Jay Foad [Thu, 29 Oct 2020 12:10:56 +0000 (12:10 +0000)]
[AMDGPU] Use pseudo instructions for readlane/writelane
This reverts r227987 "R600/SI: Determine target-specific encoding of READLANE and WRITELANE early v2".
All the codegen changes are caused by the post-RA scheduler no longer
treating readlane/writelane as scheduling barriers due to having
unmodelled side effects. (The pseudos are hasSideEffects = 0, but the
real instructions are hasSideEffects = ? which TableGen conservatively
treats as 1.)
Differential Revision: https://reviews.llvm.org/D90401
Teresa Johnson [Sun, 25 Oct 2020 05:32:03 +0000 (22:32 -0700)]
[sanitizer] Print errno for report file open failure
To help debug failures, specifically the llvm-avr-linux bot failure from
5c20d7db9f2791367b9311130eb44afecb16829c:
http://lab.llvm.org:8011/#/builders/112/builds/407/steps/5/logs/FAIL__MemProfiler-x86_64-linux-dynamic__log_path_t
Also re-enable the failing test which I temporarily disabled, to
see if this change will help identify why that particular log file can't
be opened for write on that bot (when another log file in the same
directory could earlier in the test).
Differential Revision: https://reviews.llvm.org/D90120
Scott Linder [Thu, 29 Oct 2020 15:22:15 +0000 (15:22 +0000)]
[NFC] Add more tests for DISubprogram verifier
Minimum amount of tests to cover (most) of the DISubprogram verifier
checks.
Reviewed By: vsk
Differential Revision: https://reviews.llvm.org/D90340
Change-Id: Icd25dac64f87f6dcf67ff3443eb4f95af18d05a8
Simon Pilgrim [Thu, 29 Oct 2020 14:30:42 +0000 (14:30 +0000)]
[InstCombine] visitShl - ensure inner shifts have inrange amounts
Noticed when fixing OSS Fuzz #26716
Nicolas Vasilache [Thu, 29 Oct 2020 14:02:56 +0000 (14:02 +0000)]
[mlir][Linalg] Make Linalg fusion a test pass
Linalg "tile-and-fuse" is currently exposed as a Linalg pass "-linalg-fusion" but only the mechanics of the transformation are currently relevant.
Instead turn it into a "-test-linalg-greedy-fusion" pass which performs canonicalizations to enable more fusions to compose.
This allows dropping the OperationFolder which is not meant to be used with the pattern rewrite infrastructure.
Differential Revision: https://reviews.llvm.org/D90394
Nicholas Guy [Thu, 29 Oct 2020 13:58:39 +0000 (13:58 +0000)]
[ARM] Fix IT block generation after Thumb2SizeReduce with -Oz
Fixes a regression caused by D82439, in which IT blocks were no longer being generated when -Oz is present.
Differential Revision: https://reviews.llvm.org/D88496
Nicholas Guy [Thu, 29 Oct 2020 13:56:52 +0000 (13:56 +0000)]
[ARM] Add IT block generation test
D88496 introduces some new behaviour to IT block generation,
behaviour which is not covered by the current unit tests.
This adds one to cover it
Differential Revision: https://reviews.llvm.org/D90398
Kazushi (Jam) Marukawa [Thu, 29 Oct 2020 12:17:37 +0000 (21:17 +0900)]
[VE] Change to use integrated assembly by defualt
We've implemented integrated assembler. Now, we change to use
integrated assembler by default. Update a regression test also.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90396
Louis Dionne [Thu, 29 Oct 2020 15:07:39 +0000 (11:07 -0400)]
[libc++] Remove additional uses of std::rand() missed by
63aeadb4849d
Jay Foad [Thu, 29 Oct 2020 10:20:54 +0000 (10:20 +0000)]
[AMDGPU] Remove gds operand from ds_gws_* MachineInstrs
The operand value was always 1 (except in some bad MIR tests) so it was
redundant.
Differential Revision: https://reviews.llvm.org/D90378
Jay Foad [Thu, 29 Oct 2020 09:39:28 +0000 (09:39 +0000)]
[AMDGPU] Fix double space in disassembly of s_set_gpr_idx_mode
Differential Revision: https://reviews.llvm.org/D90374
Jay Foad [Wed, 28 Oct 2020 16:59:29 +0000 (16:59 +0000)]
[AMDGPU] Fix double space in disassembly of some DPP instructions
Differential Revision: https://reviews.llvm.org/D90373
Frederik Gossen [Thu, 29 Oct 2020 13:48:07 +0000 (13:48 +0000)]
[MLIR] Support walks over regions and blocks
Add specializations for `walk` to allow traversal of regions and blocks.
