Florian Hahn [Thu, 26 May 2022 09:35:38 +0000 (10:35 +0100)]
[AArch64] Add additional tests for sinking free shuffles for FMAs.
David Spickett [Thu, 26 May 2022 09:27:23 +0000 (09:27 +0000)]
[llvm][DWARF] Move test using X86 triple into X86 tests
Fixes failure seen when building without X86 backend:
https://lab.llvm.org/buildbot/#/builders/171/builds/15124
David Green [Thu, 26 May 2022 09:16:21 +0000 (10:16 +0100)]
[AArch64] Costmodel tests for llvm.vscale intrinsics. NFC
These shows that the cost of a @llvm.vscale is indeed 1, not 10.
Fraser Cormack [Thu, 26 May 2022 09:00:33 +0000 (10:00 +0100)]
[RISCV][NFC] Add braces to 'else' to match braced 'if'
LLVM GN Syncbot [Thu, 26 May 2022 08:51:13 +0000 (08:51 +0000)]
[gn build] Port
569d6630204d
Nikolas Klauser [Wed, 25 May 2022 09:09:43 +0000 (11:09 +0200)]
[libc++] Implement ranges::equal
Reviewed By: var-const, #libc
Spies: libcxx-commits, mgorny
Differential Revision: https://reviews.llvm.org/D123681
Florian Hahn [Thu, 26 May 2022 08:46:23 +0000 (09:46 +0100)]
[SimpleLoopUnswitch] Always skip trivial select and set condition.
When updating the branch instruction outside the loopduring non-trivial
unswitching, always skip trivial selects and update the condition.
Otherwise we might create invalid IR, because the trivial select is
inside the loop, while the condition is outside the loop.
Fixes #55697.
Kito Cheng [Fri, 20 May 2022 02:47:15 +0000 (10:47 +0800)]
[RISCV] Fix state persistence bugs (PR55548)
We didn't implement RISCVELFStreamer::reset and cause some very strange
section output for attribute section...just reference D15950 to see how
ARM implement that.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D125905
Florian Hahn [Thu, 26 May 2022 07:53:33 +0000 (08:53 +0100)]
[LV] Fix indentation in tryToCreateWidenRecipe (NFC).
Jason Molenda [Thu, 26 May 2022 07:46:54 +0000 (00:46 -0700)]
Revert "Defer source path remap tilde expansion until source file use"
This reverts commit
c274b6e5830ea88d3f55d6dc1d2b99e38cf6595e.
The x86_64 debian bot got a failure with this patch,
https://lab.llvm.org/buildbot#builders/68/builds/33078
where
SymbolFile/DWARF/x86/DW_TAG_variable-DW_AT_decl_file-DW_AT_abstract_origin-crosscu1.s
is crashing here -
#2 0x0000000000425a9f SignalHandler(int) Signals.cpp:0:0
#3 0x00007f57160e9140 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14140)
#4 0x00007f570d911e43 lldb_private::SourceManager::GetFile(lldb_private::FileSpec const&) crtstuff.c:0:0
#5 0x00007f570d914270 lldb_private::SourceManager::DisplaySourceLinesWithLineNumbers(lldb_private::FileSpec const&, unsigned int, unsigned int, unsigned int, unsigned int, char const*, lldb_private::Stream*, lldb_private::SymbolContextList const*) crtstuff.c:0:0
#6 0x00007f570da662c8 lldb_private::StackFrame::GetStatus(lldb_private::Stream&, bool, bool, bool, char const*) crtstuff.c:0:0
I don't get a failure here my mac, I'll review this method more
closely tomorrow.
Jason Molenda [Thu, 26 May 2022 07:30:12 +0000 (00:30 -0700)]
Defer source path remap tilde expansion until source file use
When reading source path remappings out of a dSYM, lldb currently
does tilde expansion -- expanding the tilde-username and checking
that the destination pathname exists, for each dSYM with the path
remappings. This cost happens during lldb's initial process launch
/ load, an especially perf-sensitive time. Inside Apple, we have
dSYMs with source path remappings pointing to NFS directories where
these extra stats for every dSYM can be very expensive if the network
is slow.
This patch instead keeps the source path mapping in the original
tilde-username terms and does the tilde expansion when we need
to read a specific source file from one of the modules. We'll
be stat'ing all of those inodes to load the source file anyway,
so the fact that we do the tilde expansion on every source file
we load, it doesn't cost us significantly.
Differential Revision: https://reviews.llvm.org/D126435
rdar://
77091379
Jason Molenda [Thu, 26 May 2022 07:19:57 +0000 (00:19 -0700)]
Update three tests to realpath paths that we compare to dyld paths
I get to my work directory through a symlink, so the pathnames the
tests get for their build artifacts etc are via that symlink. There
are three tests which compare those symlink paths to a directory
received from dyld on macOS, which is the actual real pathname.
These tests have always failed for me on my dekstop but I finally
sat down to figure out why. Easy quick fix.
Fangrui Song [Thu, 26 May 2022 07:14:08 +0000 (00:14 -0700)]
[MC] Lower case the first letter of EmitCOFF* EmitWin* EmitCV*. NFC
John Paul Adrian Glaubitz [Thu, 26 May 2022 07:00:16 +0000 (00:00 -0700)]
[scudo] Link against libatomic on all MIPS targets
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D126418
Owen Anderson [Mon, 9 May 2022 05:34:43 +0000 (22:34 -0700)]
Replace the custom linked list in LeaderTableEntry with TinyPtrVector.
