Sterling Augustine [Mon, 14 Mar 2022 22:21:19 +0000 (15:21 -0700)]
Revert "[lld-macho] Avoid using bump-alloc in TrieBuider"
This reverts commit
e049a87f04cff8e81b4177097a6b2fdf37c7b148.
That commit breaks the build with errors of the form:
/usr/local/google/home/saugustine/llvm/llvm-project/lld/MachO/ExportTrie.cpp:148:11: error: definition of implicitly declared destructor
TrieNode::~TrieNode() {
Peter Klausler [Fri, 11 Mar 2022 18:17:04 +0000 (10:17 -0800)]
[flang] IEEE_ARITHMETIC must imply USE IEEE_EXCEPTIONS
The intrinsic module IEEE_ARITHMETIC must incorporate the public
names from the intrisic module IEEE_EXCEPTIONS. Rename IEEE_EXCEPTIONS
to __Fortran_ieee_exceptions so that it won't clash with the
nonintrinsic namespace, establish a new intrinic IEEE_EXCEPTIONS
module that USEs it, and add a USE to IEEE_ARITHMETIC.
Updated to use STREQUAL rather than ambiguous MATCHES in
the CMakeLists.txt file.
Differential Revision: https://reviews.llvm.org/D121490
Stanislav Mekhanoshin [Tue, 8 Mar 2022 19:14:33 +0000 (11:14 -0800)]
[AMDGPU] gfx940 memory model
Differential Revision: https://reviews.llvm.org/D121242
Andrew Litteken [Mon, 14 Mar 2022 04:44:46 +0000 (23:44 -0500)]
[IROutliner] Ensure merged PHINodes respect order and incoming blocks, not just incoming values
When matching PHINodes when margining functions the IROutliner only checks that an incoming value exists in phi node in overall function. It doesn't check the length, the order, or that the incoming block also matches. In the given example, we see that both phi nodes have the same incoming values, but from different blocks.
The fix is to to enforce stricter a match of the incoming value, and the incoming block as well when matching the created phi nodes.
Reviewers: paquette
Differential Revision: https://reviews.llvm.org/D121310
Craig Topper [Mon, 14 Mar 2022 21:30:42 +0000 (14:30 -0700)]
[InstCombine] Fold select based logic of fcmps with same operands when FMF is present.
If we have a logical and/or in select form and the true/false operand
is an fcmp with poison generating FMF, we won't be able to fold it
to an and/or instruction. This prevents us from optimizing the case
where it is a logical operation of two fcmps with identical operands.
This patch adds explicit checks for this case that doesn't rely on
converting to and/or to do the optimization. It reuses the existing
foldLogicOfFCmps, but adds a new flag to disable the other combine
that is inside that function.
FMF flags from the two FCmps are intersected using the logic added in
D121243. The FIXME has been updated to indicate that we can only use
a union for the non-select form.
This allows us to optimize cases like this from compare-fp-3.c in the
gcc torture suite with fast math.
void
test1 (float x, float y)
{
if ((x==y) && (x!=y))
link_error0();
}
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D121323
Nick Desaulniers [Mon, 14 Mar 2022 21:08:25 +0000 (14:08 -0700)]
[IRLinker] make IRLinker::AddLazyFor optional (llvm::unique_function). NFC
2 of the 3 callsite of IRMover::move() pass empty lambda functions. Just
make this parameter llvm::unique_function.
Came about via discussion in D120781. Probably worth making this change
regardless of the resolution of D120781.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D121630
Vy Nguyen [Mon, 14 Mar 2022 21:22:24 +0000 (17:22 -0400)]
[lld-macho] Avoid using bump-alloc in TrieBuider
The code can be used in multi-threads and the allocator is not thread safe.
fixes PR/54378
Reviewed By: int3, #lld-macho
Differential Revision: https://reviews.llvm.org/D121638
Simon Pilgrim [Mon, 14 Mar 2022 21:20:16 +0000 (21:20 +0000)]
[X86] Fix avx512.mask.vpshld/vpshrd tests to correctly test maskz cases
Dávid Bolvanský [Mon, 14 Mar 2022 21:19:40 +0000 (22:19 +0100)]
[AttrDocs] try to fix build
Fangrui Song [Mon, 14 Mar 2022 21:13:41 +0000 (14:13 -0700)]
[ELF] Move section assignment from initializeSymbols to postParse
https://discourse.llvm.org/t/parallel-input-file-parsing/60164
initializeSymbols currently sets Defined::section and handles non-prevailing
COMDAT groups. Move the code to the parallel postParse to reduce work from the
single-threading code path and make parallel section initialization infeasible.
Postpone reporting duplicate symbol errors so that the messages have the
section information. (`Defined::section` is assigned in postParse and another
thread may not have the information).
* duplicated-synthetic-sym.s: BinaryFile duplicate definition (very rare) now
has no section information
* comdat-binding: `%t/w.o %t/g.o` leads to an undesired undefined symbol. This
is not ideal but we report a diagnostic to inform that this is unsupported.
