Daniel Dunbar [Wed, 17 Oct 2012 16:30:54 +0000 (16:30 +0000)]
Support: Don't remove special files on signals.
- Similar to Path::eraseFromDisk(), we don't want LLVM to remove things like
/dev/null, even if it has the permission.
llvm-svn: 166105
Alexey Samsonov [Wed, 17 Oct 2012 14:04:57 +0000 (14:04 +0000)]
[ASan] unit tests: Move main() to a separate file. Fix lint
llvm-svn: 166104
Kostya Serebryany [Wed, 17 Oct 2012 13:42:04 +0000 (13:42 +0000)]
[asan] added a test for a bug in asan at -O0 introduced by r165936 (making asan a FunctionPass). The test is not actually enabled for -O0 yet (since it fails)
llvm-svn: 166103
Kostya Serebryany [Wed, 17 Oct 2012 13:40:06 +0000 (13:40 +0000)]
[asan] better debug diagnostics in asan compiler module
llvm-svn: 166102
Alexey Samsonov [Wed, 17 Oct 2012 13:39:09 +0000 (13:39 +0000)]
[ASan] Fix strchr/index tests for users who have 'char* strchr(char*,int)' instead of 'char* strchr(const char*, int)'
llvm-svn: 166101
Daniel Jasper [Wed, 17 Oct 2012 13:35:36 +0000 (13:35 +0000)]
Fix tests, which accidentally matched implicit code on specific
platforms to make buildbots happy.
llvm-svn: 166100
Alexey Samsonov [Wed, 17 Oct 2012 13:19:59 +0000 (13:19 +0000)]
[ASan] don't use attribute no_address_safety_analysis if it's unknown to compiler
llvm-svn: 166099
Alexey Samsonov [Wed, 17 Oct 2012 13:12:23 +0000 (13:12 +0000)]
[Sanitizer] symbolizer: increase the maximal number of shared libraries to 16K
llvm-svn: 166098
Chandler Carruth [Wed, 17 Oct 2012 09:40:16 +0000 (09:40 +0000)]
Remove an unneeded flag from the bitrig driver tests.
Based loosely on a patch from David Hill.
llvm-svn: 166096
Chandler Carruth [Wed, 17 Oct 2012 09:23:48 +0000 (09:23 +0000)]
This just in, it is a *bad idea* to use 'udiv' on an offset of
a pointer. A very bad idea. Let's not do that. Fixes PR14105.
Note that this wasn't *that* glaring of an oversight. Originally, these
routines were only called on offsets within an alloca, which are
intrinsically positive. But over the evolution of the pass, they ended
up being called for arbitrary offsets, and things went downhill...
llvm-svn: 166095
Daniel Jasper [Wed, 17 Oct 2012 08:52:59 +0000 (08:52 +0000)]
First version of matchers for Types and TypeLocs.
Review: http://llvm-reviews.chandlerc.com/D47
llvm-svn: 166094
Kostya Serebryany [Wed, 17 Oct 2012 08:50:53 +0000 (08:50 +0000)]
[asan] fix gcc warnings while building asan-rt
llvm-svn: 166093
Bill Wendling [Wed, 17 Oct 2012 08:08:06 +0000 (08:08 +0000)]
Marked this variable as 'used' so that LTO doesn't get rid of it.
llvm-svn: 166092
Chandler Carruth [Wed, 17 Oct 2012 07:22:16 +0000 (07:22 +0000)]
Fix a really annoying "bug" introduced in r165941. The change from that
revision makes no sense. We cannot use the address space of the *post
indexed* type to conclude anything about a *pre indexed* pointer type's
size. More importantly, this index can never be over a pointer. We are
indexing over arrays and vectors here.
