Modeling builtins as first-class function values doesn't really make sense because there's no real function value to emit, and modeling them this way complicates passes that work with builtins because they have to invent function types for builtin invocations. It's much more straightforward to have a single instruction that references the builtin by ID, along with the type information for the necessary values, type parameters, and results, so add a new "builtin" instruction that directly represents a builtin invocation. NFC yet.
Swift SVN r22690
Eliminate the intermediate top_level_code function. Now that SIL is expressive enough to express a "main" function, there's no reason for it, and this eliminates a bunch of mystery code in IRGen to thunk from main to top_level_code by reaching for hardcoded symbol names. Demystify the special code for setting up C_ARGC and C_ARGV by having SILGen look for a transparent "_didEnterMain" hook in the stdlib and emit a call to it.
Swift SVN r22525
This simplifies the code generation path for existential methods by allowing it to shared more code with the generic case, (It'll be even simpler when Sema opens the existentials for SILGen...) turning protocol_method lookups into open_existential + witness_method sequences. In this patch, we handle normal generic method lookups, but property accesses still go through protocol_method.
Swift SVN r22437
This lets us reliably print and parse opened archetypes across different compiler invocations. Using a source-related locator would be ideal, but that's complicated by the need to manufacture, print, and parse these things during SIL passes, so cop out and burn a UUID for now.
Swift SVN r22385
We currently do not serialize the body of global addressors. To make
"sil-opt swiftmodule" pass verification, we change the linkage for
the deserialized empty global addressors from public to public external.
rdar://18021024
Swift SVN r22370
This avoids a pointless copy every time an array literal is written, and will let us retire the horrible "alloc_array" instruction and globs of broken IRGen code. Implements rdar://problem/16386862, and probably fixes a bunch of bugs related to alloc_array brokenness.
Swift SVN r22289
Now the SILLinkage for functions and global variables is according to the swift visibility (private, internal or public).
In addition, the fact whether a function or global variable is considered as fragile, is kept in a separate flag at SIL level.
Previously the linkage was used for this (e.g. no inlining of less visible functions to more visible functions). But it had no effect,
because everything was public anyway.
For now this isFragile-flag is set for public transparent functions and for everything if a module is compiled with -sil-serialize-all,
i.e. for the stdlib.
For details see <rdar://problem/18201785> Set SILLinkage correctly and better handling of fragile functions.
The benefits of this change are:
*) Enable to eliminate unused private and internal functions
*) It should be possible now to use private in the stdlib
*) The symbol linkage is as one would expect (previously almost all symbols were public).
More details:
Specializations from fragile functions (e.g. from the stdlib) now get linkonce_odr,default
linkage instead of linkonce_odr,hidden, i.e. they have public visibility.
The reason is: if such a function is called from another fragile function (in the same module),
then it has to be visible from a third module, in case the fragile caller is inlined but not
the specialized function.
I had to update lots of test files, because many CHECK-LABEL lines include the linkage, which has changed.
The -sil-serialize-all option is now handled at SILGen and not at the Serializer.
This means that test files in sil format which are compiled with -sil-serialize-all
must have the [fragile] attribute set for all functions and globals.
The -disable-access-control option doesn't help anymore if the accessed module is not compiled
with -sil-serialize-all, because the linker will complain about unresolved symbols.
A final note: I tried to consider all the implications of this change, but it's not a low-risk change.
If you have any comments, please let me know.
Swift SVN r22215
Update SILGen to create SILGlobalVariable and SILGlobalAddrInst instead of
GlobalAddrInst. When we see a definition for a global variable, we create
the corrsponding SILGlobalVariable definition.
When creating SILGlobalVariable from a global VarDecl, we mangle the global
VarDecl in the same way as we mangle it at IRGen. The SILLinkage is also
set in the same way as we set it at IRGen.
At IRGen, we use the associated VarDecl for SILGlobalVariable if it exists,
to have better debugging information.
We set the initializer for SILGlobalVariable definition only.
We also handle SILGlobalAddrInst in various SILPasses, in the similar way
as we handle GlobalAddrInst.
rdar://15493694
Swift SVN r21887
This is necessary to be able to properly stash values with nontrivial lowerings, such as metatypes and functions, inside existential containers. Modify SILGen to lower values to the proper abstraction level before storing them in an existential container. Part of the fix for rdar://problem/18189508, though runtime problems still remain when trying to actually dynamicCast out a metatype from an Any container.
Swift SVN r21830
This will let the performance inliner inline a function even if the costs are too high.
This attribute is only a hint to the inliner.
If the inliner has other good reasons not to inline a function,
it will ignore this attribute. For example if it is a recursive function (which is
currently not supported by the inliner).
Note that setting the inline threshold to 0 does disable performance inlining at all and in
this case also the @inline(__always) has no effect.
Swift SVN r21452
This disables inlining at the SIL level. LLVM inlining is still enabled. We can
use this to expose one function at the SIL level - which can participate in
dominance based optimizations but which is implemented in terms of a cheap check
and an expensive check (function call) that benefits from LLVM's inlining.
Example:
The inline(late) in the example below prevents inlining of the two checks. We
can now perform dominance based optimizations on isClassOrObjExistential.
Without blocking inlining the optimizations would apply to the sizeof check
only and we would have multiple expensive function calls.
