Pavel Labath [Wed, 3 Jun 2020 10:08:45 +0000 (12:08 +0200)]
[Support] Use outs() in ToolOutputFile
Summary:
If the output filename was specified as "-", the ToolOutputFile class
would create a brand new raw_ostream object referring to the stdout.
This patch changes it to reuse the llvm::outs() singleton.
At the moment, this change should be "NFC", but it does enable other
enhancements, like the automatic stdout/stderr synchronization as
discussed on D80803.
I've checked the history, and I did not find any indication that this
class *has* to use a brand new stream object instead of outs() --
indeed, it is special-casing "-" in a number of places already, so this
change fits the pattern pretty well. I suspect the main reason for the
current state of affairs is that the class was originally introduced
(r111595, in 2010) as a raw_fd_ostream subclass, which made any other
solution impossible.
Another potential benefit of this patch is that it makes it possible to
move the raw_ostream class out of the business of special-casing "-" for
stdout handling. That state of affairs does not seem appropriate because
"-" is a valid filename (albeit hard to access with a lot of command
line tools) on most systems. Handling "-" in ToolOutputFile seems more
appropriate.
To make this possible, this patch changes the return type of
llvm::outs() and errs() to raw_fd_ostream&. Previously the functions
were constructing objects of that type, but returning a generic
raw_ostream reference. This makes it possible for new ToolOutputFile and
other code to use raw_fd_ostream methods like error() on the outs()
object. This does not seem like a bad thing (since stdout is a file
descriptor which can be redirected to anywhere, it makes sense to ask it
whether the writing was successful or if it supports seeking), and
indeed a lot of code was already depending on this fact via the
ToolOutputFile "back door".
Reviewers: dblaikie, JDevlieghere, MaskRay, jhenderson
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81078
AndreyChurbanov [Thu, 4 Jun 2020 12:39:45 +0000 (15:39 +0300)]
[openmp] Fixed nonmonotonic schedule implementation.
Differential Revision: https://reviews.llvm.org/D80942
Sanjay Patel [Thu, 4 Jun 2020 12:33:24 +0000 (08:33 -0400)]
[x86] add FileCheck / assertions to test; NFC
Anastasia Stulova [Thu, 4 Jun 2020 11:29:02 +0000 (12:29 +0100)]
[OpenCL] Add cl_khr_extended_subgroup extensions.
Added extensions and their function declarations into
the standard header.
Patch by Piotr Fusik!
Tags: #clang
Differential Revision: https://reviews.llvm.org/D79781
Vince Bridgers [Mon, 1 Jun 2020 00:03:16 +0000 (19:03 -0500)]
[analyzer] Ignore calculated indices of <= 0 in VLASizeChecker
Summary:
See https://bugs.llvm.org/show_bug.cgi?id=46128. The checker does not
yet comprehend constraints involving multiple symbols, so it's possible
to calculate a VLA size that's negative or 0. A LIT is added to catch
regressions, and this change simply bails if a VLA size of 0 or less is
calculated.
Reviewers: balazske, NoQ, martong, baloghadamsoftware, Szelethus, gamesh411
Reviewed By: balazske, NoQ, Szelethus
Subscribers: xazax.hun, szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy, Charusso, ASDenysPetrov, cfe-commits, dkrupp
Tags: #clang
Differential Revision: https://reviews.llvm.org/D80903
Sanjay Patel [Thu, 4 Jun 2020 12:17:53 +0000 (08:17 -0400)]
[PhaseOrdering] add more tests for vector reductions; NFC
More coverage for D80867.
Simon Moll [Thu, 4 Jun 2020 09:09:48 +0000 (11:09 +0200)]
[VP][Fix] canIgnoreVectorLength for scalable types
This patch fixes VPIntrinsic::canIgnoreVectorLength when used on a
VPIntrinsic with scalable vector types. Also includes new unittest cases
for the '<vscale x 1 x whatever>' and '%evl == vscale' corner cases.
David Green [Thu, 4 Jun 2020 11:31:42 +0000 (12:31 +0100)]
[ARM] FP16 conversion tests. NFC
Nico Weber [Thu, 4 Jun 2020 12:10:16 +0000 (08:10 -0400)]
[gn build] port some -Wno flags for gcc from the cmake build.
Eduardo Caldas [Thu, 4 Jun 2020 12:03:04 +0000 (14:03 +0200)]
Add support for IntegerLiteral in SyntaxTree
Reviewers: gribozavr2
Reviewed By: gribozavr2
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81135
Max Kazantsev [Thu, 4 Jun 2020 11:54:23 +0000 (18:54 +0700)]
[InstCombine][NFC] Factor out constant check
We plan to add more transforms here. Besides, this check should be
done in the beginning just from function's name.
Max Kazantsev [Thu, 4 Jun 2020 11:14:30 +0000 (18:14 +0700)]
[Test] Add test showing missing opportunity of folding ICmp(Phi(Consts...))
Georgii Rymar [Thu, 4 Jun 2020 11:22:06 +0000 (14:22 +0300)]
[ObjectYAML] - Remove unused function. NFC.
Was introduced in D81005 by mistake.
