Commit Graph

77049 Commits

Author SHA1 Message Date
John McCall 58cc69d4c1 Implement access control for overloaded functions. Suppress access control
diagnostics in "early" lookups, such as during typename checks and when building
unresolved lookup expressions.

llvm-svn: 94647
2010-01-27 01:50:18 +00:00
Evan Cheng 820a318c55 Remove a dead target hook.
llvm-svn: 94646
2010-01-27 01:44:40 +00:00
Owen Anderson 029eb7d243 Clarify the documentation of ext_vector, and add a small example. Hopefully
this will alleviate some confusion about the existence of this feature.

Comments/improvements welcome.

llvm-svn: 94645
2010-01-27 01:22:36 +00:00
Daniel Dunbar b473083fcd Support -marm by treating it as an alias for -mno-thumb.
llvm-svn: 94644
2010-01-27 01:06:14 +00:00
Chris Lattner 57f064722a ppc/linux isn't ready for this and it was an accident that it was included.
This should fix a bunch of linux buildbot failures.

llvm-svn: 94643
2010-01-27 01:02:43 +00:00
Daniel Dunbar f48d51d693 Driver/Darwin: Eliminate invalid uses of DarwinVersion -- this fixes a number of
defaults when targetting iPhoneOS (blocks, non-fragile ABI, stack protector).

llvm-svn: 94642
2010-01-27 00:57:11 +00:00
Daniel Dunbar 6d23b2f503 Driver/Darwin: Simplify target version checks.
llvm-svn: 94641
2010-01-27 00:57:03 +00:00
Daniel Dunbar 8360803a8a Driver/Darwin: Fix a few link arguments when targetting different iPhoneOS
versions to match gcc (which is matching linker changes).

llvm-svn: 94640
2010-01-27 00:56:56 +00:00
Daniel Dunbar b73bb5d085 Driver/Darwin: Switch clang -triple synthesis to use computed target information.
llvm-svn: 94639
2010-01-27 00:56:49 +00:00
Daniel Dunbar 7c870175eb Driver/Darwin: Eliminate confusing IsIPhoneOS parameter -- this was actually
just indicating whether the default target should be iPhoneOS.

llvm-svn: 94638
2010-01-27 00:56:44 +00:00
Daniel Dunbar 15c8942baf Driver/Darwin: Switch to using isTargetIPhoneOS where appropriate.
llvm-svn: 94637
2010-01-27 00:56:37 +00:00
Daniel Dunbar b92d12622c Driver/Darwin: Disallow -m{macosx,iphoneos}-version-min as -Xarch arguments.
llvm-svn: 94636
2010-01-27 00:56:31 +00:00
Daniel Dunbar 3b8e50d0ed Driver/Darwin: Track target platform more explicitly in tool chain, eventually
this should just be part of the tool chain itself once we have eliminated
argument translation.

llvm-svn: 94635
2010-01-27 00:56:25 +00:00
Victor Hernandez 477d9274bb When converting dbg.declare to dbg.value, attach promoted store's debug metadata to dbg.value
llvm-svn: 94634
2010-01-27 00:44:36 +00:00
Ted Kremenek 7ad29c72e3 Add const version of ObjCMethodDecl::getCanonicalDecl().
llvm-svn: 94633
2010-01-27 00:37:54 +00:00
Victor Hernandez 38e810d870 Linker needs to do deep-copy of function-local metadata to update references to function arguments
llvm-svn: 94632
2010-01-27 00:30:42 +00:00
Chris Lattner 2ab11001cc use existing basic block numbers instead of recomputing
a new set of them.

llvm-svn: 94631
2010-01-27 00:20:02 +00:00
Chris Lattner 9da5e1f650 Switch MSP430, CellSPU, SystemZ, Darwin/PPC, Alpha, and Sparc to
EmitFunctionHeader:

