This patch fixes the scheduler's clear() function to also clear the
ReadyList. Not doing so is a bug and results in crashes when the
ReadyList contains stale instructions, because it was never clered.
MaterializationUnits may contain arbitrary resources that need cleanup. We want
to do this outside the JIT's session lock.
This should fix a lock-order-inversion warning in clang-repl (for details see
https://github.com/llvm/llvm-project/issues/124215).
- Widen v2i8, v2i16 and v2i32 vectors so they don't cast back and forth,
and make sure that instructions with correct data unit is being used.
- Handle undef indices for VSHF when lowering VECTOR_SHUFFLE (it crashes
if such index is present).
Close https://github.com/llvm/llvm-project/issues/123815
See the comments for details. We can't get primary context arbitrarily
since the redecl may have different context and information.
There is a TODO for modules specific case, I'd like to make it after
this PR.
The current MIR cycle sinking capabilities are rather limited. It only
support sinking copies into a single successor block while obeying
limits.
This opt-in feature adds a more aggressive option, that is not limited
to the above concerns. The feature will try to "sink" by duplicating any
top-level preheader instruction (that we are sure is safe to sink) into
any user block, then does some dead code cleanup. In particular, this is
useful for high RP situations when loop bodies have control flow.
The stated return type was incorrect; this patch corrects it. More generally, it explains how the Offset and its components fits into the overall shadow mapping calculation.
The symbol string table does not have deduplication.
Here we add code to deduplicate the symbol string table.
This has a rather large size impact (20-30%) on unstripped binaries
(typically debug binaries) but no size impact on stripped
binaries(typically release binaries).
We enable deduplication by default and add a flag to disable it
(`-no-deduplicate-symbol-strings`).
This patch relands the following PRs:
* #111711
* #107350
* #111457
All of these patches were reverted due to an issue reported in
https://github.com/llvm/llvm-project/pull/111711#issuecomment-2406491485,
due to interdependencies.
---
[clang] Finish implementation of P0522
This finishes the clang implementation of P0522, getting rid
of the fallback to the old, pre-P0522 rules.
Before this patch, when partial ordering template template parameters,
we would perform, in order:
* If the old rules would match, we would accept it. Otherwise, don't
generate diagnostics yet.
* If the new rules would match, just accept it. Otherwise, don't
generate any diagnostics yet again.
* Apply the old rules again, this time with diagnostics.
This situation was far from ideal, as we would sometimes:
* Accept some things we shouldn't.
* Reject some things we shouldn't.
* Only diagnose rejection in terms of the old rules.
With this patch, we apply the P0522 rules throughout.
This needed to extend template argument deduction in order
to accept the historial rule for TTP matching pack parameter to non-pack
arguments.
This change also makes us accept some combinations of historical and P0522
allowances we wouldn't before.
It also fixes a bunch of bugs that were documented in the test suite,
which I am not sure there are issues already created for them.
This causes a lot of changes to the way these failures are diagnosed,
with related test suite churn.
The problem here is that the old rules were very simple and
non-recursive, making it easy to provide customized diagnostics,
and to keep them consistent with each other.
The new rules are a lot more complex and rely on template argument
deduction, substitutions, and they are recursive.
The approach taken here is to mostly rely on existing diagnostics,
and create a new instantiation context that keeps track of this context.
So for example when a substitution failure occurs, we use the error
produced there unmodified, and just attach notes to it explaining
that it occurred in the context of partial ordering this template
argument against that template parameter.
This diverges from the old diagnostics, which would lead with an
error pointing to the template argument, explain the problem
in subsequent notes, and produce a final note pointing to the parameter.
---
[clang] CWG2398: improve overload resolution backwards compat
With this change, we discriminate if the primary template and which partial
specializations would have participated in overload resolution prior to
P0522 changes.
We collect those in an initial set. If this set is not empty, or the
primary template would have matched, we proceed with this set as the
candidates for overload resolution.
Otherwise, we build a new overload set with everything else, and proceed
as usual.
---
[clang] Implement TTP 'reversed' pack matching for deduced function template calls.
Clang previously missed implementing P0522 pack matching
for deduced function template calls.
Previously, AArch64 used pattern matching to support
llvm.vector.(de)interleave of 2 and 4; RISC-V only supported
(de)interleave of 2.
This patch consolidates the logics in these two targets by factoring out
the common factor calculations into the InterleaveAccess Pass.
This fixes "unused-local-typedef" warnings in 9324e6a7a5.
This adds an option `--remove-note=[name/]type` to selectively delete
notes in ELF files, where `type` is the numeric value of the note type
and `name` is the name of the originator. The name can be omitted, in
which case all notes of the specified type will be removed. For now,
only `SHT_NOTE` sections that are not associated with segments are
handled. The implementation can be extended later as needed.
RFC: https://discourse.llvm.org/t/rfc-llvm-objcopy-feature-for-editing-notes/83491
Crossing BBs is not currently supported by the structures of the
vectorizer. This patch fixes instances where this was happening,
including:
- a walk of use-def operands that updates the UnscheduledSuccs counter,
- the dead instruction removal is now done per BB,
- the scheduler, which will reject bundles that cross BBs.
For Darwin/arm64 (including Apple Silicon Macs) this will enable exception
handling and stack unwinding in JIT'd code.