Catched by BB:
http://lab.llvm.org:8011/builders/clang-ppc64le-rhel/builds/4070/steps/build%20stage%201/logs/stdio
Paul Walker [Mon, 1 Jun 2020 10:09:58 +0000 (10:09 +0000)]
[FileCheck] Implement equality operators for ExpressionValue.
Subscribers: hiraditya, thopre, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81094
Yvan Roux [Thu, 4 Jun 2020 11:12:26 +0000 (13:12 +0200)]
[ARM][MachineOutliner] Remove unneeded dynamic allocation.
Simon Pilgrim [Thu, 4 Jun 2020 10:49:28 +0000 (11:49 +0100)]
[DAG] scalarizeBinOpOfSplats - extract from the source of splat vector (PR46189)
D79003/rG9fa58d1bf2f8 exposed an issue with scalarizeBinOpOfSplats that we were extracting from the splatted vector result instead of the source, the splat index is only valid for the source vector not the result, which may contain undefs, including at the splat index.
Tim Northover [Thu, 4 Jun 2020 10:15:03 +0000 (11:15 +0100)]
Revert "[DAGCombiner] avoid unnecessary indirection from SDNode/SDValue; NFCI"
This reverts commit
21dadd774f56778ef68c1ce307205dfbdacc793a.
In at least PromoteIntBinOps, they wanted to know about users of *all* values
produced by the node not just the integer being promoted. For example not
replacing chain users if the operation was a load breaks the ordering of the
DAG.
Georgii Rymar [Mon, 1 Jun 2020 09:13:02 +0000 (12:13 +0300)]
[yaml2obj] - Add a way to exclude specified sections from the section header.
This implements a new "Excluded" key that can be used
to exclude entries from section header:
```
SectionHeaderTable:
Sections:
...
Excluded:
- Name: .foo
```
Differential revision: https://reviews.llvm.org/D81005
Djordje Todorovic [Thu, 4 Jun 2020 09:59:04 +0000 (11:59 +0200)]
[CSInfo][MIPS] Describe parameter value loaded by ADDiu
Describe parameter's value loaded by MIPS ADDiu instruction.
When parameter's value is loaded into a register by mips ADDiu/DADDiu
instruction, it could be described correctly and emitted as
DW_AT_GNU_call_site_value.
Patch by Nikola Tesic
Differential revision: https://reviews.llvm.org/D78108
Florian Hahn [Thu, 4 Jun 2020 10:15:01 +0000 (11:15 +0100)]
[Sema] Remove unused matrix_begin/end helpers (NFC).
The matrix_types() helper returning an iterator range is used instead.
Simon Pilgrim [Thu, 4 Jun 2020 09:43:19 +0000 (10:43 +0100)]
[X86] Add test case for PR46189
Georgii Rymar [Thu, 4 Jun 2020 10:10:43 +0000 (13:10 +0300)]
Revert "[yaml2obj] - Allocate the file space for SHT_NOBITS sections in some cases."
This reverts commit
aa3a85cdaa4432ca389bdbf48049eaa64dc6e266.
There are problems with it. See here: https://reviews.llvm.org/D80629
Esme-Yi [Thu, 4 Jun 2020 10:09:06 +0000 (10:09 +0000)]
[PowerPC][NFC] Testing ROTL of v1i128.
Summary: A bug is reported in bugzilla-45628, where the swap_with_shift case can’t be matched to a single HW instruction xxswapd as expected. In fact the case matches the idiom of rotate, but PPC doesn’t support ROTL v1i128.
This is a NFC patch for testing ROTL with v1i128 at master.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D81073
Kadir Cetinkaya [Thu, 4 Jun 2020 10:02:49 +0000 (12:02 +0200)]
[clangd] Fix build for gcc 7.4
Vitaly Buka [Thu, 4 Jun 2020 09:38:58 +0000 (02:38 -0700)]
[StackSafety] Rename testing opts
Vitaly Buka [Tue, 2 Jun 2020 04:40:47 +0000 (21:40 -0700)]
[StackSafety,NFC] Remove SCEVRewriteVisitor
Summary: Depends on D80956.
Reviewers: eugenis
Reviewed By: eugenis
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80976
Jay Foad [Wed, 11 Mar 2020 11:33:30 +0000 (11:33 +0000)]
[AMDGPU] More accurate gfx10 latencies
Differential Revision: https://reviews.llvm.org/D81012
Jay Foad [Tue, 10 Mar 2020 13:08:58 +0000 (13:08 +0000)]
[AMDGPU] Introduce new sched classes for transcendental instructions
This is in preparation for scheduling them slightly differently on
gfx10. NFC.
Differential Revision: https://reviews.llvm.org/D81011
Kazushi (Jam) Marukawa [Thu, 4 Jun 2020 09:27:57 +0000 (11:27 +0200)]
[VE] Clean SDNodeXForm stuff
Summary:
Gather definitions of SDNodeXForm and change them to call C functions
instead of copying C expressions in td files. Doing this solved some
bugs in mimm detections.
Differential Revision: https://reviews.llvm.org/D81132
Georgii Rymar [Wed, 3 Jun 2020 15:19:05 +0000 (18:19 +0300)]
[llvm-readelf] - Do not try to read past the end of the file when dumping the the SHT_GNU_HASH.