Differential Revision: https://reviews.llvm.org/D90379
David Sherwood [Mon, 26 Oct 2020 15:12:58 +0000 (15:12 +0000)]
[SVE] Remove TypeSize comparison operators
All known instances in the code where we relied upon the TypeSize
comparison operators have now been changed to either use scalar
interger comparisons or one of the TypeSize::isKnownXY functions.
It is now safe to remove the comparison operators.
Differential Revision: https://reviews.llvm.org/D90160
Kazushi (Jam) Marukawa [Thu, 29 Oct 2020 11:57:00 +0000 (20:57 +0900)]
[VE] Add missing BCR format
Add missing "BCR %sy, 0, target" format instruction and a regression
test for this format.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90387
Nico Weber [Thu, 29 Oct 2020 14:29:53 +0000 (10:29 -0400)]
Revert "clang-format: Add a consumer to diagnostics engine"
This reverts commit
df00267f1fdb0b098dc42f1caa8a59b29c8e0e5f.
clang-format should not depend on Frontend, see comment on
https://reviews.llvm.org/D90121.
Kazushi (Jam) Marukawa [Thu, 29 Oct 2020 12:01:16 +0000 (21:01 +0900)]
[VE] Add missing symbolic branch patterns
Add missing symbolic branch patterns to a regression test.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90388
Kazushi (Jam) Marukawa [Thu, 29 Oct 2020 11:38:04 +0000 (20:38 +0900)]
[VE] Support register aliases in llvm-mc
Support register aliases in MC layer to compile existing assembly
files with clang and integrated assembler.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90383
Valentin Clement [Thu, 29 Oct 2020 13:54:31 +0000 (09:54 -0400)]
[mlir][openacc] Add if and device_type to update op
Update op is modelling the update directive (2.14.4) from the OpenACC specs.
An if condition and a device_type list can be attached to the directive. This patch add
these two information to the current op.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D90310
Valentin Clement [Thu, 29 Oct 2020 13:53:10 +0000 (09:53 -0400)]
[flang][openacc] Enforce no modifier on enter data and exit data clauses
Enter data can have the copyin clause and exit data can have the copyout clause.
Both clauses support modifier with other directive but for these two directives no modifier
are supported. This semantic check enforce this rule.
Reviewed By: kiranktp
Differential Revision: https://reviews.llvm.org/D90280
Marek Kurdej [Thu, 29 Oct 2020 13:39:09 +0000 (14:39 +0100)]
[libcxx] [docs] [NFC] Fix typo.
Daniel Kiss [Thu, 29 Oct 2020 13:03:30 +0000 (14:03 +0100)]
[libcxx] Add targets to available features.
This patch add the target-* (x86_64-*) as used elsewhere in llvm.
Reviewed By: #libc, #libc_abi, ldionne
Differential Revision: https://reviews.llvm.org/D88027
Daniel Kiss [Thu, 29 Oct 2020 13:00:40 +0000 (14:00 +0100)]
[libunwind] Fix linker flag handling in the tests.
--export-dynamic is not always available on all targets.
-funwind-tables was a duplicate in the lit.site.cfg.in.
Reviewed By: ldionne
Differential Revision: https://reviews.llvm.org/D90202
Pavel Labath [Wed, 28 Oct 2020 13:59:14 +0000 (14:59 +0100)]
[lldb] Use reverse connection method for lldb-server tests
This fixes an flakyness is all gdb-remote tests. These tests have been
(mildly) flaky since we started using "localhost" instead of 127.0.0.1
in the test suite. The reason is that lldb-server needs to create two
sockets (v4 and v6) to listen for localhost connections. The algorithm
it uses first tries to select a random port (bind(localhost:0)) for the
first address, and then bind the same port for the second one.
The creating of the second socket can fail as there's no guarantee that
port will be available -- it seems that the (linux) kernel tries to
choose an unused port for the first socket (I've had to create thousands
of sockets to reproduce this reliably), but this can apparantly fail
when the system is under load (and our test suite creates a _lot_ of
sockets).
The socket creationg operation is considered successful if it creates at
least one socket is created, but the test harness has no way of knowing
which one it is, so it can end up connecting to the wrong address.
I'm not aware of a way to atomically create two sockets bound to the
same port. One way to fix this would be to make lldb-server report the
address is it listening on instead of just the port. However, this would
be a breaking change and it's not clear to me that's worth it (the
algorithm works pretty well under normal circumstances).
Instead, this patch sidesteps that problem by using "reverse"
connections. This way, the test harness is responsible for creating the
listening socket so it can pass the address that it has managed to open.
It also results in much simpler code overall.
To preserve test coverage for the named pipe method, I've moved the
relevant code to a dedicated test. To avoid original problem, this test
passes raw addresses (as obtained by getaddrinfo(localhost)) instead of
"localhost".
