Sebastian Pop [Tue, 26 Jul 2016 00:15:08 +0000 (00:15 +0000)]
GVN-hoist: limit hoisting depth (PR28670)
This patch adds an option to specify the maximum depth in a BB at which to
consider hoisting instructions. Hoisting instructions from a deeper level is
not profitable as it increases register pressure and compilation time.
Differential Revision: https://reviews.llvm.org/D22772
llvm-svn: 276713
Evgeniy Stepanov [Tue, 26 Jul 2016 00:05:14 +0000 (00:05 +0000)]
[safestack] Fix stack guard live range.
Stack guard slot is live throughout the function.
llvm-svn: 276712
Chris Bieneman [Mon, 25 Jul 2016 23:48:14 +0000 (23:48 +0000)]
[CMake] Pass DYLD_LIBRARY_PATH as CMake variable instead of as envar
On OS X 10.11 System Integrity Protection prevents the DYLD environment variables from being set on system binaries. To work around this r276710 accepts DYLD_LIBRARY_PATH as a CMake variable and sets it directly on the archiver commands.
To make this work with bootstrapping we need to set DYLD_LIBRARY_PATH to the current stage's library directory and pass that into the next stage's configuration.
llvm-svn: 276711
Chris Bieneman [Mon, 25 Jul 2016 23:46:08 +0000 (23:46 +0000)]
[CMake] Support feeding DYLD_LIBRARY_PATH into archiver calls
OS X 10.11 has a feature named System Integrity Protection. The goal of the feature is to make system binaries immutable (even as root). One part of this is that protected binaries do not receive DYLD_* environment variables because the kernel scrubs them before process launch.
This causes problems for LTO bootstrap builds on Darwin that try to use the just-built libLTO with the host ar, ranlib, or libtool.
This patch addresses two problems.
(1) The tools themselves aren't protected binaries but the shim tools installed at / are, so we need to call xcrun -find to find libtool instead of using the one CMake finds.
(2) Some build tools (ninja and make) use /bin/sh to invoke their subprocesses. Since /bin/sh is a system binary, the kernel scrubs the DYLD envars from their environment. To work around this we need to set the environment variables as part of the archiver commands, so the envars are set by the shell process instead of on the shell process.
llvm-svn: 276710
Enrico Granata [Mon, 25 Jul 2016 23:26:24 +0000 (23:26 +0000)]
Print a warning if the directory passed to --test-subdir doesn't end up existing
llvm-svn: 276709
Xinliang David Li [Mon, 25 Jul 2016 23:12:53 +0000 (23:12 +0000)]
Define compatibility flag if not defined with -std=c++
llvm-svn: 276708
Adam Nemet [Mon, 25 Jul 2016 23:09:10 +0000 (23:09 +0000)]
[lit] Don't match tool names within new PM's <> markers
For example, stop expanding 'opt' in -passes='require<opt-remark-emit>'.
llvm-svn: 276707
Rui Ueyama [Mon, 25 Jul 2016 22:51:07 +0000 (22:51 +0000)]
Split LinkerScript::createSections into small functions.
createSections function is getting longer, so it is time to split it
into small functions. The reason why the function is long is because
it has deeply nested for-loops. This patch constructs temporary data
to reduce nesting level.
Differential Revision: https://reviews.llvm.org/D22786
llvm-svn: 276706
Rui Ueyama [Mon, 25 Jul 2016 22:41:42 +0000 (22:41 +0000)]
Fix parameter names.
match() returns true of the first argument, a target string, matches
one of the second argument, a list of glob patterns. Calling the
target string, which is not a glob pattern, "Pattern" was very confusing.
llvm-svn: 276705
Jan Vesely [Mon, 25 Jul 2016 22:36:22 +0000 (22:36 +0000)]
Make min follow the OCL 1.0 specs
OpenCL 1.0: "Returns y if y < x, otherwise it returns x. If x *and* y
are infinite or NaN, the return values are undefined."
OpenCL 1.1+: "Returns y if y < x, otherwise it returns x. If x *or* y
are infinite or NaN, the return values are undefined."
The 1.0 version is stricter so use that one.
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 276704
Michael Kuperstein [Mon, 25 Jul 2016 22:29:04 +0000 (22:29 +0000)]
Attempt to pacify windows bots, again.
llvm-svn: 276703
Rui Ueyama [Mon, 25 Jul 2016 22:26:28 +0000 (22:26 +0000)]
Do not pass InputFile to reportDiscarded().
