oota-llvm.git
9 years agoRip out X86-specific vector SDIV lowering, make the corresponding DAGCombiner transfo...
Benjamin Kramer [Sat, 26 Apr 2014 13:00:53 +0000 (13:00 +0000)]
Rip out X86-specific vector SDIV lowering, make the corresponding DAGCombiner transform work on vectors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207316 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDAGCombiner: Turn divs of vector splats into vectorized multiplications.
Benjamin Kramer [Sat, 26 Apr 2014 12:06:28 +0000 (12:06 +0000)]
DAGCombiner: Turn divs of vector splats into vectorized multiplications.

Otherwise the legalizer would just scalarize everything. Support for
mulhi in the targets isn't that great yet so on most targets we get
exactly the same scalarized output. Add a test for x86 vector udiv.

I had to disable the mulhi nodes on ARM because there aren't any patterns
for it. As far as I know ARM has instructions for getting the high part of
a multiply so this should be fixed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207315 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoX86: Custom lower v4i32 UMUL_LOHI into 2 pmuludqs.
Benjamin Kramer [Sat, 26 Apr 2014 12:06:11 +0000 (12:06 +0000)]
X86: Custom lower v4i32 UMUL_LOHI into 2 pmuludqs.

Test will follow soon.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207314 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert r206749 till a final decision about the intrinsics is made.
Michael Zolotukhin [Sat, 26 Apr 2014 09:56:41 +0000 (09:56 +0000)]
Revert r206749 till a final decision about the intrinsics is made.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207313 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Rather than removing nodes from the SCC entry set when we process
Chandler Carruth [Sat, 26 Apr 2014 09:45:55 +0000 (09:45 +0000)]
[LCG] Rather than removing nodes from the SCC entry set when we process
them, just skip over any DFS-numbered nodes when finding the next root
of a DFS. This allows the entry set to just be a vector as we populate
it from a uniqued source. It also removes the possibility for a linear
scan of the entry set to actually do the removal which can make things
go quadratic if we get unlucky.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207312 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Rotate the full SCC finding algorithm to avoid round-trips through
Chandler Carruth [Sat, 26 Apr 2014 09:28:00 +0000 (09:28 +0000)]
[LCG] Rotate the full SCC finding algorithm to avoid round-trips through
the DFS stack for leaves in the call graph. As mentioned in my previous
commit, this is particularly interesting for graphs which have high fan
out but low connectivity resulting in many leaves. For such graphs, this
can remove a large % of the DFS stack traffic even though it doesn't
make the stack much smaller.

It's a bit easier to formulate this for the full algorithm because that
one stops completely for each SCC. For example, I was able to directly
eliminate the "Recurse" boolean used to continue an outer loop from the
inner loop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207311 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Hoist the main DFS loop out of the edge removal function. This
Chandler Carruth [Sat, 26 Apr 2014 09:06:53 +0000 (09:06 +0000)]
[LCG] Hoist the main DFS loop out of the edge removal function. This
makes working through the worklist much cleaner, and makes it possible
to avoid the 'bool-to-continue-the-outer-loop' hack. Not a huge
difference, but I think this is approaching as polished as I can make
it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207310 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRecursivelyDeleteTriviallyDeadInstructions() could remove
Gerolf Hoflehner [Sat, 26 Apr 2014 05:58:11 +0000 (05:58 +0000)]
RecursivelyDeleteTriviallyDeadInstructions() could remove
more than 1 instruction. The caller need to be aware of this
and adjust instruction iterators accordingly.

rdar://16679376

Repaired r207302.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207309 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRestore CloneFunction.cpp which got accidently
Gerolf Hoflehner [Sat, 26 Apr 2014 05:43:41 +0000 (05:43 +0000)]
Restore CloneFunction.cpp which got accidently
overwritten by previous backout of r207303

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207308 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] In the incremental SCC re-formation, lift the node currently being
Chandler Carruth [Sat, 26 Apr 2014 03:36:42 +0000 (03:36 +0000)]
[LCG] In the incremental SCC re-formation, lift the node currently being
processed in the DFS out of the stack completely. Keep it exclusively in
a variable. Re-shuffle some code structure to make this easier. This can
have a very dramatic effect in some cases because call graphs tend to
look like a high fan-out spanning tree. As a consequence, there are
a large number of leaf nodes in the graph, and this technique causes
leaf nodes to never even go into the stack. While this only reduces the
max depth by 1, it may cause the total number of round trips through the
stack to drop by a lot.

