Justin Holewinski [Wed, 20 Mar 2013 14:51:01 +0000 (14:51 +0000)]
Move SDNode order propagation to SDNodeOrdering, which also fixes a missed
case of order propagation during isel.
Thanks Owen for the suggestion!
llvm-svn: 177525
Daniel Jasper [Wed, 20 Mar 2013 14:31:47 +0000 (14:31 +0000)]
Do not consider comments when adjusting to local indent style.
Before (when only reformatting "int b"):
int a; // comment
// comment
int b;
After:
int a; // comment
// comment
int b;
This also fixes llvm.org/PR15433.
llvm-svn: 177524
Dmitry Vyukov [Wed, 20 Mar 2013 14:04:23 +0000 (14:04 +0000)]
tsan: add missing stat descriptions
llvm-svn: 177523
Dmitry Vyukov [Wed, 20 Mar 2013 14:01:10 +0000 (14:01 +0000)]
tsan: call fflush(0) on exit again
llvm-svn: 177522
Daniel Jasper [Wed, 20 Mar 2013 13:53:11 +0000 (13:53 +0000)]
Dmitry Vyukov [Wed, 20 Mar 2013 13:50:47 +0000 (13:50 +0000)]
tsan: reduce size of mutexsets from 64 to 16 mutexes
overflow is handled anyway
saves memory because each thread holds 1024 mutexsets
llvm-svn: 177520
Dmitry Vyukov [Wed, 20 Mar 2013 13:49:45 +0000 (13:49 +0000)]
tsan: fix incorrect test
llvm-svn: 177519
Christian Konig [Wed, 20 Mar 2013 13:49:22 +0000 (13:49 +0000)]
pre-RA-sched: fix TargetOpcode usage
TargetOpcodes need to be treaded as Machine- and not ISD-Opcodes.
Signed-off-by: Christian König <christian.koenig@amd.com>
llvm-svn: 177518
Dmitry Vyukov [Wed, 20 Mar 2013 13:21:50 +0000 (13:21 +0000)]
tsan: special handling of .rodata (don't try to find races, don't keep shadow, dont' put into traces)
llvm-svn: 177517
Tobias Grosser [Wed, 20 Mar 2013 13:03:26 +0000 (13:03 +0000)]
Add option -polly-code-generator=none
This allows us to test Polly and the Polly optimizer without actually doing
code generation at the end. By enabling this option, we can also measure the
compile time overhead due to code generation and the cost of LLVM optimizing the
newly generated code.t
llvm-svn: 177516
Tobias Grosser [Wed, 20 Mar 2013 13:03:25 +0000 (13:03 +0000)]
Silence 'variable unused' warning in release mode
llvm-svn: 177515
Daniel Jasper [Wed, 20 Mar 2013 12:37:50 +0000 (12:37 +0000)]
Add extra indentation for multiline comparisons.
This seems to be generally more desired.
Before:
if (
aaaaaaaa &&
bbbbbbbb >
cccccccc) {}
After:
if (
aaaaaaaa &&
bbbbbbbb >
cccccccc) {}
Also: Some formatting cleanup on clang-format's files.
llvm-svn: 177514
Dmitry Vyukov [Wed, 20 Mar 2013 11:22:03 +0000 (11:22 +0000)]
tsan: fix build
llvm-svn: 177513
Dmitry Vyukov [Wed, 20 Mar 2013 10:31:53 +0000 (10:31 +0000)]
tsan: move trace header into 0x600000000000 range
eliminat thread "dead info" altogether
llvm-svn: 177512
Andy Gibbs [Wed, 20 Mar 2013 10:30:02 +0000 (10:30 +0000)]
Update source/DataFormatters/CMakeLists.txt to reflect actual source files.
llvm-svn: 177511
Dmitry Vyukov [Wed, 20 Mar 2013 10:28:36 +0000 (10:28 +0000)]
tsan: add internal_unlink() function
llvm-svn: 177510
Daniel Jasper [Wed, 20 Mar 2013 10:23:53 +0000 (10:23 +0000)]
Don't remove all indentation when in #defines.
