 17:54:25 +0000 (17:54 +0000)]
Add missing new file for r266637
llvm-svn: 266639
Xinliang David Li [Mon, 18 Apr 2016 17:48:12 +0000 (17:48 +0000)]
Update InstrProf pass creator API reference
llvm-svn: 266638
Xinliang David Li [Mon, 18 Apr 2016 17:47:38 +0000 (17:47 +0000)]
Port InstrProfiling pass to the new pass manager
Differential Revision: http://reviews.llvm.org/D18126
llvm-svn: 266637
Eugene Zelenko [Mon, 18 Apr 2016 17:40:18 +0000 (17:40 +0000)]
[CMake] Fix misleading error message in CMakeLists.txt.
Differential revision: http://reviews.llvm.org/D17117
llvm-svn: 266635
Jonathan Peyton [Mon, 18 Apr 2016 17:27:30 +0000 (17:27 +0000)]
[STATS] Remove timePair class and unused functions
llvm-svn: 266634
Jonathan Peyton [Mon, 18 Apr 2016 17:24:20 +0000 (17:24 +0000)]
[STATS] print Total_* stats on their own line
llvm-svn: 266633
Simon Pilgrim [Mon, 18 Apr 2016 17:14:04 +0000 (17:14 +0000)]
[X86][AVX] Added zero+blend vs vperm2f128 optsize tests cases (PR22984)
We should be trying to use vperm2f128 instead of zero+blend (if we're only the user of zero?) when optsize is enabled.
llvm-svn: 266632
Easwaran Raman [Mon, 18 Apr 2016 17:10:17 +0000 (17:10 +0000)]
Revert r266488.
This goes with r266477 which has been reverted.
llvm-svn: 266631
JF Bastien [Mon, 18 Apr 2016 16:33:41 +0000 (16:33 +0000)]
Lanai: fix debug build
There's currently no raw_ostream &operator<<(SimpleValueType); provided by LLVM. It could be added by refactoring utils/TableGen/CodeGenTarget.cpp:getEnumName, but that's much more work than fixing the build.
llvm-svn: 266627
Konstantin Zhuravlyov [Mon, 18 Apr 2016 16:28:23 +0000 (16:28 +0000)]
[AMDGPU] Add insert nops pass based on subtarget features instead of cl::opt
Also,
- Skip pass if machine module does not have debug info
- Minor comment changes
- Added test
Differential Revision: http://reviews.llvm.org/D19079
llvm-svn: 266626
Weiming Zhao [Mon, 18 Apr 2016 16:25:46 +0000 (16:25 +0000)]
[ARM] predefines __ELF__ macro for arm-none-eabi
Summary: predefines __ELF__ macro for arm-none-eabi
Reviewers: silviu.baranga, rengolin
Subscribers: aemerson, rengolin, cfe-commits
Differential Revision: http://reviews.llvm.org/D19225
llvm-svn: 266625
Todd Fiala [Mon, 18 Apr 2016 16:09:21 +0000 (16:09 +0000)]
fix a race is the LLDB test suite results collection
The race boiled down to this:
If a test worker queue is able to run the test inferior and
clean up before the dosep.py listener socket is spun up, and
the worker queue is the last one (as would be the case when
there's only one test rerunning in the rerun queue), then
the test suite will exit the main loop before having a chance
to process any test events coming from the test inferior or
the worker queue job control.
I found this race to be far more likely on fast hardware.
Our Linux CI is one such example. While it will show
up primarily during meta test events generated by
a worker thread when a test inferior times out or
exits with an exceptional exit (e.g. seg fault), it only
requires that the OS takes longer to hook up the
listener socket than it takes for the final test inferior
and worker thread to shut down.
See:
http://reviews.llvm.org/D19214
reviewed by:
Pavel Labath
llvm-svn: 266624
Simon Pilgrim [Mon, 18 Apr 2016 16:08:19 +0000 (16:08 +0000)]
[X86][AVX] Renamed vperm2f128 test to make it quicker to review
missed one the first time round...
llvm-svn: 266623
Tobias Grosser [Mon, 18 Apr 2016 15:46:27 +0000 (15:46 +0000)]
SCoPValidator: Use SCEVTraversal to simplify SCEVInRegionDependences
llvm-svn: 266622
Simon Pilgrim [Mon, 18 Apr 2016 15:37:45 +0000 (15:37 +0000)]
[X86][AVX] Renamed vperm2f128 tests to make it quicker to review
llvm-svn: 266621
Igor Kudrin [Mon, 18 Apr 2016 15:36:30 +0000 (15:36 +0000)]
Reapply "[Coverage] Prevent detection of false instantiations in case of macro expansion."
