Kerry McLaughlin [Fri, 13 Nov 2020 10:51:17 +0000 (10:51 +0000)]
[SVE][CodeGen] Improve codegen of scalable masked scatters
If the scatter store is able to perform the sign/zero extend of
its index, this is folded into the instruction with refineIndexType().
Additionally, refineUniformBase() will return the base pointer and index
from an add + splat_vector.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D90942
Max Kazantsev [Fri, 13 Nov 2020 10:45:42 +0000 (17:45 +0700)]
[NFC] Move code between functions as a preparation step for further improvement
Simon Pilgrim [Fri, 13 Nov 2020 10:47:56 +0000 (10:47 +0000)]
[ValueTracking] computeKnownBitsFromShiftOperator use KnownBits direct for constant shift amounts.
Let KnownBits shift handlers deal with out-of-range shift amounts.
Max Kazantsev [Fri, 13 Nov 2020 10:18:16 +0000 (17:18 +0700)]
[NFC] Refactor lambda into static function
Raphael Isemann [Fri, 13 Nov 2020 10:39:21 +0000 (11:39 +0100)]
[lldb][NFC] Fix flaky TestForwardDeclFromStdModule test
"[lldb/DataFormatters] Display null C++ pointers as nullptr" added an assumption
that the member we check for is always a nullptr, but it is actually never
initialized. That causes the test to randomly fail due to the pointer having
some random value that isn't 0.
Simon Pilgrim [Fri, 13 Nov 2020 10:20:28 +0000 (10:20 +0000)]
Fix MSVC signed/unsigned comparison warning. NFCI.
Kazushi (Jam) Marukawa [Fri, 13 Nov 2020 09:27:05 +0000 (18:27 +0900)]
[VE] Add vst intrinsic instructions
Add vst intrinsic instructions and a regression test.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91406
Max Kazantsev [Fri, 13 Nov 2020 10:00:49 +0000 (17:00 +0700)]
[NFC] Move lambdae into static functions
Florian Hahn [Fri, 13 Nov 2020 09:42:36 +0000 (09:42 +0000)]
[docs] Fix undefined reference in ORCv2 design doc.
This fixes a typo introduced in
984e87923f1096c815cef900cda0926c68286ddf
which caused the docs build to fail.
Jay Foad [Fri, 13 Nov 2020 09:41:37 +0000 (09:41 +0000)]
[AMDGPU] One more use of the new export target names. NFC.
serge-sans-paille [Fri, 9 Oct 2020 16:41:21 +0000 (18:41 +0200)]
llvmbuildectomy - replace llvm-build by plain cmake
No longer rely on an external tool to build the llvm component layout.
Instead, leverage the existing `add_llvm_componentlibrary` cmake function and
introduce `add_llvm_component_group` to accurately describe component behavior.
These function store extra properties in the created targets. These properties
are processed once all components are defined to resolve library dependencies
and produce the header expected by llvm-config.
Differential Revision: https://reviews.llvm.org/D90848
serge-sans-paille [Tue, 10 Nov 2020 13:55:25 +0000 (14:55 +0100)]
[nfc] Fix missing include
Lang Hames [Fri, 13 Nov 2020 09:22:01 +0000 (20:22 +1100)]
[ORC][examples] Fix missing includes/dependencies in more examples.
Lang Hames [Fri, 13 Nov 2020 09:09:38 +0000 (20:09 +1100)]
[ORC] Make a narrowing conversion explicit.
Max Kazantsev [Fri, 13 Nov 2020 09:02:31 +0000 (16:02 +0700)]
[Test] One more IndVars test with inverted exit condition
Max Kazantsev [Fri, 13 Nov 2020 08:42:02 +0000 (15:42 +0700)]
[IndVars] Fix branches exiting by true with invariant conditions
Forgot to invert the condition for them.
Max Kazantsev [Fri, 13 Nov 2020 08:21:52 +0000 (15:21 +0700)]
[Test] Add test with inverted branch
Stephan Herhut [Thu, 12 Nov 2020 16:09:57 +0000 (17:09 +0100)]
[mlir][bufferize] Fix buffer promotion to stack for index types
The index type does not have a bitsize and hence the size of corresponding allocations cannot be computed. Instead, the promotion pass now has an explicit option to specify the size of index.
Differential Revision: https://reviews.llvm.org/D91360
Stephan Herhut [Thu, 12 Nov 2020 17:36:14 +0000 (18:36 +0100)]
[mlir][gpu] Only transform mapped parallel loops to GPU.
This exposes a hook to configure legality of operations such that only
`scf.parallel` operations that have mapping attributes are marked as
illegal. Consequently, the transformation can now also be applied to
mixed forms.
Differential Revision: https://reviews.llvm.org/D91340
Lang Hames [Fri, 13 Nov 2020 08:02:52 +0000 (19:02 +1100)]
[examples] Fix Kaleidoscope examples after OrcJIT break-up / remote TPC commit.
