oota-llvm.git
9 years agoDon't Place Entry Safepoints Before the llvm.frameescape() Intrinsic
Philip Reames [Sun, 26 Apr 2015 19:41:23 +0000 (19:41 +0000)]
Don't Place Entry Safepoints Before the llvm.frameescape() Intrinsic

llvm.frameescape() intrinsic is not a real call. The intrinsic can only exist in the entry block. Inserting a gc.statepoint() before llvm.frameescape() may split the entry block, and push the intrinsic out of the entry block.

Patch by: Swaroop.Sridhar@microsoft.com
Differential Revision: http://reviews.llvm.org/D8910

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235820 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[bpf] fix build and remove a compiler warning in Release mode
Alexei Starovoitov [Sun, 26 Apr 2015 01:58:08 +0000 (01:58 +0000)]
[bpf] fix build and remove a compiler warning in Release mode

Patch by Brenden Blanco.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235814 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600: Remove / merge redundant testcases
Matt Arsenault [Sun, 26 Apr 2015 00:53:33 +0000 (00:53 +0000)]
R600: Remove / merge redundant testcases

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235813 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[x86] instcombine more cases of insertps into a shufflevector
Sanjay Patel [Sat, 25 Apr 2015 20:55:25 +0000 (20:55 +0000)]
[x86] instcombine more cases of insertps into a shufflevector

This is a follow-on to D8833 (insertps optimization when the zero mask is not used).

In this patch, we check for the case where the zmask is used, but both input vectors
to the insertps intrinsic are the same operand or the zmask overrides the destination
lane. This lets us replace the 2nd shuffle input operand with the zero vector.

Differential Revision: http://reviews.llvm.org/D9257

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235810 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoadd SSE run to check non-AVX codegen
Sanjay Patel [Sat, 25 Apr 2015 20:41:51 +0000 (20:41 +0000)]
add SSE run to check non-AVX codegen

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235809 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARM] Simplify code. NFC.
Benjamin Kramer [Sat, 25 Apr 2015 17:25:13 +0000 (17:25 +0000)]
[ARM] Simplify code. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235803 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[hexagon] Use range-based for loops. No functionality change intended.
Benjamin Kramer [Sat, 25 Apr 2015 14:46:53 +0000 (14:46 +0000)]
[hexagon] Use range-based for loops. No functionality change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235802 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[hexagon] Remove setHexLibcallName, it leaks memory.
Benjamin Kramer [Sat, 25 Apr 2015 14:46:46 +0000 (14:46 +0000)]
[hexagon] Remove setHexLibcallName, it leaks memory.

Just spell out the full names, it's not that much more code.
No functional change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235801 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoline endings fix
Simon Pilgrim [Sat, 25 Apr 2015 12:12:43 +0000 (12:12 +0000)]
line endings fix

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235800 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agobuild: make libunwind a proper project
Saleem Abdulrasool [Sat, 25 Apr 2015 01:47:39 +0000 (01:47 +0000)]
build: make libunwind a proper project

This allows the build infrastructure to properly detect and build libunwind.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235796 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix build error from accidental change
Andrew Kaylor [Fri, 24 Apr 2015 23:34:46 +0000 (23:34 +0000)]
Fix build error from accidental change

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235792 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Find correct cloned entry block for outlined handler functions.
Andrew Kaylor [Fri, 24 Apr 2015 23:27:32 +0000 (23:27 +0000)]
[WinEH] Find correct cloned entry block for outlined handler functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235791 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Find correct cloned entry block for outlined handler functions.
Andrew Kaylor [Fri, 24 Apr 2015 23:10:38 +0000 (23:10 +0000)]
[WinEH] Find correct cloned entry block for outlined handler functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235789 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAssert that Process::FindInEnvPath() is passed a relative path.
Nico Weber [Fri, 24 Apr 2015 22:18:46 +0000 (22:18 +0000)]
Assert that Process::FindInEnvPath() is passed a relative path.

