Commit Graph

796 Commits

Author SHA1 Message Date
Andrew Trick
6c99017588 Minor comment cleanup and code formatting.
Extracted from the functional changes for clarity.
2025-02-25 23:08:53 -08:00
Andrew Trick
75e84f1e97 Rename InteriorLiveness methods for clarity.
It was too easy to confuse the "inner" vs. "outer" value.

Use less ambigous names: visitAllUses vs visitInnerScopeUses
2025-02-25 23:08:53 -08:00
Erik Eckstein
57a236e671 InitializeStaticGlobals: support statically initializing globals which are or contain inline arrays
For example:

```
struct S {
  static let slab: Slab = [1, 2, 3, 4]
}
```

rdar://143005996
2025-02-12 22:37:49 +01:00
eeckstein
dff88f968b Merge pull request #79317 from eeckstein/remove-alloc-vector
Remove the experimental FixedArray
2025-02-12 17:03:06 +01:00
eeckstein
b6639e1782 Merge pull request #79314 from eeckstein/fix-simplify-keypath
SimplifyKeyPath: insert the correct destroy operation for an operand of a removed keypath
2025-02-12 15:18:39 +01:00
Erik Eckstein
e0b4f71af6 SIL: remove the alloc_vector instruction
It's not needed anymore, because the "FixedArray" experimental feature is replaced by inline-arrays.
2025-02-12 10:51:14 +01:00
Erik Eckstein
6407f9a0bd remove the allocVector builtin
It's not needed anymore, because the "FixedArray" experimental feature is replaced by inline-arrays.
2025-02-12 10:51:14 +01:00
Erik Eckstein
5b93eb31bf Optimizer: remove the AllocVectorLowering pass
It's not needed anymore, because the "FixedArray" experimental feature is replaced by inline-arrays.
2025-02-12 10:51:14 +01:00
Erik Eckstein
3f95ce9645 Optimizer: fix spelling of the Simplifiable protocol
Rename `Simplifyable` -> `Simplifiable`

NFC
2025-02-12 09:01:11 +01:00
Erik Eckstein
2962474b58 SimplifyKeyPath: insert the correct destroy operation for an operand of a removed keypath
If the operand is an address, we need to use `destroy_addr` and not `destroy_value`.

Fixes a compiler crash.
rdar://144662171
2025-02-12 08:35:23 +01:00
eeckstein
aa0833b0b3 Merge pull request #79110 from eeckstein/verify_mark_dependence
SIL: Fix memory behavior of mark_dependence
2025-02-11 06:39:43 +01:00
Andrew Trick
c2842e8e19 LifetimeDependenceInsertion: remove a bailout on ~Copyable
Needed to diagnose MutableSpan and OutputSpan.

For now, simply remove the bailout and TODO. The next change will introduce more
logic to force a diagnostic error in rare cases that can't be handled completely.

Fixes rdar://143584461 (Extended exclusive borrow issues with
MutableSpan and OutputSpan)
2025-02-10 09:13:27 -08:00
Andrew Trick
fa3ebb5c1e LifetimeDependence.Scope add global variable support.
This fixes functions that return @lifetime(immortal).
2025-02-10 09:11:22 -08:00
Andrew Trick
39bae0bb58 LifetimeDependenceDiagnostics: immortal dependence on global 'let' 2025-02-10 09:11:22 -08:00
Andrew Trick
c3de120ca5 LifetimeDependence: simplify and fix multiple bugs.
Functional changes:

Improved modeling of dependence on local variable scopes.

For nested modify->read accesses, only extend the read accesses.

Avoid making a read access dependent on an inout argument.
The following needs to be an error to prevent span storage from being modified:

  @lifetime(owner)
  foo(owner: inout Owner) -> Span {
    owner.span
  }

Improve usability of borrowing trivial values (UnsafePointer). Allow:

  let span = Span(buffer.baseAddress)

Ignore access scopes for trivial values.

Structural changes:

Delete the LifetimeDependenceUseDefWalker.