(See release note)
* comdat-discarded-lazy.s: %tdef.o is unextracted. The new behavior (discarded
section error) makes more sense
Depends on D120640
Reviewed By: peter.smith
Differential Revision: https://reviews.llvm.org/D120626
Stella Laurenzo [Mon, 14 Mar 2022 21:10:08 +0000 (14:10 -0700)]
NFC: Remove unterminated string from Python pyi file.
Stanislav Mekhanoshin [Fri, 11 Mar 2022 21:42:11 +0000 (13:42 -0800)]
[AMDGPU] Restrict machine copy propagation from creating unaligned classes
Fixes: SWDEV-326366
Differential Revision: https://reviews.llvm.org/D121491
Valentin Clement [Mon, 14 Mar 2022 21:06:02 +0000 (22:06 +0100)]
[flang] Lower any intrinsic
This patch lowers the `any` intrinsic function to FIR.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121609
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: mleair <leairmark@gmail.com>
Amara Emerson [Mon, 14 Mar 2022 20:47:14 +0000 (13:47 -0700)]
[GlobalISel] Fix store merging incorrectly merging volatile stores.
The existing volatile checks only handle aliasing hazards between stores,
but that isn't enough since by that point volatile stores may have already
been added to the current candidate group.
Andrew Browne [Mon, 14 Mar 2022 20:47:21 +0000 (13:47 -0700)]
Revert "[DFSan] Remove trampolines to unblock opaque pointers."
This reverts commit
84af90336fed36f7dfdc468ded39236f32bbb82e.
Dávid Bolvanský [Mon, 14 Mar 2022 20:44:59 +0000 (21:44 +0100)]
[Clang] always_inline statement attribute
Motivation:
```
int test(int x, int y) {
int r = 0;
[[clang::always_inline]] r += foo(x, y); // force compiler to inline this function here
return r;
}
```
In 2018, @kuhar proposed "Introduce per-callsite inline intrinsics" in https://reviews.llvm.org/D51200 to solve this motivation case (and many others).
This patch solves this problem with call site attribute. "noinline" statement attribute already landed in D119061. Also, some LLVM Inliner fixes landed so call site attribute is stronger than function attribute.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D120717
Andrew Browne [Tue, 8 Mar 2022 21:18:29 +0000 (13:18 -0800)]
[DFSan] Remove trampolines to unblock opaque pointers.
https://github.com/llvm/llvm-project/issues/54172
Reviewed By: pcc
Differential Revision: https://reviews.llvm.org/D121250
Florian Mayer [Fri, 11 Mar 2022 02:15:48 +0000 (18:15 -0800)]
[MTE] Add test that stack tagging does not mess up stack coloring.
Reviewed By: eugenis
Differential Revision: https://reviews.llvm.org/D121433
Shafik Yaghmour [Mon, 14 Mar 2022 20:32:03 +0000 (13:32 -0700)]
[LLDB] Applying clang-tidy modernize-use-default-member-init over LLDB
Applied modernize-use-default-member-init clang-tidy check over LLDB.
It appears in many files we had already switched to in class member init but
never updated the constructors to reflect that. This check is already present in
the lldb/.clang-tidy config.
Differential Revision: https://reviews.llvm.org/D121481
Andrew Litteken [Mon, 14 Mar 2022 04:45:09 +0000 (23:45 -0500)]
[IROutliner] Separate split PHI nodes from multiple exits by different outlinable regions.
The IR Outliner is supposed to extract the outputs contained in an external phi node and place them into a phi node contained within the outlined function. However, when the output values of two outlined functions with two different output sets are contained within the same phi node, they are counted as the same exit path when first analyzed. In reality, these create two different phi nodes, creating an inconsistency, resulting in a mismatch in the expected number of output paths and a crash. This fixes that counting when analyzing the outputs by also analyzing the incoming blocks rather than just the incoming values.
Reviewer: paquette
Differential Revision: https://reviews.llvm.org/D121313
Peter Collingbourne [Tue, 8 Mar 2022 22:48:16 +0000 (14:48 -0800)]
gn build: Add support for building with libcurl.
Differential Revision: https://reviews.llvm.org/D121260
Florian Hahn [Mon, 14 Mar 2022 19:39:30 +0000 (19:39 +0000)]
[LV] Check for users of truncated IVs, add more detailed comment.
Add missing outside user check for truncated IVs. Also hoist the code in
the helper with additional explanations.
Fixes #54370.
Amir Ayupov [Mon, 14 Mar 2022 19:27:21 +0000 (12:27 -0700)]
[X86] Fix cosmetic issues in instruction mnemonics
- Remove spurious } in invlpgb mnemonic
- Add \t between mnemonic and operands for ud1 instructions
Reviewed By: skan, craig.topper
Differential Revision: https://reviews.llvm.org/D121570
Jonas Devlieghere [Mon, 14 Mar 2022 19:14:16 +0000 (12:14 -0700)]
[lldb] Plumb process host architecture through platform selection
To allow us to select a different platform based on where the process is
running, plumb the process host architecture through platform selection.
