Cullen Rhodes [Tue, 26 May 2020 10:50:35 +0000 (10:50 +0000)]
[AArch64][SVE] Implement structured load intrinsics
Summary:
This patch adds initial support for the following instrinsics:
* llvm.aarch64.sve.ld2
* llvm.aarch64.sve.ld3
* llvm.aarch64.sve.ld4
For loading two, three and four vectors worth of data. Basic codegen is
implemented with reg+reg and reg+imm addressing modes being addressed
in a later patch.
The types returned by these intrinsics have a number of elements that is a
multiple of the elements in a 128-bit vector for a given type and N, where N is
the number of vectors being loaded, i.e. 2, 3 or 4. Thus, for 32-bit elements
the types are:
LD2 : <vscale x 8 x i32>
LD3 : <vscale x 12 x i32>
LD4 : <vscale x 16 x i32>
This is implemented with target-specific intrinsics for each variant that take
the same operands as the IR intrinsic but return N values, where the type of
each value is a full vector, i.e. <vscale x 4 x i32> in the above example.
These values are then concatenated using the standard concat_vector intrinsic
to maintain type legality with the IR.
These intrinsics are intended for use in the Arm C Language
Extension (ACLE).
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D75751
Kazushi (Jam) Marukawa [Tue, 9 Jun 2020 08:39:03 +0000 (10:39 +0200)]
[VE] Support Transfer Control Instructions in MC layer
Summary:
Add regression tests of asmparser, mccodeemitter, and disassembler for
transfer control instructions. Add FENCEI/FENCEM/FENCEC/SVOB instructions
also. Add new instruction format to represent FENCE* instructions too.
Differential Revision: https://reviews.llvm.org/D81440
hsmahesha [Tue, 9 Jun 2020 08:38:35 +0000 (14:08 +0530)]
[AMDGPU/MemOpsCluster] Implement new heuristic for computing max mem ops cluster size
Summary:
Make use of both the - (1) clustered bytes and (2) cluster length, to decide on
the max number of mem ops that can be clustered. On an average, when loads
are dword or smaller, consider `5` as max threshold, otherwise `4`. This heuristic
is purely based on different experimentation conducted, and there is no analytical
logic here.
Reviewers: foad, rampitec, arsenm, vpykhtin
Reviewed By: foad, rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, Anastasia, t-tye, hiraditya, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81085
Kazushi (Jam) Marukawa [Tue, 9 Jun 2020 08:17:20 +0000 (10:17 +0200)]
[VE] Support lowering to NND instruction
Summary:
Support NND instruction of Aurora VE. Add hasAndNot function also
for optimization.
Differential Revision: https://reviews.llvm.org/D81383
Simon Wallis [Tue, 9 Jun 2020 07:32:46 +0000 (08:32 +0100)]
[AVR][test] Remove test for naked function containing a return.
Summary:
Remove an old test for an explicit return in a naked function from test/CodeGen/AVR/return.ll
clang no longer allows a C return in a naked function.
This test is causing failure of the patch https://reviews.llvm.org/D80715
Reviewers: llvm-commits, dylanmckay
Reviewed By: dylanmckay
Subscribers: Jim
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81099
Change-Id: Id218027e520247ae480b92e7801a660fbe0cf29b
Guillaume Chatelet [Tue, 9 Jun 2020 08:05:13 +0000 (08:05 +0000)]
[Alignment] Fix deprecation message
Sameer Sahasrabuddhe [Tue, 9 Jun 2020 07:44:15 +0000 (13:14 +0530)]
[AMDGPU] Enable structurizer workarounds by default
Reviewed By: nhaehnle
Differential Revision: https://reviews.llvm.org/D81211
Kang Zhang [Tue, 9 Jun 2020 07:39:42 +0000 (07:39 +0000)]
[MachineVerifier] Add TiedOpsRewritten flag to fix verify two-address error
Summary:
Currently, MachineVerifier will attempt to verify that tied operands
satisfy register constraints as soon as the function is no longer in
SSA form. However, PHIElimination will take the function out of SSA
form while TwoAddressInstructionPass will actually rewrite tied operands
to match the constraints. PHIElimination runs first in the pipeline.
Therefore, whenever the MachineVerifier is run after PHIElimination,
it will encounter verification errors on any tied operands.
This patch adds a function property called TiedOpsRewritten that will be
set by TwoAddressInstructionPass and will control when the verifier checks
tied operands.
Reviewed By: nemanjai
Differential Revision: https://reviews.llvm.org/D80538
Kristof Beyls [Tue, 9 Jun 2020 07:11:02 +0000 (08:11 +0100)]
[AArch64] Fix branch, terminator, etc properties for BRA* instructions.
Tests relying on some of these fixes will be added for this in follow-on
patches that introduce new features that require these properties to be
correct.
Differential Revision: https://reviews.llvm.org/D81399
David Sherwood [Tue, 2 Jun 2020 10:16:23 +0000 (11:16 +0100)]
[CodeGen] Ensure callers of CreateStackTemporary use sensible alignments
In two instances of CreateStackTemporary we are sometimes promoting
alignments beyond the stack alignment. I have introduced a new function
called getReducedAlign that will return the alignment for the broken
down parts of illegal vector types. For example, on NEON a <32 x i8>
type is made up of two <16 x i8> types - in this case the sensible
alignment is 16 bytes, not 32.
In the legalization code wherever we create stack temporaries I have
started using the reduced alignments instead for illegal vector types.
I added a test to
CodeGen/AArch64/build-one-lane.ll
that tries to insert an element into an illegal fixed vector type
that involves creating a temporary stack object.
Differential Revision: https://reviews.llvm.org/D80370
Sam Parker [Tue, 26 May 2020 13:28:34 +0000 (14:28 +0100)]
[NFCI][CostModel] Unify getCmpSelInstrCost
Add cases for icmp, fcmp and select into the switch statement of the
generic getUserCost implementation with getInstructionThroughput then
calling into it. The BasicTTI and backend implementations have be set
to return a default value (1) when a cost other than throughput is
being queried.
