Philip Pfaffe [Mon, 10 Jul 2017 12:48:51 +0000 (12:48 +0000)]
[PM] Fix r307532: Get rid of a dangling reference.
Escaping lambda by-reference capture of local variable caused a dangling
reference.
llvm-svn: 307534
Hiroshi Inoue [Mon, 10 Jul 2017 12:44:25 +0000 (12:44 +0000)]
fix typos in comments and error messages; NFC
llvm-svn: 307533
Philip Pfaffe [Mon, 10 Jul 2017 10:57:55 +0000 (10:57 +0000)]
[PM] Enable registration of out-of-tree passes with PassBuilder
Summary:
This patch adds a callback registration API to the PassBuilder,
enabling registering out-of-tree passes with it.
Through the Callback API, callers may register callbacks with the
various stages at which passes are added into pass managers, including
parsing of a pass pipeline as well as at extension points within the
default -O pipelines.
Registering utilities like `require<>` and `invalidate<>` needs to be
handled manually by the caller, but a helper is provided.
Additionally, adding passes at pipeline extension points is exposed
through the opt tool. This patch adds a `-passes-ep-X` commandline
option for every extension point X, which opt parses into pipelines
inserted into that extension point.
Reviewers: chandlerc
Reviewed By: chandlerc
Subscribers: lksbhm, grosser, davide, mehdi_amini, llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D33464
llvm-svn: 307532
Javed Absar [Mon, 10 Jul 2017 10:42:55 +0000 (10:42 +0000)]
[ARM] Tidy up ARMBaseRegisterInfo implementation. NFC
Clean up ARMBaseRegisterInfo implementation a bit.
Differential Revision: https://reviews.llvm.org/D35116
llvm-svn: 307531
George Rimar [Mon, 10 Jul 2017 10:04:51 +0000 (10:04 +0000)]
[DWARF] - Rename variable. NFC.
Variable was called 'Name' and contained text
name of relocation type. Problem was that
outside of this error handling scope we already
have different 'Name' variable that contains
section name.
Change helps to avoid confusion.
llvm-svn: 307530
Gadi Haber [Mon, 10 Jul 2017 09:53:16 +0000 (09:53 +0000)]
This patch completely replaces the scheduling information for the SandyBridge architecture target by modifying the file X86SchedSandyBridge.td located under the X86 Target.
The SandyBridge architects have provided us with a more accurate information about each instruction latency, number of uOPs and used ports and I used it to replace the existing estimated SNB instructions scheduling and to add missing scheduling information.
Please note that the patch extensively affects the X86 MC instr scheduling for SNB.
Also note that this patch will be followed by additional patches for the remaining target architectures HSW, IVB, BDW, SKL and SKX.
The updated and extended information about each instruction includes the following details:
•static latency of the instruction
•number of uOps from which the instruction consists of
•all ports used by the instruction's' uOPs
For example, the following code dictates that instructions, ADC64mr, ADC8mr, SBB64mr, SBB8mr have a static latency of 9 cycles. Each of these instructions is decoded into 6 micro operations which use ports 4, ports 2 or 3 and port 0 and ports 0 or 1 or 5:
def SBWriteResGroup94 : SchedWriteRes<[SBPort4,SBPort23,SBPort0,SBPort015]> {
let Latency = 9;
let NumMicroOps = 6;
let ResourceCycles = [1,2,2,1];
}
def: InstRW<[SBWriteResGroup94], (instregex "ADC64mr")>;
def: InstRW<[SBWriteResGroup94], (instregex "ADC8mr")>;
def: InstRW<[SBWriteResGroup94], (instregex "SBB64mr")>;
def: InstRW<[SBWriteResGroup94], (instregex "SBB8mr")>;
Note that apart for the header, most of the X86SchedSandyBridge.td file was generated by a script.
Reviewers: zvi, chandlerc, RKSimon, m_zuckerman, craig.topper, igorb
Differential Revision: https://reviews.llvm.org/D35019#inline-304691
llvm-svn: 307529
George Rimar [Mon, 10 Jul 2017 09:36:44 +0000 (09:36 +0000)]
[DWARF] - Remove unused variables. NFC.
llvm-svn: 307528
Igor Breger [Mon, 10 Jul 2017 09:26:09 +0000 (09:26 +0000)]
[GlobalISel][X86] Support G_LOAD/G_STORE i1.
Summary: Support G_LOAD/G_STORE i1.
Reviewers: zvi, guyblank
Reviewed By: guyblank
Subscribers: rovka, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D35178
llvm-svn: 307527
Igor Breger [Mon, 10 Jul 2017 09:07:34 +0000 (09:07 +0000)]
[GlobalISel][X86] extend G_ZEXT support.
Summary:
Mark G_ZEXT/G_SEXT i1 to i8/i16, i8 to i16 as legal.
Support G_ZEXT i1 to i8/i16 instruction selection ( C++ code).
This patch requred to support G_LOAD/G_STORE i1.
Reviewers: zvi, guyblank
Reviewed By: guyblank
Subscribers: rovka, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D35177
llvm-svn: 307526
Kirill Bobyrev [Mon, 10 Jul 2017 09:07:23 +0000 (09:07 +0000)]
[docs] NFC: Fix links in the tutorial
r274441 introduced Chapter 10 of "Implementing a Language with LLVM" tutorial,
which caused all files in the tutorial to start using two digit numbering. But
many links were not changed and therefore appear to be broken. This patch
addresses described issue.