This patch is in preparation for D121444 which needs this functionality
to tell apart iOS binaries running on Apple Silicon vs on a remote iOS
device.
Differential revision: https://reviews.llvm.org/D121484
Fangrui Song [Mon, 14 Mar 2022 19:00:15 +0000 (12:00 -0700)]
[ELF] Set the priority of STB_GNU_UNIQUE the same as STB_WEAK
In GCC -fgnu-unique output, STB_GNU_UNIQUE symbols are always defined relative
to a section in a COMDAT group. Currently `other` cannot be STB_GNU_UNIQUE for
valid input, so this patch is NFC.
If we switch to the model that ignores COMDAT resolution when performing symbol
resolution (D120626), this will fix bogus `relocation refers to a symbol in a
discarded section` errors when mixing -fno-gnu-unique objects with -fgnu-unique
objects.
Differential Revision: https://reviews.llvm.org/D120640
Ben Barham [Fri, 11 Mar 2022 22:37:06 +0000 (14:37 -0800)]
[Support] Add const to `FileError::getFileName`
`getFileName` returns a `StringRef`, there's no reason it shouldn't be
const.
Differential Revision: https://reviews.llvm.org/D121495
Ben Barham [Fri, 11 Mar 2022 22:31:21 +0000 (14:31 -0800)]
[VFS] Rename `RedirectingFileSystem::dump` to `print`
The rest of LLVM uses `print` for the method taking the `raw_ostream`
and `dump` only for the method with no parameters. Use the same for
`RedirectingFileSystem`.
Differential Revision: https://reviews.llvm.org/D121494
Kazu Hirata [Mon, 14 Mar 2022 18:36:58 +0000 (11:36 -0700)]
[CodeGen] Remove an unused variable introduced in D121128
Fangrui Song [Mon, 14 Mar 2022 18:38:04 +0000 (11:38 -0700)]
[Support] Change zlib::compress to return void
With a sufficiently large output buffer, the only failure is Z_MEM_ERROR.
Check it and call the noreturn report_bad_alloc_error if applicable.
resize_for_overwrite may call report_bad_alloc_error as well.
Now that there is no other error type, we can replace the return type with void
and simplify call sites.
Reviewed By: ikudrin
Differential Revision: https://reviews.llvm.org/D121512
Jonas Devlieghere [Mon, 14 Mar 2022 18:22:12 +0000 (11:22 -0700)]
[lldb] Fix the Windows build after D121536
Peter Klausler [Wed, 9 Mar 2022 21:43:54 +0000 (13:43 -0800)]
[flang] LBOUND() edge case: empty dimension
LBOUND must return 1 for an empty dimension, no matter what
explicit expression might appear in a declaration or arrive in
a descriptor.
Differential Revision: https://reviews.llvm.org/D121488
Eric Schweitz [Wed, 2 Feb 2022 20:46:21 +0000 (12:46 -0800)]
Write a pass to annotate constant operands on FIR ops. This works
around the feature in MLIR's canonicalizer, which considers the semantics
of constants differently based on how they are packaged and not their
values and use. Add test.
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D121625
Andrzej Warzynski [Mon, 14 Mar 2022 17:59:55 +0000 (17:59 +0000)]
Revert "[flang] IEEE_ARITHMETIC must imply USE IEEE_EXCEPTIONS"
This reverts commit
b6a7600491d755f6f700fe245b2fc7b928264c58. It caused
the following build failure:
```
ninja: error: dependency cycle: include/flang/__fortran_ieee_exceptions.mod -> include/flang/__fortran_ieee_exceptions.mod
```
See e.g.:
* https://lab.llvm.org/buildbot/#/builders/172/builds/9595
To reproduce:
```
cmake -G Ninja \
-DLLVM_TARGETS_TO_BUILD=host \
-DCMAKE_BUILD_TYPE=Release \
-DLLVM_ENABLE_PROJECTS="clang;flang" \
../../llvm
ninja check-flang
```
Fraser Cormack [Fri, 27 Nov 2020 14:29:58 +0000 (14:29 +0000)]
[RISCV] Add MIR tests exposing missed InstAliases
The InstAlias framework cannot match registers against zero_reg, which
RVV uses to encode unmasked operations.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D92228
Louis Dionne [Mon, 14 Mar 2022 15:36:21 +0000 (11:36 -0400)]
[libc++][NFC] Remove several redundant #if _LIBCPP_STD_VER > 17 in <span>
It turns out that the whole header is only enabled in C++20 and above,
so these checks were redundant (and always true).