Differential Revision: https://reviews.llvm.org/D80550
David Sherwood [Thu, 28 May 2020 13:22:49 +0000 (14:22 +0100)]
[CodeGen,AArch64] Fix up warnings in splitStores
The code for trying to split up stores is designed for NEON vectors,
where we support arbitrary alignments. It's an optimisation designed
to improve performance by using smaller, aligned stores. However,
we currently only support 16 byte alignments for SVE vectors anyway
so we may as well bail out early.
This change fixes up remaining warnings in a couple of tests:
CodeGen/AArch64/sve-callbyref-notailcall.ll
CodeGen/AArch64/sve-calling-convention-byref.ll
Differential Revision: https://reviews.llvm.org/D80720
Stanislav Mekhanoshin [Mon, 8 Jun 2020 20:52:16 +0000 (13:52 -0700)]
[AMDGPU] Custom lowering of i64 umulo/smulo
Differential Revision: https://reviews.llvm.org/D81430
Craig Topper [Tue, 9 Jun 2020 05:23:48 +0000 (22:23 -0700)]
[X86] Autogenerate complete checks and add encoding information. NFC
I think we're using an inefficient CMP encoding for absolute
symbols that would fit in an 8-bit value. This is a pattern
priority issue in isel.
Siva Chandra Reddy [Tue, 9 Jun 2020 05:08:42 +0000 (22:08 -0700)]
[libc][NFC] Add few more missing entrypoints to the entrypoint list.
They were likely missed while rebasing.
Siva Chandra Reddy [Tue, 9 Jun 2020 00:31:04 +0000 (17:31 -0700)]
[libc][NFC] Add bzero to list of entrypoints.
A new LLVM libc specific extension standard spec has been added.
Kang Zhang [Tue, 9 Jun 2020 03:32:27 +0000 (03:32 +0000)]
[NFC][PowerPC] Remove the redundant InstAlias for OR instruction
Summary:
We have handle the InstAlias for OR instructions, but we handle it
agagin in PPCInstPrinter.cpp.
This patch is to Remove the redundant InstAlias for OR instruction.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D80502
Saleem Abdulrasool [Tue, 9 Jun 2020 02:22:02 +0000 (19:22 -0700)]
test: repair lld builder
Add the explicit `-arch x86_64` to the invocation.
Chen Zheng [Tue, 9 Jun 2020 02:43:04 +0000 (22:43 -0400)]
[APInt] set all bits for getBitsSetWithWrap if loBit == hiBit
differentiate getBitsSetWithWrap & getBitsSet when loBit == hiBit
getBitsSetWithWrap sets all bits;
getBitsSet does nothing.
Reviewed By: lkail, RKSimon, lebedev.ri
Differential Revision: https://reviews.llvm.org/D81325
Anil Mahmud [Tue, 9 Jun 2020 01:54:59 +0000 (20:54 -0500)]
[PowerPC] Fix pattern for DCBFL/DCBFLP instrinsics.
The previous implementation used "asm parser only" pseudo-instructions in their
output patterns. Those are not meant to emit code and will caused crashes when
built with -filetype=obj.
Differential Revision: https://reviews.llvm.org/D80151
Anil Mahmud [Tue, 9 Jun 2020 01:43:56 +0000 (20:43 -0500)]
[PowerPC] Remove extra instruction left by emitRLDICWhenLoweringJumpTables
The function emitRLDICWhenLoweringJumpTables in PPCMIPeephole.cpp
was supposed to convert a pair of RLDICL and RLDICR to a single RLDIC,
but it was leaving out the RLDICL instruction. This PR fixes the bug.
Differential Revision: https://reviews.llvm.org/D78063
Stefan Pintilie [Tue, 9 Jun 2020 01:20:42 +0000 (20:20 -0500)]
[PowerPC] Fix incorrect PC Relative relocations for Big Endian
Fix the incorrect PC Relative relocations for Big Endian for 34 bit offsets.
The offset should be zero for both BE and LE in this situation.
Differential Revision: https://reviews.llvm.org/D81033
Matt Arsenault [Sat, 15 Jun 2019 17:25:03 +0000 (13:25 -0400)]
Make myself code owner of InferAddressSpaces
The original author isn't active, and I've made most of the commits to
it.
Saleem Abdulrasool [Fri, 5 Jun 2020 18:18:33 +0000 (11:18 -0700)]
lld: initial pass at supporting TBD
Add support to lld to use Text Based API stubs for linking. This is
support is incomplete not filtering out platforms. It also does not
account for architecture specific API handling and potentially does not
correctly handle trees of re-exports with inlined libraries being
treated as direct children of the top level library.
Jacques Pienaar [Tue, 9 Jun 2020 01:00:35 +0000 (18:00 -0700)]
[mlir] Unranked memref type has no rank
Summary:
UnrankedMemRefType doesn't have a rank but previously this was just
checking for unranked tensor. Avoids failure later if one queries the shape
post checking if ranked.
Differential Revision: https://reviews.llvm.org/D81441
Jian Cai [Tue, 9 Jun 2020 00:56:46 +0000 (17:56 -0700)]
[AArch64] Support expression results as immediate values in mov
Summary:
This patch adds support of using the result of an expression as an
immediate value. For example,
0:
.skip 4
1:
mov x0, 1b - 0b
is assembled to
mov x0, #4
Currently it does not support expressions requiring relocation unless
explicitly specified. This fixes PR#45781.
