Commit Graph

99668 Commits

Author SHA1 Message Date
Matheus Almeida
67244395fb [mips][msa] Add fill.d instruction.
This instruction is only available on Mips64 cores
that implement the MSA ASE.

llvm-svn: 200400
2014-01-29 15:12:02 +00:00
Matheus Almeida
6fd8deacb5 [mips][msa] CHECK-DAG-ize MSA 2r_vector_scalar.ll test.
This update is a preparation for the addition of Mips64 MSA tests.

No functional changes.

llvm-svn: 200399
2014-01-29 14:32:03 +00:00
Matheus Almeida
3e07e293c7 [mips][msa] Add copy_{u,s}.d.
These instructions are only available on Mips64 cores
that implement the MSA ASE.

llvm-svn: 200398
2014-01-29 14:05:28 +00:00
Matheus Almeida
8cac709f4b [mips][msa] CHECK-DAG-ize MSA elm_copy.ll test.
This update is a preparation for the addition of Mips64 MSA tests.

No functional changes.

llvm-svn: 200395
2014-01-29 13:51:34 +00:00
Chandler Carruth
6ba48b6c38 [LPM] Fix PR18643, another scary place where loop transforms failed to
preserve loop simplify of enclosing loops.

The problem here starts with LoopRotation which ends up cloning code out
of the latch into the new preheader it is buidling. This can create
a new edge from the preheader into the exit block of the loop which
breaks LoopSimplify form. The code tries to fix this by splitting the
critical edge between the latch and the exit block to get a new exit
block that only the latch dominates. This sadly isn't sufficient.

The exit block may be an exit block for multiple nested loops. When we
clone an edge from the latch of the inner loop to the new preheader
being built in the outer loop, we create an exiting edge from the outer
loop to this exit block. Despite breaking the LoopSimplify form for the
inner loop, this is fine for the outer loop. However, when we split the
edge from the inner loop to the exit block, we create a new block which
is in neither the inner nor outer loop as the new exit block. This is
a predecessor to the old exit block, and so the split itself takes the
outer loop out of LoopSimplify form. We need to split every edge
entering the exit block from inside a loop nested more deeply than the
exit block in order to preserve all of the loop simplify constraints.

Once we try to do that, a problem with splitting critical edges
surfaces. Previously, we tried a very brute force to update LoopSimplify
form by re-computing it for all exit blocks. We don't need to do this,
and doing this much will sometimes but not always overlap with the
LoopRotate bug fix. Instead, the code needs to specifically handle the
cases which can start to violate LoopSimplify -- they aren't that
common. We need to see if the destination of the split edge was a loop
exit block in simplified form for the loop of the source of the edge.
For this to be true, all the predecessors need to be in the exact same
loop as the source of the edge being split. If the dest block was
originally in this form, we have to split all of the deges back into
this loop to recover it. The old mechanism of doing this was
conservatively correct because at least *one* of the exiting blocks it
rewrote was the DestBB and so the DestBB's predecessors were fixed. But
this is a much more targeted way of doing it. Making it targeted is
important, because ballooning the set of edges touched prevents
LoopRotate from being able to split edges *it* needs to split to
preserve loop simplify in a coherent way -- the critical edge splitting
would sometimes find the other edges in need of splitting but not
others.

Many, *many* thanks for help from Nick reducing these test cases
mightily. And helping lots with the analysis here as this one was quite
tricky to track down.

llvm-svn: 200393
2014-01-29 13:16:53 +00:00
Evgeniy Stepanov
7f5888aaec Remove -arm-enable-ehabi from Android build rules. It's enabled by default.
llvm-svn: 200391
2014-01-29 12:36:18 +00:00
Renato Golin
7e3b0bd21b Document EHABI enabled by default
llvm-svn: 200390
2014-01-29 12:04:13 +00:00
Renato Golin
6ca0034624 Enable EHABI by default
After all hard work to implement the EHABI and with the test-suite
passing, it's time to turn it on by default and allow users to
disable it as a work-around while we fix the eventual bugs that show
up.

This commit also remove the -arm-enable-ehabi-descriptors, since we
want the tables to be printed every time the EHABI is turned on
for non-Darwin ARM targets.

Although MCJIT EHABI is not working yet (needs linking with the right
libraries), this commit also fixes some relocations on MCJIT regarding
the EH tables/lib calls, and update some tests to avoid using EH tables
when none are needed.

