Commit Graph

2184 Commits

Author SHA1 Message Date
Hal Finkel
f208398528 Revert: r189565 - Add getUnrollingPreferences to TTI
Revert unintentional commit (of an unreviewed change).

Original commit message:

Add getUnrollingPreferences to TTI

Allow targets to customize the default behavior of the generic loop unrolling
transformation. This will be used by the PowerPC backend when targeting the A2
core (which is in-order with a deep pipeline), and using more aggressive
defaults is important.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189566 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-29 03:33:15 +00:00
Hal Finkel
32f258b96a Add getUnrollingPreferences to TTI
Allow targets to customize the default behavior of the generic loop unrolling
transformation. This will be used by the PowerPC backend when targeting the A2
core (which is in-order with a deep pipeline), and using more aggressive
defaults is important.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189565 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-29 03:29:57 +00:00
Richard Sandiford
a8a7099c18 Turn MipsOptimizeMathLibCalls into a target-independent scalar transform
...so that it can be used for z too.  Most of the code is the same.
The only real change is to use TargetTransformInfo to test when a sqrt
instruction is available.

The pass is opt-in because at the moment it only handles sqrt.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189097 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-23 10:27:02 +00:00
Jakub Staszak
328a4fbfbb Add some constantness.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188844 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-20 23:04:15 +00:00
Nick Lewycky
72dba254ae Fix an oversight in isPotentiallyReachable where we wouldn't do any CFG-walking
to find loops if the From and To instructions were in the same block.

Refactor the code a little now that we need to fill to start the CFG-walking
algorithm with more than one starting basic block sometimes.

Special thanks to Andrew Trick for catching an error in my understanding of
natural loops in code review.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188236 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-13 00:03:47 +00:00
Matt Arsenault
00d7baad90 Fix missing -*- C++ -*-s
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187758 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-06 00:16:21 +00:00
Nick Lewycky
81e480463d Reimplement isPotentiallyReachable to make nocapture deduction much stronger.
Adds unit tests for it too.

Split BasicBlockUtils into an analysis-half and a transforms-half, and put the
analysis bits into a new Analysis/CFG.{h,cpp}. Promote isPotentiallyReachable
into llvm::isPotentiallyReachable and move it into Analysis/CFG.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187283 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-27 01:24:00 +00:00
Tom Stellard
57e6b2d1f3 SimplifyCFG: Use parallel-and and parallel-or mode to consolidate branch conditions
Merge consecutive if-regions if they contain identical statements.
Both transformations reduce number of branches.  The transformation
is guarded by a target-hook, and is currently enabled only for +R600,
but the correctness has been tested on X86 target using a variety of
CPU benchmarks.

Patch by: Mei Ye

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187278 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-27 00:01:07 +00:00
Matt Arsenault
a1ac7ede0c Fix typo (ponted -> pointed)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186896 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-22 23:52:23 +00:00
Craig Topper
a0ec3f9b7b Use SmallVectorImpl& instead of SmallVector to avoid repeating small vector size.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186274 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 04:42:23 +00:00
Michael Gottesman
bee07bddea Fixed 80+ violation and added C++ to header.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186225 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-12 23:09:43 +00:00
Arnold Schwaighofer
c0a11edba6 TargetTransformInfo: address calculation parameter for gather/scather
Address calculation for gather/scather in vectorized code can incur a
significant cost making vectorization unbeneficial. Add infrastructure to add
cost.
Tests and cost model for targets will be in follow-up commits.

radar://14351991

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186187 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-12 19:16:02 +00:00
Michael Gottesman
4ff87aaf84 Fixed comment in header of Block Frequency Impl and added text for C++ mode.
This is a generic block implementation that works on more than machine blocks.
The C++ mode addition is a bonus due to the extra space provided.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186175 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-12 18:11:14 +00:00
Shuxin Yang
a10369920f Stylistic change.
Thank Nick for figuring out these problems.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186146 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-12 07:25:38 +00:00
Shuxin Yang
5e915e6e36 Fix a SCEV update problem.
The symptom is seg-fault, and the root cause is that a SCEV contains a SCEVUnknown
which has null-pointer to a llvm::Value.

 This is how the problem take place:
 ===================================
  1). In the pristine input IR, there are two relevant instrutions Op1 and Op2, 
     Op1's corresponding SCEV (denoted as SCEV(op1)) is a SCEVUnknown, and
     SCEV(Op2) contains SCEV(Op1).  None of these instructions are dead.

     Op1 : V1 = ...
     ...
     Op2 : V2 = ... // directly or indirectly (data-flow) depends on Op1
    
  2) Optimizer (LSR in my case) generates an instruction holding the equivalent
     value of Op1, making Op1 dead. 
     Op1': V1' = ...
     Op1: V1 = ... ; now dead)
     Op2 : V2 = ... //Now deps on Op1', but the SCEV(Op2) still contains SCEV(Op1)

  3) Op1 is deleted, and call-back function is called to reset 
     SCEV(Op1) to indicate it is invalid. However, SCEV(Op2) is not 
     invalidated as well.

