Douglas Gregor [Tue, 11 Dec 2012 22:25:26 +0000 (22:25 +0000)]
Don't use 'touch' in this test; it may be the cause of Windows-related failures.
llvm-svn: 169922
Douglas Gregor [Tue, 11 Dec 2012 22:11:52 +0000 (22:11 +0000)]
Use @import rather than @__experimental_modules_import, since the
latter is rather a mess to type.
llvm-svn: 169919
Guy Benyei [Tue, 11 Dec 2012 21:38:14 +0000 (21:38 +0000)]
Add SPIR32/SPIR64 targets to Clang
llvm-svn: 169917
Nadav Rotem [Tue, 11 Dec 2012 21:30:14 +0000 (21:30 +0000)]
PR14574. Fix a bug in the code that calculates the mask the converted PHIs in if-conversion.
llvm-svn: 169916
Tom Stellard [Tue, 11 Dec 2012 21:25:42 +0000 (21:25 +0000)]
Add R600 backend
A new backend supporting AMD GPUs: Radeon HD2XXX - HD7XXX
llvm-svn: 169915
Bill Schmidt [Tue, 11 Dec 2012 20:30:11 +0000 (20:30 +0000)]
This patch implements the general dynamic TLS model for 64-bit PowerPC.
Given a thread-local symbol x with global-dynamic access, the generated
code to obtain x's address is:
Instruction Relocation Symbol
addis ra,r2,x@got@tlsgd@ha R_PPC64_GOT_TLSGD16_HA x
addi r3,ra,x@got@tlsgd@l R_PPC64_GOT_TLSGD16_L x
bl __tls_get_addr(x@tlsgd) R_PPC64_TLSGD x
R_PPC64_REL24 __tls_get_addr
nop
<use address in r3>
The implementation borrows from the medium code model work for introducing
special forms of ADDIS and ADDI into the DAG representation. This is made
slightly more complicated by having to introduce a call to the external
function __tls_get_addr. Using the full call machinery is overkill and,
more importantly, makes it difficult to add a special relocation. So I've
introduced another opcode GET_TLS_ADDR to represent the function call, and
surrounded it with register copies to set up the parameter and return value.
Most of the code is pretty straightforward. I ran into one peculiarity
when I introduced a new PPC opcode BL8_NOP_ELF_TLSGD, which is just like
BL8_NOP_ELF except that it takes another parameter to represent the symbol
("x" above) that requires a relocation on the call. Something in the
TblGen machinery causes BL8_NOP_ELF and BL8_NOP_ELF_TLSGD to be treated
identically during the emit phase, so this second operand was never
visited to generate relocations. This is the reason for the slightly
messy workaround in PPCMCCodeEmitter.cpp:getDirectBrEncoding().
Two new tests are included to demonstrate correct external assembly and
correct generation of relocations using the integrated assembler.
Comments welcome!
Thanks,
Bill
llvm-svn: 169910
Hal Finkel [Tue, 11 Dec 2012 19:59:32 +0000 (19:59 +0000)]
Add -fslp-vectorize to enable bb-vectorize
Add -fslp-vectorize (with -ftree-slp-vectorize as an alias for gcc compatibility)
to provide a way to enable the basic-block vectorization pass. This uses the same
acronym as gcc, superword-level parallelism (SLP), also common in the literature,
to refer to basic-block vectorization.
Nadav suggested this as a follow-up to the adding of -fvectorize.
llvm-svn: 169909
Fariborz Jahanian [Tue, 11 Dec 2012 19:58:01 +0000 (19:58 +0000)]
objective-C blocks: Check for record type when deciding if
byref variable requires extended layout info. to prevent
a crash involving arrays declared __block. // rdar://
12787751
llvm-svn: 169908
Eric Christopher [Tue, 11 Dec 2012 19:42:09 +0000 (19:42 +0000)]
Update some comments.
llvm-svn: 169907
Nadav Rotem [Tue, 11 Dec 2012 18:58:10 +0000 (18:58 +0000)]
Loop Vectorize: optimize the vectorization of trunc(induction_var). The truncation is now done on scalars.
llvm-svn: 169904
Douglas Gregor [Tue, 11 Dec 2012 18:53:07 +0000 (18:53 +0000)]
Don't complain about incomplete implementations for methods that are
unavailable due to availability attributes. <rdar://problem/
12798237>
llvm-svn: 169903
Benjamin Kramer [Tue, 11 Dec 2012 18:00:22 +0000 (18:00 +0000)]
Speed up looking up static diagnostic infos.
