Lang Hames [Sun, 17 Apr 2022 15:34:22 +0000 (08:34 -0700)]
[JITLink] Suppress "symbols not found" errors for testcase.
On some platforms _ZTIi may not be present (see discussion at
https://reviews.llvm.org/rG43acef48d38e). We don't need this symbol for the
test to work, so just add -phony-externals to the testcase to suppress the
error.
Andrew Savonichev [Mon, 11 Apr 2022 18:33:04 +0000 (21:33 +0300)]
[NVPTX] Disable parens for identifiers starting with '$'
ptxas fails to parse such syntax:
mov.u64 %rd1, ($str);
fatal : Parsing error near '$str': syntax error
A new MCAsmInfo option was added because InParens parameter of
MCExpr::print is not sufficient to disable parens
completely. MCExpr::print resets it to false for a recursive call in
case of unary or binary expressions.
Targets that require parens around identifiers that start with '$'
should always pass MCAsmInfo to MCExpr::print.
Therefore 'operator<<(raw_ostream &, MCExpr&)' should be avoided
because it calls MCExpr::print with nullptr MAI.
Differential Revision: https://reviews.llvm.org/D123702
Uday Bondhugula [Sat, 16 Apr 2022 03:54:20 +0000 (09:24 +0530)]
[MLIR] Provide a way to print ops in custom form on pass failure
The generic form of the op is too verbose and in some cases not
readable. On pass failure, ops have been so far printed in generic form
to provide a (stronger) guarantee that the IR print succeeds. However,
in a large number of pass failure cases, the IR is still valid and
the custom printers for the ops will succeed. In fact, readability is
highly desirable post pass failure. This revision provides an option to
print ops in their custom/pretty-printed form on IR failure -- this
option is unsafe and there is no guarantee it will succeed. It's
disabled by default and can be turned on only if needed.
Differential Revision: https://reviews.llvm.org/D123893
Christudasan Devadasan [Wed, 13 Apr 2022 11:39:45 +0000 (17:09 +0530)]
[AMDGPU][SIFrameLowering] Refactor custom SGPR spills (NFC).
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D123666
Bill Wendling [Sun, 17 Apr 2022 06:24:48 +0000 (23:24 -0700)]
[randstruct] Enforce using a designated init for a randomized struct
A randomized structure needs to use a designated or default initializer.
Using a non-designated initializer will result in values being assigned
to the wrong fields.
Differential Revision: https://reviews.llvm.org/D123763
hsmahesha [Sun, 17 Apr 2022 05:14:40 +0000 (10:44 +0530)]
[AMDGPU] Pre-checkin updated lit tests for D123525.
Fix indentation within the lit test - agpr-copy-no-free-registers.ll.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D123809
Jacques Pienaar [Sun, 17 Apr 2022 04:59:18 +0000 (21:59 -0700)]
[mlir][docs] Add missing directory separator
Lang Hames [Sun, 17 Apr 2022 01:49:47 +0000 (18:49 -0700)]
[JITLink] Error instead of asserting on unrecognized edge kinds.
It's idiomatic to require that plugins (especially platform plugins) be
installed to handle special edge kinds. If the plugins are not installed and an
object is loaded that uses one of the special edge kinds then we want to error
out rather than asserting.
Saleem Abdulrasool [Sun, 17 Apr 2022 00:57:00 +0000 (17:57 -0700)]
Windows: correct iteration of additional search paths
This adjusts the path iteration - `paths` is a null-terminated sequence
of C strings, creating an array from a single contiguous buffer. We
would previously continue to iterate indefinitely as we did not check if
we had encountered the terminator.
Found by inspection.
Jonathan Wakely [Sat, 16 Apr 2022 21:01:05 +0000 (23:01 +0200)]
[libcxx] [test] Use LIBCPP_ASSERT for libc++-specific checks
This makes it a little easier to reuse the libc++ tests for another
std::lib (specifically libstdc++).
The regex_error::what() strings are different in other
implementations, so should be tested with LIBCPP_ASSERT so that those
checks are skipped for other implementations.
The value of ECMAScript constant is non-zero in other implementations,
and conditionally for libc++, so adjust the preprocessor condition for
that too.
Reviewed By: ldionne, #libc, philnik, Mordante
Spies: Mordante, philnik, libcxx-commits
Differential Revision: https://reviews.llvm.org/D122574
Peter Klausler [Wed, 23 Mar 2022 21:05:50 +0000 (14:05 -0700)]
[flang] Add & use a better visit() (take 2)
Adds flang/include/flang/Common/log2-visit.h, which defines
a Fortran::common::visit() template function that is a drop-in
replacement for std::visit(). Modifies most use sites in
the front-end and runtime to use common::visit().
The C++ standard mandates that std::visit() have O(1) execution
time, which forces implementations to build dispatch tables.