Now, most of this isn't really relevant for the incremental version. =]
But I wanted to prototype it first here as this variant is in ways more
complex. As long as I can get the code factored well here, I'll next
make the primary walk look the same. There are several refactorings this
exposes I think.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207306 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Special case the removal of self edges. These don't impact the SCC
Chandler Carruth [Sat, 26 Apr 2014 03:36:37 +0000 (03:36 +0000)]
[LCG] Special case the removal of self edges. These don't impact the SCC
graph in any way because we don't track edges in the SCC graph, just
nodes. This also lets us add a nice assert about the invariant that
we're working on at least a certain number of nodes within the SCC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207305 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[DAG] During DAG legalization keep opaque constants even after expanding.
Juergen Ributzka [Sat, 26 Apr 2014 02:58:04 +0000 (02:58 +0000)]
[DAG] During DAG legalization keep opaque constants even after expanding.

The included test case would return the incorrect results, because the expansion
of an shift with a constant shift amount of 0 would generate undefined behavior.

This is because ExpandShiftByConstant assumes that all shifts by constants with
a value of 0 have already been optimized away. This doesn't happen for opaque
constants and usually this isn't a problem, because opaque constants won't take
this code path - they are not supposed to. In the case that the opaque constant
has to be expanded by the legalizer, the legalizer would drop the opaque flag.
In this case we hit the limitations of ExpandShiftByConstant and create incorrect
code.

This commit fixes the legalizer by not dropping the opaque flag when expanding
opaque constants and adding an assertion to ExpandShiftByConstant to catch this
not supported case in the future.

This fixes <rdar://problem/16718472>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207304 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert commit r207302 since build failures
Gerolf Hoflehner [Sat, 26 Apr 2014 02:03:17 +0000 (02:03 +0000)]
Revert commit r207302 since build failures
have been reported.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207303 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRecursivelyDeleteTriviallyDeadInstructions() could remove
Gerolf Hoflehner [Sat, 26 Apr 2014 01:19:16 +0000 (01:19 +0000)]
RecursivelyDeleteTriviallyDeadInstructions() could remove
more than 1 instruction. The caller need to be aware of this
and adjust instruction iterators accordingly.

rdar://16679376

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207302 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[X86] Implement TargetLowering::getScalingFactorCost hook.
Quentin Colombet [Sat, 26 Apr 2014 01:11:26 +0000 (01:11 +0000)]
[X86] Implement TargetLowering::getScalingFactorCost hook.
Scaling factors are not free on X86 because every "complex" addressing mode
breaks the related instruction into 2 allocations instead of 1.

<rdar://problem/16730541>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207301 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Refactor the duplicated code I added in my last commit here into
Chandler Carruth [Sat, 26 Apr 2014 01:03:46 +0000 (01:03 +0000)]
[LCG] Refactor the duplicated code I added in my last commit here into
a helper function. Also factor the other two places where we did the
same thing into the helper function. =] Much cleaner this way. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207300 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[InstCombine][X86] Teach how to fold calls to SSE2/AVX2 packed logical shift
Andrea Di Biagio [Sat, 26 Apr 2014 01:03:22 +0000 (01:03 +0000)]
[InstCombine][X86] Teach how to fold calls to SSE2/AVX2 packed logical shift
right intrinsics.

A packed logical shift right with a shift count bigger than or equal to the
element size always produces a zero vector. In all other cases, it can be
safely replaced by a 'lshr' instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207299 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd missing include guards and missing #include, found by modules build.
Richard Smith [Sat, 26 Apr 2014 00:53:26 +0000 (00:53 +0000)]
Add missing include guards and missing #include, found by modules build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207298 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAppease the almighty buildbots.
Filipe Cabecinhas [Sat, 26 Apr 2014 00:02:37 +0000 (00:02 +0000)]
Appease the almighty buildbots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207295 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoOptimization for certain shufflevector by using insertps.
Filipe Cabecinhas [Fri, 25 Apr 2014 23:51:17 +0000 (23:51 +0000)]
Optimization for certain shufflevector by using insertps.

Summary:
If we're doing a v4f32/v4i32 shuffle on x86 with SSE4.1, we can lower
certain shufflevectors to an insertps instruction:
When most of the shufflevector result's elements come from one vector (and
keep their index), and one element comes from another vector or a memory
operand.

Added tests for insertps optimizations on shufflevector.
Added support and tests for v4i32 vector optimization.

Reviewers: nadav

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D3475

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207291 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert "blockfreq: Approximate irreducible control flow"
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 23:16:58 +0000 (23:16 +0000)]
Revert "blockfreq: Approximate irreducible control flow"

This reverts commit r207286.  It causes an ICE on the
cmake-llvm-x86_64-linux buildbot [1]:

    llvm/lib/Analysis/BlockFrequencyInfo.cpp: In lambda function:
    llvm/lib/Analysis/BlockFrequencyInfo.cpp:182:1: internal compiler error: in get_expr_operands, at tree-ssa-operands.c:1035

[1]: http://bb.pgr.jp/builders/cmake-llvm-x86_64-linux/builds/12093/steps/build_llvm/logs/stdio

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207287 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Approximate irreducible control flow
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 23:08:57 +0000 (23:08 +0000)]
blockfreq: Approximate irreducible control flow

Previously, irreducible backedges were ignored.  With this commit,
irreducible SCCs are discovered on the fly, and modelled as loops with
multiple headers.

