Commit Graph

3810 Commits

Author SHA1 Message Date
Evan Cheng
27af5c414b Only remove the edge from entry to false if false block is merged.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37503 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-07 22:31:28 +00:00
Dale Johannesen
62000eee7d Do not change the size of function arguments. PR 1489.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37496 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-07 21:07:15 +00:00
Evan Cheng
f476961ae6 ifcvt a triangle: don't merge ifcvt block with rejoin block if it can fall through to it. If merged, the resulting block is not a candidate for iterative ifcvting since it contains both predicated and non-predicated code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37487 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-07 08:13:00 +00:00
Evan Cheng
d4de6d91b2 Lots of bug fixes. Now finally in a reasonable state.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37485 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-07 02:12:15 +00:00
Owen Anderson
96dd9a8b1b Quick patch to fix the build, based on what it appears Evan meant to write.
Evan, please check that this is in fact correct.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37471 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 16:22:00 +00:00
Duncan Sands
b32edb4b22 Fold the exception actions table harder: if two typeid lists start the
same, only output one copy of the common part.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37470 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 15:37:31 +00:00
Evan Cheng
7a655479ae Lots of bug fixes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37467 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 10:16:17 +00:00
Duncan Sands
9fac0b5ff2 Additional fix for PR1422: make sure the landing pad label is placed in the
correct machine basic block - do not rely on the eh.exception intrinsic
being in the landing pad: the loop optimizers can move it out.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37463 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 10:05:18 +00:00
Evan Cheng
3d6f60ec19 If a unconditional branch is added to branch to the false path during ifcvt, the predicated block cannot be iteratively ifcvted.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37456 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 02:08:52 +00:00
Evan Cheng
cb78d67a1a Minor statistics counting bug.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37451 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 01:12:44 +00:00
Evan Cheng
e705213b67 Fix a couple of typos and be smarter about order of blocks when ifcvt a diamond.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37449 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-06 00:57:55 +00:00
Evan Cheng
993fc9594e Fix diamond shape ifcvt bugs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37444 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 23:46:14 +00:00
Evan Cheng
a1a9f408da ReplaceUsesOfBlockWith() can modify the predecessors list.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37441 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 22:03:53 +00:00
Evan Cheng
e004317c66 Do not ifcvt if either true / false path is a backedge. Not profitable in almost all cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37440 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 20:38:42 +00:00
Evan Cheng
c53ef58a35 I had a senior moment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37433 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 07:05:25 +00:00
Evan Cheng
8ed680cce3 If the predicated block requires an early exit, end the block there and add a unconditional branch to false block. AnalyzeBranch() does not understand early exits.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37430 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 01:31:40 +00:00
Evan Cheng
f5305f9cc7 Fix some subtle bugs: bug during succeessor copying; incorrectly updating states of ifcvted blocks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37429 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-05 00:07:37 +00:00
Dale Johannesen
fe7e397100 Tail merging wasn't working for predecessors of landing pads. PR 1496.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37427 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 23:52:54 +00:00
Evan Cheng
916d07cdfa Global ctors / dtors alignment shouldn't be hard-coded at 4. e.g. It could be 8 for 64-bit targets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37421 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 20:39:18 +00:00
Evan Cheng
3ec425470f Forgot to check for if iterator reached the end.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37420 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 20:33:36 +00:00
Dan Gohman
3983358555 Pass the DAG to SDNode::dump to let it do more detailed dumps in some cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37413 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 16:17:33 +00:00
Dan Gohman
575e2f4ea8 Resolve implicit alignment before computing the FoldingSet information so
that the CSE map always contains explicit alignment information. This allows
more loads to be CSE'd when there is a mix of explicit-alignment loads and
implicit-alignment loads.