Of course, I have no test case here. Neither did the original patch. =/
llvm-svn: 166091
Richard Smith [Wed, 17 Oct 2012 06:45:09 +0000 (06:45 +0000)]
Update test FIXME: The '[]' in 'delete []' is never part of a lambda.
llvm-svn: 166090
Craig Topper [Wed, 17 Oct 2012 05:15:58 +0000 (05:15 +0000)]
Remove LLVM_DELETED_FUNCTION from destructors that override non-deleted base class destructors. This isn't legal by the C++11 standard and clang now checks for it. Curiously gcc didn't catch this, possibly because of the template usage.
llvm-svn: 166089
John McCall [Wed, 17 Oct 2012 04:53:31 +0000 (04:53 +0000)]
Set a special flag in class metadata when an Objective-C class
has ivars that require destruction, but none that require anything
except zero-initialization. This is common in ARC and (when true
throughout a class hierarchy) permits the elimination of an
unnecessary message-send during allocation.
llvm-svn: 166088
John McCall [Wed, 17 Oct 2012 04:53:23 +0000 (04:53 +0000)]
Organize and rename the magic constants for class flags.
No functionality change.
llvm-svn: 166087
Michael Liao [Wed, 17 Oct 2012 03:59:18 +0000 (03:59 +0000)]
Check SSSE3 instead of SSE4.1
- All shuffle insns required, especially PSHUB, are added in SSSE3.
llvm-svn: 166086
John McCall [Wed, 17 Oct 2012 02:28:37 +0000 (02:28 +0000)]
At -O0, prefer objc_storeStrong with a null new value to the
combination of a load+objc_release; this is generally better
for tools that try to track why values are retained and
released. Also use objc_storeStrong when copying a block
(again, only at -O0), which requires us to do a preliminary
store of null in order to compensate for objc_storeStrong's
assign semantics.
llvm-svn: 166085
Michael Liao [Wed, 17 Oct 2012 02:22:27 +0000 (02:22 +0000)]
Fix setjmp on models with non-Small code model nor non-Static relocation model
- MBB address is only valid as an immediate value in Small & Static
code/relocation models. On other models, LEA is needed to load IP address of
the restore MBB.
- A minor fix of MBB in MC lowering is added as well to enable target
relocation flag being propagated into MC.
llvm-svn: 166084
Jakob Stoklund Olesen [Wed, 17 Oct 2012 01:37:59 +0000 (01:37 +0000)]
Use a SparseSet instead of a BitVector for UsedInInstr in RAFast.
This is just as fast, and it makes it possible to avoid leaking the
UsedPhysRegs BitVector implementation through
MachineRegisterInfo::addPhysRegsUsed().
llvm-svn: 166083
David Blaikie [Wed, 17 Oct 2012 00:47:58 +0000 (00:47 +0000)]
Implement C++ 10.3p16 - overrides involving deleted functions must match.
Only deleted functions may override deleted functions and non-deleted functions
may only override non-deleted functions.
llvm-svn: 166082
Jason Molenda [Wed, 17 Oct 2012 00:41:14 +0000 (00:41 +0000)]
Add code to UnwindAssemblyInstEmulation::GetNonCallSiteUnwindPlanFromAssembly
to handle an addition class of early-return instructions we find in arm code:
tail-call optimziation returns where we restore the register state from the
function entry and jump directly (not branch & link) to another function --
when that other function returns, it will return to our caller.
Previously this mid-function epilogue sequence was not being correctly detected.
We would not re-instate the prologue setup instructions for the rest of the function
so unwinds would break from that point until the end of the function.
<rdar://problem/
12502597>
llvm-svn: 166081
Nico Weber [Wed, 17 Oct 2012 00:34:34 +0000 (00:34 +0000)]
"'Might as well make it static const.' -- John McCall" -- Michael Scott
llvm-svn: 166080
Douglas Gregor [Wed, 17 Oct 2012 00:11:35 +0000 (00:11 +0000)]
Fix the handling of target options in our unit tests.
llvm-svn: 166079
Eli Friedman [Tue, 16 Oct 2012 23:55:06 +0000 (23:55 +0000)]
Move test to a more appropriate place.
llvm-svn: 166078
Eric Christopher [Tue, 16 Oct 2012 23:46:25 +0000 (23:46 +0000)]
Use a typedef to reduce some typing and reformat code accordingly.