The root of the problem was that findMainViewFileID(File, Function)
could return some ID for any given file, even though that file
was not the main file for that function.
This patch ensures that the result of this function is conformed
with the result of findMainViewFileID(Function).
This commit reapplies r266436, which was reverted by r266458,
with the .covmapping file serialized in v1 format.
Differential Revision: http://reviews.llvm.org/D18787
llvm-svn: 266620
Eric Liu [Mon, 18 Apr 2016 15:31:11 +0000 (15:31 +0000)]
Revert "Replace the use of MaxFunctionCount module flag"
This reverts commit r266477.
This commit introduces cyclic dependency. This commit has "Analysis" depend on "ProfileData",
while "ProfileData" depends on "Object", which depends on "BitCode", which
depends on "Analysis".
llvm-svn: 266619
George Rimar [Mon, 18 Apr 2016 15:16:17 +0000 (15:16 +0000)]
[ELF] - Avoid using memset for zero-initialization of struct member. NFC.
llvm-svn: 266618
Artem Tamazov [Mon, 18 Apr 2016 14:54:26 +0000 (14:54 +0000)]
[AMDGPU][llvm-mc] s_setreg* - Fix order of operands
Order should match the sp3 syntax, where destination (simm16 denoting the hwreg) is coming first.
Differential Revision: http://reviews.llvm.org/D19161
llvm-svn: 266617
Aaron Ballman [Mon, 18 Apr 2016 14:47:19 +0000 (14:47 +0000)]
Silence some "initialized but unused" warnings from MSVC -- the function being called is a static function, so there's no need for an instance variable. NFC.
llvm-svn: 266616
Nico Weber [Mon, 18 Apr 2016 13:57:08 +0000 (13:57 +0000)]
Revert 266581 (and follow-up 266588), it doesn't build on Windows.
Three problems:
1. <future> can't be easily used. If you must use it, see
include/Support/ThreadPool.h for how.
2. constexpr problems, even after 266588.
3. Move assignment operators can't be defaulted in MSVC2013.
llvm-svn: 266615
Nico Weber [Mon, 18 Apr 2016 13:54:50 +0000 (13:54 +0000)]
Unbreak building unit tests on Windows after r266595.
llvm-svn: 266614
Nico Weber [Mon, 18 Apr 2016 13:38:58 +0000 (13:38 +0000)]
Unbreak building LLVMTarget on Windows after r266595.
llvm-svn: 266613
Nico Weber [Mon, 18 Apr 2016 13:31:31 +0000 (13:31 +0000)]
Unbreak building llvm-pdbdump on Windows after r266595.
llvm-svn: 266612
Ben Craig [Mon, 18 Apr 2016 13:30:38 +0000 (13:30 +0000)]
Respect LIBCXXABI_LIBDIR_SUFFIX before an install
Prior to this patch, setting LIBCXXABI_LIBDIR_SUFFIX would confuse the
check-libcxxabi target. libc++abi.* would get output to lib instead of
lib${LIBCXXABI_LIBDIR_SUFFIX}, but the tests would look in the suffixed
directory.
Now, we match what libcxx does, and set the CMAKE_*_OUTPUT_DIRECTORY to the
LIBRARY_DIR.
http://reviews.llvm.org/D17410
llvm-svn: 266611
Rafael Espindola [Mon, 18 Apr 2016 12:58:59 +0000 (12:58 +0000)]
Fix handling of R_X86_64_GOT32.
It computes the offset to the end of .got.
llvm-svn: 266610
Rafael Espindola [Mon, 18 Apr 2016 12:44:33 +0000 (12:44 +0000)]
Simplify handling of R_X86_64_TPOFF32. NFC.
llvm-svn: 266609
Daniel Sanders [Mon, 18 Apr 2016 12:35:36 +0000 (12:35 +0000)]
[mips][ias] Prevent double-filling of delay slots by generating '.set noreorder' regions.
Summary:
When clang is given -save-temps or -via-file-asm, any inline assembly in
the source is parsed twice. Once by the compiler, and again by the
assembler. We must take care to ensure that this doesn't lead to
double-filling delay slots.
Reviewers: sdardis, vkalintiris
Subscribers: dsanders, sdardis, llvm-commits
Differential Revision: http://reviews.llvm.org/D19166
llvm-svn: 266608
Rafael Espindola [Mon, 18 Apr 2016 12:31:37 +0000 (12:31 +0000)]
Remove isGotRelative.
