This patch adds unique idenfitiers to the existing OpenMP remarks. This makes
it easier to identify the corresponding documentation for each remark that will
be hosted in the OpenMP webpage.
Depends on D105898
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105939
This patch rewrites and reworks a few of the existing remarks to make the mmore
concise and consistent prior to writing the documentation for them.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105898
A common use of `ChangeStatus` is as follows:
```
ChangeStatus Changed = ChangeStatus::UNCHANGED;
Changed |= foo();
```
where `foo` returns `ChangeStatus` as well. Currently `ChangeStatus` doesn't
support compound assignment, we have to write as
```
Changed = Changed | foo();
```
which is not that convenient.
This patch add the support for compound assignment for `ChangeStatus`. Compound
assignment is usually implemented as a member function, and binary arithmetic
operator is therefore implemented using compound assignment. However, unlike
regular C++ class, enum class doesn't support member functions. As a result, they
can only be implemented in the way shown in the patch.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D106109
In the device runtime there are many function calls to `__kmpc_is_spmd_exec_mode`
to query the execution mode of current kernels. In many cases, user programs
only contain target region executing in one mode. As a consequence, those runtime
function calls will only return one value. If we can get rid of these function
calls during compliation, it can potentially improve performance.
In this patch, we use `AAKernelInfo` to analyze kernel execution. Basically, for
each kernel (device) function `F`, we collect all kernel entries `K` that can
reach `F`. A new AA, `AAFoldRuntimeCall`, is created for each call site. In each
iteration, it will check all reaching kernel entries, and update the folded value
accordingly.
In the future we will support more function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105787
This patch adds a feature to AACallEdges AbstractAttribute that allows
users to ask if there is a unknown callee that isn't a inline assembly.
This feature is needed by some of it's users.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105992
In the device runtime there are many function calls to `__kmpc_is_spmd_exec_mode`
to query the execution mode of current kernels. In many cases, user programs
only contain target region executing in one mode. As a consequence, those runtime
function calls will only return one value. If we can get rid of these function
calls during compliation, it can potentially improve performance.
In this patch, we use `AAKernelInfo` to analyze kernel execution. Basically, for
each kernel (device) function `F`, we collect all kernel entries `K` that can
reach `F`. A new AA, `AAFoldRuntimeCall`, is created for each call site. In each
iteration, it will check all reaching kernel entries, and update the folded value
accordingly.
In the future we will support more function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105787
Just like intrinsics are not tracked for IFI.InlinedCalls, they should not be tracked for IFI.InlinedCallSites.
In the current top-of-tree this change is a NFC, but the full restrict patches (D68484) potentially trigger an read-after-free
if intrinsics are also added to the InlindeCallSites, due to a late optimization potentially removing some of the inlined intrinsics.
Also see https://lists.llvm.org/pipermail/llvm-dev/2021-July/151722.html for a discussion about the problem.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D105805
Continuing from D105763, this allows placing certain properties
about attributes in the TableGen definition. In particular, we
store whether an attribute applies to fn/param/ret (or a combination
thereof). This information is used by the Verifier, as well as the
ForceFunctionAttrs pass. I also plan to use this in LLParser,
which also duplicates info on which attributes are valid where.
This keeps metadata about attributes in one place, and makes it
more likely that it stays in sync, rather than in various
functions spread across the codebase.
Differential Revision: https://reviews.llvm.org/D105780
The patch templatize PriorityInlinerOrder so that it can accept any type priority metric.
Reviewed By: kazu
Differential Revision: https://reviews.llvm.org/D104972
As with other Attributor interfaces we often want to know if assumed
information was used to answer a query. This is important if only
known information is allowed or if known information can lead to an
early fixpoint. The users have been adjusted but none of them utilizes
the new information yet.
AllocationInfo and DeallocationInfo objects themselves are allocated
with the Attributor bump allocator and do not need to be deallocated.
That said, the sets in AllocationInfo and DeallocationInfo need to be
destroyed to avoid memory leaks.
In the spirit of TRegions [0], this patch analyzes a kernel and tracks
if it can be executed in SPMD-mode. If so, we flip the arguments of
the __kmpc_target_init and deinit call to enable the mode. We also
update the `<kernel>_exec_mode` flag to indicate to the runtime we
changed the mode to SPMD.
The code analysis is done interprocedurally by extending the
AAKernelInfo abstract attribute to track SPMD compatibility as well.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
Differential Revision: https://reviews.llvm.org/D102307
In the spirit of TRegions [0], this patch creates a custom state
machine for a generic target region based on the potentially called
parallel regions.
The code analysis is done interprocedurally via an abstract attribute
(AAKernelInfo). All outermost parallel regions are collected and we
check if there might be unknown outermost parallel regions for which
we need an indirect call. Other AAKernelInfo extensions are expected.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
Differential Revision: https://reviews.llvm.org/D101977
In the spirit of TRegions [0], this patch provides a simpler and uniform
interface for a kernel to set up the device runtime. The OMPIRBuilder is
used for reuse in Flang. A custom state machine will be generated in the
follow up patch.
The "surplus" threads of the "master warp" will not exit early anymore
so we need to use non-aligned barriers. The new runtime will not have an
extra warp but also require these non-aligned barriers.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
This was in parts extracted from D59319.
Reviewed By: ABataev, JonChesterfield
Differential Revision: https://reviews.llvm.org/D101976
When we talk to outside analyse, e.g., LVI and ScalarEvolution, we need
to be careful with the query. The particular error occurred because we
folded a PHI node before the LVI query but the context location was now
not dominated by the value anymore. This is not supported by LVI so we
have to filter these situations before we query the outside analyses.