llvm-svn: 166077
Eric Christopher [Tue, 16 Oct 2012 23:46:23 +0000 (23:46 +0000)]
Variable name cleanup.
llvm-svn: 166076
Eric Christopher [Tue, 16 Oct 2012 23:46:21 +0000 (23:46 +0000)]
Formatting and 80-col.
llvm-svn: 166075
Eric Christopher [Tue, 16 Oct 2012 23:46:19 +0000 (23:46 +0000)]
Spacing.
llvm-svn: 166074
Eli Friedman [Tue, 16 Oct 2012 23:45:15 +0000 (23:45 +0000)]
Fix pretty-printing for variables declared in a condition. Patch by Grzegorz Jablonski.
llvm-svn: 166073
Douglas Gregor [Tue, 16 Oct 2012 23:40:58 +0000 (23:40 +0000)]
Serialize TargetOptions into an AST file, and make sure that we keep
target options around so they can be accessed at any point (rather
than keeping them transient).
llvm-svn: 166072
Richard Smith [Tue, 16 Oct 2012 23:30:16 +0000 (23:30 +0000)]
DR1492: In a definition of a destructor, the exception specification must be
explicitly specified iff it was specified in the declaration.
llvm-svn: 166071
Greg Clayton [Tue, 16 Oct 2012 22:58:25 +0000 (22:58 +0000)]
API cleanup.
llvm-svn: 166070
David Blaikie [Tue, 16 Oct 2012 22:56:05 +0000 (22:56 +0000)]
PR13684: Emit vtable entries for deleted functions as __cxa_deleted_function.
This is consistent/interoperable with GCC 4.7 (& __cxa_deleted_function isn't
present in 4.4 - not sure when it got added, but you'll need something with
that function available for this to work).
llvm-svn: 166069
Jakob Stoklund Olesen [Tue, 16 Oct 2012 22:51:58 +0000 (22:51 +0000)]
Avoid rematerializing a redef immediately after the old def.
PR14098 contains an example where we would rematerialize a MOV8ri
immediately after the original instruction:
%vreg7:sub_8bit<def> = MOV8ri 9; GR32_ABCD:%vreg7
%vreg22:sub_8bit<def> = MOV8ri 9; GR32_ABCD:%vreg7
Besides being pointless, it is also wrong since the original instruction
only redefines part of the register, and the value read by the new
instruction is wrong.
The problem was the LiveRangeEdit::allUsesAvailableAt() didn't
special-case OrigIdx == UseIdx and found the wrong SSA value.
llvm-svn: 166068
Jakob Stoklund Olesen [Tue, 16 Oct 2012 22:51:55 +0000 (22:51 +0000)]
Revert r166046 "Switch back to the old coalescer for now to fix the 32 bit bit"
A fix for PR14098, including the test case is in the next commit.
llvm-svn: 166067
Manman Ren [Tue, 16 Oct 2012 22:40:48 +0000 (22:40 +0000)]
ARM ABI: fix testing case arm-abi-vector.c by using [[VAR]]
rdar://
12439123
llvm-svn: 166066
Derek Schuff [Tue, 16 Oct 2012 22:30:41 +0000 (22:30 +0000)]
Add pnaclcall convention to Native Client targets.
Because PNaCl bitcode must be target-independent, it uses some
different bitcode representations from other targets (e.g. byval and
sret for structures). This means that without additional type
information, it cannot meet some native ABI requirements for some
targets (e.g. passing structures containing unions by value on
x86-64). To allow generation of code which uses the correct native
ABIs, we also support triples such as x86_64-nacl, which uses
target-dependent IR (as opposed to le32-nacl, which uses byval and
sret).
To allow interoperation between the two types of code, this patch adds
a calling convention attribute to be used in code compiled with the
target-dependent triple, which will generate code using the le32-style
bitcode. This calling convention does not need to be explicitly
supported in the backend because it determines bitcode representation
rather than native conventions (the backend just needs to undersand
how to handle byval and sret for the Native Client OS).