The purpose of the custom linked list was to optimize for the case
of a single-element list. It turns out that TinyPtrVector handles
the same basic scenario even better, reducing the size of
LeaderTableEntry by 33%, and requiring only log2(N) allocations
as the size of the list grows. The only downside is that we have
to store the Value's and BasicBlock's in separate vectors, which
is slightly awkward in a few cases. Fortunately that ends up being
entirely encapsulated inside helper functions.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D125205
Serguei Katkov [Wed, 25 May 2022 06:37:06 +0000 (13:37 +0700)]
[GuardWidening] Remove nuw/nsw flags for hoisted instructions
When we hoist instructions over guard we must clear flags due to these flags
might be implied using this guard, so they make sense only after the guard.
As an example of the bug due to current behavior.
L is known to be in range say [0, 100)
c1 = x u< L
guard (c1)
x1 = add x, 1
c2 = x1 u< L
guard(c2)
basing on guard(c1) we can say that x1 = add nuw nsw x, 1
after guard widening we get
c1 = x u< L
x1 = add nuw nsw x, 1
c2 = x1 u< L
c = and c1, c2
guard(c)
now, basing on fact that x + 1 < L and x >= 0 due to x + 1 is nuw
we can prove that x + 1 u< L implies that x u< L, so we can just remove c1
x1 = add nuw nsw x, 1
c2 = x1 u< L
guard(c2)
But that is not correct due to we will pass x == -1 value.
Reviewed By: mkazantsev
Subscribers: llvm-commits, nikic
Differential Revision: https://reviews.llvm.org/D126354
serge-sans-paille [Wed, 25 May 2022 20:29:51 +0000 (16:29 -0400)]
[iwyu] Handle regressions in libLLVM header include
Running iwyu-diff on LLVM codebase since
7030654296a0416bd9402a0278 detected a few
regressions, fixing them.
Differential Revision: https://reviews.llvm.org/D126417
Serguei Katkov [Wed, 25 May 2022 06:37:06 +0000 (13:37 +0700)]
[GuardWidening] Add test showing incorrect behavior with nuw/nsw flags
when we move instruction over the guard we must clear nuw/nsw flags
but we do not.
Serguei Katkov [Thu, 26 May 2022 05:17:16 +0000 (12:17 +0700)]
[GuardWidening] Update test with update_test_checks.py
River Riddle [Tue, 24 May 2022 08:55:27 +0000 (01:55 -0700)]
[mlir:Printer][NFC] Add utility methods for printing escaped/hex strings
This simplifies quite a few cases where we manually duplicate the
escaping logic.
River Riddle [Sat, 21 May 2022 01:20:13 +0000 (18:20 -0700)]
[mlir:LSP] Link the test dialect into mlir-lsp-server
This allows for more easily interacting with lit files that utilize
the test dialect.
Matthias Springer [Wed, 25 May 2022 22:37:51 +0000 (00:37 +0200)]
[mlir][bufferization] Clean up imports and code comments
Differential Revision: https://reviews.llvm.org/D126427
Congzhe Cao [Thu, 26 May 2022 03:31:17 +0000 (23:31 -0400)]
[NFC][LoopCacheAnalysis] Update test cases to make sure the outputs follow the right order
In this patch we change test cases from using "CHECK" to using
"CHECK-NEXT", which is to ensure the order of loops output by
loop cache analysis is correct. After D124725 we fixed the
non-deterministic output order hence we did not use "CHECK-DAG"
anymore, and now we should really use "CHECK-NEXT" to make sure
the loops in the output loop vector follow the right order.
Reviewed By: bmahjour, #loopoptwg
Differential Revision: https://reviews.llvm.org/D124984
Chenbing Zheng [Thu, 26 May 2022 03:07:18 +0000 (11:07 +0800)]
[InstCombine] [NFC] refector foldXorOfICmps
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D126268
Chenbing Zheng [Thu, 26 May 2022 02:16:54 +0000 (10:16 +0800)]
[InstCombine] bitcast(logic(bitcast(X), bitcast(Y))) -> bitcast'(logic(bitcast'(X), Y))
This patch break foldBitCastBitwiseLogic limite the destination
must have an integer element type, and eliminate one bitcast by
doing the logic op in the type of the input that has an integer
element type.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D126184
jacquesguan [Wed, 11 May 2022 09:17:43 +0000 (09:17 +0000)]
[RISCV] Replace ISD::FP_EXTEND and ISD::FP_ROUND with RVV VL op.
This patch tries to solve the incoordination between the direct and intermediate cast caused by D123975.
This patch replaces ISD::FP_EXTEND and ISD::FP_ROUND with RVV VL op in the lowering of FP scalable vector direct cast to unify with the intermediate cast.
And it also changes the FP widenning pattern with the VL op.
Differential Revision: https://reviews.llvm.org/D125364
Alan Zhao [Wed, 25 May 2022 22:06:08 +0000 (18:06 -0400)]
[llvm-ml] Add support for the .S extension
Even though MASM files typically have the .asm extension, there are some
use cases [0] where they have the .S extension. MSVC ml assembles such
files with no problems, so llvm-ml should as well.
