Bing1 Yu [Mon, 10 Apr 2023 09:09:35 +0000 (17:09 +0800)]
[X86][AMX] Add check for potentially dereferencing null return value
Reviewed By: xiangzhangllvm
Differential Revision: https://reviews.llvm.org/D147921
Shengchen Kan [Tue, 11 Apr 2023 04:32:37 +0000 (12:32 +0800)]
[X86][mem-fold] Only run x86-fold-tables.test when target x86 is built
Avoid the LIT fail when only arm/aarch64 is built.
Alex Brachet [Tue, 11 Apr 2023 04:41:14 +0000 (04:41 +0000)]
[libc] Correctly add new files
Alex Brachet [Tue, 11 Apr 2023 04:30:50 +0000 (04:30 +0000)]
Reland "[libc] Add {,r}index"
Differential Revision: https://reviews.llvm.org/D147464
Owen Shepherd [Tue, 11 Apr 2023 04:11:48 +0000 (04:11 +0000)]
[ORC] Link `atomic` on platforms without hardware support for atomics.
Fixes error about missing reference to __atomic_fetch_add_8 when linking
LLVMOrcJit on ARMv6.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D147937
Shengchen Kan [Tue, 11 Apr 2023 02:00:18 +0000 (10:00 +0800)]
[X86][mem-fold] Speed up test by not re-generating the .inc file
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D147835
Joshua Cao [Sun, 9 Apr 2023 01:12:46 +0000 (18:12 -0700)]
[SCEV] Strengthen huge constant trip multiples.
SCEV determines that loops with trip count >=2^32 have a trip multiple
of 1 to guard against huge multiples. This patch stregthens this to
instead find the greatest power of 2 divisor that is less than the
threshold.
Differential Revision: https://reviews.llvm.org/D147868
Joshua Cao [Sun, 9 Apr 2023 01:05:26 +0000 (18:05 -0700)]
[LoopUnroll] Add script test checks for LoopUnroll/X86/mmx.ll
Joshua Cao [Sun, 9 Apr 2023 00:31:43 +0000 (17:31 -0700)]
[SCEV][NFC] Convert check to assert getSmallConstantTripMultiple()
Joshua Cao [Wed, 29 Mar 2023 03:49:39 +0000 (20:49 -0700)]
[SCEV] When computing trip count, only zext if necessary
This patch improves on https://reviews.llvm.org/D110587. To summarize
the patch, given backedge-taken count BC, trip count TC is `BC + 1`.
However, we don't know if BC we might overflow. So the patch modifies TC
computation to `1 + zext(BC)`.
This patch only adds the zext if necessary by looking at the constant
range. If we can determine that BC cannot be the max value for its
bitwidth, then we know adding 1 will not overflow, and the zext is not
needed. We apply loop guards before computing TC to get more data.
The primary motivation is to support my work on more precise trip
multiples in https://reviews.llvm.org/D141823. For example:
```
void test(unsigned n)
__builtin_assume(n % 6 == 0);
for (unsigned i = 0; i < n; ++i)
foo();
```
Prior to this patch, we had `TC = 1 + zext(-1 + 6 * ((6 umax %n) /u
6))<nuw>`. SCEV range computation is able to determine that the BC
cannot be the max value, so the zext is not needed. The result is `TC
-> (6 * ((6 umax %n) /u 6))<nuw>`. From here, we would be able to
determine that %n is a multiple of 6.
There was one change in LoopCacheAnalysis/LoopInterchange required.
Before this patch, if a loop has BC = false, it would compute `TC -> 1 +
zext(false) -> 1`, which was fine. After this patch, it computes `TC -> 1
+ false = true`. CacheAnalysis would then sign extend the `true`, which
was not the intended the behavior. I modified CacheAnalysis such that
it would only zero extend trip counts.
This patch is not NFC, but also does not change any SCEV outputs. I
would like to get this patch out first to make work with trip multiples
easier.
Differential Revision: https://reviews.llvm.org/D147117
Blue Gaston [Mon, 10 Apr 2023 22:53:47 +0000 (15:53 -0700)]
[Sanitizer][Test] Disable atos test on unknown address for ubsan
This test is failing for some cases in CI for ubsan. I suspect because
of the undefined behavior nature of the test, ubsan does not always
try to symbolicate or save the frame for the faulting address.
Specifically is failing on x86 CI. But making this unsupported on
ubsan for now.
rdar://
107846128
Differential Revision: https://reviews.llvm.org/D147977
Youngsuk Kim [Tue, 11 Apr 2023 01:41:48 +0000 (01:41 +0000)]
[MCJIT] Run ExecutionEngine/MCJIT/remote tests only if threads enabled.
When LLVM_ENABLE_THREADS=Off, disables tests under
llvm/test/ExecutionEngine/MCJIT/remote/ that require threads.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D147803
Matthias Springer [Tue, 11 Apr 2023 01:19:19 +0000 (10:19 +0900)]
[mlir][linalg] Replace `getUpperBoundForIndex` implementation
Use `reifyValueBound` instead, which is more general not hard-coded to a specific list supported ops.
Also add a `closedUB` parameter to the ValueBoundsOpInterface API.