This new common::visit() is O(log2 N) in the number of alternatives
in a variant<>, but that N tends to be small and so this change
produces a fairly significant improvement in compiler build
memory requirements, a 5-10% improvement in compiler build time,
and a small improvement in compiler execution time.
Building with -DFLANG_USE_STD_VISIT causes common::visit()
to be an alias for std::visit().
Calls to common::visit() with multiple variant arguments
are referred to std::visit(), pending further work.
This change is enabled only for GCC builds with GCC >= 9;
an earlier attempt (D122441) ran into bugs in some versions of
clang and was reverted rather than simply disabled; and it is
not well tested with MSVC. In non-GCC and older GCC builds,
common::visit() is simply an alias for std::visit().
Craig Topper [Sat, 16 Apr 2022 21:08:04 +0000 (14:08 -0700)]
[X86] Move some hasOneUse checks after checking what the opcode is.
Calling hasOneUse can be expensive on nodes with multiple results.
Especially when some results are Chains. By checking the opcode first,
we can avoid walking the uses if it isn't an interesting node,
and thus avoid calling hasOneUse on a node that might have many uses.
Found by profiling the IR given in D123857.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D123881
Lang Hames [Sat, 16 Apr 2022 21:06:27 +0000 (14:06 -0700)]
[ORC] Make IRSpeculationLayer::BaseLayer an IRLayer.
BaseLayer was originally written as an IRCompileLayer, but there was no need for
this restriction. Using IRLayer gives clients more flexibility in choosing the
underlying layer.
Andrew Litteken [Mon, 21 Mar 2022 20:56:50 +0000 (15:56 -0500)]
[IROutliner] Ensure that phi values that are passed in as arguments are remapped as arguments
Issue: https://github.com/llvm/llvm-project/issues/54430
For incoming values of phi nodes added to an outlined function to accommodate different exit paths in the function, when a value is a constant that is passed into the outlined function as an argument, we find the corresponding value in the first extracted function used to fill the overall outlined function. When this value is an argument, the corresponding value used will be the old value, prior to outlining. This patch maintains a mapping from these values to arguments, and uses this mapping to update the added phi node accordingly.
Reviewers: paquette
Recommit of
d6eb480afbc038871570fa053d772c913cd77a61
Differential Revision: https://reviews.llvm.org/D122206
eop Chen [Sat, 16 Apr 2022 19:10:58 +0000 (12:10 -0700)]
[LSR] Update outdated comment
Valentin Clement [Sat, 16 Apr 2022 19:08:16 +0000 (21:08 +0200)]
[mlir][CSE] Add ability to remove commutative operations
This patch takes advantage of the Commutative trait on operation
to remove identical commutative operations where the operands are swapped.
The second operation below can be removed since `arith.addi` is commutative.
```
%1 = arith.addi %a, %b : i32
%2 = arith.addi %b, %a : i32
```
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D123492
Nico Weber [Sat, 16 Apr 2022 16:58:04 +0000 (12:58 -0400)]
[gn build] (manually) port
f2526c1a5c6f
Aaron Ballman [Sat, 16 Apr 2022 15:11:32 +0000 (11:11 -0400)]
Revert "[randstruct] Enforce using a designated init for a randomized struct"
This reverts commit
aed923b1246ac38335b222b89594516fcf0d6385.
It causes some buildbot test failures.
Aaron Ballman [Sat, 16 Apr 2022 15:10:57 +0000 (11:10 -0400)]
Revert "[randstruct] Force errors for all platforms"
This reverts commit
2a404cdfd8bc75de593ce0e15fff0a7a0a18ec1c.
It didn't address the built bot failures.
https://lab.llvm.org/buildbot/#/builders/171/builds/13231
https://lab.llvm.org/buildbot/#/builders/186/builds/5520
Nathan James [Sat, 16 Apr 2022 08:53:32 +0000 (09:53 +0100)]
[clang-tidy] Add a Standalone diagnostics mode to clang-tidy
Adds a flag to `ClangTidyContext` that is used to indicate to checks that fixes will only be applied one at a time.
This is to indicate to checks that each fix emitted should not depend on any other fixes emitted across the translation unit.
I've currently implemented the `IncludeInserter`, `LoopConvertCheck` and `PreferMemberInitializerCheck` to use these support these modes.
Reasoning behind this is in use cases like `clangd` it's only possible to apply one fix at a time.
For include inserter checks, the include is only added once for the first diagnostic that requires it, this will result in subsequent fixes not having the included needed.
A similar issue is seen in the `PreferMemberInitializerCheck` where the `:` will only be added for the first member that needs fixing.
Fixes emitted in `StandaloneDiagsMode` will likely result in malformed code if they are applied all together, conversely fixes currently emitted may result in malformed code if they are applied one at a time.
For this reason invoking `clang-tidy` from the binary will always with `StandaloneDiagsMode` disabled, However using it as a library its possible to select the mode you wish to use, `clangd` always selects `StandaloneDiagsMode`.