Because the file from which an input section is created can be
obtained using getFile().
llvm-svn: 276702
Renato Golin [Mon, 25 Jul 2016 22:25:25 +0000 (22:25 +0000)]
[ARM] Saturation instructions are DSP-only
The saturation instructions appeared in v6T2, with DSP extensions, but they
were being accepted / generated on any, with the new introduction of the
saturation detection in the back-end. This commit restricts the usage to
DSP-enable only cores.
Fixes PR28607.
llvm-svn: 276701
David Majnemer [Mon, 25 Jul 2016 22:24:59 +0000 (22:24 +0000)]
Revert "[InstSimplify] Add support for bitcasts"
This reverts commit r276698. Clang has tests which rely on the
optimizer :(
llvm-svn: 276700
David Majnemer [Mon, 25 Jul 2016 22:04:58 +0000 (22:04 +0000)]
[InstSimplify] Add support for bitcasts
BitCasts of BitCasts can be folded away as can BitCasts which don't
change the type of the operand.
llvm-svn: 276698
Rui Ueyama [Mon, 25 Jul 2016 22:00:10 +0000 (22:00 +0000)]
Make ConstraintKind an enum class.
llvm-svn: 276697
Akira Hatanaka [Mon, 25 Jul 2016 21:58:19 +0000 (21:58 +0000)]
[Sema][ObjC] Compute the nullability of a conditional expression based
on the nullabilities of its operands.
This commit is a follow-up to r276076 and enables
computeConditionalNullability to compute the merged nullability when
the operands are objective-c pointers.
rdar://problem/
22074116
llvm-svn: 276696
Davide Italiano [Mon, 25 Jul 2016 21:47:13 +0000 (21:47 +0000)]
[LinkerScript] Refactor KEEP handling in a separate function
This will grow because I have a patch to support more complex
constructs, e.g.:
KEEP (*(EXCLUDE_FILE (*crtend.o *crtend?.o ) .dtors))
Let's make this a separate function.
llvm-svn: 276695
David Blaikie [Mon, 25 Jul 2016 21:41:42 +0000 (21:41 +0000)]
[WebAssembly] Update for Target API (TargetRegistry::RegisterMCAsmBackend) change
llvm-svn: 276694
Rui Ueyama [Mon, 25 Jul 2016 21:30:00 +0000 (21:30 +0000)]
Re-commit "Split LinkerScript::createSections".
Re-commit r276543 with a fix for buildbots.
llvm-svn: 276693
Simon Pilgrim [Mon, 25 Jul 2016 21:14:22 +0000 (21:14 +0000)]
[X86] Regenerate v2i256 shift legalization tests
llvm-svn: 276692
Simon Pilgrim [Mon, 25 Jul 2016 21:11:45 +0000 (21:11 +0000)]
[X86] Regenerate i64 shift legalization tests
llvm-svn: 276691
Tim Northover [Mon, 25 Jul 2016 21:01:29 +0000 (21:01 +0000)]
GlobalISel: add generic casts to IRTranslator
This adds LLVM's 3 main cast instructions (inttoptr, ptrtoint, bitcast) to the
IRTranslator. The first two are direct translations (with 2 MachineInstr types
each). Since LLT discards information, a bitcast might become trivial and we
emit a COPY in those cases instead.
llvm-svn: 276690
Tim Northover [Mon, 25 Jul 2016 21:01:17 +0000 (21:01 +0000)]
GlobalISel[AArch64]: support pointer types in argument lowering.
They're basically i64 for AArch64, but we'll leave them intact for stranger
targets. Also add some tests for the (very few) other cases we can handle right
now.
llvm-svn: 276689
Michael Kuperstein [Mon, 25 Jul 2016 20:52:00 +0000 (20:52 +0000)]
[PM] Port SymbolRewriter to the new PM
Differential Revision: https://reviews.llvm.org/D22703
llvm-svn: 276687
Kevin Enderby [Mon, 25 Jul 2016 20:36:36 +0000 (20:36 +0000)]
Next step along the way to getting good error messages for bad archives.
I consulted with Lang Hames on this work, and the goal was to add a bit
of "where" in the archive the error occurred along with what the error was.
