Commit Graph

343037 Commits

Author SHA1 Message Date
Thomas Lively 5286180999 [WebAssembly] Fix RegStackify and ExplicitLocals to handle multivalue
Summary:
There is still room for improvement in the handling of multivalue
nodes in both passes, but the current algorithm is at least correct
and optimizes some simpler cases. In order to make future
optimizations of these passes easier and build confidence that the
current algorithms are correct, this CL also adds a script that
automatically and exhaustively generates interesting multivalue test
cases.

Reviewers: aheejin, dschuff

Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D72902
2020-02-18 14:56:09 -08:00
Aditya Nandakumar b91d9ec0bb [GlobalISel]: Fix some non determinism exposed in CSE due to not notifying observers about mutations + add verification for CSE
https://reviews.llvm.org/D67133

While investigating some non determinism (CSE doesn't produce wrong
code, it just doesn't CSE some times) in GISel CSE on an out of tree
target, I realized that the core issue was that there were lots of code
that mutates (setReg, setRegClass etc), but doesn't notify observers
(CSE in this case but this could be any other observer). In order to
make the Observer be available in various parts of code and to avoid
having to thread it through various API, the MachineFunction now has the
observer as field. This allows it to be easily used in helper functions
such as constrainOperandRegClass.
Also added some invariant verification method in CSEInfo which can
catch these issues (when CSE is enabled).
2020-02-18 14:54:57 -08:00
Reid Kleckner 0c2b09a9b6 [IR] Lazily number instructions for local dominance queries
Essentially, fold OrderedBasicBlock into BasicBlock, and make it
auto-invalidate the instruction ordering when new instructions are
added. Notably, we don't need to invalidate it when removing
instructions, which is helpful when a pass mostly delete dead
instructions rather than transforming them.

The downside is that Instruction grows from 56 bytes to 64 bytes.  The
resulting LLVM code is substantially simpler and automatically handles
invalidation, which makes me think that this is the right speed and size
tradeoff.

The important change is in SymbolTableTraitsImpl.h, where the numbering
is invalidated. Everything else should be straightforward.

We probably want to implement a fancier re-numbering scheme so that
local updates don't invalidate the ordering, but I plan for that to be
future work, maybe for someone else.

Reviewed By: lattner, vsk, fhahn, dexonsmith

Differential Revision: https://reviews.llvm.org/D51664
2020-02-18 14:44:24 -08:00
Sid Manning cf4574299a [hexagon] Fix testcase issue with windows builder. 2020-02-18 16:36:38 -06:00
Reid Kleckner 236fcbc21a Add coding standard recommending use of qualifiers in cpp files
There is prior art for this in the code base itself, and a recent
example of this here: c45f8d4989

This came up in discussion on this review where @maskray was going the
opposite direction:
  https://reviews.llvm.org/D68772

Given that there is disagreement, we should make a choice and document
it.

Thanks to John McCall for the precise wording.

Reviewed By: MaskRay, rjmccall

Differential Revision: https://reviews.llvm.org/D74515
2020-02-18 14:08:56 -08:00
Daniel Sanders 2c8ee5329b Fix assertion on `!eq(?, 0)`
Instead of asserting, emit a proper error message
2020-02-18 14:05:55 -08:00
Thomas Lively 9d37f5afac [WebAssembly] Implement multivalue call_indirects
Summary:
Unlike normal calls, call_indirects have immediate arguments that
caused a MachineVerifier failure without a small tweak to loosen the
verifier's requirements for variadicOpsAreDefs instructions.

One nice thing about the new call_indirects is that they do not need
to participate in the PCALL_INDIRECT mechanism because their post-isel
hook handles moving the function pointer argument and adding the flags
and typeindex arguments itself.

Reviewers: aheejin

Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74191
2020-02-18 13:49:46 -08:00
Thomas Lively d51910967f Reland "[WebAssembly] Split and recombine multivalue calls for ISel"
This reverts commit 8acedb595d and
relands a prerequisite for the patch series culminating in
https://reviews.llvm.org/D74192.
2020-02-18 13:49:46 -08:00
Thomas Lively 7b64a59060 Reland "[WebAssembly][InstrEmitter] Foundation for multivalue call lowering"
This reverts commit 649aba93a2, now that
the approach started there has been shown to be workable in the patch
series culminating in https://reviews.llvm.org/D74192.
2020-02-18 13:49:46 -08:00
Davide Italiano 42cab985fd [TestTargetCommand] Remove another reference to a stale rdar.
The test passes, and the rdar is closed.
2020-02-18 13:39:37 -08:00
Davide Italiano 2560a93b70 [TestTargetCommand] `target var` without a process doesn't work on arm64e.
lldb needs to know about chains of authenticated relocations.