As a result, following command does not produce any output anymore:
$ grep -nR '<LangImpl[0-9].html>' ./docs/tutorial/
llvm-svn: 307525
Zvi Rackover [Mon, 10 Jul 2017 07:13:56 +0000 (07:13 +0000)]
X86 Intrinsics: _bit_scan_forward should not be under #ifdef __RDRND__
Summary:
The _bit_scan_forward and _bit_scan_reverse intrinsics were accidentally
masked under the preprocessor checks that prune intrinsics definitions for the
benefit of faster compile-time on Windows. This patch moves the
definitons out of that region.
Fixes pr33722
Reviewers: craig.topper, aaboud, thakis
Reviewed By: craig.topper
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D35184
llvm-svn: 307524
Hiroshi Inoue [Mon, 10 Jul 2017 06:32:52 +0000 (06:32 +0000)]
fix formatting; NFC
llvm-svn: 307523
Craig Topper [Mon, 10 Jul 2017 06:09:22 +0000 (06:09 +0000)]
[X86] Fix typo in comment. NFC
llvm-svn: 307522
Mikael Holmen [Mon, 10 Jul 2017 06:07:24 +0000 (06:07 +0000)]
[ArgumentPromotion] Change use of removed argument in llvm.dbg.value to undef
Summary:
This solves PR33641.
When removing a dead argument we must also handle possibly existing calls
to llvm.dbg.value that use the removed argument. Now we change the use
of the otherwise dead argument to an undef for some other pass to cleanup
later.
If the calls are left untouched, they will later on cause errors:
"function-local metadata used in wrong function"
since the ArgumentPromotion rewrites the code by creating a new function
with the wanted signature, but the metadata is not recreated so the new
function may then erroneously use metadata from the old function.
Reviewers: mstorsjo, rnk, arsenm
Reviewed By: rnk
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D34874
llvm-svn: 307521
Craig Topper [Mon, 10 Jul 2017 06:04:11 +0000 (06:04 +0000)]
[X86] Remove asserts from getX86CpuIDAndInfo/getX86CpuIDAndInfoEx. Restore past behavior of returning an unsupported indication to the caller instead.
These asserts could only occur if we fail to properly detect the compiler, but an assert is not a good way to do that because it doesn't work in release builds.
I wonder if we could use #error?
llvm-svn: 307520
Chandler Carruth [Mon, 10 Jul 2017 05:41:14 +0000 (05:41 +0000)]
[ADT] Fix another "oops" spotted by eddyb and reported in IRC.
This test pretty clearly should be calling 'maxnum' here. =]
llvm-svn: 307519
Eric Fiselier [Mon, 10 Jul 2017 04:32:21 +0000 (04:32 +0000)]
Fix issues with UBSAN test configuration.
On Apple the test feature 'sanitizer-new-delete' was incorrectly
getting added to the LIT feature set, which mistakenly caused tests
to be disabled when using UBSAN (the feature is only needed with ASAN/MSAN/TSAN).
llvm-svn: 307518
Eric Fiselier [Mon, 10 Jul 2017 04:16:50 +0000 (04:16 +0000)]
Work around PR31864 - ATOMIC_LLONG_LOCK_FREE is incorrect in 32 bit builds
llvm-svn: 307517
David Blaikie [Mon, 10 Jul 2017 03:04:59 +0000 (03:04 +0000)]
llvm-profdata: Reduce memory usage by using Error callback rather than member
Reduces llvm-profdata memory usage on a large profile from 7.8GB to 5.1GB.
The ProfData API now supports reporting all the errors/warnings rather
than only the first, though llvm-profdata ignores everything after the
first for now to preserve existing behavior. (if there's a desire for
other behavior, happy to implement that - but might be as well left for
a separate patch)
Reviewers: davidxl
Differential Revision: https://reviews.llvm.org/D35149
llvm-svn: 307516
Eric Fiselier [Mon, 10 Jul 2017 02:59:26 +0000 (02:59 +0000)]
Remove incorrect FIXME comment; the FIXME was addressed before the changes were committed
llvm-svn: 307515
Eric Fiselier [Mon, 10 Jul 2017 02:52:34 +0000 (02:52 +0000)]
Remove non-ascii characters introduced in r307513
llvm-svn: 307514
Eric Fiselier [Mon, 10 Jul 2017 01:27:22 +0000 (01:27 +0000)]
[coroutines] Include the implicit object parameter type when looking up coroutine_traits for member functions.
This patch was originally from Toby Allsopp, but I hijacked it and
fixed it up with his permission.
llvm-svn: 307513
Zachary Turner [Sun, 9 Jul 2017 23:32:15 +0000 (23:32 +0000)]
Don't access Python objects while not holding the GIL.
Patch by Tatyana Krasnukha
Differential Revision: https://reviews.llvm.org/D34942
llvm-svn: 307512
NAKAMURA Takumi [Sun, 9 Jul 2017 23:06:05 +0000 (23:06 +0000)]
CGSCCPassManagerTest.cpp: Fix warnings. [-Wunused-variable]
llvm-svn: 307511
Eric Fiselier [Sun, 9 Jul 2017 22:20:07 +0000 (22:20 +0000)]
Fix test failure to to new/delete ellisions
llvm-svn: 307510
Raphael Isemann [Sun, 9 Jul 2017 21:14:36 +0000 (21:14 +0000)]
[analyzer] Faster hashing of subsequences in CompoundStmts.