In order to simplify future extensions, e.g., the merge of
AAHeapToShared in to AAHeapToStack, we reorganize AAHeapToStack and the
state we keep for each malloc-like call. The result is also less
confusing as we only track malloc-like calls, not all calls. Further, we
only perform the updates necessary for a malloc-like to argue it can go
to the stack, e.g., we won't check all uses if we moved on to the
"must-be-freed" argument.
This patch also uses Attributor helps to simplify the allocated size,
alignment, and the potentially freed objects.
Overall, this is mostly a reorganization and only the use of the
optimistic helpers should change (=improve) the capabilities a bit.
Differential Revision: https://reviews.llvm.org/D104993
We have to be careful when we replace values to not use a non-dominating
instruction. It makes sense that simplification offers those as
"simplified values" but we can't manifest them in the IR without PHI
nodes. In the future we should consider potentially adding those PHI
nodes.
We should use AAValueSimplify for all value simplification, however
there was some leftover logic that predates AAValueSimplify in
AAReturnedValues. This remove the AAReturnedValues part and provides a
replacement by making AAValueSimplifyReturned strong enough to handle
all previously covered cases. Further, this improve
AAValueSimplifyCallSiteReturned to handle returned arguments.
AAReturnedValues is now much easier and the collected returned
values/instructions are now from the associated function only, making it
much more sane. We also do not have the brittle logic anymore that looks
for unresolved calls. Instead, we use AAValueSimplify to handle
recursion.
Useful code has been split into helper functions, e.g., an Attributor
interface to get a simplified value.
Differential Revision: https://reviews.llvm.org/D103860
As the `llvm::getUnderlyingObjects` helper, the optimistic version
collects objects that might be the base of a given pointer. In contrast
to the llvm variant, the optimistic one will use assumed information,
e.g., about select conditions or dead blocks, to provide a more precise
result.
Differential Revision: https://reviews.llvm.org/D103859
Not all attributes are able to handle the interprocedural step and
follow the uses into a call site. Let them be able to combine call site
uses instead. This might result in some unused values/arguments being
leftover but it removes problems where we misused "is dead" even though
it was actually "is simplified/replaced".
We explicitly check for dead values due to constant propagation in
`AAIsDeadValueImpl::areAllUsesAssumedDead` instead.
Differential Revision: https://reviews.llvm.org/D103858
Broke check-clang, see https://reviews.llvm.org/D102307#2869065
Ran `git revert -n ebbe149a6f08535ede848a531a601ae6591cfbc5..269416d41908bb670f67af689155d5ab8eea689a`
As with other Attributor interfaces we often want to know if assumed
information was used to answer a query. This is important if only
known information is allowed or if known information can lead to an
early fixpoint. The users have been adjusted but none of them utilizes
the new information yet.
In the spirit of TRegions [0], this patch analyzes a kernel and tracks
if it can be executed in SPMD-mode. If so, we flip the arguments of
the __kmpc_target_init and deinit call to enable the mode. We also
update the `<kernel>_exec_mode` flag to indicate to the runtime we
changed the mode to SPMD.
The code analysis is done interprocedurally by extending the
AAKernelInfo abstract attribute to track SPMD compatibility as well.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
Differential Revision: https://reviews.llvm.org/D102307
When we talk to outside analyse, e.g., LVI and ScalarEvolution, we need
to be careful with the query. The particular error occurred because we
folded a PHI node before the LVI query but the context location was now
not dominated by the value anymore. This is not supported by LVI so we
have to filter these situations before we query the outside analyses.
We have to be careful when we replace values to not use a non-dominating
instruction. It makes sense that simplification offers those as
"simplified values" but we can't manifest them in the IR without PHI
nodes. In the future we should consider potentially adding those PHI
nodes.
In the spirit of TRegions [0], this patch creates a custom state
machine for a generic target region based on the potentially called
parallel regions.
The code analysis is done interprocedurally via an abstract attribute
(AAKernelInfo). All outermost parallel regions are collected and we
check if there might be unknown outermost parallel regions for which
we need an indirect call. Other AAKernelInfo extensions are expected.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
Differential Revision: https://reviews.llvm.org/D101977
In the spirit of TRegions [0], this patch provides a simpler and uniform
interface for a kernel to set up the device runtime. The OMPIRBuilder is
used for reuse in Flang. A custom state machine will be generated in the
follow up patch.
The "surplus" threads of the "master warp" will not exit early anymore
so we need to use non-aligned barriers. The new runtime will not have an
extra warp but also require these non-aligned barriers.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
This was in parts extracted from D59319.
Reviewed By: ABataev, JonChesterfield
Differential Revision: https://reviews.llvm.org/D101976
In order to simplify future extensions, e.g., the merge of
AAHeapToShared in to AAHeapToStack, we reorganize AAHeapToStack and the
state we keep for each malloc-like call. The result is also less
confusing as we only track malloc-like calls, not all calls. Further, we
only perform the updates necessary for a malloc-like to argue it can go
to the stack, e.g., we won't check all uses if we moved on to the
"must-be-freed" argument.
This patch also uses Attributor helps to simplify the allocated size,
alignment, and the potentially freed objects.
Overall, this is mostly a reorganization and only the use of the
optimistic helpers should change (=improve) the capabilities a bit.