Differential Revision: https://reviews.llvm.org/D121604
Mircea Trofin [Mon, 7 Mar 2022 16:23:05 +0000 (08:23 -0800)]
[regalloc] Remove -consider-local-interval-cost
Discussed extensively on D98232. The functionality introduced in D35816
never worked correctly. In D98232, it was fixed, but, as it was
introducing a large compile-time regression, and the value of the
original patch was called into doubt, we disabled it by default
everywhere. A year later, it appears that caused no grief, so it seems
safe to remove the disabled code.
This should be accompanied by re-opening bug 26810.
Differential Revision: https://reviews.llvm.org/D121128
Peter Klausler [Fri, 11 Mar 2022 18:17:04 +0000 (10:17 -0800)]
[flang] IEEE_ARITHMETIC must imply USE IEEE_EXCEPTIONS
The intrinsic module IEEE_ARITHMETIC must incorporate the public
names from the intrisic module IEEE_EXCEPTIONS. Rename IEEE_EXCEPTIONS
to __Fortran_ieee_exceptions so that it won't clash with the
nonintrinsic namespace, establish a new intrinic IEEE_EXCEPTIONS
module that USEs it, and add a USE to IEEE_ARITHMETIC.
Differential Revision: https://reviews.llvm.org/D121490
Jonas Devlieghere [Mon, 14 Mar 2022 17:22:04 +0000 (10:22 -0700)]
[lldb] Use the debugger's asynchronous output stream for progress events.
Use the debugger's asynchronous output stream for printing progress
events. This allows the active IOHandler to be in charge of printing
them and doing the necessary synchronization.
Differential revision: https://reviews.llvm.org/D121502
zhongyunde [Mon, 14 Mar 2022 17:19:39 +0000 (01:19 +0800)]
[AArch64] Perform last active true vector combine
Test bit of lane EC-1 can use P register directly, eg:
Materialize : Idx = (add (mul vscale, NumEls), -1)
i1 = extract_vector_elt t37, Constant:i64<Idx>
... into: "ptrue p, all" + PTEST
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D121180
Stanislav Gatev [Fri, 11 Mar 2022 11:52:53 +0000 (11:52 +0000)]
[clang][dataflow] Add support for nested composite bool expressions
This is part of the implementation of the dataflow analysis framework.
See "[RFC] A dataflow analysis framework for Clang AST" on cfe-dev.
Differential Revision: https://reviews.llvm.org/D121455
Sam Clegg [Sun, 13 Mar 2022 05:11:56 +0000 (21:11 -0800)]
[WebAssembly] Remove unused method from WebAssemblyTargetStreamer. NFC
Differential Revision: https://reviews.llvm.org/D121538
Teresa Johnson [Thu, 10 Mar 2022 22:17:53 +0000 (14:17 -0800)]
[WPD] Extend checking mode to support fallback to indirect call
Extend -wholeprogramdevirt-check to support both the existing
trapping mode on an incorrect devirtualization, as well as a new
mode to fallback to an indirect call on a mismatch. The new mode is
The new mode is useful in cases where we want to enable
devirtualization but cannot fully guarantee whole program visibility
(e.g in the case where LTO has been disabled for a small set of objects
that could potentially override virtual methods without having a symbol
reference to anything in the base class including the vtable).
Remove !prof and !callees metadata (which are used by indirect call
promotion) from both the new direct call and the fallback indirect call
(so that we don't perform another round of promotion on the latter).
Also remove it from the direct call in the non-fallback cases, which
was an oversight, although it didn't seem to cause any issues. Add tests
for the metadata removal covering the various cases.
Differential Revision: https://reviews.llvm.org/D121419
Valentin Clement [Mon, 14 Mar 2022 17:15:16 +0000 (18:15 +0100)]
[flang] Lower format statement
This patch lowers the format statement.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121611
Jonas Devlieghere [Mon, 14 Mar 2022 16:47:39 +0000 (09:47 -0700)]
[llvm] Fix warning: missing submodule 'LLVM_Analysis.ScalarFuncs'
Jonas Devlieghere [Mon, 14 Mar 2022 16:26:39 +0000 (09:26 -0700)]
[lldb] Use the IOHandler's stream instead of the debugger's in PrintAsync
PrintAsync is relying on the IOHandler to print to the output/error
stream. In that context it doesn't make much sense that this is using
the debugger's streams rather than the one from the IOHandler.
Differential revision: https://reviews.llvm.org/D121536
Arthur Eubanks [Thu, 24 Feb 2022 23:51:00 +0000 (15:51 -0800)]
[OpaquePtr][AArch64] Use elementtype on ldxr/stxr
Includes verifier changes checking the elementtype, clang codegen
changes to emit the elementtype, and ISel changes using the elementtype.
Reviewed By: #opaque-pointers, nikic
Differential Revision: https://reviews.llvm.org/D120527
Valentin Clement [Mon, 14 Mar 2022 17:03:50 +0000 (18:03 +0100)]
[flang] Lower common block
This patch lowers common block variable to FIR.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121610
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Andrew Browne [Fri, 11 Mar 2022 02:58:17 +0000 (18:58 -0800)]
[DFSan] Remove use of setarch in dfsan test.
