Benjamin Kramer [Tue, 18 Dec 2012 18:40:20 +0000 (18:40 +0000)]
LoopVectorize: Emit reductions as log2(vectorsize) shuffles + vector ops instead of scalar operations.
For example on x86 with SSE4.2 a <8 x i8> add reduction becomes
movdqa %xmm0, %xmm1
movhlps %xmm1, %xmm1 ## xmm1 = xmm1[1,1]
paddw %xmm0, %xmm1
pshufd $1, %xmm1, %xmm0 ## xmm0 = xmm1[1,0,0,0]
paddw %xmm1, %xmm0
phaddw %xmm0, %xmm0
pextrb $0, %xmm0, %edx
instead of
pextrb $2, %xmm0, %esi
pextrb $0, %xmm0, %edx
addb %sil, %dl
pextrb $4, %xmm0, %esi
addb %dl, %sil
pextrb $6, %xmm0, %edx
addb %sil, %dl
pextrb $8, %xmm0, %esi
addb %dl, %sil
pextrb $10, %xmm0, %edi
pextrb $14, %xmm0, %edx
addb %sil, %dil
pextrb $12, %xmm0, %esi
addb %dil, %sil
addb %sil, %dl
llvm-svn: 170439
Eli Bendersky [Tue, 18 Dec 2012 18:21:29 +0000 (18:21 +0000)]
Get rid of the pesky -Woverloaded-virtual warning. No change in functionality.
llvm-svn: 170438
Jakob Stoklund Olesen [Tue, 18 Dec 2012 17:54:53 +0000 (17:54 +0000)]
Tighten the insert() API for bundled instructions.
The normal insert() function takes an MBB::iterator position, and
inserts a stand-alone MachineInstr as before.
The insert() function that takes an MBB::instr_iterator position can
insert instructions inside a bundle, and will now update the bundle
flags correctly when that happens.
When the insert position is between two bundles, it is unclear whether
the instruction should be appended to the previous bundle, prepended to
the next bundle, or stand on its own. The MBB::insert() function doesn't
bundle the instruction in that case, use the MIBundleBuilder class for
that.
llvm-svn: 170437
Hal Finkel [Tue, 18 Dec 2012 17:50:58 +0000 (17:50 +0000)]
Check multiple register classes for inline asm tied registers
A register can be associated with several distinct register classes.
For example, on PPC, the floating point registers are each associated with
both F4RC (which holds f32) and F8RC (which holds f64). As a result, this code
would fail when provided with a floating point register and an f64 operand
because it would happen to find the register in the F4RC class first and
return that. From the F4RC class, SDAG would extract f32 as the register
type and then assert because of the invalid implied conversion between
the f64 value and the f32 register.
Instead, search all register classes. If a register class containing the
the requested register has the requested type, then return that register
class. Otherwise, as before, return the first register class found that
contains the requested register.
llvm-svn: 170436
Marshall Clow [Tue, 18 Dec 2012 16:46:30 +0000 (16:46 +0000)]
Added static_assert to std::get<N>(std::array) calls to catch "out of bounds" calls
llvm-svn: 170435
Kostya Serebryany [Tue, 18 Dec 2012 14:56:38 +0000 (14:56 +0000)]
[sanitizer] fix LargeMmapAllocator::GetBlockBegin
llvm-svn: 170434
Dmitry Vyukov [Tue, 18 Dec 2012 14:44:44 +0000 (14:44 +0000)]
tsan: intercept fork() to prevent false race reports on fd's
llvm-svn: 170433
Guy Benyei [Tue, 18 Dec 2012 14:38:23 +0000 (14:38 +0000)]
Re-commit r170428 changes with Linux style file endings.