It misbehaves with absolute paths. (So does path::append().)
Goes with clang r235787.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235788 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoLinker: Copy over function metadata attachments
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 22:07:31 +0000 (22:07 +0000)]
Linker: Copy over function metadata attachments

Update `lib/Linker` to handle `Function` metadata attachments.  The
attachments stick with the function body.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235786 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Add assembly/bitcode support for function metadata attachments
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 22:04:41 +0000 (22:04 +0000)]
IR: Add assembly/bitcode support for function metadata attachments

Add serialization support for function metadata attachments (added in
r235783).  The syntax is:

    define @foo() !attach !0 {

Metadata attachments are only allowed on functions with bodies.  Since
they come before the `{`, they're not really part of the body; since
they require a body, they're not really part of the header.  In
`LLParser` I gave them a separate function called from `ParseDefine()`,
`ParseOptionalFunctionMetadata()`.

In bitcode, I'm using the same `METADATA_ATTACHMENT` record used by
instructions.  Instruction metadata attachments are included in a
special "attachment" block at the end of a `Function`.  The attachment
records are laid out like this:

    InstID (KindID MetadataID)+

Note that these records always have an odd number of fields.  The new
code takes advantage of this to recognize function attachments (which
don't need an instruction ID):

    (KindID MetadataID)+

This means we can use the same attachment block already used for
instructions.

This is part of PR23340.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235785 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoVerifier: Function metadata attachments require a body
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:53:27 +0000 (21:53 +0000)]
Verifier: Function metadata attachments require a body

Add a verifier check that only functions with bodies have metadata
attachments.  This should help catch bugs in frontends and
transformation passes.  Part of PR23340.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235784 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Add Function metadata attachments
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:51:02 +0000 (21:51 +0000)]
IR: Add Function metadata attachments

Add IR support for `Metadata` attachments.  Assembly and bitcode support
will follow shortly, but for now we just have unit tests.  This is part
of PR23340.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235783 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoLLParser: Simplify ParseInstructionMetadata(), NFC
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:29:36 +0000 (21:29 +0000)]
LLParser: Simplify ParseInstructionMetadata(), NFC

Remove unused `PFS` variable and take the `Instruction` by reference.
(Not really related to PR23340, but might as well clean this up while
I'm here.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235782 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[DAGCombiner] Fix the type used in canFoldInAddressingMode to account for the
Quentin Colombet [Fri, 24 Apr 2015 21:28:00 +0000 (21:28 +0000)]
[DAGCombiner] Fix the type used in canFoldInAddressingMode to account for the
right scaling.

In the function canFoldInAddressingMode, VT is computed as the type of the
destination/source of a LOAD/STORE operations, instead of the memory type of the
operation.
On targets with a scaling factor on the offset of the LOAD/STORE operations, the
function may return false for actually valid cases. This may then prevent the
selection of profitable pre or post indexed load/store operations, and instead
select pre or post indexed load/store for unprofitable cases.

Patch by Francois de Ferriere <francois.de-ferriere@st.com>!

Differential Revision: http://reviews.llvm.org/D9146

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235780 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoLLParser: Split out ParseMetadataAttachment(), NFC
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:21:57 +0000 (21:21 +0000)]
LLParser: Split out ParseMetadataAttachment(), NFC

Make the code reusable for `Function` metadata attachments (PR23340).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235778 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[opaque pointer type] Verifier/AutoUpgrade: Remove a few uses of PointerType::getElem...
David Blaikie [Fri, 24 Apr 2015 21:16:07 +0000 (21:16 +0000)]
[opaque pointer type] Verifier/AutoUpgrade: Remove a few uses of PointerType::getElementType

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235777 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmWriter: Parameterize the syntactic separator for attachments
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:06:21 +0000 (21:06 +0000)]
AsmWriter: Parameterize the syntactic separator for attachments

Parameterize the separator for attachments, since `Function` metadata
attachments (PR23340) aren't going to use a `,` (comma).  No real
functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235775 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmWriter: Only collect attachment names once per module
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 21:03:05 +0000 (21:03 +0000)]
AsmWriter: Only collect attachment names once per module

Collect metadata names once per `AssemblyWriter` instead of every time
we need to print some attachments.  Just a drive-by; this caught my eye
while I was refactoring the code in r235772.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235774 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove an unused variable to prevent -Werror build failures.
Kaelyn Takata [Fri, 24 Apr 2015 21:02:18 +0000 (21:02 +0000)]
Remove an unused variable to prevent -Werror build failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235773 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmWriter: Split out code for printing Metadata attachments, NFC
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:59:52 +0000 (20:59 +0000)]
AsmWriter: Split out code for printing Metadata attachments, NFC