Additionally, fix the implementation of the /Ta flag and add a test for
it.
[0]: https://crrev.com/c/3668287
Reviewed By: epastor
Differential Revision: https://reviews.llvm.org/D126425
Lian Wang [Thu, 19 May 2022 07:11:18 +0000 (07:11 +0000)]
[LegalizeTypes][VP] Add widen and split support for VP_TRUNCATE
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D125950
Haocong.Lu [Thu, 26 May 2022 01:54:52 +0000 (09:54 +0800)]
[RISCV][NFC] Remove solved TODO for combining constant shifts
Reviewed By: benshi001, asb
Differential Revision: https://reviews.llvm.org/D126185
Peter Klausler [Wed, 25 May 2022 23:44:58 +0000 (16:44 -0700)]
[flang] Allow forward reference to ENTRY from generic interface
The CreateEntry() function in name resolution needs to allow for the name
of an alternate entry point already having been declared in the outer scope
as the homonymous specific procedure of a generic interface; e.g.,
interface foo
module procedure foo
end interface
subroutine bar
entry foo
end subroutine
Differential Revision: https://reviews.llvm.org/D126436
V Donaldson [Wed, 25 May 2022 22:32:02 +0000 (15:32 -0700)]
[flang] Update intrinsic module source files
The f18 standard defines several intrinsic modules containing definitions
and declarations for various constants, types, and procedures. This PR adds
declarations for missing procedures in these modules.
bixia1 [Tue, 24 May 2022 23:07:31 +0000 (16:07 -0700)]
[mlir][sparse] Add more complex operations.
Support complex operations sqrt, expm1, and tanh.
Add tests.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D126393
bixia1 [Wed, 25 May 2022 19:39:37 +0000 (12:39 -0700)]
Lower complex.expm1 to standard dialect.
Add a test.
Reviewed By: pifon2a
Differential Revision: https://reviews.llvm.org/D126409
Jason Molenda [Wed, 25 May 2022 23:05:16 +0000 (16:05 -0700)]
Skip testing of watchpoint hit-count/ignore-count on multithreaded
Skip all watchpoint hit-count/ignore-count tests for multithreaded
API tests for now on arm64 Darwin.
On AArch64, insns that trigger a WP are rolled back and we are
notified. lldb needs to disable the WP, insn step, re-enable it,
then report it to the user. lldb only does this full step action
for the "selected thread", and so when a program stops with
multiple threads hitting a stop reason, some of them watchpoints,
any non-selected-thread will not be completed in this way. But
all threads with the initial watchpoint exception will have their
hit-count/ignore-counts updated. When we resume execution, the
other threads sitting at the instruction will again execute &
trigger the WP exceptoin again, repeating until we've gone through
all of the threads.
This bug is being tracked in llvm.org/pr49433 and inside apple
in rdar://
93863107
Alan Zhao [Wed, 25 May 2022 22:12:35 +0000 (18:12 -0400)]
[NFC][llvm-ml] Rename rip-relative-addressing.asm to rip_relative_addressing.asm
All the other tests here use underscores, so we should rename this file
to be consistent.
Arthur Eubanks [Wed, 25 May 2022 21:47:30 +0000 (14:47 -0700)]
[gn build] Fix D126415 to only set `llvm_have_mallinfo2 = true` on linux
Otherwise mac builds are broken, e.g. http://45.33.8.238/macm1/36056/step_4.txt
../../llvm/lib/Support/Unix/Process.inc:35:10: fatal error: 'malloc.h' file not found
#include <malloc.h>
Peter Klausler [Wed, 25 May 2022 20:32:52 +0000 (13:32 -0700)]
[flang] Fix crash from a blank BIND(C,NAME="") on subprogram
A recent change fixed the processing of BIND(C,NAME=expr) character
expressions so that they are evaluated as constants in the scope of
the subprogram. However, when the character name expression results
in an empty value after trimming, the compiler emits a warning message,
and this message is now causing a crash due to a lack of statement
context. To fix, extend the deferred processing of the BIND(C,NAME="")
so that a basic statement context exists.
Differential Revision: https://reviews.llvm.org/D126416
Arthur Eubanks [Wed, 25 May 2022 21:36:42 +0000 (14:36 -0700)]
[gn build] Manually port
bed9efed71
Alexey Bataev [Wed, 25 May 2022 14:21:55 +0000 (07:21 -0700)]
[SLP]Do not try to generate ScatterVectorize if it will be scalarized.
SLP should build ScatterVectorize nodes only if they actually end up
with masked gather rather than with scalarization. In the second
scenario better to build a gather node.
Differential Revision: https://reviews.llvm.org/D126379
NAKAMURA Takumi [Wed, 25 May 2022 21:14:16 +0000 (06:14 +0900)]
[bazel] Unset REVISION as if LLVM_APPEND_VC_REV=OFF, for now.
We could implement retrieving the revision here, but we may avoid
"Just the same but only different revision hash string".
NAKAMURA Takumi [Wed, 25 May 2022 21:10:22 +0000 (06:10 +0900)]
[bazel] Bump to 15.0.0git
Snehasish Kumar [Tue, 24 May 2022 23:58:36 +0000 (23:58 +0000)]
[memprof] Keep and display symbol names in the RawMemProfReader.
