1bfe55a378
Use ArrayRef instead of 'const std::vector' to pass around the list of basic blocks to extract.
...
llvm-svn: 140168
2011-09-20 18:42:07 +00:00
add1f17575
Update GCOVLines to provide interfaces to write line table and calculate complete length.
...
llvm-svn: 140167
2011-09-20 18:35:00 +00:00
9a2ba72c49
Fix comments.
...
llvm-svn: 140164
2011-09-20 18:24:46 +00:00
1a155a8200
Update comment.
...
llvm-svn: 140156
2011-09-20 18:05:45 +00:00
9cb1fc034b
Use StringRef instead of std::string.
...
llvm-svn: 140154
2011-09-20 17:55:19 +00:00
972df96ab1
Eliminate unnecessary copy of FileName from GCOVLines.
...
GCOVLines is always accessed through a StringMap where the key is FileName.
llvm-svn: 140151
2011-09-20 17:43:14 +00:00
b011105d6c
There is no need to write a local utility routine to find subprogram info if the utility routine is already available in DebugInfo.
...
llvm-svn: 140145
2011-09-20 15:57:19 +00:00
7cdaa3a1a8
Revert r140083 and r140084 until buildbots can be fixed.
...
llvm-svn: 140094
2011-09-19 23:30:41 +00:00
d3c9d971e6
If we are extracting a basic block that ends in an invoke call, we must also
...
extract the landing pad block. Otherwise, there will be a situation where the
invoke's unwind edge lands on a non-landing pad.
We also forbid the user from extracting the landing pad block by itself. Again,
this is not a valid transformation.
llvm-svn: 140083
2011-09-19 23:00:52 +00:00
61d7c8a065
Fix an infinite loop where a transform in InstCombiner::visitAnd claims a construct is changed when it is not. (See included testcase.)
...
Patch by Xiaoyi Guo.
llvm-svn: 140072
2011-09-19 21:58:15 +00:00
7251e41b16
[indvars] Fix PR10946: SCEV cannot handle Vector IVs.
...
llvm-svn: 140026
2011-09-19 17:54:39 +00:00
74111ee07f
Reapply r139759. Disable IV rewriting by default. See PR10916.
...
llvm-svn: 139842
2011-09-15 20:58:37 +00:00
888bea0b95
Make demanded-elt simplification for shufflevector slightly stronger. Spotted by inspection.
...
llvm-svn: 139768
2011-09-15 01:14:29 +00:00
fca43c21c3
Don't mark objc_retainBlock as nounwind. It calls user copy constructors
...
which could theoretically throw.
llvm-svn: 139710
2011-09-14 18:33:34 +00:00
d4b5e3a4d9
objc_retainBlock is not NoModRef because it can update forwarding pointers
...
in memory relevant to the optimizer. rdar://10050579.
llvm-svn: 139708
2011-09-14 18:13:00 +00:00
f9f68b816b
[indvars] Revert r139579 until 401.bzip -arch i386 miscompilation is fixed. PR10920.
...
llvm-svn: 139583
2011-09-13 05:23:49 +00:00
061d811c51
Disable IV rewriting by default. See PR10916.
...
llvm-svn: 139579
2011-09-13 03:23:21 +00:00
3de5b8e4c1
[indvars] Fix bugs in floating point IV range checks noticed by inspection.
...
llvm-svn: 139574
2011-09-13 01:59:32 +00:00
72a93e5e9b
Add comment to clarify the behavior of a helper in DSE.
...
llvm-svn: 139571
2011-09-13 01:28:59 +00:00
a93ab13e0b
Correct grammar.
...
llvm-svn: 139565
2011-09-13 00:44:16 +00:00
7c5dc122a0
Change a bunch of isVolatile() checks to check for atomic load/store as well.
...
No tests; these changes aren't really interesting in the sense that the logic is the same for volatile and atomic.
I believe this completes all of the changes necessary for the optimizer to handle loads and stores correctly. I'm going to try and come up with some additional testing, though.
llvm-svn: 139533
2011-09-12 20:23:13 +00:00
183013d8d4
Rename -disable-iv-rewrite to -enable-iv-rewrite=false in preparation for default change.
...
llvm-svn: 139517
2011-09-12 18:28:44 +00:00
c7868bf064
[disable-iv-rewrite] Allow WidenIV to handle NSW/NUW operations
...
better.
