method to initialize the members. This doesn't matter so much
for structs (the generated IR is essentially equivalent except for
small structs), but on classes, we don't want to make "new X" generate
code that knows about metadata/destructors/etc for the class X.
Also, make sure classes always have a constructor. (We haven't really
discussed the rules for implicitly declared constructors, so for now,
the rule is just "generate an implicit constructor if there is no
explicit constructor". We'll want to revisit this when we actually
design object construction.)
Swift SVN r2361
Mangling is still a hack, pending a better type AST. Fixed
a bug where arguments passed indirectly were not being destroyed
by the callee (when passed by value). Changed some of the protocol
signatures to use the generic opaque pointer type, making the
types a bit more self-documenting in the IR.
Swift SVN r2274
swift_retain calls. The pertinent difference is that the former can be
marked nocapture, allowing general LLVM optimizations more flexibility.
With this change, early-cse is able to zap 9 more instructions, and 3
more functions are able to be marked nocapture by functionattrs in the
stdlib.
Swift SVN r2043
match the actual entrypoints vended by the runtime:
1) there is no swift_slowRawAlloc
2) swift_deallocObject takes two arguments
Also, make all calls to swift_allocObject go through a common
function so that we can easily use optimized entrypoints if the
runtime provides them.
Swift SVN r1993
of protocol types to correctly handle self-assignment.
This ends up creating such a large amount of code that it's
worth extracting into its own helper function. Fortunately,
this can be the same helper function for every protocol type.
Swift SVN r1905
pointer TypeInfo more conveniently. Have this class return true for
isSingleRetainablePointer(). Further specialize the implementations
of the value witnesses so that we no longer ever require type-specific
witnesses for reference types.
Swift SVN r1900
value witnesses goes.
There are three major remaining things to do to support protocols:
- laying out the actual protocol members
- emitting witnesse for the actual protocol members
- detecting uses of the actual protocol members and funnelling
them through the witnesses as appropriate
All this work was just to let us treat protocol types as
first-class values.
Swift SVN r1899
exactly how the arguments are emitted. Introduce a new code path
which emits arguments from existing Explosions (possibly of the wrong
resilience level). Use that code path to implement getter/setter
code for MemberRefExpr l-values. This is just a checkpoint for the
latter two parts, but the ApplyExpr path is working correctly in the
new framework, which is worth a commit.
Swift SVN r1589
heap allocations it makes, and switch swift_alloc over to pass
that pointer in as well as the alignment. Also, compute
whether a type is POD during its generation and cache that in
the object, and introduce a method on TypeInfo to destroy an
object in memory.
Swift SVN r1356
use irgen::ASTVisitor, which is a subclass of the AST version
that always looks through some expressions and asserts that
others (the unchecked/error ones) can't happen. The main
obstacle to that is access control, because it's quite nice to
have the visitor be a non-local class for nesting purposes,
but that makes it impossible to use private methods anymore.
It's nice to give all the type-specific emitters private
interfaces anyway, though; that ends up being most of the
patch.
I futzed with GenClosure.cpp while I was in there, but
separating it would have been obnoxious.
Swift SVN r1267
an owner attached. Use this to implement [byref(heap)]. Force
locals to the heap if they've been referenced in a way that requires
this.
Swift SVN r1265
qualifier, making sure that variables end up so-qualified by
default. Add a RequalifyExpr to capture the act of adding
qualifiers (to form a supertype) to an l-value.
Swift SVN r1236
