There is one result per lookup symbol, so we have to advance the result iterator no matter whether it's NULL or not.
MissingSymbols variable is unused.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D91707
Distinguish objects by target properties address size, endian and machine architecture. So far we only
support x86-64 (ELFCLASS64, ELFDATA2LSB, EM_X86_64).
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D90860
Currently in some places we use signed type to represent size of an access and put explicit casts from unsigned to signed.
For example: int64_t EarlierSize = int64_t(Loc.Size.getValue());
Even though it doesn't loos bits (immidiatly) it may overflow and we end up with negative size. Potentially that cause later code to work incorrectly. A simple expample is a check that size is not negative.
I think it would be safer and clearer if we use unsigned type for the size and handle it appropriately.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D92648
Add a pass option to control the number of nested parallel loops produced by
the parallelization passes. This is useful to build end-to-end passes targeting
systems that don't need multiple parallel dimensions (e.g., CPUs typically need
only one).
Reviewed By: wsmoses, chelini
Differential Revision: https://reviews.llvm.org/D92765
The existing implementation of the affine parallelization silently copies over
the lower and upper bound maps from affine.for to affine.parallel. However, the
semantics of these maps differ between these two ops: in affine.for, a max(min)
of results is taken for the lower(upper) bound; in affine.parallel, multiple
induction variables can be defined an each result corresponds to one induction
variable. Thus the existing implementation could generate invalid IR or IR that
passes the verifier but has different semantics than the original code. Fix the
parallelization utility to emit dedicated min/max operations before the
affine.parallel in such cases. Disallow parallelization if min/max would have
been in an operation without the AffineScope trait, e.g., in another loop,
since the result of these operations is not considered a valid affine dimension
identifier and may not be properly handled by the affine analyses.
Reviewed By: wsmoses
Differential Revision: https://reviews.llvm.org/D92763
This patch adds tests that showcase a behavior that is currently buggy.
Fix in a follow-up patch.
Differential Revision: https://reviews.llvm.org/D91269
This change exposed a pre-existing issue with deserialization cycles
caused by a combination of attributes and template instantiations
violating the deserialization ordering restrictions; see PR48434 for
details.
A previous commit attempted to work around PR48434, but appears to have
only been a partial fix, and fixing this properly seems non-trivial.
Backing out for now to unblock things.
This reverts commit 98f76adf4e and
commit a64c26a47a.
Don't produce or expect any output from the infinite looping test -
doing so is a recipe for racey flakyness without a longer timeout to
ensure the output is received first, even though that doesn't seem
integral/important to the test. Instead have a plain, no output infinite
loop and check that that is caught and handled.
If for some reason the output is valuable for test coverage - the
timeout should be increased from 1 second to give the process time to
output the text, flush, and for that text to be received and buffered
before the test is timed out.
The NPM runs loop passes on loops in forward program order instead of
the legacy PM's reverse program order, causing some debug output to be
in a different order.
Don't know why under Sanitizer build(asan/msan/ubsan), the `std::unordered_map<string, ...>`'s output order is reversed, make the regression test failed.
This change creates a workaround by using sorted container to make the output deterministic.
Reviewed By: hoy, wenlei
Differential Revision: https://reviews.llvm.org/D92816
Instruction darn was introduced in ISA 3.0. It means 'Deliver A Random
Number'. The immediate number L means:
- L=0, the number is 32-bit (higher 32-bits are all-zero)
- L=1, the number is 'conditioned' (processed by hardware to reduce bias)
- L=2, the number is not conditioned, directly from noise source
GCC implements them in three separate intrinsics: __builtin_darn,
__builtin_darn_32 and __builtin_darn_raw. This patch implements the
same intrinsics. And this change also addresses Bugzilla PR39800.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D92465
The -enable-new-pm=1 translation caused loop-vectorize to run on all
functions, then instcombine, rather than all passes on one function then
the next. This caused the output of -debug-only and -print-after to be
interleaved in an unexpected way.
