Richard Sandiford [Thu, 10 Jul 2014 10:45:11 +0000 (10:45 +0000)]
[SystemZ] Fix FPR dwarf numbering
The dwarf FPR numbers are supposed to have the order F0, F2, F4, F6,
F1, F3, F5, F7, F8, etc., which matches the pairing of registers for
long doubles. E.g. a long double stored in F0 is paired with F2.
llvm-svn: 212701
Daniel Sanders [Thu, 10 Jul 2014 10:39:51 +0000 (10:39 +0000)]
[mips] Add support for -modd-spreg/-mno-odd-spreg
Differential Revision: http://reviews.llvm.org/D4432
llvm-svn: 212700
Timur Iskhodzhanov [Thu, 10 Jul 2014 10:33:48 +0000 (10:33 +0000)]
[ASan/Win] Don't hardcode ASan runtime version in ASan dll thunk
Reviewed at http://reviews.llvm.org/D4459
llvm-svn: 212699
Jason Molenda [Thu, 10 Jul 2014 10:23:01 +0000 (10:23 +0000)]
Get the inferior binary's name via the command line argument instead
of hardcoding it.
llvm-svn: 212698
Daniel Sanders [Thu, 10 Jul 2014 10:18:12 +0000 (10:18 +0000)]
Make it possible for ints/floats to return different values from getBooleanContents()
Summary:
On MIPS32r6/MIPS64r6, floating point comparisons return 0 or -1 but integer
comparisons return 0 or 1.
Updated the various uses of getBooleanContents. Two simplifications had to be
disabled when float and int boolean contents differ:
- ScalarizeVecRes_VSELECT except when the kind of boolean contents is trivially
discoverable (i.e. when the condition of the VSELECT is a SETCC node).
- visitVSELECT (select C, 0, 1) -> (xor C, 1).
Come to think of it, this one could test for the common case of 'C'
being a SETCC too.
Preserved existing behaviour for all other targets and updated the affected
MIPS32r6/MIPS64r6 tests. This also fixes the pi benchmark where the 'low'
variable was counting in the wrong direction because it thought it could simply
add the result of the comparison.
Reviewers: hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, jholewinski, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D4389
llvm-svn: 212697
Tim Northover [Thu, 10 Jul 2014 10:06:42 +0000 (10:06 +0000)]
MachO: add dummy armv6m architecture directory so clang_macho_embedded builds things.
Without some mention of armv6m in a subdirectory of builtins, the make code
doesn't even know that armv6m exists and is something it should be looking for
in the platform-specific Makefiles. This means that none of the functions
listed actually get built and we end up with an almost entirely empty
libclang_rt.a for armv6m.
Unfortunately, the assembly code in the usual arm directory has no hope of
running on armv6m, which only supports Thumb-1 (not even ARM mode), so adding
it there won't work. Realistically, we probably *will* want to put any
optimised versions in a separate directory, so creating it now is harmless.
rdar://problem/
17613576
llvm-svn: 212696
Chandler Carruth [Thu, 10 Jul 2014 09:57:36 +0000 (09:57 +0000)]
[x86] Expand the target DAG combining for PSHUFD nodes to be able to
combine into half-shuffles through unpack instructions that expand the
half to a whole vector without messing with the dword lanes.
This fixes some redundant instructions in splat-like lowerings for
v16i8, which are now getting to be *really* nice.
llvm-svn: 212695
Jason Molenda [Thu, 10 Jul 2014 09:55:19 +0000 (09:55 +0000)]
Fix test name.
llvm-svn: 212694
Viktor Kutuzov [Thu, 10 Jul 2014 09:16:58 +0000 (09:16 +0000)]
Make sanitizers' interceptors non-weak on FreeBSD
Differential Revision: http://reviews.llvm.org/D4418
llvm-svn: 212693
Chandler Carruth [Thu, 10 Jul 2014 09:16:40 +0000 (09:16 +0000)]
[x86] Tweak the v16i8 single input special case lowering for shuffles
that splat i8s into i16s.
Previously, we would try much too hard to arrange a sequence of i8s in
one half of the input such that we could unpack them into i16s and
shuffle those into place. This isn't always going to be a cheaper i8
shuffle than our other strategies. The case where it is always going to
be cheaper is when we can arrange all the necessary inputs into one half
using just i16 shuffles. It happens that viewing the problem this way
also makes it much easier to produce an efficient set of shuffles to
move the inputs into one half and then unpack them.
With this, our splat code gets one step closer to being not terrible
with the new experimental lowering strategy. It also exposes two
combines missing which I will add next.
llvm-svn: 212692
Viktor Kutuzov [Thu, 10 Jul 2014 09:11:15 +0000 (09:11 +0000)]
Only use fast stack unwinding for sanitizers failure reports on FreeBSD
Differential Revision: http://reviews.llvm.org/D4420
llvm-svn: 212691
Viktor Kutuzov [Thu, 10 Jul 2014 08:53:29 +0000 (08:53 +0000)]
Fix getting environment variables for sanitizers needs on FreeBSD
Differential Revision: http://reviews.llvm.org/D4229
llvm-svn: 212690
Viktor Kutuzov [Thu, 10 Jul 2014 08:43:39 +0000 (08:43 +0000)]
Add clang headers that fix machine-dependent definitions on FreeBSD 9.2
Differential Revision: http://reviews.llvm.org/D3908
llvm-svn: 212689
Hal Finkel [Thu, 10 Jul 2014 07:04:37 +0000 (07:04 +0000)]
A test case for not asserting in isDereferenceablePointer upon unsized types
This is the test case for r212687.
llvm-svn: 212688
Hal Finkel [Thu, 10 Jul 2014 06:06:11 +0000 (06:06 +0000)]
Fix isDereferenceablePointer not to try to take the size of an unsized type.
