Commit Graph

80790 Commits

Author SHA1 Message Date
Anders Carlsson 382a96550b Remove some VTT builder arguments that were always zero.
llvm-svn: 99576
2010-03-26 00:11:51 +00:00
Evan Cheng 7b4a1a221b Try trivial remat before the coalescer gives up on a vr / physreg coalescing for fear of tying up a physical register.
llvm-svn: 99575
2010-03-26 00:07:25 +00:00
Dale Johannesen 5d99d7fe79 Handle DEBUG_VALUE in this pass.
llvm-svn: 99573
2010-03-26 00:02:44 +00:00
Jim Grosbach 71fcb4fedd switch the flag for using NEON for SP floating point to a subtarget 'feature'.
Re-commit. This time complete with testsuite updates.

llvm-svn: 99570
2010-03-25 23:47:34 +00:00
Jim Grosbach 42bb89c7d9 need to fix 'make check' tests first. revert for a moment.
llvm-svn: 99569
2010-03-25 23:34:05 +00:00
Jim Grosbach 7fce4e39aa switch the flag for using NEON for SP floating point to a subtarget 'feature'
llvm-svn: 99568
2010-03-25 23:32:19 +00:00
Gabor Greif 6c6b2fd2b2 rename pred_const_iterator to const_pred_iterator for consistency's sake
llvm-svn: 99567
2010-03-25 23:25:28 +00:00
Johnny Chen a3617ec88a Removed instruction class NI from ARMInstrFormats.td.
It doesn't seem to be used anywhere.

llvm-svn: 99566
2010-03-25 23:11:56 +00:00
Jim Grosbach a43386ba8f switch the use-vml[as] instructions flag to a subtarget 'feature'
llvm-svn: 99565
2010-03-25 23:11:16 +00:00
Gabor Greif c78d720f02 rename use_const_iterator to const_use_iterator for consistency's sake
llvm-svn: 99564
2010-03-25 23:06:16 +00:00
Daniel Dunbar d821f4ac60 llvm-mc: Add a -mc-relax-all option, which relaxes every fixup. We always need
exactly two passes in that case, and don't ever need to recompute any layout,
so this is a nice baseline for relaxation performance.

llvm-svn: 99563
2010-03-25 22:49:09 +00:00
Douglas Gregor 63eed63312 Warn when the conversion of an integral non-type template argument to
the type of its corresponding non-type template parameter changes the
value. Previously, we were diagnosing this as an error, which was
wrong. We give reasonably nice warnings like:

test/SemaTemplate/temp_arg_nontype.cpp💯10: warning: non-type template
      argument value '256' truncated to '0' for template parameter of type
      'unsigned char'
Overflow<256> *overflow3; // expected-warning{{non-type template ...
         ^~~
test/SemaTemplate/temp_arg_nontype.cpp:96:24: note: template parameter is
      declared here
template<unsigned char C> struct Overflow;
                       ^

llvm-svn: 99561
2010-03-25 22:21:04 +00:00
Douglas Gregor 210b590562 Teach the diagnostic engine to provide more detailed information about
how to handle a diagnostic during template argument deduction, which
may be "substitution failure", "suppress", or "report". This keeps us
from, e.g., emitting warnings while performing template argument
deduction.

llvm-svn: 99560
2010-03-25 22:17:48 +00:00
John McCall e99d5f3044 When finishing a function definition, leave the function definition *after*
doing all the cleanup tasks and checks.  This gives us the proper context for
checking access to base and member destructors.

llvm-svn: 99559
2010-03-25 22:08:03 +00:00
Fariborz Jahanian f9c4585c80 Fix a code gen. bug involving generation of getter method
from properties of _Complex type. (radar 7351147).

llvm-svn: 99558
2010-03-25 21:56:43 +00:00
Johnny Chen 91d2774416 Add NVDupLnFrm and change NVDupLane class to use that format.
llvm-svn: 99557
2010-03-25 21:49:12 +00:00
Anders Carlsson 151b062fee Don't add address points for virtual primary bases that aren't primary bases in the complete class.
llvm-svn: 99555
2010-03-25 21:45:14 +00:00
John McCall 27e70baa55 Fix a very minor oversight in privileges-elevation: we were only considering
friendship for a derived class if the base class specifier was non-public,
and thus not considering friendship for non-public members of public bases.

