I am purposely doing this in SILGen rather than at the type system level to
avoid having to have to add a bunch of boilerplate to the type system. Instead
of doing that, I am in SILGen checking for the isNoImplicitCopy bit on the
ParamDecl when we emit arguments. At that point, I set on the specific
SILArgument being emitted the bit that it is no implicit copy. In terms of
printing at the SIL level, I just printed it in front of the function argument
type like @owned, e.x.:
func myFunc(_ x: @_noImplicitCopy T) -> T {
...
}
becomes:
bb0(%0 : @noImplicitCopy @owned $T):
Some notes:
* Just to be explicit, I am making it so that no implicit copy parameters by
default are always passed at +1. The reason why I think this makes sense is
that this is the natural way of working with a move only value.
* As always, one can not write no implicit copy the attribute without passing
the flag -enable-experimental-move-only so this is NFC.
rdar://83957088
This is the initial version of a buildable SIL definition in libswift.
It defines an initial set of SIL classes, like Function, BasicBlock, Instruction, Argument, and a few instruction classes.
The interface between C++ and SIL is a bridging layer, implemented in C.
It contains all the required bridging data structures used to access various SIL data structures.
In OSSA, we do not allow for address phis, but in certain cases the logic of
LoopRotate really wants them. To work around this issue, I added some code in
this PR to loop rotate that as a post-pass fixes up any address phis by
inserting address <-> raw pointer adapters and changing the address phi to
instead be of raw pointer type.
This removes the ambiguity when casting from a SingleValueInstruction to SILNode, which makes the code simpler. E.g. the "isRepresentativeSILNode" logic is not needed anymore.
Also, it reduces the size of the most used instruction class - SingleValueInstruction - by one pointer.
Conceptually, SILInstruction is still a SILNode. But implementation-wise SILNode is not a base class of SILInstruction anymore.
Only the two sub-classes of SILInstruction - SingleValueInstruction and NonSingleValueInstruction - inherit from SILNode. SingleValueInstruction's SILNode is embedded into a ValueBase and its relative offset in the class is the same as in NonSingleValueInstruction (see SILNodeOffsetChecker).
This makes it possible to cast from a SILInstruction to a SILNode without knowing which SILInstruction sub-class it is.
Casting to SILNode cannot be done implicitly, but only with an LLVM `cast` or with SILInstruction::asSILNode(). But this is a rare case anyway.
This removes the ambiguity when casting from a SingleValueInstruction to SILNode, which makes the code simpler. E.g. the "isRepresentativeSILNode" logic is not needed anymore.
Also, it reduces the size of the most used instruction class - SingleValueInstruction - by one pointer.
Conceptually, SILInstruction is still a SILNode. But implementation-wise SILNode is not a base class of SILInstruction anymore.
Only the two sub-classes of SILInstruction - SingleValueInstruction and NonSingleValueInstruction - inherit from SILNode. SingleValueInstruction's SILNode is embedded into a ValueBase and its relative offset in the class is the same as in NonSingleValueInstruction (see SILNodeOffsetChecker).
This makes it possible to cast from a SILInstruction to a SILNode without knowing which SILInstruction sub-class it is.
Casting to SILNode cannot be done implicitly, but only with an LLVM `cast` or with SILInstruction::asSILNode(). But this is a rare case anyway.
This is letting me refactor the implementation of every place that I work with
branches as "ownership phi operands" to instead work with
OwnershipPhiOperand. In a future commit, I am going to use this to add support
for optimizing structs and tuples that have multiple owned incoming values.
<rdar://problem/63950481>
If only a single field of a struct phi-argument is used, replace the argument by the field value.
br bb(%str)
bb(%phi):
%f = struct_extract %phi, #Field // the only use of %phi
use %f
is replaced with
%f = struct_extract %str, #Field
br bb(%f)
bb(%phi):
use %phi
This also works if the phi-argument is in a def-use cycle.
The new PhiExpansionPass is in the same file as the RedundantPhiEliminationPass. Therefore I renamed the source file to PhiArgumentOptimizations.cpp
Operands are generally better to return than values since the operand also
enables you to get to the terminator instruction as well. Since so much code in
the compiler already uses the getIncomingPhiValue methods, I reimplemented them
on top of the operand methods.
VS2015 had an issue with the deletion of an operator. Since VS2017 is
the minimum version that LLVM uses, we can assume that VS2017+ is in use
(_MSC_VER >= 1910). Clean up the now defunct workaround.
The client of this interface naturally expects to get back the
incoming phi value. Ignoring dominance and SIL ownership, the incoming
phi value and the block argument should be substitutable.
This method was actually returning the incoming operand for
checked_cast and switch_enum terminators, which is deeply misleading
and has been the source of bugs.
