Andy Wingo [Tue, 16 Feb 2021 12:28:44 +0000 (13:28 +0100)]
[lld][WebAssembly] Fix resolveIndirectFunctionTable for relocatable output
For relocatable output that needs the indirect function table, identify
the well-known function table. This allows us to properly fix the
limits on the imported table, and in a followup will allow the element
section to reference the indirect function table even if it's not
assigned to table number 0. Adapt tests for import reordering.
Differential Revision: https://reviews.llvm.org/D96770
Andy Wingo [Wed, 17 Feb 2021 16:20:28 +0000 (17:20 +0100)]
[WebAssembly][lld] --importTable flag only imports table if needed
Before, --importTable forced the creation of an indirect function table,
whether it was needed or not. Now it only imports a table if needed.
Differential Revision: https://reviews.llvm.org/D96872
Raphael Isemann [Thu, 18 Feb 2021 08:02:52 +0000 (09:02 +0100)]
[lldb][NFC] Silence missing case warnings in TypeSystemClang::GetTypeBitAlign due to new RISC-V V types
Those types were added in D92715. This just silences the warning but doesn't
actually add support for those types to LLDB.
Kazu Hirata [Thu, 18 Feb 2021 07:58:46 +0000 (23:58 -0800)]
[CodeGen] Use range-based for loops (NFC)
Kazu Hirata [Thu, 18 Feb 2021 07:58:44 +0000 (23:58 -0800)]
[llvm] Ensure newlines at the end of files (NFC)
This patch eliminates pesky "No newline at end of file" messages from
git diff.
Kazu Hirata [Thu, 18 Feb 2021 07:58:43 +0000 (23:58 -0800)]
[CodeGen] Use ListSeparator (NFC)
Jason Molenda [Thu, 18 Feb 2021 07:44:03 +0000 (23:44 -0800)]
Add a new Row setting to mark all un-declared regs as Undefined
Add a new state for UnwindPlan::Row which indicates that any
register not listed is not defined, and should not be found in
stack frames newer than this one and passed up the stack. Mostly
intended for use with architectural default unwind plans that are
used for jitted stack frames, where we have no unwind information
or start address. lldb has no way to tell if registers were
spilled in the jitted frame & overwritten, so passing register
values up the stack is not safe to show the user.
Architectural default unwind plans are also used as a fast unwind
plan on x86_64 in particular, and are used as the fallback unwind
plans when lldb thinks it may be able to work around a problem
which causes the unwinder to stop walking the stack early.
For fast unwind plans, when we don't find a register location in
the arch default unwind plan, we fall back to computing & using
the full unwind plan. One small part of this patch is to know that
a register marked as Undefined in the fast unwind plan is a special
case, and we should continue on to the full unwind plan to find what
the real unwind rule is for this register.
Differential Revision: https://reviews.llvm.org/D96829
<rdar://problem/
70398009>
Fangrui Song [Thu, 18 Feb 2021 07:33:13 +0000 (23:33 -0800)]
[profile] Make {__start_,__stop_}__llvm_prf_* symbols undefined weak
To make a kind of metadata section usage work, we want to drop the
`__start_/__stop_ references retain C identifier name sections` rule from LLD (see D96914).
If an application has no `__llvm_prf_data` input section surviving --gc-sections,
LLD will error for undefined hidden `{__start_,__stop_}__llvm_prf_*` from `libclang_rt.profile-*`.
Other `__llvm_prf_*` sections have similar issues.
Making the references weak can address the problem.
This probably enables the opportunity to drop zero size dummy sections in `InstrProfilingPlatformLinux.c`.
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D96936
Fangrui Song [Thu, 18 Feb 2021 07:17:20 +0000 (23:17 -0800)]
[llvm-objdump] Map STT_TLS to ST_Other (previously ST_Data)
ST_Data is used to model BFD `BFD_OBJECT`.
A STT_TLS symbol does not have the `BFD_OBJECT` flag in BFD.
This makes sense because a STT_TLS symbol is like in a different address space,
normal data/object properties do not apply on them.
With this change, a STT_TLS symbol will not be displayed as 'O'.
This new behavior matches objdump.
Differential Revision: https://reviews.llvm.org/D96735
Martin Storsjö [Wed, 4 Nov 2020 13:59:56 +0000 (15:59 +0200)]
[libcxx] Implement append and operator/ properly for windows
The root_path function has to be changed to return the parsed bit
as-is; otherwise a path like "//net" gets a root path of "//net/", as
the root name, "//net", gets the root directory (an empty string) appended,
forming "//net/". (The same doesn't happen for the root dir "c:" though.)
Differential Revision: https://reviews.llvm.org/D91178
Yang Fan [Thu, 18 Feb 2021 06:57:45 +0000 (14:57 +0800)]
[MC][ELF] Fix gcc "enumeral and non-enumeral type in conditional expression" warning (NFC)
GCC warning:
```
/llvm-project/llvm/lib/MC/ELFObjectWriter.cpp: In member function ‘uint64_t {anonymous}::ELFWriter::writeObject(llvm::MCAssembler&, const llvm::MCAsmLayout&)’:
/llvm-project/llvm/lib/MC/ELFObjectWriter.cpp:1137:38: warning: enumeral and non-enumeral type in conditional expression [-Wextra]
1137 | write(uint32_t(Group->isComdat() ? ELF::GRP_COMDAT : 0));
| ~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~
```
Stanislav Mekhanoshin [Thu, 18 Feb 2021 06:41:04 +0000 (22:41 -0800)]
[AMDGPU] Temporary remove test
Remove hsa-gfx90a-v3.s until D95638. It unexpectedly passes
on s390x.