Don't immediately give up when an add operation can't be trivially
sign/zero-extended within a loop. If it has NSW/NUW flags, generate a
new expression with sign extended (non-recurrent) operand. As before,
if SCEV says that all sign extends are loop invariant, then we can
widen the operation.
llvm-svn: 139453
2011-09-10 01:24:17 +00:00
465f42ff67
Comment formatting.
...
llvm-svn: 139375
2011-09-09 17:35:10 +00:00
1eee7f1242
Add -verify-indvars for imperfect SCEV trip count verification after indvars.
...
llvm-svn: 139169
2011-09-06 20:20:38 +00:00
c10e52a0c4
Use IRBuilder.
...
llvm-svn: 139156
2011-09-06 18:49:53 +00:00
58704ee442
Try again at r138809 (make DSE more aggressive in removing dead stores at the end of a function), now with less deleting stores before memcpy's.
...
llvm-svn: 139150
2011-09-06 18:14:09 +00:00
a098436b32
Split the init.trampoline intrinsic, which currently combines GCC's
...
init.trampoline and adjust.trampoline intrinsics, into two intrinsics
like in GCC. While having one combined intrinsic is tempting, it is
not natural because typically the trampoline initialization needs to
be done in one function, and the result of adjust trampoline is needed
in a different (nested) function. To get around this llvm-gcc hacks the
nested function lowering code to insert an additional parent variable
holding the adjust.trampoline result that can be accessed from the child
function. Dragonegg doesn't have the luxury of tweaking GCC code, so it
stored the result of adjust.trampoline in the memory GCC set aside for
the trampoline itself (this is always available in the child function),
and set up some new memory (using an alloca) to hold the trampoline.
Unfortunately this breaks Go which allocates trampoline memory on the
heap and wants to use it even after the parent has exited (!). Rather
than doing even more hacks to get Go working, it seemed best to just use
two intrinsics like in GCC. Patch mostly by Sanjoy Das.
llvm-svn: 139140
2011-09-06 13:37:06 +00:00
29192d042e
Delete trivial landing pads that just continue unwinding the caught
...
exception.
llvm-svn: 139117
2011-09-05 12:57:57 +00:00
321fb37773
Use Duncan's patch to delete the instructions in reverse order (minus the landingpad and terminator).
...
llvm-svn: 139090
2011-09-04 09:43:36 +00:00
a336e70573
Update comments to reflect reality.
...
llvm-svn: 139023
2011-09-02 18:43:33 +00:00
31b941a60d
Enable SCEV-based unrolling by default.
...
This changes loop unrolling to use the same mechanism for trip count
computation as indvars. This is a stronger check that tends to unroll
more loops. A very common side-effect is that many single iteration
loops will be removed sooner. The real goal was simply to remove
dependence on canonical IVs.
x86 is break even.
ARM performance changes to expect (+ is good):
External/SPEC/CFP2000/183.equake/183.equake +13%
SingleSource/Benchmarks/Dhrystone/fldry +21%
MultiSource/Applications/spiff/spiff +3%
SingleSource/Benchmarks/Stanford/Puzzle -14%
The Puzzle regression is actually an improvement in loop optimization
that defeats GVN: rdar://problem/10065079.
llvm-svn: 139009
2011-09-02 17:26:28 +00:00
7470fb01d0
Compare type size instead of type _store_ size to make sure that BitCastInst
...
will be valid. This fixes PR10820.
llvm-svn: 139005
2011-09-02 14:57:37 +00:00
a3ba6d3b80
Reduce indentation. No functionality change.
...
llvm-svn: 138968
2011-09-01 21:29:49 +00:00
bf8280ff27
Change worklist driven deletion to be an iterative process.
...
Duncan noticed this!
llvm-svn: 138967
2011-09-01 21:28:33 +00:00
71f5c2f158
Fix an issue with the IR sink pass found by inspection. (I'm not sure anyone is actually using this, but might as well fix it since I found the issue.)
...
llvm-svn: 138965
2011-09-01 21:21:24 +00:00
a617c32745
Resubmit with fix. Properly remove the instructions except for landingpad, which should be removed only when its invokes are.
...
llvm-svn: 138932
2011-09-01 01:28:11 +00:00
9f7cf20e60
Submitted this too early.
...
llvm-svn: 138931
2011-09-01 01:18:33 +00:00
2d1f11f743
Don't DCE the landingpad instruction.
...
The landingpad instruction can be removed only when its invokes are removed.
llvm-svn: 138930
2011-09-01 01:16:58 +00:00
770d0f0700
Make sure we aren't deleting the landingpad instruction.