<rdar://problem/37773624>
2020-02-18 13:38:27 -08:00
Davide Italiano 3f5e0501b5 [TestGlobalVariables] Remove a reference to a stale rdar.
The bug has been fixed and the rdar is closed. No need to
clutter the test.
2020-02-18 13:31:14 -08:00
Davide Italiano 0159c21119 [TestGlobalVariables] `target var` without a process doesn't work on arm64e.
lldb needs to know about chains of authenticated relocations.

<rdar://problem/37773624>
2020-02-18 13:30:51 -08:00
Jonas Devlieghere ebf9a99bbd [lldb/Plugin] Reject WASM and Hexagon in DynamicLoaderStatic
The WASM and Hexagon plugin check the ArchType rather than the OSType,
so explicitly reject those in the DynamicLoaderStatic.

Differential revision: https://reviews.llvm.org/D74780
2020-02-18 13:29:34 -08:00
MaheshRavishankar a8355b5c0f [mlir][Linalg] Allow specifiying zero-rank shaped type operands to linalg.generic ops.
Fixing a bug where using a zero-rank shaped type operand to
linalg.generic ops hit an unrelated assert. This also meant that
lowering the operation to loops was not supported. Adding roundtrip
tests and lowering to loops test for zero-rank shaped type operand
with fixes to make the test pass.

Differential Revision: https://reviews.llvm.org/D74638
2020-02-18 13:23:28 -08:00
Alex Zinenko 870c1fd4c8 [mlir] NFC: rename LLVMOpLowering to ConvertToLLVMPattern
This better reflects the nature of the class and matches the current
naming scheme.

Differential Revision: https://reviews.llvm.org/D74774
2020-02-18 22:19:58 +01:00
David Tenty 58817a0783 [clang][XCOFF] Indicate that XCOFF does not support COMDATs
Summary: XCOFF doesn't support COMDATs, so clang shouldn't emit them.

Reviewers: stevewan, sfertile, Xiangling_L

Reviewed By: sfertile

Subscribers: dschuff, aheejin, dexonsmith, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74631
2020-02-18 16:10:11 -05:00
Reid Kleckner a0a1f412fd Remove "ELF Only" from -f*-sections help text
-ffunction-sections and -fdata-sections are well supported by many
object file formats:
- ELF
- COFF
- XCOFF
- wasm
Only MachO ignores this flag.

While here, remove it from -funique-section-names. Wasm honors this
option.

Addresses PR44910.

Reviewed By: hans, aaron.ballman

Differential Revision: https://reviews.llvm.org/D74634
2020-02-18 12:59:50 -08:00
Karasev Nikita 47282b1b4b Fix PR#44620 'readability-redundant-string-cstr quick-fix causes invalid code'
static void f2(std::string&&) {}
static void f() {
	std::string const s;
	f2(s.c_str()); // readability-redundant-string-cstr previously warning
}

Skips the problematic AST pattern in the matcher.
2020-02-18 15:33:52 -05:00
Sid Manning faa889b235 [Hexagon] clang driver should consider --sysroot option
Hexagon's clang driver should consider --sysroot option when setting
up include paths.

Differential Revision: https://reviews.llvm.org/D74776
2020-02-18 14:25:55 -06:00
Louis Dionne a7dcbe90cc [libc++] Fix overly complicated test of std::span's extent
Thanks to Billy O'Neal for the patch.

Differential Revision: https://reviews.llvm.org/D73138
2020-02-18 15:04:33 -05:00
Jorge Gorbe Moya 1ae8d81147 [libunwind] Fix memory leak in handling of DW_CFA_remember_state and DW_CFA_restore_state
parseInstructions() doesn't always process the whole set of DWARF
instructions for a frame. It will stop once the target PC is reached, or
if malformed instructions are found. So, for example, if we have an
instruction sequence like this:

```
<start>
...
DW_CFA_remember_state
...
DW_CFA_advance_loc past the location we're unwinding at (pcoffset in parseInstructions() main loop)
...
DW_CFA_restore_state
<end>
```

... the saved state will never be freed, even though the
DW_CFA_remember_state opcode has a matching DW_CFA_restore_state later
in the sequence.

This change adds code to free whatever is left on rememberStack after
parsing the CIE and the FDE instructions.