This approximation specifies the headers of irreducible sub-SCCs as its
entry blocks and all nodes that are targets of a backedge within it
(excluding backedges within true sub-loops).  Block frequency
calculations act as if we insert a new block that intercepts all the
edges to the headers.  All backedges and entries to the irreducible SCC
point to this imaginary block.  This imaginary block has an edge (with
even probability) to each header block.

The result is now reasonable enough that I've added a number of
testcases for irreducible control flow.  I've outlined in
`BlockFrequencyInfoImpl.h` ways to improve the approximation.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207286 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUnbreak the gdb buildbot by not lowering dbg.declare intrinsics for arrays.
Adrian Prantl [Fri, 25 Apr 2014 23:00:25 +0000 (23:00 +0000)]
Unbreak the gdb buildbot by not lowering dbg.declare intrinsics for arrays.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207284 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMake sure that rangelists are also relative to the compile unit
Eric Christopher [Fri, 25 Apr 2014 22:23:54 +0000 (22:23 +0000)]
Make sure that rangelists are also relative to the compile unit
low_pc similar to location lists.

Fixes PR19563

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207283 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600: Fix function name printing in LowerCall
Matt Arsenault [Fri, 25 Apr 2014 22:22:01 +0000 (22:22 +0000)]
R600: Fix function name printing in LowerCall

v2: Check both ExternalSymbol and GlobalAddress

Patch by: Jan Vesely <jan.vesely@rutgers.edu>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207282 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDwarfAccelTable: Store the string symbol in the accelerator table to avoid duplicate...
David Blaikie [Fri, 25 Apr 2014 22:21:35 +0000 (22:21 +0000)]
DwarfAccelTable: Store the string symbol in the accelerator table to avoid duplicate lookup.

This also avoids the need for subtly side-effecting calls to manifest
strings in the string table at the point where items are added to the
accelerator tables.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207281 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd an -mattr option to the gold plugin to support subtarget features in LTO
Tom Roeder [Fri, 25 Apr 2014 21:46:51 +0000 (21:46 +0000)]
Add an -mattr option to the gold plugin to support subtarget features in LTO

This adds support for an -mattr option to the gold plugin and to llvm-lto. This
allows the caller to specify details of the subtarget architecture, like +aes,
or +ssse3 on x86.  Note that this requires a change to the include/llvm-c/lto.h
interface: it adds a function lto_codegen_set_attr and it increments the
version of the interface.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207279 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix missing include
Alexey Samsonov [Fri, 25 Apr 2014 21:42:35 +0000 (21:42 +0000)]
Fix missing include

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207278 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoEncapsulate the DWARF string pool in a separate type.
David Blaikie [Fri, 25 Apr 2014 21:34:35 +0000 (21:34 +0000)]
Encapsulate the DWARF string pool in a separate type.

Pulls out some more code from some of the rather monolithic DWARF
classes. Unlike the address table, the string table won't move up into
DwarfDebug - each DWARF file has its own string table (but there can be
only one address table).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207277 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[DWARF parser] Cleanup code in DWARFDebugAranges.
Alexey Samsonov [Fri, 25 Apr 2014 21:30:03 +0000 (21:30 +0000)]
[DWARF parser] Cleanup code in DWARFDebugAranges.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207276 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[DWARF parser] Cleanup code in DWARFDebugAbbrev.
Alexey Samsonov [Fri, 25 Apr 2014 21:10:56 +0000 (21:10 +0000)]
[DWARF parser] Cleanup code in DWARFDebugAbbrev.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207274 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LoopStrengthReduce] Don't trim formula that uses a subset of required registers
Adam Nemet [Fri, 25 Apr 2014 21:02:21 +0000 (21:02 +0000)]
[LoopStrengthReduce] Don't trim formula that uses a subset of required registers

Consider this use from the new testcase:

  LSR Use: Kind=ICmpZero, Offsets={0}, widest fixup type: i32
    reg({1000,+,-1}<nw><%for.body>)
    -3003 + reg({3,+,3}<nw><%for.body>)
    -1001 + reg({1,+,1}<nuw><nsw><%for.body>)
    -1000 + reg({0,+,1}<nw><%for.body>)
    -3000 + reg({0,+,3}<nuw><%for.body>)
    reg({-1000,+,1}<nw><%for.body>)
    reg({-3000,+,3}<nsw><%for.body>)

This is the last use we consider for a solution in SolveRecurse, so CurRegs is
a large set.  (CurRegs is the set of registers that are needed by the
previously visited uses in the in-progress solution.)