It is now redundant with the expression. While at it, handle both "X -
Got" and "Got - PC".
llvm-svn: 266607
Eric Liu [Mon, 18 Apr 2016 12:21:59 +0000 (12:21 +0000)]
Include SmallVector.h header in lib/Target/WebAssembly/InstPrinter/WebAssemblyInstPrinter.h
llvm-svn: 266606
Pavel Labath [Mon, 18 Apr 2016 12:18:35 +0000 (12:18 +0000)]
Attempt to fix darwin build after header refactor in llvm (r266595)
llvm-svn: 266605
Rafael Espindola [Mon, 18 Apr 2016 12:07:13 +0000 (12:07 +0000)]
Have getRelExpr handle all cases on x86.
This requires adding a few more expression types, but is already a small
simplification. Having Writer.cpp know the exact expression will also
allow further simplifications.
llvm-svn: 266604
Renato Golin [Mon, 18 Apr 2016 12:06:47 +0000 (12:06 +0000)]
[ARM] AArch32 v8 NEON is still not IEEE-754 compliant
llvm-svn: 266603
Daniel Sanders [Mon, 18 Apr 2016 12:06:15 +0000 (12:06 +0000)]
[mips][ias] Stream macro expansions to output instead of buffering them. NFC.
Summary:
This will allows us to eliminate some magic numbers from the offset operand of
branch instructions in favour of symbols and makes it possible to avoid
double-filling delay slots when clang is given -save-temps.
parseDirectiveCpRestore() is calling isIntegratedAssemblerRequired() for the
moment since correctly pushing the generation of these instructions into the
ELF target streamer is tricky enough to warrant a separate patch.
Reviewers: sdardis, vkalintiris
Subscribers: dsanders, llvm-commits, sdardis
Differential Revision: http://reviews.llvm.org/D19164
llvm-svn: 266602
Rafael Espindola [Mon, 18 Apr 2016 11:53:42 +0000 (11:53 +0000)]
Merge duplicated cases. NFC.
llvm-svn: 266601
Strahinja Petrovic [Mon, 18 Apr 2016 11:52:14 +0000 (11:52 +0000)]
[PowerPC] add comment to test
Added comment in test for soft-float operations on ppc architecture.
Test commit.
llvm-svn: 266600
Daniel Jasper [Mon, 18 Apr 2016 11:31:21 +0000 (11:31 +0000)]
clang-format: Improve heuristics to detect function declarations/definitions.
Specifically understand ellipses in parameter lists and treat trailing
reference qualifiers and the "{" as signals.
llvm-svn: 266599
Pavel Labath [Mon, 18 Apr 2016 11:01:41 +0000 (11:01 +0000)]
Fixup r266327
Fix XFAILed tests in TestThreadStates for the new signature of wait_for_running_event.
llvm-svn: 266598
Mehdi Amini [Mon, 18 Apr 2016 09:33:41 +0000 (09:33 +0000)]
Add missing header (NFC)
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266597
Mehdi Amini [Mon, 18 Apr 2016 09:26:29 +0000 (09:26 +0000)]
Add missing header (NFC)
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266596
Mehdi Amini [Mon, 18 Apr 2016 09:17:29 +0000 (09:17 +0000)]
[NFC] Header cleanup
Removed some unused headers, replaced some headers with forward class declarations.
Found using simple scripts like this one:
clear && ack --cpp -l '#include "llvm/ADT/IndexedMap.h"' | xargs grep -L 'IndexedMap[<]' | xargs grep -n --color=auto 'IndexedMap'
Patch by Eugene Kosov <claprix@yandex.ru>
Differential Revision: http://reviews.llvm.org/D19219
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266595
Mehdi Amini [Mon, 18 Apr 2016 09:08:59 +0000 (09:08 +0000)]
Add missing include for StringRef (NFC)
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266594
Marcin Koscielnicki [Mon, 18 Apr 2016 09:01:19 +0000 (09:01 +0000)]
[asan] [SystemZ] Add slop for stack address detection.
On s390, siginfo reports the faulting address with page granularity -
we need to mask off the low bits of sp before comparison.