Fix the Kaleidoscope examples after
1d0676b54c4 by explicitly creating the
SymbolStringPool.
Kai Luo [Fri, 13 Nov 2020 08:05:08 +0000 (08:05 +0000)]
[PowerPC] Add test case for negated abs. NFC.
Max Kazantsev [Fri, 13 Nov 2020 07:39:19 +0000 (14:39 +0700)]
[NFC] Refactor loop-invariant getters to return Optional
River Riddle [Fri, 13 Nov 2020 07:45:07 +0000 (23:45 -0800)]
[mlir][Pass] Remove the verifierPass now that verification is run during normal pass execution
A recent refactoring removed the need to interleave verifier passes and instead opted to verify during the normal execution of passes instead. As such, the old verify pass is no longer necessary and can be removed.
Differential Revision: https://reviews.llvm.org/D91212
Lang Hames [Fri, 13 Nov 2020 07:40:40 +0000 (18:40 +1100)]
[examples] Fix Kaleidoscope examples after OrcJIT break-up / remote TPC commit.
Fix the Kaleidoscope examples after
1d0676b54c4 by explicitly creating the
SymbolStringPool.
River Riddle [Fri, 13 Nov 2020 07:33:43 +0000 (23:33 -0800)]
[mlir][Asm] Add support for resolving operation locations after parsing has finished
This revision adds support in the parser/printer for "deferrable" aliases, i.e. those that can be resolved after printing has finished. This allows for printing aliases for operation locations after the module instead of before, i.e. this is now supported:
```
"foo.op"() : () -> () loc(#loc)
#loc = loc("some_location")
```
Differential Revision: https://reviews.llvm.org/D91227
Jason Molenda [Fri, 13 Nov 2020 07:28:24 +0000 (23:28 -0800)]
debugserver should advance pc past builtin_debugtrap insn
On x86_64, when you hit a __builtin_debugtrap instruction, you
can continue past this in the debugger. This patch has debugserver
recognize the specific instruction used for __builtin_debugtrap
and advance the pc past it, so that the user can continue execution
once they've hit one of these.
In the patch discussion, we were in agreement that it would be better
to have this knowledge up in lldb instead of depending on each
stub rewriting the pc behind the debugger's back, but that's a
larger scale change for another day.
<rdar://problem/
65521634>
Differential revision: https://reviews.llvm.org/D91238
Lang Hames [Fri, 13 Nov 2020 07:11:14 +0000 (18:11 +1100)]
[ORC][examples] Fix include and library dependence for SpeculativeJIT example.
Serge Pavlov [Fri, 24 Apr 2020 10:53:05 +0000 (17:53 +0700)]
[Driver] Add option -fproc-stat-report
The new option `-fproc-stat-info=<file>` can be used to generate report
about used memory and execution tile of each stage of compilation.
Documentation for this option can be found in `UserManual.rst`. The
option can be used in parallel builds.
Differential Revision: https://reviews.llvm.org/D78903
Lang Hames [Fri, 13 Nov 2020 06:48:27 +0000 (17:48 +1100)]
[ORC] Add dependence of OrcJIT on OrcTargetProcess.
The SelfTargetProcessControl class depends on OrcTargetProcess.
River Riddle [Fri, 13 Nov 2020 07:02:25 +0000 (23:02 -0800)]
[mlir] Remove C++17 only use of inline on constexpr variable
River Riddle [Fri, 13 Nov 2020 06:55:28 +0000 (22:55 -0800)]
[mlir][Interfaces] Add implicit casts from concrete operation types to the interfaces they implement.
This removes the need to have an explicit `cast<>` given that we always know it `isa` instance of the interface.
Differential Revision: https://reviews.llvm.org/D91304
River Riddle [Fri, 13 Nov 2020 06:43:06 +0000 (22:43 -0800)]
[mlir][DenseElementsAttr] Allow for custom floating point types in `getValues`
Some users have native c++ data types that correspond to floating point values stored within a DenseElementsAttr that do not have a corresponding native C++ data type(e.g. bfloat16/half/etc). This revision allows for such users to use those native types directly, and removes the need to go through APFloat when the much faster native value path is available.
Differential Revision: https://reviews.llvm.org/D91402
Arthur Eubanks [Fri, 13 Nov 2020 06:20:39 +0000 (22:20 -0800)]
[NFC] Removed unused variable
Obsolete as of https://reviews.llvm.org/D91046.
Akira Hatanaka [Fri, 13 Nov 2020 02:33:29 +0000 (18:33 -0800)]
[ObjC][ARC] Clear the lists of basic blocks and instructions before
continuing the loop
This fixes a bug introduced in
c6f1713c46e61bbb8ece9ac5ac329d02e7f93228.
Lang Hames [Fri, 13 Nov 2020 06:15:13 +0000 (17:15 +1100)]
[ORC] Make WrapperFunctionResult::zeroInit static
Lang Hames [Fri, 13 Nov 2020 06:12:33 +0000 (17:12 +1100)]
[ORC] Remove designated initializer.