Yang Fan [Thu, 18 Feb 2021 06:23:18 +0000 (14:23 +0800)]
[MC][ELF] Fix unused variable warning (NFC)
GCC warning:
```
/llvm-project/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp: In member function ‘virtual llvm::MCSection* llvm::TargetLoweringObjectFileELF::getSectionForLSDA(const llvm::Function&, const llvm::MCSymbol&, const llvm::TargetMachine&) const’:
/llvm-project/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp:871:8: warning: variable ‘IsComdat’ set but not used [-Wunused-but-set-variable]
871 | bool IsComdat = false;
| ^~~~~~~~
```
Vitaly Buka [Thu, 18 Feb 2021 06:09:46 +0000 (22:09 -0800)]
[NFC] Simplify msan test
Wang, Pengfei [Thu, 18 Feb 2021 03:35:34 +0000 (11:35 +0800)]
[X86] Zero AMX config buffer for non AVX512 cases.
Zero AMX config buffer for non AVX512 cases.
Differential Revision: https://reviews.llvm.org/D96927
Fangrui Song [Thu, 18 Feb 2021 04:03:16 +0000 (20:03 -0800)]
[GWP-ASan] Change sys/cdefs.h to features.h
sys/cdefs.h is a glibc internal header which is not supposed to be included by applications.
(Some libc implementations provide this file for compatibility.)
Android features.h includes sys/cdefs.h, so we can include features.h instead.
This change makes `ninja gwp_asan` build on musl.
Mehdi Chinoune [Thu, 18 Feb 2021 03:45:29 +0000 (21:45 -0600)]
[Flang] Fix compilation on MinGW-w64
Reviewed By: Meinersbur
Differential Revision: https://reviews.llvm.org/D94707
Igor Kudrin [Wed, 17 Feb 2021 14:17:46 +0000 (21:17 +0700)]
[Driver] Honor "-gdwarf-N" at any position for assembler sources
This fixes an issue when "-gdwarf-N" switch was ignored if it was given
before another debug option.
Differential Revision: https://reviews.llvm.org/D96865
Wang, Pengfei [Thu, 18 Feb 2021 03:30:12 +0000 (11:30 +0800)]
[X86] Add AVX2/SSE2 checks for AMX config buffer zeroing. NFC
Craig Topper [Thu, 18 Feb 2021 02:56:11 +0000 (18:56 -0800)]
[RISCV] Guard LowerINSERT_VECTOR_ELT against fixed vectors.
The type legalizer can call this code based on the scalar type so
we need to verify the vector type is a scalable vector.
I think due to how type legalization visits nodes, the vector type
will have already been legalized so we don't have an issue with
using MVT here like we did for EXTRACT_VECTOR_ELT.
I've added a test just in case.
Fangrui Song [Thu, 18 Feb 2021 03:22:25 +0000 (19:22 -0800)]
[profile] Add __attribute__((used)) to zero size dummy sections
D14468 added these dummy sections. This patch adds `__attribute__((used))` so
that when compiled by GCC>=11 or (expected, D96838) Clang>=13 on some ELF platforms,
these sections will get SHF_GNU_RETAIN to make sure they will not be discarded
by ld --gc-sections.
We are trying to get rid of LLD's "__start_/__stop_ references retain C identifier name sections" rule.
If LLD drops the rule in the future (we will retain compatibility for `__llvm_prf_*` for a while),
`__llvm_prf_*` will need to have the SHF_GNU_RETAIN flag, otherwise:
```
// __llvm_prf_cnts/__llvm_prf_data usually exist, but {names,vnds} may not exist.
// Such diagnostics will happen with {cnts,data} as well if no input object file is instrumented.
% clang++ -fprofile-generate a.cc -fuse-ld=lld -Wl,--gc-sections
ld.lld: error: undefined hidden symbol: __start___llvm_prf_names
>>> referenced by InstrProfilingPlatformLinux.c
>>> InstrProfilingPlatformLinux.c.o:(__llvm_profile_begin_names) in archive /tmp/RelA/lib/clang/13.0.0/lib/linux/libclang_rt.profile-x86_64.a
...
```
Differential Revision: https://reviews.llvm.org/D96902
Joseph Huber [Wed, 27 Jan 2021 15:34:01 +0000 (10:34 -0500)]
[LV] Add analysis remark for mixed precision conversions
Floating point conversions inside vectorized loops have performance
implications but are very subtle. The user could specify a floating
point constant, or call a function without realizing that it will
force a change in the vector width. An example of this behaviour is
seen in https://godbolt.org/z/M3nT6c . The vectorizer should indicate
when this happens becuase it is most likely unintended behaviour.
This patch adds a simple check for this behaviour by following floating
point stores in the original loop and checking if a floating point
conversion operation occurs.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D95539
Craig Topper [Thu, 18 Feb 2021 02:19:22 +0000 (18:19 -0800)]
[RISCV] Guard the ISD::EXTRACT_VECTOR_ELT handling in ReplaceNodeResults against fixed vectors and non-MVT types.
The type legalizer is calling this code based on the scalar type so
we need to verify the input type is a scalable vector.
The vector type has also not been legalized yet when this is called
so we need to use EVT for it.
Aart Bik [Thu, 18 Feb 2021 00:47:33 +0000 (16:47 -0800)]
[mlir][sparse] generalize sparse storage format to many more types
Rationale:
Narrower types for overhead storage yield a smaller memory footprint for
sparse tensors and thus needs to be supported. Also, more value types
need to be supported to deal with all kinds of kernels. Since the
"one-size-fits-all" sparse storage scheme implementation is used
instead of actual codegen, the library needs to be able to support
all combinations of desired types. With some crafty templating and
overloading, the actual code for this is kept reasonably sized though.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D96819
Hsiangkai Wang [Thu, 4 Feb 2021 04:57:36 +0000 (12:57 +0800)]
[Clang][RISCV] Define RISC-V V builtin types
Add the types for the RISC-V V extension builtins.