Instead of doing a binary search over the whole diagnostic table (which weighs
a whopping 48k on x86_64), use the existing enums to compute the index in the
table. This avoids loading any unneeded data from the table and avoids littering
CPU caches with it. This code is in a hot path for code with many diagnostics.
1% speedup on -fsyntax-only gcc.c, which emits a lot of warnings.
llvm-svn: 169890
Chad Rosier [Tue, 11 Dec 2012 17:46:33 +0000 (17:46 +0000)]
[driver] Add the -ftree-vectorize flag as a alias of -fvectorize to maintain
compatibility with gcc.
rdar://
12839978
llvm-svn: 169888
Matt Beaumont-Gay [Tue, 11 Dec 2012 17:37:46 +0000 (17:37 +0000)]
Fix a copypasto bug. Also rename the parameter in question to not shadow the
'file' builtin, and fix up a docstring a little.
Hat tip to Sebastian Kreft Carreno at Google for noticing the bug.
llvm-svn: 169887
Eli Bendersky [Tue, 11 Dec 2012 17:16:00 +0000 (17:16 +0000)]
Remove the RelaxAll overrule in MCAssembler::fixupNeedsRelaxation,
because that method is only getting called for MCInstFragment. These
fragments aren't even generated when RelaxAll is set, which is why the
flag reference here is superfluous. Removing it simplifies the code
with no harmful effects.
An assertion is added higher up to make sure this path is never
reached.
llvm-svn: 169886
Chad Rosier [Tue, 11 Dec 2012 17:12:28 +0000 (17:12 +0000)]
[driver] Add the -fvectorize flag to enable the loop vectorization passes.
rdar://
12839978
llvm-svn: 169885
Rafael Espindola [Tue, 11 Dec 2012 16:36:02 +0000 (16:36 +0000)]
Use an ArrayRef instead of a std::vector&.
llvm-svn: 169881
Joel Jones [Tue, 11 Dec 2012 16:10:25 +0000 (16:10 +0000)]
Add comment for load folding
llvm-svn: 169880
Dmitri Gribenko [Tue, 11 Dec 2012 15:29:37 +0000 (15:29 +0000)]
Documentation: convert Passes.html to reST.
Since now we have an autogenerated TOC, a manually written table of all passes
was removed.
Patch by Anthony Mykhailenko with small fixes by me.
llvm-svn: 169867
Alexander Kornienko [Tue, 11 Dec 2012 15:28:09 +0000 (15:28 +0000)]
StmtDumper coding style changes
Summary:
A few small coding style changes for StmtDumper, including:
- rename Dump* methods to dump*
- uninline some methods
- comment fixes
- whitespace fixes
Please review and commit if okay.
Reviewers: alexfh
Reviewed By: alexfh
CC: cfe-commits, alexfh, klimek
Differential Revision: http://llvm-reviews.chandlerc.com/D190
llvm-svn: 169866
Alexander Kornienko [Tue, 11 Dec 2012 15:20:44 +0000 (15:20 +0000)]
Rename StmtDumper::DumpSubTree() to dumpStmt()
Summary:
Also rename DumpDeclarator() to dumpDecl(). Once Decl dumping is added, these will be the two main methods of the class, so this is just for consistency in naming.
There was a DumpStmt() method already, but there was no point in having it, so I have merged it into VisitStmt(). Similarly, DumpExpr() is merged into VisitExpr().
Reviewers: alexfh
Reviewed By: alexfh
CC: cfe-commits, alexfh
Differential Revision: http://llvm-reviews.chandlerc.com/D156
llvm-svn: 169865
Kostya Serebryany [Tue, 11 Dec 2012 14:41:31 +0000 (14:41 +0000)]
[asan] more code for asan_allocator2: basic Allocate/Deallocate code
llvm-svn: 169864
NAKAMURA Takumi [Tue, 11 Dec 2012 13:14:16 +0000 (13:14 +0000)]
llvm/test/TableGen: Remove XFAIL:vg_leak in dozen of tests, according to llvm-x86_64-linux-vg_leak.
llvm-svn: 169862
Evgeniy Stepanov [Tue, 11 Dec 2012 12:52:38 +0000 (12:52 +0000)]
Add MemorySanitizer runtime to check_lint.sh.
llvm-svn: 169861
Evgeniy Stepanov [Tue, 11 Dec 2012 12:44:43 +0000 (12:44 +0000)]
[msan] Add a missing file from r169858.
llvm-svn: 169860
Evgeniy Stepanov [Tue, 11 Dec 2012 12:34:09 +0000 (12:34 +0000)]
[msan] Use explicitely aligned stores and loads with function argument shadow.