This is an example of the current behaviour failing
```lang=c++
struct Foo {
int A, B;
Foo(int D, int E) {
A = D;
B = E; // Fix Here
}
};
```
Incorrectly transformed to:
```lang=c++
struct Foo {
int A, B;
Foo(int D, int E), B(E) {
A = D;
// Fix Here
}
};
```
In `StandaloneDiagsMode`, it gets transformed to:
```lang=c++
struct Foo {
int A, B;
Foo(int D, int E) : B(E) {
A = D;
// Fix Here
}
};
```
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D97121
Vitaly Buka [Sat, 16 Apr 2022 07:18:48 +0000 (00:18 -0700)]
Revert "Treat `std::move`, `forward`, and `move_if_noexcept` as builtins."
Revert "Extend support for std::move etc to also cover std::as_const and"
Revert "Update test to handle opaque pointers flag flip."
It crashes on libcxx tests https://lab.llvm.org/buildbot/#/builders/85/builds/8174
This reverts commit
fc3090109643af8d2da9822d0f99c84742b9c877.
This reverts commit
a571f82a50416b767fd3cce0fb5027bb5dfec58c.
This reverts commit
64c045e25b8471bbb572bd29159c294a82a86a25.
Mehdi Amini [Mon, 4 Apr 2022 00:32:38 +0000 (00:32 +0000)]
Apply clang-tidy fixes for readability-identifier-naming in OpFormatGen.cpp (NFC)
Mehdi Amini [Mon, 4 Apr 2022 00:31:56 +0000 (00:31 +0000)]
Apply clang-tidy fixes for llvm-qualified-auto in OpFormatGen.cpp (NFC)
River Riddle [Wed, 19 Jan 2022 00:16:54 +0000 (16:16 -0800)]
[mlir] Add asserts when changing various MLIRContext configurations
This helps to prevent tsan failures when users inadvertantly mutate the
context in a non-safe way.
Differential Revision: https://reviews.llvm.org/D112021
chenglin.bi [Sat, 16 Apr 2022 04:29:11 +0000 (12:29 +0800)]
[Arch64][SelectionDAG] Add target-specific implementation of srem
X%C to the equivalent of X-X/C*C is not always fastest path if there is no SDIV pair exist. So check target have faster for srem only first. Add AArch64 faster path for SREM only pow2 case.
Fix https://github.com/llvm/llvm-project/issues/54649
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D122968
Fangrui Song [Sat, 16 Apr 2022 04:07:50 +0000 (21:07 -0700)]
[msan][test] Remove legacy PM style opt -foo tests
Peter Klausler [Tue, 12 Apr 2022 22:15:46 +0000 (15:15 -0700)]
[flang][runtime] Don't emit empty lines for bad writes
When an error occurs in a formatted sequential output statement
and no output was ever emitted, don't emit a blank record.
This matches the error case behavior of other Fortran compilers.
Differential Revision: https://reviews.llvm.org/D123734
Jun Zhang [Sat, 16 Apr 2022 03:27:18 +0000 (11:27 +0800)]
Add some helpers to better check Scope's kind. NFC
Signed-off-by: Jun Zhang <jun@junz.org>
Peter Klausler [Tue, 12 Apr 2022 21:08:04 +0000 (14:08 -0700)]
[flang] Accept %KIND type parameter inquiries on %RE,%IM, &c.
The x%KIND inquiry needs to be supported when 'x' is itself
a complex part reference or a type parameter inquiry.
Differential Revision: https://reviews.llvm.org/D123733
Peter Klausler [Tue, 12 Apr 2022 01:57:59 +0000 (18:57 -0700)]
[flang] Allow POINTER attribute statement on procedure interfaces
A POINTER attribute statement is allowed to add the POINTER attribute
to a procedure entity that has already been declared, e.g. with an
INTERFACE block.
Differential Revision: https://reviews.llvm.org/D123732
Fangrui Song [Sat, 16 Apr 2022 03:14:07 +0000 (20:14 -0700)]
[test] Test -Werror=foo -Wfoo & -Werror -Wno-error=foo -Wfoo
Clang now matches GCC in the two cases.
D109981 fixed the second case though it did not call out the effect.
Peter Klausler [Tue, 12 Apr 2022 01:39:02 +0000 (18:39 -0700)]
[flang] Upgrade short actual character arguments to errors
f18 was emitting a warning about short character actual arguments to
subprograms and statement functions; every other compiler considers this
case to be an error.
Differential Revision: https://reviews.llvm.org/D123731
Fangrui Song [Sat, 16 Apr 2022 03:00:56 +0000 (20:00 -0700)]
[asan][test] Remove legacy PM style opt -foo tests
Weining Lu [Sat, 9 Apr 2022 08:52:24 +0000 (16:52 +0800)]
[LoongArch] Fix typo that SP should be R3 but not R2
Peter Klausler [Tue, 12 Apr 2022 00:12:39 +0000 (17:12 -0700)]
[flang][runtime] Fix ENDFILE for formatted stream output
A predicate expression made ENDFILE statements significant
only for sequential files, but it's applicable to formatted
stream output as well.