I'll add a test-case shortly.
llvm-svn: 212687
Hal Finkel [Thu, 10 Jul 2014 05:27:53 +0000 (05:27 +0000)]
Allow isDereferenceablePointer to look through some bitcasts
isDereferenceablePointer should not give up upon encountering any bitcast. If
we're casting from a pointer to a larger type to a pointer to a small type, we
can continue by examining the bitcast's operand. This missing capability
was noted in a comment in the function.
In order for this to work, isDereferenceablePointer now takes an optional
DataLayout pointer (essentially all callers already had such a pointer
available). Most code uses isDereferenceablePointer though
isSafeToSpeculativelyExecute (which already took an optional DataLayout
pointer), and to enable the LICM test case, LICM needs to actually provide its DL
pointer to isSafeToSpeculativelyExecute (which it was not doing previously).
llvm-svn: 212686
Todd Fiala [Thu, 10 Jul 2014 05:25:39 +0000 (05:25 +0000)]
Fix a type mismatch in NativeProcessLinux that shows up in 32-bit builds.
http://llvm.org/bugs/show_bug.cgi?id=20255
llvm-svn: 212685
Saleem Abdulrasool [Thu, 10 Jul 2014 04:50:09 +0000 (04:50 +0000)]
MC: modernise for loop
Convert a for loop to range bsaed form. NFC.
llvm-svn: 212684
Saleem Abdulrasool [Thu, 10 Jul 2014 04:50:06 +0000 (04:50 +0000)]
MC: add and use an accessor for WinCFI
This adds a utility method to access the WinCFI information in bulk and uses
that to iterate rather than requesting the count and individually iterating
them. This is in preparation for restructuring WinCFI handling to enable more
clear sharing across architectures to enable unwind information emission for
Windows on ARM.
llvm-svn: 212683
Peter Collingbourne [Thu, 10 Jul 2014 04:39:40 +0000 (04:39 +0000)]
Remove move assignment operator to appease older GCCs.
llvm-svn: 212682
Todd Fiala [Thu, 10 Jul 2014 04:39:13 +0000 (04:39 +0000)]
Fixes for broken Debian build - g++ 4.7 support.
These fix the broken debian lldb build, which is using g++ 4.7.2.
TypeFormat changes:
1. stopped using the C++11 "dtor = default;" construct.
The generated default destructor in the two derived classes wanted
them to have a different throws() semantic that was causing 4.7 to
fail to generate it. I switched these to empty destructors defined
in the .cpp file.
2. Switched the m_types map from an ordered map to an unordered_map.
g++ 4.7's c++ library supports the C++11 emplace() used by TypeFormat
but the same c++ library's map impl does not. Since TypeFormat didn't
look like it depended on ordering in the map, I just switched it to
a std::unordered_map.
NativeProcessLinux - g++ 4.7 chokes on lexing the "<::" in
static_cast<::pid_t>(wpid). g++ 4.8+ and clang are fine with it.
I just put a space in between the "<" and the "::" and that cleared
it up.
llvm-svn: 212681
Chandler Carruth [Thu, 10 Jul 2014 04:34:06 +0000 (04:34 +0000)]
[x86] Initial improvements to the new shuffle lowering for v16i8
shuffles specifically for cases where a small subset of the elements in
the input vector are actually used.
This is specifically targetted at improving the shuffles generated for
trunc operations, but also helps out splat-like operations.
There is still some really low-hanging fruit here that I want to address
but this is a huge step in the right direction.
llvm-svn: 212680
Peter Collingbourne [Thu, 10 Jul 2014 04:29:06 +0000 (04:29 +0000)]
Explicitly define move constructor and move assignment operator to appease MSVC.
llvm-svn: 212679
Peter Collingbourne [Thu, 10 Jul 2014 03:55:02 +0000 (03:55 +0000)]
SpecialCaseList: use std::unique_ptr.
llvm-svn: 212678
Hao Liu [Thu, 10 Jul 2014 03:41:50 +0000 (03:41 +0000)]
[AArch64]Fix an assertion failure in DAG Combiner about concating 2 build_vector.
llvm-svn: 212677
Matt Arsenault [Thu, 10 Jul 2014 03:22:20 +0000 (03:22 +0000)]
R600/SI: Add support for llvm.convert.{to|from}.fp16
llvm-svn: 212676
Matt Arsenault [Thu, 10 Jul 2014 03:22:16 +0000 (03:22 +0000)]
Fix types in documentation.
The examples were using f32, but the IR type is called float
llvm-svn: 212675
Chandler Carruth [Thu, 10 Jul 2014 02:24:26 +0000 (02:24 +0000)]
[x86] Refactor some of the new code for lowering v16i8 shuffles to
remove duplication and make it easier to select different strategies.
