59139 Commits

Author SHA1 Message Date
Evan Cheng
499918dabf Revert 100204. It broke a bunch of tests and apparently changed what passes are run during codegen.
llvm-svn: 100207
2010-04-02 19:29:15 +00:00
Benjamin Kramer
69aee9e31b Update CMake.
llvm-svn: 100206
2010-04-02 19:09:51 +00:00
David Greene
554373897c Let's try this again. Re-apply 100143 including an apparent missing
<string> include.  For some reason the buildbot choked on this while my
builds did not.  It's probably due to a difference in system headers.

---

Add some switches helpful for debugging:

-print-before=<Pass Name>

Dump IR before running pass <Pass Name>.

-print-before-all

Dump IR before running each pass.

-print-after-all

Dump IR after running each pass.

These are helpful when tracking down a miscompilation.  It is easy to
get IR dumps and do diffs on them, etc.

To make this work well, add a new getPrinterPass API to Pass so that
each kind of pass (ModulePass, FunctionPass, etc.) can create a Pass
suitable for dumping out the kind of object the Pass works on.

llvm-svn: 100204
2010-04-02 18:46:26 +00:00
Mon P Wang
0ccf050ca3 Revert r100191 since it breaks objc in clang
llvm-svn: 100199
2010-04-02 18:43:02 +00:00
Mon P Wang
a01350755e Reapply address space patch after fixing an issue in MemCopyOptimizer.
Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)

llvm-svn: 100191
2010-04-02 18:04:15 +00:00
Dan Gohman
0b7f8ec7f7 Add an explicit keyword.
llvm-svn: 100187
2010-04-02 14:57:49 +00:00
Dan Gohman
8835a0200a Manually notify ScalarEvolution before making an operand replacement, since
it can't currently observe such changes automatically.

llvm-svn: 100186
2010-04-02 14:48:31 +00:00
Torok Edwin
13048a143f Fix SpecificBumpPtrAllocator iteration.
Need to start from (char*)(Slab+1), and not from (char*)Slab+1.
This fixes crashes in Win64 debug mode.
Thanks to Nicolas Capens!

llvm-svn: 100184
2010-04-02 13:20:51 +00:00
Duncan Sands
7a7fb5d551 Add notes about dragonegg.
llvm-svn: 100183
2010-04-02 09:23:15 +00:00
Bob Wilson
88d149d129 Recommit 100158 now that the buildbots are happy again.
llvm-svn: 100177
2010-04-02 05:09:46 +00:00
Eric Christopher
e6ad109103 Remove FIXME - if there's a better way to do this it isn't here.
llvm-svn: 100176
2010-04-02 04:32:37 +00:00
Dan Gohman
7051600e3f Revert the recent alignment changes. They're broken for -Os because,
in particular, they end up aligning strings at 16-byte boundaries, and
there's no way for GlobalOpt to check OptForSize.

llvm-svn: 100172
2010-04-02 03:04:37 +00:00
Evan Cheng
921fc2c77b After trivial coalescing, the MI being visited may have become a copy. Avoid adding it to CSE hash table since copies aren't being considered for CSE and they may be deleted.
rdar://7819990

llvm-svn: 100170
2010-04-02 02:21:24 +00:00
Dale Johannesen
5b35f2ee86 Teach AnalyzeBranch, RemoveBranch and the branch
folder to be tolerant of debug info following the
branch(es) at the end of a block.

llvm-svn: 100168
2010-04-02 01:38:09 +00:00
Chandler Carruth
db4beb9b04 Disambiguate conditional expression for newer GCCs.
llvm-svn: 100167
2010-04-02 01:31:24 +00:00
Dan Gohman
ec731a99fa Remove this initializer so that the optimizer doesn't convert
unaligned loads into aligned loads.