llvm-svn: 99554
2010-03-25 21:39:55 +00:00
John McCall e9eaf8e0f8 Handle simple friend-class decls in class templates better by ensuring that
we look for shadow friend decls in the appropriate scope before injecting
a new declaration.

llvm-svn: 99552
2010-03-25 21:28:06 +00:00
Jim Grosbach 4b3b2ef65c ARM cortex-a8 doesn't do vmla/vmls well. disable them by default for that cpu
llvm-svn: 99549
2010-03-25 20:48:50 +00:00
Johnny Chen d82f9002e4 Add NVCVTFrm (NEON Convert with fractional bits immediate) and modify N2VImm to
expect a Format arg.  N2VCvtD/N2VCvtQ are modified to use the NVCVTFrm format.

llvm-svn: 99548
2010-03-25 20:39:04 +00:00
Evan Cheng dbcf861a96 Add nounwind.
llvm-svn: 99546
2010-03-25 20:01:07 +00:00
Evan Cheng 510bda2064 Code clean up.
llvm-svn: 99544
2010-03-25 19:46:11 +00:00
Daniel Dunbar 6432bd744e MC: Stop restarting layout on every relaxation.
- Still O(N^2), just a faster form, and now its the MCAsmLayout's fault.

On the .s I am tuning against (combine.s from 403.gcc):
--
ddunbar@lordcrumb:MC$ diff stats-before.txt stats-after.txt
5,10c5,10
<    1728 assembler - Number of assembler layout and relaxation steps
<    7707 assembler - Number of emitted assembler fragments
<  120588 assembler - Number of emitted object file bytes
< 2233448 assembler - Number of evaluated fixups
<    1727 assembler - Number of relaxed instructions
< 6723845 mcexpr    - Number of MCExpr evaluations
---
>      3 assembler - Number of assembler layout and relaxation steps
>   7707 assembler - Number of emitted assembler fragments
> 120588 assembler - Number of emitted object file bytes
>  14796 assembler - Number of evaluated fixups
>   1727 assembler - Number of relaxed instructions
>  67889 mcexpr    - Number of MCExpr evaluations
--
Feel free to LOL at the -before numbers, if you like.

I am a little surprised we make more than 2 relaxation passes. It's pretty
trivial for us to do relaxation out-of-order if that would give a speedup.

llvm-svn: 99543
2010-03-25 19:35:56 +00:00
Daniel Dunbar d919276bc0 Fix -Asserts warning, again.
llvm-svn: 99542
2010-03-25 19:35:53 +00:00
Jakob Stoklund Olesen 3758ff917e Tag SSE2 integer instructions as SSEPackedInt.
llvm-svn: 99540
2010-03-25 18:52:04 +00:00
Jakob Stoklund Olesen f8d7eda663 Teach TableGen to understand X.Y notation in the TSFlagsFields strings.
Remove much horribleness from X86InstrFormats as a result. Similar
simplifications are probably possible for other targets.

llvm-svn: 99539
2010-03-25 18:52:01 +00:00
Chris Lattner fc4ec25363 fix a valgrind error on copy-constructor-synthesis.cpp, which is caused when
the custom insertion hook deletes the instruction, then we try to set dead
flags on it.  Neither the code that I added nor the code that was there 
before was safe.

llvm-svn: 99538
2010-03-25 18:49:10 +00:00
Evan Cheng a1d0a02713 Remove an unused option.
llvm-svn: 99537
2010-03-25 18:37:23 +00:00
Daniel Dunbar 0ba6a671d4 MC: Simplify main section layout process by moving alignment into LayoutSection.
llvm-svn: 99529
2010-03-25 18:16:42 +00:00
Daniel Dunbar 25d114b2b2 MC: Sink Section address assignment into LayoutSection.
llvm-svn: 99528
2010-03-25 18:16:38 +00:00
Fariborz Jahanian bb4978fbae Check for some code gen. for PR6641 test.
llvm-svn: 99526
2010-03-25 18:05:35 +00:00
John McCall 15ad0962dc Preserve type-source information in friend declarations.
llvm-svn: 99525
2010-03-25 18:04:51 +00:00
Jakob Stoklund Olesen 49e121d5e4 Add a late SSEDomainFix pass that twiddles SSE instructions to avoid domain crossings.
On Nehalem and newer CPUs there is a 2 cycle latency penalty on using a register
in a different domain than where it was defined. Some instructions have
equvivalents for different domains, like por/orps/orpd.