The NPM processes loops in forward program order, whereas the legacy PM
processes them in reverse program order. No reason to test both PMs
here, so just stick to the NPM.
getFieldOffset(layoutStartOffset) is expected to point to the first trivial
field or the one which follows non-trivial. So it must be byte aligned already.
However this is not obvious without assumptions about callers.
This patch will avoid the need in such assumptions.
Depends on D92727.
Differential Revision: https://reviews.llvm.org/D92728
The -loop-flatten legacy pass preserves loop analyses. The legacy PM
will check all passes that preserve loop analyses that they preserve
LCSSA. This implicitly involves running -loop-simplify. The test
shouldn't depend on verify flags being set in order to run
-loop-simplify, so explicitly add it. The new PM ends up not running it
otherwise.
Summary: The imm operands of some instructions are not defined accurately in td.
This is a small patch to correct these definitions.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D91603
`selectCompareBranch` was hard to understand.
Also, it was being needlessly pessimistic with the `ProduceNonFlagSettingCondBr`
case. It assumed that everything in `selectCompareBranch` would emit a TB(N)Z
or C(B)NZ. That's not true; the G_FCMP + G_BRCOND case would never emit those
instructions, and the G_ICMP + G_BRCOND case was capable of emitting an integer
compare + Bcc.
- Refactor `selectCompareBranch` into separate functions based off of what is
feeding the G_BRCOND's condition.
- Move G_BRCOND selection code from `select` to `selectCompareBranch`.
- Remove duplicated constraint code from the code originally in `select`;
`emitTestBit` already handles that, so no need to constrain twice.
- Factor out the G_FCMP + G_BRCOND case into `selectCompareBranchFedByFCmp`.
- Split the G_ICMP + G_BRCOND case into an optimization function,
`tryOptCompareBranchFedByICmp` and a general selection function,
`selectCompareBranchFedByICmp`.
- Reduce the number of things passed to `tryOptAndIntoCompareBranch`.
- Improve documentation.
- Give some variables more descriptive names.
Other than improving the code generation for functions with
speculative_load_hardening by getting the logic correct, this is NFC.
Differential Revision: https://reviews.llvm.org/D92582
I'm not sure if it would be legal by the IR reference to introduce
an addrspacecast here, since the IR reference is a bit vague on
the exact semantics, but at least for our usage of it (and I
suspect for many other's usage) it is not. For us, addrspacecasts
between non-integral address spaces carry frontend information that the
optimizer cannot deduce afterwards in a generic way (though we
have frontend specific passes in our pipline that do propagate
these). In any case, I'm sure nobody is using it this way at
the moment, since it would have introduced inttoptrs, which
are definitely illegal.
Fixes PR38375
Co-authored-by: Keno Fischer <keno@alumni.harvard.edu>
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D50010
These lit tests now requires amdgpu-registered-target since they
use clang driver and clang driver passes an LLVM option which
is available only if amdgpu target is registered.
Change-Id: I2df31967409f1627fc6d342d1ab5cc8aa17c9c0c
Our internal build bot hit a failure in llvm/test/tools/llvm-symbolizer/pdb/missing_pdb.test
because the test was checking for an error message that is emitted when a pdb file is
missing. But when the drive is mapped to a removalable drive (such as a DVD drive) in
Windows, you get a different error message which causes the test to fail.
This fixes the test by changing the drive the missing pdb is expected to be on to C:\
instead of D:\ as that is the drive historically used to install Windows and thus
if present should be a hard drive.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D92787
This is really just a workaround for a more fundamental issue in the way
we deserialize attributes. See PR48434 for details.
Also fix tablegen code generator to produce more correct indentation to
resolve buildbot issues with -Werror=misleading-indentation firing
inside the generated code.
The default exception handling isn't very user friendly and does not
point accurately to the issue. Instead we can indicate which of the
operands isn't valid and provide contextual information in the error
message.
Differential Revision: https://reviews.llvm.org/D92710
When we have a 128-bit register, emitTestBit would incorrectly narrow to 32
bits always. If the bit number was > 32, then we would need a TB(N)ZX. This
would cause a crash, as we'd have the wrong register class. (PR48379)
This generalizes `narrowExtReg` into `moveScalarRegClass`.
