Chris Lattner [Sun, 26 Jul 2009 00:51:36 +0000 (00:51 +0000)]
simplify DarwinTargetAsmInfo::SelectSectionForGlobal a bit
and make it more aggressive, we now put:
const int G2 __attribute__((weak)) = 42;
into the text (readonly) segment like gcc, previously we put
it into the data (readwrite) segment.
llvm-svn: 77104
Chris Lattner [Sun, 26 Jul 2009 00:50:43 +0000 (00:50 +0000)]
simplify some predicates, add isMergableString()
llvm-svn: 77103
Daniel Dunbar [Sun, 26 Jul 2009 00:42:33 +0000 (00:42 +0000)]
Simplify.
llvm-svn: 77102
Bob Wilson [Sun, 26 Jul 2009 00:39:34 +0000 (00:39 +0000)]
Add support for ARM Neon VREV instructions.
Patch by Anton Korzh, with some modifications from me.
llvm-svn: 77101
Daniel Dunbar [Sun, 26 Jul 2009 00:34:27 +0000 (00:34 +0000)]
Remove Value::setName(const char*, unsigned).
llvm-svn: 77100
Daniel Dunbar [Sun, 26 Jul 2009 00:17:14 +0000 (00:17 +0000)]
Remove Value::setName(const char*).
- Split into a separate patch because there is a slight functionality change,
it is no longer valid to call setName(0), which was equivalent to
setName(""). I'm hoping no one depends on this...
llvm-svn: 77099
Daniel Dunbar [Sat, 25 Jul 2009 23:55:21 +0000 (23:55 +0000)]
Remove Value::{isName, getNameRef}.
Also, change MDString to use a StringRef.
llvm-svn: 77098
Mike Stump [Sat, 25 Jul 2009 23:24:03 +0000 (23:24 +0000)]
Preserve qualifiers.
llvm-svn: 77097
Chris Lattner [Sat, 25 Jul 2009 23:21:55 +0000 (23:21 +0000)]
make SectionKind be a first-class pod struct instead of just
an enum.
llvm-svn: 77096
Ryan Flynn [Sat, 25 Jul 2009 22:29:44 +0000 (22:29 +0000)]
PR3575 - warn on declared variable or function attributes after a definition, which are currently ignored.
llvm-svn: 77095
Argyrios Kyrtzidis [Sat, 25 Jul 2009 22:16:03 +0000 (22:16 +0000)]
Refactor ObjCImplDecl::getInstanceMethod/getClassMethod into one
ObjCImplDecl::getMethod.
llvm-svn: 77094
Argyrios Kyrtzidis [Sat, 25 Jul 2009 22:15:51 +0000 (22:15 +0000)]
Refactor ObjCInterfaceDecl::lookupInstanceMethod/lookupClassMethod into one
ObjCInterfaceDecl::lookupMethod.
llvm-svn: 77093
Argyrios Kyrtzidis [Sat, 25 Jul 2009 22:15:38 +0000 (22:15 +0000)]
Refactor ObjCProtocolDecl::lookupInstanceMethod/lookupClassMethod into one
ObjCProtocolDecl::lookupMethod.
llvm-svn: 77092
Argyrios Kyrtzidis [Sat, 25 Jul 2009 22:15:22 +0000 (22:15 +0000)]
Refactor ObjCContainerDecl::getInstanceMethod/getClassMethod into one
ObjCContainerDecl::getMethod.
Avoids code duplication.
llvm-svn: 77091
Mike Stump [Sat, 25 Jul 2009 22:12:02 +0000 (22:12 +0000)]
Improve comments (Thanks Fariborz).
llvm-svn: 77090
Mike Stump [Sat, 25 Jul 2009 21:26:53 +0000 (21:26 +0000)]
Add noreturn as a type attribute, handle printing for them and handle
calls to noreturn function pointers when CFG building.
llvm-svn: 77089
Reid Kleckner [Sat, 25 Jul 2009 21:26:02 +0000 (21:26 +0000)]
Added a test and fixed a bug in BumpPtrAllocator relating to large alignment
values. Hopefully this fixes PR4622.
