1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-21 12:02:58 +02:00
Commit Graph

46630 Commits

Author SHA1 Message Date
Dan Gohman
c24a00edc2 Remove an irrelevant comment.
llvm-svn: 72132
2009-05-19 20:38:47 +00:00
Dan Gohman
ccf5417982 Add some more comments to the top of this file.
llvm-svn: 72131
2009-05-19 20:37:36 +00:00
Dan Gohman
c44e433ce7 Trim unneeded #includes.
llvm-svn: 72130
2009-05-19 20:35:26 +00:00
Dan Gohman
be817680b8 Make SCEVCallbackVH a private nested class inside ScalarEvolution, as
it's an implementation detail.

llvm-svn: 72122
2009-05-19 19:22:47 +00:00
Dan Gohman
f61ddb4c06 Create ConstantExpr GEPs the correct way. This fixes
MultiSource/Benchmarks/Prolangs-C/football and a variety of other
failures.

llvm-svn: 72120
2009-05-19 19:18:01 +00:00
Bob Wilson
5e7a2b4ff8 Follow up on new support for memory operands in ARM inline assembly.
This fixes pr4233.

llvm-svn: 72115
2009-05-19 18:33:02 +00:00
Bill Wendling
59276ca9d4 Forward decls.
llvm-svn: 72114
2009-05-19 17:52:31 +00:00
Bill Wendling
5bc220b4eb Remove 'class' specifier.
llvm-svn: 72113
2009-05-19 17:51:26 +00:00
Bill Wendling
2914f42cce Fix to compile on VS2008.
llvm-svn: 72112
2009-05-19 17:50:18 +00:00
Bob Wilson
c6726ecca5 Fix pr4058 and pr4059. Do not split i64 or double arguments between r3 and
the stack.  Patch by Sandeep Patel.

llvm-svn: 72106
2009-05-19 10:02:36 +00:00
Bob Wilson
ec676a76e7 Fix pr4091: Add support for "m" constraint in ARM inline assembly.
llvm-svn: 72105
2009-05-19 05:53:42 +00:00
Dan Gohman
1ef283d538 Remove the #ifndef NDEBUG from the FastISel debugging options. This
fixes dejagnu tests that use these options.

llvm-svn: 72094
2009-05-19 02:19:57 +00:00
Dan Gohman
922033d119 Teach SCEVExpander to expand arithmetic involving pointers into GEP
instructions. It attempts to create high-level multi-operand GEPs,
though in cases where this isn't possible it falls back to casting
the pointer to i8* and emitting a GEP with that. Using GEP instructions
instead of ptrtoint+arithmetic+inttoptr helps pointer analyses that
don't use ScalarEvolution, such as BasicAliasAnalysis.

Also, make the AddrModeMatcher more aggressive in handling GEPs.
Previously it assumed that operand 0 of a GEP would require a register
in almost all cases. It now does extra checking and can do more
matching if operand 0 of the GEP is foldable. This fixes a problem
that was exposed by SCEVExpander using GEPs.

llvm-svn: 72093
2009-05-19 02:15:55 +00:00
Cristian Cadar
9c425ac042 Test commit.
llvm-svn: 72092
2009-05-19 01:56:08 +00:00
Dan Gohman
aa59d257a4 Add some comments noting how IndVarSimplify facilitates
LoopInfo functionality.

llvm-svn: 72091
2009-05-19 01:52:46 +00:00
Owen Anderson
de192dfaee Make using an unsupported CAS size a hard error on Windows.
llvm-svn: 72088
2009-05-19 01:08:36 +00:00
Owen Anderson
421e290a35 Fix up the Windows portion of Atomic.h. This is untested, but it is my best understanding of what should work.
I'd be much obliged if someone on MSVC++ could try this out and let me know if it works.

llvm-svn: 72087
2009-05-19 01:07:40 +00:00
Owen Anderson
e4233e3349 Template CompareAndSwap function.
llvm-svn: 72086
2009-05-19 01:02:27 +00:00
Dale Johannesen
b9a41be8b0 Spacing fix.
llvm-svn: 72083
2009-05-19 00:46:42 +00:00
Mike Stump
2eb592f2f9 Fix cmake builds.
llvm-svn: 72078
2009-05-19 00:18:14 +00:00
Owen Anderson
4d289e72a2 Now that we have atomics support properly detected by configure,
use it to implement Atomic.h.

This expunges the code previously imported from libatomic_ops.

llvm-svn: 72077
2009-05-19 00:08:29 +00:00
Owen Anderson
992fe80601 Test for the presence of GCC atomic builtins at configure time. If not found,
disable building LLVM in thread-safe mode and print a nice warning.