The EH tests in the test-suite that were previously disabled on ARM
now pass with these changes, so a follow-up commit on the test-suite
will re-enable them.

llvm-svn: 200388
2014-01-29 11:50:56 +00:00
David Majnemer
b3046f7fc8 MC: Reorganize macro MC test along dialect lines
This commit seeks to do two things:
 - Run the surfeit of tests under the Darwin dialect.  This ends up
   affecting tests which assumed that spaces could deliminate arguments.
 - The GAS dialect tests should limit their surface area to things that
   could plausibly work under GAS. For example, Darwin style arguments
   have no business being in such a test.

llvm-svn: 200383
2014-01-29 09:18:43 +00:00
Kostya Serebryany
cbec3e2767 [asan] simplify a test
llvm-svn: 200378
2014-01-29 07:35:43 +00:00
Craig Topper
237e63c6ee Delete the Matchers stored in the SmallVectors in SwitchOpcodeMatcher/SwitchTypeMatcher.
llvm-svn: 200377
2014-01-29 07:06:07 +00:00
Venkatraman Govindaraju
a50ca1f645 [Sparc] Use %r_disp32 for pc_rel entries in FDE as well.
This makes MCAsmInfo::getExprForFDESymbol() a virtual function and overrides it in SparcMCAsmInfo.

llvm-svn: 200376
2014-01-29 06:59:20 +00:00
NAKAMURA Takumi
782750fa03 Revert r200340, "Add line table debug info to COFF files when using a win32 triple."
It was incompatible with --target=i686-win32.

llvm-svn: 200375
2014-01-29 06:05:38 +00:00
Rafael Espindola
fc2dbfa864 Normalize the style in llvm-nm.cpp.
It had grown fairly inconsistent. I am about to change it quite a bit to also
use the object api when handling IR files.

llvm-svn: 200374
2014-01-29 04:56:19 +00:00
Venkatraman Govindaraju
1541483555 [Sparc] Use %r_disp32 for pc_rel entries in gcc_except_table and eh_frame.
Otherwise, assembler (gas) fails to assemble them with error message "operation
combines symbols in different segments". This is because MC computes
pc_rel entries with subtract expression between labels from different sections.

llvm-svn: 200373
2014-01-29 04:51:35 +00:00
Chandler Carruth
ed726e1be7 [LPM] Fix PR18642, a pretty nasty bug in IndVars that "never mattered"
because of the inside-out run of LoopSimplify in the LoopPassManager and
the fact that LoopSimplify couldn't be "preserved" across two
independent LoopPassManagers.

Anyways, in that case, IndVars wasn't correctly preserving an LCSSA PHI
node because it thought it was rewriting (via SCEV) the incoming value
to a loop invariant value. While it may well be invariant for the
current loop, it may be rewritten in terms of an enclosing loop's
values. This in and of itself is fine, as the LCSSA PHI node in the
enclosing loop for the inner loop value we're rewriting will have its
own LCSSA PHI node if used outside of the enclosing loop. With me so
far?

Well, the current loop and the enclosing loop may share an exiting
block and exit block, and when they do they also share LCSSA PHI nodes.
In this case, its not valid to RAUW through the LCSSA PHI node.

Expected crazy test included.

llvm-svn: 200372
2014-01-29 04:40:19 +00:00
Arnold Schwaighofer
5b96c24a7a LoopVectorizer: Don't count the induction variable multiple times
When estimating register pressure, don't count the induction variable mulitple
times. It is unlikely to be unrolled. This is currently disabled and hidden
behind a flag ("enable-ind-var-reg-heur").

llvm-svn: 200371
2014-01-29 04:36:12 +00:00
Rafael Espindola
2eee1a2756 We do use pipefail these days. Update the test.
llvm-svn: 200370
2014-01-29 04:08:05 +00:00
Venkatraman Govindaraju
416f0c2389 [SparcV9] Use correct register class (I64RegClass) to hold the address of _GLOBAL_OFFSET_TABLE_ in sparcv9.
llvm-svn: 200368
2014-01-29 03:35:08 +00:00
Rafael Espindola
f6087fc40c Use a raw_stream to implement the mangler.
This is a bit more convenient for some callers, but more importantly, it is
easier to implement correctly. Doing this removes the patching of already
printed data that was used for fastcall, fixing a crash with private fastcall
symbols.

llvm-svn: 200367
2014-01-29 02:30:38 +00:00
Kevin Qin
379441a4e6 [AArch64 NEON] Lower SELECT_CC with vector operand.
When the scalar compare is between floating point and operands are
vector, we custom lower SELECT_CC to use NEON SIMD compare for
generating less instructions.

llvm-svn: 200365
2014-01-29 01:57:30 +00:00
David Woodhouse
9939400489 [ARM] Remove superfluous inline asm mode switch test
llvm-svn: 200361
2014-01-29 00:49:28 +00:00
Mark Seaborn
c53fc2913b Remove unnecessary call to pthread_mutexattr_setpshared()
The default value of this attribute is PTHREAD_PROCESS_PRIVATE, so
there's no point in calling pthread_mutexattr_setpshared() to set
that.