ReqRegs is {
  {3,+,3}<nw><%for.body>,
  {1,+,1}<nuw><nsw><%for.body>
}

This is the intersection of the regs used by any of the formulas for the
current use and CurRegs.

Now, the code requires a formula to contain *all* these regs (the comment is
simply wrong), otherwise the formula is immediately disqualified.  Obviously,
no formula for this use contains two regs so they will all get disqualified.

The fix modifies the check to allow the formula in this case.  The idea is
that neither of these formulae is introducing any new registers which is the
point of this early pruning as far as I understand.

In terms of set arithmetic, we now allow formulas whose used regs are a subset
of the required regs not just the other way around.

There are few more loops in the test-suite that are now successfully LSRed.  I
have benchmarked those and found very minimal change.

Fixes <rdar://problem/13965777>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207271 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Use the reference typedef
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 20:52:08 +0000 (20:52 +0000)]
SCC: Use the reference typedef

Actually use the `reference` typedef, and remove the private
redefinition of `pointer` since it has no users.

Using `reference` exposes a problem with r207257, which specified the
wrong `value_type` to `iterator_facade_base` (fixed that too).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207270 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThis reapplies r207235 with an additional bugfixes caught by the msan
Adrian Prantl [Fri, 25 Apr 2014 20:49:25 +0000 (20:49 +0000)]
This reapplies r207235 with an additional bugfixes caught by the msan
buildbot - do not insert debug intrinsics before phi nodes.

Debug info for optimized code: Support variables that are on the stack and
described by DBG_VALUEs during their lifetime.

Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.

This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine.ll testcase and included source

rdar://problem/16679879
http://reviews.llvm.org/D3374

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207269 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMCAssembler: Simplify implementation of const variants of getSymbolData by calling...
David Blaikie [Fri, 25 Apr 2014 20:19:11 +0000 (20:19 +0000)]
MCAssembler: Simplify implementation of const variants of getSymbolData by calling one implementation from the other.

Code review feedback by Rafael Espindola on r207124.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207266 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoBugPoint: Fix some memory leaks.
David Blaikie [Fri, 25 Apr 2014 20:15:16 +0000 (20:15 +0000)]
BugPoint: Fix some memory leaks.

Patch by Kostya Serebryany.

unique_ptr would be nice, but it's a bit too much work for an area I'm
not familiar with, nor invested in, unfortunately.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207265 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDwarfUnit: Remove unused function
David Blaikie [Fri, 25 Apr 2014 20:02:24 +0000 (20:02 +0000)]
DwarfUnit: Remove unused function

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207264 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDIE: Pass ownership of children via std::unique_ptr rather than raw pointer.
David Blaikie [Fri, 25 Apr 2014 20:00:34 +0000 (20:00 +0000)]
DIE: Pass ownership of children via std::unique_ptr rather than raw pointer.

This should reduce the chance of memory leaks like those fixed in
r207240.

There's still some unclear ownership of DIEs happening in DwarfDebug.
Pushing unique_ptr and references through more APIs should help expose
the cases where ownership is a bit fuzzy.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207263 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDIEEntry: Refer to the specified DIE via reference rather than pointer.
David Blaikie [Fri, 25 Apr 2014 19:33:43 +0000 (19:33 +0000)]
DIEEntry: Refer to the specified DIE via reference rather than pointer.

Makes some more cases (the unit tests, specifically), lexically
compatible with a change to unique_ptr.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207261 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDwarfUnit: return by reference from createAndAddDIE
David Blaikie [Fri, 25 Apr 2014 18:52:29 +0000 (18:52 +0000)]
DwarfUnit: return by reference from createAndAddDIE

Since this doesn't return ownership (the DIE has been added to the
specified parent already) nor return null, just return by reference.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207259 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Further shift logic to LoopData
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:47:04 +0000 (18:47 +0000)]
blockfreq: Further shift logic to LoopData

Move a lot of the loop-related logic that was sprinkled around the code
into `LoopData`.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207258 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Provide operator->() through iterator_facade_base
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:43:41 +0000 (18:43 +0000)]
SCC: Provide operator->() through iterator_facade_base

Use the fancy new `iterator_facade_base` to add
`scc_iterator::operator->()`.  Remove other definitions where
`iterator_facade_base` does the right thing.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207257 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoenable fast isel tablegen files for Mips
Reed Kotler [Fri, 25 Apr 2014 18:36:38 +0000 (18:36 +0000)]
enable fast isel tablegen files for Mips

Reviewers: dsanders

Reviewed By: dsanders

Differential Revision: http://reviews.llvm.org/D3498

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207256 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoReturn DIE by reference instead of pointer from DwarfUnit::getUnitDie
David Blaikie [Fri, 25 Apr 2014 18:35:57 +0000 (18:35 +0000)]
Return DIE by reference instead of pointer from DwarfUnit::getUnitDie

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207255 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Remove non-const operator*()
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:26:45 +0000 (18:26 +0000)]
SCC: Remove non-const operator*()

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207254 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDwarfUnit: Suddently, DIE references, everywhere.
David Blaikie [Fri, 25 Apr 2014 18:26:14 +0000 (18:26 +0000)]
DwarfUnit: Suddently, DIE references, everywhere.