We have unobvious issue in the condition that is used to check
that we do not read past the EOF.
The problem is that the result of "GnuHashTable->nbuckets * 4" expression is uint32.
Because of that it was still possible to overflow it and pass the check.
There was no such problem with the "GnuHashTable->maskwords * sizeof(typename ELFT::Off)"
condition, because of `sizeof` on the right (which gives 64-bits value on x64),
but I've added an explicit conversion to 64-bit value for `GnuHashTable->maskwords` too.
Differential revision: https://reviews.llvm.org/D81103
Dmitri Gribenko [Thu, 4 Jun 2020 08:35:43 +0000 (10:35 +0200)]
Make syntax tree test print the line number when it fails
Summary:
The syntax tree test uses a helper function that executes all testing
assertions. When an assertion fails, the only line number that gets
printed to the log refers to the helper function. After this change, we
would also get the line number of the EXPECT_TRUE macro invocation
(unfortunately, the line number of the last token of it, not the first
one, but there's not much I can do about it).
Reviewers: hlopko, eduucaldas
Reviewed By: hlopko, eduucaldas
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81107
Qiu Chaofan [Thu, 4 Jun 2020 08:30:17 +0000 (16:30 +0800)]
[PowerPC] Require nsz flag for c-a*b to FNMSUB
On PowerPC, FNMSUB (both VSX and non-VSX version) means -(a*b-c). But
the backend used to generate these instructions regardless whether nsz
flag exists or not. If a*b-c==0, such transformation changes sign of
zero.
This patch introduces PPC specific FNMSUB ISD opcode, which may help
improving combined FMA code sequence.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D76585
Yevgeny Rouban [Thu, 4 Jun 2020 08:34:14 +0000 (15:34 +0700)]
Extend InvokeInst !prof branch_weights metadata to unwind branches
Allow InvokeInst to have the second optional prof branch weight for
its unwind branch. InvokeInst is a terminator with two successors.
It might have its unwind branch taken many times. If so
the BranchProbabilityInfo unwind branch heuristic can be inaccurate.
This patch allows a higher accuracy calculated with both branch
weights set.
Changes:
- A new section about InvokeInst is added to
the BranchWeightMetadata page. It states the old information that
missed in the doc and adds new about the second branch weight.
- Verifier is changed to allow either 1 or 2 branch weights
for InvokeInst.
- A new test is written for BranchProbabilityInfo to demonstrate
the main improvement of the simple fix in calcMetadataWeights().
- Several new testcases are created for Inliner. Those check that
both weights are accounted for invoke instruction weight
calculation.
- PGOUseFunc::setBranchWeights() is fixed to be applicable to
InvokeInst.
Reviewers: davidxl, reames, xur, yamauchi
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80618
Raphael Isemann [Thu, 4 Jun 2020 08:27:03 +0000 (10:27 +0200)]
[lldb][NFC] Address some review feedback for D80775 ('command script delete' completion)
In the similar review D81128, Jonas pointed out some style errors that also
apply to D80775 (which is already committed). Also applying the changes
suggested there to this code.
Jan Korous [Thu, 4 Jun 2020 04:39:55 +0000 (21:39 -0700)]
Relands "[YAMLVFSWriter][Test][NFC] Add couple tests" vol. 2
This reverts commit
e4e3e41905d182c0f3d5b0b9406e3cbf2aabb30f.
Fixed dangling StringRef in test.
Gongyu Deng [Thu, 4 Jun 2020 07:53:05 +0000 (09:53 +0200)]
[lldb] tab completion for `command script delete'
Summary: Added the tab completion for `command script delete`.
Reviewers: teemperor, JDevlieghere
Reviewed By: teemperor
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D80775
Jan Korous [Thu, 4 Jun 2020 08:10:49 +0000 (01:10 -0700)]
[Support][NFC] Tests for root_name(), root_directory() and root_path()
It's literally just doc comments converted to unittests.
Yevgeny Rouban [Thu, 4 Jun 2020 07:30:58 +0000 (14:30 +0700)]
[Instruction] Remove setProfWeight()
Remove the function Instruction::setProfWeight() and make
use of Instruction::copyMetadata(.., {LLVMContext::MD_prof}).
This is correct for all use cases of setProfWeight() as it
is applied to CallBase instructions only.
This change results in prof metadata copied intact even if
the source has "VP". The old pair of calls
extractProfTotalWeight() + setProfWeight() resulted in
setting branch_weights if the source had "VP" data.
Reviewers: yamauchi, davidxl
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80987
Mikael Holmen [Thu, 4 Jun 2020 07:38:56 +0000 (09:38 +0200)]
[WebAssembly] Fix gcc warning [NFC]
gcc 7.4 complained with
../lib/Target/WebAssembly/WebAssemblyFixBrTableDefaults.cpp:125:23: warning: extra ';' [-Wpedantic]
false);
^
Sam Parker [Tue, 2 Jun 2020 11:29:42 +0000 (12:29 +0100)]
[NFCI][CostModel][AMDGPU] Simplify getUserCost
Casts and intrinsics are now handled by the default implementation
of getUserCost, so remove them from the backends switch statement.
https://reviews.llvm.org/D80994
Jean Perier [Thu, 4 Jun 2020 07:41:28 +0000 (09:41 +0200)]
[flang] Fix IsConstantExpr for division expressions
Summary:
Fortran::evaluate::IsConstantExpr did not check that the numerator
was a constant expression. This patch fixes the issue.
