* Reimplement most of the logic in Swift as an Instruction simplification and remove the old code from SILCombine
* support more cases of existential archetype replacements:
For example:
```
%0 = alloc_stack $any P
%1 = init_existential_addr %0, $T
use %1
```
is transformed to
```
%0 = alloc_stack $T
use %0
```
Also, if the alloc_stack is already an opened existential and the concrete type is known,
replace it as well:
```
%0 = metatype $@thick T.Type
%1 = init_existential_metatype %0, $@thick any P.Type
%2 = open_existential_metatype %1 : $@thick any P.Type to $@thick (@opened("X", P) Self).Type
...
%3 = alloc_stack $@opened("X", any P) Self
use %3
```
is transformed to
```
...
%3 = alloc_stack $T
use %3
```
If an apply uses an existential archetype (`@opened("...")`) and the concrete type is known, replace the existential archetype with the concrete type
1. in the apply's substitution map
2. in the arguments, e.g. by inserting address casts
For example:
```
%5 = apply %1<@opend("...")>(%2) : <τ_0_0> (τ_0_0) -> ()
```
->
```
%4 = unchecked_addr_cast %2 to $*ConcreteType
%5 = apply %1<ConcreteType>(%4) : <τ_0_0> (τ_0_0) -> ()
```
Replace `unconditional_checked_cast` to an existential metatype with an `init_existential_metatype`, it the source is a conforming type.
Note that init_existential_metatype is better than unconditional_checked_cast because it does not need to do any runtime casting.
So far a `SILCombineSimplifiable` could only replace a SILCombine visit implementation.
With the `SWIFT_SILCOMBINE_PASS_WITH_LEGACY` (to be used in Passes.def) it's possible to keep an existing C++ implementation and on top of that add a Swift Simplification pass.
Which consists of
* removing redundant `address_to_pointer`-`pointer_to_address` pairs
* optimize `index_raw_pointer` of a manually computed stride to `index_addr`
* remove or increase the alignment based on a "assumeAlignment" builtin
This is a big code cleanup but also has some functional differences for the `address_to_pointer`-`pointer_to_address` pair removal:
* It's not done if the resulting SIL would result in a (detectable) use-after-dealloc_stack memory lifetime failure.
* It's not done if `copy_value`s must be inserted or borrow-scopes must be extended to comply with ownership rules (this was the task of the OwnershipRAUWHelper).
Inserting copies is bad anyway.
Extending borrow-scopes would only be required if the original lifetime of the pointer extends a borrow scope - which shouldn't happen in save code. Therefore this is a very rare case which is not worth handling.
Canonicalize a `fix_lifetime` from an address to a `load` + `fix_lifetime`:
```
%1 = alloc_stack $T
...
fix_lifetime %1
```
->
```
%1 = alloc_stack $T
...
%2 = load %1
fix_lifetime %2
```
This transformation is done for `alloc_stack` and `store_borrow` (which always has an `alloc_stack` operand).
The benefit of this transformation is that it enables other optimizations, like mem2reg.
This peephole optimization was already done in SILCombine, but it didn't handle store_borrow.
A good opportunity to make an instruction simplification out of it.
This is part of fixing regressions when enabling OSSA modules:
rdar://140229560
* Remove dead `load_borrow` instructions (replaces the old peephole optimization in SILCombine)
* If the `load_borrow` is followed by a `copy_value`, combine both into a `load [copy]`
It hoists `destroy_value` instructions without shrinking an object's lifetime.
This is done if it can be proved that another copy of a value (either in an SSA value or in memory) keeps the referenced object(s) alive until the original position of the `destroy_value`.
```
%1 = copy_value %0
...
last_use_of %0
// other instructions
destroy_value %0 // %1 is still alive here
```
->
```
%1 = copy_value %0
...
last_use_of %0
destroy_value %0
// other instructions
```
The benefit of this optimization is that it can enable copy-propagation by moving destroys above deinit barries and access scopes.
It removes a `copy_value` where the source is a guaranteed value, if possible:
```
%1 = copy_value %0 // %0 = a guaranteed value
// uses of %1
destroy_value %1 // borrow scope of %0 is still valid here
```
->
```
// uses of %0
```
This optimization is very similar to the LoadCopyToBorrow optimization.
Therefore I merged both optimizations into a single file and renamed it to "CopyToBorrowOptimization".
Propagating array element values is done by load-simplification and redundant-load-elimination.
So ArrayElementPropagation is not needed anymore.
ArrayElementPropagation also replaced `Array.append(contentsOf:)` with individual `Array.append` calls.
This optimization is removed, because the benefit is questionably, anyway.
In most cases it resulted in a code size increase.
The optimization replaces a `load [copy]` with a `load_borrow` if possible.