Lang Hames [Wed, 11 Nov 2020 00:55:24 +0000 (11:55 +1100)]
[ORC] Break up OrcJIT library, add Orc-RPC based remote TargetProcessControl
implementation.
This patch aims to improve support for out-of-process JITing using OrcV2. It
introduces two new class templates, OrcRPCTargetProcessControlBase and
OrcRPCTPCServer, which together implement the TargetProcessControl API by
forwarding operations to an execution process via an Orc-RPC Endpoint. These
utilities are used to implement out-of-process JITing from llvm-jitlink to
a new llvm-jitlink-executor tool.
This patch also breaks the OrcJIT library into three parts:
-- OrcTargetProcess: Contains code needed by the JIT execution process.
-- OrcShared: Contains code needed by the JIT execution and compiler
processes
-- OrcJIT: Everything else.
This break-up allows JIT executor processes to link against OrcTargetProcess
and OrcShared only, without having to link in all of OrcJIT. Clients executing
JIT'd code in-process should start linking against OrcTargetProcess as well as
OrcJIT.
In the near future these changes will enable:
-- Removal of the OrcRemoteTargetClient/OrcRemoteTargetServer class templates
which provided similar functionality in OrcV1.
-- Restoration of Chapter 5 of the Building-A-JIT tutorial series, which will
serve as a simple usage example for these APIs.
-- Implementation of lazy, cross-target compilation in lli's -jit-kind=orc-lazy
mode.
Jameson Nash [Fri, 13 Nov 2020 04:49:43 +0000 (23:49 -0500)]
[AsmPrinter] fix -disable-debug-info option
This option was in a rather convoluted place, causing global parameters
to be set in awkward and undesirable ways to try to account for it
indirectly. Add tests for the -disable-debug-info option and ensure we
don't print unintended markers from unintended places.
Reviewed By: dstenb
Differential Revision: https://reviews.llvm.org/D91083
Craig Topper [Fri, 13 Nov 2020 05:46:34 +0000 (21:46 -0800)]
[X86] Use EVT::getIntegerVT instead of MVT::getIntegerVT where the type can be i2 or i4.
This was a mistake introduced in D91294. I'm not sure how to
exercise this with the existing code, but I hit it while trying
some follow up experiments.
Craig Topper [Fri, 13 Nov 2020 04:18:50 +0000 (20:18 -0800)]
[X86] When storing v1i1/v2i1/v4i1 to memory, make sure we store zeros in the rest of the byte
We can't store garbage in the unused bits. It possible that something like zextload from i1/i2/i4 is created to read the memory. Those zextloads would be legalized assuming the extra bits are 0.
I'm not sure that the code in lowerStore is executed for the v1i1/v2i1/v4i1 case. It looks like the DAG combine in combineStore may have converted them to v8i1 first. And I think we're missing some cases to avoid going to the stack in the first place. But I don't have time to investigate those things at the moment so I wanted to focus on the correctness issue.
Should fix PR48147.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D91294
Max Kazantsev [Fri, 13 Nov 2020 05:05:14 +0000 (12:05 +0700)]
[IndVars] Replace checks with invariants if we cannot remove them
If we cannot prove that the check is trivially true, but can prove that it either
fails on the 1st iteration or never fails, we can replace it with first iteration check.
Differential Revision: https://reviews.llvm.org/D88527
Reviewed By: skatkov
Richard Smith [Fri, 13 Nov 2020 04:55:21 +0000 (20:55 -0800)]
Suppress trailing template arguments equivalent to default arguments
when printing the name of a member of a class template specialization.
Mehdi Amini [Fri, 13 Nov 2020 04:43:06 +0000 (04:43 +0000)]
Fix MLIR lit test configuration after cmake Python detection change
07f1047f41d changed the CMake detection to use find_package(Python3 ...
but didn't update the lit configuration to use the expected Python3_EXECUTABLE
cmake variable to point to the interpreter path.
This resulted in an empty path on MacOS.
Philip Reames [Fri, 13 Nov 2020 03:08:45 +0000 (19:08 -0800)]
[Tests][LoopVect] Exercise basic uniform memory operand logic
Shilei Tian [Fri, 13 Nov 2020 03:27:22 +0000 (22:27 -0500)]
[OpenMP] Fixed a bug when displaying affinity
Currently the affinity format string has initial value. When users set
the format via OMP_AFFINITY_FORMAT, it will overwrite the format string. However,
when copying the format, the tailing null is missing. As a result, if the user
format string is shorter than default value, the remaining part in the default
value still makes effort. This bug is not exposed because the test case doesn't
check the end of a string. It only checks whether given output "contains" the
check string.
Reviewed By: AndreyChurbanov
Differential Revision: https://reviews.llvm.org/D91309
Michael Liao [Tue, 10 Nov 2020 14:34:14 +0000 (09:34 -0500)]
[hip] Remove the coercion on aggregate kernel arguments.