This patch implements __attribute__((pnaclcall)) to generate calls in
bitcode according to the le32 bitcode conventions, an attribute which
is accepted by any Native Client target, but issues a warning
otherwise.
llvm-svn: 166065
Manman Ren [Tue, 16 Oct 2012 22:02:26 +0000 (22:02 +0000)]
XFAIL this testing case to recover buildbot
llvm-svn: 166064
Chad Rosier [Tue, 16 Oct 2012 21:55:39 +0000 (21:55 +0000)]
[ms-inline asm] Move some logic around to simplify the interface between the
front-end and the AsmParser. No functional change intended.
llvm-svn: 166063
Jim Ingham [Tue, 16 Oct 2012 21:41:58 +0000 (21:41 +0000)]
Add the ability to set timeout & "run all threads" options both from the "expr" command and from
the SB API's that evaluate expressions.
<rdar://problem/
12457211>
llvm-svn: 166062
Michael Gottesman [Tue, 16 Oct 2012 21:29:38 +0000 (21:29 +0000)]
[InstCombine] Teach InstCombine how to handle an obfuscated splat.
An obfuscated splat is where the frontend poorly generates code for a splat
using several different shuffles to create the splat, i.e.,
%A = load <4 x float>* %in_ptr, align 16
%B = shufflevector <4 x float> %A, <4 x float> undef, <4 x i32> <i32 0, i32 0, i32 undef, i32 undef>
%C = shufflevector <4 x float> %B, <4 x float> %A, <4 x i32> <i32 0, i32 1, i32 4, i32 undef>
%D = shufflevector <4 x float> %C, <4 x float> %A, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
llvm-svn: 166061
Enrico Granata [Tue, 16 Oct 2012 21:11:14 +0000 (21:11 +0000)]
Removing the two extra GetXSize(bool) calls since we do not desire to support them long-term
llvm-svn: 166060
David Blaikie [Tue, 16 Oct 2012 21:03:19 +0000 (21:03 +0000)]
Test pure virtual vtable entries in the Itanium ABI.
(original functionality committed in r99807, refactored in r160373)
llvm-svn: 166059
Enrico Granata [Tue, 16 Oct 2012 20:57:12 +0000 (20:57 +0000)]
<rdar://problem/
12446320> Fixing an issue with our Driver where setting an immediate output would not cause suppression of the final printout. This allows effective output redirection for Python commands
llvm-svn: 166058
Jason Molenda [Tue, 16 Oct 2012 20:45:49 +0000 (20:45 +0000)]
llvm needs the OS to be set to either iOS or Mac OS X
to work properly; when doing bare-boards rom debugging
force the OS to be one of those when initializing llvm.
<rdar://problem/
12504138>
llvm-svn: 166057
David Blaikie [Tue, 16 Oct 2012 20:25:33 +0000 (20:25 +0000)]
Note deleted functions when dumping vtables.
llvm-svn: 166056
Greg Clayton [Tue, 16 Oct 2012 20:20:18 +0000 (20:20 +0000)]
Patch from Matt Kopec:
This patch fixes an issue where if lldb fails to attach to a process (ie. invalid pid) on Linux, the process monitor thread gets stuck waiting for a signal from the attach thread, which never comes due to not being signaled. It also implements StopOpThread which is used for both attach/launch cases as I'm not aware of any special handling needed for the attach case. Also, propagate 'Error' from the Detach function instead of using a bool.
llvm-svn: 166055
Chad Rosier [Tue, 16 Oct 2012 20:16:20 +0000 (20:16 +0000)]
[ms-inline asm] Add the helper function, isParseringInlineAsm(). To be used in a future commit.
llvm-svn: 166054
Jakub Staszak [Tue, 16 Oct 2012 19:52:32 +0000 (19:52 +0000)]
Simplify code. No functionality change.
llvm-svn: 166053
Manman Ren [Tue, 16 Oct 2012 19:51:48 +0000 (19:51 +0000)]
ARM ABI: fix ABI alignment issues in varargs.