Otherwise, this can become hard to read.
Before: #define A \
case 1:
After: #define A \
case 1:
llvm-svn: 177509
Alexey Samsonov [Wed, 20 Mar 2013 10:11:24 +0000 (10:11 +0000)]
[ASan] Move malloc stats collection away from AsanThreadRegistry class.
llvm-svn: 177508
Daniel Jasper [Wed, 20 Mar 2013 10:02:35 +0000 (10:02 +0000)]
Fix Makefile.
llvm-svn: 177507
Daniel Jasper [Wed, 20 Mar 2013 09:53:23 +0000 (09:53 +0000)]
Add clang-format binary to cfe.
llvm-svn: 177506
Daniel Jasper [Wed, 20 Mar 2013 09:53:18 +0000 (09:53 +0000)]
Improve formatting of function types in template parameters.
Before: A<int * (int)>;
After: A<int *(int)>;
llvm-svn: 177505
Daniel Jasper [Wed, 20 Mar 2013 09:52:50 +0000 (09:52 +0000)]
Remove clang-format from clang-tools-extra.
llvm-svn: 177504
Andy Gibbs [Wed, 20 Mar 2013 09:34:46 +0000 (09:34 +0000)]
Update source/Expression/CMakeLists.txt to reflect actual source files.
llvm-svn: 177503
Dmitry Vyukov [Wed, 20 Mar 2013 09:26:46 +0000 (09:26 +0000)]
asan/tsan: reduce per-thread memory allocator caches
256 blocks (which is 512 in reallity) seems to be too big cache,
reducing the size reduces memory consumption but does not degrade performance.
llvm-svn: 177502
Alexey Samsonov [Wed, 20 Mar 2013 09:23:28 +0000 (09:23 +0000)]
[ASan] Move GetCurrentThread/SetCurrentThread from AsanThreadRegistry class into plain functions: they don't actually use registry
llvm-svn: 177501
Dmitry Vyukov [Wed, 20 Mar 2013 09:22:13 +0000 (09:22 +0000)]
tsan: fix flaky test
llvm-svn: 177500
Dmitry Vyukov [Wed, 20 Mar 2013 09:21:49 +0000 (09:21 +0000)]
tsan: fix incorrect test
llvm-svn: 177499
Chandler Carruth [Wed, 20 Mar 2013 07:40:56 +0000 (07:40 +0000)]
Fix a silly search-and-replace goof with r177495 that only broke
non-release builds.
llvm-svn: 177498
Evgeniy Stepanov [Wed, 20 Mar 2013 07:35:52 +0000 (07:35 +0000)]
[sanitizer] Enable prctl interceptor on Android.
llvm-svn: 177497
Hans Wennborg [Wed, 20 Mar 2013 07:34:27 +0000 (07:34 +0000)]
Fix redundant comparison in gcc::Common::ConstructJob.
We were checking "Arch == llvm::Triple::x86_64 || Arch
== llvm::Triple::x86_64", but the rhs should actually check for
powerpc64.
Found while experimenting with a potential new Clang warning.
llvm-svn: 177496
Chandler Carruth [Wed, 20 Mar 2013 07:30:36 +0000 (07:30 +0000)]
[SROA] Don't preserve the IR names in release builds.
This is espcially important because the new SROA pass goes to great
lengths to provide helpful names for debugging, and as a consequence
they can become very slow to render.