Differential Revision: https://reviews.llvm.org/D123730
Shilei Tian [Sat, 16 Apr 2022 01:34:19 +0000 (21:34 -0400)]
[Clang][OpenMP] Use bitfields for flags in `OMPAtomicDirective`
As suggested in D120290.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D123862
Joseph Huber [Sat, 16 Apr 2022 01:18:56 +0000 (21:18 -0400)]
[OpenMP] Fix linting diagnostics in the linker wrapper
Summary:
A previous patch had some linter warnings that should've been addressed.
Chris Bieneman [Sat, 16 Apr 2022 01:12:52 +0000 (20:12 -0500)]
Blind stab in the dark to fix a bot failure
*fingers crossed*
Mogball [Sat, 16 Apr 2022 00:56:38 +0000 (00:56 +0000)]
[NFC] fix cmake build
Mehdi Amini [Mon, 4 Apr 2022 00:28:26 +0000 (00:28 +0000)]
Apply clang-tidy fixes for readability-identifier-naming in mlir-parser-fuzzer.cpp (NFC)
Mehdi Amini [Mon, 4 Apr 2022 00:27:44 +0000 (00:27 +0000)]
Apply clang-tidy fixes for readability-identifier-naming in DummyParserFuzzer.cpp (NFC)
Craig Topper [Sat, 16 Apr 2022 00:32:42 +0000 (17:32 -0700)]
[IR] Allow constant folding (insertelement <vscale x 2 x i32> zeroinitializer, i32 0, i32 i32 0.
Most of insertelement constant folding is blocked if the vector type
is scalable. I believe we can make an exception for inserting null
into an all zeros vector.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D123413
Mogball [Sat, 16 Apr 2022 00:36:22 +0000 (00:36 +0000)]
[mlir] Refactor LICM into a utility
LICM is refactored into a utility that is application on any region. The implementation is moved to Transform/Utils.
Joseph Huber [Fri, 25 Mar 2022 15:03:02 +0000 (11:03 -0400)]
[OpenMP] Use new offloading binary when embedding offloading images
The previous patch introduced the offloading binary format so we can
store some metada along with the binary image. This patch introduces
using this inside the linker wrapper and Clang instead of the previous
method that embedded the metadata in the section name.
Differential Revision: https://reviews.llvm.org/D122683
Joseph Huber [Wed, 13 Apr 2022 17:43:47 +0000 (13:43 -0400)]
[OpenMP] Don't manually strip sections in the linker wrapper
Summary:
The changes in D122987 ensures that the offloading sections always have
the SHF_EXCLUDE flag. This means that we do not need to manually strip
these sections for ELF or COFF targets.
Matt Arsenault [Thu, 14 Apr 2022 14:59:37 +0000 (10:59 -0400)]
llvm-reduce: Clone some of the easy function properties
Error on some of these other fields, since tracking down test cases
for all of these at once is exhausting.
Matt Arsenault [Thu, 14 Apr 2022 20:16:08 +0000 (16:16 -0400)]
MIR: Serialize a few bool function fields
Craig Topper [Sat, 16 Apr 2022 00:19:07 +0000 (17:19 -0700)]
[X86] Move hasOneUse check after opcode check. NFC
Checking opcode is cheap. hasOneUse might not be if the node has
multiple results. By checking the opcode we can rule out nodes
with multiple results we aren't interested in.
Stella Stamenova [Sat, 16 Apr 2022 00:09:05 +0000 (17:09 -0700)]
Revert "[mlir] Refactor LICM into a utility"
This reverts commit
3131f808243abe3746280e016ab9459c14d9e53b.
This commit broke the Windows mlir bot:
https://lab.llvm.org/buildbot/#/builders/13/builds/19745
Craig Topper [Sat, 16 Apr 2022 00:01:40 +0000 (17:01 -0700)]
[DAGCombiner] Move call to hasOneUse after opcode checks. NFC
Checking the opcode is cheap, counting the number of uses is not.
Chris Bieneman [Thu, 14 Apr 2022 18:37:44 +0000 (13:37 -0500)]
Add DXIL Bitcode Writer and DXIL testing
This change is a big blob of code that isn't easy to break up. It
either comes in all together as a blob, works and has tests, or it
doesn't do anything.
Logically you can think of this patch as three things:
(1) Adding virtual interfaces so the bitcode writer can be overridden
(2) Adding a new bitcode writer implementation for DXIL
(3) Adding some (optional) crazy CMake goop to build the
DirectXShaderCompiler's llvm-dis as dxil-dis for testing
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D122082
Craig Topper [Fri, 15 Apr 2022 23:33:12 +0000 (16:33 -0700)]
[DAGCombiner] Move or/xor/and opcode check in ReduceLoadOpStoreWidth before hasOneUse check.
hasOneUse is not cheap on nodes with chain results that might have
many uses. By checking the opcode first, we can avoid a costly walk
of the use list on nodes we aren't interested in.