Reviewers: DavidTruby, klausler, schweitz, PeteSteinfeld, jdoerfert, sscalpone
Reviewed By: klausler, PeteSteinfeld, sscalpone
Subscribers: llvm-commits
Tags: #llvm, #flang
Differential Revision: https://reviews.llvm.org/D81096
Kazu Hirata [Thu, 4 Jun 2020 07:40:17 +0000 (00:40 -0700)]
[Inlining] Introduce -enable-npm-pgo-inline-deferral
Summary:
Experiments show that inline deferral past pre-inlining slightly
pessimizes the performance.
This patch introduces an option to control inline deferral during PGO.
The option defaults to true for now (that is, NFC).
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: eraman, hiraditya, haicheng, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80776
Craig Topper [Thu, 4 Jun 2020 06:19:03 +0000 (23:19 -0700)]
[X86] Remove (V)MOVHPDrm patterns that involve bitcast+scalar_to_vec+loadi64.
I think these are left over from when we used to type legalize
v2f32 loads using bitcast+scalar_to_vec+loadi64 on 64-bit targets.
These days we use loadf64. If this becomes a problem a better
solution would be a DAG combine to turn it into scalar_to_vec+loadf64.
Kazushi (Jam) Marukawa [Thu, 4 Jun 2020 07:05:19 +0000 (09:05 +0200)]
[VE] Change to use EXTRACT_SUBREG instead of COPY_TO_REGCLASS
Summary:
Change to use EXTRACT_SUBREG instead of COPY_TO_REGCLASS in order to
remove unnecessary copy instructions.
Differential Revision: https://reviews.llvm.org/D81129
David Sherwood [Tue, 2 Jun 2020 13:44:24 +0000 (14:44 +0100)]
[SVE] Fix ubsan issues in DecodeIITType
In an earlier patch I removed the need for
IITDescriptor::ScalableVecArgument, which involved changing
DecodeIITType to pull out the last IIT_Info from the list. However,
it turns out this is unsafe and causes ubsan failures. I've tried to
fix this a different way by simply passing the last IIT_Info as an
additional argument to DecodeIITType.
Differential Revision: https://reviews.llvm.org/D81057
Martin Boehme [Fri, 29 May 2020 08:36:19 +0000 (10:36 +0200)]
[clang] Always allow including builtin headers in [no_undeclared_headers] modules.
Previously, this would fail if the builtin headers had been "claimed" by
a different module that wraps these builtin headers. libc++ does this,
for example.
This change adds a test demonstrating this situation; the test fails
without the fix.
Madhur Amilkanthwar [Thu, 4 Jun 2020 04:21:18 +0000 (09:51 +0530)]
Utility to dump .dot representation of SelectionDAG without firing viewer
Summary:
This patch adds support for dumping .dot
representation of SelectionDAG. It is inspired from the fact that,
a developer may want to just dump the graph at
a predictable path with a simple name to compare.
The exisitng utility (i.e. viewGraph) are overkill
for this motive hence this patch adds the requires support
while using the core routines from GraphWriter.
Example usage: DAG.dumpDotGraph("/tmp/graph.dot", "MyGraph")
will create /tmp/graph.dot file when DAG is an
object of SelectionDAG class.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D80711
Siva Chandra Reddy [Thu, 4 Jun 2020 05:39:34 +0000 (22:39 -0700)]
[libc][NFC] Add ceil[f], floor[f] and trunc[f] to the spec and config files.
Fangrui Song [Thu, 4 Jun 2020 05:05:15 +0000 (22:05 -0700)]
[gcov] Don't error 'unexpected end of memory buffe'
Fangrui Song [Thu, 4 Jun 2020 04:14:54 +0000 (21:14 -0700)]
[gcov] Make `Creating 'filename'` compatible with gcov
And clean up llvm-cov.test a bit
Philip Reames [Thu, 4 Jun 2020 03:31:17 +0000 (20:31 -0700)]
[Statepoint] Remove last of old ImmutableStatepoint code
To do so, I had to sink the old school inline operand handling into GCStatepointInst which is non ideal. This code should be removed shortly and I was able to at least clean it up a bunch.
Philip Reames [Thu, 4 Jun 2020 03:05:19 +0000 (20:05 -0700)]
[Statepoint] Delete more dead code from old wrappers
The verify() routine duplicates IR/Verifier.cpp checks, so while not technically dead it doesn't add any value either.
Philip Reames [Thu, 4 Jun 2020 02:49:25 +0000 (19:49 -0700)]
[Statepoint] Start the process of removing old interfaces
We introduced the GCStatepointInst class and have migrated almost all users of Statepoint/ImmutableStatepoint to the new API. Given downstream consumers have had a week to migrate, remove code which is now dead.