Good for between 5% and 15% of the SROA runtime on some slow test cases
such as the one in PR15412.
llvm-svn: 177495
Chandler Carruth [Wed, 20 Mar 2013 06:47:00 +0000 (06:47 +0000)]
Move the endif to the correct line so we don't have warnings about
unused statistics variables.
llvm-svn: 177494
Chandler Carruth [Wed, 20 Mar 2013 06:30:46 +0000 (06:30 +0000)]
Introduce some new statistics to help track the exact behavior of the
new SROA pass.
llvm-svn: 177493
David Blaikie [Wed, 20 Mar 2013 06:27:06 +0000 (06:27 +0000)]
Reorder the DIFile parameter in DINameSpace
Moving the DIFile parameter to immediately proceed the tag so that it will be a
common prefix with other DIScopes (once the DIFile is replaced with the raw
file/directory pair).
llvm-svn: 177492
Douglas Gregor [Wed, 20 Mar 2013 06:25:14 +0000 (06:25 +0000)]
Work-in-progress documentation on the experimental modules feature.
llvm-svn: 177491
John McCall [Wed, 20 Mar 2013 06:22:14 +0000 (06:22 +0000)]
Remove some dead code.
Patch by Stephen Lin!
llvm-svn: 177490
Hao Liu [Wed, 20 Mar 2013 06:18:06 +0000 (06:18 +0000)]
Add a test case for PR15318 fixed in r177472
llvm-svn: 177489
Nick Lewycky [Wed, 20 Mar 2013 05:59:40 +0000 (05:59 +0000)]
Don't assume the test directory is writable, use %T to find a writable
directory.
llvm-svn: 177488
David Blaikie [Wed, 20 Mar 2013 05:15:37 +0000 (05:15 +0000)]
Test DW_TAG_namespace support in the backend
This is the backend portion of a Clang test case
(clang/test/CodeGenCXX/debug-info-namespace.cpp) that was roughly/coarsely
testing LLVM.
llvm-svn: 177487
David Blaikie [Wed, 20 Mar 2013 05:14:14 +0000 (05:14 +0000)]
Provide more details for DINameSpace debug info in the IR comment annotations
Sorry for the version skew - I should've committed this before the
corresponding Clang test case.
llvm-svn: 177486
David Blaikie [Wed, 20 Mar 2013 05:12:06 +0000 (05:12 +0000)]
Make front-end debug info namespace test frontend-only & more specific without overconstraining it
The backend portion of this test will be committed to LLVM's test suite.
llvm-svn: 177485
Shankar Easwaran [Wed, 20 Mar 2013 05:10:02 +0000 (05:10 +0000)]
[ELF][Hexagon] Add Hexagon dynamic relocations
llvm-svn: 177484
Shankar Easwaran [Wed, 20 Mar 2013 05:06:40 +0000 (05:06 +0000)]
[ELF] Interp section is only needed in dynamic executables
llvm-svn: 177483
Shankar Easwaran [Wed, 20 Mar 2013 04:52:08 +0000 (04:52 +0000)]
[ELF] Order .got.plt properly
llvm-svn: 177482
Douglas Gregor [Wed, 20 Mar 2013 03:59:18 +0000 (03:59 +0000)]
Make sure that Module::ConfigMacrosExhaustive gets initialized and deserialized correctly.
This fixes regressions introduced in r177466 that caused several
module tests to fail sporadically.
llvm-svn: 177481
Richard Smith [Wed, 20 Mar 2013 03:35:02 +0000 (03:35 +0000)]
Teach statement / declaration disambiguation about C++11-style generalized initializers.
llvm-svn: 177480
Rafael Espindola [Wed, 20 Mar 2013 03:05:54 +0000 (03:05 +0000)]
Remove the Tools DensMap from the toolchain.
Each toolchain has a set of tools, but they are all of known types. It can
have a linker, an assembler, a "clang" (compile, analyze, ...) a non-clang
compiler, etc.
Instead of keeping a map, just have member variable for each type of tool.
llvm-svn: 177479
Michael Liao [Wed, 20 Mar 2013 02:33:21 +0000 (02:33 +0000)]
Fix PR15296
- Move SRA/SRL/SHL lowering support from DAG combination to DAG lowering
to support extended 256-bit integer in AVX but not AVX2.
llvm-svn: 177478
Michael Liao [Wed, 20 Mar 2013 02:28:20 +0000 (02:28 +0000)]
Mark all variable shifts needing customizing
- Prepare moving logic from DAG combining into DAG lowering. There's no
functionality change.
llvm-svn: 177477
Michael Liao [Wed, 20 Mar 2013 02:20:36 +0000 (02:20 +0000)]
Move scalar immediate shift lowering into a dedicated func
- no functionality change
llvm-svn: 177476
Nick Lewycky [Wed, 20 Mar 2013 02:14:38 +0000 (02:14 +0000)]
The flag "-coverage-function-names-in-data" is actually backwards -- we do
emit function names in .gcda files by default, and the flag turns that off!