This also allows us to remove `widenGPRBankRegIfNeeded` entirely, since
`selectCopy` correctly handles SUBREG_TO_REG etc.
This does create some codegen changes (since `selectCopy` uses the `all`
regclass variants). However, I think that these will likely be optimized away,
and we can always improve the `selectCopy` code. It looks like we should
revisit `selectCopy` at this point, and possibly refactor it into at least one
`emit` function.
Differential Revision: https://reviews.llvm.org/D92707
Build on the work started in 8f07629, and add the multiply case. In the process, more clearly describe the requirement for the operation we're looking through.
Differential Revision: https://reviews.llvm.org/D92726
Committing on behalf of thejh (Jann Horn).
As part of this change, one existing test case has to be adjusted
because it accidentally stripped the NoDeref attribute without
getting caught.
Depends on D92140
Differential Review: https://reviews.llvm.org/D92141
This patch plugs many holes in static initializer semantics, improves error
messages for default initial values and other component properties in
parameterized derived type instantiations, and cleans up several small
issues noticed during development. We now do proper scalar expansion,
folding, and type, rank, and shape conformance checking for component
default initializers in derived types and PDT instantiations.
The initial values of named constants are now guaranteed to have been folded
when installed in the symbol table, and are no longer folded or
scalar-expanded at each use in expression folding. Semantics documentation
was extended with information about the various kinds of initializations
in Fortran and when each of them are processed in the compiler.
Some necessary concomitant changes have bulked this patch out a bit:
* contextual messages attachments, which are now produced for parameterized
derived type instantiations so that the user can figure out which
instance caused a problem with a component, have been added as part
of ContextualMessages, and their implementation was debugged
* several APIs in evaluate::characteristics was changed so that a FoldingContext
is passed as an argument rather than just its intrinsic procedure table;
this affected client call sites in many files
* new tools in Evaluate/check-expression.cpp to determine when an Expr
actually is a single constant value and to validate a non-pointer
variable initializer or object component default value
* shape conformance checking has additional arguments that control
whether scalar expansion is allowed
* several now-unused functions and data members noticed and removed
* several crashes and bogus errors exposed by testing this new code
were fixed
* a -fdebug-stack-trace option to enable LLVM's stack tracing on
a crash, which might be useful in the future
TL;DR: Initialization processing does more and takes place at the right
times for all of the various kinds of things that can be initialized.
Differential Review: https://reviews.llvm.org/D92783
Committing on behalf of thejh (Jann Horn).
Given an attribute((noderef)) pointer "p" to the struct
struct s { int a[2]; };
ensure that the following expressions are treated the same way by the
noderef logic:
p->a
(*p).a
Until now, the first expression would be treated correctly (nothing is
added to PossibleDerefs because CheckMemberAccessOfNoDeref() bails out
on array members), but the second expression would incorrectly warn
because "*p" creates a PossibleDerefs entry.
Handle this case the same way as for the AddrOf operator.
Differential Revision: https://reviews.llvm.org/D92140
This attributes specifies how (or if) a given function or method will be
imported into a swift async method. rdar://70111252
Differential revision: https://reviews.llvm.org/D92742
_Nullable_result generally like _Nullable, except when being imported into a
swift async method. rdar://70106409
Differential revision: https://reviews.llvm.org/D92495
This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC.
This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path.
we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding.
Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context.
With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO.
A breakdown of noteworthy changes:
- Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack
* Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted
* Speed up by aggregating `HybridSample` into `AggregatedSamples`
* Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context. Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`.
* Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
* Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop
- `getCanonicalFnName` for callee name and name from ELF section
- Added regression test for both unwinding and profile generation
Test Plan:
ninja & ninja check-llvm
Reviewed By: hoy, wenlei, wmi
Differential Revision: https://reviews.llvm.org/D89723
Such fields will likely have offset zero making
__sanitizer_dtor_callback poisoning wrong regions.
E.g. it can poison base class member from derived class constructor.
Differential Revision: https://reviews.llvm.org/D92727
Stefan Gränitz