Commit Graph

102328 Commits

Author SHA1 Message Date
Tim Northover
e23fabc2ac ARM64: start porting regression test suite from AArch64
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206166 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 12:50:41 +00:00
Arnaud A. de Grandmaison
c81e5d6e64 Display the name of the project failing the url check in the release script
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206164 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 12:36:48 +00:00
Richard Osborne
c9fa660ddd [XCore] Don't create invalid MKMSK instructions inside loadImmediate().
Summary:
Previously loadImmediate() would produce MKMSK instructions with invalid
immediate values such as mkmsk r0, 9. Fix this by checking the mask size
is valid.

Reviewers: robertlytton

Reviewed By: robertlytton

CC: llvm-commits

Differential Revision: http://reviews.llvm.org/D3289

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206163 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 12:30:35 +00:00
NAKAMURA Takumi
1377449177 Whitespace.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206154 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 07:03:13 +00:00
NAKAMURA Takumi
9854380054 Revert r206045, "Fix shift by constants for vector."
It broke some builders, at least, i686.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206153 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 07:02:57 +00:00
Chandler Carruth
1907f75445 [Allocator] Hoist the external helper function into a namespace scope
declaration. GCC 4.7 appears to get hopelessly confused by declaring
this function within a member function of a class template. Go figure.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206152 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 06:42:56 +00:00
Chandler Carruth
273a62f9ac [cleanup] Normalize YAMLParser.h's doxygen markers with '\foo' instead
of '@foo'. The slash variant is prefered in LLVM and the file was
already using a mixture. =/

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206151 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 06:22:03 +00:00
Hal Finkel
1aee811d71 Don't assert in BasicTTI::getMemoryOpCost for non-simple types
BasicTTI::getMemoryOpCost must explicitly check for non-simple types; setting
AllowUnknown=true with TLI->getSimpleValueType is not sufficient because, for
example, non-power-of-two vector types return non-simple EVTs (not MVT::Other).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206150 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 05:59:09 +00:00
Chandler Carruth
17f9c2e35b [Allocator] Make the underlying allocator a template instead of an
abstract interface. The only user of this functionality is the JIT
memory manager and it is quite happy to have a custom type here. This
removes a virtual function call and a lot of unnecessary abstraction
from the common case where this is just a *very* thin vaneer around
a call to malloc.

Hopefully still no functionality changed here. =]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206149 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 05:11:27 +00:00
Chandler Carruth
e20c45d2d8 Fix the compile from r206147 in release builds by moving a variable
declaration outside of #ifndef NDEBUG -- its used elsewhere. Sorry for
the noise.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206148 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 04:46:30 +00:00
Chandler Carruth
cb7ead25c2 [Allocator] Switch the BumpPtrAllocator to use a vector of pointers to
slabs rather than embedding a singly linked list in the slabs
themselves. This has a few advantages:

- Better utilization of the slab's memory by not wasting 16-bytes at the
  front.
- Simpler allocation strategy by not having a struct packed at the
  front.
- Avoids paging every allocated slab in just to traverse them for
  deallocating or dumping stats.

The latter is the really nice part. Folks have complained from time to
time bitterly that tearing down a BumpPtrAllocator, even if it doesn't
run any destructors, pages in all of the memory allocated. Now it won't.
=]

Also resolves a FIXME with the scaling of the slab sizes. The scaling
now disregards specially sized slabs for allocations larger than the
threshold.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206147 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 03:55:11 +00:00
Saleem Abdulrasool
1b1e4e27e1 tools: simplify symbol handling in objdump
Rather than switching behaviour on whether a previous symbol has an auxiliary
symbol record for the next count of elements, simply iterate over the auxiliary
symbols right after processing the current symbol entry.  This makes the
behaviour much simpler to follow and similar to llvm-readobj and yaml2obj.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206146 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 02:37:28 +00:00
Saleem Abdulrasool
67635a7f8d tools: address possible non-null terminated filenames
If a filename is a multiple of 18 characters, there will be no null-terminator.
This will result in an invalid access by the constructed StringRef.  Add a test
case to exercise this and fix that handling.  Address this same vulnerability in
llvm-readobj as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206145 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 02:37:23 +00:00
Serge Pavlov
aa827a513c Use APInt arithmetic, fixed typo. Thanks to Benjamin Kramer for noticing that.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206144 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 02:20:19 +00:00
Craig Topper
01b72784c9 Add back some typecasts I accidentally lost in r206142.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206143 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 01:11:34 +00:00
Craig Topper
4ba844388c [C++11] More 'nullptr' conversion. In some cases just using a boolean check instead of comparing to nullptr.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206142 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-14 00:51:57 +00:00
Hal Finkel
b9ed50cf17 [PowerPC] [Constant Hoisting] Enable constant hoisting on PPC
Implements the various TTI functions to enable constant hoisting on PPC. The
only significant test-suite change is this:

MultiSource/Benchmarks/VersaBench/bmm/bmm - 20% speedup
(which essentially reverses the slowdown from r206120).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206141 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 23:02:40 +00:00
Saleem Abdulrasool
a94a6829c1 tools: remove duplication of coff_aux_file
Now that COFF::coff_aux_file is defined, use that rather than redefining the
type locally.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206140 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 22:54:15 +00:00
Saleem Abdulrasool
1a16898813 tools: avoid a string duplication
The auxiliary file records are contiguous and only contain the filename.
Construct a StringRef directly rather than copying to a temporary buffer.

Suggested by majnemer on IRC!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206139 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 22:54:11 +00:00
Saleem Abdulrasool
42ad510d31 MC: check machine magic when applying offset adjustments
The values for the relocation type can (and do) overlap across various
architectures.  When performing an adjustment of the emitted relocation in the
final object file, check that the file magic matches the target for which the
relocation type is valid (e.g. a I386 relocation is only applied to an X86
object file, and an AMD64 relocation is only applied to an X86_64 object file).
This was noticed while adding support for ARM WinCOFF object file emission.

A test case for this is not really possible as the values for REL32 do not
overlap on I386 and AMD64, which is why this was never noticed in practice.  The
ARM WinCOFF emission is not yet ready to merge into the tree.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206138 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 20:47:55 +00:00
Serge Pavlov
ea0ea63773 Recognize test for overflow in integer multiplication.
If multiplication involves zero-extended arguments and the result is
compared as in the patterns:

    %mul32 = trunc i64 %mul64 to i32
    %zext = zext i32 %mul32 to i64
    %overflow = icmp ne i64 %mul64, %zext
or
    %overflow = icmp ugt i64 %mul64 , 0xffffffff

then the multiplication may be replaced by call to umul.with.overflow.
This change fixes PR4917 and PR4918.

Differential Revision: http://llvm-reviews.chandlerc.com/D2814


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206137 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 18:23:41 +00:00
Hal Finkel
6a34916fbf [PowerPC] Fix rlwimi isel when mask is not constant
We had been using the known-zero values of the operand of the or to construct
the mask for an rlwimi; this is not quite correct, but fine when the mask is
constant. When the mask is constant, then the known zeros of the operand must
be a superset of the zeros in the mask. However, when the mask is not a
constant, then there might be bits in the operand that are not known to be zero
that, at runtime, might be zero in the mask. Therefore, we check that any bits
not known to be zero *are* known to be one in the mask. Otherwise, we can't
fold the mask with the or and shift.

This was revealed as a miscompile of
MultiSource/Benchmarks/BitBench/drop3/drop3 when I started experimenting with
constant hoisting.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206136 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 17:10:58 +00:00
David Blaikie
f85e6da6d0 Fix instruction debug info location during legalization
I found this from a particular GDB test suite case of inlining
(something similar is provided as a test case) but came across a few
other related cases (other callers of the same functions, and one other
instance of the same coding mistake in a separate function).

I'm not sure what the best way to test this is (let alone to cover the
other cases I discovered), so hopefully this sufficies - open to ideas.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206130 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 06:39:55 +00:00
Craig Topper
4266ae8067 [C++11] More 'nullptr' conversion or in some cases just using a boolean check instead of comparing to nullptr.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206129 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 04:57:38 +00:00
Lang Hames
4e510c10b5 [X86] unique_ptr'ify one of X86GenericDisassembler's members.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206127 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 04:09:16 +00:00
Saleem Abdulrasool
b77944d5ff tools: teach objdump about FILE aux records
Add support for file auxiliary symbol entries in COFF symbol tables.  A COFF
symbol table with a FILE entry is followed by sizeof(__FILE__) / 18 auxiliary
symbol records which contain the filename.  Read them and form the original
filename that the record contains.  Then display the name in the output.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206126 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-13 03:11:08 +00:00
Hal Finkel
f4c3a5601a [PowerPC] Implement some additional TLI callbacks
Add implementations of:
  bool isLegalICmpImmediate(int64_t Imm) const
  bool isLegalAddImmediate(int64_t Imm) const
  bool isTruncateFree(Type *Ty1, Type *Ty2) const
  bool isTruncateFree(EVT VT1, EVT VT2) const
  bool shouldConvertConstantLoadToIntImm(const APInt &Imm, Type *Ty) const