Differential Revision: https://reviews.llvm.org/D146356
Alex Brachet [Tue, 11 Apr 2023 01:21:39 +0000 (01:21 +0000)]
Revert "[libc] Add {,r}index"
This reverts commit
a0a141fcbe1dfd35032fa5c052e6906180a37fb1.
Kazu Hirata [Tue, 11 Apr 2023 01:23:07 +0000 (18:23 -0700)]
[Target] Use isNullConstant and isOneConstant (NFC)
Kazu Hirata [Tue, 11 Apr 2023 01:19:17 +0000 (18:19 -0700)]
Use APInt::getOneBitSet (NFC)
Jorge Gorbe Moya [Fri, 7 Apr 2023 20:34:16 +0000 (13:34 -0700)]
[lldb-vscode] Fix two issues with runInTerminal test.
With ptrace_scope = 1 the kernel only allows tracing descendants of a
process. When using runInTerminal, the target process is not launched
by the debugger, so we need to modify LaunchRunInTerminal to explicitly
allow tracing. This should fix a problem reported in
https://reviews.llvm.org/D84974#3903716
In order to allow only the main lldb-vscode process to attach to the
target, this change introduces a new `--debugger-pid` flag that needs
to be passed with `--launch-target` and `--comm-file`.
Also, remove a special case from the launch method in the
lldbvscode_testcase test harness. The existing test was using
stopOnEntry, so the first stop didn't happen at the expected breakpoint
unless the harness did configurationDone first.
Differential Revision: https://reviews.llvm.org/D147805
Alex Brachet [Tue, 11 Apr 2023 01:16:21 +0000 (01:16 +0000)]
[libc] Add {,r}index
Differential Revision: https://reviews.llvm.org/D147464
Alex Brachet [Tue, 11 Apr 2023 01:11:32 +0000 (01:11 +0000)]
[libc] Move str{,r}chr implementation to headers
Differential Revision: https://reviews.llvm.org/D147463
Matthias Springer [Tue, 11 Apr 2023 00:31:16 +0000 (09:31 +0900)]
[mlir][Interfaces] Better error handling for `addBound`
Do not assert if a bound could not be added. This may or may not cause an error. We have the appropriate error handling in `computeBound` (which returns a `LogicalResult`).
Differential Revision: https://reviews.llvm.org/D147769
Mehdi Amini [Tue, 11 Apr 2023 00:39:34 +0000 (18:39 -0600)]
Add tracing for pattern application in a ApplyPatternAction
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D144816
Manish Gupta [Thu, 6 Apr 2023 23:58:17 +0000 (23:58 +0000)]
[Updated commit] Fix Transpose Check in MMA.SYNC Path.
Pushed a stale commit for the same review in my previous commit.
I am updating the main-line with the latest commit including
review commits. Apologies for the redundant commit.
Differential Revision: https://reviews.llvm.org/D147749
Manish Gupta [Thu, 6 Apr 2023 23:58:17 +0000 (23:58 +0000)]
Fix Transpose Check in MMA.SYNC Path
Differential Revision: https://reviews.llvm.org/D147749
Valentin Clement [Tue, 11 Apr 2023 00:01:10 +0000 (17:01 -0700)]
[flang][NFC] Move TypeConverter.h header file to include dir
After the extraction of the TypeConverter, move the header files
to the include dir so the shared library build is fine.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D147979
Nathan James [Mon, 10 Apr 2023 23:38:37 +0000 (00:38 +0100)]
[clang-tidy][NFC] Remove redundant allOf matcher calls
Remove calls to the `allOf` matcher in all matchers that are derived from the `VariadicDynCastAllOfMatcher`.
Caslyn Tonelli [Mon, 10 Apr 2023 20:24:58 +0000 (20:24 +0000)]
[libc] Add swab implementation
Swab implementation is added to libc/src/unistd.
Differential Revision: https://reviews.llvm.org/D147970
Haowei Wu [Wed, 22 Mar 2023 23:41:47 +0000 (16:41 -0700)]
[ifs] Print an error when llvm-ifs is supplied with an unknown arch
This patch makes llvm-ifs print an error when the arch string from
'--arch' flag or the IFS file 'Arch' field is unknown.
Differential Revision: https://reviews.llvm.org/D146672
Craig Topper [Mon, 10 Apr 2023 20:36:52 +0000 (13:36 -0700)]
[TableGen] Pass TreePatternNodePtr to InlinePatternFragments by const reference. NFC
Not all paths through this function require the reference count
to be incremened. Pass by reference so we don't increase the count
unnecessarily.
Valentin Clement [Mon, 10 Apr 2023 21:29:58 +0000 (14:29 -0700)]
[flang][openacc] Add missing piece to translate to LLVM IR dialect
Add missing pieces to translate handle OpenACC dialect in the translation.
Depends on D147825
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D147828
Martin Storsjö [Wed, 5 Apr 2023 10:56:54 +0000 (13:56 +0300)]
[libcxxabi, libunwind] [test] Place output from tests under a 'test' subdir
Previously, all the output from the tests were placed directly in
the build directory. The tests produce a couple directories named
`__config_{exec,cache,src}__` which are easy to distinguish, and
the output from the individual tests were placed directly in a
directory named `Output`.
This is the same change as
736c6e246f5398331d83edd204a846cc967ad5c6, but for the libcxxabi
and libunwind test suites.