Summary: This patches improves the hashing subsequences in CompoundStmts by incrementally hashing all subsequences with the same starting position. This results in a reduction of the time for this constraint while running over SQLite from 1.10 seconds to 0.55 seconds (-50%).
Reviewers: NoQ
Reviewed By: NoQ
Subscribers: cfe-commits, xazax.hun, v.g.vassilev
Differential Revision: https://reviews.llvm.org/D34364
llvm-svn: 307509
Davide Italiano [Sun, 9 Jul 2017 19:22:48 +0000 (19:22 +0000)]
[X86] Relax an assertion when legalizing vector types.
WidenVSELECTAndMask can fold (and it folds in this case) so we
get a BUILD_VECTOR of constants as mask. convertMask() seems to
work fine when the input is a vector of constants, and we still
need to call it to extend/add elements at the end. but the current
code just asserts on anything but a SETCC or AND/OR/XOR of 2xSETCC.
This change was discussed briefly with Simon Pilgrim, who also
suggests we might consider dropping this assertion in the future.
Fixes PR33715.
llvm-svn: 307508
Craig Topper [Sun, 9 Jul 2017 17:43:11 +0000 (17:43 +0000)]
[X86] Add more feature flag bit defines to cpuid.h for gcc compatibility.
llvm-svn: 307507
Craig Topper [Sun, 9 Jul 2017 17:43:10 +0000 (17:43 +0000)]
[X86] Add __get_cpuid_count to cpuid.h. Update __get_cpuid to check the maximum level support before accessing the leaf. Rename level to leaf everywhere.
This matches gcc behavior.
llvm-svn: 307506
Casey Carter [Sun, 9 Jul 2017 17:15:49 +0000 (17:15 +0000)]
optional: Implement LWG 2900 and P0602
Differential Revision: https://reviews.llvm.org/D32385
llvm-svn: 307505
Simon Pilgrim [Sun, 9 Jul 2017 16:57:10 +0000 (16:57 +0000)]
[X86] Allow GHC calling convention to use YMM and ZMM registers
GHC 8.4 will know how to use YMM and ZMM registers for calls.
Submitted on behalf of @bgamari (Ben Gamari)
Differential Revision: https://reviews.llvm.org/D34854
llvm-svn: 307504
Dylan McKay [Sun, 9 Jul 2017 16:36:35 +0000 (16:36 +0000)]
[AVR] Fix test errors due to tied operands not matching
Broken due to r307259.
llvm-svn: 307503
Simon Pilgrim [Sun, 9 Jul 2017 16:01:04 +0000 (16:01 +0000)]
Handle ConstantExpr correctly in SelectionDAGBuilder
This change fixes a bug in SelectionDAGBuilder::visitInsertValue and SelectionDAGBuilder::visitExtractValue where constant expressions (InsertValueConstantExpr and ExtractValueConstantExpr) would be treated as non-constant instructions (InsertValueInst and ExtractValueInst). This bug resulted in an incorrect memory access, which manifested as an assertion failure in SDValue::SDValue.
Fixes PR#33094.
Submitted on behalf of @Praetonus (Benoit Vey)
Differential Revision: https://reviews.llvm.org/D34538
llvm-svn: 307502
Raphael Isemann [Sun, 9 Jul 2017 15:56:39 +0000 (15:56 +0000)]
[analyzer] Make StmtDataCollector part of the CloneDetection API
Summary: We probably want to use this useful templates in other pieces of code (e.g. the one from D34329), so we should make this public.
Reviewers: NoQ
Reviewed By: NoQ
Subscribers: cfe-commits, xazax.hun, v.g.vassilev, johannes
Differential Revision: https://reviews.llvm.org/D34880
llvm-svn: 307501
Simon Pilgrim [Sun, 9 Jul 2017 15:47:43 +0000 (15:47 +0000)]
[X86][AVX512] Regenerate AVX512VL comparison tests.
Show poor codegen on KNL targets as mentioned on D35179
llvm-svn: 307500
Tobias Grosser [Sun, 9 Jul 2017 15:47:17 +0000 (15:47 +0000)]
Remove freed InvalidDomains from InvalidDomainMap.
Summary:
Since r306667, propagateInvalidStmtDomains gets a reference to an
InvalidDomainMap. As part of the branch leading to return false, the respective
domain is freed. It is, however, not removed from the InvalidDomainMap, leaking
a pointer to a freed object which results in a use-after-free. Fix this be
removing the domain from the map before returning.
We tried to derive a test case that reliably failes, but did not succeed in
producing one. Hence, for now the failures in our LNT bots must be sufficient
to keep this issue tested.
Reviewers: grosser, Meinersbur, bollu
Subscribers: bollu, nandini12396, pollydev, llvm-commits
Differential Revision: https://reviews.llvm.org/D34971
llvm-svn: 307499
Chandler Carruth [Sun, 9 Jul 2017 13:45:11 +0000 (13:45 +0000)]
[PM] Fix a nasty bug in the new PM where we failed to properly
invalidation of analyses when merging SCCs.
While I've added a bunch of testing of this, it takes something much
more like the inliner to really trigger this as you need to have
partially-analyzed SCCs with updates at just the right time. So I've
added a direct test for this using the inliner and verifying the
domtree. Without the changes here, this test ends up finding a stale
dominator tree.