No functionality changed.
llvm-svn: 212674
Jason Molenda [Thu, 10 Jul 2014 02:21:16 +0000 (02:21 +0000)]
Add a new 'stresstest' category, set the api/multiple-debuggers test case as a stresstest.
llvm-svn: 212673
Dan Albert [Thu, 10 Jul 2014 02:20:11 +0000 (02:20 +0000)]
Add lit configs for libcxxabi tests.
This makes running libcxxabi tests on Linux _much_ easier.
Adds a check-libcxxabi target to cmake.
Also defaults to building a dynamic libc++abi. This is so that the
default options still test the libc++abi that is being built. There are
two problems with testing a static libc++abi. In the case of a
standalone build, the tests will link the system's libc++, which might
not have been built against our libc++abi. In the case of an in tree
build, libc++ will prefer a dynamic libc++abi from the system over a
static libc++abi from the output directory.
llvm-svn: 212672
Jason Molenda [Thu, 10 Jul 2014 02:17:31 +0000 (02:17 +0000)]
Add a new test in api/multiple-debuggers which tries to create 50
debug sessions simultaneously to expose race conditoin/locking
issues.
This directory has an inferior program, testprog.cpp that has a
couple of functions we can put breakpoints on.
It has a driver program, multi-process-driver.cpp, which links
against the LLDB solib and uses the SB APIs. It creates 50 pthreads,
creates a debugger on all of them, launches a debug session of the
inferior testprog, hits a couple breakpoints, walks the stack,
continues, etc., and then kills the inferior and ends the debug
session.
A pass is if all fifty debug sessions complete successfully
in the alloted time (~60 seconds).
We may need to tweak this one to work correctly on different
platforms/targets but I wanted to get it checked in to start.
llvm-svn: 212671
Alp Toker [Thu, 10 Jul 2014 02:13:29 +0000 (02:13 +0000)]
Prevent link warnings due to -Wframe-larger-than=
Although this is nominally a -W option, we actually handle it in the driver
exactly as an f-group flag that's translated directly to -mllvm.
That means f_Group (and unintuitively, not W_Group) has the semantics we want
to make it behave like a standard warning flag: no automatic forwarding, no
warning for link invocations and compile-only.
Silences diagnostics like:
[691/1545] Linking CXX executable bin/llvm-diff
clang-3.5: warning: argument unused during compilation: '-Wframe-larger-than=2048'
(Hopefully we can move towards handling these in the frontend but that'll
require some infrastructure work.)
llvm-svn: 212670
Reid Kleckner [Thu, 10 Jul 2014 01:58:55 +0000 (01:58 +0000)]
MS ABI: Fix __fastcall methods that return structs
The sret paramater consumes the register after the implicit 'this'
parameter, as with other calling conventions.
Fixes PR20278, which turned out to be very easy.
llvm-svn: 212669
Peter Collingbourne [Thu, 10 Jul 2014 01:30:39 +0000 (01:30 +0000)]
[dfsan] Handle bitcast aliases.
llvm-svn: 212668
Richard Smith [Thu, 10 Jul 2014 01:20:17 +0000 (01:20 +0000)]
Cleanup. No functionality change.
llvm-svn: 212667
Argyrios Kyrtzidis [Thu, 10 Jul 2014 01:03:37 +0000 (01:03 +0000)]
[Driver] Expose getARMCPUForMArch() function in the Driver API; NFC.
llvm-svn: 212666
Alexey Samsonov [Thu, 10 Jul 2014 00:53:20 +0000 (00:53 +0000)]
Fix configure+make build of profile runtime library
llvm-svn: 212665
Alexey Samsonov [Thu, 10 Jul 2014 00:06:53 +0000 (00:06 +0000)]
[ASan] Force the compiler to generate frame pointer in certain Mac-specific interceptors.
Patch by Kuba Brecka!
llvm-svn: 212664
Alexey Samsonov [Wed, 9 Jul 2014 23:59:40 +0000 (23:59 +0000)]
[Sanitizer] Remove SANITIZER_NEEDS_SEGV from shared sanitizer_common headers.
Otherwise, it can be accidentally redefined when we build specific sanitizer
runtime. This definition should be provided only once - when we build
sanitizer_common library.
llvm-svn: 212663
Todd Fiala [Wed, 9 Jul 2014 23:10:43 +0000 (23:10 +0000)]
Skip tests that are intermittent on Linux, fix gdb-remote port-grabbing code.
Marked skipped for Linux:
TestCallStopAndContinue
TestConvenienceVariables
TestStopHookMultipleThreads
Fixed up gdb-remote port-grabbing code to use a random port in a wide range,
and to allow that to fail more gracefully. This appears to have solved some
gdb-remote intermittent failing behavior.
llvm-svn: 212662
Chandler Carruth [Wed, 9 Jul 2014 22:53:04 +0000 (22:53 +0000)]
[SDAG] Make the new zext-vector-inreg node default to expand so targets
don't need to set it manually.
This is based on feedback from Tom who pointed out that if every target
needs to handle this we need to reach out to those maintainers. In fact,
it doesn't make sense to duplicate everything when anything other than
expand seems unlikely at this stage.
llvm-svn: 212661
Todd Fiala [Wed, 9 Jul 2014 22:03:30 +0000 (22:03 +0000)]
Skip TestCallStopAndContinue and TestCallThatRestarts on Darwin.
These are failing intermittently.
See http://llvm.org/bugs/show_bug.cgi?id=19246 for TestCallThatRestarts. Also applies to Linux.