Use explicitely aligned store and load instructions to deal with argument and
retval shadow. This matters when an argument's alignment is higher than
__msan_param_tls alignment (which is the case with __m128i).
llvm-svn: 169859
Evgeniy Stepanov [Tue, 11 Dec 2012 12:27:27 +0000 (12:27 +0000)]
[msan] MemorySanitizer runtime.
Initial commit of the MemorySanitizer runtime library.
llvm-svn: 169858
Timur Iskhodzhanov [Tue, 11 Dec 2012 12:24:41 +0000 (12:24 +0000)]
ASan: fix lint problems introduced
llvm-svn: 169857
Timur Iskhodzhanov [Tue, 11 Dec 2012 12:23:00 +0000 (12:23 +0000)]
Hopefully fix the Windows build (2)
llvm-svn: 169856
Timur Iskhodzhanov [Tue, 11 Dec 2012 12:03:06 +0000 (12:03 +0000)]
Hopefully fix the Windows build by not including intrin.h
llvm-svn: 169855
Patrik Hagglund [Tue, 11 Dec 2012 11:14:33 +0000 (11:14 +0000)]
Revert EVT->MVT changes, r169836-169851, due to buildbot failures.
llvm-svn: 169854
Chandler Carruth [Tue, 11 Dec 2012 11:05:15 +0000 (11:05 +0000)]
Holding my nose and moving the accumulation routine to GEPOperator
instead of the instruction. I've left a forwarding wrapper for the
instruction so users with the instruction don't need to create
a GEPOperator themselves.
This lets us remove the copy of this code in instsimplify.
I've looked at most of the other copies of similar code, and this is the
only one I've found that is actually exactly the same. The one in
InlineCost is very close, but it requires re-mapping non-constant
indices through the cost analysis value simplification map. I could add
direct support for this to the generic routine, but it seems overly
specific.
llvm-svn: 169853
Chandler Carruth [Tue, 11 Dec 2012 10:29:10 +0000 (10:29 +0000)]
Hoist the GEP constant address offset computation to a common home on
the GEP instruction class.
This is part of the continued refactoring and cleaning of the
infrastructure used by SROA. This particular operation is also done in
a few other places which I'll try to refactor to share this
implementation.
llvm-svn: 169852
Patrik Hagglund [Tue, 11 Dec 2012 10:24:48 +0000 (10:24 +0000)]
Change RegVT in BitTestBlock and RegsForValue, to contain MVTs,
instead of EVTs.
llvm-svn: 169851
Patrik Hagglund [Tue, 11 Dec 2012 10:20:51 +0000 (10:20 +0000)]
Change TargetLowering::getTypeForExtArgOrReturn to take and return
MVTs, instead of EVTs.
Accordingly, add bitsLT (and similar) to MVT.
llvm-svn: 169850
Patrik Hagglund [Tue, 11 Dec 2012 10:16:19 +0000 (10:16 +0000)]
Change a parameter of TargetLowering::getVectorTypeBreakdown to MVT,
from EVT.
llvm-svn: 169849
Patrik Hagglund [Tue, 11 Dec 2012 10:09:23 +0000 (10:09 +0000)]
Change TargetLowering::RegisterTypeForVT to contain MVTs, instead of
EVTs.
llvm-svn: 169848
Patrik Hagglund [Tue, 11 Dec 2012 10:05:04 +0000 (10:05 +0000)]
Change TargetLowering::TransformToType to contain MVTs, instead of
EVTs.
llvm-svn: 169847
Patrik Hagglund [Tue, 11 Dec 2012 10:00:35 +0000 (10:00 +0000)]
Change TargetLowering::getRepRegClassCostFor, getIndexedLoadAction,
getIndexedStoreAction, and addRegisterClass to take an MVT, instead
of EVT.