Differential Revision: https://reviews.llvm.org/D104993
We should use AAValueSimplify for all value simplification, however
there was some leftover logic that predates AAValueSimplify in
AAReturnedValues. This remove the AAReturnedValues part and provides a
replacement by making AAValueSimplifyReturned strong enough to handle
all previously covered cases. Further, this improve
AAValueSimplifyCallSiteReturned to handle returned arguments.
AAReturnedValues is now much easier and the collected returned
values/instructions are now from the associated function only, making it
much more sane. We also do not have the brittle logic anymore that looks
for unresolved calls. Instead, we use AAValueSimplify to handle
recursion.
Useful code has been split into helper functions, e.g., an Attributor
interface to get a simplified value.
Differential Revision: https://reviews.llvm.org/D103860
As the `llvm::getUnderlyingObjects` helper, the optimistic version
collects objects that might be the base of a given pointer. In contrast
to the llvm variant, the optimistic one will use assumed information,
e.g., about select conditions or dead blocks, to provide a more precise
result.
Differential Revision: https://reviews.llvm.org/D103859
Not all attributes are able to handle the interprocedural step and
follow the uses into a call site. Let them be able to combine call site
uses instead. This might result in some unused values/arguments being
leftover but it removes problems where we misused "is dead" even though
it was actually "is simplified/replaced".
We explicitly check for dead values due to constant propagation in
`AAIsDeadValueImpl::areAllUsesAssumedDead` instead.
Differential Revision: https://reviews.llvm.org/D103858
This patch teaches the sample profile loader to merge function
attributes after inlining functions.
Without this patch, the compiler could inline a function requiring the
512-bit vector width into its caller without merging function
attributes, triggering a failure during instruction selection.
Differential Revision: https://reviews.llvm.org/D105729
This change is intended as initial setup. The plan is to add
more semantic checks later. I plan to update the documentation
as more semantic checks are added (instead of documenting the
details up front). Most of the code closely mirrors that for
the Swift calling convention. Three places are marked as
[FIXME: swiftasynccc]; those will be addressed once the
corresponding convention is introduced in LLVM.
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D95561
There was an alias between 'simplifycfg' and 'simplify-cfg' in the
PassRegistry. That was the original reason for this patch, which
effectively removes the alias.
This patch also replaces all occurrances of 'simplify-cfg'
by 'simplifycfg'. Reason for choosing that form for the name is
that it matches the DEBUG_TYPE for the pass, and the legacy PM name
and also how it is spelled out in other passes such as
'loop-simplifycfg', and in other options such as
'simplifycfg-merge-cond-stores'.
I for some reason the name should be changed to 'simplify-cfg' in
the future, then I think such a renaming should be more widely done
and not only impacting the PassRegistry.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D105627
- ``externally_initialized`` variables would be initialized or modified
elsewhere. Particularly, CUDA or HIP may have host code to initialize
or modify ``externally_initialized`` device variables, which may not
be explicitly referenced on the device side but may still be used
through the host side interfaces. Not preserving them triggers the
elimination of them in the GlobalDCE and breaks the user code.
Reviewed By: yaxunl
Differential Revision: https://reviews.llvm.org/D105135
Before we replaced value by registering all their uses. However, as we
replace a value old uses become stale. We now replace values explicitly
and keep track of "new values" when doing so to avoid replacing only
uses in stale/old values but not their replacements.
We often need to deal with the value lattice that contains none and
undef as special values. A simple helper makes this much nicer.
Differential Revision: https://reviews.llvm.org/D103857
When we do simplification via AAPotentialValues or AAValueConstantRange
we need to simplify the operands of an instruction we deconstruct first.
This does not only improve the result, see for example range.ll, but is
required as we allow outside AAs to provide simplification rules via
callbacks. If we do ignore the simplification rules and base other
simplifications on the IR instead we can create an inconsistent state.
The remarks will trigger on some functions that are marked cold, such as the
`__muldc3` intrinsic functions. Change the remarks to avoid these functions.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105196
This patch adds additional remarks, suggesting the use of `noescape` for failed
globalization and indicating when internalization failed.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D105150
Now the option is off by default. Since we are not sure if this option
would make the compile time increase aggressively. Although we tested it
on SPEC2017, we may need to test more to make it on by default.
Reviewed By: SjoerdMeijer
Differential Revision: https://reviews.llvm.org/D104365
Currently OpenMPOpt will only check if a function is a kernel before deciding not to internalize it. Any uncalled function that gets internalized will be trivially dead in the module so this is unnnecessary.
Depends on D102423
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104890
Increase the number of attributor iterations on a GPU target. I forgot to
change this in D104416.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104920
The metadata added in D102361 introduces a module flag that we can check
to determine if the module was compiled with `-fopenmp` enables. We can
now check for the precense of this instead of scanning the call graph
for OpenMP runtime functions.
Depends on D102361
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102423
WPD currently assumes that there is a one to one correspondence between
type test assume sequences and virtual calls. However, with
-fstrict-vtable-pointers this may not be true. This ends up causing
crashes when we try to optimize a virtual call more than once (
applyUniformRetValOpt()/applyUniqueRetValOpt()/applyVirtualConstProp()/applySingleImplDevirt()).
applySingleImplDevirt() actually didn't previous crash because it would
replace the devirtualized call with the same direct call. Adding an
assert that the call is indirect causes the corresponding test to crash
with the rest of the patch.
This makes Chrome successfully build with -fstrict-vtable-pointers + WPD.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D104798
Create an internal alias with the original name for static functions
that are renamed in promoteInternals to avoid breaking inline
assembly references to them.