Found by investigating calls to hasNUsesOfValue from the example
provided in D123857.
Johannes Doerfert [Fri, 15 Apr 2022 23:29:22 +0000 (18:29 -0500)]
[Attributor][FIX] Use AttributorConfig in the unit tests too
Richard Smith [Fri, 15 Apr 2022 21:45:30 +0000 (14:45 -0700)]
Extend support for std::move etc to also cover std::as_const and
std::addressof, plus the libstdc++-specific std::__addressof.
This brings us to parity with the corresponding GCC behavior.
Remove STDBUILTIN macro that ended up not being used.
Peter Klausler [Wed, 13 Apr 2022 17:39:16 +0000 (10:39 -0700)]
[flang] Handle parameter-dependent types in PDT initializers
For parameterized derived type component initializers whose
expressions' types depend on parameter values, f18's current
scheme of analyzing the initialization expression once during
name resolution fails. For example,
type :: pdt(k)
integer, kind :: k
real :: component = real(0.0, kind=k)
end type
To handle such cases, it is necessary to re-analyze the parse
trees of these initialization expressions once for each distinct
initialization of the type.
This patch adds code to wipe an expression parse tree of its
typed expressions, and update those of its symbol table pointers
that reference type parameters, and then re-analyze that parse
tree to generate the properly typed component initializers.
Differential Revision: https://reviews.llvm.org/D123728
Johannes Doerfert [Fri, 15 Apr 2022 21:15:07 +0000 (16:15 -0500)]
[Attributor][NFCI] Introduce AttributorConfig to bundle all options
Instead of lengthy constructors we can now set the members of a
read-only struct before the Attributor is created. Should make it
clearer what is configurable and also help introducing new options in
the future. This actually added IsModulePass and avoids deduction
through the Function set size. No functional change was intended.
Bill Wendling [Fri, 15 Apr 2022 22:11:29 +0000 (15:11 -0700)]
[randstruct] Force errors for all platforms
Mogball [Fri, 15 Apr 2022 17:52:34 +0000 (17:52 +0000)]
[mlir] Refactor LICM into a utility
LICM is refactored into a utility that is application on any region. The implementation is moved to Transform/Utils.
Richard Smith [Fri, 15 Apr 2022 21:51:30 +0000 (14:51 -0700)]
Update test to handle opaque pointers flag flip.
Pavel Kosov [Fri, 15 Apr 2022 21:48:38 +0000 (23:48 +0200)]
[llvm-objdump] Implemented PrintBranchImmAsAddress for MIPS
Updated MipsInstPrinter to print absolute hex offsets for branch instructions.
It is necessary to make the llvm-objdump output close to the gnu objdump output.
This implementation is based on the implementation for RISC-V.
OS Laboratory. Huawei Russian Research Institute. Saint-Petersburg
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D123764
Vitaly Buka [Fri, 15 Apr 2022 20:54:35 +0000 (13:54 -0700)]
[msan] Set poison_in_dtor=1 by default
It's still disabled by default at compile time.
Reviewed By: kstoimenov
Differential Revision: https://reviews.llvm.org/D123875
Peter Klausler [Fri, 15 Apr 2022 20:23:16 +0000 (13:23 -0700)]
[flang] Finer control over error recovery with GetExpr()
Prior to this patch, the semantics utility GetExpr() will crash
unconditionally if it encounters a typed expression in the parse
tree that has not been set by expression semantics. This is the
right behavior when called from lowering, by which time it is known
that the program had no fatal user errors, since it signifies a
fatal internal error. However, prior to lowering, in the statement
semantics checking code, a more nuanced test should be used before
crashing -- specifically, we should not crash in the face of a
missing typed expression when in error recovery mode.
Getting this right requires GetExpr() and its helper class to have
access to the semantics context, so that it can check AnyFatalErrors()
before crashing. So this patch touches nearly all of its call sites.
Differential Revision: https://reviews.llvm.org/D123873
Richard Smith [Fri, 1 Apr 2022 01:40:40 +0000 (18:40 -0700)]
Treat `std::move`, `forward`, and `move_if_noexcept` as builtins.
We still require these functions to be declared before they can be used,
but don't instantiate their definitions unless their addresses are
taken. Instead, code generation, constant evaluation, and static
analysis are given direct knowledge of their effect.
This change aims to reduce various costs associated with these functions
-- per-instantiation memory costs, compile time and memory costs due to
creating out-of-line copies and inlining them, code size at -O0, and so
on -- so that they are not substantially more expensive than a cast.
Most of these improvements are very small, but I measured a 3% decrease
in -O0 object file size for a simple C++ source file using the standard
library after this change.
We now automatically infer the `const` and `nothrow` attributes on these
now-builtin functions, in particular meaning that we get a warning for
an unused call to one of these functions.