Eric Christopher [Thu, 4 Jun 2020 02:59:26 +0000 (19:59 -0700)]
Replace integer usage with enumeration.
Fangrui Song [Thu, 4 Jun 2020 02:58:28 +0000 (19:58 -0700)]
[gcov] Delete XFAIL: host-byteorder-big-endian
Add two to function-numbering.ll and exit-block.ll
Fangrui Song [Thu, 4 Jun 2020 02:29:21 +0000 (19:29 -0700)]
[gcov] Improve .gcno compatibility with gcov and use DataExtractor
llvm-cov.test and many Inputs/test* files contain wrong tests.
This patch rewrites a large portion of these files.
The pre-canned .gcno & .gcda are replaced by binaries produced by
clang --coverage (compatible with gcov 4.8~7)
(after some GCDAProfiling.c bugs were fixed by my previous commits).
Also make llvm-cov gcov on a little-endian host capable to parse big-endian .gcno and .gcda,
and make llvm-cov gcov on big-endian host capable to parse little-endian .gcno and .gcda
Yaxun (Sam) Liu [Fri, 1 May 2020 15:30:24 +0000 (11:30 -0400)]
[CUDA][HIP] Fix constexpr variables for C++17
constexpr variables are compile time constants and implicitly const, therefore
they are safe to emit on both device and host side. Besides, in many cases
they are intended for both device and host, therefore it makes sense
to emit them on both device and host sides if necessary.
In most cases constexpr variables are used as rvalue and the variables
themselves do not need to be emitted. However if their address is taken,
then they need to be emitted.
For C++14, clang is able to handle that since clang emits them with
available_externally linkage together with the initializer.
However for C++17, the constexpr static data member of a class or template class
become inline variables implicitly. Therefore they become definitions with
linkonce_odr or weak_odr linkages. As such, they can not have available_externally
linkage.
This patch fixes that by adding implicit constant attribute to
file scope constexpr variables and constexpr static data members
in device compilation.
Differential Revision: https://reviews.llvm.org/D79237
Xing GUO [Thu, 4 Jun 2020 01:25:26 +0000 (09:25 +0800)]
[ObjectYAML][ELF] Let the endianness of DWARF sections be inferred from FileHeader.
This patch helps infer the endianness of DWARF sections from `FileHeader`.
Reviewed By: jhenderson, grimar
Differential Revision: https://reviews.llvm.org/D81051
Xing GUO [Thu, 4 Jun 2020 00:53:40 +0000 (08:53 +0800)]
[ObjectYAML][DWARF] Support emitting the .debug_aranges section in ELFYAML.
This patch enables yaml2obj to emit the .debug_aranges section in ELFYAML.
Known issues:
- The current implementation of `debug_aranges` doesn't support emitting `segment` in the `(segment, address, length)` tuple. I will fix it in a follow-up patch.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D80972
Julian Lettner [Thu, 21 May 2020 18:14:55 +0000 (11:14 -0700)]
[Darwin] Add tests for OS version checking
Extract ParseVersion helper function for testing.
Reviewed By: delcypher
Differential Revision: https://reviews.llvm.org/D80761
Matt Arsenault [Sun, 31 May 2020 15:58:56 +0000 (11:58 -0400)]
GlobalISel: Start defining strict FP instructions
The AMDGPU lowering for unconstrained G_FDIV sometimes needs to
introduce a mode switch in the middle, so it's helpful to have
constrained instructions available to legalize this. Right now nothing
is preventing reordering of the mode switch with the other
instructions in the expansion.
Quentin Colombet [Wed, 3 Jun 2020 23:15:23 +0000 (16:15 -0700)]
[RegisterCoalescer] Update empty subranges when rematerializing
When we rematerialize a value as part of the coalescing, we may
widen the register class of the destination register.
When this happens, updateRegDefUses may create additional subranges
to account for the wider register class.
The created subranges are empty and if they are not defined by
the rematerialized instruction we clean them up.
However, if they are defined by the rematerialized instruction but
unused, we failed to flag them as dead definition and would leave
them as empty live-range.
This is wrong because empty live-ranges don't interfere with anything,
thus if we don't fix them, we would fail to account that the
rematerialized instruction clobbers some lanes.
E.g., let us consider the following pseudo code:
def.lane_low64:reg128 = ldimm
newdef:reg32 = COPY def.lane_low64_low32
When rematerialization happens for newdef, we end up with:
newdef.lane_low64:reg128 = ldimm
= use newdef.lane_low64_low32
Let's look at the live interval of newdef.
Before rematerialization, we would get:
newdef [defIdx, useIdx:0) 0@defIdx
Right after updateRegDefUses, newdef register class is widen to reg128
and the subrange definitions will be augmented to fill the subreg that
is used at the definition point, here lane_low64.
The resulting live interval would be:
newdef [newDefIdx, useIdx:0) 0@newDefIdx
* lane_low64_high32 EMPTY
* lane_low64_low32 [newDefIdx, useIdx:0)
Before this patch this would be the final status of the live interval.
Therefore we miss that lane_low64_high32 is actually live on the
definition point of newdef.