This'll make changing to unique_ptr ownership of DIEs easier since the
usages will now have '*' on them making them textually compatible
between unique_ptr and raw pointer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207253 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Change clients to use const, NFC
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:24:50 +0000 (18:24 +0000)]
SCC: Change clients to use const, NFC

It's fishy to be changing the `std::vector<>` owned by the iterator, and
no one actual does it, so I'm going to remove the ability in a
subsequent commit.  First, update the users.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207252 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Doxygen-ize comments, NFC
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:18:46 +0000 (18:18 +0000)]
SCC: Doxygen-ize comments, NFC

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207251 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert "This reapplies r207130 with an additional testcase+and a missing check for"
Adrian Prantl [Fri, 25 Apr 2014 18:18:09 +0000 (18:18 +0000)]
Revert "This reapplies r207130 with an additional testcase+and a missing check for"

This reverts commit 207235 to investigate msan buildbot breakage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207250 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Un-inline long functions
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:15:50 +0000 (18:15 +0000)]
SCC: Un-inline long functions

These are long functions that really shouldn't be inlined.  Otherwise,
no functionality change.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207249 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSCC: Remove redundant inline keywords, NFC
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 18:10:23 +0000 (18:10 +0000)]
SCC: Remove redundant inline keywords, NFC

Functions declared in line in a class are inlined by default.  There's
no reason for the `inline` keyword.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207248 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMake sure that DSUB does not duplicate the pattern of DSUBU
Reed Kotler [Fri, 25 Apr 2014 18:05:00 +0000 (18:05 +0000)]
Make sure that DSUB does not duplicate the pattern of DSUBU

Test Plan:
Run test suite to make sure there is no regression.
https://dmz-portal.mips.com/bb/builders/LLVM%20with%2064bit%20and%20delay%20slot%20optimizer%20and%20direct%20object%20emitter/builds/626

Reviewers: dsanders

Reviewed By: dsanders

Differential Revision: http://reviews.llvm.org/D3497

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207247 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM: remove @llvm.arm.sevl
Saleem Abdulrasool [Fri, 25 Apr 2014 17:51:25 +0000 (17:51 +0000)]
ARM: remove @llvm.arm.sevl

This intrinsic is no longer needed with the new @llvm.arm.hint(i32) intrinsic
which provides a generic, extensible manner for adding hint instructions.  This
functionality can now be represented as @llvm.arm.hint(i32 5).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207246 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[inline cold threshold] Command line argument for inline threshold will
Manman Ren [Fri, 25 Apr 2014 17:34:55 +0000 (17:34 +0000)]
[inline cold threshold] Command line argument for inline threshold will
override the default cold threshold.

When we use command line argument to set the inline threshold, the default
cold threshold will not be used. This is in line with how we use
OptSizeThreshold. When we want a higher threshold for all functions, we
do not have to set both inline threshold and cold threshold.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207245 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRefactor some common logic in DwarfUnit::constructVariableDIE and pass non-null DIE...
David Blaikie [Fri, 25 Apr 2014 17:32:19 +0000 (17:32 +0000)]
Refactor some common logic in DwarfUnit::constructVariableDIE and pass non-null DIE by reference to DbgVariable::setDIE

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207244 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM: provide a new generic hint intrinsic
Saleem Abdulrasool [Fri, 25 Apr 2014 17:24:24 +0000 (17:24 +0000)]
ARM: provide a new generic hint intrinsic

Introduce the llvm.arm.hint(i32) intrinsic that can be used to inject hints into
the instruction stream. This is particularly useful for generating IR from a
compiler where the user may inject an intrinsic (e.g. __yield). These are then
pattern substituted into the correct instruction which already existed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207242 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoPR19554: Fix some memory leaks in DIEHashTest.cpp
David Blaikie [Fri, 25 Apr 2014 17:07:55 +0000 (17:07 +0000)]
PR19554: Fix some memory leaks in DIEHashTest.cpp

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207240 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoReapply r207135 without modifications.
Adrian Prantl [Fri, 25 Apr 2014 17:01:04 +0000 (17:01 +0000)]
Reapply r207135 without modifications.