Unfortunately, this regresses counter-register-based loop formation because
some of the loops now end up in forms were SE cannot compute loop counts.
However, nevertheless, the test-suite results favor committing:

SingleSource/Benchmarks/BenchmarkGame/puzzle: 26% speedup
MultiSource/Benchmarks/FreeBench/analyzer/analyzer: 21% speedup
MultiSource/Benchmarks/MiBench/automotive-susan/automotive-susan: 20% speedup
SingleSource/Benchmarks/Polybench/linear-algebra/kernels/trisolv/trisolv: 19% speedup
SingleSource/Benchmarks/Polybench/linear-algebra/kernels/gesummv/gesummv: 15% speedup
MultiSource/Benchmarks/FreeBench/pcompress2/pcompress2: 2% speedup

MultiSource/Benchmarks/VersaBench/bmm/bmm: 26% slowdown

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206120 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 21:52:38 +00:00
Joerg Sonnenberger
df40801dff Fix test syntax to work with non-bash /bin/sh.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206119 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 21:13:41 +00:00
Benjamin Kramer
2c2ba4c605 Spell the specialization namespace correctly.
Not sure why clang didn't diagnose this (GCC does).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206117 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 18:45:24 +00:00
Benjamin Kramer
f6330705bd Make helper static and place random global into the llvm namespace.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206116 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 18:39:57 +00:00
Saleem Abdulrasool
2ef2fd45b8 tools: remove unnecessary copy of array_lengthof
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206115 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 18:04:27 +00:00
Benjamin Kramer
15c435a367 Retire llvm::array_endof in favor of non-member std::end.
While there make array_lengthof constexpr if we have support for it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206112 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 16:15:53 +00:00
Benjamin Kramer
479151ab7f Move MDBuilder's methods out of line.
Making them inline was a historical accident, they're neither hot nor
templated.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206109 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 14:26:59 +00:00
David Blaikie
937c689cdc PR13337: Omit DW_TAG_restrict_type when compiling for DWARF2
DWARF3 introduced DW_TAG_restrict_type, so avoid using it in prior
versions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206105 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 05:35:59 +00:00
Richard Trieu
6a871a361d Add extra checks to mvn.ll test to prevent the "f1" check from matching
on a directory name instead of the function name.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206104 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 04:47:04 +00:00
Adrian Prantl
4debb77326 Revert "Debug info: (bugfix) C++ C/Dtors can be compiled to multiple functions,"
This reverts commit 206096 while I investigate why this broke the gdb
buildbot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206103 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 04:25:02 +00:00
Juergen Ributzka
2aa7106dd6 [ARM64] Never hoist the shift value of a shift instruction.
There is no need to check if we want to hoist the immediate value of an
shift instruction. Simply return TCC_Free right away.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206101 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 02:53:51 +00:00
Juergen Ributzka
940b67465d [ARM64] Fix the cost model for cheap large constants.
Originally the cost model would give up for large constants and just return the
maximum cost. This is not what we want for constant hoisting, because some of
these constants are large in bitwidth, but are still cheap to materialize.

This commit fixes the cost model to either return TCC_Free if the cost cannot be
determined, or accurately calculate the cost even for large constants
(bitwidth > 128).

This fixes <rdar://problem/16591573>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206100 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 02:36:28 +00:00
David Blaikie
9c9ef05445 Use dwarf::Tag rather than unsigned for DIE::Tag to make debugging easier.
Nice to be able to just print out the Tag and have the debugger print
dwarf::DW_TAG_subprogram or whatever, rather than an int.

It's a bit finicky (for example DIDescriptor::getTag still returns
unsigned) because some places still handle real dwarf tags + our fake
tags (one day we'll remove the fake tags, hopefully).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206098 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 02:24:04 +00:00
Eric Christopher
ac871732b5 Remove documentation for a deleted pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206097 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 02:09:17 +00:00
Adrian Prantl
515f2f9611 Debug info: (bugfix) C++ C/Dtors can be compiled to multiple functions,
therefore, their declaration cannot have one DW_AT_linkage_name.
The specific instances however can and should have that attribute.