llvm-svn: 169846
Patrik Hagglund [Tue, 11 Dec 2012 09:57:18 +0000 (09:57 +0000)]
Change TargetLowering::findRepresentativeClass to take an MVT, instead
of EVT.
llvm-svn: 169845
Patrik Hagglund [Tue, 11 Dec 2012 09:54:23 +0000 (09:54 +0000)]
Change TargetLowering::getTypeToPromoteTo to take and return MVTs,
instead of EVTs.
llvm-svn: 169844
Patrik Hagglund [Tue, 11 Dec 2012 09:51:27 +0000 (09:51 +0000)]
Change TargetLowering::isCondCodeLegal to take an MVT, instead of EVT.
llvm-svn: 169843
Patrik Hagglund [Tue, 11 Dec 2012 09:48:14 +0000 (09:48 +0000)]
Change TargetLowering::getCondCodeAction to take an MVT, instead of
EVT.
llvm-svn: 169842
Patrik Hagglund [Tue, 11 Dec 2012 09:42:24 +0000 (09:42 +0000)]
Change TargetLowering::getTruncStoreAction to take MVTs, instead of EVTs.
llvm-svn: 169841
Patrik Hagglund [Tue, 11 Dec 2012 09:39:09 +0000 (09:39 +0000)]
Change TargetLowering::getLoadExtAction to take an MVT, instead of EVT.
llvm-svn: 169840
Patrik Hagglund [Tue, 11 Dec 2012 09:32:56 +0000 (09:32 +0000)]
Change TargetLowering::setTypeAction to take an MVT, instead fo EVT.
llvm-svn: 169839
Patrik Hagglund [Tue, 11 Dec 2012 09:31:43 +0000 (09:31 +0000)]
Change TargetLowering::getRepRegClassFor to take an MVT, instead of
EVT.
Accordingly, change RegDefIter to contain MVTs instead of EVTs.
llvm-svn: 169838
Patrik Hagglund [Tue, 11 Dec 2012 09:10:33 +0000 (09:10 +0000)]
Change TargetLowering::getRegClassFor to take an MVT, instead of EVT.
Accordingly, add helper funtions getSimpleValueType (in parallel to
getValueType) in SDValue, SDNode, and TargetLowering.
This is the first, in a series of patches.
llvm-svn: 169837
Kostya Serebryany [Tue, 11 Dec 2012 09:02:36 +0000 (09:02 +0000)]
[asan] more sceleton code for asan_allocator2
llvm-svn: 169836
Argyrios Kyrtzidis [Tue, 11 Dec 2012 07:48:23 +0000 (07:48 +0000)]
Extend stat query APIs to explicitly specify if the query is for
a file or directory, allowing just a stat call if a file descriptor
is not needed.
Doing just 'stat' is faster than 'open/fstat/close'.
This has the effect of cutting down system time for validating the input files of a PCH.
llvm-svn: 169831
Argyrios Kyrtzidis [Tue, 11 Dec 2012 07:48:14 +0000 (07:48 +0000)]
[PCH] Remove redundant stat call when validating the input files.
This became redundant after we removed the stat cache.
llvm-svn: 169830
Argyrios Kyrtzidis [Tue, 11 Dec 2012 07:48:08 +0000 (07:48 +0000)]
[PCH] When writing out the list of input files, avoid emitting duplicate
entries of the same file.
This can happen because the file was "included" multiple times and is
referenced by multiple SLocEntries.
llvm-svn: 169829
Kostya Serebryany [Tue, 11 Dec 2012 07:27:59 +0000 (07:27 +0000)]
[asan] fix the Windows build
llvm-svn: 169828
NAKAMURA Takumi [Tue, 11 Dec 2012 07:23:52 +0000 (07:23 +0000)]
clang/test/Driver/linker-opts.c: Fix XFAIL that I removed win32.
Sorry for my 3rd commit :(
llvm-svn: 169827
NAKAMURA Takumi [Tue, 11 Dec 2012 07:16:15 +0000 (07:16 +0000)]
clang/test/Driver/debug-options-as.c: Fixup not to use prefix=S. I didn't know "REQUIRES:" would match --check-prefix=S ...
llvm-svn: 169826
NAKAMURA Takumi [Tue, 11 Dec 2012 07:06:09 +0000 (07:06 +0000)]
clang/test/Driver: Introduce the feature "clang-driver", set if gcc driver is not used.
