Commit Graph

245680 Commits

Author SHA1 Message Date
Samuel Antao 5b1a89a33b Fix bug when compiling CUDA code with -emit-llvm and -o.
In this case the device code is not injected into an host action and therefore the 
user should get an error as -o can't be used when generating two outputs.

llvm-svn: 285263
2016-10-27 00:53:34 +00:00
Kostya Serebryany 94c427c23e [libFuzzer] speculatively trying to fix the Mac build; second attempt
llvm-svn: 285262
2016-10-27 00:36:38 +00:00
NAKAMURA Takumi 3f286ad1fe xray-extract.cc: Quick fix for mingw, to avoid errc::protocol_error.
errc::protocol_error is winsock-oriented and mingw doesn't include it. (MS does, though)

llvm-svn: 285261
2016-10-27 00:34:24 +00:00
Kostya Serebryany 3d945f6247 [libFuzzer] revert 285259 -- hit commit too soon
llvm-svn: 285260
2016-10-27 00:24:34 +00:00
Kostya Serebryany 15cd6b4b10 [libFuzzer] speculatively trying to fix the Mac build
llvm-svn: 285259
2016-10-27 00:22:39 +00:00
John McCall b92ab1afd5 Refactor call emission to package the function pointer together with
abstract information about the callee.  NFC.

The goal here is to make it easier to recognize indirect calls and
trigger additional logic in certain cases.  That logic will come in
a later patch; in the meantime, I felt that this was a significant
improvement to the code.

llvm-svn: 285258
2016-10-26 23:46:34 +00:00
Davide Italiano 48ef6ca0c3 [IR] Retire unused getGEPReturnType overload. NFCI.
llvm-svn: 285257
2016-10-26 23:46:16 +00:00
Tim Shen 7527752353 [APFloat] Fix APFloat::getExactInverse when the input is nullptr. This is a regression introduced by r285105.
Reviewers: kbarton, echristo, iteratee, eugenis

Subscribers: llvm-commits, mehdi_amini

Differential Revision: https://reviews.llvm.org/D26017

llvm-svn: 285256
2016-10-26 23:31:41 +00:00
Anna Zaks be7ae6684f [sanitizers] Set Darwin specific linker and compiler flags for all tests
Looks like we are missing these flags only in tsan and sanitizer-common.

This results in linker warnings in some settings as it can cause the Unit
tests to be built with a different SDK version than that was used to build
the runtime. For example, we are not setting the minimal deployment target
on the tests but are setting the minimal deployment target for the sanitizer
library, which leads to the following warning on some bots: ld: warning:
object file (sanitizer_posix_test.cc.i386.o) was built for newer OSX version
(10.12) than being linked (10.11).

Differential Revision: https://reviews.llvm.org/D25860  https://reviews.llvm.org/D25352

llvm-svn: 285255
2016-10-26 23:23:38 +00:00
Mehdi Amini 5ee58d3860 Unconditionally pass `-lto_library` to the linker on Darwin
We're only doing it with -flto currently, however it never "hurt"
to pass it, and users that are linking without -flto can get in
trouble if one of the dependency (a static library for instance)
contains bitcode.

Differential Revision: https://reviews.llvm.org/D25932

llvm-svn: 285254
2016-10-26 23:23:08 +00:00
Anna Zaks 5b2b39065c [analyzer] Report CFNumberGetValue API misuse
This patch contains 2 improvements to the CFNumber checker:
 - Checking of CFNumberGetValue misuse.
 - Treating all CFNumber API misuse errors as non-fatal. (Previously we treated errors that could cause uninitialized memory as syncs and the truncation errors as non-fatal.)

This implements a subset of functionality from https://reviews.llvm.org/D17954.

