Commit Graph

38330 Commits

Author SHA1 Message Date
Dale Johannesen 7d97662467 Remove indeterminism from a loop. We think this will
fix an occasional nonrepeatable bootstrap failure we've
been seeing on Darwin.

llvm-svn: 44202
2007-11-17 02:48:01 +00:00
Dale Johannesen 06a10df562 Fix denormal check in float->APInt conversion.
PR 1804.

llvm-svn: 44201
2007-11-17 01:02:27 +00:00
Ted Kremenek 9d89f980aa Changed implementation of Serialize::EmitDiffPtrID and
Deserialize::ReadDiffPtrID to read and emit bools instead of unsigned
integers. This should result in a nice space optimization once we have
"auto-abbreviation" generation in place.

llvm-svn: 44200
2007-11-17 00:45:37 +00:00
Ted Kremenek 4d1e79c8ac Implemented optimization for BatchEmitOwnedPtrs that we only emit one complete
SerializedPtrID, followed by the *differences* in IDs. The big idea is that
most IDs will be just be 1 off from the previous (either that or NULL, which
we encode as a difference if 0), so this will greatly reduce the encoding
space for extra IDs to just 1 bit per pointer.

So far this optimization reduces serialization of Carbon.h by only 1%, but
we aren't using any abbreviations now in the Bitcode file to properly take
advantage of this optimization.

llvm-svn: 44199
2007-11-17 00:40:45 +00:00
Evan Cheng 8e22379303 Live interval splitting:
When a live interval is being spilled, rather than creating short, non-spillable
intervals for every def / use, split the interval at BB boundaries. That is, for
every BB where the live interval is defined or used, create a new interval that
covers all the defs and uses in the BB.

This is designed to eliminate one common problem: multiple reloads of the same
value in a single basic block. Note, it does *not* decrease the number of spills
since no copies are inserted so the split intervals are *connected* through
spill and reloads (or rematerialization). The newly created intervals can be
spilled again, in that case, since it does not span multiple basic blocks, it's
spilled in the usual manner. However, it can reuse the same stack slot as the
previously split interval.

This is currently controlled by -split-intervals-at-bb.

llvm-svn: 44198
2007-11-17 00:40:40 +00:00
Evan Cheng aa72f72bd3 Shrinkfy.
llvm-svn: 44197
2007-11-17 00:31:16 +00:00
Evan Cheng 13e8b022f5 Typo.
llvm-svn: 44196
2007-11-16 23:55:08 +00:00
Dale Johannesen a674612d94 Testcase from PR 1508 (although its's somewhat
orthogonal to the main problem there)

llvm-svn: 44194
2007-11-16 23:16:35 +00:00
Tanya Lattner 1db762ba63 Fix for PR1801
llvm-svn: 44193
2007-11-16 22:44:50 +00:00
Ted Kremenek f115550f6f Added assertion in serialization of DeclRefExprs. DeclRefExprs can only
own the decl they reference if it is a FunctionDecl.  Note that his
ownership property is still considered a hack, and should be fixed.

llvm-svn: 44192
2007-11-16 19:00:35 +00:00
Ted Kremenek 50d7f6f620 Fixed bug in the serialization of FunctionDecls. We would incorrectly
query for the number of parameters for FunctionDecls that had type
FunctionTypeNoProto.

llvm-svn: 44191
2007-11-16 18:11:10 +00:00
Chris Lattner 48d52848d7 Tighten up address-of checking, implementing test/Sema/expr-address-of.c.
This fixes a bug reported by Seo Sanghyeon.