We generalize r166040 to handle ABI alignment issues for all types.
rdar://
12439123
llvm-svn: 166052
Michael Liao [Tue, 16 Oct 2012 19:49:51 +0000 (19:49 +0000)]
Check .rela instead of ELF64 for the compensation vaue resetting
llvm-svn: 166051
Jakub Staszak [Tue, 16 Oct 2012 19:39:40 +0000 (19:39 +0000)]
80-col fixup.
llvm-svn: 166050
Michael Liao [Tue, 16 Oct 2012 19:38:35 +0000 (19:38 +0000)]
Teach DAG combine to fold (trunc (fptoXi x)) to (fptoXi x)
llvm-svn: 166049
Anna Zaks [Tue, 16 Oct 2012 19:36:39 +0000 (19:36 +0000)]
[analyzer] Expose clang_version in CmpRuns
llvm-svn: 166048
Anna Zaks [Tue, 16 Oct 2012 19:36:37 +0000 (19:36 +0000)]
[analyzer] Ivar Invalidation: track ivars in continuations and
@implementation.
llvm-svn: 166047
Rafael Espindola [Tue, 16 Oct 2012 19:34:06 +0000 (19:34 +0000)]
Switch back to the old coalescer for now to fix the 32 bit bit
llvm+clang+compiler-rt bootstrap.
llvm-svn: 166046
Jakub Staszak [Tue, 16 Oct 2012 19:32:31 +0000 (19:32 +0000)]
Simplify potentially quadratic behavior while erasing elements from std::vector.
llvm-svn: 166045
Manman Ren [Tue, 16 Oct 2012 19:18:39 +0000 (19:18 +0000)]
ARM ABI: passing illegal vector types as varargs.
We expand varargs in clang and the call site is handled in the back end, it is
hard to match exactly how illegal vectors are handled in the backend. Therefore,
we legalize the illegal vector types in clang:
if (Size <= 32), legalize to i32.
if (Size == 64), legalize to v2i32.
if (Size == 128), legalize to v4i32.
if (Size > 128), use indirect.
rdar://
12439123
llvm-svn: 166043
Manman Ren [Tue, 16 Oct 2012 19:01:37 +0000 (19:01 +0000)]
ARM ABI: fix ABI alignment issues when passing legal vector types as varargs.
We create an aligned temporary space and copy the content over from ap.cur to
the temporary space. This is necessary if the natural alignment of the type is
greater than the ABI alignment.
rdar://
12439123
llvm-svn: 166040
David Blaikie [Tue, 16 Oct 2012 18:53:14 +0000 (18:53 +0000)]
Implement GCC's -Wint-to-pointer-cast.
This implementation doesn't warn on anything that GCC doesn't warn on with the
exception of templates specializations (GCC doesn't warn, Clang does). The
specific skipped cases (boolean, constant expressions, enums) are open for
debate/adjustment if anyone wants to demonstrate that GCC is being overly
conservative here. The only really obvious false positive I found was in the
Clang regression suite's MPI test - apparently MPI uses specific flag values in
pointer constants. (eg: #define FOO (void*)~0)
llvm-svn: 166039
Sean Callanan [Tue, 16 Oct 2012 18:22:04 +0000 (18:22 +0000)]
If a full path to a compiler is provided on the
command line to dotest.py, replace / with _ in
the logfile names that mention that compiler so
that we don't try to put log files in weird
places.
llvm-svn: 166038
Bill Wendling [Tue, 16 Oct 2012 18:20:09 +0000 (18:20 +0000)]
And now we can call the other 'get' method from this one and not duplicate the code.
llvm-svn: 166037
Michael Liao [Tue, 16 Oct 2012 18:14:11 +0000 (18:14 +0000)]
Support v8f32 to v8i8/vi816 conversion through custom lowering
- Add custom FP_TO_SINT on v8i16 (and v8i8 which is legalized as v8i16 due to
vector element-wise widening) to reduce DAG combiner and its overhead added
in X86 backend.