...
The landingpad instruction is required in the landing pad block. Because we're
not deleting terminating instructions, the invoke may still jump to here (see
Transforms/SCCP/2004-11-16-DeadInvoke.ll). Remove all uses of the landingpad
instruction, but keep it around until code-gen can remove the basic block.
llvm-svn: 138890
2011-08-31 20:55:20 +00:00
a45c20b049
Remove the old tail duplication pass. It is not used and is unable to update
...
ssa, so it has to be run really early in the pipeline. Any replacement
should probably use the SSAUpdater.
llvm-svn: 138841
2011-08-30 23:03:45 +00:00
e316e5b2ad
Speculatively revert r138809 in an attempt to fix DragonEgg.
...
llvm-svn: 138829
2011-08-30 21:11:06 +00:00
d708ec4c6a
When walking backwards to eliminate final stores to allocas at the end of a function, encountering an unrelated store should not cause us to give up like encountering a load does.
...
llvm-svn: 138809
2011-08-30 18:51:55 +00:00
5fc81ffbac
Fixes following the CR by Chris and Duncan:
...
Optimize chained bitcasts of the form A->B->A.
Undo r138722 and change isEliminableCastPair to allow this case.
llvm-svn: 138756
2011-08-29 19:58:36 +00:00
52600ee8c3
Bitcasts are transitive. Bitcast-Bitcast-X becomes Bitcast-X.
...
llvm-svn: 138722
2011-08-28 11:51:08 +00:00
eed1e8905a
Don't sink landingpad instructions during ind-var simplification.
...
llvm-svn: 138651
2011-08-26 20:40:15 +00:00
0655b78ccc
Address review comments.
...
- Reword comments.
- Allow undefined behavior interfering with undefined behavior.
- Add address space checks.
llvm-svn: 138619
2011-08-26 02:25:55 +00:00
fb212a6309
SimplifyCFG: If we have a PHI node that can evaluate to NULL and do a load or store to the address returned by the PHI node then we can consider this incoming value as dead and remove the edge pointing there, unless there are instructions that can affect control flow executed in between.
...
In theory this could be extended to other instructions, eg. division by zero, but it's likely that it will "miscompile" some code because people depend on div by zero not trapping. NULL pointer dereference usually leads to a crash so we should be on the safe side.
This shrinks the size of a Release clang by 16k on x86_64.
llvm-svn: 138618
2011-08-26 01:22:29 +00:00
3fb137f7ef
LSR wants to split the landing pad's critical edge. Let it do it, but use the
...
proper function to do it.
llvm-svn: 138550
2011-08-25 05:55:40 +00:00
07efd6f1e0
When inserting new instructions, use getFirstInsertionPt instead of
...
getFirstNonPHI so that it will skip over the landingpad instructions as well.
llvm-svn: 138537
2011-08-25 01:08:34 +00:00
86c5cbe613
Skip the landingpad instruction when determining the insertion point.
...
llvm-svn: 138481
2011-08-24 21:06:46 +00:00
0902a68f69
Use getFirstInsertionPt instead of getFirstNonPHI so that it skips to the proper
...
insertion place.
llvm-svn: 138473
2011-08-24 20:28:43 +00:00
d3e65e702f
Fix a crashing bug in SplitBlock when it is called on a block with no
...
dominator information even though dominators were previously computed.
Patch by Nick Sumner.
llvm-svn: 138449
2011-08-24 18:07:01 +00:00
4b8e8ce37f
Add a comment.
...
llvm-svn: 138243
2011-08-22 17:29:37 +00:00
56e1cef705
Constant pointers to objects don't need reference counting.
...
llvm-svn: 138242
2011-08-22 17:29:11 +00:00
38d813087e
If we're splitting the landing pad block and assigning it only one predecessor,
...
then don't split it a second time, since that block will be dead.
llvm-svn: 138153
2011-08-19 23:46:30 +00:00
26e19288be
The landingpad instruction isn't dead simply because it's value isn't used.
...
llvm-svn: 138102
2011-08-19 21:52:06 +00:00
4938edb02c
Make a bunch of symbols private.
...
llvm-svn: 138025
2011-08-19 01:42:18 +00:00
5a656883b1
C API functions must be able to see their extern "C" definitions, or it will be impossible to call them from C.
...
llvm-svn: 138022
2011-08-19 01:36:54 +00:00
b38940135b
Track a retain+release nesting level independently of the
...
known-incremented level, because the two concepts can be used
to prove the saftey of a retain+release removal in different
ways.
llvm-svn: 138016
2011-08-19 00:26:36 +00:00
c61f7659ba
Intelligently split the landing pad block.