This was meant to be committed yesterday, but the commit failed.  doh.

llvm-svn: 44190
2007-11-16 17:46:48 +00:00
Anton Korobeynikov b8a47244ac These were implemented
llvm-svn: 44189
2007-11-16 11:12:01 +00:00
Chris Lattner c53b18362a Fix PR1800 by correcting mistaken logic.
llvm-svn: 44188
2007-11-16 06:04:17 +00:00
Chris Lattner 1b3f58a816 fix a typo
llvm-svn: 44187
2007-11-16 05:32:05 +00:00
Anton Korobeynikov 91460e43f1 Implement codegen for flt_rounds on x86
llvm-svn: 44183
2007-11-16 01:31:51 +00:00
Anton Korobeynikov 66b91e66ec Implement necessary bits for flt_rounds gcc builtin.
Codegen bits and llvm-gcc support will follow.

llvm-svn: 44182
2007-11-15 23:25:33 +00:00
Nate Begeman bd117f06ba Basic non-power-of-2 vector support
llvm-svn: 44181
2007-11-15 21:15:26 +00:00
Chris Lattner 0640e3ef35 Fix transient failure on Lex/c90.c due to an uninit variable.
llvm-svn: 44180
2007-11-15 19:22:40 +00:00
Chris Lattner 4dcbc2090a add header file I forgot to check in
llvm-svn: 44179
2007-11-15 19:22:18 +00:00
Chris Lattner 609d413363 fix a bug Steve noticed, where a #import of the main file itself would fail.
llvm-svn: 44178
2007-11-15 19:07:47 +00:00
Anton Korobeynikov 6a7ddfdb8f Reverted r44163 per request
llvm-svn: 44177
2007-11-15 18:33:16 +00:00
Ted Kremenek 07941e21e9 Added hack when serializing DeclRefExprs. This should probably be fixed.
Some FunctionDecls do not appear at the top-level or are owned by a DeclStmt.
In calls to implicitly defined functions, a FunctionDecl is created, but only
the DeclRefExprs reference them. Since an implicitly defined function may be
called multiple times, there is no clear ownership model for such objects.

Temporary solution: when serializing out DeclRefExprs, emit an ownership bit
for the Decl. This bit is determined by querying the serializer to see if the
Decl has already been serialized. If it hasn't, emit the Decl as an owned
pointer.

I repeat: this is a hack.  This should be fixed.

llvm-svn: 44176
2007-11-15 18:26:39 +00:00
Ted Kremenek 4791e1a7bd Started separate section for method definitions for ObjC serialization.
Added missing deserialization case in Stmt::Create() switch statement.

llvm-svn: 44175
2007-11-15 18:10:29 +00:00
Daniel Berlin bb3b099f82 Fix bugs in iterator invalidation
llvm-svn: 44174
2007-11-15 18:06:49 +00:00
Steve Naroff c8fcbe09d0 No need to forward declare definition of objc_super...
llvm-svn: 44173
2007-11-15 17:06:21 +00:00
Steve Naroff e3ffc2f4a9 Finish up variadic methods/messages.
llvm-svn: 44172
2007-11-15 13:05:42 +00:00
Steve Naroff d8ea1ac576 Implement support for variadic methods (work in progress).
llvm-svn: 44171
2007-11-15 12:35:21 +00:00
Steve Naroff 05caa48fb4 Extend RewriteTest::RewriteObjCIvarRefExpr() to cope with static typing (when using -> on a type which corresponds to the implementation type).
llvm-svn: 44170
2007-11-15 11:33:00 +00:00
Steve Naroff 44864e48bd Tweak funky cast to accommodate messaging 'super'. This removes any spurious warnings.
llvm-svn: 44169
2007-11-15 10:43:57 +00:00
Steve Naroff 7fa2f040f3 Rewrite for messaging 'super'.
The code gen. results in some spurious warnings...a cast is forthcoming.