Extend the Frame struct to hold the symbol name if requested
when a RawMemProfReader object is constructed. This change updates the
tests and removes the need to pass --debug to obtain the mapping from
GUID to symbol names.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D126344
Alexey Bataev [Wed, 25 May 2022 11:37:51 +0000 (04:37 -0700)]
[SLP]Fix PR55688: Miscompile due to incorrect nuw/nsw handling.
Need to use all ReductionOps when propagating flags for the reduction
ops, otherwise transformation is not correct. Plus, need to drop nuw/nsw
flags.
Differential Revision: https://reviews.llvm.org/D126371
Eric Li [Wed, 25 May 2022 20:15:46 +0000 (20:15 +0000)]
[clang][dataflow] Fix incorrect CXXThisExpr pointee for lambdas
When constructing the `Environment`, the `this` pointee is established
for a `CXXMethodDecl` by looking at its parent. However, inside of
lambdas, a `CXXThisExpr` refers to the captured `this` coming from the
enclosing member function.
When establishing the `this` pointee for a function, we check whether
the function is a lambda, and check for an enclosing member function
to establish the `this` pointee storage location.
Differential Revision: https://reviews.llvm.org/D126413
Eric Li [Wed, 25 May 2022 19:21:08 +0000 (19:21 +0000)]
[clang][dataflow] Relax assert on existence of `this` pointee storage
Support for unions is incomplete (per
99f7d55e) and the `this` pointee
storage location is not set for unions. The assert in
`VisitCXXThisExpr` is then guaranteed to trigger when analyzing member
functions of a union.
This commit changes the assert to an early-return. Any expression may
be undefined, and so having a value for the `CXXThisExpr` is not a
postcondition of the transfer function.
Differential Revision: https://reviews.llvm.org/D126405
Arthur Eubanks [Wed, 25 May 2022 20:30:37 +0000 (13:30 -0700)]
[gn build] Set llvm_have_mallinfo2 to true by default
If you are using an old enough glibc which doesn't have mallinfo2, set
`llvm_have_mallinfo2 = false`.
At this point it's likely that most people using the gn build are
compiling against a recent enough glibc (glibc 2.33 which was released
in Feb 2021).
Reviewed By: peterwaller-arm
Differential Revision: https://reviews.llvm.org/D126415
Maksim Panchenko [Thu, 19 May 2022 20:23:40 +0000 (13:23 -0700)]
[MCDisassembler] Disambiguate Size parameter in tryAddingSymbolicOperand()
MCSymbolizer::tryAddingSymbolicOperand() overloaded the Size parameter
to specify either the instruction size or the operand size depending on
the architecture. However, for proper symbolic disassembly on X86, we
need to know both sizes, as an instruction can have two operands, and
the instruction size cannot be reliably calculated based on the operand
offset and its size. Hence, split Size into OpSize and InstSize.
For X86, the new interface allows to fix a couple of issues:
* Correctly adjust the value of PC-relative operands.
* Set operand size to zero when the operand is specified implicitly.
Differential Revision: https://reviews.llvm.org/D126101
zr33 [Wed, 25 May 2022 20:40:11 +0000 (13:40 -0700)]
[BOLT][DWARF] Convert dwarf5-df-* tests to assembly tests
Reviewed By: ayermolo
Differential Revision: https://reviews.llvm.org/D126086
Patrick Walton [Wed, 25 May 2022 20:30:49 +0000 (13:30 -0700)]
Round up zero-sized symbols to 1 byte in `.debug_aranges`.
This commit modifies the AsmPrinter to avoid emitting any zero-sized symbols to
the .debug_aranges table, by rounding their size up to 1. Entries with zero
length violate the DWARF 5 spec, which states:
> Each descriptor is a triple consisting of a segment selector, the beginning
> address within that segment of a range of text or data covered by some entry
> owned by the corresponding compilation unit, followed by the non-zero length
> of that range.
In practice, these zero-sized entries produce annoying warnings in lld and
cause GNU binutils to truncate the table when parsing it.
Other parts of LLVM, such as DWARFDebugARanges in the DebugInfo module
(specifically the appendRange method), already avoid emitting zero-sized
symbols to .debug_aranges, but not comprehensively in the AsmPrinter. In fact,
the AsmPrinter does try to avoid emitting such zero-sized symbols when labels
aren't involved, but doesn't when the symbol to emitted is a difference of two
labels; this patch extends that logic to handle the case in which the symbol is
defined via labels.
Reviewed By: dblaikie
Differential Revision: https://reviews.llvm.org/D126257
Joseph Huber [Wed, 25 May 2022 17:22:26 +0000 (13:22 -0400)]
[Clang] Introduce `--offload-link` option to perform offload device linking
The new driver uses an augmented linker wrapper to perform the device
linking phase, but to the user looks like a regular linker invocation.
Contrary to the old driver, the new driver contains all the information
necessary to produce a linked device image in the host object itself.
Currently, we infer the usage of the device linker by the user
specifying an offloading toolchain, e.g. (--offload-arch=...) or
(-fopenmp-targets=...), but this shouldn't be strictly necessary.