Differential Revision: https://reviews.llvm.org/D147628
Balaji V. Iyer [Mon, 10 Apr 2023 21:01:26 +0000 (21:01 +0000)]
[mlir][math] Expand math.floorf to truncate, compares and increments
Floorf are pushed directly to libm. This is problematic for
situations where libm is not available. This patch will break down
a floorf function to truncate followed by an increment for negative
values, if necessary.
Reviewed By: rsuderman
Differential Revision: https://reviews.llvm.org/D147966
Martin Storsjö [Fri, 31 Mar 2023 05:47:59 +0000 (08:47 +0300)]
[compiler-rt] [test] Generalize the triple regex for windows XFAILs
Don't hardcode the vendor field in the triples to "pc"; for mingw
environments the triple is often "<arch>-w64-windows-gnu".
Differential Revision: https://reviews.llvm.org/D147862
Martin Storsjö [Thu, 6 Apr 2023 10:11:46 +0000 (13:11 +0300)]
[compiler-rt] [test] [ubsan] Build .cpp files with %clangxx
This fixes linker errors in mingw configurations, where the ubsan
runtime needs the C++ standard library when instrumenting C++ code.
Differential Revision: https://reviews.llvm.org/D147861
Martin Storsjö [Wed, 5 Apr 2023 21:41:51 +0000 (21:41 +0000)]
[compiler-rt] [test] Use %clangxx for tests that use -x c++
When instrumenting C++ code, ubsan ends up referencing
the ubsan_type_hash_* object files, which require linking against
the C++ ABI library. When building with "clang -x c++", the code
is handled as C++, but the compiler still only links as if it was
C.
Change all cases of "%clang -x c++" into "%clangxx -x c++".
This fixes a lot of ubsan tests in mingw mode.
Differential Revision: https://reviews.llvm.org/D147687
Martin Storsjö [Thu, 6 Apr 2023 07:51:35 +0000 (07:51 +0000)]
[compiler-rt] [builtins] Don't use assembly floatundi*f on x86_64 mingw
The x86 assembly is entirely skipped for MSVC build configurations,
since the assembly uses GAS syntax (which MSVC proper can't
assemble, but clang-cl can). But for mingw configurations, the
assembly is used.
On x86_64 Windows, the general calling convention is entirely different
than on other platforms (passing arguments in different registers),
so we can't use this assembly for mingw targets (unless adapted to
handle that calling convention). Thus skip these few assembly routines
for Windows targets. (For i386, we can keep using the assembly
functions.)
This fixes the builtins tests that test these routines. They aren't
used in practice though, as LLVM expands code inline for doing such
conversions anyway.
Differential Revision: https://reviews.llvm.org/D147685
Martin Storsjö [Mon, 3 Apr 2023 12:52:46 +0000 (15:52 +0300)]
[libcxxabi] [test] Mark thread_local_destruction_order.pass.cpp unsupported for mingw targets
With current versions of mingw-w64, TLS destructors that are registered
while executing TLS destructors are lost and leaked.
The root cause does get fixed further upstream in mingw-w64 in
https://github.com/mingw-w64/mingw-w64/commit/
71eddccd746c56d9cde28bb5620d027d49259de9
though, but mark the test as unsupported for now. The marking can be
removed later when the version used in the CI runner has been updated
to include the fix.
Differential Revision: https://reviews.llvm.org/D147859
Martin Storsjö [Fri, 31 Mar 2023 21:52:05 +0000 (00:52 +0300)]
[libunwind] [test] Mark the signal_frame test as unsupported on Windows
Mark it as unsupported on x86_64, arm and aarch64. On i686, DWARF
is used as the default unwinding format, and there, the CFI
directives are supported.
Differential Revision: https://reviews.llvm.org/D147858
Martin Storsjö [Wed, 5 Apr 2023 10:42:22 +0000 (10:42 +0000)]
[libunwind] [SEH] Handle ExceptionContinueExecution in forced unwinding
This fixes the libcxxabi test force_unwind3.pass.cpp when run on native
Windows.
When unwinding past the main thread function into the system functions
that brought up the thread, we can hit functions whose personality
functions return ExceptionContinueExecution (instead of the regular
ExceptionContinueSearch). Interpret this as a signal to stop the
unwind.
Curiously, in this case, it does return ExceptionContinueSearch if
running within a debugger.
Differential Revision: https://reviews.llvm.org/D147739
Martin Storsjö [Tue, 4 Apr 2023 20:26:20 +0000 (20:26 +0000)]
[libunwind] [SEH] Sync LSDA and handler between unw_proc_info_t and DISPATCHER_CONTEXT
For normal C++ unwinding, we get _dispContext initialized by the
prepopulated DISPATCHER_CONTEXT in _GCC_specific_handler, which
we set with __unw_seh_set_disp_ctx.
When doing force unwinding, we step and populate the unw_proc_info_t
struct _info with getInfoFromSEH, but when we execute the handler
via the __libunwind_seh_personality wrapper function, we execute
the handler set in DISPATCHER_CONTEXT.
Whenever updating these fields in either _info or _dispContext,
sync them to the other one too.
This fixes one aspect of the libcxxabi force_unwind*.pass.cpp tests on
x86_64.