However, to handle this properly, we need to invalidate analyses
*before* merging the SCCs. After talking to Philip and Sanjoy about this
they convinced me this was the right approach. To do this, we need
a callback mechanism when merging SCCs so we can observe the cycle that
will be merged before the merge happens. This API update ended up being
surprisingly easy.
With this commit, the new PM passes the test-suite again. It hadn't
since MemorySSA was enabled for EarlyCSE as that also will find this bug
very quickly.
llvm-svn: 307498
Chandler Carruth [Sun, 9 Jul 2017 13:16:55 +0000 (13:16 +0000)]
[PM] Add unittesting of the call graph update logic with complex
dependencies between analyses.
This uncovers even more issues with the proxies and the splitting apart
of SCCs which are fixed in this patch. I discovered this while trying to
add more rigorous testing for a change I'm making to the call graph
update invalidation logic.
llvm-svn: 307497
Chandler Carruth [Sun, 9 Jul 2017 07:37:47 +0000 (07:37 +0000)]
[ADT] Fix a test case to use a correct escape for a null byte followed
by a valid octal digit.
The length argument shows that this was in fact the intent.
This was pointed out in IRC, thanks to eddyb!
llvm-svn: 307496
Craig Topper [Sun, 9 Jul 2017 07:26:14 +0000 (07:26 +0000)]
[X86] Remove check for AVX512 support from skylake-avx512 detection in getHostCPUName.
Users of getHostCPUName should also use getHostCPUFeatures which will take care of making sure avx512 is disabled if the CPU doesn't support it. This is consistent with what we do for other CPUs.
llvm-svn: 307495
Igor Breger [Sun, 9 Jul 2017 07:25:57 +0000 (07:25 +0000)]
[GlobalISel][X86] Add legalizer tests for G_LOAD/G_STORE operations. NFC.
llvm-svn: 307494
Chandler Carruth [Sun, 9 Jul 2017 07:23:27 +0000 (07:23 +0000)]
[PM] Teach PreservedAnalyses to have an `allInSet` static factory
function template to simplify building a quick object with a set marked
as preserved.
llvm-svn: 307493
Craig Topper [Sun, 9 Jul 2017 07:04:03 +0000 (07:04 +0000)]
[IR] Add Type::isIntOrIntVectorTy(unsigned) similar to the existing isIntegerTy(unsigned), but also works for vectors.
llvm-svn: 307492
Craig Topper [Sun, 9 Jul 2017 07:04:00 +0000 (07:04 +0000)]
[IR] Make use of Type::isPtrOrPtrVectorTy/isIntOrIntVectorTy/isFPOrFPVectorTy to shorten code. NFC
llvm-svn: 307491
Chandler Carruth [Sun, 9 Jul 2017 06:12:56 +0000 (06:12 +0000)]
[ADT] Add a default constructor and a bool conversion to function_ref.
The internal representation has a natural way to handle this and it
seems nicer than having to wrap this in an optional (with its own
separate flag).
This also matches how std::function works.
llvm-svn: 307490
Igor Breger [Sun, 9 Jul 2017 05:55:20 +0000 (05:55 +0000)]
[FastISel] fix a fallback diagnostic.
Summary: FastISel was marked as failed in case instruction selection succeeded.
Reviewers: qcolombet, zvi, rovka, ab
Reviewed By: zvi
Subscribers: javed.absar, ab, qcolombet, bogner, llvm-commits
Differential Revision: https://reviews.llvm.org/D34438
llvm-svn: 307489
Hiroshi Inoue [Sun, 9 Jul 2017 05:54:44 +0000 (05:54 +0000)]
fix trivial typos; NFC
sucessor -> successor
llvm-svn: 307488
Chandler Carruth [Sun, 9 Jul 2017 03:59:31 +0000 (03:59 +0000)]
[PM] Finish implementing and fix a chain of bugs uncovered by testing
the invalidation propagation logic from an SCC to a Function.
I wrote the infrastructure to test this but didn't actually use it in
the unit test where it was designed to be used. =[ My bad. Once
I actually added it to the test case I discovered that it also hadn't
been properly implemented, so I've implemented it. The logic in the FAM
proxy for an SCC pass to propagate invalidation follows the same ideas
as the FAM proxy for a Module pass, but the implementation is a bit
different to reflect the fact that it is forwarding just for an SCC.
However, implementing this correctly uncovered a surprising "bug" (it
was conservatively correct but relatively very expensive) in how we
handle invalidation when splitting one SCC into multiple SCCs. We did an
eager invalidation when in reality we should be deferring invaliadtion
for the *current* SCC to the CGSCC pass manager and just invaliating the
newly constructed SCCs. Otherwise we end up invalidating too much too
soon. This was exposed by the inliner test case that I've updated. Now,
we invalidate *just* the split off '(test1_f)' SCC when doing the CG
update, and then the inliner finishes and invalidates the '(test1_g,
test1_h)' SCC's analyses. The first few attempts at fixing this hit
still more bugs, but all of those are covered by existing tests. For
example, the inliner should also preserve the FAM proxy to avoid
unnecesasry invalidation, and this is safe because the CG update
routines it uses handle any necessary adjustments to the FAM proxy.
Finally, the unittests for the CGSCC pass manager needed a bunch of
updates where we weren't correctly preserving the FAM proxy because it
hadn't been fully implemented and failing to preserve it didn't matter.