Debug info: Let dbg.values inserted by LowerDbgDeclare inherit the location
of the dbg.value. This gets rid of tons of redundant variable DIEs in
subscopes.

rdar://problem/14874886, rdar://problem/16679936

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207236 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThis reapplies r207130 with an additional testcase+and a missing check for
Adrian Prantl [Fri, 25 Apr 2014 17:01:00 +0000 (17:01 +0000)]
This reapplies r207130 with an additional testcase+and a missing check for
AllocaInst that was missing in one location.
Debug info for optimized code: Support variables that are on the stack and
described by DBG_VALUEs during their lifetime.

Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.

This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine.ll testcase and included source

rdar://problem/16679879
http://reviews.llvm.org/D3374

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207235 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM64] When compiling for ELF in PIC mode, local symbols shouldn't go through the GOT
Tilmann Scheller [Fri, 25 Apr 2014 13:43:18 +0000 (13:43 +0000)]
[ARM64] When compiling for ELF in PIC mode, local symbols shouldn't go through the GOT

There's no need for local symbols to go through the GOT, in fact it seems GNU ld is not even emitting GOT entries for local symbols and will error out when trying to resolve a GOT relocation for a local symbol.

This bug triggers when bootstrapping clang on AArch64 Linux with -fPIC and the ARM64 backend. The AArch64 backend is not affected.

With this commit it's now possible to bootstrap clang on AArch64 Linux with the ARM64 backend (-fPIC, -O3).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207226 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM64] Handle fp128 for parameter passing on stack
Jiangning Liu [Fri, 25 Apr 2014 12:07:03 +0000 (12:07 +0000)]
[ARM64] Handle fp128 for parameter passing on stack

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207222 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM64: fix assertion in ISelDAGToDAG
Tim Northover [Fri, 25 Apr 2014 10:48:47 +0000 (10:48 +0000)]
ARM64: fix assertion in ISelDAGToDAG

Also an unused variable, so double bonus!

This should deal with PR19548.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207221 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM64] Print preferred aliases for SFBM/UBFM in InstPrinter
Bradley Smith [Fri, 25 Apr 2014 10:25:29 +0000 (10:25 +0000)]
[ARM64] Print preferred aliases for SFBM/UBFM in InstPrinter

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207219 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] During the incremental update of an SCC, switch to using the
Chandler Carruth [Fri, 25 Apr 2014 09:52:44 +0000 (09:52 +0000)]
[LCG] During the incremental update of an SCC, switch to using the
SCCMap to test for nodes that have been re-added to the root SCC rather
than a set vector. We already have done the SCCMap lookup, we juts need
to test it in two different ways. In turn, do most of the processing of
these nodes as they go into the root SCC rather than lazily. This
simplifies the final loop to just stitch the root SCC into its
children's parent sets. No functionlatiy changed.

However, this makes a few things painfully obvious, which was my intent.
=] There is tons of repeated code introduced here and elsewhere. I'm
splitting the refactoring of that code into helpers from this change so
its clear that this is the change which switches the datastructures used
around, and the other is a pure factoring & deduplication of code
change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207217 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM64] Add RUN lines for "–target arm64 –mattr=-fp-armv8" on AArch64 no-fp test.
Kevin Qin [Fri, 25 Apr 2014 09:44:20 +0000 (09:44 +0000)]
[ARM64] Add RUN lines for "–target arm64 –mattr=-fp-armv8" on AArch64 no-fp test.

This patch is a supplement of implementing predicate of FP, enabling aarch64 backend
no-fp tests on arm64 target for verification. During this, one bug is exposed and
fixed by this patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207215 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM64] Support crc predicate on ARM64.
Kevin Qin [Fri, 25 Apr 2014 09:25:42 +0000 (09:25 +0000)]
[ARM64] Support crc predicate on ARM64.

According to the specification, CRC is an optional extension of the
architecture.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207214 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] During the incremental re-build of an SCC after removing an edge,
Chandler Carruth [Fri, 25 Apr 2014 09:08:10 +0000 (09:08 +0000)]
[LCG] During the incremental re-build of an SCC after removing an edge,
remove the nodes in the SCC from the SCC map entirely prior to the DFS
walk. This allows the SCC map to represent both the state of
not-yet-re-added-to-an-SCC and added-back-to-this-SCC independently. The
first is being missing from the SCC map, the second is mapping back to
'this'. In a subsequent commit, I'm going to use this property to
simplify the new node list for this SCC.

