After submitting the DataLayout fix, some tests fail when they didn't
before. This has to do with the target essentially being ignored when
these tests were run earlier, as the --target x86-unknown-linux-gnu
only has to be correctly formed to be accepted.
Now the target triple is actually being used to get the targetmachine
earlier - before MLIR is generated - so the test that has a valid target
but not available on the platform fails.
Fix is to require x86 registered target when running those tests.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D137335
This reverts commit b4e1466c35d3ca3e04244e8e8b4ffaf0784d6d37.
This causes a crash while building the Linux kernel. See the original
Phabricator review for a reduced C and LLVM IR reproducer.
- Run auto updater for 'trunc-to-tbl.ll' and 'build-pair-isel.ll'
- Add the motivating test of D135102 to 'bitfield-insert.ll'
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D137296
PlatformDarwinKernel::GetSupportedArchitectures returns a list
of architectures that are possible for this platform; it was using
a compile-time check for the debug host to decide the list of arches
that were valid. This was copied from a codepath doing native process
debugging, and was clearly wrong for kernel debugging, but it had
not happened to cause problems so it went unnoticed for a long time.
Small NFC change to the logging messages of Target::SetArchitecture
to make them a little more explicit about how the architecture is
being modified/replaced.
Differential Revision: https://reviews.llvm.org/D137301
rdar://101690111
This allows users to restrict the transformation to a
subset of the functions in a module.
For example, a user might want to apply the transformation to
a module's entry point, but not to the calls in the module
because those calls might refer to external C functions outside
of their control.
Reviewed By: springerm, nicolasvasilache
Differential Revision: https://reviews.llvm.org/D137264
It is likley that subtargets act differently for a vector integer arithmetic instruction based on the LMUL.
This patch creates seperate SchedRead, SchedWrite, WriteRes, ReadAdvance for each relevant LMUL. It also
introduces the concept of an "UpperBound LMUL" which allows us to describe how an instruction should behave
when the LMUL is unknown. All base instructions use the UpperBound resources because they are not tied to
a specific LMUL. This gives subtargetes the flexibility to describe their own upper bounds on each vector
instruction.
I have a series of patches for the rest of the vector instruction set ready to go, but I would like to first
get feedback on the first one of the series (this one).
Differential Revision: https://reviews.llvm.org/D136730
As the other recent patches, this alters the arm_fp16 intrinsics to be
target based, not preprocessor based. Apparently arm_fp16.h is AArch64
only under clang, making this mostly trivial with the TargetGuard
infrastructure.
Differential Revision: https://reviews.llvm.org/D137256
So far the function argument attributes were only translated
for functions with bodies. This change makes sure that this
happens for functions without bodies (declarations) as well.
This is needed for https://github.com/llvm/llvm-project/issues/58579
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D137047
This adds an extra argument for specifying the pass manager's anchor op,
with a default of `any`. Previously the anchor was always defaulted to
`builtin.module`.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D136406
The pipeline string must now include the pass manager's anchor op. This
makes the parse API properly roundtrip the printed form of a pass
manager.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D136405
The pipeline string must now include the pass manager's anchor op. This
makes the parse API properly roundtrip the printed form of a pass
manager. Since this is already an API break, I also added an extra
callback argument which is used for reporting errors.
The old functionality of appending to an existing pass manager is
available through `mlirOpPassManagerAddPipeline`.
Reviewed By: mehdi_amini, ftynse
Differential Revision: https://reviews.llvm.org/D136403
Including the anchor op ensures that all pass manager settings are fully
specified, and makes the string consistent with the printed form.
Depends on D134622
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D134623
In D134622 the printed form of a pass manager is changed to include the
name of the op that the pass manager is anchored on. This updates the
`-pass-pipeline` argument format to include the anchor op as well, so
that the printed form of a pipeline can be directly passed to
`-pass-pipeline`. In most cases this requires updating
`-pass-pipeline='pipeline'` to
`-pass-pipeline='builtin.module(pipeline)'`.
This also fixes an outdated assert that prevented running a
`PassManager` anchored on `'any'`.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D134900
InstCombine does some basic store to load forwarding. One case it
currently misses is the case where the store is actually a memset.
This patch adds support for this case. This is a minimal
implementation that only handles a load at the memset base address,
without an offset.
GVN is already capable of performing this optimization. Having it
in InstCombine can help with phase ordering issues, similar to the
existing store to load forwarding.
Differential Revision: https://reviews.llvm.org/D137323
Summary:
llvm-readobj decide whether to decode the external symbol's function auxiliary entry based on whether symbol is function or not currently. But the XCOFFSymbolRef::isFunction() do not work properly when -ffunction-sections is enabled. we will not decode the function auxiliary entry based on the XCOFFSymbolRef::isFunction()
we will decode the function auxiliary entry based on following:
According to the https://www.ibm.com/docs/en/aix/7.2?topic=formats-xcoff-object-file-format#XCOFF__c0f91ad419jbau
In XCOFF32, there are only "one csect Auxiliary Entry" and "a function auxiliary symbol table entry" for the C_EXT, C_WEAKEXT, and C_HIDEXT Symbols. and By convention, the csect auxiliary entry in an XCOFF32 file must be the last auxiliary entry for any external symbol that has more than one auxiliary entry( that means for the C_EXT, C_WEAKEXT, and C_HIDEXT Symbols. if there more than one auxiliary Entries. we look the last one as csect auxiliary entry. and others auxiliary entries as function entries).