These types will be used by the RISC-V V intrinsics which require
types of the form <vscale x 1 x i64>(LMUL=1 element size=64) or
<vscale x 4 x i32>(LMUL=2 element size=32), etc. The vector_size
attribute does not work for us as it doesn't create a scalable
vector type. We want these types to be opaque and have no operators
defined for them. We want them to be sizeless. This makes them
similar to the ARM SVE builtin types. But we will have quite a bit
more types. This patch adds around 60. Later patches will add
another 230 or so types representing tuples of these types similar
to the x2/x3/x4 types in ARM SVE. But with extra complexity that
these types are combined with the LMUL concept that is unique to
RISCV.
For more background see this RFC
http://lists.llvm.org/pipermail/llvm-dev/2020-October/145850.html
Authored-by: Roger Ferrer Ibanez <roger.ferrer@bsc.es>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Differential Revision: https://reviews.llvm.org/D92715
Stanislav Mekhanoshin [Thu, 18 Feb 2021 01:59:24 +0000 (17:59 -0800)]
[AMDGPU] Fixed msan build
LoadStoreOptimizer was using uninitialized SCC value for
instructions where it is unsupported.
Eric Schweitz [Wed, 17 Feb 2021 23:34:18 +0000 (15:34 -0800)]
[flang][fir][NFC] clang-tidy change. Add include.
Differential Revision: https://reviews.llvm.org/D96912
Chen Zheng [Thu, 18 Feb 2021 01:42:45 +0000 (20:42 -0500)]
[XCOFF][NFC] make csect properties optional for getXCOFFSection
We are going to support debug sections for XCOFF. So the csect
properties are not necessary. This patch makes these properties
optional.
Reviewed By: hubert.reinterpretcast
Differential Revision: https://reviews.llvm.org/D95931
Eric Schweitz [Wed, 17 Feb 2021 23:16:04 +0000 (15:16 -0800)]
[flang][fir][NFC] Merge tablegen files.
Differential Revision: https://reviews.llvm.org/D96908
Marco Vanotti [Wed, 17 Feb 2021 21:42:40 +0000 (13:42 -0800)]
[libunwind] Add support for PC reg column in arm64
This change adds support for the dwarf PC register column in arm64, allowing
CFI directives to make use of it.
As of the last revision of the DWARF for ARM 64-bit architecture[0], the pc
register has been added as a valir register, with number 32.
This allows libunwinder to restore both pc and lr, which is useful
for stack switches and signal contexts.
[0]:
https://github.com/ARM-software/abi-aa/blob/
f52e1ad3f81254497a83578dc102f6aac89e52d0/aadwarf64/aadwarf64.rst
Reviewed By: phosek, #libunwind
Differential Revision: https://reviews.llvm.org/D96901
Joerg Sonnenberger [Thu, 18 Feb 2021 00:53:04 +0000 (01:53 +0100)]
[NetBSD] Use cortex-a8 as default CPU for ARMv7
This matches the platform default for GCC. It primarily matters when the
integrated assembler is not used as there is no default CPU defined for
ARMv7-A and GNU as is upset with -mcpu=generic.
Stanislav Mekhanoshin [Wed, 17 Feb 2021 00:34:24 +0000 (16:34 -0800)]
[AMDGPU] Mark SMRD atomics
We did not have atomic flags on SMRD, did not copy TSFlags
to real instructions, and did not have ret/noret atomic map.
At the moment it is NFC, but needed for D96469.
Differential Revision: https://reviews.llvm.org/D96823
Teresa Johnson [Wed, 3 Feb 2021 21:12:40 +0000 (13:12 -0800)]
[WPD] Add an optional checking mode for debugging devirtualization
This adds an internal option -wholeprogramdevirt-check which if enabled
will guard each devirtualization with a runtime check against the
expected target, and an invocation of a debug trap if the check fails.
This is useful for debugging WPD failures involving undefined behavior
(e.g. casting to another class type not in the inheritance chain).
Differential Revision: https://reviews.llvm.org/D95969
Nico Weber [Thu, 18 Feb 2021 00:36:36 +0000 (19:36 -0500)]
[gn build] add a comment to the goma_dir arg
Heejin Ahn [Wed, 17 Feb 2021 22:25:56 +0000 (14:25 -0800)]
[WebAssembly] Remove dependency of reference types from EH
The new spec does not have `exnref` so EH does not have dependency of
the reference types proposal anymore.
Reviewed By: dschuff
Differential Revision: https://reviews.llvm.org/D96903
Stanislav Mekhanoshin [Wed, 17 Feb 2021 21:37:46 +0000 (13:37 -0800)]
[AMDGPU] gfx90a support
Differential Revision: https://reviews.llvm.org/D96906
Rahman Lavaee [Wed, 17 Feb 2021 02:43:56 +0000 (18:43 -0800)]
[obj2yaml,yaml2obj] Add NumBlocks to the BBAddrMapEntry yaml field.
