Colin LeMahieu [Wed, 4 Feb 2015 00:09:23 +0000 (00:09 +0000)]
[Hexagon] Revert change to isCodeGenOnly = 1 in r228080
llvm-svn: 228082
Colin LeMahieu [Wed, 4 Feb 2015 00:07:26 +0000 (00:07 +0000)]
[Hexagon] Changing some isCodeGenOnly to isAsmParserOnly since we want them to asm parse but not cause decode conflicts.
llvm-svn: 228080
Owen Anderson [Wed, 4 Feb 2015 00:02:59 +0000 (00:02 +0000)]
Remove a gross usage of environment variables in MachineVerifier, replacing it with support for setting the -verify-machineinstrs flag via an environment variable in LIT.
This preserves the handy functionality of force-enabling the MachineVerifier, without the need to embed usage of environment variables in LLVM client applications.
llvm-svn: 228079
Rui Ueyama [Wed, 4 Feb 2015 00:00:21 +0000 (00:00 +0000)]
Simplify large switches.
This may be a little bit inefficient than the original code
but that should be okay as this is not really in a performance
critical pass.
http://reviews.llvm.org/D7393
llvm-svn: 228077
Justin Bogner [Tue, 3 Feb 2015 23:59:48 +0000 (23:59 +0000)]
InstrProf: Update for LLVM API change
Update for the API change in r228075
llvm-svn: 228076
Justin Bogner [Tue, 3 Feb 2015 23:59:33 +0000 (23:59 +0000)]
InstrProf: Make CounterMappingRegions less confusing to construct
Creating empty and expansion regions is awkward with the current API.
Expose static methods to make this simpler.
llvm-svn: 228075
Eric Fiselier [Tue, 3 Feb 2015 23:50:47 +0000 (23:50 +0000)]
Make test require 'linux' instead of 'linux2'
llvm-svn: 228073
Eric Fiselier [Tue, 3 Feb 2015 23:49:58 +0000 (23:49 +0000)]
add platform to LIT's available features
llvm-svn: 228071
Davide Italiano [Tue, 3 Feb 2015 23:44:33 +0000 (23:44 +0000)]
Avoid two function calls of file() when not needed.
Reported by: ruiu
llvm-svn: 228069
Arnaud A. de Grandmaison [Tue, 3 Feb 2015 23:40:24 +0000 (23:40 +0000)]
[PBQP] Provide more information in the debug prints
Based on a patch by Jonas Paulsson
llvm-svn: 228068
Jason Molenda [Tue, 3 Feb 2015 23:39:47 +0000 (23:39 +0000)]
Small fix for the "memory write -i filename" command -- if the user fails to specify the
number of bytes to write into the inferior process, the "default byte size" will be 1.
In that case, we want to copy the entire file into memory. The code was looking for
a default byte size of 0 to indicate that the user had not provided a specific # of
bytes to copy; adjust that to 1 to match the actual default value.
<rdar://problem/
18074973>
llvm-svn: 228067
Larisse Voufo [Tue, 3 Feb 2015 23:31:50 +0000 (23:31 +0000)]
PR22419: Give implicit sized deallocation functions default visibility
llvm-svn: 228066
David Majnemer [Tue, 3 Feb 2015 23:21:21 +0000 (23:21 +0000)]
Unbreak hexagon bot
llvm-svn: 228065
Philip Reames [Tue, 3 Feb 2015 23:18:47 +0000 (23:18 +0000)]
Use ImmutableCallSite for statepoint verification.
Patch by: Igor Laevsky
"This change generalizes statepoint verification to use ImmutableCallSite instead of CallInst. This will allow to easily implement invoke statepoint verification (in a following change)."
Differential Revision: http://reviews.llvm.org/D7308
llvm-svn: 228064
Nico Weber [Tue, 3 Feb 2015 23:10:18 +0000 (23:10 +0000)]
Wrap to 80 columns. No behavior change.
llvm-svn: 228063
David Majnemer [Tue, 3 Feb 2015 23:04:06 +0000 (23:04 +0000)]
CodeGen: Copy-ctorm must obey the destination's alignment requirement
We would synthesize memcpy intrinsics when emitting calls to trivial C++
constructors but we wouldn't take into account the alignment of the
destination.
llvm-svn: 228061
Rui Ueyama [Tue, 3 Feb 2015 23:00:19 +0000 (23:00 +0000)]
ELF: Support INPUT linker script directive
INPUT directive is a variant of GROUP in the sense that that specifies
a list of input files. The only difference is whether the entire file
list is wrapped with a --start-group/--end-group or not.
http://reviews.llvm.org/D7390
llvm-svn: 228060
Reid Kleckner [Tue, 3 Feb 2015 22:52:35 +0000 (22:52 +0000)]
SEH: Track users of __try so we can pick a per-func EH personality
There are four major kinds of declarations that cause code generation:
- FunctionDecl (includes CXXMethodDecl etc)
- ObjCMethodDecl
- BlockDecl
- CapturedDecl
This patch tracks __try usage on FunctionDecls and diagnoses __try usage
in other decls. If someone wants to use __try from ObjC, they can use it
from a free function, since the ObjC code will need an ObjC-style EH
personality.