Differential Revision: https://reviews.llvm.org/D147637
Martin Storsjö [Wed, 5 Apr 2023 11:12:24 +0000 (14:12 +0300)]
[libunwind] Increase the external value of _LIBUNWIND_CURSOR_SIZE for SEH/x86_64
For x86_64 Windows targets (that use SEH), _LIBUNWIND_CURSOR_SIZE
is 204; this fixes corruption in test cases that include libunwind.h
without manually defining _LIBUNWIND_IS_NATIVE_ONLY.
If the libunwind.h header is included without defining
_LIBUNWIND_IS_NATIVE_ONLY (like in the libunwind test cases), the
sizes are set to accommodate the maximum possible cursors and
contexts.
(Alternatively, __libunwind_config.h should be changed to default
to native unwinding unless cross unwinding has been requested.
Cross unwinding isn't implemented as far as I know anyway.)
Differential Revision: https://reviews.llvm.org/D147634
Alex Langford [Sat, 8 Apr 2023 05:50:45 +0000 (22:50 -0700)]
[lldb][NFC] Update syntax description for language cplusplus demangle
Also added some tests because this is completely untested.
rdar://
107780577
Differential Revision: https://reviews.llvm.org/D147841
Valentin Clement [Mon, 10 Apr 2023 20:32:37 +0000 (13:32 -0700)]
[flang][openacc] Add data operands conversion from FIR
This patch revive an old PR attempt [1] to perform the
data operands conversion needed for translation to LLVMIR.
This is currently not supporting box/class type since they will
normally not reach this pass when the proposed change in this RFC [2]
are implemented.
[1] https://github.com/flang-compiler/f18-llvm-project/pull/915
[2] https://discourse.llvm.org/t/rfc-openacc-dialect-data-operation-improvements/69825/2
Depends on D147824
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D147825
Florian Hahn [Mon, 10 Apr 2023 20:25:11 +0000 (21:25 +0100)]
[LV] Re-use already computed runtime VF in fixFixedOrderRecurrence.
This was suggested as independent cleanup in D147472.
This removes a redundant runtime VF computation when using scalable
vectors.
Florian Hahn [Mon, 10 Apr 2023 20:11:40 +0000 (21:11 +0100)]
[LV] Turn check into assert in fixFixedOrderRecurrence (NFCI).
Suggested as independent cleanup in D147567. Either VF or UF need to be
> 1. Note that if the condition would be false, the code below would use
a nullptr and crash.
Florian Hahn [Mon, 10 Apr 2023 20:08:33 +0000 (21:08 +0100)]
[VPlan] Only create extracts for recurrence exits if there are live-outs.
Move the code to collect live-out earlier and only generate extracts for
exit values if there are any live-outs that use them.
Depends on D147472.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D147567
Valentin Clement [Mon, 10 Apr 2023 20:05:37 +0000 (13:05 -0700)]
Revert "[flang][openacc] Add data operands conversion from FIR"
This reverts commit
68bcd647c9c006b707bc9a675a874658cd085d13.
Valentin Clement [Mon, 10 Apr 2023 20:05:23 +0000 (13:05 -0700)]
Revert "[flang][openacc] Add missing piece to translate to LLVM IR dialect"
This reverts commit
03289dc7af257684a4278ce4c38b0b07ad749b71.
Carlos Galvez [Sat, 8 Apr 2023 17:43:34 +0000 (17:43 +0000)]
[clang-tidy] Support specifying checks as a list in the config file
Specifying checks as a string is convenient for quickly using
clang-tidy to run a handful of checks. However it is not
suitable for projects that have a long list of enabled or
disabled checks. It is specially troublesome in case one
wants to interleave comments with the checks, to explain
why they are enabled or disabled.
Currently this can be achieved via multiline strings in YAML,
but it's error-prone. For example, comments must end with a
comma for clang-tidy to continue processing the list of globs;
a missing comma will make clang-tidy silently ignore the rest
of the list.
Instead, enable passing a native YAML list to the "Checks"
option in the config file. The implementation is done such
that the old behavior is kept: a user can pass a string
or a list. We can consider deprecating passing the checks
as a string altogether in a future release, to simplify
the internal logic of the YAML parser.
Fixes https://github.com/llvm/llvm-project/issues/51428
Differential Revision: https://reviews.llvm.org/D147876
Mircea Trofin [Fri, 7 Apr 2023 17:36:26 +0000 (10:36 -0700)]
[mlgo][inl] Interactive mode: optionally tell the default decision
This helps training algorithms that may want to sometimes replicate the
default decision. The default decision is presented as an extra feature
called `inlining_default`. It's not normally exported to save
computation time.
This is only available in interactive mode.
Differential Revision: https://reviews.llvm.org/D147794
Valentin Clement [Mon, 10 Apr 2023 19:19:26 +0000 (12:19 -0700)]
[flang][NFC] Fix invalid op result access
Fix invalid op result access. This will trigger
assertion introduced in D147883.
Reviewed By: frgossen
Differential Revision: https://reviews.llvm.org/D147959
Valentin Clement [Mon, 10 Apr 2023 19:18:37 +0000 (12:18 -0700)]
[flang][openacc] Add missing piece to translate to LLVM IR dialect
Add missing pieces to translate handle OpenACC dialect in the translation.