In C++20 onwards, we disallow taking the addresses of these functions,
per the C++20 "addressable function" rule. In earlier language modes, a
compatibility warning is produced but the address can still be taken.
The same infrastructure is extended to the existing MSVC builtin
`__GetExceptionInfo`, which is now only recognized in namespace `std`
like it always should have been.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D123345
Florian Hahn [Fri, 15 Apr 2022 20:52:46 +0000 (22:52 +0200)]
[VPlan] Handle equal address and store ops in onlyFirstLaneDemanded.
With opaque pointers, the stored value and address can be the same.
Previously the code in VPWidenMemoryInstructionRecipe::onlyFirstLaneDemanded
incorrectly considers stores with matching store and pointer operands as
only demanding the first lane, causing a crash.
Chih-Ping Chen [Wed, 6 Apr 2022 12:22:49 +0000 (08:22 -0400)]
[DebugInfo] Add a TargetFuncName field in DISubprogram for
specifying DW_AT_trampoline as a string. Also update the signature
of DIBuilder::createFunction to reflect this addition.
Differential Revision: https://reviews.llvm.org/D123697
Johannes Doerfert [Fri, 15 Apr 2022 20:29:20 +0000 (15:29 -0500)]
Revert "[Attributor] CGSCC pass should not recompute results outside the SCC"
This reverts commit
0d7f81e31315f8cda56ce6fde5ff5145e0325c51, it caused
the AMDGPU tests that use the Attributor to fail.
Lang Hames [Fri, 15 Apr 2022 19:57:19 +0000 (12:57 -0700)]
[JITLink] Add missing moves from
43acef48d38e.
River Riddle [Fri, 15 Apr 2022 19:55:15 +0000 (12:55 -0700)]
[mlir][NFC] Cleanup the TestClone pass
Fix variable naming convention and cleanup a clang-tidy warning.
River Riddle [Fri, 8 Apr 2022 04:36:40 +0000 (21:36 -0700)]
[mlir] Remove the use of FilterTypes for template metaprogramming
This technique results in an explosion in compile time, resulting from a
huge number of std::tuple/concat instatiations. This technique is replaced
by simpler metaprogramming and results in a signficant reduction in
compile time. A local debug/asan build saw a 4x speed up in the processing
of ArithmeticOps.h.inc, and given the nature of this change every dialect
should see similar reductions in compile time.
Differential Revision: https://reviews.llvm.org/D123360
Johannes Doerfert [Tue, 12 Apr 2022 21:38:11 +0000 (16:38 -0500)]
[Attributor][NFC] Introduce a flag to distinguish the scope of a query
Johannes Doerfert [Tue, 12 Apr 2022 21:24:37 +0000 (16:24 -0500)]
[Attributor] CGSCC pass should not recompute results outside the SCC
When we run the CGSCC pass we should only invest time on the SCC. We can
initialize AAs with information from the module slice but we should not
update those AAs.
Johannes Doerfert [Tue, 12 Apr 2022 21:05:32 +0000 (16:05 -0500)]
[Attributor][NFC] Code cleanup to minimize follow up changes
Johannes Doerfert [Sat, 9 Apr 2022 16:45:32 +0000 (11:45 -0500)]
[Attributor][NFC] Rename AAPotentialValues to AAPotentialConstantValues
Lang Hames [Fri, 15 Apr 2022 18:22:26 +0000 (11:22 -0700)]
[JITLink] Refactor and expand DWARF pointer encoding support.
Adds support for pointer encodings commonly used in large/static models,
including non-pcrel, sdata/udata8, indirect, and omit.
Also refactors pointer-encoding handling to consolidate error generation inside
common functions, rather than callees of those functions.
Arthur Eubanks [Fri, 15 Apr 2022 19:38:11 +0000 (12:38 -0700)]
[test][LoopDeletion] Precommit test
Arjun P [Tue, 12 Apr 2022 12:07:25 +0000 (13:07 +0100)]
[MLIR][Presburger] addSymbolicCut: fix the integral symbols heuristic to match the docs
Previously this checked if the entire symbolic numerator was divisible by the
denominator, which is never the case when this function is called. Fixed this to
check only the non-const coefficients in the numerator, which was what was
intended and documented.
Reviewed By: Groverkss
Differential Revision: https://reviews.llvm.org/D123592
Bill Wendling [Fri, 15 Apr 2022 18:37:52 +0000 (11:37 -0700)]
[randstruct] Enforce using a designated init for a randomized struct
A randomized structure needs to use a designated or default initializer.
Using a non-designated initializer will result in values being assigned
to the wrong fields.
Differential Revision: https://reviews.llvm.org/D123763
LLVM GN Syncbot [Fri, 15 Apr 2022 19:23:18 +0000 (19:23 +0000)]
[gn build] Port
721651be246e
Thomas Raoux [Fri, 15 Apr 2022 18:44:43 +0000 (18:44 +0000)]
[mlir][vector] Fix bug in extractFromBroadcast folding
extract was incorrectly folded when the source was coming from a
broadcast that was both adding new rank and broadcasting the inner
dimension.