Eventually we will want to look through CapturedDecls and track SEH
usage on the parent FunctionDecl, if present.
llvm-svn: 228058
Stephane Sezer [Tue, 3 Feb 2015 22:48:34 +0000 (22:48 +0000)]
Use basename of main executable in POSIX-DYLD on Android.
Summary:
The Android dynamic linker reports only the basename of each SO entry, so for
the above check to be successful, we need to compare it to the basename of the
main executable.
This also has a nasty side-effect when working with older version of
Android (verified on platform version 16), and debugging PIE
executables: the dynamic linker has a bug and reports the load address
of the main executable (which is a shared object, because PIE) to be 0.
We then try to update the list of loaded sections for all shared
objects, including the main executable, and set the load address to 0,
which breaks everything that relies on resolving addresses in the main
executable (breakpoints, stepping, etc). This commit also fixes that broken
behavior when debugging on older Androids. This bug doesn't happen on newer
Android versions (verified for Android L).
Test Plan: Run test suite on linux.
Reviewers: clayborg, tfiala, richard.mitton
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D7188
llvm-svn: 228057
Adam Nemet [Tue, 3 Feb 2015 22:45:39 +0000 (22:45 +0000)]
[LV] Split off memcheck block really at the first check
I've noticed this while trying to move addRuntimeCheck to LoopAccessAnalysis.
I think that the intention was to early exit from the overflow checking before
the code for the memchecks. This is the entire reason why we compute
FirstCheckInst but then we don't use that as the splitting instruction but the
final check. Looks like an oversight.
llvm-svn: 228056
Chandler Carruth [Tue, 3 Feb 2015 22:43:30 +0000 (22:43 +0000)]
[x86] Fix signed vs. unsigned comparison.
llvm-svn: 228055
Simon Pilgrim [Tue, 3 Feb 2015 22:39:28 +0000 (22:39 +0000)]
Fixed unused variable warning.
llvm-svn: 228054
David Blaikie [Tue, 3 Feb 2015 22:37:17 +0000 (22:37 +0000)]
DebugInfo: Ensure calls to functions with default arguments which themselves have default arguments, still have locations.
To handle default arguments in C++ in the debug info, we disable code
updating the debug location during the emission of default arguments.
This code was buggy in the case of default arguments which, themselves,
have default arguments - the inner default argument would re-enable
debug info when it was finished, but before the outer default argument
was finished.
This was already a bug, but got worse (because a crasher instead of just
a quality bug) with the recent improvements to debug info line quality
because... The ApplyDebugLocation scoped device would find the debug
info disabled and not save any debug location. But then in
~ApplyDebugLocation it would find the debug info had been enabled and
would then apply the no-location. Then the outer function call would be
emitted without any location. That's bad.
Arguably we could /also/ fix the ApplyDebugLocation to assert on this
situation (where debug info was disabled in the ctor and enabled in the
dtor, or the other way around) but this is at least the necessary fix
regardless.
(also, I imagine this disabling behavior might need to be in-place for
CGExprComplex and CGExprAgg too, maybe... ?)
And I seem to recall seeing some weird default arg stepping behavior
recently which might be related to this too... I'll have to look into
it.
llvm-svn: 228053
Weiming Zhao [Tue, 3 Feb 2015 22:35:58 +0000 (22:35 +0000)]
Diagnose CXX 'this' pointer reference in funcs with naked attr
Clang asserts for this pointer reference in asms of naked functions.
This patch diagnoses if this pointer reference is used.
Differential Revision: http://reviews.llvm.org/D7329
llvm-svn: 228052
DeLesley Hutchins [Tue, 3 Feb 2015 22:11:04 +0000 (22:11 +0000)]
Thread Safety Analysis: add support for before/after annotations on mutexes.
These checks detect potential deadlocks caused by inconsistent lock
ordering. The checks are implemented under the -Wthread-safety-beta flag.
This patch also replaces calls to getAttrs() with calls to attrs() throughout
ThreadSafety.cpp, which fixes the earlier issue that cause assert failures.
llvm-svn: 228051
Colin LeMahieu [Tue, 3 Feb 2015 22:09:51 +0000 (22:09 +0000)]
[Hexagon] Marking a bunch of non-encoded instructions with isCodeGenOnly = 1.
llvm-svn: 228050
Hans Wennborg [Tue, 3 Feb 2015 22:08:20 +0000 (22:08 +0000)]
[CMake] add_llvm_library: don't use .imp suffix for import libraries on Windows (PR22334)
This was added in r188351 to fix a naming conflict between the
profile_rt-static and profile_rt-shared who both ended up in
lib/profile_rt.lib.
The change also affected other libraries (like libclang), and
users are reporting that they find it surprising that there's
no longer a libclang.lib. Since the profile_rt naming conflict
doesn't seem to exist any more, I think we can remove this.