Encapsulate all logic for variable introducers within the LifetimeDependenceInsertion pass. Once mark_dependence instructions are inserted, no subsequent pass needs to think about the "root" of a dependence.

Fixes: rdar://142451725 (Escape analysis fails with mutations)
2025-02-10 09:11:22 -08:00
Andrew Trick
eae2f06dc7 Extend and improve AccessBase.findSingleInitializer
Return an Initializer rather than a tuple. This makes it easier and more robust
for clients to track dependencies on initialized memory.
2025-02-10 09:11:22 -08:00
Andrew Trick
8b39b6d16a LocalVariableUtils: record mark_dependence bad address operands
Avoid bailing out when something depends on the local variable.
2025-02-10 09:11:22 -08:00
Erik Eckstein
a88cb49ee8 SIL: define a memory-read effect on the mark_dependence base value
If the base value of a mark_dependence is an address, that memory location must be initialized at the mark_dependence.
This requires that the mark_dependence is considered to read from the base address.
2025-02-10 17:57:47 +01:00
Erik Eckstein
33761454b7 MandatoryPerformanceOptimizations: use eliminateRedundantLoads instead of optimizeMemoryAccesses to optimize redundant loads 2025-02-07 11:30:35 +01:00
Erik Eckstein
ce73deebc4 RedundantLoadElimination: add a "mandatory" redundant load elimination pass
And make `eliminateRedundantLoads` callable from other optimizations
2025-02-07 11:30:35 +01:00
Erik Eckstein
2af24703ab RedundantLoadElimination: insert mark_dependence instructions for a removed load
If the memory location depends on something, insert a dependency for the loaded value:
```
  %2 = mark_dependence %1 on %0
  %3 = load %2
```
->
```
  %2 = mark_dependence %1 on %0 // not needed anymore, can be removed eventually
  %3 = load %2
  %4 = mark_dependence %3 on %0
  // replace %3 with %4
```
2025-02-07 11:30:35 +01:00
Erik Eckstein
5acccf1802 RedundantLoadElimination: ignore memory effects of begin_access
Memory effects of begin_access are only defined to prevent the optimizer moving loads and stores across a begin_access.
But those memory effects are not relevant for RedundantLoadElimination
2025-02-07 11:30:35 +01:00
Erik Eckstein
ad2462f7f2 RedundantLoadElimination: optimize a load from a memory location which was written by copy_addr
For example:
```
  %1 = alloc_stack $B
  copy_addr %0 to [init] %1
  %3 = load [take] %1
  dealloc_stack %1
```
->
```
  %3 = load [copy] %0
```
2025-02-07 11:30:34 +01:00
Andrew Trick
e705a6d7c3 Temporarily introduce AnyInteriorPointer operand ownership.
This is necessary to fix a recent OSSA bug that breaks common occurrences on
mark_dependence [nonescaping]. Rather than reverting that change above, we make
forward progress toward implicit borrows scopes, as was the original intention.

In the near future, all InteriorPointer instructions will create an implicit
borrow scope. This means we have the option of not emitting extraneous
begin/end_borrow instructions around intructions like ref_element_addr,
open_existential, and project_box. After that, we can also migrate
GuaranteedForwarding instructions like tuple_extract and struct_extract.
2025-02-05 16:23:02 -08:00
Erik Eckstein
9b2dfbb741 DestroyHoisting: visit begin_borrows which don't have an end_borrow in dead-end blocks.
This makes sure that incomplete lifetimes are handled correctly.