llvm-svn: 77088
Fariborz Jahanian [Sat, 25 Jul 2009 21:12:28 +0000 (21:12 +0000)]
ir-gen for non-virtual base class initialization
in constructors.
llvm-svn: 77087
Chris Lattner [Sat, 25 Jul 2009 18:57:34 +0000 (18:57 +0000)]
this is (unfortunately) several changes mixed together:
1. Spell SectionFlags::Writeable as "Writable".
2. Add predicates for deriving SectionFlags from SectionKinds.
3. Sink ELF-specific getSectionPrefixForUniqueGlobal impl into
ELFTargetAsmInfo.
4. Fix SectionFlagsForGlobal to know that BSS/ThreadBSS has the
BSS bit set (the real fix for PR4619).
5. Fix isSuitableForBSS to not put globals with explicit sections
set in BSS (which was the reason #4 wasn't fixed earlier).
6. Remove my previous hack for PR4619.
llvm-svn: 77085
Chris Lattner [Sat, 25 Jul 2009 18:11:58 +0000 (18:11 +0000)]
document some invariants.
llvm-svn: 77084
Chris Lattner [Sat, 25 Jul 2009 17:57:37 +0000 (17:57 +0000)]
add the most expedient hack to fix PR4619, along with a testcase.
Thanks to Rafael for the great example.
llvm-svn: 77083
Dan Gohman [Sat, 25 Jul 2009 16:18:38 +0000 (16:18 +0000)]
Fix a few comments to say "backedge-taken count" instead of
"trip count".
llvm-svn: 77081
Dan Gohman [Sat, 25 Jul 2009 16:18:07 +0000 (16:18 +0000)]
SCEV objects are no longer reference-counted.
llvm-svn: 77080
Dan Gohman [Sat, 25 Jul 2009 16:03:55 +0000 (16:03 +0000)]
Update comments to new-style syntax.
llvm-svn: 77079
Dan Gohman [Sat, 25 Jul 2009 16:03:30 +0000 (16:03 +0000)]
When attempting to sign-extend an addrec by interpreting
the step value as unsigned, the start value and the addrec
itself still need to be treated as signed.
llvm-svn: 77078
Dan Gohman [Sat, 25 Jul 2009 16:00:54 +0000 (16:00 +0000)]
Remove spurious semicolons.
llvm-svn: 77077
Sebastian Redl [Sat, 25 Jul 2009 15:41:38 +0000 (15:41 +0000)]
Implement C++ semantics for C-style and functional-style casts. This regresses Clang extension conversions, like vectors, but allows conversions via constructors and conversion operators.
Add custom conversions to static_cast.
llvm-svn: 77076
Daniel Dunbar [Sat, 25 Jul 2009 15:26:08 +0000 (15:26 +0000)]
MultiTestRunner: Make sure to point at src dir, for out of tree builds.
Factor out routine for executing the script commands.
llvm-svn: 77075
Daniel Dunbar [Sat, 25 Jul 2009 14:46:05 +0000 (14:46 +0000)]
MultiTestRunner: Cleanup test execution & output.
- Stop writing everything to files.
- Make test output more standard.
llvm-svn: 77074
Daniel Dunbar [Sat, 25 Jul 2009 13:19:40 +0000 (13:19 +0000)]
MultiTestRunner: Drop external script and DejaGNU compatibility options for now.
- We don't use the former, and the latter doesn't actually work.
llvm-svn: 77073
Daniel Dunbar [Sat, 25 Jul 2009 13:13:06 +0000 (13:13 +0000)]
MultiTestRunner: Disable valgrind support for now, I don't feel like maintaining
it currently.
llvm-svn: 77072
Daniel Dunbar [Sat, 25 Jul 2009 12:57:15 +0000 (12:57 +0000)]
MultiTestRunner/Win32: Add SYSTEMROOT to environment, replace '&&' in .bat
files.
llvm-svn: 77071
Daniel Dunbar [Sat, 25 Jul 2009 12:47:38 +0000 (12:47 +0000)]
Make having no RUN line a failure.