Reviewers: peter.smith, ostannard, efriedma
Reviewed By: efriedma
Subscribers: nickdesaulniers, llozano, manojgupta, efriedma, ostannard, kristof.beyls, hiraditya, danielkiss, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80028
dfukalov [Sat, 6 Jun 2020 00:47:05 +0000 (03:47 +0300)]
[AMDGPU][NFC] Skip processing intrinsics that do not become real instructions
Reviewers: rampitec
Reviewed By: rampitec
Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81260
Siva Chandra Reddy [Tue, 9 Jun 2020 00:09:28 +0000 (17:09 -0700)]
[libc][NFC][Obvious] Add names to various string entrypoints.
Eli Friedman [Mon, 8 Jun 2020 23:35:10 +0000 (16:35 -0700)]
Revert "[arm][darwin] Don't generate libcalls for wide shifts on Darwin"
This reverts commit
2ba016cd5ce50a3683d3e6c2c62f00e1cccfd8b5.
This is causing a failure on the clang-cmake-armv7-full bot, and there
are outstanding review comments.
Petr Hosek [Mon, 8 Jun 2020 23:34:57 +0000 (16:34 -0700)]
[InstrProfiling] Use -fuse-ld=lld in instrprof-gc-sections test
This was accidentally omitted in 603d58b leading to a test failure
on some of the bots that don't use lld as the default linker.
George Mitenkov [Mon, 8 Jun 2020 22:20:52 +0000 (18:20 -0400)]
[MLIR][SPIRVToLLVM] Add skeleton for SPIR-V to LLVM dialect conversion
These commits set up the skeleton for SPIR-V to LLVM dialect conversion.
I created SPIR-V to LLVM pass, registered it in Passes.td, InitAllPasses.h.
Added a pattern for `spv.BitwiseAndOp` and tests for it. Integer, float
and vector types are converted through LLVMTypeConverter.
Differential Revision: https://reviews.llvm.org/D81100
KareemErgawy [Mon, 8 Jun 2020 22:18:06 +0000 (18:18 -0400)]
[MLIR][Toy] Fix a few typos in the comments/docs.
Fixes a few typos and errors in MLIR's Toy tutorial docs and comments.
Differential Revision: https://reviews.llvm.org/D81406
Petr Hosek [Sat, 13 Jul 2019 21:02:07 +0000 (14:02 -0700)]
[InstrProfiling] Use !associated metadata for counters, data and values
The !associated metadata may be attached to a global object declaration
with a single argument that references another global object. This
metadata prevents discarding of the global object in linker GC unless
the referenced object is also discarded.
Furthermore, when a function symbol is discarded by the linker, setting
up !associated metadata allows linker to discard counters, data and
values associated with that function symbol. This is not possible today
because there's metadata to guide the linker. This approach is also used
by other instrumentations like sanitizers.
Note that !associated metadata is only supported by ELF, it does not have
any effect on non-ELF targets.
Differential Revision: https://reviews.llvm.org/D76802
Petr Hosek [Mon, 8 Jun 2020 21:37:27 +0000 (14:37 -0700)]
Revert "[InstrProfiling] Use !associated metadata for counters, data and values"
This reverts commit
39b3c41b65302a969fa5507402976a255a07c158 due to
a failing associated.ll test.
Stanislav Mekhanoshin [Mon, 8 Jun 2020 15:27:07 +0000 (08:27 -0700)]
Stabilize alloca slices sort in SROA
Slice::operator<() has a non-deterministic behavior. If we have
identical slices comparison will depend on the order or operands.
Normally that does not result in unstable compilation results
because the order in which slices are inserted into the vector
is deterministic and llvm::sort() normally behaves as a stable
sort, although that is not guaranteed.
However, there is test option -sroa-random-shuffle-slices which
is used to check exactly this aspect. The vector is first randomly
shuffled and then sorted. The same shuffling happens without this
option under expensive llvm checks.
I have managed to write a test which has hit this problem.
There are no fields in the Slice class to resolve the instability.
We only have offsets, IsSplittable and Use, but neither Use nor
User have anything suitable for predictable comparison.
I have switched to stable_sort which has to be sufficient and
removed that randon shuffle option.
Differential Revision: https://reviews.llvm.org/D81310
Richard Smith [Mon, 8 Jun 2020 21:04:54 +0000 (14:04 -0700)]
Remove improper uses of DiagnosticErrorTrap and hasErrorOccurred.
DiagnosticErrorTrap is usually inappropriate because it indicates
whether an error message was rendered in a given region (and is
therefore affected by -ferror-limit and by suppression of errors if we
see an invalid declaration).
hasErrorOccurred() is usually inappropriate because it indicates
whethere an "error:" message was displayed, regardless of whether the
message was a warning promoted to an error, and therefore depends on
things like -Werror that are usually irrelevant.
Where applicable, CodeSynthesisContexts are used to attach notes to
the first diagnostic produced in a region of code, isnstead of using an
error trap and then attaching a note to whichever diagnostic happened to
be produced last (or suppressing the note if the final diagnostic is a
disabled warning!).
This is mostly NFC.
Sterling Augustine [Mon, 8 Jun 2020 21:11:14 +0000 (14:11 -0700)]
Add #includes so that ROCm.h is compilable stand-alone.
Summary:
ROCm.h had been getting the declarations for various data structures
by being #included next to them, rather than #includeing them itself.
This change fixes that by explicitly including the appropriate headers.
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81432
Yonghong Song [Sat, 6 Jun 2020 01:37:38 +0000 (18:37 -0700)]
[DebugInfo] Fix assertion for extern void type
Commit
d77ae1552fc2 ("[DebugInfo] Support to emit debugInfo
for extern variables") added support to emit debuginfo
for extern variables. Currently, only BPF target enables to
emit debuginfo for extern variables.
But if the extern variable has "void" type, the compilation will
fail.
-bash-4.4$ cat t.c
extern void bla;
void *test() {
void *x = &bla;
return x;
}
-bash-4.4$ clang -target bpf -g -O2 -S t.c
missing global variable type
!1 = distinct !DIGlobalVariable(name: "bla", scope: !2, file: !3, line: 1,
isLocal: false, isDefinition: false)
...
fatal error: error in backend: Broken module found, compilation aborted!
PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace,
preprocessed source, and associated run script.
Stack dump:
...
The IR requires a DIGlobalVariable must have a valid type and the
"void" type does not generate any type, hence the above fatal error.
Note that if the extern variable is defined as "const void", the
compilation will succeed.
-bash-4.4$ cat t.c
extern const void bla;
const void *test() {
const void *x = &bla;
return x;
}
-bash-4.4$ clang -target bpf -g -O2 -S t.c
-bash-4.4$ cat t.ll
...
!1 = distinct !DIGlobalVariable(name: "bla", scope: !2, file: !3, line: 1,
type: !6, isLocal: false, isDefinition: false)
!6 = !DIDerivedType(tag: DW_TAG_const_type, baseType: null)
...
Since currently, "const void extern_var" is supported by the
debug info, it is natural that "void extern_var" should also
be supported. This patch disabled assertion of "void extern_var"
in IR verifier and add proper guarding when emiting potential
null debug info type to dwarf types.
Differential Revision: https://reviews.llvm.org/D81131
Petr Hosek [Sat, 13 Jul 2019 21:02:07 +0000 (14:02 -0700)]
[InstrProfiling] Use !associated metadata for counters, data and values
The !associated metadata may be attached to a global object declaration
with a single argument that references another global object. This
metadata prevents discarding of the global object in linker GC unless
the referenced object is also discarded.
Furthermore, when a function symbol is discarded by the linker, setting
up !associated metadata allows linker to discard counters, data and
values associated with that function symbol. This is not possible today
because there's metadata to guide the linker. This approach is also used
by other instrumentations like sanitizers.
Note that !associated metadata is only supported by ELF, it does not have
any effect on non-ELF targets.
Differential Revision: https://reviews.llvm.org/D76802
Roman Lebedev [Mon, 8 Jun 2020 20:26:54 +0000 (23:26 +0300)]
[Support] FoldingSetNodeID::AddString(): reserve memory
Summary:
It is traditionally potentially very inefficient to not preallocate the memory,
but rely on reallocation every time you push something into vector.
For example, looking at unity build of RawSpeed
(`-O3 -g0 -emit-llvm -Xclang -disable-llvm-optzns`),
the memory story is as follows:
```
total runtime: 11.34s.
calls to allocation functions: 2694053 (237612/s)
temporary memory allocations: 645188 (56904/s)
peak heap memory consumption: 231.36MB
peak RSS (including heaptrack overhead): 397.39MB
```
Looking at details, `FoldingSetNodeID::AddString()` is noteworthy, frequently called and is allocation-heavy.
But it is quite obvious how many times we will push into `Bits` - we will push `String.size()` itself,
and then we will push once per every 4 bytes of `String` (padding last block).
And if we preallocate, we get:
```
total runtime: 11.20s.
calls to allocation functions: 2594704 (231669/s)
temporary memory allocations: 560004 (50000/s)
peak heap memory consumption: 231.36MB
peak RSS (including heaptrack overhead): 398.06MB
```
Which is a measurable win:
```
total runtime: -0.14s. # -1.23 %
calls to allocation functions: -99349 (719920/s) # -3.69 %
temporary memory allocations: -85184 (617275/s) # -13.2 % (!)
peak heap memory consumption: 0B
peak RSS (including heaptrack overhead): 0B
total memory leaked: 0B
```
Reviewers: efriedma, nikic, bkramer
Reviewed By: bkramer
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81342
Michael Liao [Mon, 8 Jun 2020 20:12:58 +0000 (16:12 -0400)]
[lld] Fix shared library build by adding the missing dependency.
Richard Smith [Mon, 8 Jun 2020 18:52:14 +0000 (11:52 -0700)]
More robust fix for crash on invalid range-based for statement.
Reliably mark the loop variable declaration in a range for as having an
invalid initializer if anything goes wrong building the initializer. We
previously based this determination on whether an error was emitted,
which is not a reliable signal due to error suppression (during error
recovery etc).
Also, properly mark the variable as having initializer errors rather
than simply marking it invalid. This is necessary to mark any structured
bindings as invalid too.
This generalizes the previous fix in
936ec89e91e2dda8b6110b1fd1f9920509d7a17b.
jasonliu [Mon, 8 Jun 2020 18:51:33 +0000 (18:51 +0000)]
[XCOFF][AIX] report_fatal_error when an overflow section is needed
If there are more than 65534 relocation entries in a single section,
we should generate an overflow section.
Since we don't support overflow section for now, we should generate
an error.
Differential revision: https://reviews.llvm.org/D81104
Valentin Clement [Mon, 8 Jun 2020 18:38:18 +0000 (14:38 -0400)]
Apply fix from D81179 only from GCC < 8
Summary: Apply workaround done in D81179 only for GCC < 8. As @klausler mentioned in D81179 we want to avoid additional checks for other compilers that do not need them.
Reviewers: DavidTruby, klausler, jdoerfert, sscalpone
Reviewed By: klausler, sscalpone
Subscribers: llvm-commits, tskeith, isuruf, klausler
Tags: #flang, #llvm
Differential Revision: https://reviews.llvm.org/D81208
Arthur Eubanks [Mon, 8 Jun 2020 19:48:03 +0000 (12:48 -0700)]
Revert "Reland D80966 [codeview] Put !heapallocsite on calls to operator new"
This reverts commit
b6e143aa5448bbe29da7b045072e66a31813bced.
Causes https://bugs.chromium.org/p/chromium/issues/detail?id=1092370#c5.
Will investigate and reland (again).
LLVM GN Syncbot [Mon, 8 Jun 2020 19:44:56 +0000 (19:44 +0000)]
[gn build] Port
bb677cacc80
Andrew Litteken [Wed, 8 Apr 2020 22:21:49 +0000 (15:21 -0700)]
[SuffixTree][MachOpt] Factoring out Suffix Tree and adding Unit Tests
This moves the SuffixTree test used in the Machine Outliner and moves it into Support for use in other outliners elsewhere in the compilation pipeline.