If the client wants to peek though casts, and enums, it should do so
explicitly. getSingleTerminatorOperand[s]() will do just that.
introduce a common superclass, SILNode.
This is in preparation for allowing instructions to have multiple
results. It is also a somewhat more elegant representation for
instructions that have zero results. Instructions that are known
to have exactly one result inherit from a class, SingleValueInstruction,
that subclasses both ValueBase and SILInstruction. Some care must be
taken when working with SILNode pointers and testing for equality;
please see the comment on SILNode for more information.
A number of SIL passes needed to be updated in order to handle this
new distinction between SIL values and SIL instructions.
Note that the SIL parser is now stricter about not trying to assign
a result value from an instruction (like 'return' or 'strong_retain')
that does not produce any.
Separate formal lowered types from SIL types.
The SIL type of an argument will depend on the SIL module's conventions.
The module conventions are determined by the SIL stage and LangOpts.
Almost NFC, but specialized manglings are broken incidentally as a result of
fixes to the way passes handle book-keeping of aruments. The mangler is fixed in
the subsequent commit.
Otherwise, NFC is intended, but quite possible do to rewriting the logic in many
places.
This in the case of insertFunctionArgument requires a ValueOwnershipKind to be
specified since we use that for transformations of function argument lists that
are only correct after the transformation is complete. This only occurs in
FunctionSignatureOptimizations.
On the other hand, createFunctionArgument is only used to construct completely
new argument lists, so we can instead just rely on the function we are in rather
than require the user to pass it in.
rdar://29791263
This means using a struct so we can put methods on the struct and using an
anonymous enum to create namespaced values. Specifically:
struct SILArgumentConvention {
enum : uint8_t {
Indirect_In,
Indirect_In_Guaranteed,
Indirect_Inout,
Indirect_InoutAliasable,
Indirect_Out,
Direct_Owned,
Direct_Unowned,
Direct_Deallocating,
Direct_Guaranteed,
} Value;
SILArgumentConvention(decltype(Value) NewValue)
: Value(NewValue) {}
operator decltype(Value)() const {
return Value;
}
ParameterConvention getParameterConvention() const {
switch (Value) {
...
}
}
bool isIndirectConvention() const {
...
}
};
This allows for:
1. Avoiding abstraction leakage via the enum type. If someone wants to use
decltype as well, I think that is enough work that the leakage is acceptable.
2. Still refer to enum cases like we are working with an enum class
(e.g. SILArgumentConvention::Direct_Owned).
3. Avoid using the anonymous type in function arguments due to an implicit
conversion.
4. And most importantly... *drum roll* add methods to our enums!
We preserve the current behavior of assuming Any ownership always and use
default arguments to hide this change most of the time. There are asserts now in
the SILBasicBlock::{create,replace,insert}{PHI,Function}Argument to ensure that
the people can only create SILFunctionArguments in entry blocks and
SILPHIArguments in non-entry blocks. This will ensure that the code in tree
maintains the API distinction even if we are not using the full distinction in
between the two.
Once the verifier is finished being upstreamed, I am going to audit the
createPHIArgument cases for the proper ownership. This is b/c I will be able to
use the verifier to properly debug the code. At that point, I will also start
serializing/printing/parsing the ownershipkind of SILPHIArguments, but lets take
things one step at a time and move incrementally.
In the process, I also discovered a CSE bug. I am not sure how it ever worked.
Basically we replace an argument with a new argument type but return the uses of
the old argument to refer to the old argument instead of a new argument.
rdar://29671437
For a long time, we have:
1. Created methods on SILArgument that only work on either function arguments or
block arguments.
2. Created code paths in the compiler that only allow for "function"
SILArguments or "block" SILArguments.
This commit refactors SILArgument into two subclasses, SILPHIArgument and
SILFunctionArgument, separates the function and block APIs onto the subclasses
(leaving the common APIs on SILArgument). It also goes through and changes all
places in the compiler that conditionalize on one of the forms of SILArgument to
just use the relevant subclass. This is made easier by the relevant APIs not
being on SILArgument anymore. If you take a quick look through you will see that
the API now expresses a lot more of its intention.
The reason why I am performing this refactoring now is that SILFunctionArguments
have a ValueOwnershipKind defined by the given function's signature. On the
other hand, SILBlockArguments have a stored ValueOwnershipKind. Rather than
store ValueOwnershipKind in both instances and in the function case have a dead
variable, I decided to just bite the bullet and fix this.
rdar://29671437
Before this commit all code relating to handling arguments in SILBasicBlock had
somewhere in the name BB. This is redundant given that the class's name is
already SILBasicBlock. This commit drops those names.