Doug, please look at decltype-crash and instantiate-function-1.mm, I'm not sure
if they are actually testing the right thing / anything.
llvm-svn: 77070
Daniel Dunbar [Sat, 25 Jul 2009 12:44:58 +0000 (12:44 +0000)]
Remove TESTDIR initialization, MultiTestRunner can find the test dirs itself.
llvm-svn: 77069
Daniel Dunbar [Sat, 25 Jul 2009 12:41:25 +0000 (12:41 +0000)]
Remove empty tests, left over from an svn revert.
llvm-svn: 77068
Daniel Dunbar [Sat, 25 Jul 2009 12:37:21 +0000 (12:37 +0000)]
Remove this Makefile, it is unused.
llvm-svn: 77067
Daniel Dunbar [Sat, 25 Jul 2009 12:23:35 +0000 (12:23 +0000)]
MultiTestRunner: Oops, clang wasn't being substituted properly. This is why the
cxx-using-declaration test case started exhibiting different behavior. It still
needs to be fixed, however...
llvm-svn: 77066
Andreas Bolka [Sat, 25 Jul 2009 12:19:58 +0000 (12:19 +0000)]
Convert DOUT to DEBUG.
llvm-svn: 77065
Daniel Dunbar [Sat, 25 Jul 2009 12:05:55 +0000 (12:05 +0000)]
MultiTestRunner: Always use absolute path names for tests.
Also, fix a function name I forgot to update.
llvm-svn: 77064
Daniel Dunbar [Sat, 25 Jul 2009 11:55:03 +0000 (11:55 +0000)]
Disable this test: cxx-using-declaration.cpp
Something strange is going on here; this test randomly started failing after
unrelated changes to TestRunner.py (even though it runs the same
script).
Argiris or Doug, can one of you investigate?
llvm-svn: 77063
Daniel Dunbar [Sat, 25 Jul 2009 11:27:37 +0000 (11:27 +0000)]
MultiTestRunner: Validate '&&' at the end of RUN lines.
- This is just to normalize, these will go away soon hopefully.
Added all the missing '&&'s that have crept in. :)
llvm-svn: 77062
Daniel Dunbar [Sat, 25 Jul 2009 10:14:19 +0000 (10:14 +0000)]
MultiTestRunner: Explicitly define child environment, the only variable we need
to import is the PATH.
llvm-svn: 77061
Daniel Dunbar [Sat, 25 Jul 2009 10:09:50 +0000 (10:09 +0000)]
Simplify JIT target selection.
- Instead of requiring targets to define a JIT quality match function, we just
have them specify if they support a JIT.
- Target selection for the JIT just gets the host triple and looks for the best
target which matches the triple and has a JIT.
llvm-svn: 77060
Daniel Dunbar [Sat, 25 Jul 2009 09:53:43 +0000 (09:53 +0000)]
MultiTestRunner.py improvements.
- Tweak output directories for temp files, derive the temporary base from the
test's parent directory name, and the test name (instead of the whole path).
llvm-svn: 77059
Daniel Dunbar [Sat, 25 Jul 2009 09:42:24 +0000 (09:42 +0000)]
MultiTestRunner.py improvements.
- Not improved: the horribly lousy name. :)
- Suppress stderr when capturing output.
- Rewrite which() to do the right PATH search instead of being lazy and
shelling out to 'which'.
- On Windows, run scripts as batch files (via 'cmd /c ...').
llvm-svn: 77058
Daniel Dunbar [Sat, 25 Jul 2009 09:01:12 +0000 (09:01 +0000)]
Change these tests to not depend as much on the name of the input.
llvm-svn: 77057
Sanjiv Gupta [Sat, 25 Jul 2009 07:48:53 +0000 (07:48 +0000)]
Mark attributes of return insn correctly. It was being assumed safe to delete in isSafeToDelete (a thing checked-in 76281).
llvm-svn: 77056
Chris Lattner [Sat, 25 Jul 2009 07:31:51 +0000 (07:31 +0000)]
remove this test. It is currently failing because we now emit the string
on darwin with ".cstring" instead of ".section __TEXT,__cstring". They
are the same and the former is better. Remove this because this is no longer
magic pixie dust in the frontend.
llvm-svn: 77055
Chris Lattner [Sat, 25 Jul 2009 07:22:20 +0000 (07:22 +0000)]
eventually we should describe string options in the data structures section
llvm-svn: 77054
Chris Lattner [Sat, 25 Jul 2009 07:16:59 +0000 (07:16 +0000)]
minor tweaks.
llvm-svn: 77053
Daniel Dunbar [Sat, 25 Jul 2009 06:49:55 +0000 (06:49 +0000)]
Add new helpers for registering targets.