Differential Revision: https://reviews.llvm.org/D90313
Georgii Rymar [Tue, 27 Oct 2020 09:15:09 +0000 (12:15 +0300)]
[llvm-readobj/elf] - Fix a crash when dumping a dynamic relocation that refer to a symbol past the EOF.
There is a possible scenario when we crash when dumping dynamic relocations.
For that we should have no section headers (to take the number of synamic symbols from)
and a dynamic relocation that refers to a symbol with an index that is too large to be in a file.
The patch fixes it.
Differential revision: https://reviews.llvm.org/D90214
Alok Kumar Sharma [Thu, 29 Oct 2020 12:11:44 +0000 (17:41 +0530)]
[DebugInfo] [NFCI] Additional test for support of DW_TAG_generic_subrange
As suggested by dstenb, additional test is added to check emission of DW_OP_consts.
Differential Revision: https://reviews.llvm.org/D89218
Clement Courbet [Thu, 29 Oct 2020 10:48:43 +0000 (11:48 +0100)]
[llvm-exegesis] Do not try to assign random registers twice.
Doing a random assignment assigns both tested (forward) and back-to-back
(backward) instructions.
When none of the tested instruction and back-to-back instruction have
implicit aliasing, we're currently trying to do a random register
asignment twice.
Fix this (see PR26418).
Differential Revision: https://reviews.llvm.org/D90380
Adam Balogh [Thu, 15 Oct 2020 13:07:48 +0000 (15:07 +0200)]
[ADT] Fix for ImmutableMapRef
The `Root` member of `ImmutableMapRef` was changed recently from a plain
pointer to `IntrusiveRefCntPtr`. However, the `Profile` member function
was not adjusted. This results in comilation error whenever the
`Profile` method is used on an `ImmutableMapRef`. This patch fixes this
issue and also adds unit tests for `ImmutableMapRef`.
Differential Revision: https://reviews.llvm.org/D89486
Georgii Rymar [Thu, 29 Oct 2020 08:55:00 +0000 (11:55 +0300)]
[yaml2obj][test] - Merge strtab-implicit-sections-*.yaml into strtab-implicit-sections.yaml and improve testing of .shstrtab
This creates `strtab-implicit-sections.yaml` and merges 2 `strtab-implicit-sections*` tests into it.
I've also added a few tests for `.shstrtab` section related to section flags.
With that we have a single place where we can test implicit string table sections and
the `.shstrtab` section in particular.
Differential revision: https://reviews.llvm.org/D90372
Simon Pilgrim [Wed, 28 Oct 2020 16:17:45 +0000 (16:17 +0000)]
[InstCombine] Add select+funnel-shift test patterns
dfukalov [Tue, 27 Oct 2020 15:17:38 +0000 (18:17 +0300)]
[MIR] Fix out of bounds access in MIRPrinter.
Fixes: SWDEV-256460
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D90239
Max Kazantsev [Thu, 29 Oct 2020 11:09:31 +0000 (18:09 +0700)]
[NFC][SCEV] Use generic predicate checkers to simplify code
Jay Foad [Thu, 29 Oct 2020 10:55:16 +0000 (10:55 +0000)]
[AMDGPU] Simplify insertNoops functions. NFC.
Alok Kumar Sharma [Thu, 29 Oct 2020 10:38:51 +0000 (16:08 +0530)]
[DebugInfo] [NFCI] Adding a missed out line in support for DW_TAG_generic_subrange.
This commit adds a missed out line in earlier commit for DW_TAG_generic_subrange.
Previous commit ID:
a6dd01afa3d5902203d04a72e0b478078f796a35
Differential Revision: https://reviews.llvm.org/D89218
Thanks markus for pointing this out.
Florian Hahn [Thu, 29 Oct 2020 09:30:37 +0000 (09:30 +0000)]
[SCEV] Match 'zext (trunc A to iB) to iY' as URem.
URem operations with constant power-of-2 second operands are modeled as
such. This patch on its own has very little impact (e.g. no changes in
CodeGen for MultiSource/SPEC2000/SPEC2006 on X86 -O3 -flto), but I'll
soon post follow-up patches that make use of it to more accurately
determine the trip multiple.
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D89821
Max Kazantsev [Thu, 29 Oct 2020 10:34:58 +0000 (17:34 +0700)]
[NFC] Add some new util functions to ICmpInst
Tres Popp [Thu, 29 Oct 2020 10:01:05 +0000 (11:01 +0100)]
[mlir] Reorder shape assuming bufferization.
The previous ordering continued to use the original assuming after
replacing it which is not allowed. Now, inline the region from the old
into the new before the replacement.
Differential Revision: https://reviews.llvm.org/D90375
Georgii Rymar [Wed, 28 Oct 2020 12:05:35 +0000 (15:05 +0300)]
[yaml2obj][test] - Merge dynsymtab-shlink.yaml to dynsym-section.yaml
This simplifies the dynsymtab-shlink.yaml test (with use of macros)
and merges it into the dynsym-section.yaml test.
