Commit Graph

83777 Commits

Author SHA1 Message Date
Douglas Gregor c278d1b3b9 Revert r103880 (thread-safe static initialization w/ exceptions),
because it's causing strange linker errors. Unfixes PR7144.

llvm-svn: 103890
2010-05-16 00:44:00 +00:00
Fariborz Jahanian 1b8b8bf25f Minor twik to my last patch. (for radar 7986354).
llvm-svn: 103889
2010-05-16 00:10:46 +00:00
Howard Hinnant e3900731f6 [rand.dist.norm.chisq]
llvm-svn: 103888
2010-05-15 23:36:00 +00:00
Fariborz Jahanian e1b45a5e6b Fix API gen for objc_msgSend property of aggregate types
in Objective-c++ mode. Fixes radar 7986354.

llvm-svn: 103887
2010-05-15 23:05:52 +00:00
Howard Hinnant 932ce81fe9 Revisited [rand.dist.bern.bin] and [rand.dist.pois.poisson] with better algorithms
llvm-svn: 103886
2010-05-15 21:36:23 +00:00
Dale Johannesen 3a366a88f2 Fix uint64->{float, double} conversion to do rounding correctly in 32-bit.
The implementation in LegalizeIntegerTypes to handle this as 
sint64->float + appropriate power of 2 is subject to double rounding,
considered incorrect by numerics people.  Use this implementation only
when it is safe.  This leads to using library calls in some cases
that produced inline code before, but it's correct now.
(EVTToAPFloatSemantics belongs somewhere else, any suggestions?)

Add a correctly rounding (though not particularly fast) conversion
that uses X87 80-bit computations for x86-32.

7885399, 5901940.  This shows up in gcc.c-torture/execute/ieee/rbug.c
in the gcc testsuite on some platforms.

llvm-svn: 103883
2010-05-15 18:51:12 +00:00
Dale Johannesen bb4656c05e Improve assertion messages.
llvm-svn: 103882
2010-05-15 18:38:02 +00:00
Anton Korobeynikov 1bf28a128b Some cheap DAG combine goodness for multiplication with a particular constant.
This can be extended later on to handle more "complex" constants.

llvm-svn: 103881
2010-05-15 18:16:59 +00:00
Douglas Gregor 58142dd8a1 When initializing thread-safe statics, put the call to
__cxa_guard_abort along the exceptional edge into (in effect) a nested
"try" that rethrows after aborting. Fixes PR7144 and the remaining
Boost.ProgramOptions failures.

llvm-svn: 103880
2010-05-15 17:55:51 +00:00
Douglas Gregor 4f0ed42601 Attempt to satisfy Release-Asserts build
llvm-svn: 103879
2010-05-15 17:28:53 +00:00
Anton Korobeynikov 2b7aace2e0 "trap" pseudo-op turned out to be apple-local.
Temporary emit it as raw bytes until it will be added to binutils as well.

llvm-svn: 103878
2010-05-15 17:19:20 +00:00
Chris Lattner 9e01b615a4 improve portability to systems that don't have round, patch by
Evzen Muller!

llvm-svn: 103877
2010-05-15 17:11:55 +00:00
Chris Lattner 93cd0f1c89 improve portability to systems that don't have powf/modf (e.g. solaris 9)
patch by Evzen Muller!

llvm-svn: 103876
2010-05-15 17:10:24 +00:00
John McCall b0efad148c Modify this comment per Doug's suggestion: we don't need to mangle protocols.
llvm-svn: 103875
2010-05-15 17:06:29 +00:00
Chris Lattner 6459bbe14a sigh
llvm-svn: 103874
2010-05-15 16:56:34 +00:00
Chris Lattner 8b3b145342 fix _mm_shuffle_pd too, thanks to Joel Falcou for pointing this out.
llvm-svn: 103873
2010-05-15 16:54:46 +00:00
Douglas Gregor 170125648c When applying the named return value optimization, we still need to
destroy the variable along the exceptional edge; it's only during
normal execution that we avoid destroying this variable.

llvm-svn: 103872
2010-05-15 16:39:56 +00:00
John McCall 8b07ec253d Substantially alter the design of the Objective C type AST by introducing
ObjCObjectType, which is basically just a pair of
  one of {primitive-id, primitive-Class, user-defined @class}
with
  a list of protocols.
An ObjCObjectPointerType is therefore just a pointer which always points to
one of these types (possibly sugared).  ObjCInterfaceType is now just a kind
of ObjCObjectType which happens to not carry any protocols.