llvm-svn: 166036
Bill Wendling [Tue, 16 Oct 2012 18:06:06 +0000 (18:06 +0000)]
Use the appropriate Attributes::get method to create an Attributes object.
llvm-svn: 166035
Joerg Sonnenberger [Tue, 16 Oct 2012 17:45:27 +0000 (17:45 +0000)]
When using a symbol with attribute weakref, search for it first and
don't try the normal GetOrCreateLLVM. The latter could drop the weak
atrtibute on the second reference, if there is no explicit declaration
of the aliasee.
llvm-svn: 166032
Owen Anderson [Tue, 16 Oct 2012 17:10:33 +0000 (17:10 +0000)]
Speculative fix the mask constants to be of type uintptr_t. I don't know of any case where the old form was incorrect, but I'm more confident that such cases don't exist in this version.
llvm-svn: 166031
Fariborz Jahanian [Tue, 16 Oct 2012 17:08:11 +0000 (17:08 +0000)]
Minor tweak to last patch along with a test case.
// rdar://
12491143
llvm-svn: 166030
Alexander Potapenko [Tue, 16 Oct 2012 16:58:10 +0000 (16:58 +0000)]
Do not call ReplaceCFAllocator() before __CFInitialize(), otherwise crashes are possible on 10.8.
Fixes http://code.google.com/p/address-sanitizer/issues/detail?id=122
llvm-svn: 166029
Alexander Potapenko [Tue, 16 Oct 2012 16:48:51 +0000 (16:48 +0000)]
Commit a leftover file from r166026: Add MACOS_VERSION_MOUNTAIN_LION.
llvm-svn: 166028
Alexander Potapenko [Tue, 16 Oct 2012 16:24:49 +0000 (16:24 +0000)]
Add MACOS_VERSION_MOUNTAIN_LION.
llvm-svn: 166026
Fariborz Jahanian [Tue, 16 Oct 2012 16:21:20 +0000 (16:21 +0000)]
fixes an objc crash involving objc bool literal on
hopelessly poorly written code after spewing several
errors. // rdar://
12491143
llvm-svn: 166025
Dmitri Gribenko [Tue, 16 Oct 2012 15:37:50 +0000 (15:37 +0000)]
Fix function parameter spelling in comments. Caught by -Wdocumentation.
llvm-svn: 166024
David Chisnall [Tue, 16 Oct 2012 15:11:55 +0000 (15:11 +0000)]
GNUstep runtime version default to 1.6, generate correct property attribute
metadata.
llvm-svn: 166023
Bill Schmidt [Tue, 16 Oct 2012 13:30:53 +0000 (13:30 +0000)]
This patch addresses PR13949.
For the PowerPC 64-bit ELF Linux ABI, aggregates of size less than 8
bytes are to be passed in the low-order bits ("right-adjusted") of the
doubleword register or memory slot assigned to them. A previous patch
addressed this for aggregates passed in registers. However, small
aggregates passed in the overflow portion of the parameter save area are
still being passed left-adjusted.
The fix is made in PPCTargetLowering::LowerCall_Darwin_Or_64SVR4 on the
caller side, and in PPCTargetLowering::LowerFormalArguments_64SVR4 on
the callee side. The main fix on the callee side simply extends
existing logic for 1- and 2-byte objects to 1- through 7-byte objects,
and correcting a constant left over from 32-bit code. There is also a
fix to a bogus calculation of the offset to the following argument in
the parameter save area.
On the caller side, again a constant left over from 32-bit code is
fixed. Additionally, some code for 1, 2, and 4-byte objects is
duplicated to handle the 3, 5, 6, and 7-byte objects for SVR4 only. The
LowerCall_Darwin_Or_64SVR4 logic is getting fairly convoluted trying to
handle both ABIs, and I propose to separate this into two functions in a
future patch, at which time the duplication can be removed.
The patch adds a new test (structsinmem.ll) to demonstrate correct
passing of structures of all seven sizes. Eight dummy parameters are
used to force these structures to be in the overflow portion of the
parameter save area.