Differential Revision: http://reviews.llvm.org/D7391
llvm-svn: 228049
Arnaud A. de Grandmaison [Tue, 3 Feb 2015 22:02:45 +0000 (22:02 +0000)]
[PBQP] Constify Graph::getEdgeNode1Id and Graph::getEdgeNode2Id
llvm-svn: 228048
Simon Pilgrim [Tue, 3 Feb 2015 21:58:29 +0000 (21:58 +0000)]
[X86][SSE] psrl(w/d/q) and psll(w/d/q) bit shifts for SSE2
Patch to match cases where shuffle masks can be reduced to bit shifts. Similar to byte shift shuffle matching from D5699.
Differential Revision: http://reviews.llvm.org/D6649
llvm-svn: 228047
Bill Schmidt [Tue, 3 Feb 2015 21:58:23 +0000 (21:58 +0000)]
[PowerPC] Implement the vpopcnt instructions for POWER8
Patch by Kit Barton.
Add the vector population count instructions for byte, halfword, word,
and doubleword sizes. There are two major changes here:
PPCISelLowering.cpp: Make CTPOP legal for vector types.
PPCRegisterInfo.td: Added v2i64 to the VRRC register
definition. This is needed for the doubleword variations of the
integer ops that were added in P8.
Test Plan
Test the instruction vpcnt* encoding/decoding in ppc64-encoding-vmx.s
Test the generation of the vpopcnt instructions for various vector
data types. When adding the v2i64 type to the Vector Register set, I
also needed to add the appropriate bit conversion patterns between
v2i64 and the existing vector types. Testing for these conversions
were also added in the test case by passing a different vector type as
a parameter into the test functions. There is also a run step that
will ensure the vpopcnt instructions are generated when the vsx
feature is disabled.
llvm-svn: 228046
Kostya Serebryany [Tue, 3 Feb 2015 21:57:32 +0000 (21:57 +0000)]
[fuzzer] Add proper dependensices to the fuzzer tests
Summary: Make sure that FileCheck is built when running check-fuzzer
Test Plan:
run on bot:
lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer
Reviewers: samsonov
Reviewed By: samsonov
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D7387
llvm-svn: 228045
Chandler Carruth [Tue, 3 Feb 2015 21:56:28 +0000 (21:56 +0000)]
[x86] Add two truly horrific test cases for the new vector shuffle
lowering. I'm prepping patches to improve these, and this will let the
delta of those patches show the improvement. =]
llvm-svn: 228044
Chandler Carruth [Tue, 3 Feb 2015 21:56:24 +0000 (21:56 +0000)]
[x86] Update the indent and layout of some tests in this file. NFC
This is just to remove voise from using the update_llc_test_checks
script.
llvm-svn: 228043
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:56:01 +0000 (21:56 +0000)]
AsmParser: Recognize DW_TAG_* constants
Recognize `DW_TAG_` constants in assembly, and output it by default for
`GenericDebugNode`.
llvm-svn: 228042
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:54:14 +0000 (21:54 +0000)]
IR: Assembly and bitcode for GenericDebugNode
llvm-svn: 228041
Marek Olsak [Tue, 3 Feb 2015 21:53:27 +0000 (21:53 +0000)]
R600/SI: Remove the -CHECK suffix from all FileCheck prefixes in LIT tests
llvm-svn: 228040
Marek Olsak [Tue, 3 Feb 2015 21:53:08 +0000 (21:53 +0000)]
R600/SI: Remove useless patterns in VALU which are already covered by SALU
Also remove hasPostISelHook=1 from V_LSHL_B32. It's defined by InstSI already.
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 228039
Marek Olsak [Tue, 3 Feb 2015 21:53:05 +0000 (21:53 +0000)]
R600/SI: Rewrite VOP1InstSI to contain a pseudo and _si opcode
What this does is that if you accidentally select these instructions on VI,
the code generation will fail, because the pseudo -> _vi mapping will be
undefined.
The idea is to be able to catch possible future bugs easily.
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 228038
Marek Olsak [Tue, 3 Feb 2015 21:53:01 +0000 (21:53 +0000)]
R600/SI: Fix B64 VALU shifts on VI
SI only has standard versions. VI only has REV versions.
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 228037
Reid Kleckner [Tue, 3 Feb 2015 21:49:15 +0000 (21:49 +0000)]
Fix Driver/mg.c test when fgrep is a shell Cygwin shell script
llvm-svn: 228036
Justin Bogner [Tue, 3 Feb 2015 21:35:49 +0000 (21:35 +0000)]
InstrProf: Remove CoverageMapping::HasCodeBefore, it isn't used
llvm-svn: 228035
Justin Bogner [Tue, 3 Feb 2015 21:35:36 +0000 (21:35 +0000)]
InstrProf: Remove CoverageMapping::HasCodeBefore, it isn't used
It's not entirely clear to me what this field was meant for, but it's
always false. Remove it.
llvm-svn: 228034
Chandler Carruth [Tue, 3 Feb 2015 21:26:45 +0000 (21:26 +0000)]
[x86] Tweak my update script to use test case function names starting
with 'stress' to indicate that the specific output isn't interesting and
relax them to only check the last instruction (a ret).