Use of setarch Was added by
https://github.com/llvm/llvm-project/commit/
f93c2b64ed381bc4bc52f4b8539026f1597f574d
Running the test now it doesn't seem necessary because:
1) Explicitly only x86_64 is supported for dfsan.
2) https://reviews.llvm.org/D111522 makes it less flakey.
Differential Revision: https://reviews.llvm.org/D121439
Andrew Litteken [Wed, 9 Mar 2022 18:33:33 +0000 (10:33 -0800)]
[IROutliner] Avoid reusing PHINodes that have already been matched when merging outlined functions' phi node blocks
When there are two external phi nodes for two different outlined regions, when compressing the created phi nodes between the two regions, the matching for the second phi node in the second region matches the first phi node created for the first region rather than the second phi node created for the first region. This adds an extra output path where there should not be one.
The fix is the ignore phi nodes that have already been matched for each region.
Reviewer: paquette
Differential Revision: https://reviews.llvm.org/D121312
LLVM GN Syncbot [Mon, 14 Mar 2022 16:52:53 +0000 (16:52 +0000)]
[gn build] Port
5e65e79bace6
Valentin Clement [Mon, 14 Mar 2022 16:40:47 +0000 (17:40 +0100)]
[flang] Lower all intrinsic
Lower the `all` intrinsic procedure.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121607
Co-authored-by: mleair <leairmark@gmail.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Jonas Devlieghere [Mon, 14 Mar 2022 16:40:24 +0000 (09:40 -0700)]
Fix the implicit module build (2/2)
This fixes the implicit module build after
b1b4b6f36695 broke the LLDB
build: https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/42084/
My previous commit didn't have the modulemap change staged.
Arthur Eubanks [Sat, 5 Mar 2022 23:06:08 +0000 (15:06 -0800)]
[NewPM] Actually recompute GlobalsAA before module optimization pipeline
RequireAnalysis<GlobalsAA> doesn't actually recompute GlobalsAA.
GlobalsAA isn't invalidated (unless specifically invalidated) because
it's self-updating via ValueHandles, but can be imprecise during the
self-updates.
Rather than invalidating GlobalsAA, which would invalidate AAManager and
any analyses that use AAManager, create a new pass that recomputes
GlobalsAA.
Fixes #53131.
Differential Revision: https://reviews.llvm.org/D121167
Arthur Eubanks [Sat, 5 Mar 2022 23:20:24 +0000 (15:20 -0800)]
[test] Add phase ordering test
Valentin Clement [Mon, 14 Mar 2022 16:38:47 +0000 (17:38 +0100)]
[flang] Lower elemental calls
This patch adds more lowering of operations sub-expression inside elemental call arguments.
It tests array contexts where an address is needed for each element (for
the argument), but part of the array sub-expression must be lowered by value
(for the operation)
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121606
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Arthur Eubanks [Thu, 24 Feb 2022 23:30:23 +0000 (15:30 -0800)]
[ValueTracking] Simplify llvm::isPointerOffset()
We still need the code after stripAndAccumulateConstantOffsets() since
it doesn't handle GEPs of scalable types and non-constant but identical
indexes.
Differential Revision: https://reviews.llvm.org/D120523
serge-sans-paille [Mon, 14 Mar 2022 16:28:38 +0000 (17:28 +0100)]
Fix issing header on z/OS
Bug introduced in
fbbc41f8dd23
Stella Stamenova [Mon, 14 Mar 2022 16:26:41 +0000 (09:26 -0700)]
[lldb] Skip Test11588 on Windows
The test was marked as XFAIL on Windows, but recent changes have made it flaky instead
Jonas Devlieghere [Mon, 14 Mar 2022 16:22:30 +0000 (09:22 -0700)]
Fix the implicit module build
This fixes the implicit module build after
b1b4b6f36695 broke the LLDB
build: https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/42084/
Jonas Devlieghere [Mon, 14 Mar 2022 16:01:53 +0000 (09:01 -0700)]
[lldb] Move ProgressEventData out of debugger and into its own file (NFC)
Move ProgressEventData out of debugger and into its own file. This is in
preparation of adding a few new type of event data for diagnostics.
Differential revision: https://reviews.llvm.org/D121506
Nikita Popov [Mon, 14 Mar 2022 16:14:16 +0000 (17:14 +0100)]
[TLI] Check that malloc argument has type size_t
DSE assumes that this is the case when forming a calloc from a
malloc + memset pair.
For tests, either update the malloc signature or change the
data layout.
Matthias Springer [Mon, 14 Mar 2022 16:05:32 +0000 (01:05 +0900)]
[mlir][vector] Implement unrolling of ReductionOp
Differential Revision: https://reviews.llvm.org/D121597
Sam Clegg [Mon, 14 Mar 2022 16:07:48 +0000 (09:07 -0700)]
[WebAssembly] Rename member in WasmYAML.h to avoid compiler warning
Followup/fix for https://reviews.llvm.org/D121349.