So this step changes ArchiveMemberHeader into a class with a pointer
to the archive header and the parent archive. Which allows the methods
in the ArchiveMemberHeader to determine which member the header is
for to include that information in the error message.
For this first step the "where" is just the offset to the member in the
archive. The next step will be a new method on ArchiveMemberHeader
to get the full name, if possible, to be use in the error message. Which
will now be possible as ArchiveMemberHeader contains a pointer to
the Archive with its string table and its size, etc. so the full name can
be determined from the header if it is valid.
Also this change adds the missing checks the archive header is actually
contained in the buffer and is not truncated, as well as if the terminating
characters are correct in the header.
And changes one error message in Archive::Child::getNext() where the
name or offset to member is now added.
llvm-svn: 276686
Jordan Rose [Mon, 25 Jul 2016 20:34:25 +0000 (20:34 +0000)]
Fix r276671 to not use a defaulted move constructor.
MSVC won't provide the body of this move constructor and assignment
operator, possibly because the copy constructor is banned. Just write
it manually.
llvm-svn: 276685
Eugene Zelenko [Mon, 25 Jul 2016 20:30:13 +0000 (20:30 +0000)]
[Clang-rename] Remove custom version, fix extra space in error message.
Also fixed some Include What You Use Warnings.
Differential revision: https://reviews.llvm.org/D22654
llvm-svn: 276684
Chris Bieneman [Mon, 25 Jul 2016 20:25:38 +0000 (20:25 +0000)]
Revert "Enable cross-compilation across architectures on android"
This reverts commit r276333.
As I commented in the review (https://reviews.llvm.org/D22415), this change isn't needed because CMAKE_C_FLAGS is implicitly added by CMake to the command line for all C source files.
With this patch enabled CMAKE_C_FLAGS is duplicated on all C sources, and applied to ASM sources, which is not ideal.
I sent an email about this to llvm-commits on the commit thread. I suspect the problem the patch author was actually seeing is that CMAKE_C_FLAGS isn't applied to ASM files, and the builtins library has quite a few of those. The correct solution there is to specify CMAKE_ASM_FLAGS with whatever flags need to be passed to the compiler when compiling ASM files.
If there are other problems with flag propagation, please let me know.
llvm-svn: 276683
Jan Vesely [Mon, 25 Jul 2016 20:17:02 +0000 (20:17 +0000)]
AMDGPU: Remove read_workdim intrinsic
Differential revision: https://reviews.llvm.org/D22732
llvm-svn: 276682
Matt Arsenault [Mon, 25 Jul 2016 20:02:54 +0000 (20:02 +0000)]
Scalarizer: Support scalarizing intrinsics
llvm-svn: 276681
Matt Arsenault [Mon, 25 Jul 2016 19:48:29 +0000 (19:48 +0000)]
AMDGPU: Make skip threshold an option
llvm-svn: 276680
Matt Arsenault [Mon, 25 Jul 2016 19:39:06 +0000 (19:39 +0000)]
AMDGPU: Fix missing verify-machineinstrs in control flow test
llvm-svn: 276679
Matt Arsenault [Mon, 25 Jul 2016 19:39:04 +0000 (19:39 +0000)]
LiveIntervals: Return index from replaceMachineInstrInMaps
Fixes weird asymmetry with insertion
llvm-svn: 276678
Matt Arsenault [Mon, 25 Jul 2016 19:39:01 +0000 (19:39 +0000)]
MachineVerifier: Fix printing nonsense for physical registers
llvm-svn: 276677
Evgeniy Stepanov [Mon, 25 Jul 2016 19:25:40 +0000 (19:25 +0000)]
Fix invalid iterator use in safestack coloring.
llvm-svn: 276676
Matt Arsenault [Mon, 25 Jul 2016 19:06:25 +0000 (19:06 +0000)]
AMDGPU: Delete dead code
llvm-svn: 276675
Chris Bieneman [Mon, 25 Jul 2016 18:54:30 +0000 (18:54 +0000)]
[CMake] Cleaning up some CMake warnings
In Bootstrap builds Clang logs some warnings. These are caused because Clang passes CLANG_STAGE and BOOTSTRAP_DEFAULT_PASSTHROUGH into the next stage's configuration.
BOOTSTRAP_DEFAULT_PASSTHROUGH shouldn't be passed, so it is renamed to _BOOTSTRAP_DEFAULT_PASSTHROUGH, to prevent passthrough.
