I'm not sure if there's a way to make this a bit more general - the
property that matters is that there's /some/ itanium ABI target the
tests can use to compile - not link or run in this case. But this seems
sufficient for the llgdb tests, so it should be sufficient (though
perhaps not necessary) for this roundtrip test.
Compilation is all that's needed here - linking helped avoid certain
false positives in llvm-dwarfdump --verify related to overlapping
functions when those functions were actually in distinct sections.
That's since been fixed, so we can adjust the test to only compile, and
not link.
The new name better suits the type.
This patch also changes the signature of the run method (it now returns a
WrapperFunctionResult), and adds runWithSPSRet methods that deserialize the
function result using SPS.
Together these chages bring this type into close alignment with its ORC runtime
counterpart.
There's another test that opens an hard-coded port to talk to debugserver
(TestPlatformSDK.py). Make sure this port and the one in that other
test are different to avoid that potential conflict.
Previously we were relying on the dynamic loader to take care of this
but it simple and correct for us to do it here instead.
Now we initialize bss segments as part of `__wasm_init_memory` at the
same time we initialize passive segments.
In addition we extent the us of `__wasm_init_memory` outside of shared
memory situations. Specifically it is now used to initialize bss
segments when the memory is imported.
Differential Revision: https://reviews.llvm.org/D112667
Functions in different sections (common in object files - inline
functions, -ffunction-sections, etc) can't overlap, so factor in the
section when diagnosing overlapping address ranges.
This removes a major false-positive when running llvm-dwarfdump on
unlinked code.
WrapperFunctionCall represents a call to a wrapper function as a pair of a
target function (as an ExecutorAddr), and an argument buffer range (as an
ExecutorAddrRange). WrapperFunctionCall instances can be serialized via
SPS to send to remote machines (only the argument buffer address range is
copied, not any buffer content).
This utility will simplify the implementation of JITLinkMemoryManager
allocation actions in the ORC runtime.
We weren't setting the listener back to the unhijacked one in this
case, so that a continue after the stop fails. It thinks the process
is still running. Also add tests for this behavior.
Differential Revision: https://reviews.llvm.org/D112747
This has no uses and the ValueObjectDynamicValue already tracks
its ownership through the parent it is passed when made. I can't
find any vestiges of the use of this API, maybe it was from some
earlier design?
Resetting the backing ivar was the only job the destructor did, so I
set that to default as well.
Differential Revision: https://reviews.llvm.org/D112677
While "null()" is accepted as a data statement constant when it
corresponds to a pointer object, "null(mold=p)" and "null(p)"
are not allowed. The current error messages simply complain
that null is not an array. This patch adds a context-sensitive
message to the effect that a data statement constant followed
by non-empty parentheses must be an array or structure constructor.
(Note that f18 can't simply special-case the name "null" when parsing
data statement constants, since programs are free to repurpose that
name as an array or derived type.)
Differential Revision: https://reviews.llvm.org/D112740
If the VL argument for a mask instruction comes from a vsetvli with
an SEW!=8, we will insert an extra vsetvli for the mask instruction
even if the SEW/LMUL ratio is the same. This requires at least one
instruction before the mask instruction that needs the SEW of the
explicit vsetvli. Otherwise, we'll just rewrite the explicit vsetvli.
We don't have much publicly documented about how build bots are maintained, and who has what responsibility with regards to the bots themselves and failures they report.
This change adds some basic framing on the topic. It is intended to simplify document long standing practice, and to avoid introducing any new policy elements. I do want to propose a couple of changes, but those will happen with their own review and broader discussion.
Differential Revision: https://reviews.llvm.org/D112755
A recent change caused some variable-length sequential formatted
output statements with record positioning at the end of a FORMAT
(e.g., FORMAT('hi',10X) to append blanks at the end of the completed
record when emitting it.
Differential Revision: https://reviews.llvm.org/D112742
Add i32x4.relaxed_trunc_f32x4_s, i32x4.relaxed_trunc_f32x4_u,
i32x4.relaxed_trunc_f64x2_s_zero, i32x4.relaxed_trunc_f64x2_u_zero.
These are only exposed as builtins, and require user opt-in.
Differential Revision: https://reviews.llvm.org/D112186
After recent changes to the Docker image, all hell broke loose and the
CI started failing. This patch marks a few tests as unsupported until
we can figure out what the issues are and fix them.
In the future, it would be ideal if the nodes could pick up the Dockerfile
present in the revision being tested, which would allow us to test changes
to the Dockerfile in the CI, like we do for all other code changes.
Differential Revision: https://reviews.llvm.org/D112737
This is relanding commit da1d1a0869 .
This patch additionally addresses failures found in buildbots & post review comments.
ARM EHABI[1] specifies the __cxa_end_cleanup to be called after cleanup.
It will call the UnwindResume.
__cxa_begin_cleanup will be called from libcxxabi while __cxa_end_cleanup is never called.
This will trigger a termination when a foreign exception is processed while UnwindResume is called
because the global state will be wrong due to the missing __cxa_end_cleanup call.
Additional test here: D109856
[1] https://github.com/ARM-software/abi-aa/blob/main/ehabi32/ehabi32.rst#941compiler-helper-functions
Reviewed By: logan
Differential Revision: https://reviews.llvm.org/D111703
`is_error_condition_enum_v` and `is_error_code_enum_v` are currently of
type `size_t`, but the standard mandates they are of type `bool`.
This is an ABI break technically since the size of these variable
templates has changed. Document it as such in the release notes.