Regenerate configure for these changes.

llvm-svn: 72075
2009-05-18 23:58:51 +00:00
Daniel Dunbar
64a0c22154 Add -submit-aux option to NewNightlyTest.pl
- If given, the argument will be run using system with the path to the sent
   data. Useful for testing nightlytest server replacements.

llvm-svn: 72070
2009-05-18 23:24:26 +00:00
Bill Wendling
a443337af0 Don't set the "location" information for inlined functions' variables.
llvm-svn: 72064
2009-05-18 23:08:55 +00:00
Bill Wendling
8e6fe772be Small code cleanup.
llvm-svn: 72057
2009-05-18 22:33:01 +00:00
Dan Gohman
e3bfb41111 Update a comment to reflect changes in the surrounding code.
llvm-svn: 72051
2009-05-18 22:15:25 +00:00
Evan Cheng
96674b2285 Add short descriptions of 'implicit' and 'parallel'.
llvm-svn: 72050
2009-05-18 22:14:45 +00:00
Bill Wendling
45cb84b4e3 RecordVariable is called each time a DECLARE node is encountered. For an inlined
function, this could be many, many times. We don't want to re-add variables to
that DIE for each time. We just want to add them once. Check to make sure that
we haven't added them already.

llvm-svn: 72047
2009-05-18 22:02:36 +00:00
Dale Johannesen
a0756109d8 Add OpSize to 16-bit ADC and SBB.
llvm-svn: 72045
2009-05-18 21:41:59 +00:00
Argyrios Kyrtzidis
ce3bd2404a Add missing file.
llvm-svn: 72042
2009-05-18 21:08:45 +00:00
Argyrios Kyrtzidis
422f3a0037 Allow the JIT ExecutionEngine to report details about the generated machine code.
Introduce a new class (MachineCodeInfo) that the JIT can fill in with details. Right now, just the address and the size of the machine code are reported.

Patch by Evan Phoenix!

llvm-svn: 72040
2009-05-18 21:06:40 +00:00
Bob Wilson
cc8d63fdef Fix CodePlacementOpt::OptimizeIntraLoopEdges so that its return value
correctly indicates whether it changed the code.

llvm-svn: 72038
2009-05-18 21:02:18 +00:00
Bob Wilson
51e9d81e72 Fix pr4202: Disable CodePlacementOpt for ARM. The ARMConstantIslandPass has
to run last because it needs to know the exact size and position of every
basic block.  Currently CodePlacementOpt is set up to run last.  It might be
worthwhile to investigate reordering these passes, but for now, let's just
make it work.

llvm-svn: 72037
2009-05-18 20:55:32 +00:00
Lang Hames
079c2769b8 New Spiller interface and trivial implementation.
llvm-svn: 72030
2009-05-18 19:03:16 +00:00
Dan Gohman
a21c85512e Revert r72025. It is possible for clients to convert between signed types
and pointer types safely if they only do so when the sizes are the same.
llvm-gcc is such a client.

llvm-svn: 72029
2009-05-18 18:55:39 +00:00
Eric Christopher
1b5625d027 Remove getType() overrides for ExtractValueInst and InsertValueInst.
Patch by John McCall.

llvm-svn: 72028
2009-05-18 18:46:57 +00:00
Bill Wendling
af2e508015 Revert last commit. It was wrong.
llvm-svn: 72026
2009-05-18 18:21:03 +00:00
Dan Gohman
20650de700 Add assertions to CastInst::getCastOpcode to catch attempted conversions
between integers and pointers when the source type is marked signed,
since inttoptr and ptrtoint always use zero-extension when the destination
is larger than the source.

llvm-svn: 72025
2009-05-18 18:18:57 +00:00
Bill Wendling
3f41e5ca13 Don't call RegionInlinedFnEnd if our optimization level isn't -O0.
llvm-svn: 72024
2009-05-18 18:17:22 +00:00
Bill Wendling
ae8a483328 Commands beginning with '--' are converted to '-f' by gcc. Blech!
llvm-svn: 72023
2009-05-18 18:09:36 +00:00
Dale Johannesen
6efc155312 Fill in the missing patterns for ADC and SBB.
Some comment cleanup.

llvm-svn: 72022
2009-05-18 17:44:15 +00:00
Douglas Gregor
1ad25cdc99 termios.h contains the winsize structure we need to determine the
width of a terminal. Don't try to get the width of a terminal if we
don't have this header.

llvm-svn: 72018
2009-05-18 17:21:34 +00:00
Dan Gohman
0a71520ce1 Rename UseTy to AccessTy, for consistency with getAccessType, and to
avoid ambiguity with the word "use" in IVStrideUse.

llvm-svn: 72012
2009-05-18 16:45:28 +00:00
Daniel Dunbar
7c5aee4f79 Silence Release-Asserts warnings.
llvm-svn: 72011
2009-05-18 16:43:04 +00:00
Dan Gohman
592d65ba06 Teach ScalarEvolution to recognize x^-1 in the case where non-demanded
bits have been stripped out by instcombine.

llvm-svn: 72010
2009-05-18 16:29:04 +00:00
Dan Gohman
09726e69b9 Delete a redundant 'else'.
llvm-svn: 72009
2009-05-18 16:17:44 +00:00
Dan Gohman
e50a960c5a Fix ScalarEvolution::isLoopGuardedByCond to accept a null Loop*, for
consistency with other routines that use a null Loop* to mean code
not contained by any loop.

llvm-svn: 72008
2009-05-18 16:03:58 +00:00
Dan Gohman
153b5acb45 Minor code cleanups. Do more of the work before the if statements
instead of within their controlling expressions.

llvm-svn: 72007
2009-05-18 15:58:39 +00:00
Dan Gohman
b24fd61296 Add assertion checks to the SCEV operator creation methods to catch
type mismatches.

llvm-svn: 72006
2009-05-18 15:44:58 +00:00
Dan Gohman
2c9bd7e0cb Make ScalarEvolution::isLoopGuardedByCond work even when the edge
entering a loop is a non-split critical edge.

llvm-svn: 72004
2009-05-18 15:36:09 +00:00