With this patch, after rematerializing, we check all the added subranges
and for the ones that are defined but empty, we flag them as dead def.
Thus, in that case, newdef would look like this:
newdef [newDefIdx, useIdx:0) 0@newDefIdx
* lane_low64_high32 [newDefIdx, newDefIdxDead) ; <-- instead of EMPTY
* lane_low64_low32 [newDefIdx, useIdx:0)
This fixes https://www.llvm.org/PR46154
Julian Lettner [Wed, 3 Jun 2020 23:59:59 +0000 (16:59 -0700)]
Eric Christopher [Wed, 3 Jun 2020 23:59:36 +0000 (16:59 -0700)]
consitfy and auto -> auto * a few places to clean up uses.
Matt Arsenault [Sun, 31 May 2020 20:18:52 +0000 (16:18 -0400)]
AArch64/GlobalISel: Fix assert on call returning 0 sized type
I don't know why this is considered valid IR, but it probably should
not be.
Matt Arsenault [Sat, 30 May 2020 15:04:15 +0000 (11:04 -0400)]
AMDGPU/GlobalISel: Handle uniform G_DYN_STACKALLOC
Matt Arsenault [Sat, 30 May 2020 14:54:43 +0000 (10:54 -0400)]
GlobalISel: Fail expansion of G_DYN_STACKALLOC for StackGrowsUp
Saleem Abdulrasool [Wed, 3 Jun 2020 19:20:35 +0000 (19:20 +0000)]
lld: add basic static library search
This is a very basic static library search addition. This is the pre-Xcode4
behaviour of searching all paths for the shared version before searching for
the static version of the library. This behaviour is supposed to be inverted
with `-search_paths_first` being the default. This adds the library search
with the intention of providing the setup to merge the paths into one path
and making it controllable by `OPT_search_paths_first`.
Vitaly Buka [Wed, 3 Jun 2020 23:12:08 +0000 (16:12 -0700)]
[StackSafety,NFC] Add statistic counters
Vitaly Buka [Wed, 3 Jun 2020 22:30:03 +0000 (15:30 -0700)]
Fix "Statistics are disabled"
There is no -DLLVM_ENABLE_STATS, only
-DLLVM_FORCE_ENABLE_STATS.
It was renamed by
6cf299cf01e4a83844126f7faf17cbeb78e88da9
Eric Christopher [Wed, 3 Jun 2020 21:20:46 +0000 (14:20 -0700)]
Fix typo in filename comment.
Philip Reames [Wed, 3 Jun 2020 22:21:11 +0000 (15:21 -0700)]
[Statepoints][CGP] Minor parameter type cleanup
Jessica Paquette [Wed, 3 Jun 2020 22:48:22 +0000 (15:48 -0700)]
[AArch64][NFC] Regenerate arm64-rev.ll
Test had some GISel stuff in it which was changed by
969d2d1ea94.
Thomas Lively [Wed, 3 Jun 2020 22:36:44 +0000 (15:36 -0700)]
[WebAssembly] Fix ISel crash in SIGN_EXTEND_INREG lowering
Summary:
The code previously assumed that the index of a vector extract was
constant, but this was not always true. This patch fixes the problem
by bailing out of the lowering if the index is nonconstant and also
replaces `static_cast`s in the lowering function with `cast`s because
the latter contain type-checking asserts that would make similar
issues easier to find and debug.
Reviewers: aheejin
Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81025
Jessica Paquette [Tue, 2 Jun 2020 18:13:56 +0000 (11:13 -0700)]
[AArch64][GlobalISel] Add selection support for rev16, rev32, and rev64
This does three things:
1) Adds G_REV16, G_REV32, and G_REV64. These are equivalent to AArch64rev16,
AArch64rev32, and AArch64rev64 respectively.
2) Adds support for producing G_REV64 in the postlegalizer combiner.
We don't legalize any of the shuffles which could give us a G_REV32 or
G_REV16 yet. Since the function for detecting the rev mask is lifted from
AArch64ISelLowering, it should work for G_REV32 and G_REV16 when we get
there.
3) Adds a selection test for a good portion of the patterns imported for the rev
family. The only ones which are not tested are the ones with bitconvert.
This also does a little cleanup, and adds a struct for shuffle vector pseudo
matchdata. This lets us still use `applyShuffleVectorPseudo` rather than adding
a new function.
It should also make it a bit easier to port some of the other masks from
AArch64ISelLowering. (e.g. `isZIP_v_undef_Mask` and friends)
Differential Revision: https://reviews.llvm.org/D81112
Julian Lettner [Wed, 3 Jun 2020 22:24:53 +0000 (15:24 -0700)]
Fix bug in newly added VersionBase::operator>=
Fixup for
ba6b1b4353e33a7a36bcbad1d1c1157826197fd2.
Philip Reames [Wed, 3 Jun 2020 22:09:03 +0000 (15:09 -0700)]
[Statepoint] Fix signed vs unsigned in index handling
As noted in a comment on D80937, all of these are specified as unsigned values, but the verifier code was using signed. Given the practical values involved, the different in range didn't matter, but we might as well clean it up.