7 files changed, 16 insertions(+), 210 deletions(-)

llvm-svn: 94630
2010-01-27 00:17:20 +00:00
Evan Cheng 640b58b8d7 Clarify what -tailcallopt option actually do.
llvm-svn: 94628
2010-01-27 00:10:09 +00:00
Jim Grosbach a3575ca846 Adjust setjmp instruction sequence to not need 32-bit alignment padding
llvm-svn: 94627
2010-01-27 00:07:20 +00:00
Evan Cheng 67a69dd2ed Eliminate target hook IsEligibleForTailCallOptimization.
Target independent isel should always pass along the "tail call" property. Change
target hook LowerCall's parameter "isTailCall" into a refernce. If the target
decides it's impossible to honor the tail call request, it should set isTailCall
to false to make target independent isel happy.

llvm-svn: 94626
2010-01-27 00:07:07 +00:00
Evan Cheng d10089a5b4 Restore to pre-94570 state.
llvm-svn: 94625
2010-01-27 00:00:57 +00:00
Chris Lattner bac7498f8e mcize label emission for functions.
llvm-svn: 94624
2010-01-26 23:53:39 +00:00
Chris Lattner fc2e376a15 use EmitLinkage for functions as well as globals. One output
change is that we now use ".linkonce discard" for global variables
instead of ".linkonce samesize".  These should be the same, just less
strict.  If anyone is interested in mcizing MCSection for COFF targets,
this should be easy to fix.

llvm-svn: 94623
2010-01-26 23:51:52 +00:00
Ted Kremenek e36bceb97d Teach RegionStore to handle initialization of incomplete arrays in structures using a compound value. Fixes <rdar://problem/7515938>.
llvm-svn: 94622
2010-01-26 23:51:00 +00:00
Chris Lattner dbf5a7617d pull linkage emission code out to a new EmitLinkage function.
llvm-svn: 94621
2010-01-26 23:47:12 +00:00
Chris Lattner 22fc38dd4f rearrange some directives, no functionality change.
llvm-svn: 94620
2010-01-26 23:41:48 +00:00
Jeffrey Yasskin 6d7deae2df Roll r94484 (avoiding RTTI problems in tests) forward again in a way that isn't
broken by setting CXXFLAGS on the command line.

llvm-svn: 94619
2010-01-26 23:30:46 +00:00
Victor Hernandez 2b17e2a452 Avoid extra calls to MD->getNumOperands()
llvm-svn: 94618
2010-01-26 23:29:09 +00:00
Evan Cheng 75766332f9 Ignore 'forced' tailcall opt in fastisel mode.
llvm-svn: 94617
2010-01-26 23:28:40 +00:00
Chris Lattner 56e9b5975d remove a noop function.
llvm-svn: 94616
2010-01-26 23:26:29 +00:00
Jakob Stoklund Olesen 0234628284 Fix inline cost predictions with SCIENCE.
After running a batch of measurements, it is clear that the inliner metrics
need some adjustments:

Own argument bonus:       20 -> 5
Outgoing argument penalty: 0 -> 5
Alloca bonus:             10 -> 5
Constant instr bonus:      7 -> 5
Dead successor bonus:     40 -> 5*(avg instrs/block)

The new cost metrics are generaly 25 points higher than before, so we may need
to move thresholds.

With this change, InlineConstants::CallPenalty becomes a political correction:

if (!isa<IntrinsicInst>(II) && !callIsSmall(CS.getCalledFunction()))
  NumInsts += InlineConstants::CallPenalty + CS.arg_size();

The code size is accurately modelled by CS.arg_size(). CallPenalty is added
because calls tend to take a long time, so it may not be worth it to inline a
function with lots of calls.