- If an aggregate argument is indirectly accessed within kernels, direct
passing results in unpromotable `alloca`, which degrade performance
significantly. InferAddrSpace pass is enhanced in
[D91121](https://reviews.llvm.org/D91121) to take the assumption that
generic pointers loaded from the constant memory could be regarded
global ones. The need for the coercion on aggregate arguments is
mitigated.
Differential Revision: https://reviews.llvm.org/D89980
Michael Kruse [Fri, 13 Nov 2020 01:31:47 +0000 (19:31 -0600)]
[Polly] Fix memory leak.
Sanjay Patel [Fri, 13 Nov 2020 00:59:56 +0000 (19:59 -0500)]
[InstCombine] fold sub of low-bit masked value from offset of same value
There might be some demanded/known bits way to generalize this,
but I'm not seeing it right now.
This came up as a regression when I was looking at a different
demanded bits improvement.
https://rise4fun.com/Alive/5fl
Name: general
Pre: ((-1 << countTrailingZeros(C1)) & C2) == 0
%a1 = add i8 %x, C1
%a2 = and i8 %x, C2
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, ~C2
Name: test 1
%a1 = add i8 %x, 192
%a2 = and i8 %x, 10
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, -11
Name: test 2
%a1 = add i8 %x, -108
%a2 = and i8 %x, 3
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, -4
Sanjay Patel [Fri, 13 Nov 2020 00:44:13 +0000 (19:44 -0500)]
[InstCombine] add tests for sub with masked bits; NFC
Rahul Joshi [Thu, 12 Nov 2020 20:05:35 +0000 (12:05 -0800)]
[MLIR] Fix standard -> LLVM conversion to fail for unsupported memref element type.
- Move isSupportedMemRefType() to ConvertToLLVMPatterns and check if the
memref element type is supported there.
Differential Revision: https://reviews.llvm.org/D91374
peter klausler [Tue, 25 Aug 2020 16:43:16 +0000 (09:43 -0700)]
[flang] Document DO CONCURRENT's problems (NFC)
Differential Revision: https://reviews.llvm.org/D86556
Jonas Devlieghere [Thu, 12 Nov 2020 20:11:47 +0000 (12:11 -0800)]
[lldb/DataFormatters] Display null C++ pointers as nullptr
Display null pointer as `nullptr`, `nil` and `NULL` for C++,
Objective-C/Objective-C++ and C respectively. The original motivation
for this patch was to display a null std::string pointer as nullptr
instead of "", but the fix seemed generic enough to be done for all
summary providers.
Differential revision: https://reviews.llvm.org/D77153
Stanislav Mekhanoshin [Tue, 10 Nov 2020 00:40:35 +0000 (16:40 -0800)]
[AMDGPU] Remove scratch rsrc from spill pseudos
Differential Revision: https://reviews.llvm.org/D91110
Nico Weber [Thu, 12 Nov 2020 23:21:22 +0000 (18:21 -0500)]
[gn build] (manually) port
410626c9b56
Sean Silva [Thu, 12 Nov 2020 20:45:05 +0000 (12:45 -0800)]
[mlir] Make tensor_to_memref op docs match reality
The previous code defined it as allocating a new memref for its result.
However, this is not how it is treated by the dialect conversion framework,
that does the equivalent of inserting and folding it away internally
(even independent of any canonicalization patterns that we have
defined).
The semantics as they were previously written were also very
constraining: Nontrivial analysis is needed to prove that the new
allocation isn't needed for correctness (e.g. to avoid aliasing).
By removing those semantics, we avoid losing that information.
Differential Revision: https://reviews.llvm.org/D91382
Sean Silva [Tue, 10 Nov 2020 00:48:00 +0000 (16:48 -0800)]
[mlir] Bufferize tensor constant ops
We lower them to a std.global_memref (uniqued by constant value) + a
std.get_global_memref to produce the corresponding memref value.
This allows removing Linalg's somewhat hacky lowering of tensor
constants, now that std properly supports this.
Differential Revision: https://reviews.llvm.org/D91306
Sean Silva [Thu, 12 Nov 2020 18:33:04 +0000 (10:33 -0800)]
[mlir] Fix subtensor_insert bufferization.
It was incorrect in the presence of a tensor argument with multiple
uses.
The bufferization of subtensor_insert was writing into a converted
memref operand, but there is no guarantee that the converted memref for
that operand is safe to write into. In this case, the same converted
memref is written to in-place by the subtensor_insert bufferization,
violating the tensor-level semantics.
I left some comments in a TODO about ways forward on this. I will be
working actively on this problem in the coming days.