Note that this doesn't yet fix the current crasher due to MemSSA finding
a stale dominator tree, but without this the fix to that crasher doesn't
really make any sense when testing because it relies on the proxy
behavior.
llvm-svn: 307487
Craig Topper [Sun, 9 Jul 2017 03:25:17 +0000 (03:25 +0000)]
[InstCombine] Speculatively implement a fix for what might be the root cause of PR33721 by making sure that we have integer types before doing select C, -1, 0 -> sext C to int
I recently changed m_One and m_AllOnes to use Constant::isOneValue/isAllOnesValue which work on floating point values too. The original implementation looked specifically for ConstantInt scalars and splats. So I'm guessing we are accidentally trying to issue sext/zexts on floating point types now.
Hopefully I figure out how to reproduce the failure from the PR soon.
llvm-svn: 307486
Simon Pilgrim [Sat, 8 Jul 2017 19:50:03 +0000 (19:50 +0000)]
[AMDGPU] Fix -Wimplicit-fallthrough warning. NFCI.
llvm-svn: 307485
Simon Pilgrim [Sat, 8 Jul 2017 19:28:24 +0000 (19:28 +0000)]
[AArch64] Fix -Wimplicit-fallthrough warnings. NFCI.
Add breaks - doesn't affect results as both GPR/FPU both check for 32/64 bit sizes. So will still default to GenericOps in the same way.
llvm-svn: 307484
Yaxun Liu [Sat, 8 Jul 2017 19:13:41 +0000 (19:13 +0000)]
Fix build failure with gcc about mixing enum and non-enum
llvm-svn: 307483
Erik Pilkington [Sat, 8 Jul 2017 18:54:08 +0000 (18:54 +0000)]
[Demangler] NFC: Don't make the parse_* functions templates
Differential revision: https://reviews.llvm.org/D35158
llvm-svn: 307482
Erik Pilkington [Sat, 8 Jul 2017 18:54:07 +0000 (18:54 +0000)]
[Demangler] NFC: Move Db struct to beginning of file
Differential revision: https://reviews.llvm.org/D35158
llvm-svn: 307481
Simon Pilgrim [Sat, 8 Jul 2017 18:42:04 +0000 (18:42 +0000)]
[ARM] Fix -Wimplicit-fallthrough warning. NFCI.
llvm-svn: 307480
Yuka Takahashi [Sat, 8 Jul 2017 17:48:59 +0000 (17:48 +0000)]
[Bash-autocompletion] Auto complete cc1 options if -cc1 is specified
Summary:
We don't want to autocomplete flags whose Flags class has `NoDriverOption` when argv[1] is not `-cc1`.
Another idea for this implementation is to make --autocomplete a cc1
option and handle it in clang Frontend, by porting --autocomplete
handler from Driver to Frontend, so that we can handle Driver options
and CC1 options in unified manner.
Differential Revision: https://reviews.llvm.org/D34770
llvm-svn: 307479
Yuka Takahashi [Sat, 8 Jul 2017 17:34:02 +0000 (17:34 +0000)]
[Bash-autocompletion] Fix a bug that -foo=bar doesn't handled properly
Summary: Fixed a bug that -foo=bar wasn't handled properly on old version of bash.
Differential Revision: https://reviews.llvm.org/D34927
llvm-svn: 307478
Max Kazantsev [Sat, 8 Jul 2017 17:17:30 +0000 (17:17 +0000)]
Re-enable "[IndVars] Canonicalize comparisons between non-negative values and indvars"
The patch was reverted due to a bug. The bug was that if the IV is the 2nd operand of the icmp
instruction, then the "Pred" variable gets swapped and differs from the instruction's predicate.
In this patch we use the original predicate to do the transformation.
Also added a test case that exercises this situation.
Differentian Revision: https://reviews.llvm.org/D35107
llvm-svn: 307477
Sanjay Patel [Sat, 8 Jul 2017 16:34:46 +0000 (16:34 +0000)]
[LoopVectorize] partly revert r307475
Bots are failing because of the additional checks.
llvm-svn: 307476
Sanjay Patel [Sat, 8 Jul 2017 16:10:42 +0000 (16:10 +0000)]
[LoopVectorize] auto-generate complete checks; NFC
I'm looking at a cmp transform in InstCombine that would affect these tests,
but it's hard to know if it makes things better or worse without seeing the
full IR. OTOH, maybe these tests shouldn't be running a bunch of transform
passes in the first place?
llvm-svn: 307475
Dimitry Andric [Sat, 8 Jul 2017 16:09:47 +0000 (16:09 +0000)]
Add a .arcconfig file for openmp.
llvm-svn: 307474
Simon Pilgrim [Sat, 8 Jul 2017 15:26:26 +0000 (15:26 +0000)]
Fix -Wimplicit-fallthrough warning. NFCI.
llvm-svn: 307473
Benjamin Kramer [Sat, 8 Jul 2017 14:14:57 +0000 (14:14 +0000)]
[CodeGen] Fold variable into assert.
Avoids warnings in Release builds.
llvm-svn: 307472
Sanjay Patel [Sat, 8 Jul 2017 14:04:48 +0000 (14:04 +0000)]
[x86] add SBB optimization for SETBE (ule) condition code
x86 scalar select-of-constants (Cond ? C1 : C2) combining/lowering is a mess
with missing optimizations. We handle some patterns, but miss logical variants.
To clean that up, we should convert all select-of-constants to logic/math and
enhance the combining for the expected patterns from that. Selecting 0 or -1
needs extra attention to produce the optimal code as shown here.