As a side effect, this corrects the case when aggregates passed in
registers are saved into the first eight doublewords of the parameter
save area: Previously they were stored left-justified, and now are
properly stored right-justified. This requires changing the expected
output of existing test case structsinregs.ll.
llvm-svn: 166022
Tobias Grosser [Tue, 16 Oct 2012 07:29:19 +0000 (07:29 +0000)]
isl scheduler: Do not fail when returning an empty band list
The bug was within isl. To fix it, we simply update the isl version that
is used by Polly. We still have some changes within Polly to be able to
write a proper test case.
Reported-by: Sameer Sahasrabuddhe <Sameer.Sahasrabuddhe@amd.com>
llvm-svn: 166021
Tobias Grosser [Tue, 16 Oct 2012 07:29:13 +0000 (07:29 +0000)]
isl-codegen: Support '<' and '>'
Previously isl always generated '<=' or '>='. However, in many cases '<' or '>'
leads to simpler code. This commit updates isl and adds the relevant code
generation support to Polly.
llvm-svn: 166020
Alexey Samsonov [Tue, 16 Oct 2012 07:22:28 +0000 (07:22 +0000)]
Transform pattern:
if (CGM.getModuleDebugInfo())
DebugInfo = CGM.getModuleDebugInfo()
into a call:
maybeInitializeDebugInfo();
This is a simplification for a possible future fix of PR13942.
llvm-svn: 166019
Stepan Dyatkovskiy [Tue, 16 Oct 2012 07:16:47 +0000 (07:16 +0000)]
Issue:
Stack is formed improperly for long structures passed as byval arguments for
EABI mode.
If we took AAPCS reference, we can found the next statements:
A: "If the argument requires double-word alignment (8-byte), the NCRN (Next
Core Register Number) is rounded up to the next even register number." (5.5
Parameter Passing, Stage C, C.3).
B: "The alignment of an aggregate shall be the alignment of its most-aligned
component." (4.3 Composite Types, 4.3.1 Aggregates).
So if we have structure with doubles (9 double fields) and 3 Core unused
registers (r1, r2, r3): caller should use r2 and r3 registers only.
Currently r1,r2,r3 set is used, but it is invalid.
Callee VA routine should also use r2 and r3 regs only. All is ok here. This
behaviour is guessed by rounding up SP address with ADD+BFC operations.
Fix:
Main fix is in ARMTargetLowering::HandleByVal. If we detected AAPCS mode and
8 byte alignment, we waste odd registers then.
P.S.:
I also improved LDRB_POST_IMM regression test. Since ldrb instruction will
not generated by current regression test after this patch.
llvm-svn: 166018
NAKAMURA Takumi [Tue, 16 Oct 2012 06:28:34 +0000 (06:28 +0000)]
Reapply r165661, Patch by Shuxin Yang <shuxin.llvm@gmail.com>.
Original message:
The attached is the fix to radar://
11663049. The optimization can be outlined by following rules:
(select (x != c), e, c) -> select (x != c), e, x),
(select (x == c), c, e) -> select (x == c), x, e)
where the <c> is an integer constant.
The reason for this change is that : on x86, conditional-move-from-constant needs two instructions;
however, conditional-move-from-register need only one instruction.
While the LowerSELECT() sounds to be the most convenient place for this optimization, it turns out to be a bad place. The reason is that by replacing the constant <c> with a symbolic value, it obscure some instruction-combining opportunities which would otherwise be very easy to spot. For that reason, I have to postpone the change to last instruction-combining phase.
The change passes the test of "make check-all -C <build-root/test" and "make -C project/test-suite/SingleSource".