Differential Revision: https://reviews.llvm.org/D25876

llvm-svn: 285253
2016-10-26 22:51:47 +00:00
Justin Bogner 31d8b7d21d llvm-objdump: Make some error messages more consistent
Most of the version of report_error were quoting the filename and
printing a colon between the file name and the error message, but this
one wasn't doing either of those. Fix the output to be more
consistent.

llvm-svn: 285252
2016-10-26 22:37:52 +00:00
Justin Lebar ebeeab87a1 [CUDA] Move device placement new definitions into a wrapper header.
Previously, these were always included -- after this change, you have to
 #include <new>, which is consistent with how things ought to work.

llvm-svn: 285251
2016-10-26 22:13:26 +00:00
Justin Lebar 6f5ec7ee88 [CUDA] Switch cuda_wrappers/complex to use a proper include guard instead of #pragma once.
This is consistent with the rest of our internal headers.

llvm-svn: 285250
2016-10-26 22:13:20 +00:00
Vedant Kumar f4df0edf3e [utils] Add a '--unified-report' option to the code coverage prep script
In --unified-report mode, a single coverage report is prepared for all
specified binaries and written to *report_dir*. This mode is compatible
with all existing script options, including the --restrict mode which is
used to limit coverage reporting to certain files or directories.

This should not break any existing users of the script.

llvm-svn: 285249
2016-10-26 22:07:39 +00:00
Vedant Kumar 7466987cb7 [utils] Use print_function in the code coverage prep script, NFC.
llvm-svn: 285248
2016-10-26 22:07:37 +00:00
Vedant Kumar 273b6dc3e2 [utils] Add an '--only-merge' option to the code coverage prep script
In --only-merge mode, the script terminates after the profile merging
step.  This makes the script less stateful: it's more natural to split
the merge out into a separate step instead of relying on the first
invocation of the script to do it.

This should not break any existing users of the script.

llvm-svn: 285247
2016-10-26 22:07:35 +00:00
Evandro Menezes ca8370396a [AArch64] Create feature set for Samsung Exynos-M2
Since Exynos-M2 improved the FP square root unit a bit over the one in
Exynos-M1, it does not benefit from using the Newton series for such
operations.

llvm-svn: 285246
2016-10-26 22:06:20 +00:00
Jonathan Peyton 3c4050d698 Fixing typos in __kmp_release_deps trace outputs
Patch by Victor Campos

Differential Revision: https://reviews.llvm.org/D25972

llvm-svn: 285244
2016-10-26 21:46:43 +00:00
Jonathan Peyton 762bc46224 Use getpagesize() instead of PAGE_SIZE macro when KMP_OS_LINUX is true
Patch by Victor Campos

Differential Revision: https://reviews.llvm.org/D26001

llvm-svn: 285243
2016-10-26 21:42:48 +00:00
Victor Leschuk a37660c669 DebugInfo: fix incorrect alignment type (NFC)
Change type of some missed DebugInfo-related alignment variables,
that are still uint64_t, to uint32_t.

Original change introduced in r284482.

llvm-svn: 285242
2016-10-26 21:32:29 +00:00
Rafael Espindola 5ceeb60c6e Have on addRegular forward to another. NFC.
Another step in making it easier to create symbols without an ELf_Sym.

llvm-svn: 285241
2016-10-26 20:57:14 +00:00
Rui Ueyama 4c1cd45965 Remove a constructor to simplify.
llvm-svn: 285240
2016-10-26 20:44:32 +00:00
Malcolm Parsons 9bd85d2286 [RecursiveASTVisitor] Visit the implicit expression of a CXXDefaultArgExpr
Summary:
The matcher
varDecl(hasDescendant(
    callExpr(hasDeclaration(functionDecl(unless(isNoThrow()))))))
didn't match calls from default arguments because the expression
for a CXXDefaultArgExpr was not visited.

Reviewers: klimek, jdennett, alexfh, aaron.ballman

Subscribers: aaron.ballman, cfe-commits

Differential Revision: https://reviews.llvm.org/D25992

llvm-svn: 285239
2016-10-26 20:39:54 +00:00
Rafael Espindola a8bf71cc4e Make the DefinedRegular constructors more regular. NFC.
All the global ones now just forward to the same constructor. This
makes it easier to construct them without creating a Elf_Sym.

llvm-svn: 285238
2016-10-26 20:34:59 +00:00
Reid Kleckner 4500f74858 [lit] Work around Windows MSys command line tokenization bug
Summary:
This will allow us to revert LLD r284768, which added spaces to get MSys
echo to print what we want.