  4) Following pass get the cached, invalid SCEV(Op2), and try to manipulate it,
     and cause segfault. 

 The fix:
 ========
 It seems there is no clean yet inexpensive fix. I write to dev-list
soliciting good solution, unforunately no ack. So, I decide to fix this 
problem in a brute-force way:

  When ScalarEvolution::getSCEV is called, check if the cached SCEV 
contains a invalid SCEVUnknow, if yes, remove the cached SCEV, and
re-evaluate the SCEV from scratch.

  I compile buch of big *.c and *.cpp, fortunately, I don't see any increase
in compile time.

 Misc:
=====
 The reduced test-case has 2357 lines of code+other-stuff, too big to commit.

 rdar://14283433


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185843 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-08 17:33:13 +00:00
Preston Briggs
99666cb8f9 extending the interface of Dependence slightly to support future work
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185241 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-28 23:34:23 +00:00
Jakob Stoklund Olesen
37cb687fcb Try to unbreak Linux buildbots.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185237 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-28 22:54:16 +00:00
Jakob Stoklund Olesen
97be1d608e Minimize precision loss when computing cyclic probabilities.
Allow block frequencies to exceed 32 bits by using the new
BlockFrequency division function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185236 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-28 22:40:43 +00:00
Jakob Stoklund Olesen
b0f8341b34 Merge isReachable into isBackedge.
Prefer using RPO.lookup() instead of RPO[] which can mutate the map.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184891 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-25 23:32:10 +00:00
Benjamin Kramer
75b5162154 BlockFrequency: Bump up the entry frequency a bit.
This is a band-aid to fix the most severe regressions we're seeing from basing
spill decisions on block frequencies, until we have a better solution.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184835 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-25 13:34:40 +00:00
Benjamin Kramer
3ade978551 Add a const version of findNearestCommonDominator to PostDom for convenience.
PR16283.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183629 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-09 15:09:30 +00:00
Jakub Staszak
326ae27c4f Remove unneeded #include.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183460 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-06 23:34:11 +00:00
Quentin Colombet
06f5ebc5a1 Loop Strength Reduce: Scaling factor cost.
Account for the cost of scaling factor in Loop Strength Reduce when rating the
formulae. This uses a target hook.

The default implementation of the hook is: if the addressing mode is legal, the
scaling factor is free.

<rdar://problem/13806271>


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183045 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 21:29:03 +00:00
Andrew Trick
616011418e Fix ScalarEvolution::ComputeExitLimitFromCond for 'or' conditions.
Fixes PR16130 - clang produces incorrect code with loop/expression at -O2.

This is a 2+ year old bug that's now holding up the release. It's a
case where we knowingly made aggressive assumptions about undefined
behavior. These assumptions are wrong when SCEV is computing a
subexpression that does not directly control the branch. With this
fix, we avoid making assumptions in those cases but still optimize the
common case. SCEV's trip count computation for exits controlled by
'or' expressions is now analagous to the trip count computation for
loops with multiple exits. I had already fixed the multiple exit case
to be conservative.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182989 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 06:43:25 +00:00
Paul Redmond
ee21b6f7b4 Add support for llvm.vectorizer metadata
- llvm.loop.parallel metadata has been renamed to llvm.loop to be more generic
  by making the root of additional loop metadata.
  - Loop::isAnnotatedParallel now looks for llvm.loop and associated
    llvm.mem.parallel_loop_access
  - document llvm.loop and update llvm.mem.parallel_loop_access
- add support for llvm.vectorizer.width and llvm.vectorizer.unroll
  - document llvm.vectorizer.* metadata
  - add utility class LoopVectorizerHints for getting/setting loop metadata
  - use llvm.vectorizer.width=1 to indicate already vectorized instead of
    already_vectorized
- update existing tests that used llvm.loop.parallel and
  llvm.vectorizer.already_vectorized

Reviewed by: Nadav Rotem


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182802 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-28 20:00:34 +00:00
Michael Kuperstein
9f5de6dadc Make BasicAliasAnalysis recognize the fact a noalias argument cannot alias another argument, even if the other argument is not itself marked noalias.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182755 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-28 08:17:48 +00:00
Diego Novillo
77226a03dc Add a new function attribute 'cold' to functions.
Other than recognizing the attribute, the patch does little else.
It changes the branch probability analyzer so that edges into
blocks postdominated by a cold function are given low weight.

Added analysis and code generation tests.  Added documentation for the
new attribute.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182638 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-24 12:26:52 +00:00
Matt Arsenault
63f3ca5da7 Add missing -*- C++ -*- to headers
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182164 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-17 21:43:39 +00:00
Sylvestre Ledru
d14743ed79 Fix a typo (ouput => output)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182090 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-17 12:31:43 +00:00
David Blaikie
453f4f0130 Use only explicit bool conversion operators
BitVector/SmallBitVector::reference::operator bool remain implicit since
they model more exactly a bool, rather than something else that can be
boolean tested.