@inline(late)
func isClassOrObjExistential(t: Type) -> Bool{
return sizeof(t) == sizeof(AnyObject) &&
swift_isClassOrObjExistential(t)
}
We do want inlining of this function to happen at the LLVM level because the
first check is constant folded away - IRGen replaces sizeof by constants.
rdar://17961249
Swift SVN r21286
Start capitalizing on some of the new diagnostic machinery in a few different ways:
- When mining constraints for type information, utilize constraints "favored" by the overload resolution process.
- When printing type variables, if the variable was created by opening a literal expression, utilize the literal
default type or conformance if possible.
- Utilize syntactic information when crafting diagnostics:
- If the constraint miner can produce a better diagnostic than the recorded failure, diagnose via constraints.
- Factor in the expression kind when choosing which types to include in a diagnostic message.
- Start customizing diagnostics based on the amount of type data available.
What does all this mean?
- Fewer type variables leaking into diagnostic messages.
- Far better diagnostics for overload resolution failures. Specifically, we now print proper argument type data
for failed function calls.
- No more "'Foo' is not convertible to 'Foo'" error messages
- A greater emphasis on type data means less dependence on the ordering of failed constraints. This means fewer
inscrutable diagnostics complaining about 'UInt8' when all the constituent expressions are of type Float.
So we still have a ways to go, but these changes should greatly improve the number of head-scratchers served up
by the type checker.
These changes address the following radars:
rdar://problem/17618403
rdar://problem/17559042
rdar://problem/17007456
rdar://problem/17559042
rdar://problem/17590992
rdar://problem/17646988
rdar://problem/16979859
rdar://problem/16922560
rdar://problem/17144902
rdar://problem/16616948
rdar://problem/16756363
rdar://problem/16338509
Swift SVN r20927
We were already effectively doing this everywhere /except/ when building
the standard library (which used -O2), so just use the model we want going
forward.
Swift SVN r20455
I am trying to enable a new ARC optimizer feature that is forcing me to
debug a test case that is affected by inlining of initializers. Plus, being
able to selectively disable inlining of the initializers is a useful
feature in general.
Swift SVN r20427
This used to pull in SIL functions from the stdlib, then try to serialize
them out again. Michael's r20375 allowed us to distinguish local functions
with shared linkage from deserialized functions with shared (now
SharedExternal) linkage, and the existing logic in the serializer took
care of skipping them.
<rdar://problem/17567391>
Swift SVN r20415
*NOTE* This linkage is different from {Public,Hidden}External in that it has no
extra semantic meaning beyond shared.
The use of this linkage is to ensure that we do not serialize deserialized
shared functions. Those shared functions can always be re-deserialized from the
original module. This prevents a whole class of bugs related to the
creation of module cross references since all references to the shared
item go straight to the original module.
<rdar://problem/17772847>
Swift SVN r20375
A while back we decided to require @IBOutlets to be optional (via ! or
?); we got as far as ripping out the implicit !'ification of
@IBOutlets, but never added the diagnostic. Diagnose this restriction
with Fix-Its.
Swift SVN r19981
enforce its own little constraints. The type checker isn't using it for
anything, and it is just clutter.
This resolves <rdar://problem/16656024> Remove @assignment from operator implementations
Swift SVN r19960
modifiers and with the func implementations of the operators. This resolves the rest of:
<rdar://problem/17527000> change operator declarations from "operator prefix" to "prefix operator" & make operator a keyword
Swift SVN r19931
eliminating the @'s from them when used on func's. This is progress towards
<rdar://problem/17527000> change operator declarations from "operator prefix" to "prefix operator" & make operator a keyword
This also consolidates rejection of custom operator definitions into one
place and makes it consistent, and adds postfix "?" to the list of rejected
operators.
This also changes the demangler to demangle weak/inout/postfix and related things
without the @.
Swift SVN r19929
This only tackles the protocol case (<rdar://problem/17510790>); it
does not yet generalize to an arbitrary "class" requirement on either
existentials or generics.
Swift SVN r19896
Mechanically add "Type" to the end of any protocol names that don't end
in "Type," "ible," or "able." Also, drop "Type" from the end of any
associated type names, except for those of the *LiteralConvertible
protocols.
There are obvious improvements to make in some of these names, which can
be handled with separate commits.
Fixes <rdar://problem/17165920> Protocols `Integer` etc should get
uglier names.
Swift SVN r19883
attribute is a "modifier" of a decl, not an "attribute" and thus shouldn't
be spelt with an @ sign. Teach the parser to parse "@foo" but reject it with
a nice diagnostic and a fixit if "foo" is a decl modifier.
Move 'dynamic' over to this (since it simplifies some code), and switch the
@optional and @required attributes to be declmodifiers (eliminating their @'s).
Swift SVN r19787
Introduce the new BooleanLiteralConvertible protocol for Boolean
literals. Take "true" and "false" as real keywords (which is most of the
reason for the testsuite churn). Make Bool BooleanLiteralConvertible
and the default Boolean literal type, and ObjCBool
BooleanLiteralConvertible. Fixes <rdar://problem/17405310> and the
recent regression that made ObjCBool not work with true/false.
Swift SVN r19728
The driver option -i now requires an input file as argument, and any
options after the input file will be treated as arguments to the
interpretted file.
This also renames the frontend option to -interpret, since it is parsed
as a flag, unlike -i. We could support -interpret in the driver if we
wanted, which would allow us to use --, but wouldn't work with shebang
scripts. For now, it's frontend-only.
Swift SVN r19718
...because you can't match them properly in switches.
In the future, we could consider allowing private enum cases in a
resilient public enum, which essentially forces the user to consider the
default case.
Swift SVN r19620