llvm-svn: 44168
2007-11-15 10:28:18 +00:00
Duncan Sands d4494352f8 This assertion was bogus.
llvm-svn: 44167
2007-11-15 09:54:37 +00:00
Evan Cheng 2c1a50455c Fix a thinko in post-allocation coalescer.
llvm-svn: 44166
2007-11-15 08:13:29 +00:00
Nate Begeman 847bfbcaf0 Break out bool/true/false support into a LangOption
llvm-svn: 44164
2007-11-15 07:30:50 +00:00
Nick Lewycky fbb24817cc Fix handling of overflow in loop calculation by adding new UDiv SCEV. This SCEV
is disabled in the sense that it will refuse to create one from a UDiv
instruction, until the code is better tested.

llvm-svn: 44163
2007-11-15 06:30:50 +00:00
Chris Lattner 0841c3b560 Fix PR1788 by taking the approach suggested by Richard Smith.
Thanks to him for his detailed analysis of the problem.

llvm-svn: 44162
2007-11-15 06:10:55 +00:00
Chris Lattner 820c1447b4 Import the boost scoped_ptr class to LLVM. This patch was prepared by
Cédric Venet.

llvm-svn: 44161
2007-11-15 05:57:06 +00:00
Nate Begeman 4c18c23eb3 Implement codegen of CXX Bool
llvm-svn: 44160
2007-11-15 05:40:03 +00:00
Chris Lattner bb31a42fef Handle "bool" in all places that touch _Bool.
This fixes code like "if((bool)x) {}" for example.

Patch by Nate Begeman.

llvm-svn: 44159
2007-11-15 05:25:19 +00:00
Owen Anderson 1b74a9a628 More templatization.
llvm-svn: 44158
2007-11-15 05:00:15 +00:00
Chris Lattner add10eecf9 many edits, patch by Kelly Wilson!
llvm-svn: 44157
2007-11-15 04:51:31 +00:00
Steve Naroff f60782b8c9 - Implement ivar rewrite (patch by Fariborz).
- RewriteMessageExpr()...make implicit casts explicit with synthesizing call (removing warnings when calling objc_msgSend()).

llvm-svn: 44156
2007-11-15 02:58:25 +00:00
Bill Wendling b3712f8146 Adding debug output during coalescing.
llvm-svn: 44154
2007-11-15 02:06:30 +00:00
Bill Wendling 8269925b1e Need to increment the iterator.
llvm-svn: 44153
2007-11-15 00:40:48 +00:00
Ted Kremenek c5f3e5371c Fixed serious bug in BatchReadOwnedPtrs where in a chain of calls to
deserialize objects if BatchReadOwnedPtrs was called more than once in the
same call chain then the second call would overwrite the SerializedPtrIDs
being used by the first call. Solved this problem by making the vector that
holds the pointer IDs local to a function call. Now BatchReadOwnedPtrs is
reentrant.

llvm-svn: 44152
2007-11-15 00:05:03 +00:00
Steve Naroff 909d666681 Refinement to previous commit. Always cast the first argument to "id"...no need to special case self.
llvm-svn: 44149
2007-11-15 00:00:21 +00:00
Steve Naroff e7f181986a Cast implicit "self" argument to "id". This removes all warnings associated with implicit references to self. It doesn't yet deal withexplicit references to self...
llvm-svn: 44148
2007-11-14 23:54:14 +00:00
Ted Kremenek cf7a20a6c4 Fixed bug in serialization of EnumConstantDecl where we improperly
"default constructed" an APSInt.  Fixed another bug in the same method
where we did not allow the NextDeclarator to be NULL.

llvm-svn: 44147
2007-11-14 23:38:09 +00:00
Ted Kremenek c77e5a126c Fixed bug when serializing QualTypes where we were serializing in the
qualifiers as part of the pointer address.

llvm-svn: 44146
2007-11-14 23:35:01 +00:00
Steve Naroff 03f2767368 Always generate a typedef for @class. The typedef removal logic that I'm removing only made sense when we were operating on preprocess files without typedef guards. Now that we have guards, it is incorrect to ever remove one (since it may increase the likelihood that the rewritten header can't be included stand alone).
llvm-svn: 44145
2007-11-14 23:02:56 +00:00