llvm-svn: 100166
2010-04-02 01:26:13 +00:00
Dan Gohman
d14fe2a5cd Update this test for the new preferred alignment heuristics.
llvm-svn: 100165
2010-04-02 01:24:08 +00:00
Bob Wilson
4b5f895975 Revert 100158 in case it is causing some of the buildbot problems.
llvm-svn: 100164
2010-04-02 01:22:49 +00:00
Sean Callanan
ea4f4644ee Updated the install location for EnhancedDisassembly
on Mac OS X to use @rpath rather than an absolute
path.  Also allowed the version to be set using an
environment variable.

llvm-svn: 100163
2010-04-02 00:53:42 +00:00
Dan Gohman
d12e6c92c5 Change variables which are exactly 16 bytes to be 16-byte-aligned too.
This fixes test/Transforms/GlobalOpt/gv-align.ll.

llvm-svn: 100161
2010-04-02 00:46:07 +00:00
Dan Gohman
ece7e3c015 Make globalopt refine global variable alignment.
llvm-svn: 100160
2010-04-02 00:14:16 +00:00
Bob Wilson
ab18888a5a Check for terminating conditions before adding PHIs to the worklists.
This is more efficient than adding them to the worklist and then ignoring them.

llvm-svn: 100158
2010-04-02 00:10:41 +00:00
Dan Gohman
5437ff1446 If the bitcode reader input stream isn't a multiple of 4 bytes, it's more
likely not a bitcode file at all, rather than being a bitcode file which
is truncated. Check for this case and issue a more relevant error message.

llvm-svn: 100156
2010-04-02 00:03:51 +00:00
Bob Wilson
1b646b39e2 Remove trailing whitespace.
llvm-svn: 100148
2010-04-01 23:06:38 +00:00
Bob Wilson
08cc4030dd Rewrite another SSAUpdater function to avoid recursion.
llvm-svn: 100147
2010-04-01 23:05:58 +00:00
Eric Christopher
77e4cc4bbd Revert r100143.
llvm-svn: 100146
2010-04-01 22:54:42 +00:00
Devang Patel
1d51b57a85 Revert r100117.
llvm-svn: 100145
2010-04-01 22:47:29 +00:00
David Greene
a358be0ef7 Add some switches helpful for debugging:
-print-before=<Pass Name>

Dump IR before running pass <Pass Name>.

-print-before-all

Dump IR before running each pass.

-print-after-all

Dump IR after running each pass.

These are helpful when tracking down a miscompilation.  It is easy to
get IR dumps and do diffs on them, etc.

To make this work well, add a new getPrinterPass API to Pass so that
each kind of pass (ModulePass, FunctionPass, etc.) can create a Pass
suitable for dumping out the kind of object the Pass works on.

llvm-svn: 100143
2010-04-01 22:43:57 +00:00
Evan Cheng
9508456bb6 In 64-bit mode, use i64 to lower memcpy / memset instead of f64.
llvm-svn: 100137
2010-04-01 20:27:45 +00:00
Devang Patel
97a60f2ce5 Do not eagerly record known previous location. DBG_VALUE may not cause a new label due to one or other reason.
llvm-svn: 100134
2010-04-01 20:22:44 +00:00
Evan Cheng
9307af84c2 Skip checking preferred alignment of GVs defined in other translation units all together.
llvm-svn: 100133
2010-04-01 20:13:28 +00:00
Evan Cheng
dc26010cc1 Add comments about DstAlign and SrcAlign.
llvm-svn: 100132
2010-04-01 20:10:42 +00:00
Bob Wilson
bfd85d79c1 Change another SSAUpdater function to avoid recursion.
llvm-svn: 100131
2010-04-01 20:04:30 +00:00
Bob Wilson
39e3a8e9f7 Simplify the code to check for existing PHIs, now that it is only used in
one place.  This removes the template function added in svn 94690.

llvm-svn: 100130
2010-04-01 19:53:48 +00:00
Bob Wilson
205385244c The SSAUpdater should avoid recursive traversals of the CFG, since that may
blow out the stack for really big functions.  Start by fixing an easy case.

llvm-svn: 100126
2010-04-01 18:46:59 +00:00
Evan Cheng
8728924812 - Avoid using floating point stores to implement memset unless the value is zero.
- Do not try to infer GV alignment unless its type is sized. It's not possible to infer alignment if it has opaque type.