Reviewers: Hubert Tong, James Henderson
Differential Revision: https://reviews.llvm.org/D136950
CVP currently only tries to simplify comparisons if there is a
constant operand. However, even if both are non-constant, we may
be able to determine the result of the comparison based on range
information.
IPSCCP is already capable of doing this, but because it runs very
early, it may miss some cases.
Differential Revision: https://reviews.llvm.org/D137253
This patch is to address the test cases in which the load has to be inserted at a right point. This happens when there is a store b/w the loads.
This patch reverts the loads merge in all cases when stores are present b/w loads and will eventually be replaced with proper fix and test cases.
Differential Revision: https://reviews.llvm.org/D137333
This patch moves the createOffloadEntriesAndInfoMetadata to OpenMPIRBuilder,
the createOffloadEntry helper function. The clang specific error handling is
invoked using a callback. This code will also be used by flang in the future.
This test caught my attention because it's the only one in JITLink that XFAILs. Running it in isolation showed that the output doesn't meet the CHECK-LINES, i.e. the block address didn't match:
```
error: CHECK-NEXT: expected string not found in input
CHECK-NEXT: block 0xfff02000 size = 0x00000001, align = 16, alignment-offset = 0
<stdin>:22:2: note: possible intended match here
block 0xfff01000 size = 0x00000001, align = 16, alignment-offset = 0
```
Though, that doesn't appear to be the reason the test XFAILs. What we really want to check here is that llvm-jitlink doesn't fail with a duplicate section error yet.
In order to avoid issues like this in the future we can match a placeholder to check for some valid address within the slab (64Kb == last 4 digits).
The patch also drops the duplicate -noexec argument, removes an empty RUN-line, fixes indentation and adds a newline at EOF.
Reviewed By: sunho
Differential Revision: https://reviews.llvm.org/D137148
On Linux the `std::function` behaved differently to that on Darwin.
This patch removes usage of `std::function` in the test but attempts
to retain the test-coverage. We mainly want function types appearing
in the template argument and function argument lists.
Also add a `char const*` overload to one of the test functions to
cover the "format function argument using ValueObject formatter" code-path.
Differential Revision: https://reviews.llvm.org/D137272
A new register class as well as a number of related subregisters are being added
to Future CPU. These registers are Dense Math Registers (DMR) and are 1024 bits
long. These regsiters can also be used in consecutive pairs which leads to a
register that is 2048 bits.
This patch also adds 7 new instructions that use these registers. More
instructions will be added in future patches.
Reviewed By: amyk, saghir
Differential Revision: https://reviews.llvm.org/D136366
[module.unit]p1 specifies that module and import are invalid components
of a module name, that module names cannot contain reserved
identifiers, and that std followed by zero or more digits is reserved.
The first issue (module and import pseudo-keywords) requires a
diagnostic, the second issue (use of reserved identifiers) does not
require a diagnostic. We diagnose both the same -- the code is ill-
formed unless the module declaration is in a system "header". This
allows STL implementations to use the reserved module names while
preventing users from stealing them out from under us.
Differential Revision: https://reviews.llvm.org/D136953
They were annotated with TrailingAnnotation, which they are not. And
that resulted in some quirky formatting in some cases.
Differential Revision: https://reviews.llvm.org/D136635
Without the patch UnwrappedLineFormatter::analyzeSolutionSpace just ran
out of possible formattings and would put everything just on one line.
The problem was that the the line break was forbidden, but putting the
conditional colon on the same line is also forbidden.
Differential Revision: https://reviews.llvm.org/D135918
The Assignment Tracking debug-info feature is outlined in this RFC:
https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir
Add the DIAssignID metadata attachment boilerplate. Includes a textual-bitcode
roundtrip test and tests that the verifier and parser catch badly formed IR.
This piece of metadata links together stores (used as an attachment) and the
yet-to-be-added llvm.dbg.assign debug intrinsic (used as an operand).
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D132222
The problem with GCNDownwardRPTracker::advanceBeforeNext is that it doesn't allow to get register pressure
after the last instruction in a MBB.
However when we track RP through the boundary of a MBB we need the state that is after the last instruction
of the MBB and before the first instruction of the successor MBB. Currently we stop traking RP in the state
'at' the last instruction of the MBB which is incorrect.
This patch fixes 27 lit tests with EXPENSIVE_CHECKS enabled.
Reviewed By: rampitec, arsenm
Differential Revision: https://reviews.llvm.org/D136927
Only add the option processing and store the result. No attributes are
added to FIR yet.
This patch follows Clang in forwarding -fno-honor-infinities as
-menable-no-infs.
Reviewed By: kiranchandramohan awarzynski vzakhari
Differential Revision: https://reviews.llvm.org/D137072