I've updated the one test case that really uses this to name the one
'stress_test' which was actually producing output we can directly check.
With this, the script doesn't introduce noise when run over the v16 test
file.
llvm-svn: 228033
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:25:34 +0000 (21:25 +0000)]
CodeGen: Update for LLVM API change in r228030
The mock tags are no longer in `dwarf::LLVMConstants`; they're in
`dwarf::Tag`.
llvm-svn: 228032
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:16:49 +0000 (21:16 +0000)]
Support: Add string => unsigned mapping for DW_TAG
Add `dwarf::getTag()` to translate from `StringRef` to `unsigned`.
llvm-svn: 228031
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:13:16 +0000 (21:13 +0000)]
Support: Re-implement dwarf::TagString() using a .def file, NFC
Also re-implements the `dwarf::Tag` enumerator. I've moved the mock
tags into the enumerator since there's no other way to do this. Really
they shouldn't be used at all (they're just a hack to identify
`MDNode`s, but we have a class hierarchy for that now).
llvm-svn: 228030
Duncan P. N. Exon Smith [Tue, 3 Feb 2015 21:08:33 +0000 (21:08 +0000)]
Support: Stop stringifying DW_TAG_{lo,hi}_user
`dwarf::TagString()` shouldn't stringify `DW_TAG_lo_user` or
`DW_TAG_hi_user`. These aren't actual tags; they're markers for the
edge of vendor-specific tag regions.
llvm-svn: 228029
Eric Fiselier [Tue, 3 Feb 2015 21:00:15 +0000 (21:00 +0000)]
Fix alignment in tests for readability.
llvm-svn: 228028
Simon Pilgrim [Tue, 3 Feb 2015 20:54:01 +0000 (20:54 +0000)]
Fixed signed/unsigned comparison warning.
llvm-svn: 228027
Colin LeMahieu [Tue, 3 Feb 2015 20:40:52 +0000 (20:40 +0000)]
[Hexagon] Converting XTYPE/SHIFT intrinsics. Cleaning out old intrinsic patterns and updating tests.
llvm-svn: 228026
Simon Pilgrim [Tue, 3 Feb 2015 20:38:52 +0000 (20:38 +0000)]
Fixed unused variable warning.
llvm-svn: 228025
Daniel Berlin [Tue, 3 Feb 2015 20:37:08 +0000 (20:37 +0000)]
Allow PRE to insert no-cost phi nodes
llvm-svn: 228024
Renato Golin [Tue, 3 Feb 2015 20:23:00 +0000 (20:23 +0000)]
Revert "Enabling testing ASAN on AArch64"
This reverts commit r227966, which turned ASAN on on AArhc64 and may be the
cause of the bots never finishing the check-all. I'll re-apply once we're
sure that bot can cope with it.
llvm-svn: 228023
Simon Pilgrim [Tue, 3 Feb 2015 20:09:18 +0000 (20:09 +0000)]
[X86][SSE] Added general integer shuffle matching for MOVQ instruction
This patch adds general shuffle pattern matching for the MOVQ zero-extend instruction (copy lower 64bits, zero upper) for all 128-bit integer vectors, it is added as a fallback test in lowerVectorShuffleAsZeroOrAnyExtend.
llvm-svn: 228022
Adrian Prantl [Tue, 3 Feb 2015 20:00:54 +0000 (20:00 +0000)]
Address review feedback for r228003.
- use named constructors
- get rid of MarkAsPrologue
llvm-svn: 228021
Reid Kleckner [Tue, 3 Feb 2015 19:51:16 +0000 (19:51 +0000)]
Revert "Thread Safety Analysis: add support for before/after annotations on mutexes."
This reverts r227997, as well as r228009. It does not pass check-clang
for me locally on Linux.
llvm-svn: 228020
Colin LeMahieu [Tue, 3 Feb 2015 19:43:59 +0000 (19:43 +0000)]
[Hexagon] Updating XTYPE/PRED intrinsics.
llvm-svn: 228019
Kostya Serebryany [Tue, 3 Feb 2015 19:42:05 +0000 (19:42 +0000)]
[fuzzer] update the include line to use the new header name
llvm-svn: 228018
Kostya Serebryany [Tue, 3 Feb 2015 19:40:53 +0000 (19:40 +0000)]
[sanitizer] move the coverage interface into a separate header, <sanitizer/coverage_interface.h>. NFC, except for the header name change. This may break existing users, but in this case it's better this way (not too many users so far)
llvm-svn: 228017
Jingyue Wu [Tue, 3 Feb 2015 19:37:06 +0000 (19:37 +0000)]
Add straight-line strength reduction to LLVM
Summary:
Straight-line strength reduction (SLSR) is implemented in GCC but not yet in
LLVM. It has proven to effectively simplify statements derived from an unrolled
loop, and can potentially benefit many other cases too. For example,
LLVM unrolls
#pragma unroll
foo (int i = 0; i < 3; ++i) {
sum += foo((b + i) * s);
}
into
sum += foo(b * s);
sum += foo((b + 1) * s);
sum += foo((b + 2) * s);
However, no optimizations yet reduce the internal redundancy of the three
expressions:
b * s
(b + 1) * s
(b + 2) * s
With SLSR, LLVM can optimize these three expressions into:
t1 = b * s
t2 = t1 + s
t3 = t2 + s
This commit is only an initial step towards implementing a series of such
optimizations. I will implement more (see TODO in the file commentary) in the
near future. This optimization is enabled for the NVPTX backend for now.