CLANG_STAGE should be passed, so I've changed the code to log it if it is set outside the if(CLANG_ENABLE_BOOTSTRAP) block. This makes the variable always used, so the warning goes away.
llvm-svn: 276674
Rong Xu [Mon, 25 Jul 2016 18:45:37 +0000 (18:45 +0000)]
[PGO] Fix profile mismatch in COMDAT function with pre-inliner
Pre-instrumentation inline (pre-inliner) greatly improves the IR
instrumentation code performance, among other benefits. One issue of the
pre-inliner is it can introduce CFG-mismatch for COMDAT functions. This
is due to the fact that the same COMDAT function may have different early
inline decisions across different modules -- that means different copies
of COMDAT functions will have different CFG checksum.
In this patch, we propose a partially renaming the COMDAT group and its
member function/variable so we have different profile counter for each
version. We will post-fix the COMDAT function and the group name with its
FunctionHash.
Differential Revision: http://reviews.llvm.org/D22600
llvm-svn: 276673
Michael Kuperstein [Mon, 25 Jul 2016 18:39:08 +0000 (18:39 +0000)]
Attempt to pacify windows bots.
llvm-svn: 276672
Jordan Rose [Mon, 25 Jul 2016 18:34:51 +0000 (18:34 +0000)]
StringSwitch cannot be copied (take 2).
This prevents StringSwitch from being used with 'auto', which is
important because the inferred type is StringSwitch rather than the
result type. This is a problem because StringSwitch stores addresses
of temporary values rather than copying or moving the value into its
own storage.
This is a compromise that still allows wrapping StringSwitch in other
temporary structures, which (unlike StringSwitch) may be non-trivial
to set up and therefore want to at least be movable. (For an example,
see QueryParser.cpp in clang-tools-extra.)
Changing this uncovered the bug in PassBuilder, also in this patch.
Clang doesn't seem to have any occurrences of the issue.
Re-commit of r276652.
llvm-svn: 276671
Daniel Berlin [Mon, 25 Jul 2016 18:19:49 +0000 (18:19 +0000)]
Revert NewGVN N^2 behavior patch
llvm-svn: 276670
Zachary Turner [Mon, 25 Jul 2016 18:18:59 +0000 (18:18 +0000)]
Add a modulemap for LLVMDebugInfoMsf.
Differential Revision: https://reviews.llvm.org/D22769
llvm-svn: 276669
Michael Kuperstein [Mon, 25 Jul 2016 18:10:54 +0000 (18:10 +0000)]
Don't use iplist in SymbolRewriter. NFC.
There didn't appear to be a good reason to use iplist in this case, a regular
list of unique_ptr works just as well.
Change made in preparation to a new PM port (since iplist is not moveable).
llvm-svn: 276668
Justin Bogner [Mon, 25 Jul 2016 18:07:14 +0000 (18:07 +0000)]
cmake: When adding lit testsuites, ignore Output directories
With in-tree builds we can get Output directories scattered among our
tests. Recursing into those to find tests doesn't make sense.
Thanks to nlewycky for noticing this!
llvm-svn: 276667
Simon Pilgrim [Mon, 25 Jul 2016 17:56:01 +0000 (17:56 +0000)]
[X86][SSE] Added 2048-bit vector comparison tests
Upper limit of what can be held in a <32 x i8> result
llvm-svn: 276666
Daniel Berlin [Mon, 25 Jul 2016 17:36:14 +0000 (17:36 +0000)]
NFC: Make a few asserts in GVNHoist do the same thing, but cheaper.
llvm-svn: 276662
Jordan Rose [Mon, 25 Jul 2016 17:28:33 +0000 (17:28 +0000)]
Revert "StringSwitch cannot be copied or moved."
This reverts commit r276652. The clang-query tool is currently
relying on this behavior. I'll try again later.
llvm-svn: 276661
Eugene Leviant [Mon, 25 Jul 2016 17:24:56 +0000 (17:24 +0000)]
[ELF] Fix bug in test case
llvm-svn: 276659
Daniel Berlin [Mon, 25 Jul 2016 17:24:27 +0000 (17:24 +0000)]
Fix N^2 instruction ordering comparisons in GVNHoist.