Differential Revision: https://reviews.llvm.org/D80586
Julian Lettner [Mon, 8 Jun 2020 19:01:14 +0000 (12:01 -0700)]
[TSan] Revert removal of ignore_interceptors_accesses flag
This flag suppresses TSan FPs on Darwin. I removed this flag
prematurely and have been dealing with the fallout ever since.
This commit puts back the flag, reverting
7d1085cb [1].
[1] https://reviews.llvm.org/D55075
Jian Cai [Mon, 8 Jun 2020 18:53:23 +0000 (11:53 -0700)]
Add a flag to debug automatic variable initialization
Summary:
Add -ftrivial-auto-var-init-stop-after= to limit the number of times
stack variables are initialized when -ftrivial-auto-var-init= is used to
initialize stack variables to zero or a pattern. This flag can be used
to bisect uninitialized uses of a stack variable exposed by automatic
variable initialization, such as http://crrev.com/c/2020401.
Reviewers: jfb, vitalybuka, kcc, glider, rsmith, rjmccall, pcc, eugenis, vlad.tsyrklevich
Reviewed By: jfb
Subscribers: phosek, hubert.reinterpretcast, srhines, MaskRay, george.burgess.iv, dexonsmith, inglorion, gbiv, llozano, manojgupta, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D77168
Nico Weber [Mon, 8 Jun 2020 19:20:16 +0000 (15:20 -0400)]
Revert "[clangd] Resolve driver symlinks, and look up unknown relative drivers in PATH."
This reverts commit
806342b8ef54ec07511d0ce5d3d1335451e952da.
Breaks check-clangd on macOS, https://reviews.llvm.org/D75414#2080076
Florian Hahn [Mon, 8 Jun 2020 18:25:14 +0000 (19:25 +0100)]
[AArch64] Fix ldst-opt of multiple disjunct subregs.
Currently aarch64-ldst-opt will incorrectly rename registers with
multiple disjunct subregisters (e.g. result of LD3). This patch updates
the canRenameUpToDef to bail out if it encounters such a register class
that contains the register to rename.
Fixes PR46105.
Reviewers: efriedma, dmgreen, paquette, t.p.northover
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D81108
Isuru Fernando [Mon, 8 Jun 2020 19:13:58 +0000 (14:13 -0500)]
[flang] Fix issue of flang/runtime/config.h not being found in out of tree builds
Summary: Fixes https://bugs.llvm.org/show_bug.cgi?id=46078
Reviewers: DavidTruby, jdoerfert, PeteSteinfeld, sscalpone, tskeith
Reviewed By: PeteSteinfeld, sscalpone, tskeith
Subscribers: mgorny, llvm-commits, tskeith
Tags: #llvm, #flang
Differential Revision: https://reviews.llvm.org/D81266
Hans Wennborg [Mon, 8 Jun 2020 16:36:56 +0000 (18:36 +0200)]
[PGO] CallPromotion: Don't try to pass sret args to varargs functions
It's not allowed by the verifier.
Differential revision: https://reviews.llvm.org/D81409
Arthur Eubanks [Thu, 28 May 2020 22:00:17 +0000 (15:00 -0700)]
Move *San module passes later in the NPM pipeline
Summary:
This fixes pr33372.cpp under the new pass manager.
ASan adds padding to globals. For example, it will change a {i32, i32, i32} to a {{i32, i32, i32}, [52 x i8]}. However, when loading from the {i32, i32, i32}, InstCombine may (after various optimizations) end up loading 16 bytes instead of 12, likely because it thinks the [52 x i8] padding is ok to load from. But ASan checks that padding should not be loaded from.
Ultimately this is an issue of *San passes wanting to be run after all optimizations. This change moves the module passes right next to the corresponding function passes.
Also remove comment that's no longer relevant, this is the last ASan/MSan/TSan failure under the NPM (hopefully...).
As mentioned in https://reviews.llvm.org/rG1285e8bcac2c54ddd924ffb813b2b187467ac2a6, NPM doesn't support LTO + sanitizers, so modified some tests that test for that.
Reviewers: leonardchan, vitalybuka
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81323
Louis Dionne [Mon, 8 Jun 2020 18:58:46 +0000 (14:58 -0400)]
[libc++] Improve tests for iterators.operations
Reduce duplication between the constexpr and the non-constexpr test cases,
and add tests for the return type of the various operations.
Craig Topper [Mon, 8 Jun 2020 18:26:11 +0000 (11:26 -0700)]
[X86] Prevent LowerSELECT from causing suboptimal codegen for __builtin_ffs(X) - 1.
LowerSELECT sees the CMP with 0 and wants to use a trick with SUB
and SBB. But we can use the flags from the BSF/TZCNT.
Fixes PR46203.
Differential Revision: https://reviews.llvm.org/D81312
Hendrik Greving [Fri, 5 Jun 2020 18:34:34 +0000 (11:34 -0700)]
[ModuloSchedule] Support instructions with > 1 destination when walking canonical use.
Fixes a minor bug that led to finding the wrong register if the definition had more
than one register destination.
Sanjay Patel [Mon, 8 Jun 2020 18:09:04 +0000 (14:09 -0400)]
[InstCombine] improve matching for sext-lshr-trunc patterns, part 2
Similar to rG42f488b63a04
This is intended to preserve the logic of the existing transform,
but remove unnecessary restrictions on uses and types.
https://rise4fun.com/Alive/oS0
Name: narrow input
Pre: C1 <= width(C1) - 24
%B = sext i8 %A
%C = lshr %B, C1
%r = trunc %C to i24
=>
%s = ashr i8 %A, trunc(umin(C1, 7))
%r = sext i8 %s to i24
Name: wide input
Pre: C1 <= width(C1) - 24
%B = sext i24 %A
%C = lshr %B, C1
%r = trunc %C to i8
=>
%s = ashr i24 %A, trunc(umin(C1, 23))
%r = trunc i24 %s to i8
Sanjay Patel [Mon, 8 Jun 2020 17:08:40 +0000 (13:08 -0400)]
[InstCombine] add tests for sext+lshr+trunc; NFC
Shows missing transforms with extra uses and vectors.