Some examples:
getBBArg() => getArgument()
BBArgList => ArgumentList
bbarg_begin() => args_begin()
Similarly to how we've always handled parameter types, we
now recursively expand tuples in result types and separately
determine a result convention for each result.
The most important code-generation change here is that
indirect results are now returned separately from each
other and from any direct results. It is generally far
better, when receiving an indirect result, to receive it
as an independent result; the caller is much more likely
to be able to directly receive the result in the address
they want to initialize, rather than having to receive it
in temporary memory and then copy parts of it into the
target.
The most important conceptual change here that clients and
producers of SIL must be aware of is the new distinction
between a SILFunctionType's *parameters* and its *argument
list*. The former is just the formal parameters, derived
purely from the parameter types of the original function;
indirect results are no longer in this list. The latter
includes the indirect result arguments; as always, all
the indirect results strictly precede the parameters.
Apply instructions and entry block arguments follow the
argument list, not the parameter list.
A relatively minor change is that there can now be multiple
direct results, each with its own result convention.
This is a minor change because I've chosen to leave
return instructions as taking a single operand and
apply instructions as producing a single result; when
the type describes multiple results, they are implicitly
bound up in a tuple. It might make sense to split these
up and allow e.g. return instructions to take a list
of operands; however, it's not clear what to do on the
caller side, and this would be a major change that can
be separated out from this already over-large patch.
Unsurprisingly, the most invasive changes here are in
SILGen; this requires substantial reworking of both call
emission and reabstraction. It also proved important
to switch several SILGen operations over to work with
RValue instead of ManagedValue, since otherwise they
would be forced to spuriously "implode" buffers.
As there are no instructions left which produce multiple result values, this is a NFC regarding the generated SIL and generated code.
Although this commit is large, most changes are straightforward adoptions to the changes in the ValueBase and SILValue classes.
In a bunch of use-cases we use stripSinglePredecessorArgs to eliminate this
case. There is no reason to assume that this is being done in the caller of
RCIdentity. Lets make sure that we handle this case here.
rdar://24156136
One common problem in swift code is the "reforming enum problem". What
happens here is that we have some enum %0 : $Optional<T> and we break it
apart and reform it as a new enum as in the following:
bb9:
...
switch_enum %0 : $Optional<T>, #Optional.None: bb10,
#Optional.Some: bb11
bb10:
%1 = enum $Optional<U>, #Optional.None
br bb12(%1 : $Optional<U>)
bb11:
%2 = some_cast_to_u %0 : ...
%3 = enum $Optional<U>, #Optional.Some, %2 : $U
br bb12(%3 : $Optional<U>)
bb12(%4 : $Optional<U>):
retain_value %0 : $Optional<T> // id %5
release_value %4 : $Optional<U> // id %6
We really would like to know that a retain on %4 is equivalent to a
retain on %0 so we can eliminate the retain, release pair. To be able to
do that safely, we need to know that along all paths %0 and %4 either:
1. Both refer to the same RCIdentity directly. An example of this is the
edge from bb11 -> bb12).
2. Both refer to the "null" RCIdentity (i.e. do not have a payload). An
example of this is the edge from bb10 -> bb12.
Only in such cases is it safe to match up %5, %6 and eliminate them. If
this is not true along all paths like in the following:
bb9:
...
cond_br %foo, bb10, bb11
bb10:
%1 = enum $Optional<U>, #Optional.None
br bb12(%1 : $Optional<U>)
bb11:
%2 = some_cast_to_u %0 : ...
%3 = enum $Optional<U>, #Optional.Some, %2 : $U
br bb12(%3 : $Optional<U>)
bb12(%4 : $Optional<U>):
retain_value %0 : $Optional<T> // id %5
release_value %4 : $Optional<U> // id %6
then we may have that %0 is always non-payloaded coming into bb12. Then
by matching up %0 and %4, if we go from bb9 -> bb11, we will lose a
retain.
Perf Changes:
TITLE..................OLD...........NEW...........NEW/OLD
LevenshteinDistance....1398195.00....1177397.00....0.84
Memset.................26541.00......23701.00......0.89
CaptureProp............5603.00.......5031.00.......0.90
ImageProc..............1281.00.......1196.00.......0.93
InsertionSort..........109828.00.....104129.00.....0.95
StringWalk.............6813.00.......7456.00.......1.09
Chars..................27182.00......30443.00......1.12
The StringWalk, Chars are both reproducible for me. When I turn back on parts of
the recursion (I took the recursion out to make this change more conservative),
the Chars regression goes away, but the StringWalk stays. I have not had a
chance to look at what is going on with StringWalk.
rdar://19724405
Swift SVN r25339