Rename the flag to make it match what it actually does. This keeps the default
format compatible with gcc 4.2.
Also add a test for this flag.
llvm-svn: 177475
David Blaikie [Wed, 20 Mar 2013 01:55:11 +0000 (01:55 +0000)]
Fix test case regression on ARM & PPC introduced r177239
llvm-svn: 177474
John McCall [Wed, 20 Mar 2013 01:53:00 +0000 (01:53 +0000)]
Don't look outside the innermost enclosing namespace when
performing unqualified lookup for a friend class declaration.
rdar://
13393749
llvm-svn: 177473
Hao Liu [Wed, 20 Mar 2013 01:46:36 +0000 (01:46 +0000)]
Fix AsmPrinter crashes with assertion. Bug 15318 in Bugzilla
llvm-svn: 177472
Reid Kleckner [Wed, 20 Mar 2013 01:40:23 +0000 (01:40 +0000)]
[ms-cxxabi] Fix assertion on unhandled function template arg types
This code was changed in r158376 to get template argument source info
for better diagnostics, but the current code asserts for any kind of
unsupported template argument before it can issue a diagnostic. This change
goes back to the Itanium implementation of isTemplate() and puts the argument
index into the diagnostic instead of a source location.
Review URL: http://llvm-reviews.chandlerc.com/D553
llvm-svn: 177471
Nick Lewycky [Wed, 20 Mar 2013 01:38:16 +0000 (01:38 +0000)]
Make clang emit linkage names in debug info for subprograms when coverage info
is enabled. Also add a new -test-coverage cc1 flag which makes testing coverage
possible and add our first clang-side coverage test.
llvm-svn: 177470
Jordan Rose [Wed, 20 Mar 2013 00:35:37 +0000 (00:35 +0000)]
[analyzer] Re-apply "Do part of the work to find shortest bug paths up front".
With the assurance that the trimmed graph does not contain cycles,
this patch is safe (with a few tweaks), and provides the performance
boost it was intended to.
Part of performance work for <rdar://problem/
13433687>.
llvm-svn: 177469
Jordan Rose [Wed, 20 Mar 2013 00:35:31 +0000 (00:35 +0000)]
[analyzer] Break cycles (optionally) when trimming an ExplodedGraph.
Having a trimmed graph with no cycles (a DAG) is much more convenient for
trying to find shortest paths, which is exactly what BugReporter needs to do.
Part of the performance work for <rdar://problem/
13433687>.
llvm-svn: 177468
David Blaikie [Wed, 20 Mar 2013 00:26:26 +0000 (00:26 +0000)]
Refactor the DIFile (2nd) parameter to DITypes to be an MDNode reference to a raw directory/file pair
This makes DIType's first non-tag parameter the same as DIFile's, allowing them
to both share the common implementation of getFilename/getDirectory in DIScope.
llvm-svn: 177467
Douglas Gregor [Wed, 20 Mar 2013 00:22:05 +0000 (00:22 +0000)]
<rdar://problem/
10796651> Introduce configuration macros into module maps.
Configuration macros are macros that are intended to alter how a
module works, such that we need to build different module variants
for different values of these macros. A module can declare its
configuration macros, in which case we will complain if the definition
of a configation macro on the command line (or lack thereof) differs
from the current preprocessor state at the point where the module is
imported. This should eliminate some surprises when enabling modules,
because "#define CONFIG_MACRO ..." followed by "#include
<module/header.h>" would silently ignore the CONFIG_MACRO setting. At
least it will no longer be silent about it.