This fixes GVNHoist's portion of PR28670.
llvm-svn: 276658
Daniel Berlin [Mon, 25 Jul 2016 17:24:22 +0000 (17:24 +0000)]
NFC: Refactor GVNHoist class so not everything is public
llvm-svn: 276657
Joel Jones [Mon, 25 Jul 2016 17:18:44 +0000 (17:18 +0000)]
[cc1as] Add MCTargetOptions argument to createAsmBackend
Allow an assembler backend to get ABI options. This is to match the changes
to http://reviews.llvm.org/D16213.
Tested with "make check-clang"
Patch by: Joel Jones
Differential Revision: https://reviews.llvm.org/D16538
llvm-svn: 276655
Joel Jones [Mon, 25 Jul 2016 17:18:28 +0000 (17:18 +0000)]
MC] Provide an MCTargetOptions to implementors of MCAsmBackendCtorTy, NFC
Some targets, notably AArch64 for ILP32, have different relocation encodings
based upon the ABI. This is an enabling change, so a future patch can use the
ABIName from MCTargetOptions to chose which relocations to use. Tested using
check-llvm.
The corresponding change to clang is in: http://reviews.llvm.org/D16538
Patch by: Joel Jones
Differential Revision: https://reviews.llvm.org/D16213
llvm-svn: 276654
Sunil Srivastava [Mon, 25 Jul 2016 17:17:06 +0000 (17:17 +0000)]
Support '#pragma once' in headers when using PCH
The '#pragma once' directive was erroneously ignored when encountered
in the header-file specified in generate-PCH-mode. This resulted in
compile-time errors in some cases with legal code, and also a misleading
warning being produced.
Patch by Warren Ristow!
Differential Revision: http://reviews.llvm.org/D19815
llvm-svn: 276653
Jordan Rose [Mon, 25 Jul 2016 17:08:24 +0000 (17:08 +0000)]
StringSwitch cannot be copied or moved.
...but most importantly, it cannot be used well with 'auto', because
the inferred type is StringSwitch rather than the result type. This
is a problem because StringSwitch stores addresses of temporary
values rather than copying or moving the value into its own storage.
Changing this uncovered the bug in PassBuilder, also in this patch.
Clang doesn't seem to have any occurrences of the issue.
llvm-svn: 276652
Alexander Kornienko [Mon, 25 Jul 2016 17:08:18 +0000 (17:08 +0000)]
Revert "MPITypeMismatchCheck for Clang-Tidy"
This reverts commit r276640. Breaks multiple buildbots.
llvm-svn: 276651
Alexander Kornienko [Mon, 25 Jul 2016 17:08:10 +0000 (17:08 +0000)]
Revert "Remove trailing spaces."
This reverts commit r276641. Breaks multiple buildbots.
llvm-svn: 276650
Rui Ueyama [Mon, 25 Jul 2016 16:51:52 +0000 (16:51 +0000)]
Remove trailing whitespaces.
llvm-svn: 276649
Elena Demikhovsky [Mon, 25 Jul 2016 16:51:00 +0000 (16:51 +0000)]
AVX-512: Fixed [US]INT_TO_FP selection for i1 vectors.
It failed with assertion before this patch.
Differential Revision: https://reviews.llvm.org/D22735
llvm-svn: 276648
Douglas Katzman [Mon, 25 Jul 2016 16:36:02 +0000 (16:36 +0000)]
[Myriad]: better compatibility with vendor source
- Accept ma{2100,2150,2150} for -mcpu
- Define more preprocessor macros
- Don't append "le/" to little-endian lib dirs
llvm-svn: 276646
Tobias Grosser [Mon, 25 Jul 2016 16:31:21 +0000 (16:31 +0000)]
GPGPU: Load GPU kernels
We embed the PTX code into the host IR as a global variable and compile it
at run-time into a GPU kernel.
llvm-svn: 276645
Wei Mi [Mon, 25 Jul 2016 16:27:34 +0000 (16:27 +0000)]
Remove useless pass from the pipeline in test/Analysis/Dominators/2007-01-14-BreakCritEdges.ll.
llvm-svn: 276644
Davide Italiano [Mon, 25 Jul 2016 16:10:40 +0000 (16:10 +0000)]
[ELF/LinkerScript] Use correct section name in a test.
llvm-svn: 276642
Alexander Kornienko [Mon, 25 Jul 2016 15:43:22 +0000 (15:43 +0000)]
Remove trailing spaces.
llvm-svn: 276641
Alexander Kornienko [Mon, 25 Jul 2016 15:43:14 +0000 (15:43 +0000)]
MPITypeMismatchCheck for Clang-Tidy
Summary:
This check verifies if buffer type and MPI (Message Passing Interface)
datatype pairs match. All MPI datatypes defined by the MPI standard (3.1)
are verified by this check. User defined typedefs, custom MPI datatypes and
null pointer constants are skipped, in the course of verification.