Refactor the code for printing `Instruction` metadata attachments so it
can be reused for `Function`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235772 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplifyCFG: Correctly handle switch lookup tables which fully cover the input type...
Hans Wennborg [Fri, 24 Apr 2015 20:57:56 +0000 (20:57 +0000)]
SimplifyCFG: Correctly handle switch lookup tables which fully cover the input type and use bit tests to check for holes

When using bit tests for hole checks, we call AddPredecessorToBlock to give the
phi node a value from the bit test block. This would break if we've
previously called removePredecessor on the default destination because the
switch is fully covered.

Test case by Mark Lacey.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235771 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Use a bitmask to access GlobalObject subclass data
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:47:23 +0000 (20:47 +0000)]
IR: Use a bitmask to access GlobalObject subclass data

Make room for more than just `Function::isMaterializable()` in the
`GlobalObject` subclass data bitfield.  Since we're treating it like a
bitfield, change `Function::Function()` to zero-out the whole thing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235770 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Extract set logic from Instruction attachments, NFC
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:36:25 +0000 (20:36 +0000)]
IR: Extract set logic from Instruction attachments, NFC

Extract the set logic for metadata attachments from `Instruction` so it
can be reused for `Function` (PR23340).

This data structure makes a `SmallVector<>` look (a little) like a map,
just doing the bare minimum to support the `Instruction` (and soon,
`Function`) metadata API.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235769 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[SEH] Implement GetExceptionCode in __except blocks
Reid Kleckner [Fri, 24 Apr 2015 20:25:05 +0000 (20:25 +0000)]
[SEH] Implement GetExceptionCode in __except blocks

This introduces an intrinsic called llvm.eh.exceptioncode. It is lowered
by copying the EAX value live into whatever basic block it is called
from. Obviously, this only works if you insert it late during codegen,
because otherwise mid-level passes might reschedule it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235768 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Use remove_if for Instruction::dropUnknownMetadata()
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:23:44 +0000 (20:23 +0000)]
IR: Use remove_if for Instruction::dropUnknownMetadata()

Technically the operations are different -- the old logic moved items
from the back into the opened-up slots, instead of the usual
`remove_if()` logic of a slow and a fast iterator -- but unless a
profile tells us otherwise I prefer the simpler logic here.  Regardless,
there shouldn't be an observable function change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235767 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Remove MDMapTy and MDPairTy typedefs, NFC
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:19:13 +0000 (20:19 +0000)]
IR: Remove MDMapTy and MDPairTy typedefs, NFC

Remove some typedefs in preparation for factoring out attachment logic
from `Instruction`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235764 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIR: Rename LLVMContextImpl::MetadataStore to InstructionMetadata
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 20:16:42 +0000 (20:16 +0000)]
IR: Rename LLVMContextImpl::MetadataStore to InstructionMetadata

Rename `MetadataStore` to the more explicit `InstructionMetadata`.  This
will make room for `FunctionMetadata` (start of PR23340).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235763 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[opaque pointer type] Add textual IR support for explicit type parameter to the invok...
David Blaikie [Fri, 24 Apr 2015 19:32:54 +0000 (19:32 +0000)]
[opaque pointer type] Add textual IR support for explicit type parameter to the invoke instruction

Same as r235145 for the call instruction - the justification, tradeoffs,
etc are all the same. The conversion script worked the same without any
false negatives (after replacing 'call' with 'invoke').

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235755 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PATCH] [Hexagon] Adding a test case for calling convention.
Sundeep Kushwaha [Fri, 24 Apr 2015 19:22:02 +0000 (19:22 +0000)]
[PATCH] [Hexagon] Adding a test case for calling convention.

http://reviews.llvm.org/D9241

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235754 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[AsmPrinter] Make AsmPrinter's OutStreamer member a unique_ptr.
Lang Hames [Fri, 24 Apr 2015 19:11:51 +0000 (19:11 +0000)]
[AsmPrinter] Make AsmPrinter's OutStreamer member a unique_ptr.