- Less boilerplate == good.
llvm-svn: 77052
Daniel Dunbar [Sat, 25 Jul 2009 06:02:13 +0000 (06:02 +0000)]
Finish migrating VMCore to StringRef/Twine based APIs.
llvm-svn: 77051
Daniel Dunbar [Sat, 25 Jul 2009 05:26:53 +0000 (05:26 +0000)]
Ok, "most clients should be unaffected" was a lie. Add notes on upgrading.
llvm-svn: 77050
Daniel Dunbar [Sat, 25 Jul 2009 04:41:11 +0000 (04:41 +0000)]
Initial update to VMCore to use Twines for string arguments.
- The only meat here is in Value.{h,cpp} the rest is essential 'const
std::string &' -> 'const Twine &'.
llvm-svn: 77048
John McCall [Sat, 25 Jul 2009 04:36:53 +0000 (04:36 +0000)]
Semantic checking for main().
Fix some invalid main() methods in the test suite that were nicely
exposed by the new checks.
llvm-svn: 77047
John McCall [Sat, 25 Jul 2009 04:35:50 +0000 (04:35 +0000)]
Flesh out the QualifierSet API.
llvm-svn: 77046
Eric Christopher [Sat, 25 Jul 2009 02:45:27 +0000 (02:45 +0000)]
Fix 80-col violations.
llvm-svn: 77045
Eric Christopher [Sat, 25 Jul 2009 02:28:41 +0000 (02:28 +0000)]
Move ExtractElementInst to ::Create instead of new. Update all uses.
llvm-svn: 77044
Dan Gohman [Sat, 25 Jul 2009 02:23:48 +0000 (02:23 +0000)]
Update to new syntax.
llvm-svn: 77043
Daniel Dunbar [Sat, 25 Jul 2009 01:55:32 +0000 (01:55 +0000)]
Rewrite examples to use DEBUG instead of DOUT.
llvm-svn: 77042
Evan Cheng [Sat, 25 Jul 2009 01:55:25 +0000 (01:55 +0000)]
80 col violation.
llvm-svn: 77041
Daniel Dunbar [Sat, 25 Jul 2009 01:43:39 +0000 (01:43 +0000)]
Tweak svn:ignore
llvm-svn: 77040
Dan Gohman [Sat, 25 Jul 2009 01:43:01 +0000 (01:43 +0000)]
Convert a few more things to use raw_ostream.
llvm-svn: 77039
Anders Carlsson [Sat, 25 Jul 2009 01:40:29 +0000 (01:40 +0000)]
More work on the constant struct builder. We can now convert the struct to a packed struct when necessary.
llvm-svn: 77038
Evan Cheng [Sat, 25 Jul 2009 01:25:08 +0000 (01:25 +0000)]
Get rid of a couple of unnecessary getOpcode calls.
llvm-svn: 77035
Dan Gohman [Sat, 25 Jul 2009 01:22:26 +0000 (01:22 +0000)]
Teach ScalarEvolution to make use of no-overflow flags when
analyzing add recurrences.
llvm-svn: 77034
Dan Gohman [Sat, 25 Jul 2009 01:13:51 +0000 (01:13 +0000)]
Convert a few more uses of llvm/Support/Streams.h to raw_ostream.
llvm-svn: 77033
Dan Gohman [Sat, 25 Jul 2009 01:13:03 +0000 (01:13 +0000)]
Instead of eagerly creating new SCEVs to replace all SCEVs that are
affected after a PHI node has been analyzed, just remove affected
SCEVs from the Scalars map, so that they'll be (lazily) recreated as
needed. This avoids creating SCEV objects that aren't actually needed.