This relands commit 4474958d3a
with a fix to a use-of-uninitialized-value error that tripped
MemorySanitizer.
Link: https://github.com/ClangBuiltLinux/linux/issues/1354
Reviewed By: nickdesaulniers, pcc
Differential Revision: https://reviews.llvm.org/D104058
This attribute uses Attributor's internal 'optimistic' call graph
information to answer queries about function call reachability.
Functions can become reachable over time as new call edges are
discovered.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104599
This patch makes PriorityInlineOrder lazily updated.
The PriorityInlineOrder would lazily update the desirability of a call site if it's decreasing.
Reviewed By: kazu
Differential Revision: https://reviews.llvm.org/D104654
This patch fixes a problem with the AAExecutionDomain attributor not
checking if it is in a valid state. This can cause it to incorrectly
return that a block is executed in a single threaded context after the
attributor failed for any reason.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D103186
After landing the globalization optimizations, the precense of globalization on
the device that was not put in shared or stack memory is a failed optimization
with performance consequences so it should indicate a missed remark.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104735
Right now the Attributor defaults to 32 fixed point iterations unless it is set
explicitly by a command line flag. This patch allows this to be configured when
the attributor instance is created. The maximum is then increased in OpenMPOpt
if the target is a kernel. This is because the globalization analysis can result
in larger iteration counts due to many dependent instances running at once.
Depends on D102444
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104416
Summary:
This patch adds support for the Attributor to emit remarks on behalf of some
other pass. The attributor can now optionally take a callback function that
returns an OptimizationRemarkEmitter object when given a Function pointer. If
this is availible then a remark will be emitted for the corresponding pass
name.
Depends on D102197
Reviewed By: sstefan1 thegameg
Differential Revision: https://reviews.llvm.org/D102444
Summary:
The changes to globalization introduced in D97680 introduce a large amount of overhead by default. The old globalization method would always ignore globalization code if executing in SPMD mode. This wasn't strictly correct as data sharing is still possible in SPMD mode. The new interface is correct but introduces globalization code even when unnecessary. This optimization will use the existing HeapToStack transformation in the attributor to allow for unneeded globalization to be replaced with thread-private stack memory. This is done using the newly introduced library instances for the RTL functions added in D102087.
Depends on D97818
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102197
Summary:
Currently the attributor needs to give up if a function has external linkage.
This means that the optimization introduced in D97818 will only apply to static
functions. This change uses the Attributor to internalize OpenMP device
routines by making a copy of each function with private linkage and replacing
the uses in the module with it. This allows for the optimization to be applied
to any regular function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102824
Summary:
The changes introduced in D97680 create a simpler interface to code that needs
to be globalized. This interface is used to simplify the globalization calls in
the middle end. We can check any globalization call that is only called by a
single thread in the team and replace it with a static shared memory buffer.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D97818
Summary:
Memory globalization is required to maintain OpenMP standard semantics for data sharing between
worker and master threads. The GPU cannot share data between its threads so must allocate global or
shared memory to store the data in. Currently this is implemented fully in the frontend using the
`__kmpc_data_sharing_push_stack` and __kmpc_data_sharing_pop_stack` functions to emulate standard
CPU stack sharing. The front-end scans the target region for variables that escape the region and
must be shared between the threads. Each variable then has a field created for it in a global record
type.
This patch replaces this functinality with a single allocation command, effectively mimicing an
alloca instruction for the variables that must be shared between the threads. This will be much
slower than the current solution, but makes it much easier to optimize as we can analyze each
variable independently and determine if it is not captured. In the future, we can replace these
calls with an `alloca` and small allocations can be pushed to shared memory.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D97680
getSpecializationCost was returning INT_MAX for a case when specialisation
shouldn't happen, but this wasn't properly checked if specialisation was
forced.
Differential Revision: https://reviews.llvm.org/D104461
This patch adds an optional PriorityInlineOrder, which uses the heap to order inlining.
The callsite which size is smaller would have a higher priority.
Reviewed By: mtrofin
Differential Revision: https://reviews.llvm.org/D104028
As a follow-up to https://reviews.llvm.org/D104129, I'm cleaning up the danling probe related code in both the compiler and llvm-profgen.
I'm seeing a 5% size win for the pseudo_probe section for SPEC2017 and 10% for Ciner. Certain benchmark such as 602.gcc has a 20% size win. No obvious difference seen on build time for SPEC2017 and Cinder.
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D104477
This patch adds an optional PriorityInlineOrder, which uses the heap to order inlining.
The callsite which size is smaller would have a higher priority.
Reviewed By: mtrofin
Differential Revision: https://reviews.llvm.org/D104028
This looks like not a practical pattern in our codebase (it could fail
in some sandbox environement).
Instead we print it via standard output, and it is controled by the
-attributor-print-call-graph, this follows a similiar pattern of attributor-print-dep.
This should fix PR50683. The wrong assumption was that we
could always know what the callee is when we replace a call site
argument with undef. We wanted to know that to remove the `noundef`
that might be attached to the argument. Since no callee means we
did the propagation on the caller site, there is no need to remove
an attribute. It is only needed if we replace all uses and therefore
pass `undef` instead of the value that was passed in otherwise.
To allow outside AAs that simplify values we need to ensure all value
simplification goes through the Attributor, not AAValueSimplify (or any
of the other AAs we have already like AAPotentialValues). This patch
also introduces an interface for the outside AAs to register
simplification callbacks for an IRPosition. To make this work as
expected we have to pass IRPositions instead of Values in
AAValueSimplify, which makes sense by itself.