Configuration macros are eventually intended to help reduce the number
of module variants that need to be built. When the list of
configuration macros for a module is exhaustive, we only need to
consider the settings for those macros when building/finding the
module, which can help isolate modules for various project-specific -D
flags that should never affect how modules are build (but currently do).
llvm-svn: 177466
Justin Holewinski [Wed, 20 Mar 2013 00:10:32 +0000 (00:10 +0000)]
Propagate DAG node ordering during type legalization and instruction selection
A node's ordering is only propagated during legalization if (a) the new node does
not have an ordering (is not a CSE'd node), or (b) the new node has an ordering
that is higher than the node being legalized.
llvm-svn: 177465
Marshall Clow [Wed, 20 Mar 2013 00:01:48 +0000 (00:01 +0000)]
Fix bug in test; found by AddressSanitizer
llvm-svn: 177464
Chad Rosier [Tue, 19 Mar 2013 23:44:03 +0000 (23:44 +0000)]
Fix pr13145 - Naming a function like a register name confuses the asm parser.
Patch by Stepan Dyatkovskiy <stpworld@narod.ru>
rdar://
13457826
llvm-svn: 177463
David Blaikie [Tue, 19 Mar 2013 23:25:22 +0000 (23:25 +0000)]
Move the DIFile operand to DITypes from the 4th operand to the 2nd.
This is another step along the way to making all DIScopes have a common prefix
which can be added to in a general manner to support using directives
(DW_TAG_imported_module).
llvm-svn: 177462
Jakob Stoklund Olesen [Tue, 19 Mar 2013 23:23:31 +0000 (23:23 +0000)]
Annotate various null idioms with SchedRW lists.
llvm-svn: 177461
Jakob Stoklund Olesen [Tue, 19 Mar 2013 23:23:29 +0000 (23:23 +0000)]
Annotate SSE float conversions with SchedRW lists.
llvm-svn: 177460
Jakob Stoklund Olesen [Tue, 19 Mar 2013 23:23:26 +0000 (23:23 +0000)]
Annotate X86InstrCMovSetCC.td with SchedRW lists.
llvm-svn: 177459
Eric Christopher [Tue, 19 Mar 2013 23:10:26 +0000 (23:10 +0000)]
Formatting fixups.
llvm-svn: 177458
David Blaikie [Tue, 19 Mar 2013 23:10:14 +0000 (23:10 +0000)]
Simplify/generalize some debug info test cases
Mostly, try to depend on the annotation comments more so these tests are more
legible, brief, and agnostic to schema changes in the future (sure, they're not
agnostic to changes to the comment annotations but since they're easier to read
they should be easier to update if that happens).
llvm-svn: 177457
Sean Callanan [Tue, 19 Mar 2013 23:03:21 +0000 (23:03 +0000)]
Updated the IRExecutionUnit to keep local copies
of the data it writes down into the process even
if the process doesn't exist. This will allow
the IR interpreter to access static data allocated
on the expression's behalf.
Also cleaned up object ownership in the
IRExecutionUnit so that allocations are created
into the allocations vector. This avoids needless
data copies.
<rdar://problem/
13424594>
llvm-svn: 177456
Jordan Rose [Tue, 19 Mar 2013 23:01:57 +0000 (23:01 +0000)]
[analyzer] Add an integer version of the Circle tests in uninit-vals.m.
A floating-point version is nice for testing unknown values, but it's
good to be able to check all parts of the structure as well.
Test change only, no functionality change.
llvm-svn: 177455
Enrico Granata [Tue, 19 Mar 2013 22:58:48 +0000 (22:58 +0000)]
The formatters for std::shared_ptr, std::weak_ptr, std::list, std::vector and std::map as provided by libc++ are now written in C++ instead of Python
std::deque is still in Python but is much less commonly used
llvm-svn: 177454
Anna Zaks [Tue, 19 Mar 2013 22:38:09 +0000 (22:38 +0000)]
[analyzer] Do not believe lazy binding when symbolic region types do not match
This fixes a crash when analyzing LLVM that was exposed by r177220 (modeling of
trivial copy/move assignment operators).