However, I am more than happy to push it to the standard optimization pipeline
after more thorough performance tests.
Test Plan: test/StraightLineStrengthReduce/slsr.ll
Reviewers: eliben, HaoLiu, meheff, hfinkel, jholewinski, atrick
Reviewed By: jholewinski, atrick
Subscribers: karthikthecool, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D7310
llvm-svn: 228016
Colin LeMahieu [Tue, 3 Feb 2015 19:36:59 +0000 (19:36 +0000)]
[Hexagon] Updating XTYPE/PERM intrinsics.
llvm-svn: 228015
Simon Pilgrim [Tue, 3 Feb 2015 19:34:09 +0000 (19:34 +0000)]
[X86][AVX2] Enabled shuffle matching for the AVX2 zero extension (128bit -> 256bit) vpmovzx* instructions.
Differential Revision: http://reviews.llvm.org/D7251
llvm-svn: 228014
Ben Langmuir [Tue, 3 Feb 2015 19:28:37 +0000 (19:28 +0000)]
Make the default module cache user-specific
Appends the username to the first component (after the temp dir) of the
module cache path. If the username contains a character that shouldn't
go into a path (for now conservatively allow [a-zA-Z0-9_]), we fallback
to the user id.
llvm-svn: 228013
Rafael Espindola [Tue, 3 Feb 2015 19:25:53 +0000 (19:25 +0000)]
Fix duplicated symbol error.
llvm-svn: 228012
Rafael Espindola [Tue, 3 Feb 2015 19:20:26 +0000 (19:20 +0000)]
Fix typo in test/CodeGen/X86/sibcall.ll (pr22331).
llvm-svn: 228011
Colin LeMahieu [Tue, 3 Feb 2015 19:15:11 +0000 (19:15 +0000)]
[Hexagon] Adding missing vector multiply instruction encodings. Converting multiply intrinsics and updating tests.
llvm-svn: 228010
Reid Kleckner [Tue, 3 Feb 2015 19:04:26 +0000 (19:04 +0000)]
thread safety: Add move ctor to BeforeInfo to fix MSVC build
MSVC cannot infer move ctors yet.
llvm-svn: 228009
Sanjay Patel [Tue, 3 Feb 2015 18:54:00 +0000 (18:54 +0000)]
Merge consecutive 16-byte loads into one 32-byte load (PR22329)
This patch detects consecutive vector loads using the existing
EltsFromConsecutiveLoads() logic. This fixes:
http://llvm.org/bugs/show_bug.cgi?id=22329
This patch effectively reverts the tablegen additions of D6492 /
http://reviews.llvm.org/rL224344 ...which in hindsight were a horrible hack.
The test cases that were added with that patch are simply modified to load
from varying offsets of a base pointer. These loads did not match the existing
tablegen patterns.
A happy side effect of doing this optimization earlier is that we can now fold
the load into a math op where possible; this is shown in some of the updated
checks in the test file.
Differential Revision: http://reviews.llvm.org/D7303
llvm-svn: 228006
Greg Fitzgerald [Tue, 3 Feb 2015 18:47:37 +0000 (18:47 +0000)]
Don't assume LIT_EXECUTABLE points to a Python script, take 2
Before this patch, the CMake build assumed LIT_EXECUTABLE pointed
to a Python script, not an executable. If you were to pass in an
executable, such as the result of py2exe on lit.py, the build would
fall over.
With this patch, the CMake build assumes LIT_EXECUTABLE is an
executable. You can continue setting it to lit.py, but it will
now use its shebang to find a Python interpreter.
Differential Revision: http://reviews.llvm.org/D7315
llvm-svn: 228005
Sanjay Patel [Tue, 3 Feb 2015 18:47:32 +0000 (18:47 +0000)]
remove variable names from comments; NFC
I didn't bother to fix the self-referential definitions and grammar
because my eyes started to bleed.
llvm-svn: 228004
Adrian Prantl [Tue, 3 Feb 2015 18:40:42 +0000 (18:40 +0000)]
Merge ArtificialLocation into ApplyDebugLocation and make a clear
distinction between the different use-cases. With the previous default
behavior we would occasionally emit empty debug locations in situations
where they actually were strictly required (= on invoke insns).
We now have a choice between defaulting to an empty location or an
artificial location.