As discussed in D95511, this allows us to encode invalid BBAddrMap
sections to be used in more rigorous testing.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D96831
Eric Schweitz [Wed, 17 Feb 2021 23:27:55 +0000 (15:27 -0800)]
[flang][fir][NFC] clang-tidy change
Differential Revision: https://reviews.llvm.org/D96911
LLVM GN Syncbot [Wed, 17 Feb 2021 23:33:31 +0000 (23:33 +0000)]
[gn build] Port
7397905ab0a0
Rong Xu [Wed, 17 Feb 2021 22:19:36 +0000 (14:19 -0800)]
[SampleFDO] Third Try: Refactor SampleProfile.cpp
Apply the patch for the third time after fixing buildbot failures.
Refactor SampleProfile.cpp to use the core code in CodeGen.
The main changes are:
(1) Move SampleProfileLoaderBaseImpl class to a header file.
(2) Split SampleCoverageTracker to a head file and a cpp file.
(3) Move the common codes (common options and callsiteIsHot())
to the common cpp file.
(4) Add inline keyword to avoid duplicated symbols -- they will
be removed later when the class is changed to a template.
Differential Revision: https://reviews.llvm.org/D96455
Teresa Johnson [Mon, 15 Feb 2021 17:19:50 +0000 (09:19 -0800)]
[gold] Match lld WPD behavior for shared library symbols and add test
lld already marks shared library defs as ExportDynamic, which prevents
potentially unsafe devirtualization of symbols defined in shared
libraries. Match that behavior in the gold plugin, and add the same
test.
Depends on D96721.
Differential Revision: https://reviews.llvm.org/D96722
AndreyChurbanov [Wed, 17 Feb 2021 23:27:49 +0000 (02:27 +0300)]
[OpenMP] fix race condition in test
Jon Chesterfield [Wed, 17 Feb 2021 23:02:16 +0000 (23:02 +0000)]
[libomptarget][amdgcn] Remove lookup of .language msgpack field
Rob Suderman [Wed, 17 Feb 2021 22:01:32 +0000 (14:01 -0800)]
[MLIR][TOSA] Expand Tosa int types to I8 and I16
Tosa integers should include I8 and I16 values.
Differential Revision: https://reviews.llvm.org/D96900
Patrick Oppenlander [Wed, 17 Feb 2021 22:04:36 +0000 (17:04 -0500)]
[libc++abi] Add builtins to dynamic library link
Otherwise libc++abi.so fails to link on arm with undefined references to
some __aeabi_ builtins.
Differential Revision: https://reviews.llvm.org/D96574
Jessica Paquette [Wed, 17 Feb 2021 19:54:51 +0000 (11:54 -0800)]
[GlobalISel] Implement computeKnownBits for G_ASSERT_SEXT
Implementation is the same as G_SEXT_INREG.
Differential Revision: https://reviews.llvm.org/D96899
Jessica Paquette [Wed, 17 Feb 2021 19:35:46 +0000 (11:35 -0800)]
[GlobalISel] Implement computeNumSignBits for G_ASSERT_SEXT
Same implementation as G_SEXT_INREG.
Add a testcase to combine-sext-inreg for a concrete example, and a testcase
to KnownBitsTest.
Differential Revision: https://reviews.llvm.org/D96897
Fangrui Song [Wed, 17 Feb 2021 21:49:41 +0000 (13:49 -0800)]
[Driver] Clean up some Separate form options
Drop the `Separate` form of `-fmodule-name X`, `-fprofile-remapping-file X`, and `-frewrite-map-file X`.
To the best of my knowledge they are not used. Their conventional Joined forms (`-fFOO=`) should be used instead.
`-fdebug-compilation-dir X` is used in several places, e.g. chromium/infra/goma.
It is also advertised in http://blog.llvm.org/2019/11/deterministic-builds-with-clang-and-lld.html
So we keep it but make the EQ form canonical and the Separate form an alias.
Differential Revision: https://reviews.llvm.org/D96886
AndreyChurbanov [Wed, 17 Feb 2021 21:38:18 +0000 (00:38 +0300)]
[OpenMP][NFC] replaced 'dependencies' with 'dependences' in comments and debug prints
peter klausler [Wed, 17 Feb 2021 18:24:14 +0000 (10:24 -0800)]
[flang] Warn about useless explicit typing of intrinsics
Fortran 2018 explicitly permits an ignored type declaration
for the result of a generic intrinsic function. See the comment
added to Semantics/expression.cpp for an explanation of why this
is somewhat dangerous and worthy of a warning.
Differential Revision: https://reviews.llvm.org/D96879
Yusra Syeda [Wed, 17 Feb 2021 21:06:02 +0000 (16:06 -0500)]
[SystemZ] Separate LoZ ELF specifics in tablegen.
Separate the LoZ ELF calling convention in tablegen.
This will make it easier to add the z/OS ABI in future patches.
Reviewed By: uweigand
Differential Revision: https://reviews.llvm.org/D96867
Jessica Paquette [Wed, 17 Feb 2021 18:57:10 +0000 (10:57 -0800)]
[GlobalISel] Add G_ASSERT_SEXT
This adds a G_ASSERT_SEXT opcode, similar to G_ASSERT_ZEXT. This instruction
signifies that an operation was already sign extended from a smaller type.
This is useful for functions with sign-extended parameters.
E.g.
```
define void @foo(i16 signext %x) {
...
}
```
This adds verifier, regbankselect, and instruction selection support for
G_ASSERT_SEXT equivalent to G_ASSERT_ZEXT.
Differential Revision: https://reviews.llvm.org/D96890
Aaron Green [Wed, 17 Feb 2021 21:00:44 +0000 (13:00 -0800)]
Bugfix for collecting features from very small DSOs.
During unit tests, it was observed that crafting an artificially small DSO could cause OOB memory to be accessed. This change fixes that (but again, the affected DSOs are unlikely to ever occur outside unit tests).