See http://llvm.org/bugs/show_bug.cgi?id=20274 for TestCallStopAndContinue.
llvm-svn: 212660
Todd Fiala [Wed, 9 Jul 2014 21:24:41 +0000 (21:24 +0000)]
Marked TestTargetAPI.test_launch_new_process_and_redirect_stdout_with_* as XFAIL on Darwin.
See http://llvm.org/bugs/show_bug.cgi?id=20273
llvm-svn: 212659
Alexander Kornienko [Wed, 9 Jul 2014 21:21:39 +0000 (21:21 +0000)]
Removed the "-fix" suffix from tests that verify both fixes and messages.
llvm-svn: 212658
Todd Fiala [Wed, 9 Jul 2014 21:21:39 +0000 (21:21 +0000)]
TestRealDefinition tests marked XFAIL on Darwin.
http://llvm.org/bugs/show_bug.cgi?id=20272
llvm-svn: 212657
Todd Fiala [Wed, 9 Jul 2014 21:19:03 +0000 (21:19 +0000)]
Marked failing TestObjCDynamicValue tests as XFAIL on Darwin.
See http://llvm.org/bugs/show_bug.cgi?id=20271
llvm-svn: 212656
Todd Fiala [Wed, 9 Jul 2014 21:15:46 +0000 (21:15 +0000)]
Marked failing TestObjCDynamicSBType tests XFAIL on Darwin.
See http://llvm.org/bugs/show_bug.cgi?id=20270
llvm-svn: 212655
Todd Fiala [Wed, 9 Jul 2014 21:10:13 +0000 (21:10 +0000)]
Marked failing TestHiddenIvars tests as XFAIL on Darwin.
See http://llvm.org/bugs/show_bug.cgi?id=20269
llvm-svn: 212654
Alexander Kornienko [Wed, 9 Jul 2014 21:09:26 +0000 (21:09 +0000)]
Check messages and fixes in the same test.
Summary: This reduces duplication of test code and improves locality of checks.
Reviewers: sbenza, djasper
Reviewed By: sbenza, djasper
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4437
llvm-svn: 212653
Nick Kledzik [Wed, 9 Jul 2014 21:04:24 +0000 (21:04 +0000)]
Move GOTPass and StubsPass from Core to MachO
llvm-svn: 212652
Marshall Clow [Wed, 9 Jul 2014 21:04:22 +0000 (21:04 +0000)]
Add a reset method to the counting predicates in the test suite
llvm-svn: 212651
Todd Fiala [Wed, 9 Jul 2014 21:02:52 +0000 (21:02 +0000)]
Marked TestObjCMethods failing tests XFAIL, fixed cleanup code, removed now-passing expected failure markers.
All tests matching '-p TestObjCMethods' now are marked correctly for MacOSX, and some
error classes have been removed in cleanup code looking for files that
might not exist due to previous failure.
See http://llvm.org/bugs/show_bug.cgi?id=20267
llvm-svn: 212650
David Blaikie [Wed, 9 Jul 2014 21:02:41 +0000 (21:02 +0000)]
Recommit r212203: Don't try to construct debug LexicalScopes hierarchy for functions that do not have top level debug information.
Reverted by Eric Christopher (Thanks!) in r212203 after Bob Wilson
reported LTO issues. Duncan Exon Smith and Aditya Nandakumar helped
provide a reduced reproduction, though the failure wasn't too hard to
guess, and even easier with the example to confirm.
The assertion that the subprogram metadata associated with an
llvm::Function matches the scope data referenced by the DbgLocs on the
instructions in that function is not valid under LTO. In LTO, a C++
inline function might exist in multiple CUs and the subprogram metadata
nodes will refer to the same llvm::Function. In this case, depending on
the order of the CUs, the first intance of the subprogram metadata may
not be the one referenced by the instructions in that function and the
assertion will fail.
A test case (test/DebugInfo/cross-cu-linkonce-distinct.ll) is added, the
assertion removed and a comment added to explain this situation.
Original commit message:
If a function isn't actually in a CU's subprogram list in the debug info
metadata, ignore all the DebugLocs and don't try to build scopes, track
variables, etc.
While this is possibly a minor optimization, it's also a correctness fix
for an incoming patch that will add assertions to LexicalScopes and the
debug info verifier to ensure that all scope chains lead to debug info
for the current function.
Fix up a few test cases that had broken/incomplete debug info that could
violate this constraint.
Add a test case where this occurs by design (inlining a
debug-info-having function in an attribute nodebug function - we want
this to work because /if/ the nodebug function is then inlined into a
debug-info-having function, it should be fine (and will work fine - we
just stitch the scopes up as usual), but should the inlining not happen
we need to not assert fail either).