llvm-svn: 100118
2010-04-01 18:19:11 +00:00
Devang Patel
52c509b85d Skip instructions until new scope is seen.
llvm-svn: 100117
2010-04-01 17:32:01 +00:00
Devang Patel
e55fdcf347 Cosmetic changes.
Update comment, rename a local variable.

llvm-svn: 100116
2010-04-01 17:16:48 +00:00
Benjamin Kramer
1ef6689d58 Remove accidental include and add a comment.
llvm-svn: 100107
2010-04-01 14:39:55 +00:00
Benjamin Kramer
71eaf27165 Various improvements to MemoryBuffer::getFile:
- Use a RAII object to close the FD.
- Use sys::StrError instead of thread-unsafe strerror calls.
- Recover gracefully if read returns zero. This works around an issue on
  DragonFlyBSD where /dev/null has an st_size of 136 but we can't read 136 bytes
  from it.

llvm-svn: 100106
2010-04-01 14:35:22 +00:00
Evan Cheng
202b6327c4 Add -mcpu to memcpy / memset tests to ensure they behave the same on all hosts / targets.
llvm-svn: 100101
2010-04-01 08:25:26 +00:00
Gabor Greif
0ed92c6e98 Introduce ImmutableCallSite, useful for contexts where no mutation
is necessary. Inherits from new templated baseclass CallSiteBase<>
which is highly customizable. Base CallSite on it too, in a configuration
that allows full mutation.
Adapt some call sites in analyses to employ ImmutableCallSite.

llvm-svn: 100100
2010-04-01 08:21:08 +00:00
Nick Lewycky
2440202f2b Clean up this file a little, no functionality change. This is a subset of my
patch back in r94322.

llvm-svn: 100097
2010-04-01 07:34:00 +00:00
Chris Lattner
b27da4e9b4 switch IRBuilder to use NewDebugLoc for locations instead
of raw mdnodes.  This allows frontends to specify debug
locations without ever creating an MDNode for the DILocation.

This requires a corresponding clang/llvm-gcc change which
I'll try to commit as simultaneously as possible.

llvm-svn: 100095
2010-04-01 06:31:45 +00:00
Chris Lattner
c7c8dc16f1 add comments, don't require inlined-at to be specified.
llvm-svn: 100092
2010-04-01 06:29:56 +00:00
Evan Cheng
562bb43207 Fix sdisel memcpy, memset, memmove lowering:
1. Makes it possible to lower with floating point loads and stores.
2. Avoid unaligned loads / stores unless it's fast.
3. Fix some memcpy lowering logic bug related to when to optimize a
   load from constant string into a constant.
4. Adjust x86 memcpy lowering threshold to make it more sane.
5. Fix x86 target hook so it uses vector and floating point memory
   ops more effectively.
rdar://7774704

llvm-svn: 100090
2010-04-01 06:04:33 +00:00
Evan Cheng
e60ac0f6a6 Nehalem unaligned memory access is fast.
llvm-svn: 100089
2010-04-01 05:58:17 +00:00
Chris Lattner
96eb1c7f8f Switch the representation of the location in instruction from
being a TrackingVH<MDNode> to a NewDebugLoc, shrinking 
sizeof(Instruction) a lot, and providing clients the ability
to deal with locations in terms of NewDebugLoc instead of
having to deal with Metadata.  This is still fully compatible
with all clients that *do* use MDNodes for everything of
course.

No functionality change.

llvm-svn: 100088
2010-04-01 05:23:13 +00:00
Chris Lattner
b77a53bb60 include header.
llvm-svn: 100087
2010-04-01 05:20:21 +00:00
Chris Lattner
b0cddaebe0 rewrite handling of forward ref'd instruction metadata
to used deferred resolution instead of creating a temporary
node + rauw.  There is no reason to create the temporary
mdnode, then do rauw, then destroy it.

llvm-svn: 100086
2010-04-01 05:14:45 +00:00