Reviewed By: morehouse, charco
Differential Revision: https://reviews.llvm.org/D94507
Teresa Johnson [Mon, 15 Feb 2021 01:48:34 +0000 (17:48 -0800)]
[WPD][lld] Test handling of vtable definition from shared libraries
Adds a lld test for a case that the handling added for dynamically
exported symbols in
1487747e990ce9f8851f3d92c3006a74134d7518 already
fixes. Because isExportDynamic returns true when the symbol is
SharedKind with default visibility, it will treat as dynamically
exported and block devirtualization when the definition of a vtable
comes from a shared library. This is desireable as it is dangerous to
devirtualize in that case, since there could be hidden overrides in the
shared library. Typically that happens when the shared library header
contains available externally definitions, which applications can
override. An example is std::error_category, which is overridden in LLVM
and causing failures after a self build with WPD enabled, because
libstdc++ contains hidden overrides of the virtual base class methods.
The regular LTO case in the new test already worked, but there are
2 fixes in this patch needed for the index-only case and the hybrid
LTO case. For the index-only case, WPD should not simply ignore
available externally vtables. A follow on fix will be made to clang to
emit type metadata for those vtables, which the new test is modeling.
For the hybrid case, we need to ensure when the module is split that any
llvm.*used globals are cloned to the regular LTO split module so
available externally vtable definitions are not prematurely deleted.
Another follow on fix will add the equivalent gold test, which requires
a small fix to the plugin to treat symbols in dynamic libraries the same
way lld already is.
Differential Revision: https://reviews.llvm.org/D96721
Sriraman Tallam [Wed, 17 Feb 2021 20:34:37 +0000 (12:34 -0800)]
Basic block sections should enable not function sections implicitly.
Basic block sections enables function sections implicitly, this is not needed
and is inefficient with "=list" option.
We had basic block sections enable function sections implicitly in clang. This
is particularly inefficient with "=list" option as it places functions that do
not have any basic block sections in separate sections. This causes unnecessary
object file overhead for large applications.
This patch disables this implicit behavior. It only creates function sections
for those functions that require basic block sections.
This patch is the second of two patches and this patch removes the implicit
enabling of function sections with basic block sections in clang.
Differential Revision: https://reviews.llvm.org/D93876
Nico Weber [Wed, 17 Feb 2021 20:29:39 +0000 (15:29 -0500)]
fix comment typo to cycle bots
Heejin Ahn [Mon, 15 Feb 2021 15:32:51 +0000 (07:32 -0800)]
[WebAssemblly] Fix EHPadStack update in fixCallUnwindMismatches
Updating `EHPadStack` with respect to `TRY` and `CATCH` instructions
have to be done after checking all other conditions, not before. Because
we did this before checking other conditions, when we encounter `TRY`
and we want to record the current mismatching range, we already have
popped up the entry from `EHPadStack`, which we need to access to record
the range.
The `baz` call in the added test needs try-delegate because the previous
TRY marker placement for `quux` was placed before `baz`, because `baz`'s
return value was stackified in RegStackify. If this wasn't stackified
this try-delegate is not strictly necessary, but at the moment it is not
easy to identify cases like this. I plan to transfer `nounwind`
attributes from the LLVM IR to prevent cases like this. The call in the
test does not have `unwind` attribute in order to test this bug, but in
many cases of this pattern the previous call has `nounwind` attribute.
Reviewed By: tlively
Differential Revision: https://reviews.llvm.org/D96711
Mircea Trofin [Wed, 17 Feb 2021 20:07:27 +0000 (12:07 -0800)]
[NFC][RegAlloc] InlineSpiller::Original is a Register
Nico Weber [Wed, 17 Feb 2021 19:47:02 +0000 (14:47 -0500)]
[gn build] make WindowsManifestMerger.cpp build fine with sysroot
This already works in the cmake build.
Differential Revision: https://reviews.llvm.org/D96889
Zixu Wang [Wed, 17 Feb 2021 19:44:43 +0000 (11:44 -0800)]
[NFC][clang] Bump up DIAG_SIZE_SEMA for downstream diagnostics
Bump DIAG_SIZE_SEMA up by 500 to accommodate extra downstream diagnostics
Differential Revision: https://reviews.llvm.org/D96888
Craig Topper [Wed, 17 Feb 2021 19:37:06 +0000 (11:37 -0800)]
[RISCV] Localize RISCVZvlssegTable to RISCVISelDAGToDAG.cpp, the only place it is used.
peter klausler [Wed, 17 Feb 2021 18:11:29 +0000 (10:11 -0800)]
[flang] Ensure that intrinsic procedures are PURE &/or ELEMENTAL
The intrinsic procedure table properly classify the various
intrinsics, but the PURE and ELEMENTAL attributes that these
classifications imply don't always make it to the utility
predicates that test symbols for them, leading to spurious
error messages in some contexts. So set those attribute flags
as appropriate in name resolution, using a new function to
isolate the tests.
An alternate solution, in which the predicates would query
the intrinsic procedure table for these attributes on demand,
was something I also tried, so that this information could
come directly from an authoritative source; but it would have
required references to the intrinsic table to be passed along
on too many seemingly unrelated APIs and ended up looking messy.
Several symbol table tests needed to have their expected outputs
augmented with the PURE and ELEMENTAL flags. Some bogus messages
that were flagged as such in test/Semantics/doconcurrent01.f90 were
removed, since they are now correctly not emitted.
Differential Revision: https://reviews.llvm.org/D96878
Derek Schuff [Wed, 17 Feb 2021 00:35:27 +0000 (16:35 -0800)]
[WebAssembly] Do not use EHCatchret symbols with wasm EH
D94835 added support for WinEH to export public symbols pointing to
basic blocks which are catchret targets for use with Windows CET.