Also, rewrite the associated def-use walking code to be non-recursive
and to continue traversing past Instructions that don't have an
entry in the Scalars map.
llvm-svn: 77032
Evan Cheng [Sat, 25 Jul 2009 01:11:46 +0000 (01:11 +0000)]
I've lost my mind. PR4572 has not been fixed.
llvm-svn: 77031
Fariborz Jahanian [Sat, 25 Jul 2009 01:08:28 +0000 (01:08 +0000)]
Some code refactoring per Daniel's feedback.
llvm-svn: 77030
Dan Gohman [Sat, 25 Jul 2009 00:48:42 +0000 (00:48 +0000)]
Make AliasAnalysis and related classes use
getAnalysisIfAvailable<TargetData>().
llvm-svn: 77028
Daniel Dunbar [Sat, 25 Jul 2009 00:43:31 +0000 (00:43 +0000)]
One more getName -> getNameStr
llvm-svn: 77027
Evan Cheng [Sat, 25 Jul 2009 00:39:37 +0000 (00:39 +0000)]
Another TODO.
llvm-svn: 77026
Jeffrey Yasskin [Sat, 25 Jul 2009 00:33:57 +0000 (00:33 +0000)]
Add a missing ilist_node.h #include to SparseBitVector, and add a very short
test for it. The test is by no means complete, but it tests the problem I was
fixing.
llvm-svn: 77025
Evan Cheng [Sat, 25 Jul 2009 00:33:29 +0000 (00:33 +0000)]
Change Thumb2 jumptable codegen to one that uses two level jumps:
Before:
adr r12, #LJTI3_0_0
ldr pc, [r12, +r0, lsl #2]
LJTI3_0_0:
.long LBB3_24
.long LBB3_30
.long LBB3_31
.long LBB3_32
After:
adr r12, #LJTI3_0_0
add pc, r12, +r0, lsl #2
LJTI3_0_0:
b.w LBB3_24
b.w LBB3_30
b.w LBB3_31
b.w LBB3_32
This has several advantages.
1. This will make it easier to optimize this to a TBB / TBH instruction +
(smaller) table.
2. This eliminate the need for ugly asm printer hack to force the address
into thumb addresses (bit 0 is one).
3. Same codegen for pic and non-pic.
4. This eliminate the need to align the table so constantpool island pass
won't have to over-estimate the size.
Based on my calculation, the later is probably slightly faster as well since
ldr pc with shifter address is very slow. That is, it should be a win as long
as the HW implementation can do a reasonable job of branch predict the second
branch.
llvm-svn: 77024
Sean Callanan [Sat, 25 Jul 2009 00:30:51 +0000 (00:30 +0000)]
MemoryObject - Abstract base class for contiguous addressable memory.
Necessary for cases in which the memory is in another process, in a
file, or on a remote machine.
The primary use for this is the llvm-mc disassemblers, so that they
can be targeted at arbitrary objects, not just in-process memory.
llvm-svn: 77023
Anders Carlsson [Sat, 25 Jul 2009 00:26:00 +0000 (00:26 +0000)]
Whoops, did not mean to check this in.
llvm-svn: 77022
Anders Carlsson [Sat, 25 Jul 2009 00:24:56 +0000 (00:24 +0000)]
More work on the constant struct builder. We now try to layout all constant structs but throw away the result.
llvm-svn: 77021
Evan Cheng [Sat, 25 Jul 2009 00:24:40 +0000 (00:24 +0000)]
Remove a duplicated test.
llvm-svn: 77020
Daniel Dunbar [Sat, 25 Jul 2009 00:23:56 +0000 (00:23 +0000)]
More migration to raw_ostream, the water has dried up around the iostream hole.