Add OpenCL images as clang builtin types.
llvm-svn: 170432
Guy Benyei [Tue, 18 Dec 2012 14:30:41 +0000 (14:30 +0000)]
Revert changes from r170428, as I accidentally changed the line endings of these files to Windows style.
llvm-svn: 170431
Dmitry Vyukov [Tue, 18 Dec 2012 14:19:24 +0000 (14:19 +0000)]
tsan: warn about reports from signal handlers
llvm-svn: 170430
Dmitry Vyukov [Tue, 18 Dec 2012 12:35:31 +0000 (12:35 +0000)]
tsan: add signalfd() and inotify_init() interceptors
llvm-svn: 170429
Guy Benyei [Tue, 18 Dec 2012 12:30:03 +0000 (12:30 +0000)]
Add OpenCL images as clang builtin types.
llvm-svn: 170428
Dmitry Vyukov [Tue, 18 Dec 2012 12:20:55 +0000 (12:20 +0000)]
tsan: add io_sync flag that controls amount of IO synchronization
llvm-svn: 170427
Dmitry Vyukov [Tue, 18 Dec 2012 12:19:50 +0000 (12:19 +0000)]
tsan: remove TSAN_OPTIONS from the script
The runtime skips atexit sleep if there are no threads now,
so it must be fast w/o it.
Allows to specify own TSAN_OPTIONS for the tests.
llvm-svn: 170426
David Chisnall [Tue, 18 Dec 2012 10:24:07 +0000 (10:24 +0000)]
Remove obsolete FIXMEs.
llvm-svn: 170425
Alexey Samsonov [Tue, 18 Dec 2012 09:57:34 +0000 (09:57 +0000)]
[Sanitizer] Expose StackTrace::GetPreviousInstructionPc() to get PC of call instruction from return address
llvm-svn: 170424
Richard Smith [Tue, 18 Dec 2012 09:30:21 +0000 (09:30 +0000)]
ubsan: Demangle class names, and be more informative when a reinterpret_cast
has got us to the wrong offset within an object.
llvm-svn: 170423
Sebastian Pop [Tue, 18 Dec 2012 08:56:51 +0000 (08:56 +0000)]
return -1 when polly::getNumberOfIterations returns -1
llvm-svn: 170422
Sebastian Pop [Tue, 18 Dec 2012 07:46:13 +0000 (07:46 +0000)]
isl: vector code generation based on ISL ast
Original patch by Tobias Grosser, slightly modified by Sebastian Pop.
llvm-svn: 170420
Sebastian Pop [Tue, 18 Dec 2012 07:46:06 +0000 (07:46 +0000)]
change interface for isStride
isStride now takes a partial schedule as input.
Patch from Tobias Grosser <tobias@grosser.es>.
llvm-svn: 170419
Kostya Serebryany [Tue, 18 Dec 2012 07:32:16 +0000 (07:32 +0000)]
[asan] add some colors to asan output if printing to tty (following ubsan)
llvm-svn: 170418
Dmitry Vyukov [Tue, 18 Dec 2012 06:57:34 +0000 (06:57 +0000)]
tsan: describe "file descriptor" location
llvm-svn: 170417
Nadav Rotem [Tue, 18 Dec 2012 06:37:12 +0000 (06:37 +0000)]
Enable the loop vectorizer.
llvm-svn: 170416
Richard Smith [Tue, 18 Dec 2012 06:30:32 +0000 (06:30 +0000)]
ubsan: When diagnosing something wrong somewhere in memory, emit a note
pointing at the bad location and a snippet of nearby memory values. This is
strictly best-effort; reading these bytes to display the note could lead to a
seg fault, and that's explicitly OK.
llvm-svn: 170415
Nadav Rotem [Tue, 18 Dec 2012 05:50:54 +0000 (05:50 +0000)]
Rename the test so that we can add additional vectors-of-pointers tests
into the same file in the future.
llvm-svn: 170414
Richard Smith [Tue, 18 Dec 2012 05:49:49 +0000 (05:49 +0000)]
Don't escape %s in printf usage message; it's not written by printf.
llvm-svn: 170413
Nadav Rotem [Tue, 18 Dec 2012 05:23:31 +0000 (05:23 +0000)]
SROA: Replace calls to getScalarSizeInBits to DataLayout's API because
getScalarSizeInBits could not handle vectors of pointers.
llvm-svn: 170412
NAKAMURA Takumi [Tue, 18 Dec 2012 05:14:30 +0000 (05:14 +0000)]
test/Misc/dev-fd-fs.c: Mark this as XFAIL:freebsd (since r169831).