Differential revision: https://reviews.llvm.org/D90301
Krasimir Georgiev [Thu, 29 Oct 2020 10:27:54 +0000 (11:27 +0100)]
clang-format: Add a consumer to diagnostics engine
Contributed by dmikis (Kirill Dmitrenko)!
Otherwise problems like trying to format readonly file in-place led to crashes.
I've added reviewers by looking at `git blame` and other reviews to the changed file, so may have missed someone.
Reviewed By: krasimir
Differential Revision: https://reviews.llvm.org/D90121
Kazushi (Jam) Marukawa [Wed, 28 Oct 2020 11:13:45 +0000 (20:13 +0900)]
[VE] Add missing vector regression test
I forgot to add a regression test for VMAXX instruction when I added
it. So, I'm adding it now.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90356
Kazushi (Jam) Marukawa [Wed, 28 Oct 2020 11:11:40 +0000 (20:11 +0900)]
[VE] Add vector control instructions
Add LVL/SVL/SMVL/LVIX isntructions. Add regression tests too.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90355
David Spickett [Tue, 27 Oct 2020 11:55:27 +0000 (11:55 +0000)]
[lldb] Correct --help output for qemu rootfs script
It was printing "Usage:" twice.
Reviewed By: omjavaid
Differential Revision: https://reviews.llvm.org/D90225
David Zarzycki [Thu, 29 Oct 2020 09:53:47 +0000 (05:53 -0400)]
[lldb] Unbreak the build after a recent PowerPC change
40dd4d5233d9f81705a24d91b48d2620e487b89d introduced two new types.
Max Kazantsev [Thu, 29 Oct 2020 09:29:45 +0000 (16:29 +0700)]
[SCEV][NFC] Use general predicate checkers in monotonicity check
This makes the code more compact and readable.
Georgii Rymar [Wed, 28 Oct 2020 10:34:28 +0000 (13:34 +0300)]
[yaml2obj] - Improve handling of SectionHeaderTable::NoHeaders flag.
When `NoHeaders` is set, we still have following issues:
1) We emit the `.shstrtab` implicit section of size 1 (empty string table).
2) We still align the start of the section header table, what affects the output size.
3) We still write section header table bytes.
This patch fixes all of these issues.
Differential revision: https://reviews.llvm.org/D90295
David Green [Thu, 29 Oct 2020 09:13:23 +0000 (09:13 +0000)]
[InterleaveAccess] Recognise Interleave loads through binary operations
Instcombine will currently sink identical shuffles though vector binary
operations. This is probably generally useful, but can break up the code
pattern we use to represent an interleaving load group. This patch
reverses that in the InterleaveAccessPass to re-recognise the pattern of
shuffles sunk past binary operations and folds them back if an
interleave group can be created.
Differential Revision: https://reviews.llvm.org/D89489
Sam McCall [Tue, 27 Oct 2020 09:58:34 +0000 (10:58 +0100)]
[CMake] Support inter-proto dependencies in generate_protos.
Differential Revision: https://reviews.llvm.org/D90215
Max Kazantsev [Thu, 29 Oct 2020 08:27:21 +0000 (15:27 +0700)]
[NFC][SCEV] Refactor monotonic predicate checks to return enums instead of bools
This patch gets rid of output parameter which is not needed for most users
and prepares this API for further refactoring.
Sam McCall [Fri, 9 Oct 2020 13:17:26 +0000 (15:17 +0200)]
[clangd] Support CodeActionParams.only
Differential Revision: https://reviews.llvm.org/D89126
Vitaly Buka [Thu, 29 Oct 2020 08:09:39 +0000 (01:09 -0700)]
Revert "[NFC][TSAN] Logs to debug test script on bot"
Done with debugging. Script didn't work because of low limit on open
files on the bot.
This reverts commit
220293da53b5049ded67b1a251bd85069fa068e7.
Christian Sigg [Thu, 29 Oct 2020 07:17:27 +0000 (08:17 +0100)]
[mlir][gpu] Fix leaked stream and module when lowering gpu.launch_func to runtime calls.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D90370
Georgii Rymar [Wed, 28 Oct 2020 15:11:28 +0000 (18:11 +0300)]
[llvm-objdump][test] - Stop using precompiled binary in MachO/disassemble-g-dsym.test
This removes Inputs/libbogus11.a
Initially I've removed it in D90013, but had to restore it, because BB found this
test is using it.
I've updated the test to use YAMLs, added comment and one more possible error check.
Differential revision: https://reviews.llvm.org/D90312
Cameron Finucane [Thu, 29 Oct 2020 06:47:38 +0000 (23:47 -0700)]
[libFuzzer] Remove InterruptHandler from Fuchsia implementation
As implemented, the `InterruptHandler` thread was spinning trying to
`select()` on a null "stdin", wasting a significant amount of CPU for no
benefit. As Fuchsia does not have a native concept of stdin (or POSIX
signals), this commit simply removes this feature entirely.