```
%1 = load [copy] %0
// no writes to %0
destroy_value %1
```
->
```
%1 = load_borrow %0
// no writes to %0
end_borrow %1
```
The new implementation uses alias-analysis (instead of a simple def-use walk), which is much more powerful.
rdar://115315849
MandatoryPerformanceOptimizations already did most of the vtable specialization work.
So it makes sense to remove the VTableSpecializerPass completely and do everything in MandatoryPerformanceOptimizations.
The main changes are:
*) Rewrite everything in swift. So far, parts of memory-behavior analysis were already implemented in swift. Now everything is done in swift and lives in `AliasAnalysis.swift`. This is a big code simplification.
*) Support many more instructions in the memory-behavior analysis - especially OSSA instructions, like `begin_borrow`, `end_borrow`, `store_borrow`, `load_borrow`. The computation of end_borrow effects is now much more precise. Also, partial_apply is now handled more precisely.
*) Simplify and reduce type-based alias analysis (TBAA). The complexity of the old TBAA comes from old days where the language and SIL didn't have strict aliasing and exclusivity rules (e.g. for inout arguments). Now TBAA is only needed for code using unsafe pointers. The new TBAA handles this - and not more. Note that TBAA for classes is already done in `AccessBase.isDistinct`.
*) Handle aliasing in `begin_access [modify]` scopes. We already supported truly immutable scopes like `begin_access [read]` or `ref_element_addr [immutable]`. For `begin_access [modify]` we know that there are no other reads or writes to the access-address within the scope.
*) Don't cache memory-behavior results. It turned out that the hit-miss rate was pretty bad (~ 1:7). The overhead of the cache lookup took as long as recomputing the memory behavior.
Changes in this CR add part of the, Swift based, Autodiff specific
closure specialization optimization pass. The pass does not modify any
code nor does it even exist in any of the optimization pipelines. The
rationale for pushing this partially complete optimization pass upstream
is to keep up with the breaking changes in the underlying Swift based
compiler infrastructure.
The reason why I am doing this is that I am going to be adding support for
preconcurrency imports to TransferNonSendable. That implies that we can have
preconcurrency import suppression in the SIL pipeline and thus that emitting the
diagnostic in Sema is too early.
To do this, I introduced a new module pass called
DiagnoseUnnecessaryPreconcurrencyImports that runs after the SILFunction pass
TransferNonSendable. The reason why I use a module pass is to ensure that
TransferNonSendable has run on all functions before we attempt to emit these
diagnostics. Then in that pass, we iterate over all of the modules functions and
construct a uniqued array of SourceFiles for these functions. Then we iterate
over the uniqued SourceFiles and use the already constructed Sema machinery to
emit the diagnostic using the source files.
rdar://126928265
Compute, update and handle borrowed-from instruction in various utilities and passes.
Also, used borrowed-from to simplify `gatherBorrowIntroducers` and `gatherEnclosingValues`.
Replace those utilities by `Value.getBorrowIntroducers` and `Value.getEnclosingValues`, which return a lazily computed Sequence of borrowed/enclosing values.
* Let the customBits and lastInitializedBitfieldID share a single uint64_t. This increases the number of available bits in SILNode and Operand from 8 to 20. Also, it simplifies the Operand class because no PointerIntPairs are used anymore to store the operand pointer fields.
* Instead make the "deleted" flag a separate bool field in SILNode (instead of encoding it with the sign of lastInitializedBitfieldID). Another simplification
* Enable important invariant checks also in release builds by using `require` instead of `assert`. Not catching such errors in release builds would be a disaster.
* Let the Swift optimization passes use all the available bits and not only a fixed amount of 8 (SILNode) and 16 (SILBasicBlock).
We often look at the SIL output of -sil-print-function and may want to debug a specific pass
after looking at the output.
-sil-break-before-pass-count=<pass_number> will allow to automatically break in the debugger
after <pass_count> of passes are run.
Example:
From -sil-print-function dump:
"SIL function after #6680, stage MidLevel,Function, pass 38: RedundantLoadElimination"
-Xllvm -sil-break-before-pass-count=6680 will break before running this pass in the debugger
InstructionRange uses 3 sets. Some algorithms need two active ranges at the top-level. Additionally, utilities often have few levels of nesting that each require a block set.
I am doing this since region based isolation hit the same issue that the move
checker did. So it makes sense to refactor the functionality into its own pass
and move it into a helper pass that runs before both.
It is very conservative and only stubifies functions that the specialization
passes explicitly mark as this being ok to be done to.
For years, optimizer engineers have been hitting a common bug caused by passes
assuming all SILValues have a parent function only to be surprised by SILUndef.
Generally we see SILUndef not that often so we see this come up later in
testing. This patch eliminates that problem by making SILUndef uniqued at the
function level instead of the module level. This ensures that it makes sense for
SILUndef to have a parent function, eliminating this possibility since we can
define an API to get its parent function.
rdar://123484595