Differential Revision: http://reviews.llvm.org/D19112
llvm-svn: 266593
Eric Fiselier [Mon, 18 Apr 2016 06:58:50 +0000 (06:58 +0000)]
Mark P0152R1 as complete.
llvm-svn: 266592
Eric Fiselier [Mon, 18 Apr 2016 06:51:33 +0000 (06:51 +0000)]
Mark LWG issue 2469 as done. Also simplify try_emplace and insert_or_assign implementations in unordered_map
llvm-svn: 266591
Eric Fiselier [Mon, 18 Apr 2016 06:17:30 +0000 (06:17 +0000)]
Implement LWG issue 2219 - support reference_wrapper in INVOKE
llvm-svn: 266590
Mohit K. Bhakkad [Mon, 18 Apr 2016 05:27:42 +0000 (05:27 +0000)]
[LLDB][MIPS] Fix TestConcurrentEvents
Patch by Nitesh Jain
Reviewers: clayborg
Subscribers: jaydeep, bhushan, mohit.bhakkad, sagar
Differential Revision: http://reviews.llvm.org/D18389
llvm-svn: 266589
Lang Hames [Mon, 18 Apr 2016 05:22:32 +0000 (05:22 +0000)]
[Orc] Tweak some of the new RPC code to silence a warning (extraneous ';') and
MSVC errors related to constexpr.
llvm-svn: 266588
Eric Fiselier [Mon, 18 Apr 2016 02:54:00 +0000 (02:54 +0000)]
Add hash specializations for __int128_t. Fixes LWG issue 2119
llvm-svn: 266587
Eric Fiselier [Mon, 18 Apr 2016 02:31:05 +0000 (02:31 +0000)]
Add tests for LWG issue 2361
llvm-svn: 266586
Eric Fiselier [Mon, 18 Apr 2016 01:40:45 +0000 (01:40 +0000)]
Fix LWG issue 2345 - Add insert(value_type&&)
llvm-svn: 266585
Rafael Espindola [Mon, 18 Apr 2016 01:34:20 +0000 (01:34 +0000)]
Delete the now redundant pointsToLocalDynamicGotEntry.
llvm-svn: 266584
Rafael Espindola [Mon, 18 Apr 2016 01:29:15 +0000 (01:29 +0000)]
Have getRelExpr return the final values for tls gd.
llvm-svn: 266583
Duncan P. N. Exon Smith [Mon, 18 Apr 2016 01:24:58 +0000 (01:24 +0000)]
BitcodeWriter: WorkList => Worklist, NFC
I have no idea how I chose two different spellings in the space of a
couple of weeks, but now I can't remember what to use where. Choose
"Worklist".
llvm-svn: 266582
Lang Hames [Mon, 18 Apr 2016 01:06:49 +0000 (01:06 +0000)]
[ORC] Generalize the ORC RPC utils to support RPC function return values and
asynchronous call/handle. Also updates the ORC remote JIT API to use the new
scheme.
The previous version of the RPC tools only supported void functions, and
required the user to manually call a paired function to return results. This
patch replaces the Procedure typedef (which only supported void functions) with
the Function typedef which supports return values, e.g.:
Function<FooId, int32_t(std::string)> Foo;
The RPC primitives and channel operations are also expanded. RPC channels must
support four new operations: startSendMessage, endSendMessage,
startRecieveMessage and endRecieveMessage, to handle channel locking. In
addition, serialization support for tuples to RPCChannels is added to enable
multiple return values.
The RPC primitives are expanded from callAppend, call, expect and handle, to:
appendCallAsync - Make an asynchronous call to the given function.
callAsync - The same as appendCallAsync, but calls send on the channel when
done.
callSTHandling - Blocking call for single-threaded code. Wraps a call to
callAsync then waits on the result, using a user-supplied
handler to handle any callbacks from the remote.
callST - The same as callSTHandling, except that it doesn't handle
callbacks - it expects the result to be the first return.
expect and handle - as before.
handleResponse - Handle a response from the remote.
waitForResult - Wait for the response with the given sequence number to arrive.
llvm-svn: 266581
Rafael Espindola [Mon, 18 Apr 2016 00:28:57 +0000 (00:28 +0000)]
Change getRelExpr to return the final value for TLSLD. NFC.
llvm-svn: 266580
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 23:30:31 +0000 (23:30 +0000)]
Linker: Share a single Metadata map for the lifetime of IRMover
Cache the result of mapping metadata nodes between instances of IRLinker
(i.e., for the lifetime of IRMover). There shouldn't be any real
functional change here, but this should give a major speedup. I had
loaned this to Mehdi when he tested performance of r266446, and the two
patches together gave a 10x speedup in metadata mapping.
llvm-svn: 266579
Rafael Espindola [Sun, 17 Apr 2016 23:20:08 +0000 (23:20 +0000)]
Make CreateTargetMachine as small as possible.