...
We have to be careful when splitting the landing pad block, because the
landingpad instruction is required to remain as the first non-PHI of an invoke's
unwind edge. To retain this, we split the block into two blocks, moving the
predecessors within the loop to one block and the remaining predecessors to the
other. The landingpad instruction is cloned into the new blocks.
llvm-svn: 138015
2011-08-19 00:09:22 +00:00
ca7d309623
Add SplitLandingPadPredecessors().
...
SplitLandingPadPredecessors is similar to SplitBlockPredecessors in that it
splits the current block and attaches a set of predecessors to the new basic
block. However, it differs from SplitBlockPredecessors in that it's specifically
designed to handle landing pad blocks.
Two new basic blocks are created: one that is has the vector of predecessors as
its predecessors and one that has the remaining predecessors as its
predecessors. Those two new blocks then receive a cloned copy of the landingpad
instruction from the original block. The landingpad instructions are joined in a
PHI, etc. Like SplitBlockPredecessors, it updates the LLVM IR, AliasAnalysis,
DominatorTree, DominanceFrontier, LoopInfo, and LCCSA analyses.
llvm-svn: 138014
2011-08-19 00:05:40 +00:00
2b31c45e8e
Use 'getFirstInsertionPt' when trying to insert new instructions during LICM.
...
llvm-svn: 138008
2011-08-18 23:42:36 +00:00
c57b58cc40
Make it clear that this code is iterating in reverse order through the array.
...
llvm-svn: 137985
2011-08-18 21:27:42 +00:00
b15d6eb93b
Revert r137871. The loop simplify pass should require all exits from a loop that
...
aren't from an indirect branch need to be dominated by the loop header.
llvm-svn: 137981
2011-08-18 21:10:01 +00:00
b267e2a7ec
Split out the updating of PHI nodes after splitting the BB into a separate
...
function.
llvm-svn: 137979
2011-08-18 20:51:04 +00:00
ec3823dcb7
Use this fantzy ArrayRef thing to pass in the list of predecessors.
...
llvm-svn: 137978
2011-08-18 20:39:32 +00:00
74acf9f501
The edge from DISubprogram to DICompileUnit has been removed in recent versions
...
of debug info.
llvm-svn: 137972
2011-08-18 19:07:42 +00:00
6029135af9
Use static instead of anonymous namespace.
...
llvm-svn: 137959
2011-08-18 17:57:57 +00:00
0a693f47ee
Split out the analysis updating code into a helper function. No intended
...
functionality change.
llvm-svn: 137926
2011-08-18 05:25:23 +00:00
53771ba07c
Dramatically speedup codegen prepare by a) avoiding use of dominator tree and b) doing a separate pass over dbg.value instructions.
...
llvm-svn: 137908
2011-08-18 00:50:51 +00:00
2b21d86cfe
Do not use DebugInfoFinder. Extract debug info directly from llvm.dbg.cu named mdnode.
...
llvm-svn: 137890
2011-08-17 22:49:38 +00:00
9a468153e1
Atomic load/store handling for the passes using memdep (GVN, DSE, memcpyopt).
...
llvm-svn: 137888
2011-08-17 22:22:24 +00:00
8bbcbedeaf
Disable PRE for landing pads.
...
PRE needs the landing pads to have their critical edges split. Doing this for a
landing pad is non-trivial. Abandon the attempt to perform PRE when we come
across a landing pad. (Reviewed by Owen!)
llvm-svn: 137876
2011-08-17 21:32:02 +00:00
79a6873d9c
Increment the insertion iterator to beyond the landingpad instruction.
...
llvm-svn: 137872
2011-08-17 21:21:31 +00:00
39257d6b5c
Don't optimize the landing pad exit block.
...
One way to exit the loop is through an unwind edge. However, that may involve
splitting the critical edge of the landing pad, which is non-trivial. Prevent
the transformation from rewriting the landing pad exit loop block.
llvm-svn: 137871
2011-08-17 21:20:43 +00:00
2dfbcc4506
Assert that we aren't trying to split the critical edge of a landing pad. Doing
...
so requires more care than this generic algorithm should handle.
llvm-svn: 137866
2011-08-17 21:04:05 +00:00
a9ee09f4be
Revert r137655. There is some question about whether the 'landingpad'
...
instruction should be marked as potentially reading and/or writing memory.
llvm-svn: 137863
2011-08-17 20:36:44 +00:00
d7749be2d7
Silly mistake from r137777; restore significant isStructTy() checks. While here, be a bit more defensive
...
with unknown instructions.