Christopher Tetreault [Mon, 8 Jun 2020 18:07:25 +0000 (11:07 -0700)]
[SVE] Eliminate calls to default-false VectorType::get() from llvm-stress
Reviewers: efriedma, kmclaughlin, sdesmalen, MaskRay, JDevlieghere
Reviewed By: sdesmalen
Subscribers: tschuett, rkruppe, psnobl, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81201
Chris Jackson [Mon, 8 Jun 2020 17:44:11 +0000 (18:44 +0100)]
[DebugInfo] Reduce SalvageDebugInfo() functions
- Now all SalvageDebugInfo() calls will mark undef if the salvage
attempt fails.
Reviewed by: vsk, Orlando
Differential Revision: https://reviews.llvm.org/D78369
Yaxun (Sam) Liu [Mon, 8 Jun 2020 17:42:52 +0000 (13:42 -0400)]
Rename arg name in __clang_hip_math.h
__sptr is a keyword for ms-extension. Change it from __sptr to __sinptr.
Saleem Abdulrasool [Sat, 6 Jun 2020 02:54:58 +0000 (19:54 -0700)]
lld: improve the `-arch` handling for MachO
Use the default target triple configured by the user to determine the
default architecture for `ld64.lld`. Stash the architecture in the
configuration as when linking against TBDs, we will need to filter out
the symbols based upon the architecture. Treat the Haswell slice as it
is equivalent to `x86_64` but with the extra Haswell extensions (e.g.
AVX2, FMA3, BMI1, etc). This will make it easier to add new
architectures in the future.
This change also changes the failure mode where an invalid `-arch`
parameter will result in the linker exiting without further processing.
Florian Hahn [Mon, 8 Jun 2020 17:29:17 +0000 (18:29 +0100)]
[AArch64] Add a ldst-opt test with undef operands (NFC).
This patch adds a test to check that we do not use an undef renamable
register for renaming the other operand in a LDP instruction, as
suggested in D81108.
Jan-Willem Maessen [Mon, 8 Jun 2020 17:38:16 +0000 (18:38 +0100)]
[NFC] Fix quadratic LexicalScopes::constructScopeNest
We sometimes have functions with large numbers of sibling basic
blocks (usually with an error path exit from each one). This was
triggering the qudratic behavior in this function - after visiting
each child llvm would re-scan the parent from the beginning again. We
modify the work stack to record the next index to be worked on
alongside the pointer. This avoids the need to linearly search for
the next unfinished child.
Differential Revision: https://reviews.llvm.org/D80029
Christopher Tetreault [Mon, 8 Jun 2020 17:12:08 +0000 (10:12 -0700)]
[SVE] Eliminate calls to default-false VectorType::get() from CodeGen
Reviewers: efriedma, c-rhodes, david-arm, spatel, craig.topper, aqjune, paquette, arsenm, gchatelet
Reviewed By: spatel, gchatelet
Subscribers: wdng, tschuett, hiraditya, rkruppe, psnobl, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80313
Hiroshi Yamauchi [Tue, 2 Jun 2020 17:50:59 +0000 (10:50 -0700)]
[PGO][PGSO] Enable non-cold code size opts under non-partial-profile sample PGO.
Summary: Following up D78949.
Reviewers: davidxl
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81020
LLVM GN Syncbot [Mon, 8 Jun 2020 16:55:22 +0000 (16:55 +0000)]
[gn build] Port
550b5995233
zoecarver [Mon, 8 Jun 2020 16:47:48 +0000 (09:47 -0700)]
[NFC] [libcxx] Remove shared_ptr's no-rvalue unique_ptr converting constructor.
All compilers supported by libc++ have rvalues in C++03 mode so, there is no need for this non-rvalue overload.
Differential Revision: https://reviews.llvm.org/D80881
Jonas Devlieghere [Mon, 8 Jun 2020 16:46:34 +0000 (09:46 -0700)]
[Support] Replace 'DisableColors' boolean with 'ColorMode' enum
Replace the DisableColors with a ColorMode which can be set to Auto,
Enabled and Disabled. The purpose of this change is to make it possible
to ignore the command line option not only for disabling colors, but
also for enabling them.
Differential revision: https://reviews.llvm.org/D81056
Matt Arsenault [Mon, 8 Jun 2020 16:46:55 +0000 (12:46 -0400)]
AMDGPU/GlobalISel: Precommit regenerated check lines
The update_*test_checks scripts miss new stuff added at the end of
lines. Regenerate checks so the new mode register operands don't show
up in the diff of a future patch.
David Goldman [Fri, 5 Jun 2020 14:57:33 +0000 (10:57 -0400)]
[Sema][CodeComplete][ObjC] Don't include arrow/dot fixits
Summary:
Exempt ObjC from arrow/dot fixits since this has limited value for
Objective-C, where properties (referenced by dot syntax) are normally
backed by ivars (referenced by arrow syntax).
In addition, the current implementation doesn't properly mark
the fix it condition for Objective-C.
This was initially added in https://reviews.llvm.org/D41537
for C++ and then later C, don't believe the Objective-C changes
were intentional.