Original message since r165661:
My previous change has a bug: I negated the condition code of a CMOV, and go ahead creating a new CMOV using the *ORIGINAL* condition code.
llvm-svn: 166017
Bill Wendling [Tue, 16 Oct 2012 06:10:45 +0000 (06:10 +0000)]
Cleanup whitespace.
llvm-svn: 166016
Owen Anderson [Tue, 16 Oct 2012 06:04:27 +0000 (06:04 +0000)]
Fix a bug in the set(I,E)/reset(I,E) methods that I recently added. The boundary condition for checking if I and E were in the same word were incorrect, and, beyond that, the mask computation was not using a wide enough constant.
llvm-svn: 166015
Craig Topper [Tue, 16 Oct 2012 06:01:50 +0000 (06:01 +0000)]
Move X86MCInstLower class definition into implementation file. It's not needed outside.
llvm-svn: 166014
Bill Wendling [Tue, 16 Oct 2012 06:01:44 +0000 (06:01 +0000)]
Cleanup whitespace.
llvm-svn: 166013
Bill Wendling [Tue, 16 Oct 2012 05:57:28 +0000 (05:57 +0000)]
Have AttributesImpl defriend the Attributes class.
llvm-svn: 166012
Bill Wendling [Tue, 16 Oct 2012 05:55:09 +0000 (05:55 +0000)]
Have AttrBuilder defriend the Attributes class.
llvm-svn: 166011
Bill Wendling [Tue, 16 Oct 2012 05:23:44 +0000 (05:23 +0000)]
Use the Attributes::get method which takes an AttrVal value directly to simplify the code a bit. No functionality change.
llvm-svn: 166010
Bill Wendling [Tue, 16 Oct 2012 05:23:31 +0000 (05:23 +0000)]
Use the Attributes::get method which takes an AttrVal value directly to simplify the code a bit. No functionality change.
llvm-svn: 166009
Bill Wendling [Tue, 16 Oct 2012 05:22:28 +0000 (05:22 +0000)]
Put simple c'tors inline.
llvm-svn: 166008
Bill Wendling [Tue, 16 Oct 2012 05:20:51 +0000 (05:20 +0000)]
Pass in the context to the Attributes::get method.
llvm-svn: 166007
Kostya Serebryany [Tue, 16 Oct 2012 04:50:32 +0000 (04:50 +0000)]
[asan] fix lint
llvm-svn: 166006
Jason Molenda [Tue, 16 Oct 2012 02:39:21 +0000 (02:39 +0000)]
Add a new ABI plugin method which specifies whether the architecture
must push something on the stack for a function call or not. In
x86, the stack pointer is decremented when the caller's pc is saved
on the stack. In arm, the stack pointer and frame pointer don't
necessarily have to change for a function call, although most
functions need to use some stack space during their execution.
Use this information in the RegisterContextLLDB to detect invalid
unwind scenarios more accurately.
<rdar://problem/
12348574>
llvm-svn: 166005
Craig Topper [Tue, 16 Oct 2012 02:21:30 +0000 (02:21 +0000)]
Fix filename in file header.
llvm-svn: 166004
Rafael Espindola [Tue, 16 Oct 2012 01:13:06 +0000 (01:13 +0000)]
Fix the cpu name and add -verify-machineinstrs.
llvm-svn: 166003
Jordan Rose [Tue, 16 Oct 2012 00:47:25 +0000 (00:47 +0000)]
[analyzer] ObjCContainersASTChecker: minor cleanup and an extra test case.
Follow-up to r165838, which fixed a potential crash.
llvm-svn: 166002
Andrew Trick [Tue, 16 Oct 2012 00:22:51 +0000 (00:22 +0000)]
misched: Added handleMove support for updating all kill flags, not just for allocatable regs.
This is a medium term workaround until we have a more robust solution
in the form of a register liveness utility for postRA passes.
llvm-svn: 166001
Jim Ingham [Tue, 16 Oct 2012 00:09:33 +0000 (00:09 +0000)]
Patch from Matt Kopec <matt.kopec@intel.com> to fix the problem that if two breakpoints were set on consecutive addresses, the continue from the
first breakpoint would skip the second.
llvm-svn: 166000
Jakob Stoklund Olesen [Tue, 16 Oct 2012 00:05:06 +0000 (00:05 +0000)]
Remove unused BitVectors from getAllocatableSet().
llvm-svn: 165999