Jessica Paquette [Tue, 2 Jun 2020 16:30:04 +0000 (09:30 -0700)]
[AArch64][GlobalISel] Select uzp1 and uzp2
Porting the mask stuff for uzp1 and uzp2 from AArch64ISelLowering.
Add two custom opcodes: G_UZP1 and G_UZP2.
Produce them in the post-legalizer combiner when the mask checks out.
Tests:
- postlegalizer-combiner-uzp.mir verifies that we create G_UZP1 and G_UZP2.
The testcases that check that we create them come from neon-perm.ll.
- select-uzp.mir verifies that we can select G_UZP1 and G_UZP2.
Differential Revision: https://reviews.llvm.org/D81049
Hanhan Wang [Wed, 3 Jun 2020 22:00:33 +0000 (15:00 -0700)]
[mlir][StandardToSPIRV] Handle i1 case for lowering std.zexti to SPIR-V.
Differential Revision: https://reviews.llvm.org/D80965
Philip Reames [Wed, 3 Jun 2020 21:56:24 +0000 (14:56 -0700)]
Introduce a "gc-live" bundle for the gc arguments of a statepoint
Currently, gc.relocates are defined in terms of indices into the statepoint's operand list. Given the gc args are at the end of a variable length list of operands, this makes interpreting their indices by hand a tad challenging. We can simplify the statepoint sequence and improve readability quite a bit by pulling these new operands into their own named operand bundle.
This patch defines a new operand bundle tag "gc-live". The semantics of the bundle are the same as the existing gc arguments of a statepoint. This patch simply introduces the definition and codegen for the bundle, future patches will migrate RS4GC to emitting the new form.
Interestingly, with this done and the recent migration to using deopt and gc-transition bundles, we really don't have much left in the statepoint itself. It really looks like the existing ID and flags fields are redundant; we have (existing!) attributes for all of them. I think we'll be able to reduce the gc.statepoint signature to simply a wrapped call (e.g. actual target and actual arguments).
Differential Revision: https://reviews.llvm.org/D80937
Hanhan Wang [Wed, 3 Jun 2020 21:55:10 +0000 (14:55 -0700)]
[mlir][Linalg] Add support for fusion between indexed_generic ops and tensor_reshape ops
Summary:
The fusion for tensor_reshape is embedding the information to indexing maps,
thus the exising pattenr also works for indexed_generic ops.
Depends On D80347
Differential Revision: https://reviews.llvm.org/D80348
Hanhan Wang [Wed, 3 Jun 2020 21:58:19 +0000 (14:58 -0700)]
[mlir][Linalg] Add support for fusion between indexed_generic ops and generic ops on tensors.
Summary:
Different from the fusion between generic ops, indices are involved. In this
context, we need to re-map the indices for producer since the fused op is built
on consumer's perspective. This patch supports all combination of the fusion
between indexed_generic ops and generic ops, which includes tests case:
1) generic op as producer and indexed_generic op as consumer.
2) indexed_generic op as producer and generic op as consumer.
3) indexed_generic op as producer and indexed_generic op as consumer.
Differential Revision: https://reviews.llvm.org/D80347
Kadir Cetinkaya [Wed, 3 Jun 2020 21:50:26 +0000 (23:50 +0200)]
[clangd] Fix forgotten propagation of AsnycPreamble flag
aartbik [Wed, 3 Jun 2020 21:13:22 +0000 (14:13 -0700)]
[mlir] [VectorOps] Use 'vector.flat_transpose' for 2-D 'vector.tranpose'
Summary:
Progressive lowering of vector.transpose into an operation that
is closer to an intrinsic, and thus the hardware ISA. Currently
under the common vector transform testing flag, as we prepare
deploying this transformation in the LLVM lowering pipeline.
Reviewers: nicolasvasilache, reidtatge, andydavis1, ftynse
Reviewed By: nicolasvasilache, ftynse
Subscribers: mehdi_amini, rriddle, jpienaar, shauheen, antiagainst, nicolasvasilache, arpith-jacob, mgester, lucyrfox, liufengdb, stephenneuendorffer, Joonsoo, grosul1, frgossen, Kayjukh, jurahul, llvm-commits
Tags: #llvm, #mlir
Differential Revision: https://reviews.llvm.org/D80772
Jonas Devlieghere [Wed, 3 Jun 2020 21:53:30 +0000 (14:53 -0700)]
[lldb/Test] Python <3.5 requires **kwargs to come last
Thanks Martin Böhme for pointing this out.
Jonas Devlieghere [Wed, 3 Jun 2020 20:14:08 +0000 (13:14 -0700)]
[lldb/Test] Remove un(used|needed|maintained) files from lldbsuite.
Jim Ingham [Wed, 3 Jun 2020 21:29:22 +0000 (14:29 -0700)]
Add a test for preserving state on the non-expr thread across expression evaluation.
There may be another test that tests this but I couldn't find one.