Reviewers: sammccall, yvvan
Subscribers: jfb, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81263
Kadir Cetinkaya [Mon, 8 Jun 2020 16:17:42 +0000 (18:17 +0200)]
[clangd] Drop unnecessary FS modifications in FindSymbolsTests
Fangrui Song [Mon, 8 Jun 2020 16:32:30 +0000 (09:32 -0700)]
Reland D80979 [clang] Implement VectorType logic not operator
With a fix to use -triple %itanium_abi_triple
Differential Revision: https://reviews.llvm.org/D80979
Ilya Biryukov [Mon, 8 Jun 2020 16:20:10 +0000 (19:20 +0300)]
Updates to the 'CLion Integration' section in ClangFormat docs
This commit updates the 'CLion Integration' section in ClangFormat docs.
Key changes:
- clang-format is enabled automatically when there is a config file;
- formatting now works for indentations;
- if clang-format is enabled without a config file, CLion suggests creating it based on the IDE settings or uses the LLVM style by default.
Patch by Marina Kalashina!
Reviewers: sylvestre.ledru, ilya-biryukov
Reviewed By: ilya-biryukov
Subscribers: ilya-biryukov, klimek, MyDeveloperDay, sammccall, gribozavr2
Differential Revision: https://reviews.llvm.org/D80721
Jonas Devlieghere [Mon, 8 Jun 2020 15:47:55 +0000 (08:47 -0700)]
[Support] Move color handling from raw_fd_ostream to raw_ostream
Move the color handling code from raw_fd_ostream to raw_ostream. This
makes it possible to use colors with any ostream when enabled. The
existing behavior where only raw_fd_ostream supports colors by default
remains unchanged.
Differential revision: https://reviews.llvm.org/D81110
Sanjay Patel [Mon, 8 Jun 2020 15:37:43 +0000 (11:37 -0400)]
[InstCombine] improve matching for sext-lshr-trunc patterns
This is intended to preserve the logic of the existing transform,
but remove unnecessary restrictions on uses and types.
https://rise4fun.com/Alive/pYfR
Pre: C1 <= width(C1) - 8
%B = sext i8 %A
%C = lshr %B, C1
%r = trunc %C to i8
=>
%r = ashr i8 %A, trunc(umin(C1, 7))
David Goldman [Fri, 5 Jun 2020 18:00:13 +0000 (14:00 -0400)]
[ObjC] Fix AST serialization for pseudo-strong parameters
This bit was assumed to be always false for ParmVarDecls, but attribute
objc_externally_retained now can produce it.
Differential revision: https://reviews.llvm.org/D74417
Alexander Belyaev [Mon, 8 Jun 2020 15:48:01 +0000 (17:48 +0200)]
[mlir] Lower Shape binary ops (AddOp, MulOp) to Standard.
Differential Revision: https://reviews.llvm.org/D81344
Haojian Wu [Mon, 8 Jun 2020 12:52:20 +0000 (14:52 +0200)]
[AST] Fix a clang crash on an invalid for-range statement.
Summary:
crash stack:
```
llvm-project/clang/lib/AST/ASTContext.cpp:2248: clang::TypeInfo clang::ASTContext::getTypeInfoImpl(const clang::Type *) const: Assertion `!A->getDeducedType().isNull() && "cannot request the size of an undeduced or dependent auto type"' failed.
PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace, preprocessed source, and associated run script.
Stack dump:
#0 0x00000000025bb0bf llvm::sys::PrintStackTrace(llvm::raw_ostream&) llvm-project/llvm/lib/Support/Unix/Signals.inc:564:13
#1 0x00000000025b92b0 llvm::sys::RunSignalHandlers() llvm-project/llvm/lib/Support/Signals.cpp:69:18
#2 0x00000000025bb535 SignalHandler(int) llvm-project/llvm/lib/Support/Unix/Signals.inc:396:3
#3 0x00007f9ef9298110 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14110)
#4 0x00007f9ef8d72761 raise /build/glibc-M65Gwz/glibc-2.30/signal/../sysdeps/unix/sysv/linux/raise.c:51:1
#5 0x00007f9ef8d5c55b abort /build/glibc-M65Gwz/glibc-2.30/stdlib/abort.c:81:7
#6 0x00007f9ef8d5c42f get_sysdep_segment_value /build/glibc-M65Gwz/glibc-2.30/intl/loadmsgcat.c:509:8
#7 0x00007f9ef8d5c42f _nl_load_domain /build/glibc-M65Gwz/glibc-2.30/intl/loadmsgcat.c:970:34
#8 0x00007f9ef8d6b092 (/lib/x86_64-linux-gnu/libc.so.6+0x34092)
#9 0x000000000458abe0 clang::ASTContext::getTypeInfoImpl(clang::Type const*) const llvm-project/clang/lib/AST/ASTContext.cpp:0:5
```
Reviewers: sammccall
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81384
Sam McCall [Thu, 28 May 2020 12:24:30 +0000 (14:24 +0200)]
[clangd] Resolve driver symlinks, and look up unknown relative drivers in PATH.
Summary:
This fixes a reported bug: if clang and libc++ are installed under
/usr/lib/llvm-11/... but there'- a symlink /usr/bin/clang++-11, then a
compile_commands.json with "/usr/bin/clang++-11 -stdlib=libc++" would previously
look for libc++ under /usr/include instead of /usr/lib/llvm-11/include.
The PATH change makes this work if the compiler is just "clang++-11" too.
As this is now doing IO potentially on every getCompileCommand(), we cache
the results for each distinct driver.
While here:
- Added a Memoize helper for this as multithreaded caching is a bit noisy.
- Used this helper to simplify QueryDriverDatabase and reduce blocking there.
(This makes use of the fact that llvm::Regex is now threadsafe)
Reviewers: kadircet
Subscribers: jyknight, ormris, ilya-biryukov, MaskRay, jkorous, arphaman, jfb, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D75414
Sanjay Patel [Mon, 8 Jun 2020 15:15:13 +0000 (11:15 -0400)]
[InstCombine] reduce code duplication in visitTrunc(); NFC
Sanjay Patel [Mon, 8 Jun 2020 13:33:00 +0000 (09:33 -0400)]
[InstCombine] add tests for sext+lshr+trunc; NFC
Shows missing transforms with extra uses and vectors.