Reviewed By: aarongreen
Differential Revision: https://reviews.llvm.org/D89266
Fangrui Song [Thu, 29 Oct 2020 06:58:31 +0000 (23:58 -0700)]
[Reassociate][test] Delete improper -NOT patterns to work with -enable-new-pm=1
The two tests rely on LegacyInlinerBase::doFinalization to remove
Function::isDefTriviallyDead() functions. The new PM does not have the behavior.
The -NEXT patterns checking the emptiness are actually sufficient.
Note, reassociate-deadinst.ll has become stale - it no longer catches
the problem r285380 intended. Unfortunately it is difficult to craft a
new test because it is actually pretty difficult to break it with
`MadeChange = true;` all over the file.
Serge Pavlov [Fri, 23 Oct 2020 11:36:16 +0000 (18:36 +0700)]
[FPEnv] Tests for rounding properties of constant evalution
These are moved from D88498.
Differential Revision: https://reviews.llvm.org/D90026
Fangrui Song [Thu, 29 Oct 2020 05:58:20 +0000 (22:58 -0700)]
[opt] Pin -stats-json & -opt-bisect-limit tests to -enable-new-pm=0
-stats-json requires `TimeReagion PassTimer(getPassTimer(...))` in the legacy
PM. The loss of functionality is not critical because we have the similar -time-passes.
Johannes Doerfert [Thu, 29 Oct 2020 05:38:40 +0000 (00:38 -0500)]
[Attributor][NFC] Rerun update test script
Johannes Doerfert [Thu, 29 Oct 2020 05:13:27 +0000 (00:13 -0500)]
[Attributor][FIX] Properly promote arguments pointers to arrays
When we promote pointer arguments we did compute a wrong offset and use
a wrong type for the array case.
Bug reported and reduced by Whitney Tsang <whitneyt@ca.ibm.com>.
River Riddle [Thu, 29 Oct 2020 05:00:55 +0000 (22:00 -0700)]
[mlir][SymbolTable] Small optimization to walking symbol references
* Check region count for unknown symbol tables first, as it is a faster check
* Add an accessor to MutableDictionaryAttr to get the internal dictionary without creating a new one if it is empty. This avoids an otherwise unnecessary lookup of an MLIRContext.
River Riddle [Thu, 29 Oct 2020 04:48:48 +0000 (21:48 -0700)]
[mlir][Inliner] Add a `wouldBeCloned` flag to each of the `isLegalToInline` hooks.
Often times the legality of inlining can change depending on if the callable is going to be inlined in-place, or cloned. For example, some operations are not allowed to be duplicated and can only be inlined if the original callable will cease to exist afterwards. The new `wouldBeCloned` flag allows for dialects to hook into this when determining legality.
Differential Revision: https://reviews.llvm.org/D90360
River Riddle [Thu, 29 Oct 2020 04:48:38 +0000 (21:48 -0700)]
[mlir][Inliner] Add a new hook for checking if it is legal to inline a callable into a call
In certain situations it isn't legal to inline a call operation, but this isn't something that is possible(at least not easily) to prevent with the current hooks. This revision adds a new hook so that dialects with call operations that shouldn't be inlined can prevent it.
Differential Revision: https://reviews.llvm.org/D90359
Ben Shi [Thu, 29 Oct 2020 03:49:21 +0000 (11:49 +0800)]
[NFC][clang][AVR] Add more devices
Reviewed By: dylanmckay
Differential Revision: https://reviews.llvm.org/D88352
Ben Shi [Thu, 29 Oct 2020 02:54:17 +0000 (10:54 +0800)]
[NFC][AVR] Improve device list
Reviewed By: dylanmckay
https://reviews.llvm.org/D87968
Fangrui Song [Thu, 29 Oct 2020 02:11:19 +0000 (19:11 -0700)]
[Debugify] Move global namespace functions into llvm::
Also move exportDebugifyStats from tools/opt to Debugify.cpp
Vedant Kumar [Thu, 29 Oct 2020 01:55:46 +0000 (18:55 -0700)]
Revert "[DebugInfo] Fix legacy ZExt emission when FromBits >= 64 (PR47927)"
This reverts commit
99053462216cf835eb3ae063942c618d9609de87.
It breaks the compiler-rt build, see https://reviews.llvm.org/D89838
Vedant Kumar [Thu, 29 Oct 2020 01:55:37 +0000 (18:55 -0700)]
Revert "[DebugInfo] Shorten legacy [s|z]ext dwarf expressions"
This reverts commit
2ce36ebca544dd71075a7818ff4070da5667603b. It depends
on https://reviews.llvm.org/D89838, which needs to be reverted.