It is a pity that we have to create a TargetMachine once per thread,
so at least make that code as small as possible.
llvm-svn: 266578
Craig Topper [Sun, 17 Apr 2016 22:49:46 +0000 (22:49 +0000)]
[X86] Be explicit about calls to setOperationAction for AVX2 and AVX512 rather than just looping over all vector types and conditinally matching them. NFC
llvm-svn: 266577
JF Bastien [Sun, 17 Apr 2016 21:29:01 +0000 (21:29 +0000)]
Revert "NFC: unify clang / LLVM atomic ordering"
This reverts commit
537951f2f16d6a8542571c7722fcbae07d4e62c2.
Causes an assert in:
test/Transforms/AtomicExpand/SPARC/libcalls.ll
(Ordering2 != AtomicOrdering::NotAtomic && "expect atomic MO")
Bot:
http://lab.llvm.org:8080/green/job/clang-stage1-cmake-RA-incremental_check/21724/testReport/junit/LLVM/Transforms_AtomicExpand_SPARC/libcalls_ll/
I'm not getting this assert on my local debug build, but I'll revert
just to be sure.
llvm-svn: 266576
JF Bastien [Sun, 17 Apr 2016 21:28:50 +0000 (21:28 +0000)]
Revert "NFC: unify clang / LLVM atomic ordering"
This reverts commit
b0495df9eae2824bee830cc4c94f5441f0d4cbc9.
Same as for the corresponding LLVM revert, an assert seems to fire.
llvm-svn: 266575
JF Bastien [Sun, 17 Apr 2016 21:01:09 +0000 (21:01 +0000)]
NFC: unify clang / LLVM atomic ordering
Summary:
Depends on http://reviews.llvm.org/D18875
This makes the C11 / C++11 *ABI* atomic ordering accessible from LLVM, as discussed in http://reviews.llvm.org/D18200#inline-151433
Reviewers: jyknight, reames
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D18876
llvm-svn: 266574
JF Bastien [Sun, 17 Apr 2016 21:00:57 +0000 (21:00 +0000)]
NFC: unify clang / LLVM atomic ordering
Summary: This makes the C11 / C++11 *ABI* atomic ordering accessible from LLVM, as discussed in http://reviews.llvm.org/D18200#inline-151433
Reviewers: jyknight, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D18875
llvm-svn: 266573
Richard Smith [Sun, 17 Apr 2016 20:58:01 +0000 (20:58 +0000)]
When building with LLVM_ENABLE_MODULES, put the module cache into the build
directory. This is important for build bots to avoid filling up /tmp.
llvm-svn: 266571
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 20:11:09 +0000 (20:11 +0000)]
Transforms: Try harder to fix bootstrap after r266565
This catches two nullptr insertions into the ValueMap I missed in
r266567. I missed CloneFunction becuase it never calls RemapInstruction
directly. Here's one of the still-failing bots:
http://lab.llvm.org:8011/builders/clang-x64-ninja-win7/builds/11496
llvm-svn: 266570
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 19:40:20 +0000 (19:40 +0000)]
Linker: Don't double-schedule appending variables
Add an assertion to ValueMapper that prevents double-scheduling of
GlobalValues to remap, and fix the one place it happened. There are
tons of tests that fail with this assertion in place and without the
code change, so I'm not adding another.
Although it looks related, r266563 was, indeed, removing dead code.
AFAICT, this cross-file double-scheduling started in r266510 when the
cross-file recursion was removed.
llvm-svn: 266569
Davide Italiano [Sun, 17 Apr 2016 19:38:57 +0000 (19:38 +0000)]
[ParallelCG] SmallVector<char> -> SmallString.
llvm-svn: 266568
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 19:26:49 +0000 (19:26 +0000)]
Transforms: Fix bootstrap after r266565
Apparently there isn't test coverage for all of these. I'd appreciate
if someone with could reproduce and send me something to reduce, but for
now I've just looked for users of RemapInstruction and MapValue and
ensured they don't accidentally insert nullptr. Here is one of the
bootstraps that caught:
http://lab.llvm.org:8011/builders/clang-x64-ninja-win7/builds/11494
llvm-svn: 266567
Rafael Espindola [Sun, 17 Apr 2016 18:56:49 +0000 (18:56 +0000)]
Fix a use after move.
llvm-svn: 266566
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 18:53:24 +0000 (18:53 +0000)]
ValueMapper: Don't allow explicit null mappings of Values, NFC
As a follow-up to r123058, assert that there are no null mappings in the
ValueMap instead of just ignoring them when they are there. There were
a couple of accidental insertions in CloneFunction so I cleaned those up
(caught by testcases).
llvm-svn: 266565
Rafael Espindola [Sun, 17 Apr 2016 18:42:27 +0000 (18:42 +0000)]
Keep only the splitCodegen version that takes a factory.