Differential Revision: https://reviews.llvm.org/D123867
Alexandre Ganea [Mon, 11 Apr 2022 21:22:53 +0000 (17:22 -0400)]
[Support][cmake] Fix snmalloc integration. NFC.
When using LLVM_INTEGRATED_CRT_ALLOC, fix compiling with the latest snmalloc at ToT (https://github.com/microsoft/snmalloc).
Xiang Li [Fri, 15 Apr 2022 19:09:10 +0000 (14:09 -0500)]
[HLSL][clang][Driver] Support target profile command line option.
The target profile option(/T) decide the shader model when compile hlsl.
The format is shaderKind_major_minor like ps_6_1.
The shader model is saved as llvm::Triple is clang/llvm like
dxil-unknown-shadermodel6.1-hull.
The main job to support the option is translating ps_6_1 into
shadermodel6.1-pixel.
That is done inside tryParseProfile at HLSL.cpp.
To integrate the option into clang Driver, a new DriverMode DxcMode is
created. When DxcMode is enabled, OSType for TargetTriple will be
forced into Triple::ShaderModel. And new ToolChain HLSLToolChain will
be created when OSType is Triple::ShaderModel.
In HLSLToolChain, ComputeEffectiveClangTriple is overridden to call
tryParseProfile when targetProfile option is set.
To make test work, Fo option is added and .hlsl is added for active
-xhlsl.
Reviewed By: beanz
Differential Revision: https://reviews.llvm.org/D122865
Patch by: Xiang Li <python3kgae@outlook.com>
Arjun P [Tue, 12 Apr 2022 13:20:32 +0000 (14:20 +0100)]
[MLIR][Presburger][Simplex] moveRowUnknownToColumn: support the row sample value being zero
When the sample value is zero, everything is the same except that failure to
pivot does not imply emptiness. So, leave it to the user to mark as empty if
necessary, if they know the sample value is strictly negative. This is needed
for an upcoming symbolic lexmin heuristic.
Reviewed By: Groverkss
Differential Revision: https://reviews.llvm.org/D123604
William S. Moses [Fri, 15 Apr 2022 19:14:06 +0000 (15:14 -0400)]
[MLIR][ClonePass] Attempt fix for anonymous pass name
Eli Friedman [Thu, 14 Apr 2022 22:20:37 +0000 (15:20 -0700)]
Fix size of flexible array initializers, and re-enable assertions.
In D123649, I got the formula for getFlexibleArrayInitChars slightly
wrong: the flexible array elements can be contained in the tail padding
of the struct. Fix the formula to account for that.
With the fixed formula, we run into another issue: in some cases, we
were emitting extra padding for flexible arrray initializers. Fix
CGExprConstant so it uses a packed struct when necessary, to avoid this
extra padding.
Differential Revision: https://reviews.llvm.org/D123826
Zequan Wu [Fri, 15 Apr 2022 19:08:38 +0000 (12:08 -0700)]
rdzhabarov [Fri, 15 Apr 2022 18:59:03 +0000 (18:59 +0000)]
[mlir] Fix BUILD issues and dependencies.
Differential Revision: https://reviews.llvm.org/D123868
Zequan Wu [Fri, 15 Apr 2022 18:59:00 +0000 (11:59 -0700)]
[LLDB][NativePDB] Don't create inlined function parameters when it's malformed.
Johannes Doerfert [Wed, 13 Apr 2022 19:39:15 +0000 (14:39 -0500)]
[DWARF][FIX] Handle the use of multiple registers gracefully
Certain applications crashed for us with the AMDGPU backend. While this
is not a proper fix it allows us to compile the code for now. I left a
TODO for someone that understands DWARF.
Differential Revision: https://reviews.llvm.org/D123717
Johannes Doerfert [Wed, 13 Apr 2022 17:35:34 +0000 (12:35 -0500)]
[AMDGPU][FIX] Proper load-store-vectorizer result with opaque pointers
The original code relied on the fact that we needed a bitcast
instruction (for non constant base objects). With opaque pointers there
might not be a bitcast. Always check if reordering is required instead.
Fixes: https://github.com/llvm/llvm-project/issues/54896
Differential Revision: https://reviews.llvm.org/D123694
William S. Moses [Fri, 15 Apr 2022 18:32:44 +0000 (14:32 -0400)]
[NFC] Update comments
Aaron Ballman [Fri, 15 Apr 2022 18:03:46 +0000 (14:03 -0400)]
Fix an edge case in determining is a function has a prototype
Given the declaration:
typedef void func_t(unsigned);
__attribute__((noreturn)) func_t func;
we would incorrectly determine that `func` had no prototype because the
`noreturn` attribute would convert the underlying type directly into a
FunctionProtoType, but the declarator for `func` itself was not one for
a function with a prototype. This adds an additional check for when the
declarator is a type representation for a function with a prototype.