Also, in SelectionDAG::FindModifiedNodeSlot, add the operands to the
FoldingSetNodeID before the load/store information instead of after, so
that it matches what is done elsewhere.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37411 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 15:49:41 +00:00
Evan Cheng
b6665f62c5 Let IfConverter loose. Allow more aggressive subsumptions; reorder basic blocks to expose more ifcvt opportunities; code clean up and fixes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37409 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 06:47:22 +00:00
Evan Cheng
0370fad74b Move ReplaceUsesOfBlockWith() out of BranchFolding into a MachineBasicBlock general facility.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37408 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-04 06:44:01 +00:00
Duncan Sands
09085fe63d The semantics of invoke require that we always jump to the unwind block
(landing pad) when an exception unwinds through the call.  This doesn't
quite match the way the dwarf unwinder works: by default it only jumps to
the landing pad if the catch or filter specification matches, and otherwise
it keeps on unwinding.  There are two ways of specifying to the unwinder
that it should "always" (more on why there are quotes here later) jump to
the landing pad: follow the specification by a 0 typeid, or follow it by
the typeid for the NULL typeinfo.  GCC does the first, and this patch makes
LLVM do the same as gcc.  However there is a problem: the unwinder performs
optimizations based on C++ semantics (it only expects destructors to be
run if the 0 typeid fires - known as "cleanups"), meaning it assumes that no
exceptions will be raised and that the raised exception will be reraised
at the end of the cleanup code.  So if someone writes their own LLVM code
using the exception intrinsics they will get a nasty surprise if they don't
follow these rules.  The other possibility of using the typeid corresponding
to NULL (catch-all) causes the unwinder to make no assumptions, so this is
probably what we should use in the long-run.  However since we are still
having trouble getting exception handling working properly, for the moment
it seems best to closely imitate GCC.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37399 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-02 17:16:06 +00:00
Duncan Sands
73ef58ab92 Integrate exception filter support and exception catch support. This
simplifies the code in DwarfWriter, allows for multiple filters and
makes it trivial to specify filters accompanied by cleanups or catch-all
specifications (see next patch).  What a deal!  Patch blessed by Anton.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37398 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-02 16:53:42 +00:00
Zhou Sheng
02031c0ff8 Make LowerCTPOP() support arbitrary bitwidth integer type.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37397 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-02 04:10:33 +00:00
Dale Johannesen
8e63bf375d Fix CorrectExtraCFGEdges to allow for multiple LandingPad targets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37394 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-02 00:08:15 +00:00
Dale Johannesen
b468163869 Fancier algorithm in tail-merge comment implemented, so remove comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37393 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 23:04:28 +00:00
Dale Johannesen
a5a2117a46 Implement smarter algorithm for choosing which blocks to tail-merge.
See test/CodeGen/X86/test-pic-jtbl.ll for a case where it works well;
shaves another 10K off our favorite benchmark.  I was hesitant about
this because of compile speed, but seems to do OK on a bootstrap.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37392 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 23:02:45 +00:00
Evan Cheng
b5a0690788 Correctly mark early-exit on the false path.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37387 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 20:29:21 +00:00
Duncan Sands
398898cc6a Since TypeInfos are passed as i8 pointers, a NULL TypeInfo should be passed
as a null i8 pointer not as a 0 i32.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37383 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 08:18:30 +00:00
Evan Cheng
8c52938e44 Ifcvt triangle: don't ifcvt 'true' BB if it has other predecessors; don't merge 'false' BB if it has other predecessors.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37382 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 07:41:07 +00:00
Evan Cheng
fe57a7e4df Remove a bogus check. Even terminators in a ifcvt need to be predicated. Unconditional branches can usually be converted to conditional ones.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37380 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 00:55:26 +00:00
Evan Cheng
5f70218c75 Allow multiple ifcvt candidates to share children blocks; add some debugging code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37379 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-01 00:12:12 +00:00
Dale Johannesen
a52dd15137 Arrange for only 1 of multiple branches to landing pad to be kept.
Do not remove empty landing pads (EH table needs to be updated)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37375 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-31 21:54:00 +00:00
Evan Cheng
f15d44cc10 Fix a typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37374 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-31 20:53:33 +00:00
Chris Lattner
3a4205367d Fix the asmprinter so that a globalvalue can specify an explicit alignment
smaller than the preferred alignment, but so that the target can actually
specify a minimum alignment if needed.  This fixes some objc protocol
failures Devang tracked down.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37373 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-31 18:57:45 +00:00
Lauro Ramos Venancio
8d410b6938 Fix PR1424.
When a function has FP, the register scavenging spill slot offset already
was calculated.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37371 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-31 18:27:58 +00:00
Evan Cheng
8258210ed4 Change traversal order to bottom up in preparation for more aggressive if-conversion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37365 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-30 19:49:19 +00:00
Chris Lattner
4e7e6cd13a Fix CodeGen/PowerPC/2007-05-30-dagcombine-miscomp.ll, and PR1473.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37362 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-30 16:30:06 +00:00
Dale Johannesen
44008c5918 Changed per review comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37355 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-30 00:32:01 +00:00
Dale Johannesen
95ef406e0f Make stable_sort in tail merging actually be stable (it never was, but didn't
matter until my last change).  Reenable tail merging by default.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37354 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-29 23:47:50 +00:00
Evan Cheng
58fbb9f5ba Don't merge in tail block of a diamond if it has more than one predecessors after if-conversion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37353 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-29 23:37:20 +00:00
Evan Cheng
c8ed9badfa If there is an empty block between a source and its successor block, it still requires a unconditional branch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37344 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-29 22:31:16 +00:00
Evan Cheng
f277ee4be7 Add missing const qualifiers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37341 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-29 18:35:22 +00:00
Zhou Sheng
924361aef9 Correct the logic in LowerPartSet which cleared the bits from 0 to low-1.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37331 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-26 03:43:13 +00:00
Chris Lattner
3adf951364 tighten up recursion depth again
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37330 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-25 02:19:06 +00:00
Evan Cheng
4bec8ae694 Silly boog.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37328 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-25 00:59:01 +00:00
Dale Johannesen
035fdeb370 Blocks that cond-br and uncond-br/fallthrough to same block should have
only one successor.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37324 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-24 18:31:55 +00:00