Differential Revision: https://reviews.llvm.org/D91371
Jessica Paquette [Tue, 3 Nov 2020 19:08:08 +0000 (11:08 -0800)]
[AArch64][GlobalISel] Select CSINC and CSINV for G_SELECT with constants
Select the following:
- G_SELECT cc, 0, 1 -> CSINC zreg, zreg, cc
- G_SELECT cc 0, -1 -> CSINV zreg, zreg cc
- G_SELECT cc, 1, f -> CSINC f, zreg, inv_cc
- G_SELECT cc, -1, f -> CSINV f, zreg, inv_cc
- G_SELECT cc, t, 1 -> CSINC t, zreg, cc
- G_SELECT cc, t, -1 -> CSINC t, zreg, cc
(IR example: https://godbolt.org/z/YfPna9)
These correspond to a bunch of the AArch64csel patterns in AArch64InstrInfo.td.
Unfortunately, it doesn't seem like we can import patterns that use NZCV like
those ones do. E.g.
```
def : Pat<(AArch64csel GPR32:$tval, (i32 1), (i32 imm:$cc), NZCV),
(CSINCWr GPR32:$tval, WZR, (i32 imm:$cc))>;
```
So we have to manually select these for now.
This replaces `selectSelectOpc` with an `emitSelect` function, which performs
these optimizations.
Differential Revision: https://reviews.llvm.org/D90701
Kazushi (Jam) Marukawa [Mon, 9 Nov 2020 21:50:15 +0000 (06:50 +0900)]
[VE] Support vld intrinsics
Add intrinsics for vector load instructions. Add a regression test also.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91332
Sanjay Patel [Thu, 12 Nov 2020 22:15:26 +0000 (17:15 -0500)]
[LoopVectorize] regenerate test checks; NFC
shafik [Thu, 12 Nov 2020 19:22:02 +0000 (11:22 -0800)]
[LLDB] Fix handling of bit-fields in a union
When parsing DWARF and laying out bit-fields we don't properly take into account when they are in a union, they will all have a zero offset.
Differential Revision: https://reviews.llvm.org/D91118
Sanjay Patel [Thu, 12 Nov 2020 22:01:38 +0000 (17:01 -0500)]
[PhaseOrdering] regenerate test checks; NFC
Sanjay Patel [Thu, 12 Nov 2020 21:37:13 +0000 (16:37 -0500)]
[InstCombine] add tests for low-mask-of-add; NFC
Amy Huang [Wed, 7 Aug 2019 23:38:58 +0000 (16:38 -0700)]
Some updates/fixes to the creduce script.
This was motivated by changes to llvm's `not --crash` disabling symbolization
but I ended up removing `not` from the script entirely because it
returns differently depending on whether clang "crashes" or exits for some
other reason. The script had to choose between calling `not` and `not --crash`
and sometimes it was wrong.
The script also now disables symbolization when we don't read the stack
trace because symbolizing is kind of slow.
Differential Revision: https://reviews.llvm.org/D91372
Stanislav Mekhanoshin [Thu, 12 Nov 2020 21:04:33 +0000 (13:04 -0800)]
[AMDGPU] Enable multi-dword flat scratch load/stores
Differential Revision: https://reviews.llvm.org/D91384
Stella Laurenzo [Wed, 11 Nov 2020 18:02:32 +0000 (10:02 -0800)]
[mlir][Python] Fix 'unreferenced local variable' warning on MSVC.
Differential Revision: https://reviews.llvm.org/D91282
Florian Hahn [Thu, 12 Nov 2020 20:13:13 +0000 (20:13 +0000)]
[PatternMatch] Add single index InsertValue matcher.
This patch adds a new matcher for single index InsertValue instructions,
similar to the existing matcher for ExtractValue.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D91352
Alexey Bataev [Thu, 12 Nov 2020 19:42:30 +0000 (11:42 -0800)]
[OPENMP]Fix PR47790: segfault in frontend while parsing Objective-C with OpenMP.
Need to check if the sema is actually finishing a function decl.
Differential Revision: https://reviews.llvm.org/D91376
Kazushi (Jam) Marukawa [Thu, 12 Nov 2020 11:41:04 +0000 (20:41 +0900)]
[VE] Disable -fsigaddr option for VE
VE needs to support integrated assembler and "nas". This "nas"
doesn't recognize ".sigaddr" pseudo mnemonics, so need to disable
it. This patch disable it on VE by default. Also add a regression
test for that.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91350
peter klausler [Thu, 12 Nov 2020 19:57:23 +0000 (11:57 -0800)]
[flang] Include source information in an invalid file-unit-number message
An io-unit that is an internal-file-variable is syntactically identical
to a file-unit-number expression that is a variable reference. An
ambiguous unit is initially parsed as an internal-file-variable. If
semantic analysis determines that the unit is not of character type,
it is rewritten as an internal-file-variable. This modification must
retain source coordinate information.