It is not set at targetting cygming. See PR12920.
llvm-svn: 169824
Hao Liu [Tue, 11 Dec 2012 06:25:18 +0000 (06:25 +0000)]
revert the test change
llvm-svn: 169823
Kostya Serebryany [Tue, 11 Dec 2012 06:23:10 +0000 (06:23 +0000)]
[asan] make ThreadNamesTest more flexible
llvm-svn: 169822
Hao Liu [Tue, 11 Dec 2012 06:22:54 +0000 (06:22 +0000)]
A newbie try a test commit
llvm-svn: 169821
NAKAMURA Takumi [Tue, 11 Dec 2012 05:53:54 +0000 (05:53 +0000)]
[CMake] Remove dependencies to intrinsics_gen I introduced in r169724.
llvm-svn: 169819
NAKAMURA Takumi [Tue, 11 Dec 2012 05:53:43 +0000 (05:53 +0000)]
llvm/Target/TargetMachine.h: Remove two dependent headers.
-#include "llvm/Target/TargetTransformImpl.h"
-#include "llvm/TargetTransformInfo.h"
llvm-svn: 169818
NAKAMURA Takumi [Tue, 11 Dec 2012 05:53:37 +0000 (05:53 +0000)]
llvm/tools: Add #include "llvm/TargetTransformInfo.h"
llvm-svn: 169817
NAKAMURA Takumi [Tue, 11 Dec 2012 05:53:24 +0000 (05:53 +0000)]
BackendUtil.cpp: Add #include "llvm/TargetTransformInfo.h"
llvm/Target/TargetMachine.h will not provide "llvm/TargetTransformInfo.h" any more.
llvm-svn: 169816
Rafael Espindola [Tue, 11 Dec 2012 05:52:24 +0000 (05:52 +0000)]
Add tests for a corner case of when to relax all instructions:
We don't want to relax all instructions in
$ clang -c test.s
since most users don't pass -O when using the driver to assemble.
On the other hand, -save-temps should not change the output unnecessary, so in
$ clang -c test.c -save-temps
we should relax all instructions.
llvm-svn: 169815
Jyotsna Verma [Tue, 11 Dec 2012 05:12:25 +0000 (05:12 +0000)]
Use multiclass for new-value store instructions with MEMri operand.
llvm-svn: 169814
Nadav Rotem [Tue, 11 Dec 2012 04:55:10 +0000 (04:55 +0000)]
Fix PR14565. Don't if-convert loops that have switch statements in them.
llvm-svn: 169813
Rafael Espindola [Tue, 11 Dec 2012 03:10:43 +0000 (03:10 +0000)]
Change some functions to take const pointers.
llvm-svn: 169812
Evan Cheng [Tue, 11 Dec 2012 02:31:57 +0000 (02:31 +0000)]
Stylistic tweak.
llvm-svn: 169811
Jim Ingham [Tue, 11 Dec 2012 02:31:48 +0000 (02:31 +0000)]
Broadcast an event when the selected thread is changed.
<rdar://problem/
10976636>
llvm-svn: 169810
Enrico Granata [Tue, 11 Dec 2012 02:17:22 +0000 (02:17 +0000)]
<rdar://problem/
12639506>
Make sure that the user's choice of a format for dumping aggregate types is persisted to child members
llvm-svn: 169809
Jason Molenda [Tue, 11 Dec 2012 01:20:57 +0000 (01:20 +0000)]
Update comments around ABISysV_x86_64::RegisterIsCalleeSaved to
make it clear that this is the list of callee-saved registers,
add reference to the ABI doc that this comes from.
llvm-svn: 169807
Greg Clayton [Tue, 11 Dec 2012 01:20:51 +0000 (01:20 +0000)]
<rdar://problem/
12842032>
Don't load __LINKEDIT segments when dynamically loading kexts.
llvm-svn: 169806
Richard Smith [Tue, 11 Dec 2012 01:14:52 +0000 (01:14 +0000)]
PR14558: Compute triviality of special members (etc) at the end of the class
definition, rather than at the end of the definition of the set of nested
classes. We still defer checking of the user-specified exception specification
to the end of the nesting -- we can't check that until we've parsed the
in-class initializers for non-static data members.