Differential Revision: https://reviews.llvm.org/D66904
2020-02-18 11:57:18 -08:00
serge-sans-paille b8bea9346a Revert "Support -fuse-ld=lld for riscv"
This reverts commit dd230142d8.

Failures:

http://lab.llvm.org:8011/builders/llvm-clang-win-x-armv7l/builds/4749
http://lab.llvm.org:8011/builders/llvm-clang-win-x-aarch64/builds/4752
2020-02-18 20:56:02 +01:00
Simon Pilgrim d6eef0614f [TargetLowering] Add SimplifyMultipleUseDemandedBits 'all elements' helper wrapper. NFC. 2020-02-18 19:53:50 +00:00
Yaxun (Sam) Liu 36f480f22c Revert "[CUDA][HIP][OpenMP] Add lib/Sema/UsedDeclVisitor.h after D70172"
This reverts commit c7fa409bca.
2020-02-18 14:45:34 -05:00
Yaxun (Sam) Liu bcadb1f2e6 Revert "[CUDA][HIP][OpenMP] Emit deferred diagnostics by a post-parsing AST travese"
This reverts commit 1b978ddba0.
2020-02-18 14:45:34 -05:00
River Riddle 94a4ca4bf3 [mlir] Add a TypeRange class that functions similar to ValueRange.
Summary: This class wraps around the various different ways to construct a range of Type, without forcing the materialization of that range into a contiguous vector.

Differential Revision: https://reviews.llvm.org/D74646
2020-02-18 11:37:24 -08:00
Jonas Devlieghere 2d146aa2a2 [lldb/Plugin] Generate LLDB_PLUGIN_DECLARE with CMake
Generate the LLDB_PLUGIN_DECLARE macros with CMake and a def file. I'm
landing D73067 in pieces so I can bisect what exactly is breaking the
Windows bot.
2020-02-18 11:29:56 -08:00
Alexandre Ganea 6f846c8504 Improve comments after 8404aeb56a. 2020-02-18 14:25:21 -05:00
Craig Topper 89ab5c69c8 [X86] Add a helper function to pull some repeated code out of combineGatherScatter. NFC 2020-02-18 11:10:40 -08:00
Fangrui Song 13a97305ba [JumpThreading] Skip unconditional PredBB when threading jumps through two basic blocks
Fixes https://bugs.llvm.org/show_bug.cgi?id=44922 (caused by 4698bf145d)

ThreadThroughTwoBasicBlocks assumes PredBBBranch is conditional. The following code can segfault.

  AddPHINodeEntriesForMappedBlock(PredBBBranch->getSuccessor(1), PredBB, NewBB,
                                  ValueMapping);

We can also allow unconditional PredBB, but the produced code is not
better.

Reviewed By: kazu

Differential Revision: https://reviews.llvm.org/D74747
2020-02-18 11:01:46 -08:00
LLVM GN Syncbot e706656af0 [gn build] Port c9e93c84f6 2020-02-18 18:45:25 +00:00
Tyker c9e93c84f6 Add Query API for llvm.assume holding attributes
Reviewers: jdoerfert, sstefan1, uenoku

Reviewed By: jdoerfert

Subscribers: mgorny, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D72885
2020-02-18 19:42:07 +01:00
Huihui Zhang 8ee0e1dc02 [NFC] Silence compiler warning [-Wmissing-braces]. 2020-02-18 10:37:12 -08:00
Stanislav Mekhanoshin dd4766451e [AMDGPU] Use generated RegisterPressureSets enum
Differential Revision: https://reviews.llvm.org/D74671
2020-02-18 10:34:03 -08:00
Matt Arsenault f4d3765fd9 CodeGen: Move undef_tied_input declaration
This doesn't belong in ARM specific code since it's generally
recognized by tablegen.
2020-02-18 10:33:10 -08:00
Jonas Devlieghere 884a58948b [lldb/Plugin] Unconditionally initialize DynamicLoaderDarwinKernel
Other plugins depend on DynamicLoaderDarwinKernel and which means we
cannot conditionally enable/build this plugin based on the target
platform. This means that it will be past of the list of plugins
initialized once that's autogenerated.
2020-02-18 10:32:06 -08:00
Nico Weber 711caee533 [gn build] (manually) port fc69967a4b 2020-02-18 13:29:13 -05:00
Stanislav Mekhanoshin b2a958a013 [TBLGEN] Emit register pressure set enum
Differential Revision: https://reviews.llvm.org/D74649
2020-02-18 10:09:05 -08:00
Kostya Kortchinsky fc69967a4b [scudo][standalone] Shift some data from dynamic to static
Summary:
Most of our larger data is dynamically allocated (via `map`) but it
became an hindrance with regard to init time, for a cost to benefit
ratio that is not great. So change the `TSD`s, `RegionInfo`, `ByteMap`
to be static.