Wasm currently doesn't support public symbols to non-function code
addresses (they get treated like new functions in asm but then don't
lower to object files correctly).
It created them unconditionally for all catchret targets.
This change disables those symbols unless the exceptionHandlingType
is WinEH (since they aren't used with ExceptionHandling::Wasm)
Differential Revision: https://reviews.llvm.org/D96824
Craig Topper [Wed, 17 Feb 2021 19:11:58 +0000 (11:11 -0800)]
[RISCV] Use bits<7> instead of bits<11> for the EEW field size in the RISCVZvlsseg searchable table. NFCI
We only support 8, 16, 32, and 64 for EEW. These only need 7 bits
to represent.
William S. Moses [Wed, 17 Feb 2021 18:54:42 +0000 (13:54 -0500)]
[SROA] Amend failing test from D95826
Nico Weber [Wed, 17 Feb 2021 18:48:10 +0000 (13:48 -0500)]
sysroot.py: add support for non-darwin platforms
CMAKE_SYSROOT works fine here, and `sysroot.py make-fake`
borders on trivial here, but I suppose it's still nice
to have a consistent script to set these up across platforms.
And these are the platforms where we can do real sysroot management one
day.
Differential Revision: https://reviews.llvm.org/D96882
LLVM GN Syncbot [Wed, 17 Feb 2021 18:30:02 +0000 (18:30 +0000)]
[gn build] Port
c28622fbf363
Alex Zinenko [Mon, 15 Feb 2021 09:52:19 +0000 (10:52 +0100)]
[mlir] silence unused-function warnings in table-generated code
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D96695
Vedant Kumar [Wed, 17 Feb 2021 18:20:40 +0000 (10:20 -0800)]
Revert "[SampleFDO] Reapply: Refactor SampleProfile.cpp"
Revert "[SampleFDO] Add missing #includes to unbreak modules build after D96455"
This reverts commit
c73cbf218a289029cc0b54183c3cf79454ecc76f.
Revert "[SampleFDO] Fix MSVC "namespace uses itself" warning (NFC)"
This reverts commit
a23e6b321ca623b83252f8b1e06a2ad4fc441f89.
Revert "[SampleFDO] Reapply: Refactor SampleProfile.cpp"
This reverts commit
6fd5ccff72eeaffcb3b3ba2696282015aab755bc.
Still seeing link failures when building llc (or other tools), due to
the new SampleProfileLoaderBaseImpl.h containing definitions that get
duplicated across multiple TU's.
```
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::findEquivalenceClasses(llvm::Function&)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::buildEdges(llvm::Function&)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::computeDominanceAndLoopInfo(llvm::Function&)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::getFunctionLoc(llvm::Function&)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::getBlockWeight(llvm::BasicBlock const*)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::printBlockWeight(llvm::raw_ostream&, llvm::BasicBlock const*) const' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::printBlockEquivalence(llvm::raw_ostream&, llvm::BasicBlock const*)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
duplicate symbol 'llvm::SampleProfileLoaderBaseImpl::printEdgeWeight(llvm::raw_ostream&, std::__1::pair<llvm::BasicBlock const*, llvm::BasicBlock const*>)' in:
tools/llc/CMakeFiles/llc.dir/llc.cpp.o
lib/libLLVMInstCombine.a(InstCombineVectorOps.cpp.o)
```
Heejin Ahn [Wed, 17 Feb 2021 13:27:19 +0000 (05:27 -0800)]
[WebAssembly] Change catch_all's opcode
We decided to change `catch_all`'s opcode from 0x05, which is the same
as `else`, to 0x19, to avoid some complicated handling in the tools.
See: https://github.com/WebAssembly/exception-handling/issues/147
Reviewed By: sbc100
Differential Revision: https://reviews.llvm.org/D96863
Craig Topper [Wed, 17 Feb 2021 06:58:14 +0000 (22:58 -0800)]
[RISCV] Merge the handlers for masked and unmasked segment loads/stores.
A lot of the code for the masked and unmasked is the same. This
patch adds a boolean to handle the differences so we can share
the code.
Differential Revision: https://reviews.llvm.org/D96841
Craig Topper [Wed, 17 Feb 2021 03:46:44 +0000 (19:46 -0800)]
[RISCV] Merge the vsetvli and vsetvlimax intrinsic selection
These have very similar code just with a different number of
operands and handling for vsetivl.
Differential Revision: https://reviews.llvm.org/D96834
Eugene Zhulenev [Wed, 17 Feb 2021 16:34:33 +0000 (08:34 -0800)]
[mlir] Add fma operation to std dialect
Will remove `vector.fma` operation in the followup CLs.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D96801
Dmitry Vyukov [Wed, 17 Feb 2021 16:43:16 +0000 (17:43 +0100)]
tsan: fix mmap_lots test
If tsan runtime will try to allocate something during exit handling,
the allocation will fail because there is no VA whatsoever.
It's observed to fail with the following error in some cases:
failed to allocate 0x1000 (4096) bytes of DTLS_NextBlock.
So terminate the process immediately.