Reviewers: ruiu, inglorion, rafael

Subscribers: modocache, llvm-commits

Differential Revision: https://reviews.llvm.org/D26009

llvm-svn: 285237
2016-10-26 20:29:27 +00:00
Rui Ueyama e784ee7348 Fix build breakage.
llvm-svn: 285236
2016-10-26 20:27:38 +00:00
Rui Ueyama 5f7e6e6cfb Add comments.
llvm-svn: 285235
2016-10-26 20:26:29 +00:00
Casey Carter be65b24a55 Cleanup nonportable behavior in tests for std::any
Fixes MS issues 63, 64, and 65.

test/std/utilities/any/any.class/any.cons/move.pass.cpp:
* "Moves are always destructive" is not a portable assumption; check with LIBCPP_ASSERT.

test/std/utilities/any/any.class/any.cons/value.pass.cpp:
* The standard does not forbid initializing std::any from any pointer-to-function type. Remove the non-conforming "DecayTag" test.

test/std/utilities/any/any.class/any.modifiers/swap.pass.cpp:
* Self-swap is not specified to perform no moves; check with LIBCPP_ASSERT.

Differential Revision: https://reviews.llvm.org/D26007

llvm-svn: 285234
2016-10-26 20:18:13 +00:00
Ehsan Amiri 00b6e3bcc5 [PPC] Remove testcase from incorrect directory
During my last commit this testcase was put in an incorrect directory. Removing
it. Will put it in the right directory when I can verify everything is correct.

llvm-svn: 285233
2016-10-26 20:16:59 +00:00
Tim Northover a9cc385664 ARM: don't rely on push/pop reglists being in order when folding SP adjust.
It would be a very nice invariant to rely on, but unfortunately it doesn't
necessarily hold (and the causes of mis-sorted reglists appear to be quite
varied) so to be robust the frame lowering code can't assume that the first
register in the list is also the first one that actually gets pushed.

Should fix an issue where we were turning something like:

    push {r8, r4, r7, lr}
    sub sp, #24

into nonsense like:

    push {r2, r3, r4, r5, r6, r7, r8, r4, r7, lr}

llvm-svn: 285232
2016-10-26 20:01:00 +00:00
Nemanja Ivanovic 275853e777 Do not assume that FP vector operands are never legalized by expanding
This patch ensures that if a floating point vector operand is legalized by
expanding, it is legalized through the stack rather than by calling
DAGTypeLegalizer::IntegerToVector which will cause a failure since the operand
is a non-integer type.

This fixes PR 30715.

llvm-svn: 285231
2016-10-26 19:51:35 +00:00
Nemanja Ivanovic 3de0a385c9 [PowerPC] Implement vector_insert_exp builtins - clang portion
This patch corresponds to review https://reviews.llvm.org/D25956.
Committing on behalf of Zaara Syeda.

llvm-svn: 285229
2016-10-26 19:27:11 +00:00
Sanjoy Das 01969218a4 Simplify `x >=u x >> y` and `x >=u x udiv y`
Summary:
Extends InstSimplify to handle both `x >=u x >> y` and `x >=u x udiv y`.

This is a folloup of rL258422 and
https://github.com/rust-lang/rust/pull/30917 where llvm failed to
optimize away the bounds checking in a binary search.

Patch by Arthur Silva!

Reviewers: sanjoy

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D25941

llvm-svn: 285228
2016-10-26 19:18:43 +00:00
Chad Rosier 4447d7a816 Revert "[AliasSetTracker] Make AST smarter about intrinsics that don't actually affect memory."
This reverts commit r285191.