This patch introduces a new option `--offload-link` to tell
the driver to use the offloading linker instead. So a compilation flow
can now look like this,
```
clang foo.cu --offload-new-driver -fgpu-rdc --offload-arch=sm_70 -c
clang foo.o --offload-link -lcudart
```
I was considering if this could be merged into the `-fuse-ld` option,
but because the device linker wraps over the users linker it would
conflict with that. In the future it's possible to merge this into `lld`
completely or `gold` via a plugin and we would use this option to
enable the device linking feature. Let me know what you think for this.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D126398
Matthias Springer [Wed, 25 May 2022 20:25:27 +0000 (22:25 +0200)]
[mlir][bufferization][NFC] Remove dead code
There were two copies of AlwaysCopyAnalysisState. (Must have been a merge conflict mistake...)
Differential Revision: https://reviews.llvm.org/D126414
Zequan Wu [Wed, 18 May 2022 01:23:35 +0000 (18:23 -0700)]
[LLDB][NativePDB] Check for missing type info to avoid crash.
NativePDB often assumes that all debug info are available.
This is one step to make it more pervasive.
Differential Revision: https://reviews.llvm.org/D125844
Joe Nash [Tue, 10 May 2022 18:46:57 +0000 (14:46 -0400)]
[AMDGPU] gfx11 FLAT Instructions
MachineCode Support for FLAT type instructions
Contributors:
Sebastian Neubauer <sebastian.neubauer@amd.com>
Patch 12/N for upstreaming of AMDGPU gfx11 architecture.
Depends on D125989
Reviewed By: rampitec, #amdgpu
Differential Revision: https://reviews.llvm.org/D125992
Thomas Preud'homme [Wed, 25 May 2022 18:37:15 +0000 (19:37 +0100)]
Fix conversion error to Expected
On Ubuntu 18.04 with GCC 7.5 Intel trace code fails to build due to
failure to convert from
lldb_private::process_linux::IntelPTPerThreadProcessTraceUP to
Expected<lldb_private::process_linux::IntelPTPerThreadProcessTraceUP>.
This commit explicitely marks those unique_ptr values as being moved
which fixes the conversion error.
Reviewed By: wallace
Differential Revision: https://reviews.llvm.org/D126402
Alex Lorenz [Tue, 24 May 2022 20:41:39 +0000 (13:41 -0700)]
[clang] Fix the begin location of concepts specialization expression
The concept specialization expression should start at the location of
the nested qualifiers when it has nested qualifiers.
This ensures that libclang reports correct source ranges that include
all subexpressions when visiting the expression.
Differential Revision: https://reviews.llvm.org/D126332
Chia-hung Duan [Wed, 25 May 2022 19:21:48 +0000 (19:21 +0000)]
[mlir][PDLL] Allow numeric result indexing for unregistered op
If we don't specify the result index while matching operand with the
result of certain operation, it's supposed to match all the results of
the operation with the operand. For registered op, it's easy to do that
by either indexing with number or name. For unregistered op, this commit
enables the numeric result indexing for this use case.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D126330
Joe Nash [Wed, 27 Apr 2022 13:17:25 +0000 (09:17 -0400)]
[AMDGPU] gfx11 vinterp instructions MC support
A new instruction encoding. Some of these instructions were previously VOP3
encoded.
Contributors:
Carl Ritson <carl.ritson@amd.com>
Patch 11/N for upstreaming of AMDGPU gfx11 architecture.
Depends on D125824
Reviewed By: critson
Differential Revision: https://reviews.llvm.org/D125989
Joe Nash [Mon, 25 Apr 2022 17:58:14 +0000 (13:58 -0400)]
[AMDGPU] gfx11 export instructions
Contributors:
Jay Foad <jay.foad@amd.com>
Dmitry Preobrazhensky <d-pre@mail.ru>
Patch 10/N for upstreaming of AMDGPU gfx11 architecture.
Depends on D125822
Reviewed By: dp
Differential Revision: https://reviews.llvm.org/D125824
Christian Sigg [Wed, 25 May 2022 11:41:53 +0000 (13:41 +0200)]
[LLVM] Add rcp.approx.ftz.f32 intrinsic
Split out from https://reviews.llvm.org/D126158.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D126369
Paul Kirth [Mon, 23 May 2022 18:22:40 +0000 (18:22 +0000)]
[compiler-rt][test] Fix flake in symbolize_stack test
Addresses tests flakes described in
https://github.com/llvm/llvm-project/issues/55460
The test being updated can fail in FileCheck to match when given long
enough stack traces. This can be problematic when file system paths
become long enough to cause the majority of the long function name to
become truncated. We found in our CI that the truncated output would
often fail to match, thereby causing the test to fail when it should not.
Here we change the test to match on sybolizer output that should be more
reliable than matching inside the long function name.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D126102
Peter Klausler [Wed, 25 May 2022 16:19:19 +0000 (09:19 -0700)]
[flang] Address regression (calls to assumed-length character function dummy procedures)
A recent fix beefed up semantics checking to catch the case of a call
to an external assumed-length character function; this check has false
positives in the case of an assumed-length character function that is
a dummy procedure. These do have a length that is passed in extra
compiler-created arguments. This patch refines the check and undoes some
changes to tests.