When we look up a lazy binding for “Builder”, we see the direct binding of Loc at offset 0.
Previously, we believed the binding, which led to a crash. Now, we do not believe it as
the types do not match.
llvm-svn: 177453
Howard Hinnant [Tue, 19 Mar 2013 22:16:57 +0000 (22:16 +0000)]
Marshall Clow found this memory problem in strstream using -fsanitize=address on the test suite.
llvm-svn: 177452
Chad Rosier [Tue, 19 Mar 2013 22:13:05 +0000 (22:13 +0000)]
Dead code.
llvm-svn: 177451
Chad Rosier [Tue, 19 Mar 2013 22:12:47 +0000 (22:12 +0000)]
Dead code.
llvm-svn: 177450
Jordan Rose [Tue, 19 Mar 2013 22:10:44 +0000 (22:10 +0000)]
[analyzer] Add a test case for diagnostic suppression on a graph with cycles.
(see previous commit)
llvm-svn: 177449
Jordan Rose [Tue, 19 Mar 2013 22:10:35 +0000 (22:10 +0000)]
Revert "[analyzer] Do part of the work to find shortest bug paths up front."
The whole reason we were doing a BFS in the first place is because an
ExplodedGraph can have cycles. Unfortunately, my removeErrorNode "update"
doesn't work at all if there are cycles.
I'd still like to be able to avoid doing the BFS every time, but I'll come
back to it later.
This reverts r177353 /
481fa5071c203bc8ba4f88d929780f8d0f8837ba.
llvm-svn: 177448
Matt Kopec [Tue, 19 Mar 2013 22:00:10 +0000 (22:00 +0000)]
Mark TestChangeValueAPI.py as expected fail with gcc as it still reproduces.
llvm-svn: 177447
Chad Rosier [Tue, 19 Mar 2013 21:58:18 +0000 (21:58 +0000)]
[ms-inline asm] Move the immediate asm rewrite into the target specific
logic as a QOI cleanup. No functional change. Tests already in place.
rdar://
13456414
llvm-svn: 177446
Quentin Colombet [Tue, 19 Mar 2013 21:46:49 +0000 (21:46 +0000)]
Update global merge pass according to Duncan's advices:
- Remove useless includes
- Change misleading comments
- Move code into doFinalization
llvm-svn: 177445
Tobias Grosser [Tue, 19 Mar 2013 21:44:07 +0000 (21:44 +0000)]
ScopDetect: Remove some redundant semicolons
llvm-svn: 177444
Howard Hinnant [Tue, 19 Mar 2013 21:34:48 +0000 (21:34 +0000)]
This is an optimization which produces improved launching time. There should be no functionality change. Clients should see no ABI differences.
llvm-svn: 177443
Jakob Stoklund Olesen [Tue, 19 Mar 2013 21:16:56 +0000 (21:16 +0000)]
Annotate X86InstrCompiler.td with SchedRW lists.
Add a new WriteZero SchedWrite type for the common dependency-breaking
instructions that clear a register.
llvm-svn: 177442
Chad Rosier [Tue, 19 Mar 2013 21:12:57 +0000 (21:12 +0000)]
Test case for r177439 and r177440.
llvm-svn: 177441
Chad Rosier [Tue, 19 Mar 2013 21:12:14 +0000 (21:12 +0000)]
[ms-inline asm] Remove the brackets from X86Operand in the IR. These will be
added back in by X86AsmPrinter::printIntelMemReference() during codegen.