Differential revision: https://reviews.llvm.org/D91375
Joe Pletcher [Thu, 12 Nov 2020 20:37:35 +0000 (12:37 -0800)]
[fuzzer] Add Windows Visual C++ exception intercept
Adds a new option, `handle_winexcept` to try to intercept uncaught
Visual C++ exceptions on Windows. On Linux, such exceptions are handled
implicitly by `std::terminate()` raising `SIBABRT`. This option brings the
Windows behavior in line with Linux.
Unfortunately this exception code is intentionally undocumented, however
has remained stable for the last decade. More information can be found
here: https://devblogs.microsoft.com/oldnewthing/
20100730-00/?p=13273
Reviewed By: morehouse, metzman
Differential Revision: https://reviews.llvm.org/D89755
peter klausler [Thu, 12 Nov 2020 20:02:09 +0000 (12:02 -0800)]
[flang] Recognize END FILE as ENDFILE in free form source
The ENDFILE statement may be spelled as two words.
Differential revision: https://reviews.llvm.org/D91377
Arthur Eubanks [Thu, 5 Nov 2020 18:25:35 +0000 (10:25 -0800)]
[NFC][NewPM] Reuse PassBuilder callbacks with -O0
This removes lots of duplicated code which was necessary before
https://reviews.llvm.org/D89158.
Now we can use PassBuilder::runRegisteredEPCallbacks().
This is mostly sanitizers.
There is likely more that can be done to simplify, but let's start with this.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D90870
Alexey Bataev [Thu, 12 Nov 2020 18:20:22 +0000 (10:20 -0800)]
[OPENMP]Fix PR48076: Check map types array before accessing its front.
Need to check if there are map types for the components before trying to
access them when trying to modify type mappings for combined partial
mappings.
Differential Revision: https://reviews.llvm.org/D91370
Jay Foad [Wed, 11 Nov 2020 19:29:53 +0000 (19:29 +0000)]
[AMDGPU] Fix scheduling of exp pos4
Also fix a similar issue in SIInsertWaitcnts, but I don't think that fix
has any effect in practice.
Differential Revision: https://reviews.llvm.org/D91290
Jay Foad [Wed, 11 Nov 2020 19:21:02 +0000 (19:21 +0000)]
[AMDGPU] Define and use names for export targets. NFC.
Differential Revision: https://reviews.llvm.org/D91289
Jianzhou Zhao [Thu, 12 Nov 2020 19:47:39 +0000 (19:47 +0000)]
[msan] Break the getShadow loop after matching an argument
Reviewed-by: eugenis
Differential Revision: https://reviews.llvm.org/D91320
Zbigniew Sarbinowski [Thu, 12 Nov 2020 19:40:35 +0000 (14:40 -0500)]
[SystemZ][ZOS] libcxx - no posix memalign
The unavailability of posix_memalign on z/OS forces us to define _LIBCPP_HAS_NO_LIBRARY_ALIGNED_ALLOCATION'. The use of posix_memalign is being used in libcxx/src/new.cpp.
Reviewed By: #libc, ldionne
Differential Revision: https://reviews.llvm.org/D90178
Nikita Popov [Sat, 7 Nov 2020 18:02:49 +0000 (19:02 +0100)]
[BasicAA] Remove checks for GEP decomposition limit reached
The GEP aliasing code currently checks for the GEP decomposition
limit being reached (i.e., we did not reach the "final" underlying
object). As far as I can see, these checks are not necessary. It is
perfectly fine to work with a GEP whose base can still be further
decomposed.
Looking back through the commit history, these checks were originally
introduced in
1a444489e9d90915cfdda0720489893896ef1503. However, I
believe that the problem this was intended to address was later
properly fixed with
1726fc698ccb85fe4bb23c200a50f28b57fc53cb, and
the checks are no longer necessary since then (and were not the
right fix in the first place).
Differential Revision: https://reviews.llvm.org/D91010
Anatoliy Tomilov [Thu, 12 Nov 2020 19:42:52 +0000 (11:42 -0800)]
[Frontend] Treat .cuh files as CUDA source files
to synchronize with tools/clang-format/git-clang-format
tra: Keeping them in sync does have a minor benefit of not raising a question why the two maps are different.
Differential Revision: https://reviews.llvm.org/D91034
peter klausler [Thu, 12 Nov 2020 19:38:20 +0000 (11:38 -0800)]
fix clang build
Louis Dionne [Thu, 12 Nov 2020 19:34:35 +0000 (14:34 -0500)]
[libc++] NFC: Remove symbol from ABI list changelog that was never added
The `posix_memalign@GLIBC_2.2.5` symbol can't have been added by r284206,
because it doesn't show up in the corresponding ABI list. It's also not
defined in libc++, so that wouldn't make sense. It must have made it into
that comment by mistake.
Rahul Joshi [Thu, 12 Nov 2020 18:46:44 +0000 (10:46 -0800)]
[NFC] Switch printFunctionLikeOp and parseFunctionLikeOp to only support "inline" visibility.
- Remove the default valued arguments from these functions.
- Besides FuncOp, looks like no other in-tree op is using these functions.