Attempt to verify that all of these IR forms are logically equivalent:
http://rise4fun.com/Alive/plxs
Earlier steps in this series:
rL306040
rL306072
rL307404 (D34652)
As acknowledged in the earlier review, there's a possibility that some Intel
uarch would prefer to produce an xor to clear the fake register operand with
sbb %eax, %eax. This will likely need to be addressed in a separate pass.
llvm-svn: 307471
Yaxun Liu [Sat, 8 Jul 2017 13:24:52 +0000 (13:24 +0000)]
CodeGen: Fix address space of global variable
Certain targets (e.g. amdgcn) require global variable to stay in global or constant address
space. In C or C++ global variables are emitted in the default (generic) address space.
This patch introduces virtual functions TargetCodeGenInfo::getGlobalVarAddressSpace
and TargetInfo::getConstantAddressSpace to handle this in a general approach.
It only affects IR generated for amdgcn target.
Differential Revision: https://reviews.llvm.org/D33842
llvm-svn: 307470
Kamil Rytarowski [Sat, 8 Jul 2017 11:27:56 +0000 (11:27 +0000)]
[Solaris] get rid of _RESTRICT_KYWD warning during the build
Summary:
(re)definition of _RESTRICT_KYWD rightfully causes a warning message during the Solaris build.
This hack is not needed if build compiler is properly configured (.e.g /usr/bin/gcc) so just remove it.
Reviewers: ro, mgorny, krytarowski, joerg
Reviewed By: joerg
Subscribers: quenelle, llvm-commits
Patch by Fedor Sergeev (Oracle).
Differential Revision: https://reviews.llvm.org/D35054
llvm-svn: 307469
Gabor Horvath [Sat, 8 Jul 2017 08:23:52 +0000 (08:23 +0000)]
[analyzer] Fix a path in the developer manual
Patch by: Reka Nikolett Kovacs
llvm-svn: 307468
Craig Topper [Sat, 8 Jul 2017 06:44:36 +0000 (06:44 +0000)]
[X86] In getHostCPUName, remove some code that changes some AMD CPU names based on features not being enabled.
The CPU name is really just used for scheduler and other microarchitectural optimizations. The feature flags should be determined by getHostCPUFeatures which should always be used with getHostCPUName. Trying to alter CPU name strings to control features just isn't practical.
Most of these types of things were removed from Intel CPUs a while ago.
This is part of my plan to bring compiler-rt's cpu_model.c file up to date with the equivalent functionality in libgcc. A lot of the code in that file is copied from Host.cpp and we want to keep them reasonably in sync.
llvm-svn: 307467
Craig Topper [Sat, 8 Jul 2017 06:44:35 +0000 (06:44 +0000)]
[X86] Correct the BDVER4 model numbers to include 0x70-0x7f.
According to wikipedia and some other googling suggests these should also be considered as BDVER4.
llvm-svn: 307466
Craig Topper [Sat, 8 Jul 2017 06:44:34 +0000 (06:44 +0000)]
[X86] Minor formatting fix. NFC
llvm-svn: 307465
Kuba Mracek [Sat, 8 Jul 2017 05:18:19 +0000 (05:18 +0000)]
Update message that Main Thread Checker produces.
llvm-svn: 307464
Craig Topper [Sat, 8 Jul 2017 05:16:14 +0000 (05:16 +0000)]
[X86] Use 'unsigned' instead of 'unsigned int' for consistency in the X86 portion of Host.cpp.
llvm-svn: 307463
Craig Topper [Sat, 8 Jul 2017 05:16:13 +0000 (05:16 +0000)]
[X86] Cleanup some CPUID usage in getAvailableFeatures.
We should make sure leaf 1 is available before accessing it. Same with leaf 0x80000001.
llvm-svn: 307462
Eric Fiselier [Sat, 8 Jul 2017 04:18:41 +0000 (04:18 +0000)]
Fix filesystem build on platforms with weird time_t types.
32-bit powerpc provides a 64 bit time_t type and older ppc64 systems
provide time_t as a floating point type. This caused problems when building
operations.cpp since operations.cpp contained compile time tests for conversions
between time_t and filesystem time type.
When these tests failed they caused the libc++ build to fail as well. This is unfortunate.
This patch moves the tests out of the source file and into the test suite. It also
expands the tests to allow testing of the weird time_t configurations on all platforms.
llvm-svn: 307461
Eric Beckmann [Sat, 8 Jul 2017 03:06:10 +0000 (03:06 +0000)]
Revert "Revert "Revert "Revert "Switch external cvtres.exe for llvm's own resource library.""""
This reverts commit
147f45ff24456aea59575fa4ac16c8fa554df46a.
Revert "Revert "Revert "Revert "Replace trivial use of external rc.exe by writing our own .res file.""""
This reverts commit
61a90a67ed54a1f0dfeab457b65abffa129569e4.
The patches were intially reverted because they were causing a failure
on CrWinClangLLD. Unfortunately, this was done haphazardly and didn't
compile, so the revert was reverted again quickly to fix this. One that
was done, the revert of the revert was itself reverted. This allowed me
to finally fix the actual bug in r307452. This patch re-enables the
code path that had originally been causing the bug, now that it (should)
be fixed.
llvm-svn: 307460
Zachary Turner [Sat, 8 Jul 2017 02:54:19 +0000 (02:54 +0000)]
Delete the pdb diff test.