- Some clients which used DOUT have moved to DEBUG. We are deprecating the
"magic" DOUT behavior which avoided calling printing functions when the
statement was disabled. In addition to being unnecessary magic, it had the
downside of leaving code in -Asserts builds, and of hiding potentially
unnecessary computations.
llvm-svn: 77019
Evan Cheng [Sat, 25 Jul 2009 00:13:11 +0000 (00:13 +0000)]
ARM code emitter can't handle Thumb2 instructions yet. So don't even try.
llvm-svn: 77018
Daniel Dunbar [Fri, 24 Jul 2009 23:54:34 +0000 (23:54 +0000)]
Tweak, raw_ostream is a ostream, not iostream replacement
llvm-svn: 77017
Daniel Dunbar [Fri, 24 Jul 2009 23:42:33 +0000 (23:42 +0000)]
Fix build for GCC 4.0?
llvm-svn: 77016
Daniel Dunbar [Fri, 24 Jul 2009 23:23:46 +0000 (23:23 +0000)]
Fix compile with 4.4 (I hope?); PR4617.
llvm-svn: 77015
Andreas Bolka [Fri, 24 Jul 2009 23:19:28 +0000 (23:19 +0000)]
Forward-declare raw_ostream.
llvm-svn: 77014
Owen Anderson [Fri, 24 Jul 2009 23:12:58 +0000 (23:12 +0000)]
Update for LLVM API change.
llvm-svn: 77012
Owen Anderson [Fri, 24 Jul 2009 23:12:02 +0000 (23:12 +0000)]
Revert the ConstantInt constructors back to their 2.5 forms where possible, thanks to contexts-on-types. More to come.
llvm-svn: 77011
Daniel Dunbar [Fri, 24 Jul 2009 23:04:51 +0000 (23:04 +0000)]
CodingStandards: Emphasize use of raw_ostream more.
- Chris, please approve.
llvm-svn: 77010
Dan Gohman [Fri, 24 Jul 2009 23:01:30 +0000 (23:01 +0000)]
AliasAnalysis wants sizes in address-units, not bits.
llvm-svn: 77009
Evan Cheng [Fri, 24 Jul 2009 22:42:45 +0000 (22:42 +0000)]
Forgot this test earlier.
llvm-svn: 77007
Evan Cheng [Fri, 24 Jul 2009 22:42:22 +0000 (22:42 +0000)]
Fix these tests.
llvm-svn: 77006
Eli Friedman [Fri, 24 Jul 2009 21:56:17 +0000 (21:56 +0000)]
Fix assert assembling zero-argument constant GEP.
There's still a strict-aliasing violation here, but I don't feel like
dealing with that right now...
llvm-svn: 77005
Evan Cheng [Fri, 24 Jul 2009 20:47:38 +0000 (20:47 +0000)]
Uh. It would be useful to actually print the operand.
llvm-svn: 77004
Douglas Gregor [Fri, 24 Jul 2009 20:34:43 +0000 (20:34 +0000)]
Template instantiation for static data members that are defined out-of-line.
Note that this also fixes a bug that affects non-template code, where we
were not treating out-of-line static data members are "file-scope" variables,
and therefore not checking their initializers.
llvm-svn: 77002
Fariborz Jahanian [Fri, 24 Jul 2009 20:33:35 +0000 (20:33 +0000)]
Fixes broken test by accident.
llvm-svn: 77001
Fariborz Jahanian [Fri, 24 Jul 2009 20:28:49 +0000 (20:28 +0000)]
More Sema check for constructor's member initializer along
with type conversion to fix ir-gen crash.
llvm-svn: 77000
Chris Lattner [Fri, 24 Jul 2009 20:27:11 +0000 (20:27 +0000)]
fix some predicates
llvm-svn: 76999
Ted Kremenek [Fri, 24 Jul 2009 20:26:31 +0000 (20:26 +0000)]
In the "use of floating point variable as loop counter" check, check
if the DeclRefExpr is a float, not just either argument.
llvm-svn: 76998
Chris Lattner [Fri, 24 Jul 2009 20:14:10 +0000 (20:14 +0000)]
change SectionKindForGlobal from being a public (and
previously virtual) function to being a static function
in the .cpp file.
llvm-svn: 76997
Ted Kremenek [Fri, 24 Jul 2009 20:03:11 +0000 (20:03 +0000)]
Have 'clang --analyze' run syntactic static analysis security checks.
llvm-svn: 76996