Differential Revision: https://reviews.llvm.org/D126390
James Y Knight [Wed, 25 May 2022 18:19:59 +0000 (14:19 -0400)]
C++ DR2394: Const-default-constructible for members.
Const class members may be initialized with a defaulted default
constructor under the same conditions it would be allowed for a const
object elsewhere.
Differential Revision: https://reviews.llvm.org/D126170
Philip Reames [Wed, 25 May 2022 18:07:31 +0000 (11:07 -0700)]
Reapply "[RISCV] Enable strict assertions in InsertVSETVLI data flow"
be2cb8 fixes the case which triggered the revert. Reapply, and let's see if anything else falls out.
Original commit message:
These asserts are believed to hold after several recent miscompiles have been fixed. If you see an assertion failure on this change, please toggle the default back and make sure you file a bug with a reproducer. We may have as yet uncaught miscompiles lurking in this code.
Differential Revision: https://reviews.llvm.org/D125271
Shoaib Meenai [Wed, 25 May 2022 17:22:10 +0000 (10:22 -0700)]
[pseudo] Fix pseudo-gen usage when cross-compiling
Use the LLVM build system's cross-compilation support for the tool, so
that the build works for both host and cross-compilation scenarios.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D126397
Daniel Grumberg [Mon, 16 May 2022 10:46:36 +0000 (11:46 +0100)]
[clang][extract-api] Don't emit symbols prefixed with an underscore
These symbols are understood to not be used for client API consumption
by convention so they should not appear in the generated symbol graph.
Differential Revision: https://reviews.llvm.org/D125678
Krzysztof Parzyszek [Wed, 25 May 2022 13:46:25 +0000 (06:46 -0700)]
DWARFVerifier: Change vector of IntervalMap to vector of unique_ptrs
This is a workaround for compilation issue on FreeBSD. See comments in
https://reviews.llvm.org/rG0d8cb8b399ad for more information.
This fixes https://github.com/llvm/llvm-project/issues/55414.
Differential Revision: https://reviews.llvm.org/D125611
Matthias Springer [Wed, 25 May 2022 17:39:57 +0000 (19:39 +0200)]
[mlir][bufferize] Remove hoisting functionality from One-Shot Bufferize
The same functionality is already provided by `-buffer-hoisting` and `-buffer-loop-hoisting`.
Differential Revision: https://reviews.llvm.org/D126251
Sanjay Patel [Wed, 25 May 2022 17:43:13 +0000 (13:43 -0400)]
[SLP] add minimum test for miscompile (PR55688); NFC
Philip Reames [Wed, 25 May 2022 17:37:26 +0000 (10:37 -0700)]
[riscv] Remove mutation of prior vsetvli from insertion dataflow
This moves mutation entirely out of the main algorithm.
The immediate trigger is that we hit another case of the same issue I thought we'd fixed in 72925d9. It turns out we hadn't considered the cross block case.
As a brief summary, the issue being fixed is that if we mutate a previous vsetvli in phase 3, there's a possibility that some later use of that vsetvli changes "compatibility". In the cross_block_mutate test, this later vsetvli occurs in another block (and is thus visit order dependent too!). This causes us to fail strict asserts. (To be explicit, the current on by default workaround should compensate. It's only when we turn that off that we have problems.)
Now, I want to explicitly call out an alternate workaround. We could leave the mutation in phase 3, and simplify restrict it to the case where the previous vsetvli's GPR result is unused. That covers the case we've actually seen. (I'll note that codegen regressions with a simple form of this were significant. We might have to check specifically for the use outside block case to keep them reasonable, which complicates the workaround slightly.)
Personally, I'm at the point where I want the mutation pulled out just for robustness sake. I'm worried there's yet one more form of this bug we haven't thought about.
The other motivation for this change is that it does give us a couple of minor codegen wins. None appear to be hugely significant, but improvements never hurt right?
Differential Revision: https://reviews.llvm.org/D125270
Aaron Ballman [Wed, 25 May 2022 17:45:17 +0000 (13:45 -0400)]
Removing this as part of the revert done in
69da3b6aead2e7a18a2578aad661d6d36b8d30cf
This appears to have been added in a follow-up commit that I missed.
Aaron Ballman [Wed, 25 May 2022 17:33:30 +0000 (13:33 -0400)]
Revert "[OpenMP] atomic compare fail : Parser & AST support"
This reverts commit
232bf8189ef7d574a468bd5bfd1e84e962f7f16e.
It broke the sanitize buildbot: https://lab.llvm.org/buildbot/#/builders/5/builds/24074
It also reproduces on Windows debug builds as a crash.
Mike Rice [Tue, 24 May 2022 20:57:41 +0000 (13:57 -0700)]
[OpenMP] Add diagnostic for unterminated 'omp [begin] declare target'
Warns when end-of-file is reached without seeing all matching
'omp end declare target' directives. The diagnostic shows the
location of the related begin directive.
Differential Revision: https://reviews.llvm.org/D126331
Joseph Huber [Wed, 25 May 2022 17:32:25 +0000 (13:32 -0400)]
[Clang][Docs] Document `-Xoffload-linker` flag
Summary:
I added the `-Xoffload-linker` flag and did not provide additional
documentation. This patch adds it.