Specifically, this fixes a bug caused by a missing debug location when
emitting C++ EH cleanup blocks from within an artificial function, such as
an ObjC destroy helper function.
rdar://problem/
19670595
llvm-svn: 228003
Adrian Prantl [Tue, 3 Feb 2015 18:40:38 +0000 (18:40 +0000)]
Add documentation to ApplyDebugLocation.
llvm-svn: 228002
Alexey Samsonov [Tue, 3 Feb 2015 18:40:34 +0000 (18:40 +0000)]
[ASan] Run tests with both static and dynamic runtime on Windows by default.
llvm-svn: 228001
Manman Ren [Tue, 3 Feb 2015 18:39:15 +0000 (18:39 +0000)]
[LTO API] split lto_codegen_compile to lto_codegen_optimize and
lto_codegen_compile_optimized. Also add lto_api_version.
Before this commit, we can only dump the optimized bitcode after running
lto_codegen_compile, but it includes some impacts of running codegen passes,
one example is StackProtector pass. We will get assertion failure when running
llc on the optimized bitcode, because StackProtector is effectively run twice.
After splitting lto_codegen_compile, the linker can choose to dump the bitcode
before running lto_codegen_compile_optimized.
lto_api_version is added so ld64 can check for runtime-availability of the new
API.
rdar://
19565500
llvm-svn: 228000
Hans Wennborg [Tue, 3 Feb 2015 18:31:29 +0000 (18:31 +0000)]
Fix ProgramFiles path for 64-bit Windows installer
If we are building an 64bit installer on Windows we have to adjust the
Program Files path otherwise it uses the wrong Program Files (x86)
directory. Related CMake bug report
http://public.kitware.com/Bug/view.php?id=14211
Patch by Ismail Dönmez!
llvm-svn: 227999
Zachary Turner [Tue, 3 Feb 2015 18:26:00 +0000 (18:26 +0000)]
Fix compilation failure on Windows.
llvm-svn: 227998
DeLesley Hutchins [Tue, 3 Feb 2015 18:17:48 +0000 (18:17 +0000)]
Thread Safety Analysis: add support for before/after annotations on mutexes.
These checks detect potential deadlocks caused by inconsistent lock
ordering. The checks are implemented under the -Wthread-safety-beta flag.
llvm-svn: 227997
Greg Fitzgerald [Tue, 3 Feb 2015 18:16:47 +0000 (18:16 +0000)]
Revert "Don't assume LIT_EXECUTABLE points to a Python script"
This reverts r227994
llvm-svn: 227996
Colin LeMahieu [Tue, 3 Feb 2015 18:16:28 +0000 (18:16 +0000)]
[Hexagon] Converting complex number intrinsics and adding tests.
llvm-svn: 227995
Greg Fitzgerald [Tue, 3 Feb 2015 18:02:04 +0000 (18:02 +0000)]
Don't assume LIT_EXECUTABLE points to a Python script
Before this patch, the CMake build assumed LIT_EXECUTABLE pointed
to a Python script, not an executable. If you were to pass in an
executable, such as the result of py2exe on lit.py, the build would
fall over.
With this patch, the CMake build assumes LIT_EXECUTABLE is an
executable. You can continue setting it to lit.py, but it will
now use its shebang to find a Python interpreter.
Differential Revision: http://reviews.llvm.org/D7315
llvm-svn: 227994
Colin LeMahieu [Tue, 3 Feb 2015 18:01:45 +0000 (18:01 +0000)]
[Hexagon] Adding vector intrinsics for alu32/alu and xtype/alu.
llvm-svn: 227993
Adam Nemet [Tue, 3 Feb 2015 17:59:53 +0000 (17:59 +0000)]
[LoopVectorize] Fix rebase glitch in r227751
LoopVectorizationLegality::{getNumLoads,getNumStores} should forward to
LoopAccessAnalysis now.
Thanks to Takumi for noticing this!
llvm-svn: 227992
Jingyue Wu [Tue, 3 Feb 2015 17:57:38 +0000 (17:57 +0000)]
Remove usernames from TODOs, NFC
making the style consistent with the rest
llvm-svn: 227991
Marek Olsak [Tue, 3 Feb 2015 17:38:12 +0000 (17:38 +0000)]
R600/SI: Don't generate non-existent LSHL, LSHR, ASHR B32 variants on VI
This can happen when a REV instruction is commuted.
The trick is not to define the _vi versions of instructions, which has these
consequences:
- code generation will always fail if a pseudo cannot be lowered
(very useful to catch bugs where an unsupported instruction somehow makes
it to the printer)
- ability to query if a pseudo can be lowered, which is done in commuteOpcode
to prevent REV from commuting to non-REV on VI
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227990
Marek Olsak [Tue, 3 Feb 2015 17:38:05 +0000 (17:38 +0000)]
R600/SI: Remove VOP2_REV definitions from target-specific instructions
The getCommute* functions are only used with pseudos, so this commit doesn't
change anything.
The issue with missing non-rev versions of shift instructions on VI will fixed
separately.