Fixes PR10687.
llvm-svn: 137836
2011-08-17 18:10:43 +00:00
0793eb4c46
A bunch of misc fixes to SCCPSolver::ResolvedUndefsIn, including a fix to stop
...
making random bad assumptions about instructions which are not explicitly listed.
Includes fix for rdar://9956541, a version of "undef ^ undef should return
0 because it's easier than arguing with users".
llvm-svn: 137777
2011-08-16 22:06:31 +00:00
56f2f21254
Minor bug in SCCP found by inspection. (I don't think it's possible to hit this with a normal pass pipeline, but fixing for completeness.)
...
llvm-svn: 137755
2011-08-16 21:12:35 +00:00
8ddfc09e7a
Use the getFirstInsertionPt() method instead of getFirstNonPHI + an 'isa<>'
...
check for a LandingPadInst.
llvm-svn: 137745
2011-08-16 20:45:24 +00:00
55d875fa1c
I think there was some confusion about what I meant. :-) Replacing the comment.
...
llvm-svn: 137743
2011-08-16 20:41:17 +00:00
719a72f34c
Add a mechanism for optimisation plugins to register passes that all front ends can use without needing to be aware of the plugin (or the plugin be aware of the front end).
...
Before 3.0, I'd like to add a mechanism for automatically loading a set of plugins from a config file. API suggestions welcome...
llvm-svn: 137717
2011-08-16 13:58:41 +00:00
be33e8d58d
A few places where we want to skip the landingpad instruction for insertion.
...
llvm-svn: 137712
2011-08-16 04:52:55 +00:00
a917d4f9b4
Revert a bit of r137667; the logic in question can safely handle atomic load/store.
...
llvm-svn: 137702
2011-08-16 01:28:22 +00:00
bd39703456
After talking with Bill, it seems like the LandingPad handling here is likely
...
to be wrong (or at least somewhat suspect). Leave a FIXME for Bill.
llvm-svn: 137694
2011-08-16 00:41:37 +00:00
b8f30de527
Minor comment fixes.
...
llvm-svn: 137693
2011-08-16 00:20:11 +00:00
0ffdf2ea0b
Update SimplifyCFG for atomic operations.
...
This commit includes a mention of the landingpad instruction, but it's not
changing the behavior around it. I think the current behavior is correct,
though. Bill, can you double-check that?
llvm-svn: 137691
2011-08-15 23:59:28 +00:00
01a67111d1
Add comments and test for atomic load/store and mem2reg.
...
llvm-svn: 137690
2011-08-15 23:55:52 +00:00
5a18b7c7c7
In places where it's using "getFirstNonPHI", skip the landingpad instruction if necessary.
...
llvm-svn: 137679
2011-08-15 23:19:54 +00:00
91d4e9edec
Don't sink the instruction to before a landingpad instruction.
...
llvm-svn: 137672
2011-08-15 22:53:05 +00:00
211e348eaa
Update inter-procedural optimizations for atomic load/store.
...
llvm-svn: 137667
2011-08-15 22:16:46 +00:00
8bc586e770
Update instcombine for atomic load/store.
...
llvm-svn: 137664
2011-08-15 22:09:40 +00:00
e86965ee19
Duncan pointed out that the LandingPadInst might read memory. (It might also
...
write to memory.) Marking it as such makes some checks for immobility go away.
llvm-svn: 137655
2011-08-15 21:14:31 +00:00
4d05198d1f
Fix llvm::CloneModule to correctly clone globals. Patch per bug report by Simon Moll on llvmdev.
...
llvm-svn: 137654
2011-08-15 21:05:06 +00:00
91386c7be4
Atomic load/store support in LICM.
...
llvm-svn: 137648
2011-08-15 20:52:09 +00:00
d9fb470758
The "landingpad" instruction will never be "trivially" dead.
...
llvm-svn: 137642
2011-08-15 20:10:51 +00:00
dd94d3426b
Don't try to sink the landingpad instruction. It's immobile.
...
llvm-svn: 137629
2011-08-15 18:23:40 +00:00
88294cdbe0
Mark the SCC as "might unwind" if we run into a 'resume' instruction.
...
llvm-svn: 137627
2011-08-15 18:22:00 +00:00
Bill Wendling