This is failing on Linux for unknown reasons, so I need to
get the bots green while I investigate.
llvm-svn: 307459
Richard Trieu [Sat, 8 Jul 2017 02:04:42 +0000 (02:04 +0000)]
[ODRHash] Support FriendDecl
llvm-svn: 307458
Eric Christopher [Sat, 8 Jul 2017 01:03:29 +0000 (01:03 +0000)]
Remove a variable that was only used in asserts and had a duplicate copy in something we did use anyhow.
llvm-svn: 307457
Craig Topper [Sat, 8 Jul 2017 00:47:44 +0000 (00:47 +0000)]
[X86] Move AVX512VPOPCNTDQ in __builtin_cpu_support's enum to match trunk gcc.
There are two other features before it that we don't currently support in the the frontend or backend so I left placeholders to keep the encoding correct.
I think the compiler-rt implementation of this feature is even further out of date.
llvm-svn: 307456
Richard Smith [Sat, 8 Jul 2017 00:37:59 +0000 (00:37 +0000)]
Fix warnings introduced by r307434.
llvm-svn: 307455
Jason Molenda [Sat, 8 Jul 2017 00:12:15 +0000 (00:12 +0000)]
The x86 instruction unwinder can be asked to disassemble non-instruction
blocks of memory, and if the final bytes of that block look like a long
x86 instruction, it can cause the llvm disassembler to read past the end
of the buffer. Use the maximum allowed instruction length that we pass
to the llvm disassembler as a way to limit this to the size of the buffer.
An example of how to trigger this is when lldb does a function call, it
puts a breakpoint on the beginning of main() and uses that as the return
address from the function call. When we stop at that location, lldb may
try to find the first frame up the stack. Because this is on the first
instruction of a function, it will get the word-size value at the stack
pointer and assume that this was the caller's pc value. But this is random
stack memory and could point to anything - an object in memory, something
in the data section, whatever. And if we have a symbol for that thing,
we'll try to disassemble it.
This was leading to infrequent crashes in customer scenarios; figured out
what was happening with address sanitizer.
<rdar://problem/
30463256>
llvm-svn: 307454
Eric Beckmann [Fri, 7 Jul 2017 23:23:53 +0000 (23:23 +0000)]
Add name offset flags, for parity with cvtres.exe.
Summary:
The original cvtres.exe sets the high bit when an identifier offset
points to a string. Even though this is not mentioned in the spec, and
in fact does not seem to cause errors with most cases, for some reason
this causes a failure in Chromium where the new resource file is not
verified as a new version. This patch sets this high bit flag, and also
adds a test case to check that the output of our library is always
identical to original cvtres.
Reviewers: zturner, ruiu
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D35099
llvm-svn: 307452
Craig Topper [Fri, 7 Jul 2017 23:16:26 +0000 (23:16 +0000)]
[InstCombine] Make InstCombine's IRBuilder be passed by reference everywhere
Previously the InstCombiner class contained a pointer to an IR builder that had been passed to the constructor. Sometimes this would be passed to helper functions as either a pointer or the pointer would be dereferenced to be passed by reference.
This patch makes it a reference everywhere including the InstCombiner class itself so there is more inconsistency. This a large, but mechanical patch. I've done very minimal formatting changes on it despite what clang-format wanted to do.
llvm-svn: 307451
Eric Fiselier [Fri, 7 Jul 2017 23:02:30 +0000 (23:02 +0000)]
Fix diagnostic in verify test to match new Clang output
llvm-svn: 307450
Dehao Chen [Fri, 7 Jul 2017 22:40:37 +0000 (22:40 +0000)]
Revert r307445 as it breaks on certain platforms.
llvm-svn: 307449
Kuba Mracek [Fri, 7 Jul 2017 22:40:13 +0000 (22:40 +0000)]
Make sure SANITIZER_MIN_OSX_VERSION is defined before using it.
llvm-svn: 307448
Reid Kleckner [Fri, 7 Jul 2017 22:04:29 +0000 (22:04 +0000)]
[MS] Don't statically initialize dllimport member function pointers
Summary:
r306137 made dllimport pointers to member functions non-constant. This
is correct because a load must be executed to resolve any dllimported
data. However, r306137 did not account for the use of dllimport member
function pointers used as template arguments.
This change re-lands r306137 with a template instantiation fix.
This fixes PR33570.
Reviewers: rnk, majnemer
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D34714
llvm-svn: 307446
Dehao Chen [Fri, 7 Jul 2017 22:01:47 +0000 (22:01 +0000)]
Add sample PGO integration test to cover profile annotation and inlining.
Summary: The patch makes the integration test cover major sample PGO components.
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: sanjoy, cfe-commits
Differential Revision: https://reviews.llvm.org/D34725
llvm-svn: 307445
Lang Hames [Fri, 7 Jul 2017 21:51:11 +0000 (21:51 +0000)]
Add testcase for r305850.
Accidentally left this out of the original commit.
llvm-svn: 307444
Kuba Mracek [Fri, 7 Jul 2017 21:17:29 +0000 (21:17 +0000)]
Update buildgo.sh to pass -isysroot on Darwin.
llvm-svn: 307443
Lei Huang [Fri, 7 Jul 2017 21:12:35 +0000 (21:12 +0000)]
[PowerPC] NFC : Common up definitions of isIntS16Immediate and update parameter to int16_t
llvm-svn: 307442
Ed Maste [Fri, 7 Jul 2017 21:06:05 +0000 (21:06 +0000)]
remove deprecated register storage class specifier
While importing libomp into the FreeBSD base system we encountered
Clang warnings that "'register' storage class specifier is deprecated
and incompatible with C++1z [-Wdeprecated-register]".