I don't want to suppress this to be excluded with the feature "dev-fd-fs". This should work also on freebsd.
llvm-svn: 170411
NAKAMURA Takumi [Tue, 18 Dec 2012 05:08:12 +0000 (05:08 +0000)]
llvm/test/MC/ELF/comp-dir.s: Appease MSYS Bash.
llvm-svn: 170410
Jason Molenda [Tue, 18 Dec 2012 04:39:43 +0000 (04:39 +0000)]
Add a new qProcessInfo packet to debugserver.
This can be used by lldb to ask for information
about the process debugserver is attached to/launched.
Particularly useful on a 64-bit x86 Mac system which
can run 32-bit or 64-bit user-land processes.
llvm-svn: 170409
Rafael Espindola [Tue, 18 Dec 2012 04:29:34 +0000 (04:29 +0000)]
Inline hasAtomicCopyHelperAPI.
llvm-svn: 170408
Richard Smith [Tue, 18 Dec 2012 04:23:18 +0000 (04:23 +0000)]
ubsan: if the frontend didn't provide us with a source location, try to work
one out from the return address. Currently, we can only resolve this address to
a file and line number if we have an external symbolizer.
llvm-svn: 170407
Rafael Espindola [Tue, 18 Dec 2012 04:18:55 +0000 (04:18 +0000)]
Merge storage classes even when contexts don't match.
This fixes the storage class of extern decls that are merged with file level
statics. The patch also fixes the linkage computation so that they are
considered internal.
llvm-svn: 170406
Rafael Espindola [Tue, 18 Dec 2012 03:54:29 +0000 (03:54 +0000)]
Now that the uninitialized variable has been fixed in llvm, add this test back,
but remove unneeded options and fix the comment.
llvm-svn: 170405
Rafael Espindola [Tue, 18 Dec 2012 03:35:05 +0000 (03:35 +0000)]
Initialize NoRedZone and remove unused default values.
llvm-svn: 170404
Richard Smith [Tue, 18 Dec 2012 03:04:38 +0000 (03:04 +0000)]
Fix typo (thanks to Jordan for spotting it!).
llvm-svn: 170403
Eli Friedman [Tue, 18 Dec 2012 02:37:32 +0000 (02:37 +0000)]
The underlying type for an enum should be an integer type, not another enum.
(This change only affects ObjC.)
<rdar://problem/
12857117>.
llvm-svn: 170402
Sean Silva [Tue, 18 Dec 2012 02:15:05 +0000 (02:15 +0000)]
docs: Nuke AnalyzerRegions.rst.
As per Ted's advice. It can be brought back from version control if
needed.
This also fixes a Sphinx warning.
llvm-svn: 170401
Jim Ingham [Tue, 18 Dec 2012 02:03:49 +0000 (02:03 +0000)]
Adding events when watchpoints are set or changed.
<rdar://problem/
11597849>
llvm-svn: 170400
Anders Carlsson [Tue, 18 Dec 2012 01:29:20 +0000 (01:29 +0000)]
When warning about a missing prototype because a function declaration is missing 'void', insert a fixit to add the void.
llvm-svn: 170399
Eli Bendersky [Tue, 18 Dec 2012 00:53:36 +0000 (00:53 +0000)]
Cleanup comment and formatting
llvm-svn: 170398
Eli Friedman [Tue, 18 Dec 2012 00:52:36 +0000 (00:52 +0000)]
More conservative fix for <rdar://problem/
12847524> (a crash printing diagnostic ranges).
I'm not really happy with this fix, but I'm confident it's correct.
llvm-svn: 170397
Sean Callanan [Tue, 18 Dec 2012 00:50:17 +0000 (00:50 +0000)]
Removed a possible use of unchecked NULL from
the AST result synthesizer's logging.
<rdar://problem/
12889053>
llvm-svn: 170396
Jakob Stoklund Olesen [Tue, 18 Dec 2012 00:46:39 +0000 (00:46 +0000)]
Repair bundles that were broken by removing and reinserting the first
instruction.