llvm-svn: 169805
Chad Rosier [Tue, 11 Dec 2012 00:51:36 +0000 (00:51 +0000)]
Add a triple to this test.
llvm-svn: 169803
Chandler Carruth [Tue, 11 Dec 2012 00:36:57 +0000 (00:36 +0000)]
Fix a miscompile in the DAG combiner. Previously, we would incorrectly
try to reduce the width of this load, and would end up transforming:
(truncate (lshr (sextload i48 <ptr> as i64), 32) to i32)
to
(truncate (zextload i32 <ptr+4> as i64) to i32)
We lost the sext attached to the load while building the narrower i32
load, and replaced it with a zext because lshr always zext's the
results. Instead, bail out of this combine when there is a conflict
between a sextload and a zext narrowing. The rest of the DAG combiner
still optimize the code down to the proper single instruction:
movswl 6(...),%eax
Which is exactly what we wanted. Previously we read past the end *and*
missed the sign extension:
movl 6(...), %eax
llvm-svn: 169802
Paul Redmond [Tue, 11 Dec 2012 00:36:43 +0000 (00:36 +0000)]
move X86-specific test
This test case uses -mcpu=corei7 so it belongs in CodeGen/X86
Reviewed by: Nadav
llvm-svn: 169801
Han Ming Ong [Tue, 11 Dec 2012 00:24:58 +0000 (00:24 +0000)]
Removed debugging code.
llvm-svn: 169799
Bill Wendling [Tue, 11 Dec 2012 00:23:07 +0000 (00:23 +0000)]
Fix grammar-o.
llvm-svn: 169798
Chad Rosier [Tue, 11 Dec 2012 00:18:02 +0000 (00:18 +0000)]
Fall back to the selection dag isel to select tail calls.
This shouldn't affect codegen for -O0 compiles as tail call markers are not
emitted in unoptimized compiles. Testing with the external/internal nightly
test suite reveals no change in compile time performance. Testing with -O1,
-O2 and -O3 with fast-isel enabled did not cause any compile-time or
execution-time failures. All tests were performed on my x86 machine.
I'll monitor our arm testers to ensure no regressions occur there.
In an upcoming clang patch I will be marking the objc_autoreleaseReturnValue
and objc_retainAutoreleaseReturnValue as tail calls unconditionally. While
it's theoretically true that this is just an optimization, it's an
optimization that we very much want to happen even at -O0, or else ARC
applications become substantially harder to debug.
Part of rdar://
12553082
llvm-svn: 169796
Anna Zaks [Tue, 11 Dec 2012 00:17:53 +0000 (00:17 +0000)]
[analyzer] Don't generate a summary for "freeWhenDone" if method is
inlined.
Fixes a false positive that occurs if a user writes their own
initWithBytesNoCopy:freeWhenDone wrapper.
llvm-svn: 169795
Michael J. Spencer [Mon, 10 Dec 2012 23:52:34 +0000 (23:52 +0000)]
[CMake] Fix external llvm build. It needs to know where llvm-tblgen is.
llvm-svn: 169794
Eric Christopher [Mon, 10 Dec 2012 23:34:43 +0000 (23:34 +0000)]
Refactor out the abbreviation handling into a separate class that
controls each of the abbreviation sets (only a single one at the
moment) and computes offsets separately as well for each set
of DIEs.
No real function change, ordering of abbreviations for the skeleton
CU changed but only because we're computing in a separate order. Fix
the testcase not to care.
llvm-svn: 169793
Enrico Granata [Mon, 10 Dec 2012 23:30:25 +0000 (23:30 +0000)]
<rdar://problem/
12709976>
Adding a summary for NSError
llvm-svn: 169792
Evan Cheng [Mon, 10 Dec 2012 23:21:26 +0000 (23:21 +0000)]
Some enhancements for memcpy / memset inline expansion.
1. Teach it to use overlapping unaligned load / store to copy / set the trailing
bytes. e.g. On 86, use two pairs of movups / movaps for 17 - 31 byte copies.
2. Use f64 for memcpy / memset on targets where i64 is not legal but f64 is. e.g.
x86 and ARM.