This patch reorders the code in DwarfUnit::getOrCreateSubprogramDIE()
to emit linkage names for C/Dtors.

rdar://problem/16362674.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206096 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 01:44:42 +00:00
Jim Grosbach
d023e14a5e llc: Add support for -mcpu=native.
When -mcpu=native is passed, autodetect the host CPU and pass that
as the CPU name to the TargetMachine factory method.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206095 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 01:34:31 +00:00
Jim Grosbach
6bb00df864 X86: Remove TargetMachine CPU auto-detection.
This logic is properly in the realm of whatever is creating the
TargetMachine. This makes plain 'llc foo.ll' consistent across
heterogenous machines.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206094 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 01:34:29 +00:00
Hal Finkel
e13b87c08d Reenable use of TBAA during CodeGen
We had disabled use of TBAA during CodeGen (even when otherwise using AA)
because the ptrtoint/inttoptr used by CGP for address sinking caused BasicAA to
miss basic type punning that it should catch (and, thus, we'd fail to override
TBAA when we should).

However, when AA is in use during CodeGen, CGP now uses normal GEPs and
bitcasts, instead of ptrtoint/inttoptr, when doing address sinking. As a
result, BasicAA should be able to make us do the right thing in the face of
type-punning, and it seems safe to enable use of TBAA again. self-hosting seems
fine on PPC64/Linux on the P7, with TBAA enabled and -misched=shuffle.

Note: We still don't update TBAA when merging stack slots, although because
BasicAA should now catch all such cases, this is no longer a blocking issue.
Nevertheless, I plan to commit code to deal with this properly in the near
future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206093 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 01:26:00 +00:00
Hal Finkel
24517d023f Add the ability to use GEPs for address sinking in CGP
The current memory-instruction optimization logic in CGP, which sinks parts of
the address computation that can be adsorbed by the addressing mode, does this
by explicitly converting the relevant part of the address computation into
IR-level integer operations (making use of ptrtoint and inttoptr). For most
targets this is currently not a problem, but for targets wishing to make use of
IR-level aliasing analysis during CodeGen, the use of ptrtoint/inttoptr is a
problem for two reasons:
  1. BasicAA becomes less powerful in the face of the ptrtoint/inttoptr
  2. In cases where type-punning was used, and BasicAA was used
     to override TBAA, BasicAA may no longer do so. (this had forced us to disable
     all use of TBAA in CodeGen; something which we can now enable again)

This (use of GEPs instead of ptrtoint/inttoptr) is not currently enabled by
default (except for those targets that use AA during CodeGen), and so aside
from some PowerPC subtargets and SystemZ, there should be no change in
behavior. We may be able to switch completely away from the ptrtoint/inttoptr
sinking on all targets, but further testing is required.

I've doubled-up on a number of existing tests that are sensitive to the
address sinking behavior (including some store-merging tests that are
sensitive to the order of the resulting ADD operations at the SDAG level).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206092 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 00:59:48 +00:00
Chad Rosier
a5b2772d31 [AArch64] Implement the isLegalAddressingMode and getScalingFactorCost APIs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206089 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-12 00:14:23 +00:00
Duncan P. N. Exon Smith
23a6033104 blockfreq: Document BlockFrequencyInfo terminology
Documents terminology used in the forthcoming rewrite of
BlockFrequencyInfo.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206086 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-11 23:21:07 +00:00
Duncan P. N. Exon Smith
670060dddf blockfreq: Rename BlockProbInfoT to BranchProbabilityInfoT
No functionality change.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206085 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-11 23:21:04 +00:00
Duncan P. N. Exon Smith
3c371e995a blockfreq: Remove unnecessary template parameters
Moves redundant template parameters into an implementation detail of
BlockFrequencyInfoImpl.

No functionality change.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206084 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-11 23:21:02 +00:00
Duncan P. N. Exon Smith
e9139c68d4 blockfreq: Rename BlockFrequencyImpl to BlockFrequencyInfoImpl
This is a shared implementation class for BlockFrequencyInfo and
MachineBlockFrequencyInfo, not for BlockFrequency, a related (but
distinct) class.

No functionality change.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206083 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-11 23:20:58 +00:00