Alter a rather large number of use sites to use ObjCObjectType instead of
ObjCInterfaceType.  Store an ObjCInterfaceType as a pointer on the decl rather
than hashing them in a FoldingSet.  Remove some number of methods that are no
longer used, at least after this patch.

By simplifying ObjCObjectPointerType, we are now able to easily remove and apply
pointers to Objective-C types, which is crucial for a certain kind of ObjC++
metaprogramming common in WebKit.

llvm-svn: 103870
2010-05-15 11:32:37 +00:00
Chandler Carruth 75142e6bfc Fix an GCC warning that seems to have actually caught a bug (!!!) in
a condition's grouping. Every other use of Allocatable.test(Hint) groups it the
same way as it is indented, so move the parentheses to agree with that
grouping.

llvm-svn: 103869
2010-05-15 10:23:23 +00:00
Evan Cheng 9e688cbcc9 Model 128-bit vld lane with REG_SEQUENCE.
llvm-svn: 103868
2010-05-15 07:53:37 +00:00
Douglas Gregor 290c93ec0d Implement a simple form of the C++ named return value optimization for
return statements. We perform NRVO only when all of the return
statements in the function return the same variable. Fixes some link
failures in Boost.Interprocess (which is relying on NRVO), and
probably improves performance for some C++ applications.

llvm-svn: 103867
2010-05-15 06:46:45 +00:00
Jakob Stoklund Olesen 84ce290822 Calculate liveness on the fly for local registers.
When working top-down in a basic block, substituting physregs for virtregs, the use-def chains are kept up to date. That means we can recognize a virtreg kill by the use-def chain becoming empty.

This makes the fast allocator independent of incoming kill flags.

llvm-svn: 103866
2010-05-15 06:09:08 +00:00
Douglas Gregor 6fd1b1802f Implement semantic analysis and an AST representation for the named
return value optimization. Sema marks return statements with their
NRVO candidates (which may or may not end up using the NRVO), then, at
the end of a function body, computes and marks those variables that
can be allocated into the return slot.

I've checked this locally with some debugging statements (not
committed), but there won't be any tests until CodeGen comes along.

llvm-svn: 103865
2010-05-15 06:01:05 +00:00
Chris Lattner 6625c7028e add type casts to _mm_shuffle_ps so the front-end applies
implicit conversions where needed.  This fixes an error reported
on cfe-dev, not really worth a testcase though.

llvm-svn: 103864
2010-05-15 05:53:53 +00:00
Nick Lewycky b35818eb25 Teach the always inliner to release its inline cost estimates, like the basic
inliner did in r103653. Why does the always inliner even bother with cost
estimates anyways?

llvm-svn: 103858
2010-05-15 04:26:25 +00:00
Nick Lewycky 002a45eb64 Clean up, no functional change.
llvm-svn: 103857
2010-05-15 03:41:58 +00:00
Jakob Stoklund Olesen 3ced97c0a0 Add reg_nodbg_iterator
llvm-svn: 103856
2010-05-15 02:52:58 +00:00
Evan Cheng 3d214cdfaf v4i64 and v8i64 are only synthesizable when NEON is available.
llvm-svn: 103855
2010-05-15 02:20:21 +00:00
Evan Cheng 4cad68eb34 Allow TargetLowering::getRegClassFor() to be called on illegal types. Also
allow target to override it in order to map register classes to illegal
but synthesizable types. e.g. v4i64, v8i64 for ARM / NEON.

llvm-svn: 103854
2010-05-15 02:18:07 +00:00
Nick Kledzik 749dc33dff <rdar://problem/7987491> Libcompiler_rt isn't platform aware
llvm-svn: 103852
2010-05-15 01:42:04 +00:00
Evan Cheng 0cbd11dfb2 Model 64-bit lane vld with REG_SEQUENCE.
llvm-svn: 103851
2010-05-15 01:36:29 +00:00
Evan Cheng e26e56e72b A partial re-def instruction may be a copy.
llvm-svn: 103850
2010-05-15 01:35:44 +00:00
Douglas Gregor 222cf0efbc Recognize when the named return value optimization applies in a
"return" statement and mark the corresponding CXXConstructExpr as
elidable. Teach CodeGen that eliding a temporary is different from
eliding an object construction.