See: http://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_mutexattr_getpshared.html

This removes some ifdefs that tend to need to be extended for other
platforms (e.g. for NaCl).

Note that this call was in the first implementation of Mutex, added in
r22403, so it doesn't appear to have been added in response to a
performance problem.

Differential Revision: http://llvm-reviews.chandlerc.com/D2633

llvm-svn: 200360
2014-01-29 00:20:44 +00:00
David Majnemer
e40cc64844 MC: Clean up error paths in AsmParser::parseMacroArgument
Use an RAII object Instead of inserting a call to
AsmLexer::setSkipSpace(true) in all error paths.

No functional change.

llvm-svn: 200358
2014-01-29 00:07:39 +00:00
Rafael Espindola
9ade36c920 Make createObjectFile's signature a bit less error prone.
This will be better with c++11, but right now file_magic converts to bool,
which makes the api really easy to misuse.

llvm-svn: 200357
2014-01-29 00:02:26 +00:00
David Woodhouse
0cc2f6368a [Sparc] Fix breakage in r200345
Oops. Don't do build tests on patches like that with --enable-targets=x86_64

llvm-svn: 200355
2014-01-28 23:38:16 +00:00
Rafael Espindola
4550ecf07b Finish bringing file_type to the llvm style (other than its name).
I assume that the name is file_type because it is the name of a c++11 type that
we will use once we convert, but at least our current implementation can look
like llvm code.

Thanks to David Blakie for the push.

llvm-svn: 200354
2014-01-28 23:36:39 +00:00
Nick Kledzik
9c5a7fd806 Be more explicit about which overloaded variant to use. Caught by ASan!
llvm-svn: 200353
2014-01-28 23:25:13 +00:00
Rafael Espindola
64d9f0edcd Convert to the inner enum so the compiler can warn about it in switches.
llvm-svn: 200352
2014-01-28 23:15:56 +00:00
David Woodhouse
69762f7e1e Tests for mode switching
1. test that inlineasm works
2. test that relaxable instructions are re-encoded in the correct mode.

llvm-svn: 200351
2014-01-28 23:13:30 +00:00
David Woodhouse
6c8fefd999 Delete MCSubtargetInfo data members from target MCCodeEmitter classes
The subtarget info is explicitly passed to the EncodeInstruction
method and we should use that subtarget info to influence any
encoding decisions.

llvm-svn: 200350
2014-01-28 23:13:25 +00:00
David Woodhouse
a79a37b435 Propagate MCSubtargetInfo through TableGen's getBinaryCodeForInstr()
llvm-svn: 200349
2014-01-28 23:13:18 +00:00
David Woodhouse
4a4c611e36 Explictly pass MCSubtargetInfo to MCCodeEmitter::EncodeInstruction()
llvm-svn: 200348
2014-01-28 23:13:07 +00:00
David Woodhouse
b9294892a1 Keep the MCSubtargetInfo in the MCRelxableFragment class.
Needed to fix PR18303 to correctly re-encode the instruction if it
is relaxed.

We keep a copy of the MCSubtargetInfo to make sure that we are not
effected by future changes to the subtarget info coming from the
assembler (e.g. when parsing .code 16 directived).

llvm-svn: 200347
2014-01-28 23:12:53 +00:00
David Woodhouse
d7e33ceb85 Modify MCObjectStreamer EmitInstTo* interface
Add MCSubtargetInfo parameter
virtual void EmitInstToFragment(const MCInst &Inst, const MCSubtargetInfo &);
virtual void EmitInstToData(const MCInst &Inst, const MCSubtargetInfo &);

llvm-svn: 200346
2014-01-28 23:12:49 +00:00
David Woodhouse
5d0b529d58 Change MCStreamer EmitInstruction interface to take subtarget info
llvm-svn: 200345
2014-01-28 23:12:42 +00:00
Timur Iskhodzhanov
dbb8fbd46f Disable the COFF tests on non-X86 archs
llvm-svn: 200341
2014-01-28 21:47:33 +00:00
Timur Iskhodzhanov
53b4a3ded1 Add line table debug info to COFF files when using a win32 triple.
Reviewed at http://llvm-reviews.chandlerc.com/D2232