Previously, this following example
void t() {
int i;
__asm mov eax, [i]
}
would generate the below assembly
mov eax, dword ptr [[eax]]
which resulted in a fatal error when compiling. Test case coming on the
clang side.
rdar://
13444264
llvm-svn: 177440
Chad Rosier [Tue, 19 Mar 2013 21:11:56 +0000 (21:11 +0000)]
[ms-inline asm] Create a helper function, CreateMemForInlineAsm, that creates
an X86Operand, but also performs a Sema lookup and adds the sizing directive
when appropriate. Use this when parsing a bracketed statement. This is
necessary to get the instruction matching correct as well. Test case coming
on clang side.
rdar://
13455408
llvm-svn: 177439
Tobias Grosser [Tue, 19 Mar 2013 21:11:25 +0000 (21:11 +0000)]
IndependentBlocks: Add a couple of test cases.
llvm-svn: 177438
Bill Wendling [Tue, 19 Mar 2013 21:03:22 +0000 (21:03 +0000)]
Register the GCOV writeout functions so that they're emitted serially.
We don't want to write out >1000 files at the same time. That could make things
prohibitively expensive. Instead, register the "writeout" function so that it's
emitted serially.
<rdar://problem/
12439551>
llvm-svn: 177437
Bill Wendling [Tue, 19 Mar 2013 21:01:19 +0000 (21:01 +0000)]
Add a way to register and execute "writeout" functions.
It may be prohibitively expensive to write out >1000 files at the same time. So
we would rather emit them serially. These functions allow the GCOV
implementation to register the functions that writeout the GCOV information per
compile unit. At exit, they are written.
<rdar://problem/
12439551>
llvm-svn: 177436
Fariborz Jahanian [Tue, 19 Mar 2013 20:27:40 +0000 (20:27 +0000)]
Objective-C [qoi]: Provide improved parse diagnostics when
closing rbrace is missing in an ObjC class declaration.
Can do beter than this, but it involves addition of
overhead which will be present in correct code.
// rdar://6854840
llvm-svn: 177435
Hal Finkel [Tue, 19 Mar 2013 20:22:32 +0000 (20:22 +0000)]
Add a comment to the CodeGen/PowerPC/asym-regclass-copy.ll test
llvm-svn: 177434
Tobias Grosser [Tue, 19 Mar 2013 20:15:19 +0000 (20:15 +0000)]
ScopDetection: Add a couple of test cases
llvm-svn: 177433
Arnaud A. de Grandmaison [Tue, 19 Mar 2013 20:00:22 +0000 (20:00 +0000)]
IndVarSimplify: do not recompute an IV value outside of the loop if :
- it is trivially known to be used inside the loop in a way that can not be optimized away
- there is no use outside of the loop which can take advantage of the computation hoisting
llvm-svn: 177432
Ulrich Weigand [Tue, 19 Mar 2013 19:53:27 +0000 (19:53 +0000)]
Add missing mayLoad flag to LHAUX8 and LWAUX.
All pre-increment load patterns need to set the mayLoad flag (since
they don't provide a DAG pattern).
This was missing for LHAUX8 and LWAUX, which is added by this patch.
llvm-svn: 177431
Ulrich Weigand [Tue, 19 Mar 2013 19:52:30 +0000 (19:52 +0000)]
Rewrite LHAU8 pattern to use standard memory operand.
As opposed to to pre-increment store patterns, the pre-increment
load patterns were already using standard memory operands, with
the sole exception of LHAU8.
As there's no real reason why LHAU8 should be different here,
this patch simply rewrites the pattern to also use a memri
operand, just like all the other patterns.
llvm-svn: 177430
Ulrich Weigand [Tue, 19 Mar 2013 19:52:04 +0000 (19:52 +0000)]
Rewrite pre-increment store patterns to use standard memory operands.
Currently, pre-increment store patterns are written to use two separate
operands to represent address base and displacement:
stwu $rS, $ptroff($ptrreg)
This causes problems when implementing the assembler parser, so this
commit changes the patterns to use standard (complex) memory operands
like in all other memory access instruction patterns:
stwu $rS, $dst
To still match those instructions against the appropriate pre_store
SelectionDAG nodes, the patch uses the new feature that allows a Pat
to match multiple DAG operands against a single (complex) instruction
operand.