Instructions on how to apply the check can be found at: https://github.com/0ax1/MPI-Checker/tree/master/examples
Reviewers: alexfh
Subscribers: cfe-commits
Projects: #clang-tools-extra
Patch by Alexander Droste!
Differential Revision: https://reviews.llvm.org/D21962
llvm-svn: 276640
Alexander Kornienko [Mon, 25 Jul 2016 15:27:16 +0000 (15:27 +0000)]
MPI-Checker: move MPIFunctionClassifier.h
Summary:
This patch moves the MPIFunctionClassifier header to `clang/include/clang/StaticAnalyzer/Checkers`,
in order to make it accessible in other parts of the architecture.
Reviewers: dcoughlin, zaks.anna
Subscribers: alexfh, cfe-commits
Patch by Alexander Droste!
Differential Revision: https://reviews.llvm.org/D22671
llvm-svn: 276639
Krzysztof Parzyszek [Mon, 25 Jul 2016 14:42:11 +0000 (14:42 +0000)]
[Hexagon] Add target feature to generate long calls
llvm-svn: 276638
Johannes Doerfert [Mon, 25 Jul 2016 12:48:45 +0000 (12:48 +0000)]
[GSoC] Add PolyhedralInfo pass - new interface to polly analysis
Adding a new pass PolyhedralInfo. This pass will be the interface to Polly.
Initially, we will provide the following interface:
- #IsParallel(Loop *L) - return a bool depending on whether the loop is
parallel or not for the given program order.
Patch by Utpal Bora <cs14mtech11017@iith.ac.in>
Differential Revision: https://reviews.llvm.org/D21486
llvm-svn: 276637
Tobias Grosser [Mon, 25 Jul 2016 12:47:39 +0000 (12:47 +0000)]
GPGPU: Emit data-transfer code
Also factor out getArraySize() to avoid code dupliciation and reorder some
function arguments to indicate the direction into which data is transferred.
llvm-svn: 276636
Tobias Grosser [Mon, 25 Jul 2016 12:47:33 +0000 (12:47 +0000)]
GPGPU: Complete code to allocate and free device arrays
At the beginning of each SCoP, we allocate device arrays for all arrays
used on the GPU and we free such arrays after the SCoP has been executed.
llvm-svn: 276635
Tobias Grosser [Mon, 25 Jul 2016 12:47:28 +0000 (12:47 +0000)]
GPURuntime: Add missing debug output
llvm-svn: 276634
Tobias Grosser [Mon, 25 Jul 2016 12:47:25 +0000 (12:47 +0000)]
GPURuntime: Fix typo in docu
llvm-svn: 276633
Tobias Grosser [Mon, 25 Jul 2016 12:47:22 +0000 (12:47 +0000)]
GPURuntime: Drop polly_cleanupGPGPUResources
This function is currently unused and won't be used in this form again. Instead
of freeing many unrelated items at the same time, we will instead explicitly
call free function from the host-IR we generate for each object we want to free.
These specific free functions will be added together with the corresponding
host-IR generation code.
llvm-svn: 276632
Johannes Doerfert [Mon, 25 Jul 2016 12:40:59 +0000 (12:40 +0000)]
[GSoC] Do not process SCoPs with infeasible runtime context
Do not process SCoPs with infeasible runtime context in the new
ScopInfoWrapperPass. Do not compute dependences for such SCoPs in the new
DependenceInfoWrapperPass.
Patch by Utpal Bora <cs14mtech11017@iith.ac.in>
Differential Revision: https://reviews.llvm.org/D22402
llvm-svn: 276631
Eugene Leviant [Mon, 25 Jul 2016 10:59:20 +0000 (10:59 +0000)]
[ELF] Test case for correct input section order, when sections are listed in linker script
llvm-svn: 276630
Sam Parker [Mon, 25 Jul 2016 10:11:00 +0000 (10:11 +0000)]
[ARM] Improve longMAC codegen test
Added thumb targets and dataflow checks to the longMAC test.
