lldb-vsdode was communicating the list of modules to the IDE with events, which in practice ended up having some drawbacks
- when debugging large targets, the number of these events were easily 10k, which polluted the messages being transmitted, which caused the following: a harder time debugging the messages, a lag after terminated the process because of these messages being processes (this could easily take several seconds). The latter was specially bad, as users were complaining about it even when they didn't check the modules view.
- these events were rarely used, as users only check the modules view when something is wrong and they try to debug things.
After getting some feedback from users, we realized that it's better to not used events but make this simply a request and is triggered by users whenever they needed.
This diff achieves that and does some small clean up in the existing code.
Differential Revision: https://reviews.llvm.org/D94033
This adds cost modelling for the inloop vectorization added in
745bf6cf44. Up until now they have been modelled as the original
underlying instruction, usually an add. This happens to works OK for MVE
with instructions that are reducing into the same type as they are
working on. But MVE's instructions can perform the equivalent of an
extended MLA as a single instruction:
%sa = sext <16 x i8> A to <16 x i32>
%sb = sext <16 x i8> B to <16 x i32>
%m = mul <16 x i32> %sa, %sb
%r = vecreduce.add(%m)
->
R = VMLADAV A, B
There are other instructions for performing add reductions of
v4i32/v8i16/v16i8 into i32 (VADDV), for doing the same with v4i32->i64
(VADDLV) and for performing a v4i32/v8i16 MLA into an i64 (VMLALDAV).
The i64 are particularly interesting as there are no native i64 add/mul
instructions, leading to the i64 add and mul naturally getting very
high costs.
Also worth mentioning, under NEON there is the concept of a sdot/udot
instruction which performs a partial reduction from a v16i8 to a v4i32.
They extend and mul/sum the first four elements from the inputs into the
first element of the output, repeating for each of the four output
lanes. They could possibly be represented in the same way as above in
llvm, so long as a vecreduce.add could perform a partial reduction. The
vectorizer would then produce a combination of in and outer loop
reductions to efficiently use the sdot and udot instructions. Although
this patch does not do that yet, it does suggest that separating the
input reduction type from the produced result type is a useful concept
to model. It also shows that a MLA reduction as a single instruction is
fairly common.
This patch attempt to improve the costmodelling of in-loop reductions
by:
- Adding some pattern matching in the loop vectorizer cost model to
match extended reduction patterns that are optionally extended and/or
MLA patterns. This marks the cost of the reduction instruction correctly
and the sext/zext/mul leading up to it as free, which is otherwise
difficult to tell and may get a very high cost. (In the long run this
can hopefully be replaced by vplan producing a single node and costing
it correctly, but that is not yet something that vplan can do).
- getExtendedAddReductionCost is added to query the cost of these
extended reduction patterns.
- Expanded the ARM costs to account for these expanded sizes, which is a
fairly simple change in itself.
- Some minor alterations to allow inloop reduction larger than the highest
vector width and i64 MVE reductions.
- An extra InLoopReductionImmediateChains map was added to the vectorizer
for it to efficiently detect which instructions are reductions in the
cost model.
- The tests have some updates to show what I believe is optimal
vectorization and where we are now.
Put together this can greatly improve performance for reduction loop
under MVE.
Differential Revision: https://reviews.llvm.org/D93476
In LoopInterchange, `findInnerReductionPhi()` looks for reduction
variables, which cannot be constants. Update it to return early in that
case.
This also addresses a blocker for removing use-lists from ConstantData,
whose users could be spread across arbitrary modules in the same
LLVMContext.
Differential Revision: https://reviews.llvm.org/D94712
This fixes the final (I think?) reference invalidation in `SmallVector`
that we need to fix to align with `std::vector`. (There is still some
left in the range insert / append / assign, but the standard calls that
UB for `std::vector` so I think we don't care?)
For POD-like types, reimplement `emplace_back()` in terms of
`push_back()`, taking a copy even for large `T` rather than lose the
realloc optimization in `grow_pod()`.
For other types, split the grow operation in three and construct the new
element in the middle.
- `mallocForGrow()` calculates the new capacity and returns the result
of `safe_malloc()`. We only need a single definition per
`SmallVectorBase` so this is defined in SmallVector.cpp to avoid code
size bloat. Moving this part of non-POD grow to the source file also
allows the logic to be easily shared with `grow_pod`, and
`report_size_overflow()` and `report_at_maximum_capacity()` can move
there too.
- `moveElementsForGrow()` moves elements from the old to the new
allocation.
- `takeAllocationForGrow()` frees the old allocation and saves the
new allocation and capacity .
`SmallVector:assign(size_type, const T&)` also uses the split-grow
operations for non-POD, but it also has a semantic change when not
growing. Previously, assign would start with `clear()`, and so the old
elements were destructed and all elements of the new vector were
copy-constructed (potentially invalidating references). The new
implementation skips destruction and uses copy-assignment for the prefix
of the new vector that fits. The new semantics match what libc++ does
for `std::vector::assign()`.