Keith Smiley [Sun, 13 Mar 2022 17:02:02 +0000 (10:02 -0700)]
clang-tidy: discover binaries in build dir
This changes the clang-tidy script to discover binaries you've built
locally without having to pass them.
Differential Revision: https://reviews.llvm.org/D100692
Sam Clegg [Thu, 10 Mar 2022 02:11:44 +0000 (18:11 -0800)]
[WebAssembly] Second phase of implemented extended const proposal
This change continues to lay the ground work for supporting extended
const expressions in the linker.
The included test covers object file reading and writing and the YAML
representation.
Differential Revision: https://reviews.llvm.org/D121349
Nikita Popov [Mon, 14 Mar 2022 15:54:07 +0000 (16:54 +0100)]
[SLPVectorizer] Handle external load/store pointer uses with opaque pointers
In this case we may not generate a bitcast, so the new load/store
becomes the external user.
Ayush Sahay [Mon, 14 Mar 2022 15:25:13 +0000 (20:55 +0530)]
[lldb] Require native for command-thread-siginfo.test
command-thread-siginfo.test employs a subject with a call to wait, and
thus requires system-linux. However, it's possible to target non-Linux
platforms despite operating on Linux hosts. So, have it require native
too.
Reviewed By: mgorny, labath
Differential Revision: https://reviews.llvm.org/D121487
Florian Hahn [Mon, 14 Mar 2022 15:47:40 +0000 (15:47 +0000)]
[LV] Remove dead Loop argument from emitMinimumVector... (NFC)
The argument is not used, remove it.
Michael Kruse [Mon, 14 Mar 2022 13:39:25 +0000 (08:39 -0500)]
[polly] Introduce -polly-print-* passes to replace -analyze.
The `opt -analyze` option only works with the legacy pass manager and might be removed in the future, as explained in llvm.org/PR53733. This patch introduced -polly-print-* passes that print what the pass would print with the `-analyze` option and replaces all uses of `-analyze` in the regression tests.
There are two exceptions: `CodeGen\single_loop_param_less_equal.ll` and `CodeGen\loop_with_condition_nested.ll` use `-analyze on the `-loops` pass which is not part of Polly.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D120782
Thomas Raoux [Wed, 9 Mar 2022 00:08:11 +0000 (00:08 +0000)]
[mlir][vector] Add unrolling pattern for multidim_reduce op
Implement the vectorLoopUnroll interface for MultiDimReduceOp and add a
pattern to do the unrolling following the same interface other vector
unroll patterns.
Differential Revision: https://reviews.llvm.org/D121263
Nikita Popov [Mon, 14 Mar 2022 14:36:32 +0000 (15:36 +0100)]
[Verifier] Verify llvm.access.group metadata
According to LangRef, an access scope must have zero operands and
be distinct. The access group may either be a single access scope
or a list of access scopes.
LoopInfo may assert if this is not the case.
Nemanja Ivanovic [Fri, 11 Mar 2022 17:19:20 +0000 (11:19 -0600)]
[PowerPC] Add missed VSX shuffles instead of Altivec ones
VSX introduced some permute instructions that are direct
replacements for Altivec ones except they can target all
the VSX registers. We have added code generation for most
of these but somehow missed the low/hi word merges (XXMRG[LH]W).
This caused some additional spills on some large
computationally intensive code.
This patch simply adds the missed patterns.
Marek Kurdej [Mon, 14 Mar 2022 15:01:24 +0000 (16:01 +0100)]
Revert "[clang-format] Correctly format variable templates."
This reverts commit
a140b7104fdae0d9eff5b18efbc784754e0ca274.
It provoked the bug https://github.com/llvm/llvm-project/issues/54374.
Simon Moll [Mon, 14 Mar 2022 13:32:19 +0000 (14:32 +0100)]
[VE] v256f32|64 fma isel
llvm.fma|fmuladd vp.fma isel and tests
Reviewed By: kaz7
Differential Revision: https://reviews.llvm.org/D121477
Nico Weber [Sun, 13 Mar 2022 14:35:38 +0000 (10:35 -0400)]
[lld-link] Tweak winsysroottest.test to have passing links on happy path
Previously, the test checked for a "undefined symbol" error
(instead of the "could not open std*.lib" which would happen without
the flag).
Instead, use /entry: so that the link succeeds.
No behavior change, but maybe makes the test a bit easier to understand.
Differential Revision: https://reviews.llvm.org/D121553
Aaron Ballman [Mon, 14 Mar 2022 14:13:01 +0000 (10:13 -0400)]
Silence -Wlogical-op-parentheses and fix a logic bug while doing so
Tue Ly [Mon, 14 Mar 2022 14:04:39 +0000 (10:04 -0400)]
[libc] Include -150 to the special cases at the beginning of exp2f function.