Differential Revision: https://reviews.llvm.org/D22684
llvm-svn: 276629
Simon Dardis [Mon, 25 Jul 2016 09:57:28 +0000 (09:57 +0000)]
[mips] Optimize materialization of i64 constants
Avoid MipsAnalyzeImmediate usage if the constant fits in an 32-bit
integer. This allows us to generate the same instructions for the
materialization of the same constants regardless the width of their
type.
Patch by: Vasileios Kalintiris
Contributions by: Simon Dardis
Reviewers: Daniel Sanders
Differential Review: https://reviews.llvm.org/D21689
llvm-svn: 276628
Roman Gareev [Mon, 25 Jul 2016 09:42:53 +0000 (09:42 +0000)]
Apply all necessary tilings and interchangings to get a macro-kernel
This is the second patch to apply the BLIS matmul optimization pattern
on matmul kernels
(http://www.cs.utexas.edu/users/flame/pubs/TOMS-BLIS-Analytical.pdf).
BLIS implements gemm as three nested loops around a macro-kernel, plus
two packing routines. The macro-kernel is implemented in terms
of two additional loops around a micro-kernel. The micro-kernel
is a loop around a rank-1 (i.e., outer product) update. In this change
we create the BLIS macro-kernel by applying a combination of tiling
and interchanging. In subsequent changes we will implement the packing
transformation.
Reviewed-by: Tobias Grosser <tobias@grosser.es>
Differential Revision: http://reviews.llvm.org/D21491
llvm-svn: 276627
Sam Parker [Mon, 25 Jul 2016 09:29:24 +0000 (09:29 +0000)]
[ARM] Small refactor of Thumb2 SMLA insts
Follow up to r276624. Changes bits 22-20 to be parameters to
instruction class.
Differential Revision: https://reviews.llvm.org/D22562
llvm-svn: 276626
Oliver Stannard [Mon, 25 Jul 2016 09:21:56 +0000 (09:21 +0000)]
[libunwind][ARM] Add support for Thumb1 targets
The Thumb1 version of the code for saving and restoring the unwind
context has a few bugs which prevent it from working:
* It uses the STM instruction without writeback, which is not valid for Thumb1
(It was introduced in Thumb2).
* It only saves/restores the low 8 registers, the sp and the lr, so if a
program uses r8-r12 they will not be correctly restored when throwing an
exception.
There aren't currently any Thumb1 build-bots to test this, but we have
been successfully running the libc++abi and libc++ test suites on
Cortex-M0 models, as well as some other test suites that use C++
exceptions on a downstream version of libunwind with this patch applied.
Differential Revision: https://reviews.llvm.org/D22292
llvm-svn: 276625
Sam Parker [Mon, 25 Jul 2016 09:20:20 +0000 (09:20 +0000)]
[ARM] Enable ISel of SMMLS for ARM and Thumb2
Use ISelDAGToDAG to recognise the SMMLS instruction pattern.
Differential Revision: https://reviews.llvm.org/D22562
llvm-svn: 276624
Tobias Grosser [Mon, 25 Jul 2016 09:16:01 +0000 (09:16 +0000)]
GPGPU: initialize GPU context and simplify the corresponding GPURuntime interface.
There is no need to expose the selected device at the moment. We also pass back
pointers as return values, as this simplifies the interface.
llvm-svn: 276623
Tobias Grosser [Mon, 25 Jul 2016 09:15:57 +0000 (09:15 +0000)]
IslNodeBuilder: Make finalize() virtual
This allows the finalization routine of the IslNodeBuilder to be overwritten
by derived classes. Being here, we also drop the unnecessary 'Scop' postfix
and the unnecessary 'Scop' parameter.
llvm-svn: 276622
Tobias Grosser [Mon, 25 Jul 2016 09:15:53 +0000 (09:15 +0000)]
GPURuntime: Check for debug-mode early on
Before this change, the debug statements in polly_initDevice would all be
skipped, as debug-mode would only be enabled _after_ they have already been run.
llvm-svn: 276621
Kirill Bobyrev [Mon, 25 Jul 2016 08:56:11 +0000 (08:56 +0000)]
[clang-rename] fix typos in tests
TemplateFunctionFindBy{Declaration|Use}.cpp contained typos and therefore were
failing. clang-rename passes these tests after typos are fixed.
llvm-svn: 276620
George Rimar [Mon, 25 Jul 2016 08:29:46 +0000 (08:29 +0000)]
[ELF] - Linkerscript: implemented output section [address] attribute.
