This patch teaches `isProjectedPermutation` and `inverseAndBroadcastProjectedPermutation`
utilities to deal with maps representing an explicit broadcast, e.g., (d0, d1) -> (d0, 0).
This extension is needed to enable vectorization of such explicit broadcast in Linalg.
Reviewed By: pifon2a, nicolasvasilache
Differential Revision: https://reviews.llvm.org/D111563
This patch upstream the cfg conversion pass. This pass
rewrite FIR loop-like operation to a CFG.
This patch is part of the upstreaming effort from fir-dev branch.
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
Reviewed By: schweitz
Differential Revision: https://reviews.llvm.org/D111095
If we have an instruction which produces poison only when flags are specified on the instruction, then we know that freezing the operands and dropping flags is equivalent to freezing the result. If we know those flags don't result in any undefined behavior being executed, then there's no point in preserving the flags as we gain no knowledge by having them.
This patch extends the existing propagation logic which sinks freeze to single potential non-poison operands to allow dropping of flags when we know the freeze is the sole use of the instruction with poison flags.
The main value is that we tend to sink freezes towards the phi in IV cycles where the incoming value to the phi is the freeze of an IV increment. This will in turn (in a future patch), let us fold the freeze through the phi into the loop preheader. Motivated by eliminating need for CanonicalizeFreezeInLoops for the clearly profitable cases from onephi.ll test case in the test directory.
Differential Revision: https://reviews.llvm.org/D111675
On AIX, the system assembler does not support the extended mnemonics
dcbtt and dcbtstt. This patch stops them from being emitted on
AIX and emits the base mnemonics instead, dcbt X, X, 16 and
dcbtstt X, X, 16 respectively.
Differential revision: https://reviews.llvm.org/D111258
f341161689 added a task dispatcher for async handlers, but didn't add a
TaskDispatcher::shutdown call to SelfExecutorProcessControl or SimpleRemoteEPC.
This patch adds the missing call, which ensures that we don't destroy the
dispatcher while tasks are still running.
This should fix the use-after-free crash seen in
https://lab.llvm.org/buildbot/#/builders/5/builds/13063
Simplify the test case to make it easier to look at. Change from auto-generated
checks to targeted manual checks to reduce sensitivity to register allocation
and scheduling changes.
Differential Revision: https://reviews.llvm.org/D111333
Average pool assumed the same input/output type. Result type for integers
is always an i32, should be updated appropriately.
Reviewed By: GMNGeoffrey
Differential Revision: https://reviews.llvm.org/D111590
Noticed in code review
4318028cd2 (commitcomment-57738034)
But the issue had already been fixed in
943b304848 due to a code checking tool
(PVS studio) identification, but that lacked test coverage.
Refactor this test a little bit too by using more CHECK-SAME to help the
checks fail sooner (rather than, if the addrx or sizes are wrong, having
that check bind to a much later output line - and then fail due to the
implicit-check-nots, which don't provide a lot of information about
where the intended check was likely to land) & more informatively.
Also modify the test to be more robust (current IR generation doesn't
include call sites for callees that are only declared but not defined -
so the test case couldn't be regenerated - add a function definition (&
optnone attribute) so it doesn't depend on call sites for
declared-but-not-defined functions)
This patch fixes another crash revealed by PR51614:
when *deciding* to vectorize with masked interleave groups, check if the access
is reverse (which is currently not supported).
Differential Revision: https://reviews.llvm.org/D108900
This adds strncat to llvm libc. In addition, an error was found with
strcat and that was fixed.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D111583
Adapt CodegenStartegy to used the vector transfer lowering patterns by default.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D111649
unlock corrupted backets by using s set by loop to nullptr.
Also StackDepot supports iterating without locking.
Reviewed By: dvyukov
Differential Revision: https://reviews.llvm.org/D111599
This patch fixes a couple of small oversights in the documentation for
the datalayout specification:
* The v and f specifications are subject to the same constraints on <size>
as i is.