Nikita Popov [Mon, 14 Mar 2022 13:34:01 +0000 (14:34 +0100)]
[SCCP][IR] Landing pads are not safe to remove
For landingpads with {} type, SCCP ended up dropping them, because
we considered them as safe to remove.
Tue Ly [Fri, 11 Mar 2022 15:08:47 +0000 (10:08 -0500)]
[libc] Implement exp2f function that is correctly rounded for all rounding modes.
Implement exp2f function that is correctly rounded for all rounding modes.
Reviewed By: sivachandra, zimmermann6
Differential Revision: https://reviews.llvm.org/D121463
Aaron Ballman [Mon, 14 Mar 2022 13:22:37 +0000 (09:22 -0400)]
Implement literal suffixes for _BitInt
WG14 adopted N2775 (http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2775.pdf)
at our Feb 2022 meeting. This paper adds a literal suffix for
bit-precise types that automatically sizes the bit-precise type to be
the smallest possible legal _BitInt type that can represent the literal
value. The suffix chosen is wb (for a signed bit-precise type) which
can be combined with the u suffix (for an unsigned bit-precise type).
The preprocessor continues to operate as-if all integer types were
intmax_t/uintmax_t, including bit-precise integer types. It is a
constraint violation if the bit-precise literal is too large to fit
within that type in the context of the preprocessor (when still using
a pp-number preprocessing token), but it is not a constraint violation
in other circumstances. This allows you to make bit-precise integer
literals that are wider than what the preprocessor currently supports
in order to initialize variables, etc.
Erich Keane [Thu, 10 Mar 2022 21:31:52 +0000 (13:31 -0800)]
Have cpu-specific variants set 'tune-cpu' as an optimization hint
Due to various implementation constraints, despite the programmer
choosing a 'processor' cpu_dispatch/cpu_specific needs to use the
'feature' list of a processor to identify it. This results in the
identified processor in source-code not being propogated to the
optimizer, and thus, not able to be tuned for.
This patch changes to use the actual cpu as written for tune-cpu so that
opt can make decisions based on the cpu-as-spelled, which should better
match the behavior expected by the programmer.
Note that the 'valid' list of processors for x86 is in
llvm/include/llvm/Support/X86TargetParser.def. At the moment, this list
contains only Intel processors, but other vendors may wish to add their
own entries as 'alias'es (or with different feature lists!).
If this is not done, there is two potential performance issues with the
patch, but I believe them to be worth it in light of the improvements to
behavior and performance.
1- In the event that the user spelled "ProcessorB", but we only have the
features available to test for "ProcessorA" (where A is B minus
features),
AND there is an optimization opportunity for "B" that negatively affects
"A", the optimizer will likely choose to do so.
2- In the event that the user spelled VendorI's processor, and the
feature
list allows it to run on VendorA's processor of similar features, AND
there
is an optimization opportunity for VendorIs that negatively affects
"A"s,
the optimizer will likely choose to do so. This can be fixed by adding
an
alias to X86TargetParser.def.
Differential Revision: https://reviews.llvm.org/D121410
Matthias Springer [Mon, 14 Mar 2022 13:06:11 +0000 (22:06 +0900)]
[mlir][bufferization] Update public MLIR documentation
Differential Revision: https://reviews.llvm.org/D121071
Florian Hahn [Mon, 14 Mar 2022 13:00:02 +0000 (13:00 +0000)]
[LV] Remove dead Loop argument from emitSCEVChecks. (NFC)
The argument is not used, remove it.
Nikita Popov [Mon, 14 Mar 2022 12:52:59 +0000 (13:52 +0100)]
[CoroSplit] Avoid self-replacement
With opaque pointers, the bitcast might be a no-op, and this can
end up trying to replace a value with itself, which is illegal.
Egor Zhdan [Fri, 11 Mar 2022 15:57:01 +0000 (15:57 +0000)]
[Clang][Sema] Avoid crashing for `__builtin_memcpy_inline` with an array argument
This change teaches the Sema logic for `__builtin_memcpy_inline` to implicitly convert arrays passed as arguments to pointers, similarly to regular `memcpy`.
This code will no longer cause a compiler crash:
```
void f(char *p) {
char s[1] = {0};
__builtin_memcpy_inline(p, s, 1);
}
```
rdar://
88147527
Differential Revision: https://reviews.llvm.org/D121475
Florian Hahn [Mon, 14 Mar 2022 12:21:20 +0000 (12:21 +0000)]
[LV] Do not set insert point in completeLoopSkeleton. (NFCI)
The insertion point for the builder used during VPlan code generation is
set during code generation. Setting the insert point here is dead code
and can be removed.
Nikita Popov [Mon, 14 Mar 2022 12:03:04 +0000 (13:03 +0100)]
[DeadArgElim] Guard against function type mismatch
If the call function type and function type don't match, we should
consider the function live (there is effectively a bitcast
sitting in between).