In theory, I think this also makes the contract for orphaning a node
from the graph slightly less confusing. Now it is also orphaned from the
SCC graph. Still, this isn't quite right either, and so I'm not adding
test cases here. I'll add test cases for the behavior of orphaning nodes
when the code *actually* supports it. The change here is mostly
incidental, my goal is simplifying the algorithm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207213 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Rather than doing a linear time SmallSetVector removal of each
Chandler Carruth [Fri, 25 Apr 2014 09:08:05 +0000 (09:08 +0000)]
[LCG] Rather than doing a linear time SmallSetVector removal of each
child from the worklist, wait until we actually need to pop another
element off of the worklist and skip over any that were already visited
by the DFS. This also enables swapping the nodes of the SCC into the
worklist. No functionality changed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207212 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Remove a completely unnecessary loop. It wasn't even doing any
Chandler Carruth [Fri, 25 Apr 2014 06:45:06 +0000 (06:45 +0000)]
[LCG] Remove a completely unnecessary loop. It wasn't even doing any
thing, just mucking up the code. I feel bad that I even wrote this loop.
Very sorry. The diff is huge because of the indent change, but I promise
all this is doing is realizing that the outer two loops were actually
the exact same loops, and we didn't need two of them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207202 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LCG] Now that the loop structure of the core SCC finding routine is
Chandler Carruth [Fri, 25 Apr 2014 06:38:58 +0000 (06:38 +0000)]
[LCG] Now that the loop structure of the core SCC finding routine is
factored into a more reasonable form, replace the tail call with
a simple outer-loop continuation. It's sad that C++ makes this so
awkward to write, but it seems more direct and clear than the tail call
at this point.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207201 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoX86: convert object streamer selection to a switch
Saleem Abdulrasool [Fri, 25 Apr 2014 06:29:36 +0000 (06:29 +0000)]
X86: convert object streamer selection to a switch

Change the object streamer selection to a switch from a series of if conditions.
Rather than defaulting to ELF, require that an ELF format is requested.  The
Windows/!ELF is maintained as MachO would have been selected first and will
still provide a MachO format.  Add an assertion that if COFF is requested that
the target platform is Windows as only WinCOFF object emission is currently
supported.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207200 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[python] Fix getting section contents.
Anders Waldenborg [Fri, 25 Apr 2014 06:25:15 +0000 (06:25 +0000)]
[python] Fix getting section contents.

The returnvalue was handled as c_char_p which ment that ctypes
handled it as a NUL-terminated string making it cut the contents
at first NUL (or even worse - overrunning the buffer if it doesn't
contain a NUL).

Differential Revision: http://reviews.llvm.org/D3474

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207199 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd missing cpp file header
David Blaikie [Fri, 25 Apr 2014 06:22:32 +0000 (06:22 +0000)]
Add missing cpp file header

Code review feedback from Paul Robinson on r207022

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207198 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[C++] Use 'nullptr'. Target edition.
Craig Topper [Fri, 25 Apr 2014 05:30:21 +0000 (05:30 +0000)]
[C++] Use 'nullptr'. Target edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207197 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[C++] Use 'nullptr'. Transforms edition.
Craig Topper [Fri, 25 Apr 2014 05:29:35 +0000 (05:29 +0000)]
[C++] Use 'nullptr'. Transforms edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207196 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Only one mass distribution per node
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:43 +0000 (04:38 +0000)]
blockfreq: Only one mass distribution per node

Remove the concepts of "forward" and "general" mass distributions, which
was wrong.  The split might have made sense in an early version of the
algorithm, but it's definitely wrong now.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207195 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Document assertion
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:40 +0000 (04:38 +0000)]
blockfreq: Document assertion

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207194 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Use better branch weights in multiexit test
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:37 +0000 (04:38 +0000)]
blockfreq: Use better branch weights in multiexit test

The branch weights were even before.  Make them different.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207193 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Clean up irreducible testcases
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:35 +0000 (04:38 +0000)]
blockfreq: Clean up irreducible testcases

Strip irreducible testcases to pure control flow.  The function calls
made the branch weights more believable but cluttered it up a lot.
There isn't going to be any constant analysis here, so just use dumb
branch logic to clarify the important parts.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207192 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Document high-level functions
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:32 +0000 (04:38 +0000)]
blockfreq: Document high-level functions

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207191 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Remove dead code
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:30 +0000 (04:38 +0000)]
blockfreq: Remove dead code

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207190 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Scale LoopData::Scale on the way down
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:27 +0000 (04:38 +0000)]
blockfreq: Scale LoopData::Scale on the way down

Rather than scaling loop headers and then scaling all the loop members
by the header frequency, scale `LoopData::Scale` itself, and scale the
loop members by it.  It's much more obvious what's going on this way,
and doesn't cost any extra multiplies.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207189 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: unwrapLoopPackage() => unwrapLoop()
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:25 +0000 (04:38 +0000)]
blockfreq: unwrapLoopPackage() => unwrapLoop()

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207188 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Pass the Loop directly into unwrapLoopPackage()
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:23 +0000 (04:38 +0000)]
blockfreq: Pass the Loop directly into unwrapLoopPackage()

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207187 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Unwrap from Loops
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:20 +0000 (04:38 +0000)]
blockfreq: Unwrap from Loops