This makes it much easier to see that all created TargetMachines are
equivalent.
llvm-svn: 266564
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 18:21:47 +0000 (18:21 +0000)]
IRMover: Remove dead code, NFC
llvm-svn: 266563
Craig Topper [Sun, 17 Apr 2016 17:37:33 +0000 (17:37 +0000)]
Declare MVT::SimpleValueType as an int8_t sized enum. This removes 400 bytes from TargetLoweringBase and probably other places.
This required changing several places to print VT enums as strings instead of raw ints since the proper method to use to print became ambiguous. This is probably an improvement anyway.
This also appears to save ~8K from an x86 self host build of llc.
llvm-svn: 266562
Faisal Vali [Sun, 17 Apr 2016 17:32:04 +0000 (17:32 +0000)]
Implement CWG 941 - explicit specializations of deleted function templates
template<class T> void f(T) = delete;
template<> void f(int); // OK.
f(3); // OK
Implementation strategy:
When an explicit specialization of a function template, a member function template or a member function of a class template is declared, clang first implicitly instantiates the declaration of a specialization from the templated-entity being explicitly specialized (since their signatures must be the same) and then links the explicit specialization being declared as a redeclaration of the aforementioned specialization.
The problem was that when clang 'implicitly instantiates' the initial specialization, it marks the corresponding FunctionDecl as deleted if the corresponding templated-entity was deleted, rather than waiting to see whether the explicit specialization being declared provides a non-deleted body. (The eager marking of delete has advantages during overload resolution I suppose, where we don't have to try and instantiate a definition of the function to see if it is deleted).
The present fix entails recognizing that when clang knows that an explicit specialization is being declared (for whichever templated-entity), the prior implicit instantiation should not inherit the 'deleted' status, and so we reset it to false.
I suppose an alternative fix (amongst others) could consider creating a new context (ExplicitSpecializationDeclarationSubstitution or some such) that is checked during template-argument-deduction and final substitution, and avoid inheriting the deleted status during declaration substitution. But while conceptually cleaner, that would be a slightly more involved change (as could be some of the other alternatives: such as avoid tagging implicit specializations as deleted, and check their primary templates for the deleted status where needed), and so I chose a different path. Hopefully it'll prove to not be a bad choice.
llvm-svn: 266561
Simon Pilgrim [Sun, 17 Apr 2016 15:10:42 +0000 (15:10 +0000)]
[X86][SSE] Added 16i8 -> 8i64 sext test
Shows poor codegen for AVX2
llvm-svn: 266560
Simon Pilgrim [Sun, 17 Apr 2016 11:34:18 +0000 (11:34 +0000)]
[X86] Added TODO comment for target shuffle mask decoding of bitcasted masks
llvm-svn: 266559
Ismail Donmez [Sun, 17 Apr 2016 09:01:51 +0000 (09:01 +0000)]
Fix shared build after r266484
llvm-svn: 266558
Asaf Badouh [Sun, 17 Apr 2016 08:28:40 +0000 (08:28 +0000)]
[X86] Remove unneeded variables
no functional change.
ExtraLoad and WrapperKind are been used only if (OpFlags == X86II::MO_GOTPCREL).
Differential Revision: http://reviews.llvm.org/D18942
llvm-svn: 266557
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 07:45:08 +0000 (07:45 +0000)]
DebugInfo: Make DICompositeTypes distinct most of the time
Since elements of most kinds of DICompositeType have back references,
most are involved in uniquing cycles. Except via the ODR 'identifier:'
field, which doesn't care about the storage type (see r266549),
they have no hope of being uniqued.
Distinct nodes are far more efficient, so use them for most kinds of
DICompositeType definitions (i.e., when DIType::isForwardDecl is false).
The exceptions:
- DW_TAG_array_type, since their elements never have back-references
and they never have ODR 'identifier:' fields;
- DW_TAG_enumeration_type when there is no ODR 'identifier:' field,
since their elements usually don't have back-references.
This breaks the last major uniquing cycle I'm aware of in the debug info
graph. The impact won't be enormous for C++ because references to
ODR-uniqued nodes still use string-based DITypeRefs; but this should
prevent a regression in C++ when we drop the string-based references.