LICM appears to rely on the Alias Set Tracker hitting lifetime markers to prevent
code from being moved outside of the original scope.

llvm-svn: 285227
2016-10-26 19:18:19 +00:00
Enrico Granata 73418dfe61 Fix an issue where frame variable -s <varname> would not show the scope even though the user asked for it
Part of rdar://28434047

llvm-svn: 285226
2016-10-26 19:17:49 +00:00
Nemanja Ivanovic 0f45998bc6 [PowerPC] Implement vec_insert_exp builtins - llvm portion
This revision corresponds to review: https://reviews.llvm.org/D25957.
Committing on behalf of Zaara Syeda.

llvm-svn: 285225
2016-10-26 19:03:40 +00:00
Rafael Espindola cb09daab0a Call _exit.
As the state of lld gets more complicated, shutting down gets more
expensive.

In a normal lld run we can just call _exit immediately after renaming
the temporary output file. We still want the ability to run a full
shutdown since that is useful for detecting memory leaks.

This patch adds a --full-shutdown flag and changes lit to use it.

llvm-svn: 285224
2016-10-26 18:59:00 +00:00
Kostya Serebryany 2fabecaee3 [libFuzzer] simplify TracePC::HandleTrace even further. Also, when dealing with -exit_on_src_pos, symbolize every PC only once
llvm-svn: 285223
2016-10-26 18:52:04 +00:00
Chad Rosier 96e5e16acb Fix test from r285217.
llvm-svn: 285222
2016-10-26 18:49:16 +00:00
Rafael Espindola 99558efed6 Pass a InputSectionData to classoff.
This allows a non template class to hold input sections.

llvm-svn: 285221
2016-10-26 18:44:57 +00:00
Rui Ueyama 8fcc3afdd1 Move a helper function that doesn't depend on a class out of the class.
llvm-svn: 285220
2016-10-26 18:28:08 +00:00
Rui Ueyama 5ce977ccd0 Define a helper function to demangle symbols.
llvm-svn: 285219
2016-10-26 18:28:06 +00:00
Nemanja Ivanovic 85a28dcc5d [PPC] Implement vector reverse elements builtins (vec_reve)
This patch corresponds to review https://reviews.llvm.org/D25906.
Committing on behalf of Tony Jiang.

llvm-svn: 285218
2016-10-26 18:25:45 +00:00
Chad Rosier 0c621fda0d [AArch64] Avoid materializing constant 1 when generating cneg instructions.
Instead of

 cmp w0, #1
 orr w8, wzr, #0x1
 cneg w0, w8, ne

we now generate

 cmp w0, #1
 csinv w0, w0, wzr, eq

PR28965

llvm-svn: 285217
2016-10-26 18:15:32 +00:00
Enrico Granata 0cd95447db Actually use = delete to mark constructors and operators we want to not exist. Just declaring them as private works, but it can confuse certain tools as it doesn't actually declare intent, and the C++11 way is more expressive anyway, so this is pure win
rdar://28960209

llvm-svn: 285216
2016-10-26 18:12:52 +00:00
Dan Gohman 68a423bf84 [WebAssembly] Update the README.txt.
Update the README.txt with newer information, add a link to the Emscripten
page explaining the current easiest way to use the LLVM wasm backend, and
mention that other ways of using the LLVM wasm backend are in development.

llvm-svn: 285215
2016-10-26 17:44:09 +00:00
Nirav Dave e2369aafb9 [MC] Fix comma typo in .loc parsing
llvm-svn: 285214
2016-10-26 17:28:58 +00:00
Casey Carter 9450d5e262 Silence unused parameter warnings in archetypes.hpp
Reviewed at: https://reviews.llvm.org/D25958

llvm-svn: 285213
2016-10-26 17:22:25 +00:00
Robert Lougher 660f2f9560 Reapply: "Remove debug location from common tail when tail-merging"
This reapplies revision 285093.  Original commit message:

The branch folding pass tail merges blocks into a common-tail.  However, the
tail retains the debug information from one of the original inputs to the
merge (chosen randomly).  This is a problem for sampled-based PGO, as hits
on the common-tail will be attributed to whichever block was chosen,
irrespective of which path was actually taken to the common-tail.

This patch fixes the issue by nulling the debug location for the common-tail.

Differential Revision: https://reviews.llvm.org/D25742

llvm-svn: 285212
2016-10-26 17:01:47 +00:00