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227989
Marek Olsak [Tue, 3 Feb 2015 17:38:01 +0000 (17:38 +0000)]
R600/SI: Trivial instruction definition corrections for VI (v2)
- V_MAC_LEGACY_F32 exists on VI, but it's VOP3-only.
- Define CVT_PK opcodes which are different between SI and VI. These are
unused. The idea is to define all chip differences.
v2: keep V_MUL_LO_U32
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227988
Marek Olsak [Tue, 3 Feb 2015 17:37:57 +0000 (17:37 +0000)]
R600/SI: Determine target-specific encoding of READLANE and WRITELANE early v2
These are VOP2 on SI and VOP3 on VI, and their pseudos are neither, which can
be a problem. In order to make isVOP2 and isVOP3 queries behave as expected,
the encoding must be determined first.
This doesn't fix any known issue, but better safe than sorry.
v2: add and use getMCOpcodeFromPseudo
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227987
Marek Olsak [Tue, 3 Feb 2015 17:37:52 +0000 (17:37 +0000)]
R600/SI: Fix dependency between instruction writing M0 and S_SENDMSG on VI (v2)
This fixes a hang when using an empty geometry shader.
v2: - don't add s_nop when followed by s_waitcnt
- comestic changes
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227986
Sanjay Patel [Tue, 3 Feb 2015 17:13:04 +0000 (17:13 +0000)]
Fix program crashes due to alignment exceptions generated for SSE memop instructions (PR22371).
r224330 introduced a bug by misinterpreting the "FeatureVectorUAMem" bit.
The commit log says that change did not affect anything, but that's not correct.
That change allowed SSE instructions to have unaligned mem operands folded into
math ops, and that's not allowed in the default specification for any SSE variant.
The bug is exposed when compiling for an AVX-capable CPU that had this feature
flag but without enabling AVX codegen. Another mistake in r224330 was not adding
the feature flag to all AVX CPUs; the AMD chips were excluded.
This is part of the fix for PR22371 ( http://llvm.org/bugs/show_bug.cgi?id=22371 ).
This feature bit is SSE-specific, so I've renamed it to "FeatureSSEUnalignedMem".
Changed the existing test case for the feature bit to reflect the new name and
renamed the test file itself to better reflect the feature.
Added runs to fold-vex.ll to check for the failing codegen.
Note that the feature bit is not set by default on any CPU because it may require a
configuration register setting to enable the enhanced unaligned behavior.
llvm-svn: 227983
Nico Weber [Tue, 3 Feb 2015 17:06:08 +0000 (17:06 +0000)]
Implement jump scope SEHmantic analysis.
Thou shall not jump into SEH blocks. Jumping out of SEH __try and __excepts
is A-ok. Jumping out of __finally blocks is B-ok (msvc doesn't error about it,
but warns that it has undefined behavior).
I've checked that clang's behavior with this patch matches msvc's behavior.
We don't have the warning on jumping out of a __finally yet, see the FIXME
in the test. clang also currently crashes on codegen for a jump out of a
__finally block, see PR22414 comment 7.
I also added a few tests for the interaction of indirect jumps and SEH blocks.
MSVC doesn't support indirect jumps, so there's no way to know if clang behave
the same way as msvc here. clang's behavior with this patch does make sense
to me, but maybe it could be argued that it should be more permissive (see
FIXME in the indirect jump tests -- shout if you have an opinion on this).
llvm-svn: 227982
Bill Schmidt [Tue, 3 Feb 2015 16:57:38 +0000 (16:57 +0000)]
Disable 32-bit tests in tls-pic.ll until they can be repaired
llvm-svn: 227981
Bill Schmidt [Tue, 3 Feb 2015 16:33:55 +0000 (16:33 +0000)]
Further revise too-restrictive test CodeGen/PowerPC/tls-pic.ll
llvm-svn: 227980
Rafael Espindola [Tue, 3 Feb 2015 16:33:53 +0000 (16:33 +0000)]
Use CLANG_LIBDIR_SUFFIX when looking for the gold plugin.
Patch by İsmail Dönmez!
llvm-svn: 227979
Bill Schmidt [Tue, 3 Feb 2015 16:29:52 +0000 (16:29 +0000)]
Further revise too-restrictive test CodeGen/PowerPC/tls-pic.ll
llvm-svn: 227978
Bill Schmidt [Tue, 3 Feb 2015 16:24:05 +0000 (16:24 +0000)]
Revise too-restrictive test CodeGen/PowerPC/tls-pic.ll
llvm-svn: 227977
Bill Schmidt [Tue, 3 Feb 2015 16:16:01 +0000 (16:16 +0000)]
[PowerPC] Yet another approach to __tls_get_addr
This patch is a third attempt to properly handle the local-dynamic and
global-dynamic TLS models.
In my original implementation, calls to __tls_get_addr were hidden
from view until the asm-printer phase, at which point the underlying
branch-and-link instruction was created with proper relocations. This
mostly worked well, but I used some repellent techniques to ensure
that the TLS_GET_ADDR nodes at the SD and MI levels correctly received
input from GPR3 and produced output into GPR3. This proved to work
badly in the presence of multiple TLS variable accesses, with the
copies to and from GPR3 being scheduled incorrectly and generally
creating havoc.