AsmPrinter owns the OutStreamer, so an owning pointer makes sense here. Using a
reference for this is crufty.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235752 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert changes to LTO test case since llvm-lto can't handle textual IR inputs
David Blaikie [Fri, 24 Apr 2015 18:13:27 +0000 (18:13 +0000)]
Revert changes to LTO test case since llvm-lto can't handle textual IR inputs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235738 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSkip extra LLVM IR assemble/disassemble steps in some tests
David Blaikie [Fri, 24 Apr 2015 18:06:09 +0000 (18:06 +0000)]
Skip extra LLVM IR assemble/disassemble steps in some tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235736 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[opaque pointer type] bitcode: add explicit callee type to invoke instructions
David Blaikie [Fri, 24 Apr 2015 18:06:06 +0000 (18:06 +0000)]
[opaque pointer type] bitcode: add explicit callee type to invoke instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235735 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix LoopInterchange/reductions.ll test for debug builds
Andrew Kaylor [Fri, 24 Apr 2015 17:39:16 +0000 (17:39 +0000)]
Fix LoopInterchange/reductions.ll test for debug builds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235734 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoTeach AArch64\lit.local.cfg the new triple names windows-gnu and windows-msvc.
Yaron Keren [Fri, 24 Apr 2015 17:14:16 +0000 (17:14 +0000)]
Teach AArch64\lit.local.cfg the new triple names windows-gnu and windows-msvc.

Tests were failing when built with -DLLVM_DEFAULT_TARGET_TRIPLE=i686-pc-windows-gnu.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235733 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[CMake] Fix for PR 23328: LLVM_OPTIMIZED_TABLEGEN broken
Chris Bieneman [Fri, 24 Apr 2015 17:09:20 +0000 (17:09 +0000)]
[CMake] Fix for PR 23328: LLVM_OPTIMIZED_TABLEGEN broken

In CMake dependencies can be filenames or targets, and targets can't be filenames. The Ninja generator handles filename dependencies because it generates targets for every output file from a command. For example:

add_custom_command(OUTPUT foo.txt COMMAND touch foo.txt)

With the Ninja generator this generates a target foo.txt, but with the Makefile generator it doesn't. This is probably because Ninja explicitly requires these hard dependency ties, and Make just behaves oddly in general.

To fix this we need to make the tablegen actions depend on a target rather than a filename.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235732 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoLinker: Update -override testcase to check callers
Duncan P. N. Exon Smith [Fri, 24 Apr 2015 16:56:24 +0000 (16:56 +0000)]
Linker: Update -override testcase to check callers

Check that `@main` is calling `@foo2` (the renamed internal function),
not the `@foo` with external linkage that's been pulled in from the
override file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235730 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSwitch lowering: fix APInt overflow causing infinite loop / OOM
Hans Wennborg [Fri, 24 Apr 2015 16:53:55 +0000 (16:53 +0000)]
Switch lowering: fix APInt overflow causing infinite loop / OOM

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235729 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Split the landingpad BB instead of cloning it
Reid Kleckner [Fri, 24 Apr 2015 16:22:19 +0000 (16:22 +0000)]
[WinEH] Split the landingpad BB instead of cloning it

This means we don't have to RAUW the landingpad instruction and
landingpad BB, which is a nice win.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235725 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix typo in comment.
Diego Novillo [Fri, 24 Apr 2015 15:46:41 +0000 (15:46 +0000)]
Fix typo in comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235723 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUse the cleaner syntx value initialization to zero initialize POD structs.
Yaron Keren [Fri, 24 Apr 2015 15:39:47 +0000 (15:39 +0000)]
Use the cleaner syntx value initialization to zero initialize POD structs.
Suggestion from David Blaikie!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235721 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSilence clang warning: private field 'data_' is not used.
Yaron Keren [Fri, 24 Apr 2015 15:10:15 +0000 (15:10 +0000)]
Silence clang warning: private field 'data_' is not used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235720 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSilence clang warning: missing field 'Dr0' initializer.
Yaron Keren [Fri, 24 Apr 2015 14:26:27 +0000 (14:26 +0000)]
Silence clang warning: missing field 'Dr0' initializer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235719 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[mips][FastISel] Specify which types we handle for integer extension.
Vasileios Kalintiris [Fri, 24 Apr 2015 13:48:19 +0000 (13:48 +0000)]
[mips][FastISel] Specify which types we handle for integer extension.