Note that the following is another possible implementation:
```
void assign(size_type NumElts, ValueParamT Elt) {
std::fill_n(this->begin(), std::min(NumElts, this->size()), Elt);
this->resize(NumElts, Elt);
}
```
The downside of this simpler implementation is that if the vector has to
grow there will be `size()` redundant copy operations.
(I had planned on splitting this patch up into three for committing
(after getting performance numbers / initial review), but I've realized
that if this does for some reason need to be reverted we'll probably
want to revert the whole package...)
Differential Revision: https://reviews.llvm.org/D94739
This recommits 71ed4b6ce5 with
the polarity of some of the pattern corrected.
Original commit message:
The custom expansion of select operations in the RISC-V backend
interferes with the matching of cmov instructions. Legalizing
select when the Zbt extension is available solves that problem.
Reviewed By: luismarques, craig.topper
Differential Revision: https://reviews.llvm.org/D93767
This is NFC-intended and removes the "OperationData"
class which had become nothing more than a recurrence
(reduction) type.
I adjusted the matching logic to distinguish
instructions from non-instructions - that's all that
the "IsLeafValue" member was keeping track of.
Fixes PR48523. When the linker errors with "output file too large",
one question that comes to mind is how the section sizes differ from
what they were previously. Unfortunately, this information is lost
when the linker exits without writing the output file. This change
makes it so that the error message includes the sizes of the largest
sections.
Reviewed By: MaskRay, grimar, jhenderson
Differential Revision: https://reviews.llvm.org/D94560
If a function doesn't contain loops and does not call non-willreturn
functions, then it is willreturn. Loops are detected by checking
for backedges in the function. We don't attempt to handle finite
loops at this point.
Differential Revision: https://reviews.llvm.org/D94633
`X86AsmParser::ParseIntelExpression` has a while loop. In the body,
calls to MCAsmLexer::UnLex can force a reallocation in the MCAsmLexer's
`CurToken` SmallVector, invalidating saved references to
`MCAsmLexer::getTok()`.
`const MCAsmToken &Tok` is such a saved reference, and this moves it
from outside the while loop to inside the body, fixing a
use-after-realloc.
`Tok` will still be reused across calls to `Lex()`, each of which
effectively destroys and constructs the pointed-to token. I'm a bit
skeptical of this usage pattern, but it seems broadly used in the
X86AsmParser (and others) so I'm leaving it alone (for now).
Somehow this bug was exposed by https://reviews.llvm.org/D94739,
resulting in test failures in dot-operator related tests in
llvm/test/tools/llvm-ml. I suspect the exposure path is related to
optimizer changes from splitting up the grow operation, but I haven't
dug all the way in. Regardless, there are already tests in tree that
cover this; they might fail consistently if we added ASan
instrumentation to SmallVector.
Differential Revision: https://reviews.llvm.org/D95112
Summary:
Prior to D91261 the information checked the OMP_MAP_TARGET_PARAM flag, change this as it has been removed. The INFO macro was changed to accept a flag as input to make conditionally printing information easier.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D95133
Defaulted destructor was treated inconsistently, compared to other
compiler-generated functions.
When Sema::IdentifyCUDATarget() got called on just-created dtor which didn't
have implicit __host__ __device__ attributes applied yet, it would treat it as a
host function. That happened to (sometimes) hide the error when dtor referred
to a host-only functions.
Even when we had identified defaulted dtor as a HD function, we still treated it
inconsistently during selection of usual deallocators, where we did not allow
referring to wrong-side functions, while it is allowed for other HD functions.
This change brings handling of defaulted dtors in line with other HD functions.
Differential Revision: https://reviews.llvm.org/D94732
The place-holding implementation of C_LOC just didn't work
when used with our more complete semantic checking, specifically
in the case of a polymorphic argument; convert it to an external
function with an implicit interface. C_ASSOCIATED needs to be
a generic interface with specific implementations for C_PTR and
C_FUNPTR.
Differential Revision: https://reviews.llvm.org/D94714
Profiling has been recently implemented in libomptarget (D93055). This patch enables time profiling support for libomptarget in libomp, to support profiling of multi-threaded execution of offloaded regions.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D94855
All Fortran options should be set in `CompilerInstance` (via its
`CompilerInvocation`) before any of `FrontendAction` is entered -
that's one of the tasks of the driver. However, this is a bit tricky
with fixed and free from detection introduced in
https://reviews.llvm.org/D94228.
Fixed-free form detection needs to happen:
* before any frontend action (we need to specify `isFixedForm` in
`Fortran::parser::Options` before running any actions)
* separately for every input file (we might be compiling multiple
Fortran files, some in free form, some in fixed form)
In other words, we need this to happen early (before any
`FrontendAction`), but not too early (we need to know what the current
input file is). In practice, `isFixedForm` can only be set later
than other options (other options are inferred from compiler flags). So
we can't really set all of them in one place, which is not ideal.