All of the political corrections are in the InlineConstants namespace:
IndirectCallBonus, CallPenalty, LastCallToStaticBonus, ColdccPenalty,
NoreturnPenalty.

llvm-svn: 94615
2010-01-26 23:21:56 +00:00
Chris Lattner c7b91156e4 now that enough stuff is constified, move function header printing
logic up from X86 into the common code.  The other targets will 
hopefully start using this soon.

llvm-svn: 94614
2010-01-26 23:18:44 +00:00
Chris Lattner 76555b513c constify a bunch of dwarf stuff now that the registerinfo method
is constified.

llvm-svn: 94613
2010-01-26 23:18:02 +00:00
Chris Lattner f13e5f7323 constify a method argument.
llvm-svn: 94612
2010-01-26 23:15:09 +00:00
Evan Cheng c35b5a123b Allow some automatic tailcall optimization without changing ABI.
llvm-svn: 94611
2010-01-26 23:13:04 +00:00
Evan Cheng 1b7b9e852c Delete blank lines that bug me.
llvm-svn: 94610
2010-01-26 23:07:57 +00:00
Ted Kremenek 6296e0990b Move 'LocalCheckers.h' to the 'Checkers' subdirectory.
llvm-svn: 94609
2010-01-26 22:59:55 +00:00
Ted Kremenek b29faead69 Move 'include/Checkers/PathSensitive/Checkers' directory to 'include/Checkers'.
llvm-svn: 94608
2010-01-26 22:56:15 +00:00
Fariborz Jahanian d59baba9fd Patch to add 'volatile' to first argument of all
sync compare_and_swap to match them with gcc.
It also fixes a few incorrect argument defitions.
(fixes radar 7540291).

llvm-svn: 94607
2010-01-26 22:48:42 +00:00
Mike Stump ded0a406b4 Refine the non-virtual this adjustments for thunks by using the offset
to the declaring class from the nearest virtual base class.  WIP.

llvm-svn: 94606
2010-01-26 22:44:01 +00:00
Chris Lattner 3ba6da49f5 call emitconstantpool and emitjumptable like other targets.
llvm-svn: 94601
2010-01-26 22:06:58 +00:00
Devang Patel 670ca2bcbf Before existing NamedMDNode entry in the symbol table, remove any existing entry with the same name.
llvm-svn: 94600
2010-01-26 22:03:41 +00:00
Sebastian Redl 833ef45b1d Bring some semblance of order into Decl.h and Decl.cpp. While at it, fix some typo comments and remove an unused and unimplemented function prototype. No functionality change.
llvm-svn: 94599
2010-01-26 22:01:41 +00:00
Chris Lattner b657c4cdc3 emit jump table an alias ".set" directives through MCStreamer as
assignments.

.set x, a-b

is the same as:

x = a-b

llvm-svn: 94596
2010-01-26 21:53:08 +00:00
Chris Lattner 37bc78a5e2 fix CastInst::castIsValid to reject aggregate types, fixing PR6153:
llvm-as: t.ll:1:25: error: invalid cast opcode for cast from '[4 x i8]' to '[1 x i32]'
@x = constant [1 x i32] bitcast ([4 x i8] c"abcd" to [1 x i32])
                        ^

llvm-svn: 94595
2010-01-26 21:51:43 +00:00
Devang Patel 74da6a0480 Remve unnecessary include.
llvm-svn: 94594
2010-01-26 21:42:58 +00:00
Devang Patel 1973df2e3f Use AssertingVH, just to be paranoid.
llvm-svn: 94593
2010-01-26 21:39:14 +00:00
Mike Stump a04ecfb701 Refine the non-virtual this adjustments for thunks by using the offset
to the declaring class from the nearest virtual base class.  WIP.

This fixes 40% of all the problems remaining in one of my testcases.

llvm-svn: 94592
2010-01-26 21:35:27 +00:00
Jakob Stoklund Olesen 87256d8fe1 Revert test polarity to match comment and desired outcome. Remove undeserved bonus.
A GEP with all constant indices is already considered free by
analyzeBasicBlock(), so don't give it an extra bonus in
CountCodeReductionForAlloca().

This patch should remove a small positive bias toward inlining functions with
variable-index GEPs, and remove a smaller negative bias from functions with
all-constant index GEPs.

llvm-svn: 94591
2010-01-26 21:31:35 +00:00