We now enforce via an assert that each witness table is unique and that
every protocol conformance that is referenceable from SIL must have a
witness table matched to it.
Also, I taught the linker that it should deserialize witness tables for
InitExistentialRefInst instructions, something that was missed before.
Swift SVN r17283
r15824 overrode SILVerifier::visitSILBasicBlock without calling up to the super definition, causing us to never actually verify any instructions for the past three weeks. Awesome. Patch up the latent bugs that have crept in, except for three devirtualizer tests that fail. This doesn't reenable the verifier because I don't want to cause crashes until all the regressions have been cleared up.
Swift SVN r17121
around in the deserializer's .
1,2d
1i
Bump the reference count of a SILFunction as long as it's sitting
around in the deserializer's cache.
If we deserialize a function, then delete it, then deserialize
another reference to it, we'll end up pointing to a deleted
function. This was causing the bug where IRGen would assert
with "no order number for SIL function definition?"; I can't
seem to find a radar for that, though.
Also, deserialization test cases are hard to write and probably
inherently unstable.
Swift SVN r15908
Fixes a problem where CSE would coalesce string_literals with the same text but different encodings when they happened to land in the same hash bucket.
Swift SVN r15603
alloc_ref_dynamic allocates an instance of a class type based on the
value in its metatype operand. Start emitting these instructions for
the allocating constructor of a complete object initializer (not yet
tested) and for the allocating constructor synthesized for an imported
Objective-C init method.
Still missing:
- IRGen still does the same thing as alloc_ref right now. That
change will follow.
- There are devirtualization opportunities when we know the value of
the metatype that would turn an alloc_ref_dynamic into an alloc_ref;
I'm not planning to do this optimization.
Swift SVN r14560
Riding off of project_existential[_ref] was convenient, but the
resuls are used quite differently. Note that open_existential[_ref]
still don't print/parse reasonably yet.
Swift SVN r13878
which is the address of the string data. Have SILGen compute and produce the
isASCII bit as an integer literal, and remove the logic from IRGen.
This overall approach is simpler and enables better SIL-level optimizations.
Swift SVN r13363
give 'let' declarations debug info. This is part of rdar://15785053.
IRGen is just stubbed out and nothing generates these yet (other than
the sil parser).
Swift SVN r12179
Split 'destructive_switch_enum_addr' into separate 'switch_enum_addr' and 'take_enum_data_addr' instructions. This should unblock some optimization work we would like to do with enums.
Swift SVN r12015
SIL instruction. Implement IR-gen support for same. Fix
the parsing of SIL string literals to properly unescape them.
SIL-gen still emits UTF8 literals unconditionally.
Swift SVN r11904
The current implementation depends on the target basic block to record the number of argumens, but this is a bad idea because it makes it difficult to modify the argument list.
Swift SVN r11597
The code here is a simple checklist. It performs the following checks in order:
1. Quick check to see if the kind/number of operands/number of types match up.
2. Comparison all of the types/operands.
3. Finally use a switch to perform specialized per instruction comparisons on
state not represented in the instructions types/operands (i.e. vardecls,
structdecls, etc.).
Additionally the switch is setup so that instructions which have not been fully
implemented in the switch will cause an unreachable to be hit. Once all
instructions have isIdenticalTo implementations (implying the switch is
covered), I will remove the unreachable.
*NOTE* The primary use case for this now is CSE. In CSE we can control which
instructions we process so during the review process I am only going to process
literals for the sake of simplicity (and thus this patch contains only code for
checking literals).
Swift SVN r10743
a FuncDecl. This makes it much more straight-forward for SIL passes to
introduce a new one - without doing name lookup in the builtin module!
Swift SVN r10694
Like GlobalAddrInst, but for SILGlobalVariables. These would become the same instruction when SILGlobalVariable can replace AST-level global references.
Swift SVN r10510
Instead of hardcoding Builtin.Word to be an alias for Builtin.Int64, make it its own type of abstract pointer width.
- Change BuiltinIntegerType's width representation to accommodate abstract widths.
- In the AST and in SIL, store values of the types as the greatest supported size for the abstract width (64 bits for a pointer).
- Add some type safety to the ([sz]ext|trunc)(OrBitCast)? builtins that they're used appropriately given the upper and lower bounds of the abstract sizes they're working with.
- Now that Builtin.Word is a distinct type, give it its own mangling.
- In IRGen, lower pointer-sized BuiltinIntegerType appropriately for the target, and truncate lowered SIL values if necessary.
Fixes <rdar://problem/15367913>.
Swift SVN r10467
(the same way alloc_box returns two) instead of returning a tuple.
This eliminates a ton of tuple_extract instructions, which just
bloat the generated SIL. This resolves rdar://15378135.
Swift SVN r10416
SILFunction that it references. Use this in the mandatory inlining
pass to remove deserialized transparent functions, to clean up the
-emit-sil output of the compiler (and presumably speed up compile
time). This implements rdar://15272652
Swift SVN r9699
Make ApplyInst and PartialApplyInst directly take substitutions for generic functions instead of trying to stage out substitutions separately. The legacy reasons for doing this are gone.
Swift SVN r8747
This mirrors the behavior of project_existential and simplifies some special cases in SILGen. It unfortunately makes dynamic_lookup sequences a bit noisier because of the need to explicitly cast the projection from DynamicLookup.Self to Builtin.ObjCPointer, but I think this modeling is more solid and will fit better with my planned redesign of archetype_method/protocol_method.
Swift SVN r8572
getBitWidthForCase() is called from the destructor of the object and it
used to rely on Operand being around. However, dropAllReferences(), which
is run before instruction destruction makes the operand invalid at the
time of instruction death. (This results in a crash if we were to delete a
SwitchInstInt instruction.)
Swift SVN r8472