The most common (non-buggy) case are where such objects are used as
return expressions in bool-returning functions or as boolean function
arguments. In those cases I've used (& added if necessary) a named
function to provide the equivalent (or sometimes negative, depending on
convenient wording) test.

One behavior change (YAMLParser) was made, though no test case is
included as I'm not sure how to reach that code path. Essentially any
comparison of llvm::yaml::document_iterators would be invalid if neither
iterator was at the end.

This helped uncover a couple of bugs in Clang - test cases provided for
those in a separate commit along with similar changes to `operator bool`
instances in Clang.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181868 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-15 07:36:59 +00:00
Tobias Grosser
d03fdfb97c RegionInfo: Add helpers to replace entry/exit recursively
Contributed by: Star Tan <tanmx_star@yeah.net>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179157 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-10 06:54:49 +00:00
Nadav Rotem
8e4df489d0 Revert r176408 and r176407 to address PR15540.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179111 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 18:16:05 +00:00
Arnold Schwaighofer
6bf4f67641 CostModel: Add parameter to instruction cost to further classify operand values
On certain architectures we can support efficient vectorized version of
instructions if the operand value is uniform (splat) or a constant scalar.
An example of this is a vector shift on x86.

We can efficiently support

for (i = 0 ; i < ; i += 4)
  w[0:3] = v[0:3] << <2, 2, 2, 2>

but not

for (i = 0; i < ; i += 4)
  w[0:3] = v[0:3] << x[0:3]

This patch adds a parameter to getArithmeticInstrCost to further qualify operand
values as uniform or uniform constant.

Targets can then choose to return a different cost for instructions with such
operand values.

A follow-up commit will test this feature on x86.

radar://13576547

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178807 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-04 23:26:21 +00:00
Andrew Trick
e74c2e86cb Fix SCEV forgetMemoizedResults should search and destroy backedge exprs.
Fixes PR15570: SEGV: SCEV back-edge info invalid after dead code removal.

Indvars creates a SCEV expression for the loop's back edge taken
count, then determines that the comparison is always true and
removes it.

When loop-unroll asks for the expression, it contains a NULL
SCEVUnknkown (as a CallbackVH).

forgetMemoizedResults should invalidate the loop back edges expression.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177986 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-26 03:14:53 +00:00
Rafael Espindola
80c6a66bbf Add std prefixes to fix the build with xlc.
Patch by Kai <kai@redstar.de>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177574 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-20 21:00:22 +00:00
Jakub Staszak
02a2d4fb9e Remove unneeded #includes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177351 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-18 23:33:44 +00:00
Jakub Staszak
ec3bb4b660 Remove unneeded #includes. Use forward declarations instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176783 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-10 00:34:01 +00:00
Jakub Staszak
e939a066c8 Remove unneeded #includes. Use forward declarations instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176782 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-10 00:20:16 +00:00
Jakub Staszak
64bf55af6f Remove unneeded const_cast.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176771 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-09 18:24:26 +00:00
Jakub Staszak
209cb5b56b Use forward declaration instead of #include.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176770 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-09 18:05:34 +00:00
David Blaikie
423de3f047 Remove -print-dbginfo as it is unused & bitrotten.
This pass hasn't been touched in two years & would fail with assertions against
the current debug info metadata format (the only test case for it still uses a
many-versions old debug info metadata format)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176707 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-08 18:17:46 +00:00
Shuxin Yang
985dac6579 Memory Dependence Analysis (not mem-dep test) take advantage of "invariant.load" metadata.
The "invariant.load" metadata indicates the memory unit being accessed is immutable.
A load annotated with this metadata can be moved across any store.

As I am not sure if it is legal to move such loads across barrier/fence, this
change dose not allow such transformation.

rdar://11311484

Thank Arnold for code review.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176562 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-06 17:48:48 +00:00
Jakub Staszak
62c102360a Remove duplicated forward declaration.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176536 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-06 00:04:32 +00:00
Jakub Staszak
f84606732c Fix a few typos in comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176519 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-05 22:05:16 +00:00
Jakub Staszak
fe0c244633 Add some constantness.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176518 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-05 22:01:15 +00:00
Jakub Staszak
aa4e2aea9e std::distance() == 0 means that iterators are equal. No functionality change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176516 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-05 21:53:57 +00:00
Arnold Schwaighofer
6936ebd700 Use the right number of slashes in comment string
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176504 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-05 19:04:12 +00:00
Arnold Schwaighofer
1c9bdf1e5e Clarify comment for function getObjectSize
Clarify that we mean the object starting at the pointer to the end of the
underlying object and not the size of the whole allocated object.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176491 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-05 16:53:24 +00:00
Nuno Lopes
b443a0aeac recommit r172363 & r171325 (reverted in r172756)
This adds minimalistic support for PHI nodes to llvm.objectsize() evaluation

fingers crossed so that it does break clang boostrap again..

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176408 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-02 11:36:24 +00:00
Nuno Lopes
2bc689c846 add getUnderlyingObjectSize()
this is similar to getObjectSize(), but doesnt subtract the offset
tweak the BasicAA code accordingly (per PR14988)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176407 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-02 11:23:34 +00:00