Darwin supports two unwind-info formats: DWARF eh-frames and compact-unwind. On
Darwin/x86-64 compilers usually produce both by default, and ORC supported
exceptions and unwinding via eh-frames (same as on Linux), discarding the
redundant compact-unwind info. On Darwin/arm64 compilers typically default to
producing compact-unwind only, with DWARF eh-frames as a fallback for functions
that can't be described in compact-unwind. Since ORC did not previously support
the compact-unwind format and eh-frames were not present ORC was unable to
handle exceptions or unwinding by default in Darwin/arm64 JIT'd code.
This patch enables support for the compact-unwind-info format, and contains
three major moving parts:
(1) The JITLink CompactUnwindManager class is responsible for transforming the
__compact_unwind records produced by the linker into the __unwind_info
tables that libunwind parses during unwinding. To enable this the
CompactUnwindManager class provides three JITLink passes: The
prepareForPrune pass that splits the __compact_unwind section into
single-record blocks, allowing unused records to be dead-stripped; the
processAndReserveUnwindInfo pass that reserves space for the final
__unwind_info section, and the writeUnwindInfo pass that writes the
__unwind_info section.
(2) The OrcTargetProcess UnwindInfoManager class maintains a table of
registered JIT'd __unwind_info and __eh_frame sections, and handles
requests from libunwind for unwind info sections (by registering a callback
with libunwind's __unw_add_find_dynamic_unwind_sections function).
(3) The Orc UnwindInfoRegistrationPlugin, which scans LinkGraphs for
__unwind_info and __eh_frame sections to register with the
UnwindInfoManager.
This commit adds the CompactUnwindManager passes to the default JITLink
pipelines for Darwin/arm64 and Darwin/x86-64, and UnwindInfoManager intances to
the SelfExecutorProcessControl class (when built for apple platforms) and the
llvm-jitlink-executor tool.
The LLJIT class will now create an UnwindInfoRegistrationPlugin when targeting
a process running on Darwin if it detects that an UnwindInfoManager is
available to handle the registrations.
The ORC runtime macho_platform class already supported libunwind callbacks, so
out-of-process execution and unwinding support will work when loading the ORC
runtime.
The llvm-jitlink tool will only support compact-unwind when the orc-runtime is
loaded, as the UnwindInfoRegistrationPlugin requires access to an IR compiler
to load a helper module and llvm-jitlink does not provide an IR compiler.
#122466 had an unexpected side effect that,
`EnableFeaturesForWasmEHSjLj` and `BanIncompatibleOptionsForWasmEHSjLj`
can be called multiple times now, every time a Wasm EH flag
(`-fwasm-exceptions`, `-wasm-enable-eh`, `-wasm-enable-sjlj`, ..) was
checked and handled. This resulted in unnecessarily adding the same
feature-enabling arguments multiple times to the command line, for
example, `-target-feature +exception-handling` could be added as many as
three times, which didn't cause any errors but unnecessary. Also we ran
`BanIncompatibleOptionsForWasmEHSjLj` more than once, which was harmless
but unnecessary.
This guards these functions with a static variable so that we only run
them once.
This remove some erroneous debug info from tests that should address the
test failures that showed up when the this was previously committed.
This reverts commit 6716ce8b641f0e42e2343e1694ee578b027be0c4.
This adds an option `--remove-note=[name/]type` to selectively delete
notes in ELF files, where `type` is the numeric value of the note type
and `name` is the name of the originator. The name can be omitted, in
which case all notes of the specified type will be removed. For now,
only `SHT_NOTE` sections that are not associated with segments are
handled. The implementation can be extended later as needed.
RFC: https://discourse.llvm.org/t/rfc-llvm-objcopy-feature-for-editing-notes/83491
Introduced stack buffer overflow, see #120272.
`getScaledReduction` can return empty vector, and there is not check for
that.
This reverts commit c9b7303b9b18129c4ee6b56aaa2a0a9f59be2d09.
This reverts commit caf0540b91b0fee31353dc7049ae836e0f814cff.
Update string_utils' string_length to work with char* or wchar_t*, so that it
may be reusable when implementing wmemchr, wcspbrk, wcsrchr, wcsstr.
Link: #121183
Link: #124027
Co-authored-by: Nick Desaulniers <ndesaulniers@google.com>
---------
Co-authored-by: Tristan Ross <tristan.ross@midstall.com>
Provide isDead interface for access to ad-hoc isDead queries.
LivePhysRegs is optional: if not provided, pessimistically check
deadness of a single MI without doing the LivePhysReg walk; if provided
it is assumed to be at the position of MI.
Remove duplicate print of TheBitsInitPool.
I don't know who or what uses this information. I happened to notice
TheBitsInitPool was printed twice which lead to auditing the whole list.
We have two forceExpandWideMUL functions. One takes the low and high
half of 2 inputs and calculates the low and high half of their product.
This does not calculate the full 2x width product.
The other signature takes 2 inputs and calculates the low and high half
of their full 2x width product. Previously it did this by sign/zero
extending the inputs to create the high bits and then calling the other
function.
We can instead copy the algorithm from the other function and use the
Signed flag to determine whether we should do SRA or SRL. This avoids
the need to multiply the high part of the inputs and add them to the
high half of the result. This improves the generated code for signed
multiplication.
This should improve the performance of #123262. I don't know yet how
close we will get to gcc.