This is just a baby step toward NRVO.

llvm-svn: 103849
2010-05-15 00:13:29 +00:00
Daniel Dunbar 36d4d1541c C++/ABI/x86_64: Member pointers should be classified as INTEGER.
llvm-svn: 103843
2010-05-15 00:00:37 +00:00
Daniel Dunbar 4bd95c65e7 C++/ABI/i386: Member function pointers should be passed by value.
llvm-svn: 103842
2010-05-15 00:00:30 +00:00
Ted Kremenek be4556f550 Change install_name for libclang.dylib to '@rpath/libclang.dylib'. Fixes <rdar://problem/7987039>.
llvm-svn: 103837
2010-05-14 23:50:33 +00:00
Evan Cheng 8c2d062ea6 Teach two-address pass to do some coalescing while eliminating REG_SEQUENCE
instructions.

e.g.
%reg1026<def> = VLDMQ %reg1025<kill>, 260, pred:14, pred:%reg0
%reg1027<def> = EXTRACT_SUBREG %reg1026, 6
%reg1028<def> = EXTRACT_SUBREG %reg1026<kill>, 5
...
%reg1029<def> = REG_SEQUENCE %reg1028<kill>, 5, %reg1027<kill>, 6, %reg1028, 7, %reg1027, 8, %reg1028, 9, %reg1027, 10, %reg1030<kill>, 11, %reg1032<kill>, 12

After REG_SEQUENCE is eliminated, we are left with:

%reg1026<def> = VLDMQ %reg1025<kill>, 260, pred:14, pred:%reg0
%reg1029:6<def> = EXTRACT_SUBREG %reg1026, 6
%reg1029:5<def> = EXTRACT_SUBREG %reg1026<kill>, 5

The regular coalescer will not be able to coalesce reg1026 and reg1029 because it doesn't
know how to combine sub-register indices 5 and 6. Now 2-address pass will consult the
target whether sub-registers 5 and 6 of reg1026 can be combined to into a larger
sub-register (or combined to be reg1026 itself as is the case here). If it is possible, 
it will be able to replace references of reg1026 with reg1029 + the larger sub-register
index.

llvm-svn: 103835
2010-05-14 23:21:14 +00:00
Dan Gohman 62c70559bc Add some comments about undef.
llvm-svn: 103834
2010-05-14 23:01:02 +00:00
Evan Cheng cb78e5558b Model VST*_UPD and VST*oddUPD pair with REG_SEQUENCE.
llvm-svn: 103833
2010-05-14 22:54:52 +00:00
Dan Gohman 88fb253562 Fast ISel trivially coalesces away no-op casts, so check for this when
setting kill flags.

llvm-svn: 103832
2010-05-14 22:53:18 +00:00
Jakob Stoklund Olesen 089e9421d2 Don't bother spilling before a return
llvm-svn: 103831
2010-05-14 22:40:43 +00:00
Jakob Stoklund Olesen cdef6bc8de RegAllocLocal can count copies too
llvm-svn: 103830
2010-05-14 22:40:40 +00:00
Bill Wendling 0160e55893 SystemZ really does mean "has calls" and not just "adjusts stack." Go ahead and
replace the check with the appropriate predicate. Modify the testcase to reflect
the correct code. (It should be saving callee-saved registers on the stack
allocated by the calling fuction.)

llvm-svn: 103829
2010-05-14 22:17:42 +00:00
Jakob Stoklund Olesen b16013936b Track allocatable instead of reserved regs, and never take an unallocatable hint.
llvm-svn: 103828
2010-05-14 22:02:56 +00:00
Dan Gohman 2f277c866d Don't set kill flags for instructions which the scheduler has cloned.
llvm-svn: 103827
2010-05-14 22:01:14 +00:00
Dan Gohman 062a97f0f6 BR is a barrier.
llvm-svn: 103826
2010-05-14 22:00:27 +00:00
Daniel Dunbar 0d8ca9e91a Driver: Make -fverbose-asm the default for platforms where the integrated
assembler defaults to on.

llvm-svn: 103825
2010-05-14 22:00:22 +00:00
Bill Wendling 1713d95874 Several tail call tests apparently rely upon this being "adjusts stack" instead
of "has calls". That's probably wrong, but it needs further
investigation. Revert to the original behavior until this is settled.

llvm-svn: 103824
2010-05-14 21:58:35 +00:00
Jakob Stoklund Olesen e68b814c8c Avoid scanning the long tail of physreg operands on calls
llvm-svn: 103823
2010-05-14 21:55:52 +00:00
Devang Patel 36debf8046 Do not forget to mark prcessed arguments.
llvm-svn: 103822
2010-05-14 21:55:50 +00:00