Fixes an ownership violation.
rdar://144065326
2025-02-04 12:59:40 +01:00
Arnold Schwaighofer
c95c45201e Merge pull request #78996 from aschwaighofer/checked_cast_br_nightmares
SIL: Update formal source type in SimplifyRefCasts when possible
2025-02-02 08:27:01 -08:00
Arnold Schwaighofer
0fe3cf0096 SIL: Update formal source type in SimplifyRefCasts when possible
We use the formal source type do decide whether a checked_cast_br is
known to succeed/fail. If we don't update it we loose that optimization

That is:

```
  checked_cast_br AnyObject in %2 : X to X, bb1, bb2
```

Will not be simplified even though the operand and the destintation type
matches.
2025-01-29 09:07:21 -08:00
Kuba (Brecka) Mracek
7c685ba36b Merge pull request #78915 from kubamracek/embedded-string-switch-with-cache
[embedded] Support _findStringSwitchCaseWithCache in Embedded Swift
2025-01-28 15:56:22 -08:00
Kuba Mracek
e1d22b5076 [embedded] Avoid changing lookupStdlibFunction, load findStringSwitchCaseWithCache in MPO instead 2025-01-28 10:03:14 -08:00
Erik Eckstein
1f534e1c6f DeinitDevirtualizer: bail out for C++ move-only types.
We cannot de-virtualize C++ destructor calls of C++ move-only types because we cannot get its destructor in SIL.
Fixes a miscompile.
2025-01-27 09:59:48 +01:00
Kuba Mracek
92b0c5f5d0 [embedded] Support _findStringSwitchCaseWithCache in Embedded Swift 2025-01-25 09:09:47 -08:00
Erik Eckstein
3ec5d7de24 SIL: replace the is_escaping_closure instruction with destroy_not_escaped_closure
The problem with `is_escaping_closure` was that it didn't consume its operand and therefore reference count checks were unreliable.
For example, copy-propagation could break it.
As this instruction was always used together with an immediately following `destroy_value` of the closure, it makes sense to combine both into a `destroy_not_escaped_closure`.
It
1. checks the reference count and returns true if it is 1
2. consumes and destroys the operand
2025-01-24 19:23:27 +01:00
Erik Eckstein
09f6f4f9b6 AutodiffClosureSpecialization: workaround a problem with OSSA.
Disable the closure specializer optimization for closures in OSSA until the underlying problem is fixed.
https://github.com/swiftlang/swift/issues/78847

Part of rdar://140229560
2025-01-23 16:10:32 +01:00
eeckstein
2af03dd177 Merge pull request #78768 from eeckstein/fix-borrowed-from-updater
SIL: when updating borrowed-from instructions, remove the old enclosing values first
2025-01-22 18:40:26 +01:00
Andrew Trick
5066dd9f90 Merge pull request #78716 from atrick/fix-lifedep-out
Fix LifetimeDependenceDiagnostics for @out dependencies
2025-01-22 00:17:50 -08:00
Erik Eckstein
5dc67fab3f SIL Verifier: check that the enclosing values of a borrowed-from instructions are valid 2025-01-21 08:21:26 +01:00
Erik Eckstein
012aa6227d SIL: when updating borrowed-from instructions, remove the old enclosing values first
When an optimization updates borrowed-from instruction it might be necessary to remove the old enclosing values from the borrowed-from instructions.
An optimization might transform the SIL in a way that an existing enclosing value is not valid anymore.
Fixes a compiler crash:
rdar://142991910
2025-01-21 08:21:26 +01:00
Erik Eckstein
97db5d8a80 AliasAnalysis: handle global let-variables in ImmutableScope
Global let-variables are immutable, except in functions which initialize them.
This brings back handling of global let-variables in alias analysis, which was removed in the previous commit.
2025-01-20 08:50:56 +01:00
Erik Eckstein
a312732f84 AliasAnalysis: consider memory effects of a consume/destroy of a class on it's let-fields
Although a let-field can never be mutated, a release or consume of the class must be considered as writing to such a field.

This change removes the special handling of let-fields in two places, where they don't belong.
Class fields are handled by ImmutableScope anyway.
Handling of global let-variable is temporarily removed by this commit.