If we simplify values we sometimes end up with type mismatches. If the
value is a constant we can often cast it though to still allow
propagation. The logic is now put into a helper and it replaces some
ad hoc things we did before.
This also introduces the AA namespace for abstract attribute related
functions and types.
Differential Revision: https://reviews.llvm.org/D103856
If the target stack is not accessible between different running
"threads" we have to make sure not to create allocas for mallocs
that might be used by multiple "threads". The "use check" is
sufficient to prevent this but if we apply the "free check" we have
to make sure the pointer is not communicated to others before
the free is reached.
Differential Revision: https://reviews.llvm.org/D98608
The initial use for AAExecutionDomain was to determine if a single
thread executes a block. While this is sometimes informative most
of the time, and for other reasons, we actually want to know if it
is the "initial thread". Thus, the thread that started execution on
the current device. The deduction needs to be adjusted in a follow
up as the methods we use right not are looking for the OpenMP thread
id which is resets whenever a thread enters a parallel region. What
we basically want is to look for `llvm.nvvm.read.ptx.sreg.ntid.x` and
equivalent functions.
We invalidated AAReachabilityImpl directly which is not helpful and
confusing as we still used it regardless. We now avoid invalidating it
(not needed anyway) and add checks for the state. This has by itself no
actual effect but prepares for later extensions.
This attribute computes the optimistic live call edges using the attributor
liveness information. This attribute will be used for deriving a
inter-procedural function reachability attribute.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D104059
The original implementation calculating UserBonus uses operator ^, which means XOR in C++
language.
At the first glance of reviewing, I thought it should be power, my bad.
It doesn't make sense to use XOR here. So I believe it should be a
carelessness as I made.
Test Plan: check-all
Reviewed By: SjoerdMeijer
Differential Revision: https://reviews.llvm.org/D104282
This adds support for functions outlined by the IR Outliner to be
recognized by the debugger. The expected behavior is that it will
skip over the instructions included in that section. This is due to the
fact that we can not say which of the original locations the
instructions originated from.
These functions will show up in the call stack, but you cannot step
through them.
Reviewers: paquette, vsk, djtodoro
Differential Revision: https://reviews.llvm.org/D87302
<string> is currently the highest impact header in a clang+llvm build:
https://commondatastorage.googleapis.com/chromium-browser-clang/llvm-include-analysis.html
One of the most common places this is being included is the APInt.h header, which needs it for an old toString() implementation that returns std::string - an inefficient method compared to the SmallString versions that it actually wraps.
This patch replaces these APInt/APSInt methods with a pair of llvm::toString() helpers inside StringExtras.h, adjusts users accordingly and removes the <string> from APInt.h - I was hoping that more of these users could be converted to use the SmallString methods, but it appears that most end up creating a std::string anyhow. I avoided trying to use the raw_ostream << operators as well as I didn't want to lose having the integer radix explicit in the code.
Differential Revision: https://reviews.llvm.org/D103888
This adds a function specialization pass to LLVM. Constant parameters
like function pointers and constant globals are propagated to the callee by
specializing the function.
This is a first version with a number of limitations:
- The pass is off by default, so needs to be enabled on the command line,
- It does not handle specialization of recursive functions,
- It does not yet handle constants and constant ranges,
- Only 1 argument per function is specialised,
- The cost-model could be further looked into, and perhaps related,
- We are not yet caching analysis results.
This is based on earlier work by Matthew Simpson (D36432) and Vinay Madhusudan.
More recently this was also discussed on the list, see:
https://lists.llvm.org/pipermail/llvm-dev/2021-March/149380.html.
The motivation for this work is that function specialisation often comes up as
a reason for performance differences of generated code between LLVM and GCC,
which has this enabled by default from optimisation level -O3 and up. And while
this certainly helps a few cpu benchmark cases, this also triggers in real
world codes and is thus a generally useful transformation to have in LLVM.
Function specialisation has great potential to increase compile-times and
code-size. The summary from some investigations with this patch is:
- Compile-time increases for short compile jobs is high relatively, but the
increase in absolute numbers still low.
- For longer compile-jobs, the extra compile time is around 1%, and very much
in line with GCC.
- It is difficult to blame one thing for compile-time increases: it looks like
everywhere a little bit more time is spent processing more functions and
instructions.
- But the function specialisation pass itself is not very expensive; it doesn't
show up very high in the profile of the optimisation passes.
The goal of this work is to reach parity with GCC which means that eventually
we would like to get this enabled by default. But first we would like to address
some of the limitations before that.
Differential Revision: https://reviews.llvm.org/D93838
Summary:
The current implementation of AANoFreeFloating will incorrectly list floating
point loads and stores as may-free. This prevents other attributor instances
like HeapToStack from pushing some allocations to the stack.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D103975
This patch abstract Calls in Inliner:run() to InlineOrder.
With this patch, it's possible to customize the inlining order,
e.g. use queue or priority queue.
Reviewed By: kazu
Differential Revision: https://reviews.llvm.org/D103315
This patch abstract Calls in Inliner:run() to InlineOrder.
With this patch, it's possible to customize the inlining order, i.e. use queue or priority queue.
Reviewed By: kazu
Differential Revision: https://reviews.llvm.org/D103315
This patch was split from https://reviews.llvm.org/D102246
[SampleFDO] New hierarchical discriminator for Flow Sensitive SampleFDO
This is for llvm-profdata part of change. It sets the bit masks for the
profile reader in llvm-profdata. Also add an internal option
"-fs-discriminator-pass" for show and merge command to process the profile
offline.