Differential Revision: https://reviews.llvm.org/D91369
Nico Weber [Thu, 12 Nov 2020 18:54:24 +0000 (13:54 -0500)]
Revert "[gn build] (semi-manually) port
173b51169b8"
This reverts commit
37a1336de722c6920a24e8cd4278e396402f1b2a.
173b51169b8 was reverted in 777ca48.
Louis Dionne [Wed, 21 Oct 2020 15:11:45 +0000 (11:11 -0400)]
[libc++] Instantiate additional <iostream> members in the dylib
This commit adds new explicit instantiations for some classes in <iostream>
in the library. This is done after noticing that many programs that use
streams end up containing weak definitions of these classes, which has a
negative impact on both code size and load times (due to the need to
resolve weak symbols at load time). Note that we are just adding the
additional explicit instantiations for the `char` specializations, since
the `wchar_t` specializations are not used as often, and as a result there
wouldn't be a clear benefit.
This change is not an ABI break, since we are just adding additional
symbols.
Differential Revision: https://reviews.llvm.org/D90677
Louis Dionne [Thu, 12 Nov 2020 18:36:18 +0000 (13:36 -0500)]
Revert "[SystemZ][ZOS] Porting the time functions within libc++ to z/OS"
This reverts commit
173b51169b838. That commit was applied incorrectly,
and undid previous changes. That was clearly not intended.
Craig Topper [Thu, 12 Nov 2020 15:54:29 +0000 (07:54 -0800)]
[MSP430] Remove unused MVT::Glue output from MSP430ISD::SELECT_CC nodes.
Follow up from a similar patch on RISCV
637f19c36b323cc3ab597f6ef138db53be395949
Nothing reads this Glue value that I could see. The SDNode def in
the td file does not have the SDNPOutGlue flag so I don't think
this glue would get properly propagated to MachineSDNodes if it
was used.
peter klausler [Tue, 10 Nov 2020 23:13:02 +0000 (15:13 -0800)]
[flang] Implement runtime support for basic ALLOCATE/DEALLOCATE
Add error reporting infrastructure and support for ALLOCATE
and DEALLOCATE statements of intrinsic types without SOURCE=
or MOLD=.
Differential revision: https://reviews.llvm.org/D91215
Nathan James [Thu, 12 Nov 2020 18:19:11 +0000 (18:19 +0000)]
[clang-tidy] Merge options inplace instead of copying
Changed `ClangTidyOptions::mergeWith` to operate on the instance instead of returning a copy. The old mergeWith method has been renamed to merge and marked as nodiscard, to aid in disambiguating which one is which.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D91184
Aaron Ballman [Thu, 12 Nov 2020 18:06:30 +0000 (10:06 -0800)]
Non-implicit attribute creation requires a source range; NFC
There are two factory functions used to create a semantic attribute,
Create() and CreateImplicit(). CreateImplicit() does not need to
specify the source range of the attribute since it's an implicitly-
generated attribute. The same logic does not apply to Create(), so
this removes the default argument from those declarations to avoid
accidentally creating a semantic attribute without source location
information.
Fangrui Song [Thu, 12 Nov 2020 17:59:43 +0000 (09:59 -0800)]
[ELF] Don't consider SHF_ALLOC ".debug*" sections debug sections
Fixes PR48071
* The Rust compiler produces SHF_ALLOC `.debug_gdb_scripts` (which normally does not have the flag)
* `.debug_gdb_scripts` sections are removed from `inputSections` due to --strip-debug/--strip-all
* When processing --gc-sections, pieces of a SHF_MERGE section can be marked live separately
`=>` segfault when marking liveness of a `.debug_gdb_scripts` which is not split into pieces (because it is not in `inputSections`)
This patch circumvents the problem by not treating SHF_ALLOC ".debug*" as debug sections (to prevent --strip-debug's stripping)
(which is still useful on its own).
Reviewed By: grimar
Differential Revision: https://reviews.llvm.org/D91291
Kevin P. Neal [Thu, 12 Nov 2020 17:49:39 +0000 (12:49 -0500)]
[FPEnv][Clang][Driver] Use MarshallingInfoFlag for -fexperimental-strict-floating-point
As of D80952 we are disabling strict floating point on all hosts except
those that are explicitly listed as supported. Use of strict floating point
on other hosts requires use of the -fexperimental-strict-floating-point
flag. This is to avoid bugs like "https://bugs.llvm.org/show_bug.cgi?id=45329"
(which has an incorrect link in the previous review).
In the review for D80952 I was asked to mark the -fexperimental option as
a MarshallingInfoFlag. This patch does exactly that.