Mircea Trofin [Wed, 28 Oct 2020 00:22:30 +0000 (17:22 -0700)]
[ThinLTO] Fix .llvmcmd emission
llvm::EmbedBitcodeInModule needs (what used to be called) EmbedMarker
set, in order to emit .llvmcmd. EmbedMarker is really about embedding the
command line, so renamed the parameter accordingly, too.
This was not caught at test because the check-prefix was incorrect, but
FileCheck does not report that when multiple prefixes are provided. A
separate patch will address that.
Differential Revision: https://reviews.llvm.org/D90278
Derek Schuff [Thu, 29 Oct 2020 00:15:58 +0000 (17:15 -0700)]
[WebAssembly] Add support for DWARF type units
Since Wasm comdat sections work similarly to ELF, we can use that mechanism
to eliminate duplicate dwarf type information in the same way.
Differential Revision: https://reviews.llvm.org/D88603
Haruki Imai [Thu, 29 Oct 2020 00:05:32 +0000 (17:05 -0700)]
[mlir] Convert raw data in dense element attributes for big-endian machines.
This patch fixes a bug [[ https://bugs.llvm.org/show_bug.cgi?id=46091 | 46091 ]]
Raw data for the `dense-element attribute` is written in little endian (LE) format.
This commit converts the format to big endian (BE) in ʻAttribute Parser` on the
BE machine. Also, when outputting on a BE machine, the BE format is converted
to LE in "AsmPrinter".
Differential Revision: https://reviews.llvm.org/D80695
River Riddle [Wed, 28 Oct 2020 23:46:38 +0000 (16:46 -0700)]
[mlir] Optimize the parsing of ElementsAttr hex strings
This revision optimizes the parsing of hex strings by using the checked variant of llvm::fromHex, and adding a specialized method to Token for extracting hex strings. This leads a large decrease in compile time when parsing large hex constants (one example: 2.6 seconds -> 370 miliseconds)
Differential Revision: https://reviews.llvm.org/D90266
River Riddle [Wed, 28 Oct 2020 23:46:31 +0000 (16:46 -0700)]
[llvm][StringExtras] Use a lookup table for `hexDigitValue`
This method is at the core of the conversion from hex to binary, and using a lookup table great improves the compile time of hex conversions.
Context: In MLIR we use hex strings to represent very large constants in the textual format of the IR. These changes lead to a large decrease in compile time when parsing these constants (>1 second -> 350 miliseconds).
Differential Revision: https://reviews.llvm.org/D90320
River Riddle [Wed, 28 Oct 2020 23:46:25 +0000 (16:46 -0700)]
[llvm][StringExtras] Add a fail-able version of `fromHex`
This revision adds a fail-able/checked version of `fromHex` that fails when the input string contains a non-hex character. This removes the need for users to have a separate check for if the string contains all hex digits. This becomes very costly for large hex strings given that checking if a string contains only hex digits is effectively the same as just converting it in the first place.
Context: In MLIR we use hex strings to represent very large constants in the textual format of the IR. These changes lead to a large decrease in compile time when parsing these constants (2 seconds -> 1 second).
Differential Revision: https://reviews.llvm.org/D90265
Jonas Devlieghere [Wed, 28 Oct 2020 23:22:15 +0000 (16:22 -0700)]
Revert "[AppleObjCRuntimeV2] Force lazily allocated class names to be resolved."
We're no longer convinced that this is needed and we have no test
coverage to disprove that. Backing out of this change until we're
convinced otherwise.
Kazushi (Jam) Marukawa [Wed, 28 Oct 2020 04:00:21 +0000 (13:00 +0900)]
[VE] Add vector mask operation instructions
Add VFMK/VFMS/VFMF/ANDM/ORM/XORM/EQVM/NNDM/NEGM/PCVM/LZVM/TOVM
isntructions. Add regression tests too. Also add new patterns
to parse VFMK/VFMS/VFMF mnemonics.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90297
Amy Huang [Tue, 27 Oct 2020 18:35:57 +0000 (11:35 -0700)]
Recommit "[CodeView] Emit static data members as S_CONSTANTs."
We used to only emit static const data members in CodeView as
S_CONSTANTS when they were used; this patch makes it so they are always emitted.
This changes CodeViewDebug.cpp to find the static const members from the
class debug info instead of creating DIGlobalVariables in the IR
whenever a static const data member is used.
Bug: https://bugs.llvm.org/show_bug.cgi?id=47580
Differential Revision: https://reviews.llvm.org/D89072
This reverts commit
504615353f31136dd6bf7a971b6c236fd70582be.