llvm-svn: 200340
2014-01-28 21:33:27 +00:00
Owen Anderson
616a1ad334 Use the proper SDK when building iOS sim builds of LLVM and/or Mac OS X builds.
llvm-svn: 200335
2014-01-28 19:57:18 +00:00
Nick Kledzik
23d28d4e2a fix templates to work with pre c++11
llvm-svn: 200333
2014-01-28 19:33:09 +00:00
Matheus Almeida
6447ece792 [mips] Fix ELF header flags.
As opposed to GCC/GAS the default ABI for Mips64 is n64.
Compatibility bit should be set if o32 ABI is used when targeting Mips64.

llvm-svn: 200332
2014-01-28 19:24:11 +00:00
Nick Kledzik
92d782a92e Add BumpPtrAllocator::allocateCopy() utilities
Makes it easy to use BumpPtrAllocator to make a copy of StringRef strings.

llvm-svn: 200331
2014-01-28 19:21:27 +00:00
Gautam Chakrabarti
dcd42d2079 [NVPTX] Fix emitting aggregate parameters
The code was missing the case for aggregate parameters and
hence was emitting them as .b0 type. Also fixed a couple
of comments.

llvm-svn: 200325
2014-01-28 18:35:29 +00:00
Andrea Di Biagio
47b83fb85b [X86] Add extra rules for combining vselect dag nodes into movsd.
This improves the fix committed at revision 199683 adding the
following new target specific combine rules:

1) fold (v4i32: vselect <0,0,-1,-1>, A, B) ->
        (v4i32 (bitcast (movsd (v2i64 (bitcast A)), (v2i64 (bitcast B))) ))

2) fold (v4f32: vselect <0,0,-1,-1>, A, B) ->
        (v4f32 (bitcast (movsd (v2f64 (bitcast A)), (v2f64 (bitcast B))) ))

3) fold (v4i32: vselect <-1,-1,0,0>, A, B) ->
        (v4i32 (bitcast (movsd (v2i64 (bitcast B)), (v2i64 (bitcast A))) ))

4) fold (v4f32: vselect <-1,-1,0,0>, A, B) ->
        (v4f32 (bitcast (movsd (v2i64 (bitcast B)), (v2i64 (bitcast A))) ))

llvm-svn: 200324
2014-01-28 18:14:21 +00:00
Adrian Prantl
5d80581891 typo
llvm-svn: 200323
2014-01-28 18:13:47 +00:00
Rafael Espindola
e8856107f0 Fix pr14893.
When simplifycfg moves an instruction, it must drop metadata it doesn't know
is still valid with the preconditions changes. In particular, it must drop
the range and tbaa metadata.

The patch implements this with an utility function to drop all metadata not
in a white list.

llvm-svn: 200322
2014-01-28 16:56:46 +00:00
Aaron Ballman
ba4fc9f4eb The llvm_headers_do_not_build project needs to be excluded from the default build, otherwise it gets built (at least in Visual Studio 2013).
Thanks to chapuni200000 for help with this in IRC!

llvm-svn: 200321
2014-01-28 16:07:10 +00:00
Andrea Di Biagio
9e72586184 [DAGCombiner] Avoid introducing an illegal build_vector when folding a sign_extend.
Make sure that we don't introduce illegal build_vector dag nodes
when trying to fold a sign_extend of a build_vector.

This fixes a regression introduced by r200234.
Added test CodeGen/X86/fold-vector-sext-crash.ll
to verify that llc no longer crashes with an assertion failure
due to an illegal build_vector of type MVT::v4i64.

Thanks to Ilia Filippov for spotting this regression and for
providing a reproducible test case.

llvm-svn: 200313
2014-01-28 12:53:56 +00:00
NAKAMURA Takumi
d6ac1d50c8 [CMake] llvm_update_compile_flags(name) doesn't require source files. TARGET PROPERTY SOURCES has them.
llvm-svn: 200311
2014-01-28 11:40:04 +00:00
Iain Sandoe
f16c0c72d6 Provide a stub Target Streamer implementation for PPC MachO
At present, this handles .tc (error) and needs to be expanded to deal properly with .machine

llvm-svn: 200309
2014-01-28 11:03:17 +00:00