Output section description in SECTIONS looks like that:
section [address] [(type)] :
...
{
...
}
Patch implements support of address atribute.
Differential revision: https://reviews.llvm.org/D22689
llvm-svn: 276619
Jonas Hahnfeld [Mon, 25 Jul 2016 08:04:26 +0000 (08:04 +0000)]
Update description for CLANG_DEFAULT_CXX_STDLIB and add comment. NFC
We agreed to call it "platform default" instead of "architecture default".
(see D17286)
llvm-svn: 276618
Tobias Grosser [Mon, 25 Jul 2016 08:03:08 +0000 (08:03 +0000)]
GPURuntime: Drop timing functionality (some leftover II)
llvm-svn: 276617
Roman Gareev [Mon, 25 Jul 2016 07:27:59 +0000 (07:27 +0000)]
[NFC] Refactor creation of the BLIS mirco-kernel and improve documentation
Reviewed-by: Tobias Grosser <tobias@grosser.es>
llvm-svn: 276616
Craig Topper [Mon, 25 Jul 2016 07:20:35 +0000 (07:20 +0000)]
[AVX512] Add load folding support for the unmasked forms of the FMA instructions.
llvm-svn: 276615
Craig Topper [Mon, 25 Jul 2016 07:20:31 +0000 (07:20 +0000)]
[AVX512] Add some additional patterns so that we can fold broadcast loads in the first argument of an FMADD/FMSUB/FNMADD/FNMSUB/FMADDSUB/FMSUBADD node. Also add patterns to support all combinations of the broadcast input and the preserved input for masked versions.
llvm-svn: 276614
Craig Topper [Mon, 25 Jul 2016 07:20:28 +0000 (07:20 +0000)]
[AVX512] Cleanup FMA operand order in patterns to match the VEX versions and to really be 213, 231, and 132.
llvm-svn: 276613
Tobias Grosser [Mon, 25 Jul 2016 07:11:49 +0000 (07:11 +0000)]
GPURuntime: Drop timing functionality (some leftover)
llvm-svn: 276612
Tobias Grosser [Mon, 25 Jul 2016 07:10:45 +0000 (07:10 +0000)]
GPURuntime: Drop timing functionality
This functionality won't be used in the current iteration. Drop it for now to
reduce the surface of the library. We can always add it back in when we need
it again.
llvm-svn: 276611
Sean Silva [Mon, 25 Jul 2016 05:57:59 +0000 (05:57 +0000)]
Cleanup : Reformat PartialInliner.cpp to have current LLVM style conventions
Modify the variable names and code style to be that of modern LLVM.
Patch by River Riddle!
Differential Revision: https://reviews.llvm.org/D22743
llvm-svn: 276610
Sean Silva [Mon, 25 Jul 2016 05:00:00 +0000 (05:00 +0000)]
Fix : Partial Inliner requires AssumptionCacheTracker
The public InlineFunction utility assumes that the passed in
InlineFunctionInfo has a valid AssumptionCacheTracker.
Patch by River Riddle!
Differential Revision: https://reviews.llvm.org/D22706
llvm-svn: 276609
Eric Fiselier [Mon, 25 Jul 2016 04:56:32 +0000 (04:56 +0000)]
Remove use of C++1z static assert in C++11 test
llvm-svn: 276608
David Majnemer [Mon, 25 Jul 2016 04:47:45 +0000 (04:47 +0000)]
Update test to account for r276604
llvm-svn: 276607
David Majnemer [Mon, 25 Jul 2016 04:47:44 +0000 (04:47 +0000)]
[MSVCToolChain] Guard hardcoded Windows paths with LLVM_ON_WIN32
Paths like C:/foo will never work on UNIX platforms, don't bother
implicitly adding them to the search path.
llvm-svn: 276606
Eric Fiselier [Mon, 25 Jul 2016 04:32:07 +0000 (04:32 +0000)]
Implement the std::pair parts of "Improving pair and tuple". Completes N4387.
llvm-svn: 276605