This wouldn't have been reasonable until r266549, when composite types
stopped relying on being uniqued by structural equivalence to prevent
blow-ups at LTO time.
llvm-svn: 266556
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 07:35:38 +0000 (07:35 +0000)]
LTO: Merge debug info types when linking bitcode
Make sure lld enables ODR type uniquing for debug info when invoking
LTO, as a follow-up to LLVM r266549.
llvm-svn: 266555
Craig Topper [Sun, 17 Apr 2016 07:25:39 +0000 (07:25 +0000)]
[AVX512] ISD::MUL v2i64/v4i64 should only be legal if DQI and VLX features are enabled.
llvm-svn: 266554
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 06:42:30 +0000 (06:42 +0000)]
IR: Fix type-refs in testcase from r266548
There's a hole in the verifier right now: if a module has no compile
units, it never checks that all the string-based DITypeRefs get
resolved. As a result, this testcase didn't fail the verifier, even
there were references to `!"has-uuid"` instead of `!"uuid"` (the former
was a composite type's 'name:' field, the latter its 'identifier:'
field).
I'm currently working on removing string-based type refs entirely, and
this testcase started failing (because the upgrade script can't resolve
the type refs). Rather than fixing the (about-to-be-removed) hole in
the verifier, I'm just going to fix the test so that my upgrade script
handles it.
llvm-svn: 266553
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 05:41:09 +0000 (05:41 +0000)]
IR: Use getRawScope() when verifying
Fix a couple of places in the Verifier that call `getScope()` instead of
`getRawScope()`. Both DIDerivedType::getScope and
DICompositeType::getScope return a DITypeRef right now (which wraps a
Metadata*) so I don't think there's currently an observable bug. I
found this because a future commit that will change them to cast to
DIScope*.
llvm-svn: 266552
Sanjoy Das [Sun, 17 Apr 2016 04:30:43 +0000 (04:30 +0000)]
Fix a typo in rL265762
I accidentally replaced `mayBeOverridden` with `!isInterposable`.
Remove the negation and add a test case that would've caught this.
Many thanks to HÃ¥kan Hjort for spotting this!
llvm-svn: 266551
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 03:59:37 +0000 (03:59 +0000)]
Revert "use range loop, try to make comments more meaningful; NFCI"
This reverts commit r266541 since it introduces a use-after-free:
http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-bootstrap/builds/11471
llvm-svn: 266550
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 03:58:21 +0000 (03:58 +0000)]
IR: Use an explicit map for debug info type uniquing
Rather than relying on the structural equivalence of DICompositeType to
merge type definitions, use an explicit map on the LLVMContext that
LLParser and BitcodeReader consult when constructing new nodes.
Each non-forward-declaration DICompositeType with a non-empty
'identifier:' field is stored/loaded from the type map, and the first
definiton will "win".
This map is opt-in: clients that expect ODR types from different modules
to be merged must call LLVMContext::ensureDITypeMap.
- Clients that just happen to load more than one Module in the same
LLVMContext won't magically merge types.
- Clients (like LTO) that want to continue to merge types based on ODR
identifiers should opt-in immediately.
I have updated LTOCodeGenerator.cpp, the two "linking" spots in
gold-plugin.cpp, and llvm-link (unless -disable-debug-info-type-map) to
set this.
With this in place, it will be straightforward to remove the DITypeRef
concept (i.e., referencing types by their 'identifier:' string rather
than pointing at them directly).
llvm-svn: 266549
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 02:30:20 +0000 (02:30 +0000)]
IR: Use ODR to unique DICompositeType members
Merge members that are describing the same member of the same ODR type,
even if other bits differ. If the file or line differ, we don't care;
if anything else differs, it's an ODR violation (and we still don't
really care).
For DISubprogram declarations, this looks at the LinkageName and Scope.
For DW_TAG_member instances of DIDerivedType, this looks at the Name and
Scope. In both cases, we know that the Scope follows ODR rules if it
has a non-empty identifier.
llvm-svn: 266548
Craig Topper [Sun, 17 Apr 2016 01:34:37 +0000 (01:34 +0000)]
[Target] Reduce size of the LoadExtActions array in TargetLoweringBase by half. Saving ~18K bytes from the array.
llvm-svn: 266547
Craig Topper [Sun, 17 Apr 2016 01:34:35 +0000 (01:34 +0000)]
[Target] Remove checks for Simple VTs before calling routines that can handle Extended VTs too. NFC
llvm-svn: 266546
Craig Topper [Sun, 17 Apr 2016 01:34:32 +0000 (01:34 +0000)]
[Target] Fix an assertion that should have been updated when the code below it was changed in r251033.
llvm-svn: 266545
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 00:45:00 +0000 (00:45 +0000)]
LangRef: Removed some outdated text about DIDerivedType
This text is also incorrect (much like r266540). It looks like I missed
updating some of what I moved from SourceLevelDebugging.rst in r232566.
llvm-svn: 266544
Duncan P. N. Exon Smith [Sun, 17 Apr 2016 00:26:17 +0000 (00:26 +0000)]
Linker: Clarify test/Linker/type-unique-odr-a.ll, NFC
Split up the long RUN and clarify the CHECK lines:
- Explicitly confirm there are no other subprograms inside of "A".