The SSEDomainFix pass tries to minimize the number of domain crossings by
changing between equvivalent opcodes where possible.

This is a work in progress, in particular the pass doesn't do anything yet. SSE
instructions are tagged with their execution domain in TableGen using the last
two bits of TSFlags. Note that not all instructions are tagged correctly. Life
just isn't that simple.

The SSE execution domain issue is very similar to the ARM NEON/VFP pipeline
issue handled by NEONMoveFixPass. This pass may become target independent to
handle both.

llvm-svn: 99524
2010-03-25 17:25:00 +00:00
Daniel Dunbar 3f540c0d7d Remove support for nand atomic builtins. They are inconsistently implemented in
gcc, and the common expectation seems to be that they are unused. If and when
someone cares we can add them back with well documented demantics.

llvm-svn: 99522
2010-03-25 17:13:09 +00:00
Johnny Chen 45ab3f3ccf Added a new instruction class NVDupLane to be inherited by VDUPLND and VDUPLNQ,
instead of the current N2V.  Format of NVDupLane instances are set to NEONFrm
currently.

llvm-svn: 99518
2010-03-25 17:01:27 +00:00
Anders Carlsson a5736bde2e Use the new vtable layout code for construction vtables.
llvm-svn: 99516
2010-03-25 16:49:53 +00:00
Douglas Gregor 6c689cc75c Use Daniel's trick for XFAIL'd tests
llvm-svn: 99515
2010-03-25 16:40:13 +00:00
Bob Wilson e543e7fcb1 Reapply Kevin's change 94440, now that Chris has fixed the limitation on
opcode values fitting in one byte (svn r99494).

llvm-svn: 99514
2010-03-25 16:36:14 +00:00
John Thompson 957816fbf3 PS3 needs __PPC__. Should this be in the PPC target?
llvm-svn: 99513
2010-03-25 16:18:32 +00:00
Daniel Dunbar eb4bc7ffd3 Sketch a few Clang release notes.
llvm-svn: 99512
2010-03-25 16:09:18 +00:00
Douglas Gregor 66767ea93c Add another test for weird substitutions into function types during template argument deduction
llvm-svn: 99510
2010-03-25 15:42:11 +00:00
Douglas Gregor da61afaf9d Improve our handling of local instantiation scopes in two related ways:
- When substituting template arguments as part of template argument
    deduction, introduce a new local instantiation scope.
  - When substituting into a function prototype type, introduce a new
    "temporary" local instantiation scope that merges with its outer
    scope but also keeps track of any additions it makes, removing
    them when we exit that scope.

Fixes PR6700, where we were getting too much mixing of local
instantiation scopes due to template argument deduction that
substituted results into function types.

llvm-svn: 99509
2010-03-25 15:38:42 +00:00
Anders Carlsson a4147148ff When -fdump-vtable-layouts is specified, construction vtable initializers will be generated using the new vtable layout code. (The code is still not completely in place but this is a huge step forward).
llvm-svn: 99508
2010-03-25 15:26:28 +00:00
Devang Patel 95cd4b9c0a Add comment.
llvm-svn: 99507
2010-03-25 15:09:44 +00:00
John Criswell 4e61b255e2 Fixed spurious warning problem noticed by Daniel Dunbar. The configure script
now configures prerequisite projects individually but also ignores them in the
big project switch statement to avoid the incorrect warning.

llvm-svn: 99506
2010-03-25 13:59:09 +00:00
Nuno Lopes b23f8f9dd5 add dump() method to CanQual for debugging purposes
llvm-svn: 99505
2010-03-25 13:19:42 +00:00
Daniel Dunbar 951459740c MC/Mach-O: Switch to MCSectionData::getOrdinal.
llvm-svn: 99504
2010-03-25 08:08:54 +00:00
Evan Cheng 1889440b52 Scheduler assumes SDDbgValue nodes are in source order. That's true currently. But add an assertion to verify it.
llvm-svn: 99501
2010-03-25 07:16:57 +00:00
Daniel Dunbar 4108802628 MC: Explicity track section and fragment ordinals.
llvm-svn: 99500
2010-03-25 07:10:11 +00:00