Summary:
Perform integer extension only when the destination type is one of
i8, i16 & i32 and when the source type is i1, i8 or i16. For other
combinations we fall back to SelectionDAG.

This fixes the test MultiSource/Benchmarks/7zip that was failing in our
out-of-tree MIPS buildbots.

Reviewers: dsanders

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D9243

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235718 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemoving dead code; NFC. This code was triggering a C4718 warning (recursive call...
Aaron Ballman [Fri, 24 Apr 2015 12:51:45 +0000 (12:51 +0000)]
Removing dead code; NFC. This code was triggering a C4718 warning (recursive call has no side effects, deleting) with MSVC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235717 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[BitcodeReader] Fix asserts when we read a non-vector type for insert/extract/shuffle
Filipe Cabecinhas [Fri, 24 Apr 2015 11:30:15 +0000 (11:30 +0000)]
[BitcodeReader] Fix asserts when we read a non-vector type for insert/extract/shuffle

Added some additional checking for vector types + tests.

Bug found with AFL fuzz.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235710 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoCorrect extractelement constant folding
Pawel Bylica [Fri, 24 Apr 2015 07:42:35 +0000 (07:42 +0000)]
Correct extractelement constant folding

Summary: Constant folding of extractelement with out-of-bound index produces undef also for indexes bigger than 64bit (instead of crash assert failure as before)

Test Plan: Unit tests included.

Reviewers: majnemer

Reviewed By: majnemer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D9225

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235700 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix APInt long division algorithm
Pawel Bylica [Fri, 24 Apr 2015 07:38:39 +0000 (07:38 +0000)]
Fix APInt long division algorithm

Summary: This patch fixes step D4 of Knuth's division algorithm implementation. Negative sign of the step result was not always detected due to incorrect "borrow" handling.

Test Plan: Unit test that reveals the bug included.

Reviewers: chandlerc, yaron.keren

Reviewed By: yaron.keren

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D9196

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235699 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen] Don't leak Expanders and Operators in SetTheory.
Craig Topper [Fri, 24 Apr 2015 06:49:44 +0000 (06:49 +0000)]
[TableGen] Don't leak Expanders and Operators in SetTheory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235697 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen] Fix all remaining memory leaks of Init and RecTy objects.
Craig Topper [Fri, 24 Apr 2015 05:38:48 +0000 (05:38 +0000)]
[TableGen] Fix all remaining memory leaks of Init and RecTy objects.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235696 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoResurrect r235688
Jingyue Wu [Fri, 24 Apr 2015 04:22:39 +0000 (04:22 +0000)]
Resurrect r235688

We should skip vector types which are not SCEVable.

test/CodeGen/NVPTX/sched2.ll passes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235695 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert r235688
Jingyue Wu [Fri, 24 Apr 2015 03:26:11 +0000 (03:26 +0000)]
Revert r235688

Seems breaking builds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235690 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NVPTX] Emits "generic()" depending on the original address space
Jingyue Wu [Fri, 24 Apr 2015 02:57:30 +0000 (02:57 +0000)]
[NVPTX] Emits "generic()" depending on the original address space

Summary:
Fixes a bug in the NVPTX codegen. The code used to miss necessary "generic()"
on aggregates of addrspacecasts.

Test Plan: addrspacecast-gvar.ll

Reviewers: eliben, jholewinski

Reviewed By: jholewinski

Subscribers: jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D9130

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235689 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NVPTX] enable NaryReassociate in NVPTX
Jingyue Wu [Fri, 24 Apr 2015 02:54:06 +0000 (02:54 +0000)]
[NVPTX] enable NaryReassociate in NVPTX

Summary:
We run NaryReassociate right after SLSR because SLSR enables many
opportunities for NaryReassociate. For example, in nary-slsr.ll

  foo((a + b) + c);
  foo((a + b * 2) + c);
  foo((a + b * 3) + c);   // 2 muls and 6 adds

after SLSR:

  ab = a + b;
  foo(ab + c);
  ab2 = ab + b;
  foo(ab2 + c);
  ab3 = ab2 + b;
  foo(ab3 + c);           // 6 adds

after NaryReassociate:

  abc = (a + b) + c;
  foo(abc);
  ab2c = abc + b;
  foo(ab2c);
  ab3c = ab2c + b;
  foo(ab3c);              // 4 adds