This isn't strictly necessary at the moment because Thumb2SizeReduction
also copies all MI flags from the old instruction to the new. However, a
future patch will make that kind of direct flag tampering illegal.
llvm-svn: 170395
Eric Christopher [Tue, 18 Dec 2012 00:42:26 +0000 (00:42 +0000)]
Formatting.
llvm-svn: 170394
Bill Wendling [Tue, 18 Dec 2012 00:37:10 +0000 (00:37 +0000)]
Use the correct return type (size_t) for these builtins.
<rdar://problem/
12646344>
llvm-svn: 170393
Bill Wendling [Tue, 18 Dec 2012 00:34:49 +0000 (00:34 +0000)]
Add an expected parameter for the size of the destination.
<rdar://problem/
12622659>
llvm-svn: 170392
Eric Christopher [Tue, 18 Dec 2012 00:31:10 +0000 (00:31 +0000)]
Add support for passing the main file name down to the assembler
for location information.
Part of PR14624
llvm-svn: 170391
Eric Christopher [Tue, 18 Dec 2012 00:31:01 +0000 (00:31 +0000)]
Add support for passing -main-file-name all the way through to
the assembler.
Part of PR14624
llvm-svn: 170390
Eric Christopher [Tue, 18 Dec 2012 00:30:54 +0000 (00:30 +0000)]
Cleanup formatting and whitespace.
llvm-svn: 170389
Richard Smith [Tue, 18 Dec 2012 00:23:12 +0000 (00:23 +0000)]
Rein ubsan's vptr sanitizer back a bit. Per core issue 453, binding a reference
to an object outside its lifetime does not have undefined behavior.
llvm-svn: 170388
Richard Smith [Tue, 18 Dec 2012 00:22:45 +0000 (00:22 +0000)]
Rein ubsan's vptr sanitizer back a bit. Per core issue 453, binding a reference
to an object outside its lifetime does not have undefined behavior.
llvm-svn: 170387
Jakob Stoklund Olesen [Tue, 18 Dec 2012 00:13:11 +0000 (00:13 +0000)]
Extract a method, no functional change intended.
Sadly, this costs us a perfectly good opportunity to use 'goto'.
llvm-svn: 170385
Jakob Stoklund Olesen [Mon, 17 Dec 2012 23:55:38 +0000 (23:55 +0000)]
Tighten up the erase/remove API for bundled instructions.
Most code is oblivious to bundles and uses the MBB::iterator which only
visits whole bundles. MBB::erase() operates on whole bundles at a time
as before.
MBB::remove() now refuses to remove bundled instructions. It is not safe
to remove all instructions in a bundle without deleting them since there
is no way of returning pointers to all the removed instructions.
MBB::remove_instr() and MBB::erase_instr() will now update bundle flags
correctly, lifting individual instructions out of bundles while leaving
the remaining bundle intact.
The MachineInstr convenience functions are updated so
eraseFromParent() erases a whole bundle as before
eraseFromBundle() erases a single instruction, leaving the rest of its bundle.
removeFromParent() refuses to operate on bundled instructions, and
removeFromBundle() lifts a single instruction out of its bundle.
These functions will no longer accidentally split or coalesce bundles -
bundle flags are updated to preserve the existing bundling, and explicit
bundleWith* / unbundleFrom* functions should be used to change the
instruction bundling.
This API update is still a work in progress. I am going to update APIs
first so they maintain bundle flags automatically when possible. Then
I'll add stricter verification of the bundle flags.
llvm-svn: 170384
Reed Kotler [Mon, 17 Dec 2012 23:41:45 +0000 (23:41 +0000)]
EmitDebugLabel should by default be the same as EmitLabel everywhere.
It must be explicity set in MCPureStreamer because otherwise it will
inherit incorrectly from the parent.
llvm-svn: 170383
Eli Bendersky [Mon, 17 Dec 2012 22:50:56 +0000 (22:50 +0000)]
fix indentation
llvm-svn: 170381
Chad Rosier [Mon, 17 Dec 2012 22:35:29 +0000 (22:35 +0000)]
[arm fast-isel] Minor cleanup. No functional change intended.
llvm-svn: 170379
Rafael Espindola [Mon, 17 Dec 2012 22:23:47 +0000 (22:23 +0000)]
Fix isThisDeclarationADefinition for extern following tentative.