Fixes a miscompile.
rdar://142996449
2025-01-20 08:50:56 +01:00
Andrew Trick
cc07b4ba11 Fix LifetimeDependenceDiagnostics for @out dependencies.
Record a forwarding mark_dependence as a local access. This is necessary because
we now emit a mark_dependence for @out arguments, which will be the starting
point for diagnostics:

  %out = alloc_stack
  apply %f(%owned, %out) : $(Owner) -> @lifetime(borrow 0) @out View
  %unused = mark_dependence [unresolved] %out on %owner
  %dependentValue = load %out

This mark_dependence has no uses. Instead, it simply records the dependency of
the in-memory value on the owner. Consequently, simply walking the uses of
LifetimeDependence.dependentValue does fails to diagnose any escapes. Instead,
if the dependentValue is an address-type mark_dependence, treat it as a local
access to the address that it forwards. Then we find any reachable uses of that
local variable as a potential escape.

Fixes rdar://143040479
(Borrow diagnostics not triggered for @out return values)
2025-01-17 11:19:04 -08:00
Andrew Trick
afda6453a5 Improve LocalVariableAccessInfo: recognize more full-assignments.
Ignore marker instructions for the purpose of determining whether a store is a
full assignment:

  %a = alloc_stack
  %m = moveonlywrapper_to_copyable_addr %a
  store %0 to [init] %m // <=== full assignemt
2025-01-17 11:19:04 -08:00
Andrew Trick
c839e265df [NFC] Update an incomplete comment. 2025-01-15 16:53:03 -08:00
Andrew Trick
c8bec28e6a LifetimeDependence.Scope: recognize @in_guaranteed dependencies.
Unlike @in, treat @in_guaranteed like a caller-side dependence
scope because there is not need to look for the end of the lifetime in the
current function.

Completely fixes rdar://142847915 (Crash during lifetime checking
while building new swift standard library `Span`-related features)
2025-01-14 15:38:54 -08:00
Andrew Trick
a8caedfbec LifetimeDependenceInsertion: fix dependence on store_borrow.
For a lifetime dependent call that depends on a temporary store_borrow, the
generated mark_dependendence should be on the stored value, not the stack
location.

  %temp = alloc_stack $AnyObject
  %sb = store_borrow %arg to %temp
  apply %10(%out, %sb)
  mark_dependence [unresolved] %out on %arg
  end_borrow %sb

Fixes rdar://142847915 (Crash during lifetime checking while
building new swift standard library `Span`-related features)
2025-01-14 00:11:17 -08:00
Andrew Trick
6849b66bb3 LifetimeDependenceDiagnostics; remove a bootstrapping hack.
This temporary hack was preventing diagnostics from kicking in. This could even
result in invalid SIL after the diagnostic failed to trigger.
2025-01-13 08:30:41 -08:00
Andrew Trick
0326c98f72 LifetimeDependenceScopeFixup: handle unsafe addressors
Handle unsafe addressors just like projections of stored properties rather than
as separate variable accesses.
2025-01-11 15:40:22 -08:00
Andrew Trick
b0f2ca03a7 AccessUtils: allow mark_deps to be tracked by the EnclosingScope.
This encourages AccessPathWalker clients to handle enclosing mark_deps. In
some cases, it is necessary. The accessBaseWithScopes API now provides both
nested begin_access and mark_dependence.
2025-01-11 15:40:22 -08:00
eeckstein
8ec8a1229a Merge pull request #78543 from eeckstein/fix-copy-to-borrow-optimization
Fix two problems in the copy-to-borrow optimization
2025-01-10 19:44:00 +01:00
Erik Eckstein
e215ae990b CopyToBorrowOptimization: correctly handle unchecked_enum_data
A `unchecked_enum_data` which extracts a trivial payload out of a non-trivial enum ends the lifetime of its owned operand.

Fixes an ownership verification error
rdar://142644731
2025-01-10 09:08:31 +01:00
Erik Eckstein
a4c675ec88 CopyToBorrowOptimization: fix insertion order of end_borrows
In liferange exit blocks the end_borrows were inserted in the wrong order.

Fixes an ownership verification error
rdar://142632741
2025-01-10 08:56:42 +01:00