Austin Kerbow [Wed, 28 Oct 2020 21:38:41 +0000 (14:38 -0700)]
[AMDGPU] Add Reset function to GCNHazardRecognizer
Reset the tracked emitted instructions when starting scheduling on a new
region.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D90347
Mehdi Amini [Wed, 28 Oct 2020 23:28:36 +0000 (23:28 +0000)]
Make the post-commit review expectations more explicit with respect to revert
See http://lists.llvm.org/pipermail/llvm-dev/2016-March/096529.html for
context.
Reviewed By: silvas, rengolin, echristo, dexonsmith, gribozavr2
Differential Revision: https://reviews.llvm.org/D89995
Christopher Di Bella [Wed, 28 Oct 2020 23:16:17 +0000 (16:16 -0700)]
[Sema] adds basic -Wfree-nonheap-object functionality
Checks to make sure that stdlib's (std::)free is being appropriately
used. Presently checks for the following misuses:
- free(&stack_object)
- free(stack_array)
Differential Revision: https://reviews.llvm.org/D89988
Duncan P. N. Exon Smith [Tue, 20 Oct 2020 22:53:51 +0000 (18:53 -0400)]
ModuleManager: Simplify lookupModuleFile by only setting the out parameter once, NFC
Differential Revision: https://reviews.llvm.org/D89835
Gaurav Jain [Wed, 28 Oct 2020 17:37:15 +0000 (10:37 -0700)]
[NFC] Use [MC]Register in CSE & LICM
Differential Revision: https://reviews.llvm.org/D90327
Peter Collingbourne [Tue, 20 Oct 2020 21:36:34 +0000 (14:36 -0700)]
Reland "hwasan: Disable operator {new,delete} interceptors when interceptors are disabled."
There was a discrepancy in the gn build which is now fixed.
Differential Revision: https://reviews.llvm.org/D89827
Peter Collingbourne [Wed, 28 Oct 2020 22:41:44 +0000 (15:41 -0700)]
gn build: Define HWASAN_WITH_INTERCEPTORS=1 for hwasan_new_delete.cpp as well.
Duncan P. N. Exon Smith [Wed, 28 Oct 2020 22:39:14 +0000 (18:39 -0400)]
Fix includes in llvm/Support/FileSystem/UniqueID.h after
23ed570af1cc165afea1b70a533a4a39d6656501
Not sure why this worked for me, but some of the bots pointed out I
copied the wrong includes from FileSystem.h in
23ed570af1cc165afea1b70a533a4a39d6656501. Fixed.
Aaron Puchert [Wed, 28 Oct 2020 22:29:41 +0000 (23:29 +0100)]
[Sema] Let getters assert that trailing return type exists, NFCI
This was requested in the review of D90129.
Aaron Puchert [Wed, 28 Oct 2020 22:23:09 +0000 (23:23 +0100)]
Better source location for -Wignored-qualifiers on trailing return types
We collect the source location of a trailing return type in the parser,
improving the location for regular functions and providing a location
for lambdas, where previously there was none.
Fixes PR47732.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D90129
Keith Smiley [Wed, 28 Oct 2020 21:56:03 +0000 (14:56 -0700)]
[llvm-install-name-tool] Quote passed rpath args in error messages
This diff refactors error reporting to make it more clear
what arguments were passed to llvm-install-name-tool.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D90080
Craig Disselkoen [Wed, 28 Oct 2020 20:48:22 +0000 (16:48 -0400)]
C API: support scalable vectors
This adds support for scalable vector types in the C API and in
llvm-c-test, and also adds a test to ensure that llvm-c-test can properly
roundtrip operations involving scalable vectors.
While creating this diff, I discovered that the C API cannot properly roundtrip
_constant expressions_ involving shufflevector / scalable vectors, but that
seems to be a separate enough issue that I plan to address it in a future diff
(unless reviewers feel it should be addressed here).
Differential Revision: https://reviews.llvm.org/D89816
Kostya Kortchinsky [Mon, 26 Oct 2020 21:54:22 +0000 (14:54 -0700)]
[GWP-ASan] Abstract the thread local variables access
In a similar fashion to D87420 for Scudo, this CL introduces a way to
get thread local variables via a platform-specific reserved TLS slot,
since Fuchsia doesn't support ELF TLS from the libc itself.
If needing to use this, a platform will have to define
`GWP_ASAN_HAS_PLATFORM_TLS_SLOT` and provide `gwp_asan_platform_tls_slot.h`
which will define a `uint64_t *getPlatformGwpAsanTlsSlot()` function
that will return the TLS word of storage.
I snuck in a couple of cleanup items as well, moving some static
functions to anonymous namespace for consistency.
Differential Revision: https://reviews.llvm.org/D90195
Jay Foad [Tue, 27 Oct 2020 12:29:11 +0000 (12:29 +0000)]
[AMDGPU] Allow some modifiers on VOP3B instructions
V_DIV_SCALE_F32/F64 are VOP3B encoded so they can't use the ABS src
modifier, but they can still use NEG and the usual output modifiers.