This patch also moved setDiscriminatorMaskedBitFrom() to
SampleProfileReader::create() to simplify the interface.
Differential Revision: https://reviews.llvm.org/D103550
This patch changes the `isKnownHeapToStack` and `isAssumedHeapToStack`
member functions to return if a function call is going to be altered by
HeapToStack.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D103574
This patch was split from https://reviews.llvm.org/D102246
[SampleFDO] New hierarchical discriminator for Flow Sensitive SampleFDO
This is mainly for ProfileData part of change. It will load
FS Profile when such profile is detected. For an extbinary format profile,
create_llvm_prof tool will add a flag to profile summary section.
For other format profiles, the users need to use an internal option
(-profile-isfs) to tell the compiler that the profile uses FS discriminators.
This patch also simplified the bit API used by FS discriminators.
Differential Revision: https://reviews.llvm.org/D103041
This is split off from D102002, and I think it is clear that
the difference in behavior was not intended. Options were
added to SimplifyCFG over time, but different chunks of
the pass pipelines were not kept in sync.
AIX use `__ssp_canary_word` instead of `__stack_chk_guard`.
This patch update the target hook to use correct symbol,
so that the basic stackprotect feature can work.
The traceback will be handled in follow up patch.
Reviewed By: #powerpc, shchenz
Differential Revision: https://reviews.llvm.org/D103100
When the lower type test pass is invoked a second time with
DropTypeTests set to true, it expects that all remaining type tests feed
assume instructions, which are removed along with the type tests.
In some cases the llvm.assume might have been merged with another one,
i.e. from a builtin_assume instruction, in which case the type test
would actually feed a phi that in turn feeds the merged assume
instruction. In this case we can simply replace that operand of the phi
with "true" before removing the type test.
Differential Revision: https://reviews.llvm.org/D103073
Beside the `comdat any` deduplication feature, instrumentations use comdat to
establish dependencies among a group of sections, to prevent section based
linker garbage collection from discarding some members without discarding all.
LangRef acknowledges this usage with the following wording:
> All global objects that specify this key will only end up in the final object file if the linker chooses that key over some other key.
On ELF, for PGO instrumentation, a `__llvm_prf_cnts` section and its associated
`__llvm_prf_data` section are placed in the same GRP_COMDAT group. A
`__llvm_prf_data` is usually not referenced and expects the liveness of its
associated `__llvm_prf_cnts` to retain it.
The `setComdat(nullptr)` code (added by D10679) in InternalizePass can break the
use case (a `__llvm_prf_data` may be dropped with its associated `__llvm_prf_cnts` retained).
The main goal of this patch is to fix the dependency relationship.
I think it makes sense for InternalizePass to internalize a comdat and thus
suppress the deduplication feature, e.g. a relocatable link of a regular LTO can
create an object file affected by InternalizePass.
If a non-internal comdat in a.o is prevailed by an internal comdat in b.o, the
a.o references to the comdat definitions will be non-resolvable (references
cannot bind to STB_LOCAL definitions in b.o).
On PE-COFF, for a non-external selection symbol, deduplication is naturally
suppressed with link.exe and lld-link. However, this is fuzzy on ELF and I tend
to believe the spec creator has not thought about this use case (see D102973).
GNU ld and gold are still using the "signature is name based" interpretation.
So even if D102973 for ld.lld is accepted, for portability, a better approach is
to rename the comdat. A comdat with one single member is the common case,
leaving the comdat can waste (sizeof(Elf64_Shdr)+4*2) bytes, so we optimize by
deleting the comdat; otherwise we rename the comdat.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D103043
We really ought to support no_sanitize("coverage") in line with other
sanitizers. This came up again in discussions on the Linux-kernel
mailing lists, because we currently do workarounds using objtool to
remove coverage instrumentation. Since that support is only on x86, to
continue support coverage instrumentation on other architectures, we
must support selectively disabling coverage instrumentation via function
attributes.
Unfortunately, for SanitizeCoverage, it has not been implemented as a
sanitizer via fsanitize= and associated options in Sanitizers.def, but
rolls its own option fsanitize-coverage. This meant that we never got
"automatic" no_sanitize attribute support.
Implement no_sanitize attribute support by special-casing the string
"coverage" in the NoSanitizeAttr implementation. To keep the feature as
unintrusive to existing IR generation as possible, define a new negative
function attribute NoSanitizeCoverage to propagate the information
through to the instrumentation pass.
Fixes: https://bugs.llvm.org/show_bug.cgi?id=49035
Reviewed By: vitalybuka, morehouse
Differential Revision: https://reviews.llvm.org/D102772
If we simplify values we sometimes end up with type mismatches. If the
value is a constant we can often cast it though to still allow
propagation. The logic is now put into a helper and it replaces some
ad hoc things we did before.
This also introduces the AA namespace for abstract attribute related
functions and types.
We have seen various problems when the call graph was not updated or
the updated did not succeed because it involved functions outside the
SCC. This patch adds assertions and checks to avoid accidentally
changing something outside the SCC that would impact the call graph.
It also prevents us from reanalyzing functions outside the current
SCC which could cause problems on its own. Note that the transformations
we do might cause the CG to be "more precise" but the original one would
always be a super set of the most precise one. Since the call graph is
by nature an approximation, it is good enough to have a super set of all
call edges.
The constant value lattice looks like this
```
<None>
|
<undef>
/ | \
... <0> ...
\ | /
<unknown>
```
We did not account for the undef and assumed a value meant we could not
change anymore. Now we actually check if we have the same value as
before, which will signal CHANGED to the users when we go from undef to
a specific constant.