Reviewed-in: https://reviews.llvm.org/D96874
Vedant Kumar [Wed, 17 Feb 2021 18:01:09 +0000 (10:01 -0800)]
[SampleFDO] Add missing #includes to unbreak modules build after D96455
Bot: http://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/28999
```
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Transforms/Utils/SampleProfileLoaderBaseImpl.h:124:19: error: missing '#include "llvm/Analysis/PostDominators.h"'; 'PostDominatorTree' must be declared before it is used
std::unique_ptr<PostDominatorTree> PDT;
^
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Analysis/PostDominators.h:28:7: note: declaration here is not visible
class PostDominatorTree : public PostDomTreeBase<BasicBlock> {
^
While building module 'LLVM_Transforms' imported from /Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/lib/Transforms/CFGuard/CFGuard.cpp:15:
In file included from <module-includes>:191:
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Transforms/Utils/SampleProfileLoaderBaseImpl.h:125:19: error: missing '#include "llvm/Analysis/LoopInfo.h"'; 'LoopInfo' must be declared before it is used
std::unique_ptr<LoopInfo> LI;
^
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Analysis/LoopInfo.h:1079:7: note: declaration here is not visible
class LoopInfo : public LoopInfoBase<BasicBlock, Loop> {
^
While building module 'LLVM_Transforms' imported from /Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/lib/Transforms/CFGuard/CFGuard.cpp:15:
In file included from <module-includes>:191:
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Transforms/Utils/SampleProfileLoaderBaseImpl.h:149:3: error: missing '#include "llvm/Analysis/OptimizationRemarkEmitter.h"'; 'OptimizationRemarkEmitter' must be declared before it is used
OptimizationRemarkEmitter *ORE = nullptr;
^
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/include/llvm/Analysis/OptimizationRemarkEmitter.h:33:7: note: declaration here is not visible
class OptimizationRemarkEmitter {
^
/Users/buildslave/jenkins/workspace/lldb-cmake/llvm-project/llvm/lib/Transforms/CFGuard/CFGuard.cpp:15:10: fatal error: could not build module 'LLVM_Transforms'
```
Jonas Devlieghere [Wed, 17 Feb 2021 17:30:07 +0000 (09:30 -0800)]
[lldb] Improve error message for modules with dots or dashes
LLDB does not like to import Python files with dashes or dots in their
name. While the former are technically allowed, it is discouraged. Dots
are allowed for subpackages but not in module names. This patch improves
the user experience by printing a useful error.
Before this patch:
error: module importing failed: SyntaxError('invalid syntax',
('<string>', 1, 11, 'import foo-bar\n'))
After this patch:
error: module importing failed: Python discourages dashes in module
names: foo-bar
rdar://
74263511
[1] https://www.python.org/dev/peps/pep-0008/#package-and-module-names
Differential revision: https://reviews.llvm.org/D96833
Martin Storsjö [Sun, 1 Nov 2020 21:39:03 +0000 (23:39 +0200)]
[libcxx] Implement is_absolute properly for windows
Differential Revision: https://reviews.llvm.org/D91177
Marianne Mailhot-Sarrasin [Wed, 17 Feb 2021 17:14:03 +0000 (12:14 -0500)]
[Pipeliner] Fixed optimization remarks and debug dumps Initiation
Interval value
The II value was incremented before exiting the loop, and therefor when
used in the optimization remarks and debug dumps it did not reflect the
initiation interval actually used in Schedule.
Differential Revision: https://reviews.llvm.org/D95692
William S. Moses [Mon, 1 Feb 2021 23:16:17 +0000 (18:16 -0500)]
[SROA] Propagate correct TBAA/TBAA Struct offsets
SROA does not correctly account for offsets in TBAA/TBAA struct metadata.
This patch creates functionality for generating new MD with the corresponding
offset and updates SROA to use this functionality.
Differential Revision: https://reviews.llvm.org/D95826
David Green [Wed, 17 Feb 2021 16:56:28 +0000 (16:56 +0000)]
[NPM][LTO] Update buildLTODefaultPipeline to be more in-line with the old pass manager
The NPM LTO pipeline has a lot of fixme's and missing passes, causing a
lot of regressions after the switch in c70737b. Notably unrolling and
vectorization were both disabled, but many other passes are missing
compared to the old pass manager. This attempt to enable the most
obvious missing passes like the unroller, vectorization and other loop
passes, fixing the existing FIXME comments.
Differential Revision: https://reviews.llvm.org/D96780
Ta-Wei Tu [Wed, 17 Feb 2021 16:55:03 +0000 (00:55 +0800)]
[NFC] Refactor LoopInterchange into a loop-nest pass
This is the preliminary patch of converting `LoopInterchange` pass to a loop-nest pass and has no intended functional change.
Changes that are not loop-nest related are split to D96650.
Reviewed By: Whitney
Differential Revision: https://reviews.llvm.org/D96644
Nico Weber [Wed, 17 Feb 2021 16:52:24 +0000 (11:52 -0500)]
fix comment typo to cycle bots
Sjoerd Meijer [Wed, 17 Feb 2021 15:32:04 +0000 (15:32 +0000)]
[LSR] Add a flag that overrides the target's preferred addressing mode
This adds a new flag -lsr-preferred-addressing-mode to override the target's
preferred addressing mode. It replaces flag -lsr-backedge-indexing, which is
equivalent to preindexed addressing that is one of the options that
-lsr-preferred-addressing-mode accepts.
Differential Revision: https://reviews.llvm.org/D96855
Nico Weber [Wed, 17 Feb 2021 16:49:23 +0000 (11:49 -0500)]
fix comment typo to cycle bots
Louis Dionne [Wed, 17 Feb 2021 16:19:37 +0000 (11:19 -0500)]
[libc++] Allow retries in a few more flaky tests
Sanjay Patel [Wed, 17 Feb 2021 15:16:12 +0000 (10:16 -0500)]
[InstCombine] fold fcmp-of-copysign idiom
As discussed in:
https://llvm.org/PR49179
...this pattern shows up in library code.