Nicolai Hähnle [Tue, 10 May 2022 15:10:36 +0000 (10:10 -0500)]
AMDGPU/GISel: Factor out AMDGPURegisterBankInfo::buildReadFirstLane
A later change will add a 3rd user, so factoring out the common code
seems useful.
Reorganizing the executeInWaterfallLoop causes some more COPYs to be
generated, but those all fold away during instruction selection.
Generating the comparisons uses generic instructions over machine
instructions now which admittedly shouldn't make a difference
(though it should make it easier to move the waterfall loop generation
to another place).
(Resubmit with missing test added.)
Differential Revision: https://reviews.llvm.org/D125324
Nicolai Hähnle [Wed, 25 May 2022 17:03:23 +0000 (12:03 -0500)]
Revert "AMDGPU/GISel: Factor out AMDGPURegisterBankInfo::buildReadFirstLane"
This reverts commit
2a28467e5389c4d741d1825fadd39ae84ecaa5dc.
Zequan Wu [Fri, 29 Apr 2022 23:38:56 +0000 (16:38 -0700)]
Reland "[llvm-pdbutil] Add options to only dump symbol record at specified offset and its parents or children with spcified depth."
This reverts commit
cfb4e782520ce59602a34732386ebbdf58136cfb.
Kazu Hirata [Wed, 25 May 2022 16:49:05 +0000 (09:49 -0700)]
[lldb] Fix an unused function warning
This patch fixes:
.../llvm-project/lldb/source/Host/common/PseudoTerminal.cpp:106:20:
error: unused function 'use_ptsname' [-Werror,-Wunused-function]
Logan Chien [Tue, 24 May 2022 01:04:30 +0000 (18:04 -0700)]
[mlir] Fix Tensor_InsertSliceOp description
This commit fixes `Tensor_InsertSliceOp` `sizes` inputs/attributes
description.
Before this commit, the description says the `sizes` inputs/attributes
denote the size of the return type. But according to the
`InsertSliceOpConstantArgumentFolder` in
`lib/Dialect/Tensor/IR/TensorOps.cpp`, the `sizes` inputs/attributes
actually denote the size of the source type.
I had an off-line discussion with the authors of `TensorOps.td` and
`TensorOps.cpp`. We concluded that it was a typo in the Op description.
This commit updates the Op description to match the actual usage.
Differential Revision: https://reviews.llvm.org/D126264
Nicolai Hähnle [Tue, 10 May 2022 15:10:36 +0000 (10:10 -0500)]
AMDGPU/GISel: Factor out AMDGPURegisterBankInfo::buildReadFirstLane
A later change will add a 3rd user, so factoring out the common code
seems useful.
Reorganizing the executeInWaterfallLoop causes some more COPYs to be
generated, but those all fold away during instruction selection.
Generating the comparisons uses generic instructions over machine
instructions now which admittedly shouldn't make a difference
(though it should make it easier to move the waterfall loop generation
to another place).
Differential Revision: https://reviews.llvm.org/D125324
Takafumi Arakaki [Wed, 25 May 2022 16:12:11 +0000 (16:12 +0000)]
Allow pointer types for atomicrmw xchg
This adds support for pointer types for `atomic xchg` and let us write
instructions such as `atomicrmw xchg i64** %0, i64* %1 seq_cst`. This
is similar to the patch for allowing atomicrmw xchg on floating point
types: https://reviews.llvm.org/D52416.
Differential Revision: https://reviews.llvm.org/D124728
Craig Topper [Wed, 25 May 2022 16:16:07 +0000 (09:16 -0700)]
[RISCV] Preserve fast math flags in lowerVPOp.
Update test to check MIR after finalize-isel instead of debug output.
This is of course not the only place we should preserve FMF, but
it's the most obvious one.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D126306
Craig Topper [Wed, 25 May 2022 16:11:11 +0000 (09:11 -0700)]
[X86] Add isSimple check to the load combine in combineExtractVectorElt.
I think we need to be sure the load isn't volatile before we
duplicate and shrink it.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D126353
Anubhab Ghosh [Wed, 25 May 2022 16:01:12 +0000 (09:01 -0700)]
[llvm-objcopy][ObjectYAML][mips] Add MIPS specific ELF section indexes
This fixes https://github.com/llvm/llvm-project/issues/53998
and displays correct information in obj2yaml for SHN_MIPS_*
sections according to
https://refspecs.linuxfoundation.org/elf/mipsabi.pdf
Reviewed By: jhenderson, MaskRay
Differential Revision: https://reviews.llvm.org/D123902
Martin Boehme [Fri, 20 May 2022 12:04:44 +0000 (14:04 +0200)]
[clang] Don't parse MS attributes in `ParseExportDeclaration()`.
As @rsmith commented on https://reviews.llvm.org/D111548: "That looks like it's
simply a bug as far as I can tell, and that call can be removed. MS attributes
will be parsed as part of the decl specifier sequence as needed and don't need
to be parsed as declaration attributes."
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D126062
Philip Reames [Wed, 25 May 2022 15:00:27 +0000 (08:00 -0700)]
[RISCV] Hoist VSETVLI vlmax, vtype out of scalable loops
This is a straight forward extension of the PRE transform introduced in D124869 to handle the VLMAX case.