This fixes, among other things, the bug exposed by @ipccp4 in
`value-simplify.ll`.
The state of AAPotentialValues tracks if undef is contained. It should
fold undef into the first non-undef value. However we missed a case
before. There was also a shadowing definition of two variables that
caused trouble. The test exposes both problems.
Add options -[no-]offload-lto and -foffload-lto=[thin,full] for controlling
LTO for offload compilation. Allow LTO for AMDGPU target.
AMDGPU target does not support codegen of object files containing
call of external functions, therefore the LLVM module passed to
AMDGPU backend needs to contain definitions of all the callees.
An LLVM option is added to allow function importer to import
functions with noinline attribute.
HIP toolchain passes proper LLVM options to lld to make sure
function importer imports definitions of all the callees.
Reviewed by: Teresa Johnson, Artem Belevich
Differential Revision: https://reviews.llvm.org/D99683
GlobalOpt can slice structs/arrays and change GEPs in the process,
but it was not updating alignments for load/store users. This
eventually causes the crashing seen in:
https://llvm.org/PR49661https://llvm.org/PR50253
On x86, this required SLP+codegen to create an aligned vector
store on an invalid address. The bugs would be easier to
demonstrate on a target with stricter alignment requirements.
I'm not sure if this is a complete solution. The alignment
updating code is adapted from InstCombine, so I assume that
part is tested and good.
Differential Revision: https://reviews.llvm.org/D102552
Summary:
Currently, only `OptimizationRemarks` can be emitted using a Function.
Add constructors to allow this for `OptimizationRemarksAnalysis` and
`OptimizationRemarkMissed` as well.
Reviewed By: jdoerfert thegameg
Differential Revision: https://reviews.llvm.org/D102784
Sample profile loader can be run in both LTO prelink and postlink. Currently the counts annoation in postilnk doesn't fully overwrite what's done in prelink. I'm adding a switch (`-overwrite-existing-weights=1`) to enable a full overwrite, which includes:
1. Clear old metadata for calls when their parent block has a zero count. This could be caused by prelink code duplication.
2. Clear indirect call metadata if somehow all the rest targets have a sum of zero count.
3. Overwrite branch weight for basic blocks.
With a CS profile, I was seeing #1 and #2 help reduce code size by preventing post-sample ICP and CGSCC inliner working on obsolete metadata, which come from a partial global inlining in prelink. It's not expected to work well for non-CS case with a less-accurate post-inline count quality.
It's worth calling out that some prelink optimizations can damage counts quality in an irreversible way. One example is the loop rotate optimization. Due to lack of exact loop entry count (profiling can only give loop iteration count and loop exit count), moving one iteration out of the loop body leaves the rest iteration count unknown. We had to turn off prelink loop rotate to achieve a better postlink counts quality. A even better postlink counts quality can be archived by turning off prelink CGSCC inlining which is not context-sensitive.
Reviewed By: wenlei, wmi
Differential Revision: https://reviews.llvm.org/D102537
Summary:
The OpenMP runtime functions don't always provide unique thread ID's to
determine if a basic block is truly single-threaded. Change the implementation
to only check NVPTX intrinsics for now.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102700
With prelink inlining, pseudo probes with same ID can come from different inline contexts. Such probes should not share samples and their factors should be fixed up separately.
I'm seeing 0.3% speedup for SPEC2017 overall. Benchmark 631.deepsjeng_s benefits the most, about 4%.
Reviewed By: wenlei, wmi
Differential Revision: https://reviews.llvm.org/D102429
This patch makes it possible to do call site specific deductions
for AAValueSimplification and AAIsDead.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D84722
findIndirectCallFunctionSamples will leave Sum uninitialized if it returns an empty vector, we don't really use Sum in this case (but we do make a copy that isn't used either) - so ensure we initialize the value to zero to at least silence the static analysis warning.
Currently we didn't support multiple return type, we work around to use error_code to represent:
1) The dangling probe.
2) Ignore the weight of non-probe instruction
While merging the instructions' weight for the whole BB, it will filter out the error code. But If all instructions of the BB give error_code, the outside logic will mark it as a BB requiring the inference algorithm to infer its weight. This is different from the zero value which will be treated as a cold block.
Fix one place that if we can't find the FunctionSamples in the profile data which indicates the BB is cold, we choose to return zero.
Also refine the comments.
Reviewed By: hoy, wenlei
Differential Revision: https://reviews.llvm.org/D102007
Summary:
This patch prevents the Attributor instances made in the CGSCC pass from
deleting functions. This prevents the attributor from changing the call
graph while OpenMPOpt is working with it.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102363
GlobalOpt implements a heap SROA (SROA for an malloc allocatated struct or array
of structs) which is largely undertested (heap-sra-[1234].ll are basically the
same test with very little difference) and does not trigger at all when
bootstrapping clang (it only supports the case of one single store).
The heap SROA implementation causes PR50027 (GEP is not properly handled; crash or miscompile).
Just drop the implementation. I have deleted some obviously duplicated tests
but kept `heap-sra-[12]{,-no-nullopt}.ll`.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D102257
Make sure the alignment of the generated operations matches the
alignment of the byval argument. Previously, we were just ignoring
alignment and getting lucky.
While I'm here, also delete the unnecessary "tail" handling.
Passing a pointer to a byval argument to a "tail" call is UB, so
rewriting to an alloca doesn't require any special handling.
Differential Revision: https://reviews.llvm.org/D89819
This is better no-functional-change-intended than the 1st attempt.