An extern declaration following a tentative definition should not itself be
considered a tentative definition.
Fixes pr14614.
llvm-svn: 170377
Nick Kledzik [Mon, 17 Dec 2012 22:11:17 +0000 (22:11 +0000)]
Fix some integer constant warnings by using a suffix
llvm-svn: 170376
Chandler Carruth [Mon, 17 Dec 2012 21:44:50 +0000 (21:44 +0000)]
Add a triple to this test -- it has to be an ELF platform...
llvm-svn: 170374
Chandler Carruth [Mon, 17 Dec 2012 21:40:04 +0000 (21:40 +0000)]
Fix PR14625 by teaching the driver to detect PWD for assembly files.
This also requires adding support to -cc1as for passing the detecting
PWD down through LLVM's debug info (which in turn required the LLVM
change in r170371).
The test case is weak (we only test the driver behavior) because there
is currently to infrastructure for running cc1as in the test suite. So
those four lines are untested (much like all other lines in that file),
but we have a test for the same pattern using llvm-mc in the LLVM
repository.
llvm-svn: 170373
Daniel Malea [Mon, 17 Dec 2012 21:32:49 +0000 (21:32 +0000)]
Mention how to do a (Linux) release build on the build instructions page.
Also, a few minor fixes:
- paragraph formatting
- capitalize "Python" where appropriate
llvm-svn: 170372
Chandler Carruth [Mon, 17 Dec 2012 21:32:42 +0000 (21:32 +0000)]
Prepare LLVM to fix PR14625, exposing a hook in MCContext to manage the
compilation directory.
This defaults to the current working directory, just as it always has,
but now an assembler can choose to override it with a custom directory.
I've taught llvm-mc about this option and added a test case.
llvm-svn: 170371
Han Ming Ong [Mon, 17 Dec 2012 20:53:19 +0000 (20:53 +0000)]
<rdar://problem/
12890948>
Send thread name using hex encoding.
llvm-svn: 170370
Nick Kledzik [Mon, 17 Dec 2012 20:43:53 +0000 (20:43 +0000)]
re-enable test cases now that traits work with g++. Fix some g++ warnings
llvm-svn: 170369
Michael Ilseman [Mon, 17 Dec 2012 20:40:14 +0000 (20:40 +0000)]
Remove trailing whitespace
llvm-svn: 170368
Michael Ilseman [Mon, 17 Dec 2012 20:37:55 +0000 (20:37 +0000)]
Removed trailing whitespace
llvm-svn: 170367
Argyrios Kyrtzidis [Mon, 17 Dec 2012 20:19:56 +0000 (20:19 +0000)]
[c-index-test] Fix warnings about unused results from chdir().
Patch from Edwin Vane!
llvm-svn: 170366
Argyrios Kyrtzidis [Mon, 17 Dec 2012 20:10:50 +0000 (20:10 +0000)]
Just use begin()/end() instead of "&Vector[0], &Vector[0]+Vector.size()".
No functionality change.
llvm-svn: 170365
Argyrios Kyrtzidis [Mon, 17 Dec 2012 20:10:49 +0000 (20:10 +0000)]
Don't include the header outside the include guards, it defeats the purpose of the include guards.
llvm-svn: 170364
Argyrios Kyrtzidis [Mon, 17 Dec 2012 20:10:43 +0000 (20:10 +0000)]
[parser] Push a semi token for recovery only when it is actually missing.
llvm-svn: 170363
Anna Zaks [Mon, 17 Dec 2012 20:08:54 +0000 (20:08 +0000)]
[analyzer] Tweak the NumFunctionsAnalyzed stat so that it's more useful.
llvm-svn: 170362
Anna Zaks [Mon, 17 Dec 2012 20:08:51 +0000 (20:08 +0000)]
[analyzer] Implement "do not inline large functions many times"
performance heuristic
After inlining a function with more than 13 basic blocks 32 times, we
are not going to inline it anymore. The idea is that inlining large
functions leads to drastic performance implications. Since the function
has already been inlined, we know that we've analyzed it in many
contexts.