llvm-svn: 212649
Todd Fiala [Wed, 9 Jul 2014 20:45:09 +0000 (20:45 +0000)]
Marked failing test XFAIL for TestRegisterVariables.test_with_dsym_and_run_command on Darwin
See http://llvm.org/bugs/show_bug.cgi?id=20266
llvm-svn: 212648
Todd Fiala [Wed, 9 Jul 2014 20:42:14 +0000 (20:42 +0000)]
Marked failing Darwin TestProcessLaunch tests as XFAIL
See http://llvm.org/bugs/show_bug.cgi?id=20265
llvm-svn: 212647
Todd Fiala [Wed, 9 Jul 2014 20:38:27 +0000 (20:38 +0000)]
TestDataFormatterStdVector.test_with_dsym_and_run_command marked XFAIL on Darwin
See http://llvm.org/bugs/show_bug.cgi?id=20264
llvm-svn: 212646
Todd Fiala [Wed, 9 Jul 2014 20:33:51 +0000 (20:33 +0000)]
Disable TestDataFormatterStdMap.test_with_dsym_and_run_command on Darwin
See http://llvm.org/bugs/show_bug.cgi?id=20263
llvm-svn: 212645
Greg Clayton [Wed, 9 Jul 2014 20:18:54 +0000 (20:18 +0000)]
Make sure the "command regex add" has a unique name for editline history purposes.
llvm-svn: 212644
Alexey Samsonov [Wed, 9 Jul 2014 19:40:08 +0000 (19:40 +0000)]
Decouple llvm::SpecialCaseList text representation and its LLVM IR semantics.
Turn llvm::SpecialCaseList into a simple class that parses text files in
a specified format and knows nothing about LLVM IR. Move this class into
LLVMSupport library. Implement two users of this class:
* DFSanABIList in DFSan instrumentation pass.
* SanitizerBlacklist in Clang CodeGen library.
The latter will be modified to use actual source-level information from frontend
(source file names) instead of unstable LLVM IR things (LLVM Module identifier).
Remove dependency edge from ClangCodeGen/ClangDriver to LLVMTransformUtils.
No functionality change.
llvm-svn: 212643
Alexey Samsonov [Wed, 9 Jul 2014 19:29:10 +0000 (19:29 +0000)]
Don't check lint for SpecialCaseList.cpp
llvm-svn: 212642
Tim Northover [Wed, 9 Jul 2014 19:14:34 +0000 (19:14 +0000)]
Use simpler constructor for range adapter.
It is a good idea, it's slightly clearer and simpler. Unfortunately
the headline news is: we save one line!
llvm-svn: 212641
Matt Arsenault [Wed, 9 Jul 2014 19:12:07 +0000 (19:12 +0000)]
Add trunc (select c, a, b) -> select c (trunc a), (trunc b) combine.
Do this if the truncate is free and the select is legal.
llvm-svn: 212640
Todd Fiala [Wed, 9 Jul 2014 19:00:21 +0000 (19:00 +0000)]
Mark failing tests in TestDataFormatterObjC on Darwin as XFAIL
See http://llvm.org/bugs/show_bug.cgi?id=20260 for more details.
llvm-svn: 212639
Jim Grosbach [Wed, 9 Jul 2014 18:55:52 +0000 (18:55 +0000)]
AArch64: Better codegen for storing to __fp16.
Storing will generally be immediately preceded by rounding from an f32
or f64, so make sure to match those patterns directly to convert into the
FPR16 register class directly rather than going through the integer GPRs.
This also eliminates an extra step in the convert-from-f64 path
which was first converting to f32 and then to f16 from there.
rdar://
17594379
llvm-svn: 212638
Jim Grosbach [Wed, 9 Jul 2014 18:55:49 +0000 (18:55 +0000)]
Change an assert() to a diagnostic.
llvm-svn: 212637
Benjamin Kramer [Wed, 9 Jul 2014 18:53:57 +0000 (18:53 +0000)]
TargetRegisterInfo: Remove function that fell out of use years ago.
llvm-svn: 212636
Cameron McInally [Wed, 9 Jul 2014 18:29:55 +0000 (18:29 +0000)]
Update ReleaseNotes to mention Atomic NAND semantic changes.
llvm-svn: 212635
Adam Nemet [Wed, 9 Jul 2014 18:22:33 +0000 (18:22 +0000)]
[X86] AVX512: Enable it in the Loop Vectorizer
This lets us experiment with 512-bit vectorization without passing
force-vector-width manually.
The code generated for a simple integer memset loop is properly vectorized.
Disassembly is still broken for it though :(.
llvm-svn: 212634
Louis Gerbarg [Wed, 9 Jul 2014 17:54:32 +0000 (17:54 +0000)]
Make AArch64FastISel::EmitIntExt explicitly check its source and destination types
This is a follow up to r212492. There should be no functional difference, but
this patch makes it clear that SrcVT must be an i1/i8/16/i32 and DestVT must be
an i8/i16/i32/i64.
rdar://
17516686
llvm-svn: 212633
Sanjay Patel [Wed, 9 Jul 2014 17:49:58 +0000 (17:49 +0000)]
removed duplicate testcase
llvm-svn: 212632
David Majnemer [Wed, 9 Jul 2014 17:15:52 +0000 (17:15 +0000)]
Sema: Allow aliases to have incomplete type
gcc supports this behavior and it is pervasively used inside the Linux
kernel.
Note that both gcc and clang will reject code that attempts to do this
in a C++ language mode.
This fixes PR17998.
llvm-svn: 212631
Zachary Turner [Wed, 9 Jul 2014 16:42:27 +0000 (16:42 +0000)]
Dont' use a random probe & alloc strategy for the IRMemoryMap.
The current strategy for host allocation is to choose a random
address and attempt to allocate there, eventually failing if the
allocation cannot be satisfied.