As noted in D102002, there were at least 2 diffs that went
unchecked in pass manager regressions tests: different pass
parameters (SimplifyCFG) and an extension point/callback.
Those should be lifted from the original code blocks correctly
now.
This reverts commit fefcb1f878.
It was supposed to be NFC, but as noted in the post-commit
comments in D102002, that was not true: SimplifyCFG uses
different parameters and there's a difference in an
extension point / callback.
Printing pass manager invocations is fairly verbose and not super
useful.
This allows us to remove DebugLogging from pass managers and PassBuilder
since all logging (aside from analysis managers) goes through
instrumentation now.
This has the downside of never being able to print the top level pass
manager via instrumentation, but that seems like a minor downside.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D101797
Summary:
Add the AAExecutionDomainInfo attributor instance to OpenMPOpt.
This will infer information relating to domain information that an
instruction might be expecting in. Right now this only includes a very
crude check for instructions that will be executed by the master thread
by comparing a thread-id function with a constant zero.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D101578
Hoisting and sinking instructions out of conditional blocks enables
additional vectorization by:
1. Executing memory accesses unconditionally.
2. Reducing the number of instructions that need predication.
After disabling early hoisting / sinking, we miss out on a few
vectorization opportunities. One of those is causing a ~10% performance
regression in one of the Geekbench benchmarks on AArch64.
This patch tires to recover the regression by running hoisting/sinking
as part of a SimplifyCFG run after LoopRotate and before LoopVectorize.
Note that in the legacy pass-manager, we run LoopRotate just before
vectorization again and there's no SimplifyCFG run in between, so the
sinking/hoisting may impact the later run on LoopRotate. But the impact
should be limited and the benefit of hosting/sinking at this stage
should outweigh the risk of not rotating.
Compile-time impact looks slightly positive for most cases.
http://llvm-compile-time-tracker.com/compare.php?from=2ea7fb7b1c045a7d60fcccf3df3ebb26aa3699e5&to=e58b4a763c691da651f25996aad619cb3d946faf&stat=instructions
NewPM-O3: geomean -0.19%
NewPM-ReleaseThinLTO: geoman -0.54%
NewPM-ReleaseLTO-g: geomean -0.03%
With a few benchmarks seeing a notable increase, but also some
improvements.
Alternative to D101290.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D101468
The ModulePassManager should already have taken care of all analysis
invalidation. Without this change, upcoming changes will cause more
invalidation than necessary.
Reviewed By: mtrofin
Differential Revision: https://reviews.llvm.org/D101320
Pseudo probe distribution factor is used to scale down profile samples to avoid misleading the counts inference due to the usage of "maximum" in `getBlockWeight`. For callsites, the scaling down can come from code duplication prior to the sample profile loader (prelink or postlink), or due to the indirect call promotion in sample loader inliner. This patch fixes an issue in sample loader ICP where the leftover indirect callsite scaling down causes the loss of non-promoted call target samples unexpectedly. While the scaling down is to favor BFI/BPI with accurate an callsite count, it doesn't fit in the current distribution factor that represents code duplication changes. Ideally, we would need two factors, one is for code duplication, the other is for ICP. However this seems over complicated. I'm going to trade one usage (callsite counts) for the other (call target counts).
Seeing perf win on one benchmark (mcf) of SPEC2017 with others unchanged.
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D100993
Both the alias and aliasee linkage are important.
PR27866 provides some background.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D99629
This revision simplifies Clang codegen for parallel regions in OpenMP GPU target offloading and corresponding changes in libomptarget: SPMD/non-SPMD parallel calls are unified under a single `kmpc_parallel_51` runtime entry point for parallel regions (which will be commonized between target, host-side parallel regions), data sharing is internalized to the runtime. Tests have been auto-generated using `update_cc_test_checks.py`. Also, the revision contains changes to OpenMPOpt for remark creation on target offloading regions.
Reviewed By: jdoerfert, Meinersbur
Differential Revision: https://reviews.llvm.org/D95976
Discovered during attributor testing comparing stats with
and without the attributor. Willreturn should not be inferred
for nonexact definitions.
Differential Revision: https://reviews.llvm.org/D100988
FunctionAnalysisManagerCGSCCProxy should not be preserved if any of its
keys may be invalid. Since we are not removing/adding functions in
FuncAttrs, it's fine to preserve it.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D100893
Summary:
This patch registers OpenMPOpt as a Module pass in addition to a CGSCC
pass. This is so certain optimzations that are sensitive to intact
call-sites can happen before inlining. The old `openmpopt` pass name is
changed to `openmp-opt-cgscc` and `openmp-opt` calls the Module pass.
The current module pass only runs a single check but will be expanded in
the future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D99202
Previously we would use the type of the pointee to determine what to
cast the result of constant folding a load. To aid with opaque pointer
types, we should explicitly pass the type of the load rather than
looking at pointee types.
ConstantFoldLoadThroughBitcast() converts the const prop'd value to the
proper load type (e.g. [1 x i32] -> i32). Instead of calling this in
every intermediate step like bitcasts, we only call this when we
actually see the global initializer value.
In some existing uses of this API, we don't know the exact type we're
loading from immediately (e.g. first we visit a bitcast, then we visit
the load using the bitcast). In those cases we have to manually call
ConstantFoldLoadThroughBitcast() when simplifying the load to make sure
that we cast to the proper type.
Reviewed By: dblaikie
Differential Revision: https://reviews.llvm.org/D100718
Joseph Huber