The following metrics are used:
- Large function is a function with more than 13 basic blocks (we
should switch to another metric, like cyclomatic complexity)
- We consider that we've inlined a function many times if it's been
inlined 32 times. This number is configurable with -analyzer-config
max-times-inline-large=xx
This heuristic addresses a performance regression introduced with
inlining on one benchmark. The analyzer on this benchmark became 60
times slower with inlining turned on. The heuristic allows us to analyze
it in 24% of the time. The performance improvements on the other
benchmarks I've tested with are much lower - under 10%, which is
expected.
llvm-svn: 170361
Chad Rosier [Mon, 17 Dec 2012 19:59:43 +0000 (19:59 +0000)]
[arm fast-isel] Fast-isel only handles simple VTs, so make sure the necessary
checks are in place. Some minor cleanup as well.
llvm-svn: 170360
Nick Kledzik [Mon, 17 Dec 2012 19:02:05 +0000 (19:02 +0000)]
Use different trait techniques to be compatible with g++
llvm-svn: 170355
David Chisnall [Mon, 17 Dec 2012 18:54:24 +0000 (18:54 +0000)]
Reapply r170344, this time without forgetting to commit the header changes.
llvm-svn: 170354
Chandler Carruth [Mon, 17 Dec 2012 18:48:07 +0000 (18:48 +0000)]
Fix another SROA crasher, PR14601.
This was a silly oversight, we weren't pruning allocas which were used
by variable-length memory intrinsics from the set that could be widened
and promoted as integers. Fix that.
llvm-svn: 170353
Benjamin Kramer [Mon, 17 Dec 2012 18:02:43 +0000 (18:02 +0000)]
Revert "Added support for new property helpers (GNUstep runtime)."
This reverts commit r170344. Doesn't even compile.
llvm-svn: 170351
Tim Northover [Mon, 17 Dec 2012 17:59:35 +0000 (17:59 +0000)]
Query section for whether it should be executable.
llvm-svn: 170350
Tim Northover [Mon, 17 Dec 2012 17:59:32 +0000 (17:59 +0000)]
Teach MachO which sections contain code
llvm-svn: 170349
Daniel Malea [Mon, 17 Dec 2012 17:40:07 +0000 (17:40 +0000)]
Initialize m_done to false in Driver constructor to avoid premature exit bug (spotted on Linux)
Patch by Chia-Hung Duan!
llvm-svn: 170348
Evgeniy Stepanov [Mon, 17 Dec 2012 16:30:05 +0000 (16:30 +0000)]
[msan] Fix lint warning.
llvm-svn: 170347
Dmitry Vyukov [Mon, 17 Dec 2012 16:28:15 +0000 (16:28 +0000)]
tsan: say what thread had created a thread in reports
llvm-svn: 170346
Richard Osborne [Mon, 17 Dec 2012 16:28:02 +0000 (16:28 +0000)]
Add instruction encodings / disassembly support for l2r instructions.
llvm-svn: 170345
David Chisnall [Mon, 17 Dec 2012 15:59:29 +0000 (15:59 +0000)]
Added support for new property helpers (GNUstep runtime).
llvm-svn: 170344
Tom Stellard [Mon, 17 Dec 2012 15:14:56 +0000 (15:14 +0000)]
R600: enable S_*N2_* instructions
They seem to work fine.
Patch by: Christian König
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
llvm-svn: 170343
Tom Stellard [Mon, 17 Dec 2012 15:14:54 +0000 (15:14 +0000)]
R600: BB operand support for SI
Patch by: Christian König
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
llvm-svn: 170342
Tom Stellard [Mon, 17 Dec 2012 15:14:53 +0000 (15:14 +0000)]
R600: remove nonsense setPrefLoopAlignment
The Align parameter is a power of two, so 16 results in 64K
alignment. Additional to that even 16 byte alignment doesn't
make any sense, so just remove it.
Patch by: Christian König
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
llvm-svn: 170341
Kostya Serebryany [Mon, 17 Dec 2012 14:57:25 +0000 (14:57 +0000)]
[asan] asan_allocator2: implement malloc_usable_size and fix 32-bit
llvm-svn: 170339
Chandler Carruth [Mon, 17 Dec 2012 14:51:24 +0000 (14:51 +0000)]
Teach the rewriting of memcpy calls to support subvector copies.