- Remove checks for "bar" and "baz", which were just implicitly
checking that there were no other subprograms inside of "A".
This prepares for adding a RUN line which links the two files in the
opposite direction.
llvm-svn: 266543
Duncan P. N. Exon Smith [Sat, 16 Apr 2016 23:42:04 +0000 (23:42 +0000)]
IR: Add a configuration point for MDNodeInfo::isEqual, NFC
This commit has no functionality change, but it adds a configuration
point for MDNodeInfo::isEqual to allow custom uniquing of subclasses of
MDNode, minimizing the diff of a follow-up.
llvm-svn: 266542
Sanjay Patel [Sat, 16 Apr 2016 23:26:16 +0000 (23:26 +0000)]
use range loop, try to make comments more meaningful; NFCI
llvm-svn: 266541
Duncan P. N. Exon Smith [Sat, 16 Apr 2016 22:46:47 +0000 (22:46 +0000)]
LangRef: Fix some bugs in debug info descriptions
Fix descriptions of DICompositeType and DIDerivedType.
llvm-svn: 266540
Duncan P. N. Exon Smith [Sat, 16 Apr 2016 22:26:04 +0000 (22:26 +0000)]
IR: Remove extra blank line, NFC
llvm-svn: 266539
Duncan P. N. Exon Smith [Sat, 16 Apr 2016 22:25:36 +0000 (22:25 +0000)]
LTO: Use a common LibLTOCodeGenerator::init, NFC
llvm-svn: 266538
Duncan P. N. Exon Smith [Sat, 16 Apr 2016 21:44:08 +0000 (21:44 +0000)]
ValueMapper: Separate mapping of distinct and uniqued nodes (again)
Since the result of a mapped distinct node is known up front, it's more
efficient to map them separately from uniqued nodes. This commit pulls
them out of the post-order traversal and stores them in a worklist to be
remapped at the top-level.
This is essentially reapplying r244181 ("ValueMapper: Rotate distinct
node remapping algorithm") to the new iterative algorithm from r265456
("ValueMapper: Rewrite Mapper::mapMetadata without recursion").
Now that the traversal logic only handles uniqued MDNodes, it's much
simpler to inline it all into MDNodeMapper::createPOT (I've killed the
MDNodeMapper::push and MDNodeMapper::tryToPop helpers and localized the
traversal worklist).
The resulting high-level algorithm for MDNodeMapper::map now looks like
this:
- Distinct nodes are immediately mapped and added to
MDNodeMapper::DistinctWorklist using MDNodeMapper::mapDistinctNode.
- Uniqued nodes are mapped via MDNodeMapper::mapTopLevelUniquedNode,
which traverses the transitive uniqued subgraph of a node to
calculate uniqued node mappings in bulk.
- This is a simplified version of MDNodeMapper::map from before
this commit (originally r265456) that doesn't traverse through
any distinct nodes.
- Distinct nodes are added to MDNodeMapper::DistinctWorklist via
MDNodeMapper::mapDistinctNode.
- This uses MDNodeMapper::createPOT to fill a
MDNodeMapper::UniquedGraph (a post-order traversal and side
table), UniquedGraph::propagateChanges to track which uniqued
nodes need to change, and MDNodeMapper::mapNodesInPOT to create
the uniqued nodes.
- Placeholders for forward references are now only needed when
there's a uniquing cycle (a cycle of uniqued nodes unbroken by
distinct nodes). This is the key functionality change that
we're reintroducing (from r244181). As of r265456, a temporary
forward reference might be needed for any cycle that involved
uniqued nodes.
- After mapping the first node appropriately, MDNodeMapper::map works
through MDNodeMapper::DistinctWorklist. For each distinct node, its
operands are remapped with MDNodeMapper::mapDistinctNode and
MDNodeMapper::mapTopLevelUniquedNode until all nodes have been
mapped.
Sadly there's nothing observable I can test here; no real functionality
change, just a compile-time speedup from reduced malloc traffic.
llvm-svn: 266537