All changes in this patch are NFCs (hence no new tests). Quick summary:
* move fixed/free form detection from `FrontendAction::ExecuteAction` to
`CompilerInstance::ExecuteAction`
* add a bool flag in `FrontendInputFile` to mark a file as fixed/free
form
* updated a few comments
Differential Revision: https://reviews.llvm.org/D95042
Define OrderedOp and UnorderedOp instructions in SPIR-V and convert
cmpf operations with `ord` and `uno` tag to these instructions
respectively.
Differential Revision: https://reviews.llvm.org/D95098
The SPIR-V spec uses OpSpecConstantOp. Using an inconsistent name
makes the dialect generation scripts fail. Update to use the right
operation name, and fix the auto generation scripts as well.
Differential Revision: https://reviews.llvm.org/D95097
This pass is required to get correct codegen for image instructions with
the tfe or lwe bits set.
Differential Revision: https://reviews.llvm.org/D95132
Pretty similar to D95058, this patch added forward declaration for
CUDA atomic functions. We already have definitions with right mangled names in
internal CUDA headers so the forward declaration here can work properly.
Reviewed By: jdoerfert, JonChesterfield
Differential Revision: https://reviews.llvm.org/D95085
Make this look more like the DAG handling and move to common code.
I also noticed AArch64 seems to not be properly adding the
physreg:virtreg mapping to the function live ins.
Allow parsing generated mir with custom pseudo source value tokens.
Also rename pseudo source values to have more meaningful names.
Differential Revision: https://reviews.llvm.org/D94768
It turns out the vectorizer calls the getIntrinsicInstrCost functions
with a scalar return type and vector VF. This updates the costmodel to
handle that, still producing the correct vector costs.
A vectorizer test is added to show it vectorizing at the correct factor
again.
This patch makes sure that diagnostics from the prescanner are reported
when running `flang-new -E` (i.e. only the preprocessor phase is
requested). More specifically, the `PrintPreprocessedAction` action is
updated.
With this patch we make sure that the `f18` and `flang-new` provide
identical output when running the preprocessor and the prescanner
generates diagnostics.
Differential Revision: https://reviews.llvm.org/D94782
Summary:
The custom mapper API did not previously support the mapping names added previously. This means they were not present if a user requested debugging information while using the mapper functions. This adds basic support for passing the mapped names to the runtime library.
Reviewers: jdoerfert
Differential Revision: https://reviews.llvm.org/D94806
GCC/libstdc++ before 6.1 can't handle scoped enums as unordered_map keys. LLVM
(and some build) bots officially support some GCC 5.x versions, so this patch
just makes the enum unscoped until we can require GCC 6.x.
In https://llvm.org/PR48810 , we are crashing while trying to
propagate attributes from mempcpy (returns void*) to memcpy
(returns nothing - void).
We can avoid the crash by removing known incompatible
attributes for the void return type.
I'm not sure if this goes far enough (should we just drop all
attributes since this isn't the same function?). We also need
to audit other transforms in LibCallSimplifier to make sure
there are no other cases that have the same problem.
Differential Revision: https://reviews.llvm.org/D95088
Add DemandedElts support inside the TRUNCATE analysis.
REAPPLIED - this was reverted by @hans at rGa51226057fc3 due to an issue with vector shift amount types, which was fixed in rG935bacd3a724 and an additional test case added at rG0ca81b90d19d
Differential Revision: https://reviews.llvm.org/D56387
As noticed on D56387, for vectors we must always correctly adjust the shift amount type during truncation (not just after legalization). We were getting away with it as we currently only accepted scalars via the dyn_cast<ConstantSDNode>.
The original patch got reverted as a dependency of cf1c774d6a .
That patch got relanded so it's also necessary to reland this patch.
Original summary:
After cf1c774d6a, Clang seems to generate code
that is more similar to icc/Clang, so we can use the same line numbers for
all compilers in this test.
The DSYM variant of this test is failing since D94890. But as we explicitly
try to disable the DSYM generation in the makefile and build the archive on
our own, I don't see why we even need to run the DSYM version of the test.
This patch disables the generated derived versions of this test for the
different debug information containers (which includes the failing DSYM one).
Currently when LLDB has enough data in the debug information to import the `std` module,
it will just try to import it. However when debugging libraries where the sources aren't
available anymore, importing the module will generate a confusing diagnostic that
the module couldn't be built.
For the fallback mode (where we retry failed expressions with the loaded module), this
will cause the second expression to fail with a module built error instead of the
actual parsing issue in the user expression.
This patch adds checks that ensures that we at least have any source files in the found
include paths before we try to import the module. This prevents the module from being
loaded in the situation described above which means we don't emit the bogus 'can't
import module' diagnostic and also don't waste any time retrying the expression in the
fallback mode.
For the unit tests I did some refactoring as they now require a VFS with the files in it
and not just the paths. The Python test just builds a binary with a fake C++ module,
then deletes the module before debugging.
Fixes rdar://73264458
Reviewed By: JDevlieghere
Differential Revision: https://reviews.llvm.org/D95096
Walter Erquinigo