Depends on D147825
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D147828
Valentin Clement [Mon, 10 Apr 2023 19:17:42 +0000 (12:17 -0700)]
[flang][openacc] Add data operands conversion from FIR
This patch revive an old PR attempt [1] to perform the
data operands conversion needed for translation to LLVMIR.
This is currently not supporting box/class type since they will
normally not reach this pass when the proposed change in this RFC [2]
are implemented.
[1] https://github.com/flang-compiler/f18-llvm-project/pull/915
[2] https://discourse.llvm.org/t/rfc-openacc-dialect-data-operation-improvements/69825/2
Depends on D147824
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D147825
Roy Jacobson [Mon, 10 Apr 2023 04:26:30 +0000 (07:26 +0300)]
[Clang] Fix cast to BigIntType in hasUniqueObjectRepresentations
A bad QualType cast caused us not to detect _BigInt types if they were wrapped inside sugar types like SubstTemplateTypeParm.
Fix https://github.com/llvm/llvm-project/issues/62019
Reviewed By: cjdb
Differential Revision: https://reviews.llvm.org/D147904
Aiden Grossman [Mon, 10 Apr 2023 19:08:27 +0000 (19:08 +0000)]
[Docs][llvm-exegesis] Add documentation for --use-dummy-perf-counters
This patch adds in documentation for the --use-dummy-perf-counters
option (introduced in D146301).
Reviewed By: kpdev42
Differential Revision: https://reviews.llvm.org/D147842
ziqingluo-90 [Mon, 10 Apr 2023 18:54:11 +0000 (11:54 -0700)]
[-Wunsafe-buffer-usage] Improve fix-its for local variable declarations with null pointer initializers
For a local pointer declaration of the form `T * p = 0` or `T * p = std::nullptr`,
Before this patch, we generate fix-its that convert the declaration to
`std::span<T> p{nullptr, <# placeholder #>}`, in cases where `p` is
used in some unsafe operations. This patch improves the fix-its to
result in a simpler form `std::span<T> p`. It gets rid of the
placeholder and keeps the result concise.
Reviewed by: NoQ (Artem Dergachev)
Differential Revision: https://reviews.llvm.org/D143680
Jordan Rupprecht [Mon, 10 Apr 2023 18:42:58 +0000 (11:42 -0700)]
[bazel][NFC] Replace @bazel_tools//src/conditions with equivalent @platforms rules
Followup to D147354. Using src/conditions is discouraged in favor of using rules from the `@platforms` repository directly.
This replaces three conditions:
* `@bazel_tools//src/conditions:windows` -> `@platforms//os:windows`
* `@bazel_tools//src/conditions:darwin` -> `@platforms//os:macos`
* `@bazel_tools//src/conditions:freebsd` -> `@platforms//os:freebsd`
`llvm/config.bzl` has a non-trivial OS+CPU selection config, so that is omitted from this patch. There is intentionally no equivalent for that in `@platforms` because every project will have their own opinions about what a platform is, and it is not feasible for the bazel selection list to include every possible combination. The recommended idiom there is for projects to define their own supported platforms list, e.g. in a separate BUILD file or platform mapping.
Reviewed By: GMNGeoffrey
Differential Revision: https://reviews.llvm.org/D147948
Craig Topper [Mon, 10 Apr 2023 18:29:01 +0000 (11:29 -0700)]
[TableGen] Simplify how commuted variants are generated in GenerateVariantsOf. NFC
We don't need to copy the ChildVariants vector into a new vector.
We're using std::move on every entry we copy so they clearly
aren't needed again. We can swap entries in the vector and reuse
it instead.
Nico Weber [Mon, 10 Apr 2023 18:10:59 +0000 (14:10 -0400)]
[gn build] Port
eb65912e41a1
Changpeng Fang [Mon, 10 Apr 2023 17:53:33 +0000 (10:53 -0700)]
AMDGPU: Created a subclass for the return address operand in the tail call return instruction
Summary:
This is to avoid using the callee saved registers for the return address
of the tail call return instruction.
Reviewers:
arsenm, cdevadas
Differential Revision:
https://reviews.llvm.org/D147096
Paul Kirth [Mon, 13 Feb 2023 18:18:33 +0000 (18:18 +0000)]
[clang][driver] Pass `-femulated-tls` through to the linker in LTO mode
Currently the driver does not propagate the `-f[no-]emulated-tls` flags
to the linker under LTO. This can be surprising when the platform
defaults differ from the flags being passed. A related discussion can be
found in https://reviews.llvm.org/D143619. While the focus there was
RISC-V support, the root cause was that setting `-femualted-tls` and
`-flto` when compiling with Clang resulted in missing symbols because
the platform defaults for Android differed from the flags being passed
to Clang.
This patch changes the Clang driver's behavior to pass the emulated-tls
flags through to the linker when compiling with LTO/ThinLTO.
Reviewed By: phosek, vit9696
Differential Revision: https://reviews.llvm.org/D147834
Nikolas Klauser [Sun, 19 Mar 2023 22:10:37 +0000 (23:10 +0100)]
[libc++] Move __errc to __system_error/errc.h
This file was added before we started granularizing the headers, but is essentially just a granularized header. This moves the header to the correct place.