The test changes here look quite positive. This surprised me until I realized that all the tests are using @llvm.vscale to figure out the VLMAX, not the llvm.riscv.vsetvlmax intrinsic. If they'd used the later, these would have been full redundancy cases and fully handled by the data flow. I'm not really sure if use of vscale here is representative or not. If it is, we should probably look at using VSETVLI to lower vscale rather than a raw read of vlenb and some math.
Differential Revision: https://reviews.llvm.org/D126338
Philip Reames [Wed, 25 May 2022 14:58:11 +0000 (07:58 -0700)]
[RISCV] Restructure comment and add clarifying assert to getFrameIndexReference [NFC]
Differential Revision: https://reviews.llvm.org/D126088
Lei Zhang [Wed, 25 May 2022 14:53:35 +0000 (10:53 -0400)]
[mlir][spirv] Fix capability check for 64-bit element types
Using 64-bit integer/float type in interface storage classes would
require Int64/Float64 capability, per the Vulkan spec:
```
shaderInt64 specifies whether 64-bit integers (signed and unsigned) are
supported in shader code. If this feature is not enabled, 64-bit integer
types must not be used in shader code. This also specifies whether
shader modules can declare the Int64 capability. Declaring and using
64-bit integers is enabled for all storage classes that SPIR-V allows
with the Int64 capability.
```
This is different from, say, 16-bit element types, where:
```
shaderInt16 specifies whether 16-bit integers (signed and unsigned) are
supported in shader code. If this feature is not enabled, 16-bit integer
types must not be used in shader code. This also specifies whether
shader modules can declare the Int16 capability. However, this only
enables a subset of the storage classes that SPIR-V allows for the Int16
SPIR-V capability: Declaring and using 16-bit integers in the Private,
Workgroup (for non-Block variables), and Function storage classes is
enabled, while declaring them in the interface storage classes (e.g.,
UniformConstant, Uniform, StorageBuffer, Input, Output, and
PushConstant) is not enabled.
```
Reviewed By: hanchung
Differential Revision: https://reviews.llvm.org/D126256
Lei Zhang [Wed, 25 May 2022 14:45:26 +0000 (10:45 -0400)]
[mlir][scf] Retain existing attributes in scf.for transforms
These attributes can carry useful information, e.g., pipelines
might use them to organize and chain patterns.
Reviewed By: hanchung
Differential Revision: https://reviews.llvm.org/D126320
Joe Loser [Tue, 24 May 2022 20:40:08 +0000 (14:40 -0600)]
[libc++][test] Verify std::ranges::rbegin, crbegin, rend, crend are CPOs
Uncomment the tests to ensure `std::ranges::rbegin` and friends are indeed
customization points objects.
Differential Revision: https://reviews.llvm.org/D126325
Philip Reames [Wed, 25 May 2022 14:41:45 +0000 (07:41 -0700)]
[riscv] Add coverage for fixed length vector loops using LMUL
Ivan Kosarev [Wed, 25 May 2022 14:15:06 +0000 (15:15 +0100)]
[TableGen] Undeprecate 'field' when used with the CodeEmitterGen backend.
Differential Revision: https://reviews.llvm.org/D126290
Michał Górny [Wed, 25 May 2022 11:04:44 +0000 (13:04 +0200)]
[llvm-rc] Avoid which(1) dependency in windres-prefix.test
Rely on lit substitution to provide the path to llvm-windres instead
of redundantly calling which(1) with the subtituted absolute path.
This fixes test failure on the happy systems without which(1) installed.
Differential Revision: https://reviews.llvm.org/D126366
Groverkss [Wed, 25 May 2022 14:06:35 +0000 (19:36 +0530)]
[MLIR][Presburger] Add inverse to IntegerRelation
This patch adds support for obtaining inverse of a relation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D126327
Nikita Popov [Wed, 25 May 2022 14:05:11 +0000 (16:05 +0200)]
[PhaseOrdering] Add test for unprofitable loop load PRE backedge splitting (NFC)
Groverkss [Wed, 25 May 2022 14:02:21 +0000 (19:32 +0530)]
[MLIR][Presburger] Add getDomainSet, getRangeSet to IntegerRelation
This patch adds support for obtaining a set corresponding to the domain/range
of the relation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D126326
Louis Dionne [Wed, 25 May 2022 14:04:47 +0000 (10:04 -0400)]
[libc++] Remove conditional include
Simon Moll [Wed, 25 May 2022 13:08:31 +0000 (15:08 +0200)]
[VP][fix] Don't discard masks in reductions
When expanding VP reductions to non VP-code, the reduction pass was
ignoring the mask before. Fix this by keeping the mask and selecting
neutral elements where the mask is zero.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D126362
Paul Robinson [Wed, 25 May 2022 13:46:12 +0000 (06:46 -0700)]
[PS5] Default to -fno-rtti
Louis Dionne [Tue, 24 May 2022 15:14:23 +0000 (11:14 -0400)]
[libc++] Use Python subprocess instead of libc++'s own utilities
Once we move off entirely from the legacy testing framework, this will
allow removing a bunch of code.
Differential Revision: https://reviews.llvm.org/D126303
Haojian Wu [Wed, 25 May 2022 13:25:03 +0000 (15:25 +0200)]
Fix unused-variable warning, NFC.