When unwrapping loops, just visit the loops rather than all nodes.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207186 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Separate unwrapLoops() from finalizeMetrics()
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:17 +0000 (04:38 +0000)]
blockfreq: Separate unwrapLoops() from finalizeMetrics()

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207185 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: LoopData::MemberList => NodeList
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:15 +0000 (04:38 +0000)]
blockfreq: LoopData::MemberList => NodeList

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207184 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Expose getPackagedNode()
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:12 +0000 (04:38 +0000)]
blockfreq: Expose getPackagedNode()

Make `getPackagedNode()` a member function of
`BlockFrequencyInfoImplBase` so that it's available for templated code.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207183 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Store the header with the members
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:09 +0000 (04:38 +0000)]
blockfreq: Store the header with the members

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207182 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Encapsulate LoopData::Header
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:06 +0000 (04:38 +0000)]
blockfreq: Encapsulate LoopData::Header

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207181 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Embed Loop hierarchy in LoopData
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:03 +0000 (04:38 +0000)]
blockfreq: Embed Loop hierarchy in LoopData

Continue refactoring to make `LoopData` first-class.  Here I'm making
the `LoopData` hierarchy explicit, instead of bouncing back and forth
with `WorkingData`.  This simplifies the logic and better matches the
`LoopInfo` design.  (Eventually, `LoopInfo` should be restructured so
that it supports this pass, and `LoopData` can be removed.)

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207180 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Use LoopData directly
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:38:01 +0000 (04:38 +0000)]
blockfreq: Use LoopData directly

Instead of passing around loop headers, pass around `LoopData` directly.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207179 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Stop using range-based for to traverse Loops
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:37:58 +0000 (04:37 +0000)]
blockfreq: Stop using range-based for to traverse Loops

A follow-up commit will need the actual iterators.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207178 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoblockfreq: Use a std::list for Loops
Duncan P. N. Exon Smith [Fri, 25 Apr 2014 04:30:06 +0000 (04:30 +0000)]
blockfreq: Use a std::list for Loops

As pointed out by David Blaikie in code review, a `std::list<T>` is
simpler than a `std::vector<std::unique_ptr<T>>`.  Another option is a
`std::deque<T>` (which allocates in chunks), but I'd like to leave open
the option of inserting in the middle of the sequence for handling
irreducible control flow on the fly.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207177 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[C++] Use 'nullptr'. Tools edition.
Craig Topper [Fri, 25 Apr 2014 04:24:47 +0000 (04:24 +0000)]
[C++] Use 'nullptr'. Tools edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207176 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAllow vectorization of bit intrinsics in BB Vectorizer.
Karthik Bhat [Fri, 25 Apr 2014 03:33:48 +0000 (03:33 +0000)]
Allow vectorization of bit intrinsics in BB Vectorizer.
This patch adds support for vectorization of  bit intrinsics such as bswap,ctpop,ctlz,cttz.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207174 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoProfileData: Treat missing function counts as malformed
Justin Bogner [Fri, 25 Apr 2014 02:45:33 +0000 (02:45 +0000)]
ProfileData: Treat missing function counts as malformed

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207172 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoChange llvm-config --ldflags to report ${CMAKE_CXX_LINK_FLAGS}
Reid Kleckner [Fri, 25 Apr 2014 01:44:20 +0000 (01:44 +0000)]
Change llvm-config --ldflags to report ${CMAKE_CXX_LINK_FLAGS}

Should fix PR19526.

When Oscar added this code in the intial CMake build system port, he had
a TODO saying that ${CMAKE_SHARED_LINKER_FLAGS} was probably wrong.  I
agree.  I'm using ${CMAKE_CXX_LINK_FLAGS} to point LLVM at my custom
installation of gcc 4.recent, so that seems more correct.  With this
change, I can build creduce against an installed clang, and it picks up
the write flags from --ldflags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207171 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix quadratic performance during debug compression due to sections x symbols iteration.
David Blaikie [Fri, 25 Apr 2014 00:48:01 +0000 (00:48 +0000)]
Fix quadratic performance during debug compression due to sections x symbols iteration.

When fixing the symbols in each compressed section we were iterating
over all symbols for each compressed section. In extreme cases this
could snowball severely (5min uncompressed -> 35min compressed) due to
iterating over all symbols for each compressed section (large numbers of
compressed sections can be generated by DWARF type units).

To address this, build a map of the symbols in each section ahead of
time, and access that map if a section is being compressed. This brings
compile time for the aforementioned example down to ~6 minutes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207167 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert "This reapplies r207130 with an additional testcase+and a missing check for"
Adrian Prantl [Fri, 25 Apr 2014 00:42:50 +0000 (00:42 +0000)]
Revert "This reapplies r207130 with an additional testcase+and a missing check for"
Typo in testcase.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207166 91177308-0d34-0410-b5e6-96231b3b80d8