The C standard only guarantees that RAND_MAX >= 32767, so for
platforms that use a very small RAND_MAX allocations will fail
with very high probability. On such platforms (Windows is one),
you can reproduce this trivially by running lldb, typing "expr (3)"
and then hitting enter you see a failure. Failures generally
happen with a frequency of about 1 failure every 5 evaluations.
There is no good reason that allocations need to look like "real"
pointers, so this patch changes the allocation scheme to simply
jump straight to the end and grab a free chunk of memory.
Reviewed By: Sean Callanan
Differential Revision: http://reviews.llvm.org/D4300
llvm-svn: 212630
Sanjay Patel [Wed, 9 Jul 2014 16:34:54 +0000 (16:34 +0000)]
Fix for PR20059 (instcombine reorders shufflevector after instruction that may trap)
In PR20059 ( http://llvm.org/pr20059 ), instcombine eliminates shuffles that are necessary before performing an operation that can trap (srem).
This patch calls isSafeToSpeculativelyExecute() and bails out of the optimization in SimplifyVectorOp() if needed.
Differential Revision: http://reviews.llvm.org/D4424
llvm-svn: 212629
Zachary Turner [Wed, 9 Jul 2014 16:32:07 +0000 (16:32 +0000)]
Fix tests broken by the OptionValidator changes.
The getopt library has a structure called option (lowercase). We
have a structure called Option (uppercase). previously the two
structures had exactly the same definitions, and we were doing a
C-style cast of an Option* to an option*. C-style casts don't
bother to warn you when you cast to unrelated types, but in the
original OptionValidator patch I modified the definition of Option.
This patch fixes the errors by building an array of option
structures and filling it out the correct way before passing it to
the getopt library.
This also fixes one other source of test failures: an uninitialized
read that occurs due to not initializing a field of the
OptionDefinition.
Reviewed By: Todd Fiala
Differential Revision: http://reviews.llvm.org/D4425
llvm-svn: 212628
Zachary Turner [Wed, 9 Jul 2014 16:31:49 +0000 (16:31 +0000)]
Revert "Fix broken tests due to new error output."
This reverts commit
ec7c94f8e6860968d384b578e5564a9c55c80b4a and
re-enables OptionValidators.
llvm-svn: 212627
Daniel Sanders [Wed, 9 Jul 2014 16:03:10 +0000 (16:03 +0000)]
Add Imagination Technologies to the vendors in llvm::Triple
Summary: This is a pre-requisite for supporting the mips-img-linux-gnu triple in clang.
Differential Revision: http://reviews.llvm.org/D4435
llvm-svn: 212626
Daniel Sanders [Wed, 9 Jul 2014 14:21:18 +0000 (14:21 +0000)]
[mips][mips64r6] Implement -mips32r6 and -mips64r6 aliases to -march=mips32r6 and -march=mips64r6
Differential Revision: http://reviews.llvm.org/D4434
llvm-svn: 212625
Alp Toker [Wed, 9 Jul 2014 14:12:55 +0000 (14:12 +0000)]
Prospective legacy build system fix following r212620
llvm-svn: 212624
Alp Toker [Wed, 9 Jul 2014 14:09:52 +0000 (14:09 +0000)]
Remove dead code from r212620
llvm-svn: 212622
Alp Toker [Wed, 9 Jul 2014 14:06:35 +0000 (14:06 +0000)]
Fix 'source-level' hyphenations
llvm-svn: 212621
Alp Toker [Wed, 9 Jul 2014 14:05:11 +0000 (14:05 +0000)]
cc1as: consolidate option flags with cc1 and eliminate duplication
The clang -cc1as options are nearly a strict subset of -cc1. Instead of
duplicating the definitions and documentation, let's go ahead and share the
definitions in a similar way the current handling of combined driver and
frontend flags, eliminating some of the vestigial legacy surrounding the
assembler subcommand.
llvm-svn: 212620
Daniel Sanders [Wed, 9 Jul 2014 13:56:23 +0000 (13:56 +0000)]
[mips][mips64r6] Define _MIPS_FPSET, __mips_fpr, and __mips_nan2008 correctly on MIPS32r6/MIPS64r6
Summary:
This removes the need to pass -mnan=2008 explicitly to be able to compile
the test-suite for MIPS32r6/MIPS64r6.
Differential Revision: http://reviews.llvm.org/D4433
llvm-svn: 212619
Daniel Sanders [Wed, 9 Jul 2014 13:43:19 +0000 (13:43 +0000)]
[mips] clz is defined to give 32 for zero. Similarly, dclz gives 64.
Summary:
While debugging another issue, I noticed that Mips currently specifies that the
count leading zero builtins are undefined when the input is zero. The
architecture specifications say that the clz and dclz instructions write 32 or
64 respectively when given zero.
This doesn't fix any bugs that I'm aware of but it may improve optimisation in
some cases.
Differential Revision: http://reviews.llvm.org/D4431
llvm-svn: 212618
Daniel Jasper [Wed, 9 Jul 2014 13:07:57 +0000 (13:07 +0000)]
clang-format: Fix behavior around pointer-to-member invocations.