3. When memcpy from a constant string, do *not* replace the load with a constant
if it's not possible to materialize an integer immediate with a single
instruction (required a new target hook: TLI.isIntImmLegal()).
4. Use unaligned load / stores more aggressively if target hooks indicates they
are "fast".
5. Update ARM target hooks to use unaligned load / stores. e.g. vld1.8 / vst1.8.
Also increase the threshold to something reasonable (8 for memset, 4 pairs
for memcpy).
This significantly improves Dhrystone, up to 50% on ARM iOS devices.
rdar://
12760078
llvm-svn: 169791
Arnold Schwaighofer [Mon, 10 Dec 2012 23:02:41 +0000 (23:02 +0000)]
Optimistically analyse Phi cycles
Analyse Phis under the starting assumption that they are NoAlias. Recursively
look at their inputs.
If they MayAlias/MustAlias there must be an input that makes them so.
Addresses bug 14351.
llvm-svn: 169788
Han Ming Ong [Mon, 10 Dec 2012 23:02:33 +0000 (23:02 +0000)]
<rdar://problem/
12850287>
When there is XPC connection error, we will report it now.
llvm-svn: 169787
Lang Hames [Mon, 10 Dec 2012 22:49:11 +0000 (22:49 +0000)]
Defer call to InitSections until after MCContext has been initialized. If
InitSections is called before the MCContext is initialized it could cause
duplicate temporary symbols to be emitted later (after context initialization
resets the temporary label counter).
llvm-svn: 169785
Anshuman Dasgupta [Mon, 10 Dec 2012 22:45:57 +0000 (22:45 +0000)]
Fix PR14568: Avoid the DFA packetizer from making an invalid read
beyond array bounds.
No test case since I cannot reproduce an ICE with this bug. According
to Carlos -- the bug reporter -- a segfault occurs only when LLVM is
compiled with a specific version of GCC.
llvm-svn: 169783
Greg Clayton [Mon, 10 Dec 2012 22:26:34 +0000 (22:26 +0000)]
<rdar://problem/
11844604>
When displaying function.name-with-args format will now print "varname=<unavailable>" instead of omitting argument names and values when there is an error reading the value.
llvm-svn: 169781
Eric Christopher [Mon, 10 Dec 2012 22:25:41 +0000 (22:25 +0000)]
Rearrange vars and make comments more obvious.
llvm-svn: 169780
Eric Christopher [Mon, 10 Dec 2012 22:25:38 +0000 (22:25 +0000)]
Remove blank line at top of file.
llvm-svn: 169779
Bill Wendling [Mon, 10 Dec 2012 22:10:53 +0000 (22:10 +0000)]
Remove until I can fix this.
llvm-svn: 169778
Eric Christopher [Mon, 10 Dec 2012 22:00:20 +0000 (22:00 +0000)]
Fix a coding style nit.
llvm-svn: 169776
Bill Wendling [Mon, 10 Dec 2012 21:48:41 +0000 (21:48 +0000)]
Add support to Darwin for the -export_dynamic' flag to the linker.
llvm-svn: 169775
Nadav Rotem [Mon, 10 Dec 2012 21:45:01 +0000 (21:45 +0000)]
Enable the loop vectorizer only on O2 and above. (Still disabled by default)
llvm-svn: 169774
Tom Stellard [Mon, 10 Dec 2012 21:41:58 +0000 (21:41 +0000)]
LegalizeDAG: Allow type promotion of scalar loads
llvm-svn: 169773
Tom Stellard [Mon, 10 Dec 2012 21:41:54 +0000 (21:41 +0000)]
LegalizeDAG: Allow type promotion for scalar stores
llvm-svn: 169772
Nadav Rotem [Mon, 10 Dec 2012 21:39:02 +0000 (21:39 +0000)]
Split the LoopVectorizer into H and CPP.
llvm-svn: 169771
Bill Wendling [Mon, 10 Dec 2012 21:33:45 +0000 (21:33 +0000)]
Revert r169656.
The linker will call `lto_codegen_add_must_preserve_symbol' on all globals that
should be kept around. The linker will pretend that a dylib is being created.
<rdar://problem/
12528059>
llvm-svn: 169770
Bill Wendling [Mon, 10 Dec 2012 21:15:22 +0000 (21:15 +0000)]
Use correct flags for this test.
llvm-svn: 169768