In r221703, I addressed that problem by representing the calls to
__tls_get_addr as true calls during instruction lowering. This had
the advantage of removing all of the bad hacks and relying on the
existing call machinery to properly glue the copies in place. It
looked like this was going to be the right way to go.
However, as a side effect of the recent discovery of problems with
linker optimizations for TLS, we discovered cases of suboptimal code
generation with this strategy. The problem comes when tls_get_addr is
called for the same address, and there is a resulting CSE
opportunity. It turns out that in such cases MachineCSE will common
the addis/addi instructions that set up the input value to
tls_get_addr, but will not common the calls themselves. MachineCSE
does not have any machinery to common idempotent calls. This is
perfectly sensible, since presumably this would be done at the IR
level, and introducing calls in the back end isn't commonplace. In
any case, we end up with two calls to __tls_get_addr when one would
suffice, and that isn't good.
I presumed that the original design would have allowed commoning of
the machine-specific nodes that hid the __tls_get_addr calls, so as
suggested by Ulrich Weigand, I went back to that design and cleaned it
up so that the copies were properly held together by glue
nodes. However, it turned out that this didn't work either...the
presence of copies to physical registers kept the machine-specific
nodes from being commoned also.
All of which leads to the design presented here. This is a return to
the original design, except that no attempt is made to introduce
copies to and from GPR3 during instruction lowering. Virtual registers
are used until prior to register allocation. At that point, a special
pass is run that identifies the machine-specific nodes that hide the
tls_get_addr calls and introduces the copies to and from GPR3 around
them. The register allocator then coalesces these copies away. With
this design, MachineCSE succeeds in commoning tls_get_addr calls where
possible, and we get nice optimal code generation (better than GCC at
the moment, which does not common these calls).
One additional problem must be dealt with: After introducing the
mentions of the physical register GPR3, the aggressive anti-dependence
breaker sees opportunities to improve scheduling by selecting a
different register instead. Flags must be used on the instruction
descriptions to tell the anti-dependence breaker to keep its hands in
its pockets.
One thing missing from the original design was recording a definition
of the link register on the GET_TLS_ADDR nodes. Doing this was found
to be insufficient to force a stack frame to be created, which led to
looping behavior because two different LR values were stored at the
same address. This appears to have been an oversight in
PPCFrameLowering::determineFrameLayout(), which is repaired here.
Because MustSaveLR() returns true for calls to builtin_return_address,
this changed the expected behavior of
test/CodeGen/PowerPC/retaddr2.ll, which now stacks a frame but
formerly did not. I've fixed the test case to reflect this.
There are existing TLS tests to catch regressions; the checks in
test/CodeGen/PowerPC/tls-store2.ll proved to be too restrictive in the
face of instruction scheduling with these changes, so I fixed that
up.
I've added a new test case based on the PrettyStackTrace module that
demonstrated the original problem. This checks that we get correct
code generation and that CSE of the calls to __get_tls_addr has taken
place.
llvm-svn: 227976
Rafael Auler [Tue, 3 Feb 2015 16:08:57 +0000 (16:08 +0000)]
Make ELFLinkingContext own LinkerScript buffers
Currently, no one owns script::Parser buffers, but yet ELFLinkingContext gets
updated with StringRef pointers to data inside Parser buffers. Since this buffer
is locally owned inside GnuLdDriver::evalLinkerScript(), as soon as this
function finishes, all pointers in ELFLinkingContext that comes from linker
scripts get invalid. The problem is that we need someone to own linker scripts
data structures and, since ELFLinkingContext transports references to linker
scripts data, we can simply make it also own all linker scripts data.
Differential Revision: http://reviews.llvm.org/D7323
llvm-svn: 227975
Eric Fiselier [Tue, 3 Feb 2015 16:04:45 +0000 (16:04 +0000)]
Mark <experimental/system_error> as complete
llvm-svn: 227974
Eric Fiselier [Tue, 3 Feb 2015 16:03:24 +0000 (16:03 +0000)]
[libcxx] Add <experimental/system_error>
Summary:
This patch just adds the variable templates in <experimental/system_error>.
see: https://rawgit.com/cplusplus/fundamentals-ts/v1/fundamentals-ts.html#syserror
Reviewers: jroelofs, danalbert, K-ballo, mclow.lists
Reviewed By: mclow.lists
Subscribers: chandlerc, cfe-commits
Differential Revision: http://reviews.llvm.org/D7353
llvm-svn: 227973
Sanjay Patel [Tue, 3 Feb 2015 15:37:18 +0000 (15:37 +0000)]
Improve test to actually check for a folded load.
This test was checking for lack of a "movaps" (an aligned load)
rather than a "movups" (an unaligned load). It also included
a store which complicated the checking.
Add specific CPU runs to prevent subtarget feature flag overrides
from inhibiting this optimization.
llvm-svn: 227972