There are several potential generalizations as noted,
but we need to be careful that we get FP special-values
right, and it's not clear how much variation we should
expect to see from this exact idiom.
Sanjay Patel [Wed, 17 Feb 2021 14:20:04 +0000 (09:20 -0500)]
[InstCombine] add tests for fcmp-of-copysign; NFC
https://llvm.org/PR49179
Nico Weber [Wed, 17 Feb 2021 15:03:02 +0000 (10:03 -0500)]
build: Add LLVM_WINSYSROOT to make setting /winsysroot easy on Win
Also add a script for sysroot management. For now, it can only create
fake sysroots that just symlink to local folders. This is useful for
testing.
Differential Revision: https://reviews.llvm.org/D96868
Hanhan Wang [Wed, 17 Feb 2021 14:55:10 +0000 (06:55 -0800)]
[mlir][StandardToSPIRV] Add support for lowering trunci to SPIR-V to i1 types.
Add a pattern to converting some value to a boolean. spirv.S/UConvert does not
work on i1 types. Thus, the pattern is lowered to cmpi + select.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D96851
Raphael Isemann [Wed, 17 Feb 2021 15:11:33 +0000 (16:11 +0100)]
[lldb][NFC] Delete deleted const char* overloads of SetValueFromString
This came up during the review of D96817 because those deleted overloads force
the caller to explicitly create a StringRef when passing a string literal.
It seems they were added as some kind of help while migrating the code base to
StringRef in D24847, but I don't think they have any use these days and make
these functions awkward to use.
This patch just removes all the deleted overloads.
Reviewed By: tatyana-krasnukha
Differential Revision: https://reviews.llvm.org/D96861
Jonny Farley [Tue, 16 Feb 2021 16:34:35 +0000 (16:34 +0000)]
[Fuzzer][Test] Use %python substitution for trace-malloc-unbalanced.test
This test was found to fail for some of our downstream builds, on
computers where python was not on the default $PATH. Therefore
add a %python substitution to use sys.executable, based on similar
solutions for python calls in tests elsewhere in LLVM.
Differential Revision: https://reviews.llvm.org/D96799
Sidharth Baveja [Wed, 17 Feb 2021 14:54:57 +0000 (14:54 +0000)]
[PowerPC][AIX] Enable Shrinkwrapping on 32 and 64 bit AIX.
Summary:
Currently Shrinkwrap is not enabled on AIX.
This patch enables shrink wrap on 32 and 64 bit AIX, and 64 bit ELF.
Reviewed By: sfertile, nemanjai
Differential Revision: https://reviews.llvm.org/D95094
Haojian Wu [Wed, 17 Feb 2021 14:40:12 +0000 (15:40 +0100)]
[clangd] IndexedFiles should include Fils from RefSlab and RelationSlab.
This looks like an oversight.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D96845
Haojian Wu [Wed, 17 Feb 2021 07:39:52 +0000 (08:39 +0100)]
[clangd] Pass file when possible to resolve URI.
Some URI scheme needs the hint path to do a correct resolution, we pass
one of the open files as hint path.
This is not perfect, and it might not work for opening files across
project, but it would fix a bug with our internal scheme.
in the long run, removing URIs from all the index internals is a more proper fix.
Differential Revision: https://reviews.llvm.org/D96844
Sean Fertile [Wed, 17 Feb 2021 14:22:43 +0000 (09:22 -0500)]
[PowerPC] Handle FP physical register in inline asm constraint.
Do not defer to the base class when the register constraint is a
physical fpr. The base class will select SPILLTOVSRRC as the register
class and register allocation will fail on subtargets without VSX
registers.
Differential Revision: https://reviews.llvm.org/D91629
David Green [Wed, 17 Feb 2021 14:21:09 +0000 (14:21 +0000)]
[ARM] Add MVE abs costs
Similar to min/max, this increases the accuracy of abs intrinsics costs
under MVE.
Sven van Haastregt [Wed, 17 Feb 2021 14:17:43 +0000 (14:17 +0000)]
[OpenCL] Support enum and typedef args in TableGen BIFs
Add enum and typedef argument support to `-fdeclare-opencl-builtins`,
which was the last major missing feature.
Adding the remaining missing builtins is left as future work.
Differential Revision: https://reviews.llvm.org/D96051
Piotr Sobczak [Wed, 17 Feb 2021 09:56:29 +0000 (10:56 +0100)]
[AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32
* Update skip-if-dead.ll with tests for wave32.
* Fix the crash in verifier in one newly enabled test by adding
missing fixImplicitOperands in branch insertion code.
```
*** Bad machine code: Using an undefined physical register ***
- function: test_kill_divergent_loop
- basic block: %bb.2 bb (0xad96308)
- instruction: S_CBRANCH_VCCNZ %bb.1, implicit $vcc_lo
- operand 1: implicit $vcc_lo
LLVM ERROR: Found 1 machine code errors.
```
* Simplify "cbranch_kill" to not use interp instructions.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D96793
Sanne Wouda [Mon, 8 Feb 2021 16:13:28 +0000 (16:13 +0000)]
Use LoopRotate PrepareForLTO stage in NPM
The PrepareForLTO stage of LoopRotate tries to avoid unrolling loops
with calls that might be inlined later. See D94232 where this was
introduced.
We didn't catch all occurances of the LoopRotatePass in the New Pass
Manager, so the original regression in astar returned with the pass
manager switch.
Weiwei Li [Wed, 17 Feb 2021 14:00:28 +0000 (09:00 -0500)]
[mlir][spirv] Add spv.GLSL.FrexpStruct
co-authored-by: Alan Liu <alanliu.yf@gmail.com>
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D96527