This also cleans up a bit of the memcpy call rewriting by sinking some
irrelevant code further down and making the call-emitting code a bit
more concrete.
Previously, memcpy of a subvector would actually miscompile (!!!) the
copy into a single vector element copy. I have no idea how this ever
worked. =/ This is the memcpy half of PR14478 which we probably weren't
noticing previously because it didn't actually assert.
The rewrite relies on the newly refactored insert- and extractVector
functions to do the heavy lifting, and those are the same as used for
loads and stores which makes the test coverage a bit more meaningful
here.
llvm-svn: 170338
Daniel Jasper [Mon, 17 Dec 2012 14:34:14 +0000 (14:34 +0000)]
Add basic support for splitting before function calls if it can't be
avoided.
This required a minor modification of the memoization as now the
"CurrentPenalty" depends on whether or not we break before the current
token. Therefore, the CurrentPenalty should not be memoized but added
after retrieving a value from memory. This should not affect the runtime
behavior.
llvm-svn: 170337
Patrik Hagglund [Mon, 17 Dec 2012 14:30:06 +0000 (14:30 +0000)]
Revert/correct some FastISel changes in r170104 (EVT->MVT for
TargetLowering::getRegClassFor).
Some isSimple() guards were missing, or getSimpleVT() were hoisted too
far, resulting in asserts on valid LLVM assembly input.
llvm-svn: 170336
Evgeniy Stepanov [Mon, 17 Dec 2012 14:28:00 +0000 (14:28 +0000)]
Optimize tree walking in markAliveBlocks.
Check whether a BB is known as reachable before adding it to the worklist.
This way BB's with multiple predecessors are added to the list no more than
once.
llvm-svn: 170335
Richard Osborne [Mon, 17 Dec 2012 14:23:54 +0000 (14:23 +0000)]
Add instruction encodings for PEEK and ENDIN.
Previously these were marked with the wrong format.
llvm-svn: 170334
Chandler Carruth [Mon, 17 Dec 2012 14:03:01 +0000 (14:03 +0000)]
Fix a secondary bug I introduced while fixing the first part of PR14478.
The first half of fixing this bug was actually in r170328, but was
entirely coincidental. It did however get me to realize the nature of
the bug, and adapt the test case to test more interesting behavior. In
turn, that uncovered the rest of the bug which I've fixed here.
This should fix two new asserts that showed up in the vectorize nightly
tester.
llvm-svn: 170333
Richard Osborne [Mon, 17 Dec 2012 13:55:49 +0000 (13:55 +0000)]
Fix parameter name in prototypes in XCoreDisassembler.
llvm-svn: 170332
Chandler Carruth [Mon, 17 Dec 2012 13:51:03 +0000 (13:51 +0000)]
Hoist a convertValue call to the two paths where it is needed.
I noticed this while looking at r170328. We only ever do a vector
rewrite when the alloca *is* the vector type, so it's good to not paper
over bugs here by doing a convertValue that isn't needed.
llvm-svn: 170331
Richard Osborne [Mon, 17 Dec 2012 13:50:04 +0000 (13:50 +0000)]
Add instruction encodings / disassembly support for rus instructions.
llvm-svn: 170330
Kostya Serebryany [Mon, 17 Dec 2012 13:43:47 +0000 (13:43 +0000)]
[asan] asan_allocator2: don't use TLS and fix calloc
llvm-svn: 170329
Chandler Carruth [Mon, 17 Dec 2012 13:41:21 +0000 (13:41 +0000)]
Hoist the insertVector helper to be a static helper.
This will allow its use inside of memcpy rewriting as well. This routine
is more complex than extractVector, and some of its uses are not 100%
where I want them to be so there is still some work to do here.
While this can technically change the output in some cases, it shouldn't
be a change that matters -- IE, it can leave some dead code lying around
that prior versions did not, etc.
Yet another step in the refactorings leading up to the solution to the
last component of PR14478.
llvm-svn: 170328