Reviewed By: #libc, EricWF
Spies: libcxx-commits, arichardson, mikhail.ramalho
Differential Revision: https://reviews.llvm.org/D146395
Augusto Noronha [Fri, 7 Apr 2023 22:46:33 +0000 (15:46 -0700)]
[lldb] Implement SymbolFile::GetCompileOptions
Implement SymbolFile::GetCompileOptions, which returns a map from
compilation units to compilation arguments associated with that unit.
Differential Revision: https://reviews.llvm.org/D147748
Kazu Hirata [Mon, 10 Apr 2023 17:11:41 +0000 (10:11 -0700)]
Fix warnings
This patch fixes:
mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp:1334:51:
warning: suggest parentheses around ‘&&’ within ‘||’ [-Wparentheses]
lld/wasm/Writer.cpp:250:39: warning: suggest parentheses around ‘&&’
within ‘||’ [-Wparentheses]
Michael Jones [Mon, 10 Apr 2023 17:01:31 +0000 (10:01 -0700)]
[libc][obvious] fix new assert typo
Jeremy Kun [Mon, 10 Apr 2023 15:57:34 +0000 (08:57 -0700)]
Add support for math.ctlz in convert-math-to-funcs
This change adds a software implementation of the `math.ctlz` operation
and includes it in `--convert-math-to-funcs`.
This is my first change to MLIR, so please bear with me as I'm still learning
the idioms of the codebase.
The context for this change is that I have some larger scale project in which
I'd like to lower from a mix of MLIR dialects to CIRCT, but many of the CIRCT
passes don't support the `math` dialect.
I noticed the content of `convert-math-to-funcs` was limited entirely to
the `pow` functions, but otherwise provided the needed structure to implement
this feature with minimal changes.
Highlight of the changes:
- Add a dependence on the SCF dialect for this lower. I could have lowered
directly to cf, following the pow lowerings in the same pass, but I felt it
was not necessary given the existing support for lowering scf to cf.
- Generalize the DenseMap storing op implementations: modify the callback
function hashmap to be keyed by both OperationType (for me this effectively
means the name of the op being implemented in software) and the type
signature of the resulting function.
- Implement the ctlz function as a loop. I had researched a variety of
implementations that claimed to be more efficient (such as those based on a
de Bruijn sequence), but it seems to me that the simplest approach would make
it easier for later compiler optimizations to do a better (platform-aware)
job optimizing this than I could do by hand.
Questions I had for the reviewer:
- [edit: found mlir-cpu-runner and added two tests] What would I add to the filecheck invocation to actually run the resulting MLIR on a value and assert the output is correct? I have done this manually with the C implementation but I'm not confident my port to MLIR is correct.
- Should I add a test for a vectorized version of this lowering? I followed suit with the ` VecOpToScalarOp` but I admit I don't fully understand what it's doing.
Reviewed By: vzakhari
Differential Revision: https://reviews.llvm.org/D146261
Tue Ly [Fri, 7 Apr 2023 01:45:45 +0000 (21:45 -0400)]
[libc][math] Update range reduction step for logf and reduce its latency.
Simplify the range reduction steps by choosing the reduction constants
carefully so that the reduced arguments v = r*m_x - 1 and v^2 are exact in double
precision, even without FMA instructions, and -2^-8 <= v < 2^-7. This allows the
polynomial evaluations to be parallelized more efficiently.
Reviewed By: santoshn, zimmermann6
Differential Revision: https://reviews.llvm.org/D147755
Michael Jones [Fri, 7 Apr 2023 21:02:39 +0000 (14:02 -0700)]
[libc] clean up ctype negative handling
The ctype functions will sometimes be passed negative numbers, such as
EOF. Some of the previous implementations didn't handle these properly.
Additionally, the tests did not check any negative numbers. These
problems have been fixed.
This patch fixes https://github.com/llvm/llvm-project/issues/62000
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D147813
Aaron Ballman [Mon, 10 Apr 2023 16:53:50 +0000 (12:53 -0400)]
Fix wording in a release note; NFC
This change happened due to post-commit feedback on the review.
Frederik Gossen [Mon, 10 Apr 2023 16:30:27 +0000 (12:30 -0400)]
Fix invalid op result access in SparseTensorCodegen
This trigges an assertion introduced in https://reviews.llvm.org/D147883
Differential Revision: https://reviews.llvm.org/D147950
Jay Foad [Mon, 6 Mar 2023 16:33:49 +0000 (16:33 +0000)]
[AMDGPU] Fix AddedComplexity for s_buffer_load patterns. NFCI.
We set AddedComplexity = 100 for s_load patterns to prefer them over
global loads, but for s_buffer_load patterns there is no need to do
this and it was quietly overriding the AddedComplexity of each
individual GCNPat that is defined inside SMLoad_Pattern (but in practice
that did not appear to make any difference).
Differential Revision: https://reviews.llvm.org/D145396
Bing1 Yu [Mon, 10 Apr 2023 16:14:04 +0000 (00:14 +0800)]
Change dyn_cast to cast
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D147923
Spenser Bauman [Mon, 10 Apr 2023 15:55:54 +0000 (15:55 +0000)]
[mlir][tosa] Lowering of tosa.gather operations with dynamic dimensions
The existing TOSA->Linalg lowering of tosa.gather only supports gathers
with either a static shape or a single dynamic batch dimension.