Before:
(
aaaaaaaaaa->*
bbbbbbb)(
aaaaaaaaaaaaaaaaaaaaaaaaaaa(
aaaaaaaaaaaaaaaaaaaaaaaaaaa));
After:
(
aaaaaaaaaa->*bbbbbbb)(
aaaaaaaaaaaaaaaaaaaaaaaaaaa(
aaaaaaaaaaaaaaaaaaaaaaaaaaa));
llvm-svn: 212617
Tim Northover [Wed, 9 Jul 2014 13:03:54 +0000 (13:03 +0000)]
[all]: Use range-based ArgList adapter instead of filtered_begin/filtered_end
Some of those loops were pretty monstrous.
llvm-svn: 212616
Tim Northover [Wed, 9 Jul 2014 13:03:37 +0000 (13:03 +0000)]
Generic: add range-adapter for option parsing.
I want to use it in lld, but while I'm here I'll update LLVM uses.
llvm-svn: 212615
Chandler Carruth [Wed, 9 Jul 2014 12:36:54 +0000 (12:36 +0000)]
[x86] Fix a bug in my new zext-vector-inreg DAG trickery where we were
not widening the input type to the node sufficiently to let the ext take
place in a register.
This would in turn result in a mysterious bitcast assertion failure
downstream. First change here is to add back the helpful assert I had in
an earlier version of the code to catch this immediately.
Next change is to add support to the type legalization to detect when we
have widened the operand either too little or too much (for whatever
reason) and find a size-matched legal vector type to convert it to
first. This can also fail so we get a new fallback path, but that seems
OK.
With this, we no longer crash on vec_cast2.ll when using widening. I've
also added the CHECK lines for the zero-extend cases here. We still need
to support sign-extend and trunc (or something) to get plausible code
for the other two thirds of this test which is one of the regression
tests that showed the most scalarization when widening was
force-enabled. Slowly closing in on widening being a viable legalization
strategy without it resorting to scalarization at every turn. =]
llvm-svn: 212614
Simon Atanasyan [Wed, 9 Jul 2014 12:31:11 +0000 (12:31 +0000)]
[Mips] Make rel-dynamic-08.test test case independent from external input files.
llvm-svn: 212613
Chandler Carruth [Wed, 9 Jul 2014 11:13:16 +0000 (11:13 +0000)]
Sink two variables only used in an assert into the assert itself. Should
fix the release builds with Werror.
llvm-svn: 212612
Benjamin Kramer [Wed, 9 Jul 2014 11:12:39 +0000 (11:12 +0000)]
X86: When lowering v8i32 himuls use the correct shuffle masks for AVX2.
Turns out my trick of using the same masks for SSE4.1 and AVX2 didn't work out
as we have to blend two vectors. While there remove unecessary cross-lane moves
from the shuffles so the backend can lower it to palignr instead of vperm.
Fixes PR20118, a miscompilation of vector sdiv by constant on AVX2.
llvm-svn: 212611
Chandler Carruth [Wed, 9 Jul 2014 10:58:18 +0000 (10:58 +0000)]
[x86] Add a ZERO_EXTEND_VECTOR_INREG DAG node and use it when widening
vector types to be legal and a ZERO_EXTEND node is encountered.
When we use widening to legalize vector types, extend nodes are a real
challenge. Either the input or output is likely to be legal, but in many
cases not both. As a consequence, we don't really have any way to
represent this situation and the prior code in the widening legalization
framework would just scalarize the extend operation completely.
This patch introduces a new DAG node to represent doing a zero extend of
a vector "in register". The core of the idea is to allow legal but
different vector types in the input and output. The output vector must
have fewer lanes but wider elements. The operation is defined to zero
extend the low elements of the input to the size of the output elements,
and drop all of the high elements which don't have a corresponding lane
in the output vector.
It also includes generic expansion of this node in terms of blending
a zero vector into the high elements of the vector and bitcasting
across. This in turn yields extremely nice code for x86 SSE2 when we use
the new widening legalization logic in conjunction with the new shuffle
lowering logic.
There is still more to do here. We need to support sign extension, any
extension, and potentially int-to-float conversions. My current plan is
to continue using similar synthetic nodes to model each of these
transitions with generic lowering code for each one.
However, with this patch LLVM already reaches performance parity with
GCC for the core C loops of the x264 code (assuming you disable the
hand-written assembly versions) when compiling for SSE2 and SSE3
architectures and enabling the new widening and lowering logic for
vectors.
Differential Revision: http://reviews.llvm.org/D4405
llvm-svn: 212610
Tobias Grosser [Wed, 9 Jul 2014 10:50:10 +0000 (10:50 +0000)]
clang-format polly to avoid buildbot noise
llvm-svn: 212609
Daniel Sanders [Wed, 9 Jul 2014 10:47:26 +0000 (10:47 +0000)]
[mips][mips64r6] Correct select patterns that have the condition or true/false values backwards
Summary: This bug caused SingleSource/Regression/C/uint64_to_float and SingleSource/UnitTests/2002-05-02-CastTest3 to fail (among others).
Differential Revision: http://reviews.llvm.org/D4388
llvm-svn: 212608
Daniel Sanders [Wed, 9 Jul 2014 10:40:20 +0000 (10:40 +0000)]
[mips][mips64r6] Correct cond names in the cmp.cond.[ds] instructions
Summary:
It seems we accidentally read the wrong column of the table MIPS64r6 spec
and used the names for c.cond.fmt instead of cmp.cond.fmt.
Differential Revision: http://reviews.llvm.org/D4387
llvm-svn: 212607
Chandler Carruth [Wed, 9 Jul 2014 10:36:42 +0000 (10:36 +0000)]
[x86] Initialize a pointer to null to fix a bug in r212602.