LLVM GN Syncbot [Wed, 3 Jun 2020 21:16:03 +0000 (21:16 +0000)]
[gn build] Port
e636e6b79ac
LLVM GN Syncbot [Wed, 3 Jun 2020 21:16:02 +0000 (21:16 +0000)]
[gn build] Port
7f50c15be5c
Craig Topper [Wed, 3 Jun 2020 18:57:03 +0000 (11:57 -0700)]
[X86] Separate X86_CPU_TYPE_COMPAT_WITH_ALIAS from X86_CPU_TYPE_COMPAT. NFC
Add a separate X86_CPU_TYPE_COMPAT_ALIAS that carries alias string
and the enum from X86_CPU_TYPE_COMPAT.
Nico Weber [Wed, 3 Jun 2020 21:05:34 +0000 (17:05 -0400)]
[gn build] Fix link with ld.bfd and gold.
lld's symbol resolution algorithm makes it not depend on
the order of object files and libraries, but ld.bfd and
gold require listing dependencies later on the link line.
Put {{libs}} after {{inputs}} so that e.g. -lpthreads
appears after the object files, not before it.
Differential Revision: https://reviews.llvm.org/D81035
Thomas Lively [Wed, 3 Jun 2020 21:04:59 +0000 (14:04 -0700)]
Reland "[WebAssembly] Eliminate range checks on br_tables"
This reverts commit
755a89591528b692315ad0325347e2fd4637271b.
Although I was not able to reproduce any test failures locally,
aheejin was able to reproduce them and found a fix, applied here.
Dan Gohman [Tue, 21 Apr 2020 17:35:10 +0000 (10:35 -0700)]
[WebAssembly] Add support for -mexec-model=reactor
This adds a -mexec-model= command-line flag. The default value is "command"
which is no change from the current behavior. The other option is "reactor"
which enables the WASI Reactor ABI:
https://github.com/WebAssembly/WASI/blob/master/design/application-abi.md
Differential Revision: https://reviews.llvm.org/D62922
Christopher Tetreault [Wed, 3 Jun 2020 20:35:41 +0000 (13:35 -0700)]
[SVE] Eliminate calls to default-false VectorType::get() from IR
Reviewers: efriedma, kmclaughlin, sdesmalen, dexonsmith, dblaikie
Reviewed By: efriedma
Subscribers: tschuett, hiraditya, rkruppe, psnobl, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80261
Konrad Kleine [Wed, 3 Jun 2020 20:53:43 +0000 (16:53 -0400)]
[clang-tidy]: Added modernize-replace-disallow-copy-and-assign-macro
Summary:
This check finds macro expansions of `DISALLOW_COPY_AND_ASSIGN(Type)` and
replaces them with a deleted copy constructor and a deleted assignment operator.
Before the `delete` keyword was introduced in C++11 it was common practice to
declare a copy constructor and an assignment operator as a private members. This
effectively makes them unusable to the public API of a class.
With the advent of the `delete` keyword in C++11 we can abandon the
`private` access of the copy constructor and the assignment operator and
delete the methods entirely.
Migration example:
```
lang=dif
class Foo {
private:
- DISALLOW_COPY_AND_ASSIGN(Foo);
+ Foo(const Foo &) = delete;
+ const Foo &operator=(const Foo &) = delete;
};
```
Reviewers: alexfh, hokein, aaron.ballman, njames93
Reviewed By: njames93
Subscribers: Eugene.Zelenko, mgorny, xazax.hun, cfe-commits
Tags: #clang, #clang-tools-extra
Differential Revision: https://reviews.llvm.org/D80531
Matt Arsenault [Tue, 19 Mar 2019 17:50:22 +0000 (13:50 -0400)]
RegAllocFast: Record internal state based on register units
Record internal state based on register units. This is often more
efficient as there are typically fewer register units to update
compared to iterating over all the aliases of a register.
Original patch by Matthias Braun, but I've been rebasing and fixing it
for almost 2 years and fixed a few bugs causing intermediate failures
to make this patch independent of the changes in
https://reviews.llvm.org/D52010.
Matt Arsenault [Wed, 3 Jun 2020 20:09:06 +0000 (16:09 -0400)]
X86: Switch fastregalloc test to use generated checks
Julian Lettner [Thu, 14 May 2020 21:52:35 +0000 (14:52 -0700)]
[Darwin] Improve runtime OS version checks
Use a struct to represent numerical versions instead of encoding release
names in an enumeration. This avoids the need to extend the enumeration
every time there is a new release.
Rename `GetMacosVersion() -> GetMacosAlignedVersion()` to better reflect
how this is used on non-MacOS platforms.
Reviewed By: delcypher
Differential Revision: https://reviews.llvm.org/D79970
Lukas Sommer [Wed, 3 Jun 2020 20:32:49 +0000 (16:32 -0400)]
Replace separator in OpenMP variant name mangling.
Summary:
Nvidia PTX does not allow `.` to appear in identifiers, so OpenMP variant mangling now uses `$` to separate segments of the mangled name for variants of functions declared via `declare variant`.
Reviewers: jdoerfert, Hahnfeld
Reviewed By: jdoerfert
Subscribers: yaxunl, guansong, sstefan1, cfe-commits
Tags: #openmp, #clang
Differential Revision: https://reviews.llvm.org/D80439
Davide Italiano [Wed, 3 Jun 2020 19:21:35 +0000 (12:21 -0700)]
[IRInterpreter] Unused. Drive-by cleanup. NFCI.