Approved by Hal Finkel.
llvm-svn: 177429
Ulrich Weigand [Tue, 19 Mar 2013 19:51:09 +0000 (19:51 +0000)]
Extend TableGen instruction selection matcher to improve handling
of complex instruction operands (e.g. address modes).
Currently, if a Pat pattern creates an instruction that has a complex
operand (i.e. one that consists of multiple sub-operands at the MI
level), this operand must match a ComplexPattern DAG pattern with the
correct number of output operands.
This commit extends TableGen to alternatively allow match a complex
operands against multiple separate operands at the DAG level.
This allows using Pat patterns to match pre-increment nodes like
pre_store (which must have separate operands at the DAG level) onto
an instruction pattern that uses a multi-operand memory operand,
like the following example on PowerPC (will be committed as a
follow-on patch):
def STWU : DForm_1<37, (outs ptr_rc:$ea_res), (ins GPRC:$rS, memri:$dst),
"stwu $rS, $dst", LdStStoreUpd, []>,
RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
def : Pat<(pre_store GPRC:$rS, ptr_rc:$ptrreg, iaddroff:$ptroff),
(STWU GPRC:$rS, iaddroff:$ptroff, ptr_rc:$ptrreg)>;
Here, the pair of "ptroff" and "ptrreg" operands is matched onto the
complex operand "dst" of class "memri" in the "STWU" instruction.
Approved by Jakob Stoklund Olesen.
llvm-svn: 177428
Ulrich Weigand [Tue, 19 Mar 2013 19:50:30 +0000 (19:50 +0000)]
Fix sub-operand size mismatch in tocentry operands.
The tocentry operand class refers to 64-bit values (it is only used in 64-bit,
where iPTR is a 64-bit type), but its sole suboperand is designated as 32-bit
type. This causes a mismatch to be detected at compile-time with the TableGen
patch I'll check in shortly.
To fix this, this commit changes the suboperand to a 64-bit type as well.
llvm-svn: 177427
Ulrich Weigand [Tue, 19 Mar 2013 19:49:52 +0000 (19:49 +0000)]
Remove an invalid and unnecessary Pat pattern from the X86 backend:
def : Pat<(load (i64 (X86Wrapper tglobaltlsaddr :$dst))),
(MOV64rm tglobaltlsaddr :$dst)>;
This pattern is invalid because the MOV64rm instruction expects a
source operand of type "i64mem", which is a subclass of X86MemOperand
and thus actually consists of five MI operands, but the Pat provides
only a single MI operand ("tglobaltlsaddr" matches an SDnode of
type ISD::TargetGlobalTLSAddress and provides a single output).
Thus, if the pattern were ever matched, subsequent uses of the MOV64rm
instruction pattern would access uninitialized memory. In addition,
with the TableGen patch I'm about to check in, this would actually be
reported as a build-time error.
Fortunately, the pattern does in fact never match, for at least two
independent reasons.
First, the code generator actually never generates a pattern of the
form (load (X86Wrapper (tglobaltlsaddr))). For most combinations of
TLS and code models, (tglobaltlsaddr) represents just an offset that
needs to be added to some base register, so it is never directly
dereferenced. The only exception is the initial-exec model, where
(tglobaltlsaddr) refers to the (pc-relative) address of a GOT slot,
which *is* in fact directly dereferenced: but in that case, the
X86WrapperRIP node is used, not X86Wrapper, so the Pat doesn't match.
Second, even if some patterns along those lines *were* ever generated,
we should not need an extra Pat pattern to match it. Instead, the
original MOV64rm instruction pattern ought to match directly, since
it uses an "addr" operand, which is implemented via the SelectAddr
C++ routine; this routine is supposed to accept the full range of
input DAGs that may be implemented by a single mov instruction,
including those cases involving ISD::TargetGlobalTLSAddress (and
actually does so e.g. in the initial-exec case as above).
To avoid build breaks (due to the above-mentioned error) after the
TableGen patch is checked in, I'm removing this Pat here.
llvm-svn: 177426