This partially reverts
3b99f12a4e6f "AMDGPU: Remove modifiers from v_div_scale_*".
Differential Revision: https://reviews.llvm.org/D90296
Florian Hahn [Wed, 28 Oct 2020 21:49:33 +0000 (21:49 +0000)]
[InstCombine] Do not introduce bitcasts for swifterror arguments.
The following constraints hold for swifterror values:
A swifterror value (either the parameter or the alloca) can only
be loaded and stored from, or used as a swifterror argument.
This patch updates instcombine to not try to convert a bitcast of a
function into a bitcast of a swifterror argument.
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D90258
Jay Foad [Wed, 28 Oct 2020 16:00:59 +0000 (16:00 +0000)]
[AMDGPU] Fix double space in disassembly of SDWA instructions with vcc
Differential Revision: https://reviews.llvm.org/D90317
Jay Foad [Wed, 28 Oct 2020 16:12:21 +0000 (16:12 +0000)]
[TableGen] Treat reg as isolated in reg$foo (but not in ${foo}reg)
D9844 fixed a problem where the ss suffix in the AsmString "cmp${cc}ss"
was recognised as the X86 SS register, by only recognising a token as a
register name if it is "isolated", i.e. surrounded by separator
characters.
In the AMDGPU backend there are many operands like $clamp which expand
to an optional string " clamp" including the preceding space, so we want
to have AsmStrings including sequences like "vcc$clamp" where vcc is a
register name.
This patch relaxes the rules for an isolated token, to say that it's OK
if the token is immediately followed by a '$'.
Differential Revision: https://reviews.llvm.org/D90315
Florian Hahn [Wed, 28 Oct 2020 21:20:52 +0000 (21:20 +0000)]
[AArch64] Improve lowering of insert_vector_elt with 0.0 consts.
When moving +0.0 into a float vector, we can use to vi*gpr variants of
INS.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D90176
Austin Kerbow [Wed, 28 Oct 2020 19:22:23 +0000 (12:22 -0700)]
[AMDGPU] Fix inserting combined s_nop in bundles
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D90334
Philip Reames [Wed, 28 Oct 2020 21:29:03 +0000 (14:29 -0700)]
[Deref] Use maximum trip count instead of exact trip count
When trying to prove that a memory access touches only dereferenceable memory across all iterations of a loop, use the maximum exit count rather than an exact one. In many cases we can't prove exact exit counts whereas we can prove an upper bound.
The test included is for a single exit loop with a min(C,V) exit count, but the true motivation is support for multiple exits loops. It's just really hard to write a test case for multiple exits because the vectorizer (the primary user of this API), bails far before this. For multiple exits, this allows a mix of analyzeable and unanalyzable exits when only analyzeable exits are needed to prove deref.
Richard Smith [Wed, 28 Oct 2020 21:27:38 +0000 (14:27 -0700)]
PR48002: Fix injection of elaborated-type-specifiers within local
classes into the enclosing block scope.
We weren't properly detecting whether the name would be injected into a
block scope in the case where it was lexically declared in a local
class.
Sean Silva [Mon, 26 Oct 2020 23:46:54 +0000 (16:46 -0700)]
[mlir] Properly handle recursive bufferization for scf.for/scf.if
This fixes a subtle issue, described in the comment starting with
"Clone the op without the regions and inline the regions from the old op",
which prevented this conversion from working on non-trivial examples.
Differential Revision: https://reviews.llvm.org/D90203
Michael Park [Mon, 26 Oct 2020 16:34:22 +0000 (12:34 -0400)]
[libc++] Re-apply the switch-based std::variant implementation
This commit is a mass re-application of the following commits:
7d15ece79c16dc3237fc514ff56a69e3d58fbd39
e0ec7a02064968c7df11713689107148b4efb993
02197f7e50b938f8167b17b89bdf7c55feff4339
a175a96517c5d9dc05ba13a6481b1b031a53a22f
Those were temporarily reverted in
057028ed391f8, and never re-applied.
Re-committed by @ldionne (author edited for credit).
Differential Revision: https://reviews.llvm.org/D90168
Louis Dionne [Thu, 22 Oct 2020 20:34:19 +0000 (16:34 -0400)]
[llvm-lit] Improve the error message when make_paths_relative() fails
Previously, if make_paths_relative() failed due to some reason, it would
happily keep going and set the ${out_pathlist} to the standard output
of the command, which would be the empty string if the command failed.
This can lead to issues that are difficult to diagnose, since the calling
code will usually try to keep going with a variable that was set to the
empty string.
Differential Revision: https://reviews.llvm.org/D89985
LLVM GN Syncbot [Wed, 28 Oct 2020 20:46:36 +0000 (20:46 +0000)]
[gn build] Port
23ed570af1c