Shengchen Kan [Mon, 8 Jun 2020 13:43:24 +0000 (21:43 +0800)]
[TEST] TreeTest.cpp - Add a comma to avoid build error with -werror
Summary:
The macro `INSTANTIATE_TEST_CASE_P` is defined as
```
\# define INSTANTIATE_TEST_CASE_P(prefix, test_case_name, generator, ...) \
...
```
If we build the test case with -werror, we will get an error like
```
error: ISO C++11 requires at least one argument for the "..." in a
variadic macro
testing::ValuesIn(TestClangConfig::allConfigs()));
^
```
This patch fixes that.
Reviewers: gribozavr, hlopko, eduucaldas, gribozavr2
Reviewed By: gribozavr2
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81388
Aaron Puchert [Mon, 8 Jun 2020 14:30:06 +0000 (16:30 +0200)]
Thread safety analysis: Add note for double unlock
Summary:
When getting a warning that we release a capability that isn't held it's
sometimes not clear why. So just like we do for double locking, we add a
note on the previous release operation, which marks the point since when
the capability isn't held any longer.
We can find this previous release operation by looking up the
corresponding negative capability.
Reviewers: aaron.ballman, delesley
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D81352
Aaron Puchert [Mon, 8 Jun 2020 14:29:53 +0000 (16:29 +0200)]
Thread safety analysis: Support deferring locks
Summary:
The standard std::unique_lock can be constructed to manage a lock without
initially acquiring it by passing std::defer_lock as second parameter.
It can be acquired later by calling lock().
To support this, we use the locks_excluded attribute. This might seem
like an odd choice at first, but its consistent with the other
annotations we support on scoped capability constructors. By excluding
the lock we state that it is currently not in use and the function
doesn't change that, which is exactly what the constructor does.
Along the way we slightly simplify handling of scoped capabilities.
Reviewers: aaron.ballman, delesley
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D81332
Aaron Puchert [Mon, 8 Jun 2020 14:58:12 +0000 (16:58 +0200)]
Fix build after removing llvm/CodeGen/GlobalISel/Types.h
Removing the include was probably just forgotten after
f13ba22227ea221af55bdad43c94e9cc43ef0926.
Guillaume Chatelet [Mon, 8 Jun 2020 07:25:57 +0000 (07:25 +0000)]
[Alignment][NFC] Deprecate dead code from CallingConvLower.h
Summary: This is a followup on D81196.
Reviewers: courbet
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81362
Dmitry Vyukov [Mon, 8 Jun 2020 14:45:39 +0000 (16:45 +0200)]
tsan: add OpenBSD support for Go
With the race_debug_openbsd_amd64.syso file created via this diff,
Go's race detector is able to detect a race in the example code
link: https://golang.org/doc/articles/race_detector.html
Reviewed-in: https://reviews.llvm.org/D80469
Author: qbit (Aaron Bieber)
Xing GUO [Mon, 8 Jun 2020 14:48:32 +0000 (22:48 +0800)]
[ObjectYAML] Add support for error handling in DWARFYAML. NFC.
This patch intends to be an NFC-patch. Test cases will be added in a follow-up patch.
Reviewed By: jhenderson, grimar
Differential Revision: https://reviews.llvm.org/D81356
Isuru Fernando [Mon, 8 Jun 2020 14:08:35 +0000 (15:08 +0100)]
[flang] Use LLVM's flags
Summary:
The only difference is that LLVM_ENABLE_WERROR is set to OFF
by default, but we enable this in a standalone flang build
This commit fixes some windows issues with the flags
Reviewers: DavidTruby, jdoerfert, sscalpone
Reviewed By: DavidTruby, sscalpone
Subscribers: ormris, richard.barton.arm, mehdi_amini, Meinersbur, ChinouneMehdi, tskeith, mgorny, llvm-commits
Tags: #llvm, #flang
Differential Revision: https://reviews.llvm.org/D78306
Matt Arsenault [Mon, 8 Jun 2020 14:25:57 +0000 (10:25 -0400)]
GlobalISel: Remove dead include
Sam Parker [Mon, 8 Jun 2020 14:25:03 +0000 (15:25 +0100)]
[CostModel] Follow-up to buildbot fix
Adding type checks into the other backends that call
getTypeLegalizationCost.
Differential Revision: https://reviews.llvm.org/D80984
Matt Arsenault [Mon, 8 Jun 2020 14:18:33 +0000 (10:18 -0400)]
GlobalISel: Remove deprecated methods
These have very few users to begin with, and have been marked
deprecated for 2 months which should be long enough for out of tree
targets.
David Sherwood [Thu, 4 Jun 2020 14:53:06 +0000 (15:53 +0100)]
[CodeGen] Fix nullptr crash in tryConvertSVEWideCompare
When the input to a wide compare instruction is a DUP or SPLAT_VECTOR
node we should deal with cases where the DUP/SPLAT_VECTOR input
operand is not an immediate value. I've fixed the code to return
SDValue() in such cases and added a couple of tests - one each to
represent the signed and unsigned cases.
Differential Revision: https://reviews.llvm.org/D81167
Matt Arsenault [Mon, 8 Jun 2020 14:04:19 +0000 (10:04 -0400)]
GlobalISel: Use Register
Matt Arsenault [Mon, 8 Jun 2020 14:05:56 +0000 (10:05 -0400)]
GlobalISel: Remove unused header
Matt Arsenault [Sat, 6 Jun 2020 18:19:45 +0000 (14:19 -0400)]
GlobalISel: Add dump method to LLT
Matt Arsenault [Mon, 8 Jun 2020 13:49:34 +0000 (09:49 -0400)]
GlobalISel: Make it clearer that regbank/class are mutually exclusive
Matt Arsenault [Mon, 8 Jun 2020 01:56:42 +0000 (21:56 -0400)]
GlobalISel: Simplify debug printing