Fixes https://bugs.llvm.org/show_bug.cgi?id=50755
Reviewed By: ldionne, Quuxplusone, #libc, var-const
Differential Revision: https://reviews.llvm.org/D112553
Fixes non-determinisctic order of XOR instructions created after
5a7a458306. The order of call argument evaluation is not
defined, so create one Value before the call.
timm isn't the common case, and TImmLeafs should make it clear what
they are. We're adding a plain ImmLeaf for 0_65535, so rename
i64_imm0_65535 to timm64_0_65535, and imm32_0_7 to timm32_0_7.
`__stack_chk_guard` is a global variable that has no uses before the LLVM code generation phase (how it is defined is platform-dependent). LTO needs to preserve this symbol for that reason. Currently, legacy LTO API preserves it by hardcoding the logic in Internalizer, but this symbol is not preserved by regular LTO API in thinlink phase. This patch marks `__stack_chk_guard` used during IR symbol table writing since this is how builtin functions are preserved by thinlink by using `RuntimeLibcalls.def`.
Reviewed By: MaskRay, tejohnson
Differential Revision: https://reviews.llvm.org/D112595
Move the section collecting `AlwaysPreserved` up before any
`maybeInternalize` is called. Otherwise, functions in `AlwaysPreserved` (in this case, `__stack_chk_fail`)
are not preserved.
Reviewed By: MaskRay, tejohnson
Differential Revision: https://reviews.llvm.org/D112684
In function convertInstTo3Addr, after converting a two address instruction into
three address instruction, only the last new instruction is inserted into
DistanceMap. This is wrong, DistanceMap should track all instructions from the
beginning of current MBB to the working instruction. When a two address
instruction is converted to three address instruction, multiple instructions may
be generated (usually an extra COPY is generated), all of them should be
inserted into DistanceMap.
Similarly when unfolding memory operand in function tryInstructionTransform
DistanceMap is not maintained correctly.
Differential Revision: https://reviews.llvm.org/D111857
Add deduction guides to `valarray` and `scoped_allocator_adaptor`. This largely
finishes implementation of the paper:
* deduction guides for other classes mentioned in the paper were
implemented previously (see the list below);
* deduction guides for several classes contained in the proposal
(`reference_wrapper`, `lock_guard`, `scoped_lock`, `unique_lock`,
`shared_lock`) were removed by [LWG2981](https://wg21.link/LWG2981).
Also add deduction guides to the synopsis for the few classes (e.g. `pair`)
where they were missing.
The only part of the paper that isn't fully implemented after this patch is
making sure certain deduction guides don't participate in overload resolution
when given incorrect template parameters.
List of significant commits implementing the other parts of P0433 (omitting some
minor fixes):
* [pair](af65856eec)
* [basic_string](6d9f750dec)
* [array](0ca8c0895c)
* [deque](dbb6f8a817)
* [forward_list](e076700b77)
* [list](4a227e582b)
* [vector](df8f754792)
* [queue/stack/priority_queue](5b8b8b5dce)
* [basic_regex](edd5e29cfe)
* [optional](f35b4bc395)
* [map/multimap](edfe8525de)
* [set/multiset](e20865c387)
* [unordered_set/unordered_multiset](296a80102a)
* [unordered_map/unordered_multimap](dfcd4384cb)
* [function](e1eabcdfad)
* [tuple](1308011e1b)
* [shared_ptr/weak_ptr](83564056d4)
Additional notes:
* It was revision 2 of the paper that was voted into the Standard.
P0433R3 is a separate paper that is not part of the Standard.
* The paper also mandates removing several `make_*_searcher` functions
(e.g. `make_boyer_moore_searcher`) which are currently not implemented
(except in `experimental/`).
* The `__cpp_lib_deduction_guides` feature test macro from the paper was
accidentally omitted from the Standard.
Differential Revision: https://reviews.llvm.org/D112510
This extends `optimizeCompareInstr` to re-use previous comparison
results if the previous comparison was with an immediate that was 1
bigger or smaller. Example:
CMP x, 13
...
CMP x, 12 ; can be removed if we change the SETg
SETg ... ; x > 12 changed to `SETge` (x >= 13) removing CMP
Motivation: This often happens because SelectionDAG canonicalization
tends to add/subtract 1 often when optimizing for fallthrough blocks.
Example for `x > C` the fallthrough optimization switches true/false
blocks with `!(x > C)` --> `x <= C` and canonicalization turns this into
`x < C + 1`.
Differential Revision: https://reviews.llvm.org/D110867
`X86InstrInfo::optimizeCompareInstr` would only optimize a `SUB`
followed by a `CMP` in `isRedundantFlagInstr`. This extends the code to
also look for other combinations like `CMP`+`CMP`, `TEST`+`TEST`, `SUB
x,0`+`TEST`.
- Change `isRedundantFlagInstr` to run `analyzeCompareInstr` on the
candidate instruction and compare the results. This normalizes things
and gives consistent results for various comparisons (`CMP x, y`,
`SUB x, y`) and immediate cases (`TEST x, x`, `SUB x, 0`,
`CMP x, 0`...).
- Turn `isRedundantFlagInstr` into a member function so it can call
`analyzeCompare`. - We now also check `isRedundantFlagInstr` even if
`IsCmpZero` is true, since we now have cases like `TEST`+`TEST`.
Differential Revision: https://reviews.llvm.org/D110865
Refactor ConnectionFileDescriptor to improve code reuse for different
types of sockets. Unify method naming.
While at it, remove some (now-)dead code from Socket.
Differential Revision: https://reviews.llvm.org/D112495
This patch updates recipe creation to ensure all
VPWidenIntOrFpInductionRecipes are in the header block. At the moment,
new induction recipes can be created in different blocks when trying to
optimize casts and induction variables.
Having all induction recipes in the header makes it easier to
analyze/transform them in VPlan.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D111300
Alex Lorenz