Differential Revision: https://reviews.llvm.org/D88987
Sanjay Patel [Thu, 12 Nov 2020 16:48:16 +0000 (11:48 -0500)]
[InstCombine] add tests for mask of sext-in-reg; NFC
Jamie Schmeiser [Thu, 12 Nov 2020 17:39:10 +0000 (17:39 +0000)]
Reland: Introduce -dot-cfg-mssa option which creates dot-cfg style file with mssa comments included in source
Summary:
Expand the print-memoryssa and print<memoryssa> passes with a new hidden
option -cfg-dot-mssa that names a file. When set, a dot-cfg style file
will be generated into the named file with the memoryssa comments retained
and those blocks containing them shown in light pink. The option does
nothing in isolation.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: asbirlea (Alina Sbirlea), dblaikie (David Blaikie)
Differential Revision: https://reviews.llvm.org/D90638
Nico Weber [Thu, 12 Nov 2020 17:15:09 +0000 (12:15 -0500)]
[gn build] (semi-manually) port
173b51169b8
Alexander Kornienko [Thu, 12 Nov 2020 17:14:06 +0000 (18:14 +0100)]
Fix unused variable warning in release builds
Simon Pilgrim [Thu, 12 Nov 2020 16:56:40 +0000 (16:56 +0000)]
[ValueTracking] Update computeKnownBitsFromShiftOperator callbacks to take KnownBits shift amount. NFCI.
We were creating this internally, but will need to support general KnownBits amounts as part of D90479.
Fangrui Song [Thu, 12 Nov 2020 16:56:12 +0000 (08:56 -0800)]
[ELF] Make SORT_INIT_PRIORITY support .ctors.N
Input sections `.ctors/.ctors.N` may go to either the output section `.init_array` or the output section `.ctors`:
* output `.ctors`: currently we sort them by name. This patch changes to sort by priority from high to low. If N in `.ctors.N` is in the form of %05u, there is no semantic difference. Actually GCC and Clang do use %05u. (In the test `ctors_dtors_priority.s` and Gold's test `gold/testsuite/script_test_14.s`, we can see %03u, but they are not really produced by compilers.)
* output `.init_array`: users can provide an input section description `SORT_BY_INIT_PRIORITY(.init_array.* .ctors.*)` to mix `.init_array.*` and `.ctors.*`. This can make .init_array.N and .ctors.(65535-N) interchangeable.
With this change, users can mix `.ctors.N` and `.init_array.N` in `.init_array` (PR44698 and PR48096) with linker scripts. As an example:
```
SECTIONS {
.init_array : {
*(SORT_BY_INIT_PRIORITY(.init_array.* .ctors.*))
*(.init_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .ctors)
}
} INSERT AFTER .fini_array;
SECTIONS {
.fini_array : {
*(SORT_BY_INIT_PRIORITY(.fini_array.* .dtors.*))
*(.fini_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .dtors)
}
} INSERT BEFORE .init_array;
```
Reviewed By: psmith
Differential Revision: https://reviews.llvm.org/D91187
Fangrui Song [Thu, 12 Nov 2020 16:53:11 +0000 (08:53 -0800)]
[ELF] Sort by input order within an input section description
According to
https://sourceware.org/binutils/docs/ld/Input-Section-Basics.html#Input-Section-Basics
for `*(.a .b)`, the order should match the input order:
* for `ld 1.o 2.o`, sections from 1.o precede sections from 2.o
* within a file, `.a` and `.b` appear in the section header table order
This patch implements the behavior. The interaction with `SORT*` and --sort-section is:
Matched sections are ordered by radix sort with the keys being `(SORT*, --sort-section, input order)`,
where `SORT*` (if present) is most significant.
> Note, multiple `SORT*` within an input section description has undocumented and
> confusing behaviors in GNU ld:
> https://sourceware.org/pipermail/binutils/2020-November/114083.html
> Therefore multiple `SORT*` is not the focus for this patch but
> this patch still strives to have an explainable behavior.
As an example, we partition `SORT(a.*) b.* c.* SORT(d.*)`, into
`SORT(a.*) | b.* c.* | SORT(d.*)` and perform sorting within groups. Sections
matched by patterns between two `SORT*` are sorted by input order. If
--sort-alignment is given, they are sorted by --sort-alignment, breaking tie by
input order.
This patch also allows a section to be matched by multiple patterns, previously
duplicated sections could occupy more space in the output and had erroneous zero bytes.
The patch is in preparation for support for
`*(SORT_BY_INIT_PRIORITY(.init_array.* .ctors.*)) *(.init_array .ctors)`,
which will allow LLD to mix .ctors*/.init_array* like GNU ld (gold's --ctors-in-init-array)
PR44698 and PR48096
Reviewed By: grimar, psmith
Differential Revision: https://reviews.llvm.org/D91127
Fangrui Song [Thu, 12 Nov 2020 16:46:53 +0000 (08:46 -0800)]
[ELF] Support multiple SORT in an input section description
The second `SORT` in `*(SORT(...) SORT(...))` is incorrectly parsed as a file pattern.
Fix the bug by stopping at `SORT*` in `readInputSectionsList`.
Reviewed By: grimar
Differential Revision: https://reviews.llvm.org/D91180