This should restore GCC hosts (which happen to put the bad stuff into
the pointer) and MSan, etc.
llvm-svn: 212606
Daniel Sanders [Wed, 9 Jul 2014 10:21:59 +0000 (10:21 +0000)]
[mips][mips64r6] Use JALR for indirect branches instead of JR (which is not available on MIPS32r6/MIPS64r6)
Summary:
This completes the change to use JALR instead of JR on MIPS32r6/MIPS64r6.
Reviewers: jkolek, vmedic, zoran.jovanovic, dsanders
Reviewed By: dsanders
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D4269
llvm-svn: 212605
Daniel Sanders [Wed, 9 Jul 2014 10:16:07 +0000 (10:16 +0000)]
[mips][mips64r6] Use JALR for returns instead of JR (which is not available on MIPS32r6/MIPS64r6)
Summary:
RET, and RET_MM have been replaced by a pseudo named PseudoReturn.
In addition a version with a 64-bit GPR named PseudoReturn64 has been
added.
Instruction selection for a return matches RetRA, which is expanded post
register allocation to PseudoReturn/PseudoReturn64. During MipsAsmPrinter,
this PseudoReturn/PseudoReturn64 are emitted as:
- (JALR64 $zero, $rs) on MIPS64r6
- (JALR $zero, $rs) on MIPS32r6
- (JR_MM $rs) on microMIPS
- (JR $rs) otherwise
On MIPS32r6/MIPS64r6, 'jr $rs' is an alias for 'jalr $zero, $rs'. To aid
development and review (specifically, to ensure all cases of jr are
updated), these aliases are temporarily named 'r6.jr' instead of 'jr'.
A follow up patch will change them back to the correct mnemonic.
Added (JALR $zero, $rs) to MipsNaClELFStreamer's definition of an indirect
jump, and removed it from its definition of a call.
Note: I haven't accounted for MIPS64 in MipsNaClELFStreamer since it's
doesn't appear to account for any MIPS64-specifics.
The return instruction created as part of eh_return expansion is now expanded
using expandRetRA() so we use the right return instruction on MIPS32r6/MIPS64r6
('jalr $zero, $rs').
Also, fixed a misuse of isABI_N64() to detect 64-bit wide registers in
expandEhReturn().
Reviewers: jkolek, vmedic, mseaborn, zoran.jovanovic, dsanders
Reviewed By: dsanders
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D4268
llvm-svn: 212604
Daniel Sanders [Wed, 9 Jul 2014 10:07:36 +0000 (10:07 +0000)]
Add ability to emit internal instruction representation to CodeGen assembly output.
Summary:
This patch re-uses the implementation of 'llvm-mc -show-inst' and makes it
available to llc as 'llc -asm-show-inst'.
This is necessary to test parts of MIPS32r6/MIPS64r6 without resorting to
'llc -filetype=obj' tests. For example, on MIPS32r2 and earlier we use the
'jr $rs' instruction for indirect branches and returns. On MIPS32r6, we no
longer have 'jr $rs' and use 'jalr $zero, $rs' instead. The catch is that,
on MIPS32r6, 'jr $rs' is an alias for 'jalr $zero, $rs' and is the preferred
way of writing this instruction. As a result, all MIPS ISA's emit 'jr $rs' in
their assembly output and the assembler encodes this to different opcodes
according to the ISA.
Using this option, we can check that the MCInst really is a JR or a JALR by
matching the emitted comment. This removes the need for a 'llc -filetype=obj'
test.
Reviewers: rafael, dsanders
Reviewed By: dsanders
Subscribers: zoran.jovanovic, llvm-commits
Differential Revision: http://reviews.llvm.org/D4267
llvm-svn: 212603
Chandler Carruth [Wed, 9 Jul 2014 10:06:58 +0000 (10:06 +0000)]
[x86] Re-apply a variant of the x86 side of r212324 now that the rest
has settled without incident, removing the x86-specific and overly
strict 'isVectorSplat' routine in favor of generic and more powerful
splat detection.
The primary motivation and result of this is that the x86 backend can
now see through splats which contain undef elements. This is essential
if we are using a widening form of legalization and I've updated a test
case to also run in that mode as before this change the generated code
for the test case was completely scalarized.
This version of the patch much more carefully handles the undef lanes.
- We aren't overly conservative about them in the shift lowering
(where we will never use the splat itself).
- One place where the splat would have been re-used by the existing code
now explicitly constructs a new constant splat that will be safe.
- The broadcast lowering is much more reasonable with undefs by doing
a correct check of whether the splat is the only user of a loaded
value, checking that the splat actually crosses multiple lanes before
using a broadcast, and handling broadcasts of non-constant splats.
As a consequence of the last bullet, the weird usage of vpshufd instead
of vbroadcast is gone, and we actually can lower an AVX splat with
vbroadcastss where before we emitted a really strange pattern of
a vector load and a manual splat across the vector.
llvm-svn: 212602
Alexey Bataev [Wed, 9 Jul 2014 10:00:31 +0000 (10:00 +0000)]
Fixed incompatibility with MSVC builds in range-based for loops in DataRecursiveASTVisitor and RecursiveASTVisitor for OpenMP constructs.
llvm-svn: 212601