Zequan Wu [Fri, 15 Apr 2022 17:36:06 +0000 (10:36 -0700)]
[LLDB][NativePDB] Fix subfield_register_simple_type.s test
Mogball [Tue, 12 Apr 2022 00:50:20 +0000 (00:50 +0000)]
[mlir] Update LICM to support Graph Regions
Changes the algorithm of LICM to support graph regions (no guarantee of topologically sorted order). Also fixes an issue where ops with recursive side effects and regions would not be hoisted if any nested ops used operands that were defined within the nested region.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D122465
Fangrui Song [Fri, 15 Apr 2022 17:26:43 +0000 (10:26 -0700)]
[PGO] Remove legacy PM passes
Legacy PM for optimization pipeline was deprecated in 13.0.0 and Clang dropped
legacy PM support in D123609. This change removes legacy PM passes for PGO so
that downstream projects won't be able to use it. It seems appropriate to start
removing such "add-on" features like instrumentations, before we remove more
stuff after 15.x is branched.
I have checked many LLVM users and only ldc[1] uses the legacy PGO pass.
[1]: https://github.com/ldc-developers/ldc/issues/3961
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D123834
William S. Moses [Sat, 26 Mar 2022 18:42:15 +0000 (14:42 -0400)]
[MLIR] Fix operation clone
Operation clone is currently faulty.
Suppose you have a block like as follows:
```
(%x0 : i32) {
%x1 = f(%x0)
return %x1
}
```
The test case we have is that we want to "unroll" this, in which we want to change this to compute `f(f(x0))` instead of just `f(x0)`. We do so by making a copy of the body at the end of the block and set the uses of the argument in the copy operations with the value returned from the original block.
This is implemented as follows:
1) map to the block arguments to the returned value (`map[x0] = x1`).
2) clone the body
Now for this small example, this works as intended and we get the following.
```
(%x0 : i32) {
%x1 = f(%x0)
%x2 = f(%x1)
return %x2
}
```
This is because the current logic to clone `x1 = f(x0)` first looks up the arguments in the map (which finds `x0` maps to `x1` from the initialization), and then sets the map of the result to the cloned result (`map[x1] = x2`).
However, this fails if `x0` is not an argument to the op, but instead used inside the region, like below.
```
(%x0 : i32) {
%x1 = f() {
yield %x0
}
return %x1
}
```
This is because cloning an op currently first looks up the args (none), sets the map of the result (`map[%x1] = %x2`), and then clones the regions. This results in the following, which is clearly illegal:
```
(%x0 : i32) {
%x1 = f() {
yield %x0
}
%x2 = f() {
yield %x2
}
return %x2
}
```
Diving deeper, this is partially due to the ordering (how this PR fixes it), as well as how region cloning works. Namely it will first clone with the mapping, and then it will remap all operands. Since the ordering above now has a map of `x0 -> x1` and `x1 -> x2`, we end up with the incorrect behavior here.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D122531
Peter Klausler [Wed, 6 Apr 2022 22:55:53 +0000 (15:55 -0700)]
[flang] Fix Symbol::Rank for ProcEntityDetails
When a procedure pointer or procedure dummy argument has a
defined interface, the rank of the pointer (or dummy) is the
rank of the interface.
Also tweak code discovered in shape analysis when investigating
this problam so that it returns a vector of emptied extents rather
than std::nullopt when the extents are not scope-invariant, so that
the rank can at least be known.
Differential Revision: https://reviews.llvm.org/D123727
jfurtek [Fri, 15 Apr 2022 16:41:56 +0000 (16:41 +0000)]
[mlir][ods][NFC] Move enum attribute definitions from OpBase.td to EnumAttr.td
This diff moves `EnumAttr` tablegen definitions (specifically, `IntEnumAttr` and
`BitEnumAttr`-related classes) from `OpBase.td` to `EnumAttr.td`. No
functionality is changed.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D123551
Alex Zinenko [Fri, 15 Apr 2022 10:00:55 +0000 (12:00 +0200)]
[mlir] Support opaque types in LLVM IR -> MLIR translation
LLVM IR is moving towards adoption of opaque pointer types. These require extra
information to be passed when constructing some operations, in particular GEP
and Alloca. Adapt the builders of said operations and modify the translation
code to handle both opaque and non-opaque pointers.
This incidentally adds the translation for Alloca alignment and fixes the translation
of struct-related GEP indices that must be constant.
Reviewed By: wsmoses
Differential Revision: https://reviews.llvm.org/D123792
Aaron Ballman [Fri, 15 Apr 2022 15:45:52 +0000 (11:45 -0400)]
Properly identify builtins in a diagnostic note
When emitting a "conflicting types" warning for a function declaration,
it's more clear to diagnose the previous declaration specifically as
being a builtin if it one.
Shilei Tian [Fri, 15 Apr 2022 15:39:04 +0000 (11:39 -0400)]
Clean up `OMPAtomicDirective::Create`