This change extends support to arbitrary number of dynamic dimensions on
both the values and indices of the gather operation.
Reviewed By: rsuderman
Differential Revision: https://reviews.llvm.org/D147810
Valentin Clement [Mon, 10 Apr 2023 15:33:52 +0000 (08:33 -0700)]
[flang][NFC] Extract TypeConverter from header file
Extract the type converter definitions into its own .cpp
file so the type converter can be used by passes.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D147824
Corentin Jabot [Mon, 10 Apr 2023 09:52:30 +0000 (11:52 +0200)]
[Clang] Check type-constraints applied to placeholder types
(and deduced return types)
Previously, only type constraints applied to type parameter
were semantically checked.
A diagnostic would still be emitted on instantiation, but it was
too late, lacked clarity, and was inconsistent with type parameters.
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D147925
Alvin Wong [Sun, 2 Apr 2023 18:26:30 +0000 (02:26 +0800)]
[asan][test] Fix two test cases on Windows
Fixes global-location-nodebug and heavy_uar_test for both MSVC and MinGW
targets.
Differential Revision: https://reviews.llvm.org/D147445
Alvin Wong [Sun, 2 Apr 2023 15:19:23 +0000 (23:19 +0800)]
[sanitizer][win] Fix `Atexit()` on MinGW asan_dynamic runtime
Some functions of asan depends on `Atexit()` handlers. On Windows, this
is implemented in
ad3ec82bb1c7217b0187a1e16bb22642e194ce94 to queue the
handlers in a vector then register them with `atexit()` only after the
CRT is fully initialized. However, this is broken on MinGW with
asan_dynamic runtime due to different initialization order. This change
fixes the issue by making sure that `Atexit()` can call `atexit()`
directly past the pre-initialization phase.
This fixes two asan test cases on MinGW.
Differential Revision: https://reviews.llvm.org/D147413
Piotr Zegar [Mon, 10 Apr 2023 14:41:32 +0000 (14:41 +0000)]
[clang-tidy] Added IgnoreVirtual option to misc-unused-parameters
Added option to ignore virtual methods in this check.
This allows to quickly get rid of big number of
potentialy false-positive issues without inserting
not-needed comments.
Fixes #55665.
Reviewed By: njames93
Differential Revision: https://reviews.llvm.org/D147918
Anna Thomas [Thu, 6 Apr 2023 18:42:14 +0000 (14:42 -0400)]
[LoopPredication] Fix where we generate widened condition. PR61963
Loop predication's predicateLoopExit pass does two incorrect things:
It sinks the widenable call into the loop, thereby converting an invariant condition to a variant one
It widens the widenable call at a branch thereby converting the branch into a loop-varying one.
The latter is problematic when the branch may have been loop-invariant
and prior optimizations (such as indvars) may have relied on this
fact, and updated the deopt state accordingly.
Now, when we widen this with a loop-varying condition, the deopt state
is no longer correct.
https://github.com/llvm/llvm-project/issues/61963 fixed.
Differential Revision: https://reviews.llvm.org/D147662
Anna Thomas [Mon, 10 Apr 2023 14:28:19 +0000 (10:28 -0400)]
Simplify test with deopt state in D147662. NFC
Jake Egan [Mon, 10 Apr 2023 13:43:31 +0000 (09:43 -0400)]
[NFC][tests] Disable new gmodules-preferred-name-* tests on AIX and z/OS
Objective-C is not supported on AIX and z/OS. Precedent here: D109060
skc7 [Mon, 10 Apr 2023 13:07:17 +0000 (18:37 +0530)]
[AMDGPU][NFC] Regenerate test checks for merge-tbuffer.mir
Aaron Ballman [Mon, 10 Apr 2023 12:45:56 +0000 (08:45 -0400)]
Fix the buildbots after
636dd1e8a1782e22f9bdee640428ed5c50a4a4f2
This addresses issues found by:
http://45.33.8.238/linux/103908/step_4.txt
https://lab.llvm.org/buildbot/#/builders/216/builds/19607
and others
Piotr Zegar [Mon, 10 Apr 2023 12:42:24 +0000 (12:42 +0000)]
[clang-tidy][NFC] Remove trailing spaces from ReleaseNotes.rst
Cleanup some trailing whitespaces from ReleaseNotes.rst
Aaron Ballman [Mon, 10 Apr 2023 12:29:21 +0000 (08:29 -0400)]
Make explicit the single-argument constructors of AttributeCommonInfo; NFC
The single-argument constructors of this class were not marked explicit
and that led to some incorrect uses that slipped under the radar (see
changes in SemaDeclAttr.cpp). This makes the constructors explicit,
changes the benignly incorrect uses, and updates the tablegen code to
emit the correct support code to call the explicit constructors.
While this does correct a misuse, that incorrect usage could not be
observed except via a debugger and so no additional tests are added.
Differential Revision: https://reviews.llvm.org/D147661
Shengchen Kan [Mon, 10 Apr 2023 12:21:06 +0000 (20:21 +0800)]
[TableGen] Fix compile fail when compiling llvm with MSVC due to typo in
931418667063
Florian Hahn [Mon, 10 Apr 2023 12:02:44 +0000 (13:02 +0100)]
[VPlan] Use VPLiveOut to update FOR live-out users.