Differential Revision: https://reviews.llvm.org/D35124
llvm-svn: 307441
David Blaikie [Fri, 7 Jul 2017 21:02:59 +0000 (21:02 +0000)]
ProfData: Fix some unchecked Errors in unit tests
The 'NoError' function was meant to be used as the input to
ASSERT/EXPECT_TRUE, but it is easy to forget this (it could be annotated
with nodiscard to help this) so many sites that look like they're checked
are not (& silently discard the failure). Only one site actually has an
Error sneaking out this way and I've replaced that one with a
FIXME+consumeError.
The rest of the code has been modified to use the EXPECT_THAT_ERROR
macros Zach introduced a while back. Between the options available this
seems OK/good/something to standardize on - though it's difficult to
build a matcher that could handle checking for a specific llvm::Error
result, so those remain using the custom ErrorEquals (& the nodiscard
added to ensure it is not misused as it was previous to this patch). It
could still be generalized a bit further (even not as far as a matcher,
but at least support multiple kinds of Error, etc) & added to the
general Error utility header.
llvm-svn: 307440
Dehao Chen [Fri, 7 Jul 2017 21:01:00 +0000 (21:01 +0000)]
Increase the import-threshold for crtical functions.
Summary: For interative sample-pgo, if a hot call site is inlined in the profiling binary, we should inline it in before profile annotation in the backend. Before that, the compile phase first collects all GUIDs that needs to be imported and creates virtual "hot" call edge in the summary. However, "hot" is not good enough to guarantee the callsites get inlined. This patch introduces "critical" call edge, and assign much higher importing threshold for those edges.
Reviewers: tejohnson
Reviewed By: tejohnson
Subscribers: sanjoy, mehdi_amini, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D35096
llvm-svn: 307439
Dehao Chen [Fri, 7 Jul 2017 20:53:17 +0000 (20:53 +0000)]
Enable the new PM + SamlePGO + ThinLTO testing.
Summary: This patch should be enabled after https://reviews.llvm.org/D34895
Reviewers: chandlerc, tejohnson, davidxl
Reviewed By: tejohnson
Subscribers: sanjoy, mehdi_amini, inglorion, eraman, cfe-commits
Differential Revision: https://reviews.llvm.org/D34896
llvm-svn: 307438
Dehao Chen [Fri, 7 Jul 2017 20:53:10 +0000 (20:53 +0000)]
Add sample PGO support to ThinLTO new pass manager.
Summary:
For SamplePGO + ThinLTO, because profile annotation is done twice at both PrepareForThinLTO pipeline and backend compiler, the following changes are needed at the PrepareForThinLTO phase to ensure the IR is not changed dramatically. Otherwise the profile annotation will be inaccurate in the backend compiler.
* disable hot-caller heuristic
* disable loop unrolling
* disable indirect call promotion
This will unblock the new PM testing for sample PGO (tools/clang/test/CodeGen/pgo-sample-thinlto-summary.c), which will be covered in another cfe patch.
Reviewers: chandlerc, tejohnson, davidxl
Reviewed By: tejohnson
Subscribers: sanjoy, mehdi_amini, Prazek, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D34895
llvm-svn: 307437
Zachary Turner [Fri, 7 Jul 2017 20:25:39 +0000 (20:25 +0000)]
[PDB] More changes to bring lld PDBs to parity with MSVC.
1) Don't write a /src/headerblock stream. This appears to be
written conditionally by MSVC, but it's not clear what the
condition is. For now, just remove it since we dont' know
what it is anyway and the particular pdb we've checked in
for the test doesn't have one.
2) Write a valid timestamp for the PDB file signature. This
leads to non-reproducible builds, but it matches the default
behavior of link, so it should be out default as well. If
we need reproducibility, we should add a separate command
line option for it that is off by default.
3) Write an empty FPO stream. MSVC seems to always write an
FPO stream. This change makes the stream directory match
up, although we still need to make the contents of the FPO
stream match.
llvm-svn: 307436
Anna Thomas [Fri, 7 Jul 2017 20:12:32 +0000 (20:12 +0000)]
[LoopUnrollRuntime] Support multiple exit blocks unrolling when prolog remainder generated
With the NFC refactoring in rL307417 (git SHA 987dd01), all the logic
is in place to support multiple exit/exiting blocks when prolog
remainder is generated.
This patch removed the assert that multiple exit blocks unrolling is only
supported when epilog remainder is generated.
Also, added test runs and checks with PROLOG prefix in
runtime-loop-multiple-exits.ll test cases.
llvm-svn: 307435
Richard Smith [Fri, 7 Jul 2017 20:04:28 +0000 (20:04 +0000)]
[modules ts] Basic for module linkage.
In addition to the formal linkage rules, the Modules TS includes cases where
internal-linkage symbols within a module interface unit can be referenced from
outside the module via exported inline functions / templates. We give such
declarations "module-internal linkage", which is formally internal linkage, but
results in an externally-visible symbol.
llvm-svn: 307434
Craig Topper [Fri, 7 Jul 2017 19:56:23 +0000 (19:56 +0000)]
[PatternMatch] Implemenet m_SignMask using Constant::isMinSignedValue instead of doing splat detection and analyzing the resulting APInt.
llvm-svn: 307433