Nikita Popov [Mon, 14 Mar 2022 11:46:10 +0000 (12:46 +0100)]
[GVN] Check load type in select PRE
This is no longer implicitly guaranteed with opaque pointers.
Björn Schäpers [Sun, 13 Mar 2022 20:25:11 +0000 (21:25 +0100)]
[clang-format] Fix crash on asm block with label
Fixes https://github.com/llvm/llvm-project/issues/54349
Differential Revision: https://reviews.llvm.org/D121559
Björn Schäpers [Sun, 13 Mar 2022 20:07:01 +0000 (21:07 +0100)]
[clang-format][NFC] Left renamed to OpeningBrace...
in TokenAnnotator::parseBrace. Left is misleading, because we have a
loop and Left does not move.
Also return early.
Differential Revision: https://reviews.llvm.org/D121558
Björn Schäpers [Sun, 13 Mar 2022 10:55:24 +0000 (11:55 +0100)]
[clang-format] Fix crash on invalid requires expression
Fixes https://github.com/llvm/llvm-project/issues/54350
Differential Revision: https://reviews.llvm.org/D121550
Björn Schäpers [Sun, 13 Mar 2022 20:49:15 +0000 (21:49 +0100)]
[clang-format][NFC] Rename Left to OpeningParen...
in TokenAnnotator::parseParens(). Left is misleading since we have a
loop and Left is not adjusted.
Differential Revision: https://reviews.llvm.org/D121557
Benoit Jacob [Mon, 14 Mar 2022 10:07:33 +0000 (11:07 +0100)]
Expose ScalarizerPass options to C++ (not just commandline)
Context: I needed this for https://github.com/google/iree/pull/8474 .
I found that TSan instrumentation expects vector sizes to be <= 16,
and in my project (IREE) we have tests with higher vector sizes.
That left some test functions uninstrumented, resulting in crashes as
instrumented code called into them.
Differential Revision: https://reviews.llvm.org/D121182
Marek Kurdej [Mon, 14 Mar 2022 10:59:41 +0000 (11:59 +0100)]
[clang-format] Clean up UnwrappedLineParser::parseRecord. NFC.
Simon Pilgrim [Mon, 14 Mar 2022 10:57:17 +0000 (10:57 +0000)]
[X86] Update AVX512 VBMI2 VL intrinsic tests to avoid adds
As noticed in D119654, by adding the masked intrinsics results together we can end up with the selects being canonicalized away from the intrinsic - this isn't what we want to test here so replace with a insertvalue chain into a aggregate instead to retain all the results.
gysit [Mon, 14 Mar 2022 10:45:04 +0000 (10:45 +0000)]
[mlir][linalg] Replace linalg.fill by OpDSL variant.
The revision removes the linalg.fill operation and renames the OpDSL generated linalg.fill_tensor operation to replace it. After the change, all named structured operations are defined via OpDSL and there are no handwritten operations left.
A side-effect of the change is that the pretty printed form changes from:
```
%1 = linalg.fill(%cst, %0) : f32, tensor<?x?xf32> -> tensor<?x?xf32>
```
changes to
```
%1 = linalg.fill ins(%cst : f32) outs(%0 : tensor<?x?xf32>) -> tensor<?x?xf32>
```
Additionally, the builder signature now takes input and output value ranges as it is the case for all other OpDSL operations:
```
rewriter.create<linalg::FillOp>(loc, val, output)
```
changes to
```
rewriter.create<linalg::FillOp>(loc, ValueRange{val}, ValueRange{output})
```
All other changes remain minimal. In particular, the canonicalization patterns are the same and the `value()`, `output()`, and `result()` methods are now implemented by the FillOpInterface.
Depends On D120726
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D120728
Valentin Clement [Mon, 14 Mar 2022 10:49:50 +0000 (11:49 +0100)]
[flang][NFC] Use TODO with location
Use the TODO macro in `flang/Lower/Todo.h` with the converter location.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D121582
Kazushi (Jam) Marukawa [Sat, 12 Mar 2022 02:21:51 +0000 (11:21 +0900)]
[VE] Support more intrinsics
Support new intrinsics for following instrauctions.
- VLDZ, VPCNT, VBRV
- LCR, SCR, TSCR, FIDCR
- FENCE
Also clean the intrinsics implementation of a following instruction.
- SVOB
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D121509
Simon Moll [Mon, 14 Mar 2022 09:39:30 +0000 (10:39 +0100)]
[VE] v256i32|64 reduction isel and tests
and|add|or|xor|smax v256i32|64 isel and tests for vp and vector.reduce
intrinsics
Reviewed By: kaz7
Differential Revision: https://reviews.llvm.org/D121469
Valentin Clement [Mon, 14 Mar 2022 09:41:50 +0000 (10:41 +0100)]
[flang][NFC] Add todo in CallInterface
Add a todo for assumed shape dummy argument with VALUE attribute
since this is not implemented yet.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D121581