Instead of iterating over all LCSSA phis in the exit block, collect all
LiveOut users of the FOR splice VPInstruction and only update those
users.
Building on top of D147471, this removes an access to the cost model
after VPlan execution.
Depends on D147471.
Reviewed By: Ayal, michaelmaitland
Differential Revision: https://reviews.llvm.org/D147472
Congcong Cai [Mon, 10 Apr 2023 11:42:32 +0000 (13:42 +0200)]
[clang-tidy] avoid colon in namespace cause false positve
Refactor the Namespaces with NamespaceDecl[][].
First level stores non nested namepsace.
Second level stores nested namespace.
Reviewed By: PiotrZSL
Differential Revision: https://reviews.llvm.org/D147893
mmarjano [Fri, 7 Apr 2023 11:27:27 +0000 (13:27 +0200)]
[AMDGPU] Extend tbuffer_load_format merge
Add support for merging _IDXEN and _BOTHEN variants of
TBUFFER_LOAD_FORMAT instruction.
Vlad Serebrennikov [Mon, 10 Apr 2023 10:17:04 +0000 (13:17 +0300)]
[clang] Mark CWG536 as N/A
[[https://wg21.link/p1787 | P1787]]: CWG536 (long partially resolved) is resolved by specifying that unqualified lookup occurs for any kind of unqualified-id as an id-expression.
Wording: An unqualified name is a name that <...>. Unless otherwise specified, such a name undergoes unqualified name lookup from the point where it appears. ([basic.lookup.unqual]/4)
Reviewed By: #clang-language-wg, aaron.ballman
Differential Revision: https://reviews.llvm.org/D147564
Piotr Zegar [Mon, 10 Apr 2023 09:50:00 +0000 (09:50 +0000)]
[clang-tidy][NFC] Rename test files for simplify-boolean-expr to match check name
Change simplify-bool-expr -> simplify-boolean-expr
Max Kazantsev [Mon, 10 Apr 2023 09:37:10 +0000 (16:37 +0700)]
[SCEV] Improve AddRecs' range computation in Expensive Range Sharpening mode
Apply loop guards to AddRec's start in range computation for
non-self-wrapping AddRecs.
According to CT measurements, this has a wide negative compile time impact,
so we hold it in expensive range sharpening mode where it's not so critical.
However, we need to find a way to share benefits of this mode with default mode.
Patch by Aleksandr Popov!
Differential Revision: https://reviews.llvm.org/D147557
Reviewed By: mkazantsev
Luo, Yuanke [Mon, 10 Apr 2023 07:37:05 +0000 (15:37 +0800)]
[Coverity] avoid array overflow when use -1 as index.
Craig Topper [Mon, 10 Apr 2023 06:13:51 +0000 (23:13 -0700)]
[TableGen] Pass size to std::vector constructor instead of using resize. NFC
Craig Topper [Mon, 10 Apr 2023 04:39:57 +0000 (21:39 -0700)]
[TableGen] Make TreePatternNode::InlinePatternFragments a static method. NFC
Previously we were passing 'this' and the std::shared_ptr version
of 'this'. This replaces all uses of 'this' with the shared_ptr and
makes the method static.
Craig Topper [Mon, 10 Apr 2023 03:53:45 +0000 (20:53 -0700)]
[TableGen] Move vectors into DAGInstruction instead of copying them. NFC
Max Kazantsev [Mon, 10 Apr 2023 06:57:51 +0000 (13:57 +0700)]
[NFC][IRCE] Do not store latch exit count
It is not actually used for any computations. Its only purpose is to
check that the loop is finite and find out the type of computed exit
count. Refactor code so that we only store this type.
skc7 [Tue, 28 Mar 2023 18:05:02 +0000 (23:35 +0530)]
[AMDGPU] Introduce SIInstrWorklist to process instructions in moveToVALU
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D147168
Noah Goldstein [Mon, 10 Apr 2023 00:23:30 +0000 (19:23 -0500)]
[X86] Add inst fixup for `unpckps` -> `unpckdq`.
`unpckps` has the same performance as `unpckpd` (only port5) wereas
`unpckdq` can run on p15 on some newer architectures.
`unpckdq` is in the integer domain, so only do the transform if the
target has no bypass delay on shuffles (SKL+).
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D147729
Noah Goldstein [Mon, 10 Apr 2023 00:23:26 +0000 (19:23 -0500)]
[X86] Add inst fixup for `unpckpd` -> `unpckqdq`.
`unpckqdq` seems to be treated as a shuffle from bypass delay
perspective (which makes sense it appears to have shared shuffle units
for all micro-arch).
`unpckqdq` is slightly preferable to `shufpd` as it saves 1-byte of
code size and can be used to replace the micro-fused `rm` version. So,
if the target has no bypass delay, we should do `unpckpd` ->
`unpckqdq` instead of `shufpd.
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D147728
Noah Goldstein [Mon, 10 Apr 2023 00:23:23 +0000 (19:23 -0500)]
[X86] Fix perf bug in `permilps` -> `shufd` in X86FixupInstTuning.
We shouldn't do the transformation if we either have bypass delay OR
the new opcode has worse performance. Previous code was incorrectly
using AND.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D147727