* The p[n] specification didn't mark <idx> as optional, despite
being documented and parsed as such.
* Similarly, none of the alignment specifications require <pref>.
If another inlining session came after a ModuleInlinerWrapperPass, the
advisor alanysis would still be cached, but its Result would be cleared.
We need to clear both.
This addresses PR52118
Differential Revision: https://reviews.llvm.org/D111586
Clarify the message provided when the analyzer catches the use of memory
that is allocated with size zero.
Differential Revision: https://reviews.llvm.org/D111655
CFGBuilder::addStmt() implicitly passes AddStmtChoice::AlwaysAdd
to Visit() already, so this should have no behavior change.
Differential Revision: https://reviews.llvm.org/D111570
If I remember correctly this wasn't done previously because dim used to
be in the memref dialect.
Differential Revision: https://reviews.llvm.org/D111651
The new pass walks kernel's pointer arguments, then loads from them.
If a loaded value is a pointer and loaded pointer is unmodified in
the kernel before the load, then promote loaded pointer to global.
Then recursively continue.
Differential Revision: https://reviews.llvm.org/D111464
This may not be obvious, but Alive2 agrees:
https://alive2.llvm.org/ce/z/Ld9qNT
If the mul has "nsw", then -1 * INT_MIN is poison, so the
negate can also have "nsw" because 0 - INT_MIN is poison.
If the mul has "nuw", then that means the "OtherOp" can only
be 0 or 1 (anything else multiplied by 0xfff... would wrap).
So the replacement negate must be "nsw" because it is either
"0-0" or "0-1".
This is another regression noticed with a planned follow-up
to D111410.
As noted in https://github.com/halide/Halide/pull/6302,
we hilariously fail to match PAVG if we even as much
as look at it the wrong way.
In this particular case, the problem stems from the fact that
`PAVG` root (def) is a `trunc`, and leafs (uses) are `zext`'s,
and InstCombine really loves to get rid of both of these,
for example replace them with a bit mask. So we may not have
said `zext`.
Instead of checking for that + type match,
i think we should rely on the actual active type,
as per the knownbits.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D111571
As reported in https://bugs.llvm.org/show_bug.cgi?id=48145, name resolution for omp critical construct was failing. This patch adds functionality to help that name resolution as well as implementation to catch name mismatches.
The following semantic restrictions are therefore handled here:
- If a name is specified on a critical directive, the same name must also be specified on the end critical directive
- If no name appears on the critical directive, no name can appear on the end critical directive
- If a name appears on either the start critical directive or the end critical directive
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D110502
This patch continues unblocking optimizations that are blocked by pseudo probe instrumentation.
Not exactly like DbgIntrinsics, PseudoProbe intrinsic has other attributes (such as mayread, maywrite, mayhaveSideEffect) that can block optimizations. The issues fixed are:
- Flipped default param of getFirstNonPHIOrDbg API to skip pseudo probes
- Unblocked CSE by avoiding pseudo probe from clobbering memory SSA
- Unblocked induction variable simpliciation
- Allow empty loop deletion by treating probe intrinsic isDroppable
- Some refactoring.
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D110847
To get proper wrap-around behavior for the various kind parameter
values of the optional COUNT= and COUNT_MAX= dummy arguments to
the intrinsic subroutine SYSTEM_CLOCK, add an extra argument to
the APIs for lowering to pass the integer kind of the actual argument.
Avoid confusion by requiring that both actual arguments have the same
kind when both are present. The results of the runtime functions
remain std::int64_t and lowering should still convert them before
storing to the actual argument variables.
Rework the implementation a bit to accomodate the dynamic
specification of the kind parameter, and to clean up some coding
issues with preprocessing and templates.
Use the kind of the COUNT=/COUNT_MAX= actual arguments to determine
the clock's resolution, where possible, in conformance with other
Fortran implementations.
Differential Revision: https://reviews.llvm.org/D111281
Vitaly Buka