Additionally, for reclaiming, we used mapped & unmapped a buffer each
time, which is costly. It turns out that we can have a static buffer,
and that there isn't much contention on it.

One of the other things changed here, is that we hard set the number
of TSDs on Android to the maximum number, as there could be a
situation where cores are put to sleep and we could miss some.

Subscribers: mgorny, #sanitizers, llvm-commits

Tags: #sanitizers, #llvm

Differential Revision: https://reviews.llvm.org/D74696
2020-02-18 09:38:50 -08:00
Miloš Stojanović 7603bfb4b0 Revert "[llvm-exegesis] Improve error reporting in Assembler.cpp"
This reverts https://reviews.llvm.org/rG63bb9fee525f
due to buildbot failures:
http://lab.llvm.org:8011/builders/clang-ppc64le-rhel/builds/1389
2020-02-18 18:35:21 +01:00
Mikhail Maltsev 63809d365e [ARM,MVE] Add vbrsrq intrinsics family
Summary:
This patch adds a new MVE intrinsics family, `vbrsrq`: vector bit
reverse and shift right. The intrinsics are compiled into the VBRSR
instruction. Two new LLVM IR intrinsics were also added: arm.mve.vbrsr
and arm.mve.vbrsr.predicated.

Reviewers: simon_tatham, dmgreen, ostannard, MarkMurrayARM

Reviewed By: simon_tatham

Subscribers: kristof.beyls, hiraditya, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74721
2020-02-18 17:31:21 +00:00
Florian Hahn e32522ca17 [SLPVectorizer] Do not assume extracelement idx is a ConstantInt.
The index of an ExtractElementInst is not guaranteed to be a
ConstantInt. It can be any integer value. Check explicitly for
ConstantInts.

The new test cases illustrate scenarios where we crash without
this patch. I've also added another test case to check the matching
of extractelement vector ops works.

Reviewers: RKSimon, ABataev, dtemirbulatov, vporpo

Reviewed By: ABataev

Differential Revision: https://reviews.llvm.org/D74758
2020-02-18 18:16:06 +01:00
Jan Kratochvil aa3e99dc85 [lldb] [nfc] Separate DIERef vs. user_id_t: GetForwardDeclClangTypeToDie()
Reasons are the same as for D74637.

Differential Revision: https://reviews.llvm.org/D74690
2020-02-18 18:09:39 +01:00
Nikita Popov f37e899fd7 [VectorUtils] Accept IRBuilderBase; NFC 2020-02-18 18:02:04 +01:00
Nikita Popov ec6c623ff9 [SimplifyLibCalls] Accept IRBuilderBase; NFC 2020-02-18 17:59:07 +01:00
Nikita Popov 28ffe38bba [LoopUtils] Accept IRBuilderBase; NFC 2020-02-18 17:58:46 +01:00
Nikita Popov ed6d30b517 [BuildLibCalls] Accept IRBuilderBase; NFC
Accept IRBuilderBase instead of IRBuilder<>. Remove dependency
on IRBuilder from header.
2020-02-18 17:58:16 +01:00
Nikita Popov 1ab37fad61 [InstCombine] Fix worklist management when simplifying demanded bits
When simplifying demanded bits, we currently only report the
instruction on which SimplifyDemandedBits was called as changed.
However, this is a recursive call, and the actually modified
instruction will usually be further up the chain. Additionally,
all the intermediate instructions should also be revisited,
as additional combines may be possible after the demanded bits
simplification. We fix this by explicitly adding them back to the
worklist.

Differential Revision: https://reviews.llvm.org/D72944
2020-02-18 17:55:40 +01:00
Nikita Popov c9540fe59b [InstCombine] Fix multi-use handling in cttz transform
The select-of-cttz transform can currently duplicate cttz intrinsics
and zext/trunc ops. The cause is that it unnecessarily duplicates
the intrinsic and the zext/trunc when setting the "undef_on_zero"
flag to false. However, it's always legal to set the flag from true
to false, so we can make this replacement even if there are extra users.

Differential Revision: https://reviews.llvm.org/D74685
2020-02-18 17:55:00 +01:00