Test Plan: nary-slsr.ll

Reviewers: jholewinski, eliben

Reviewed By: eliben

Subscribers: jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D9066

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235688 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix verifier error when producing v_madmk_f32
Matt Arsenault [Fri, 24 Apr 2015 01:57:58 +0000 (01:57 +0000)]
R600/SI: Fix verifier error when producing v_madmk_f32

Copy the kill flags when swapping the operands.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235687 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoImprove isTriviallyReMaterializable() documentation.
Matthias Braun [Fri, 24 Apr 2015 01:15:27 +0000 (01:15 +0000)]
Improve isTriviallyReMaterializable() documentation.

This should make it clear under which narrow circumstances implicit
physreg uses are okay when rematerializing and prevent people from
accidentally allowing too much when overriding
isReallyTriviallyReMaterializable() even with the weaker assert in the
RegisterCoalescer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235679 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/RegisterCoalescer: Enable more rematerialization/add missing testcase
Matthias Braun [Fri, 24 Apr 2015 00:25:50 +0000 (00:25 +0000)]
R600/RegisterCoalescer: Enable more rematerialization/add missing testcase

This enables the rematerialization of some R600 MOV instructions in the
RegisterCoalescer and adds a testcase for r235668.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235675 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix a couple of typos in comments.
Michael Zolotukhin [Fri, 24 Apr 2015 00:10:27 +0000 (00:10 +0000)]
Fix a couple of typos in comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235674 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRegisterCoalescer: implicit phsreg uses are fine when rematerializing
Matthias Braun [Fri, 24 Apr 2015 00:01:37 +0000 (00:01 +0000)]
RegisterCoalescer: implicit phsreg uses are fine when rematerializing

The target hooks should have already checked them. This change is
necessary to enable the remateriailzation on R600.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235673 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert accidentally committed "MC: Allow targets to stop symbol name quoting"
Matt Arsenault [Thu, 23 Apr 2015 23:34:51 +0000 (23:34 +0000)]
Revert accidentally committed "MC: Allow targets to stop symbol name quoting"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235672 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Special case v_mov_b32 as really rematerializable
Matt Arsenault [Thu, 23 Apr 2015 23:34:48 +0000 (23:34 +0000)]
R600/SI: Special case v_mov_b32 as really rematerializable

This should be fixed to properly understand all rematerializable
instructions while ignoring implicit reads of exec.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235671 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMC: Allow targets to stop symbol name quoting
Matt Arsenault [Thu, 23 Apr 2015 23:34:05 +0000 (23:34 +0000)]
MC: Allow targets to stop symbol name quoting

Currently symbol names are printed in quotes if it contains something
outside of the arbitrary set of characters that isAcceptableChar tests
for. On somem targets, it is never OK to print a symbol name in quotes
so allow targets to opt out of this behavior.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235670 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRegisterCoalescer: Avoid unnecessary register class widening for some rematerializations
Matthias Braun [Thu, 23 Apr 2015 23:24:36 +0000 (23:24 +0000)]
RegisterCoalescer: Avoid unnecessary register class widening for some rematerializations

I couldn't provide a testcase as none of the public targets has wide
register classes with alot of subregisters and at the same time an
instruction which "ReMaterializable" and "AsCheapAsAMove" (could
probably be added for R600).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235668 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRe-commit "[SEH] Remove the old __C_specific_handler code now that WinEHPrepare works"
Reid Kleckner [Thu, 23 Apr 2015 23:22:33 +0000 (23:22 +0000)]
Re-commit "[SEH] Remove the old __C_specific_handler code now that WinEHPrepare works"

This reverts commit r235617.

r235649 should have addressed the problems.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235667 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix modules build post-r235612.
Richard Smith [Thu, 23 Apr 2015 23:22:26 +0000 (23:22 +0000)]
Fix modules build post-r235612.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235666 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Support register name prefixes for vector registers
Hal Finkel [Thu, 23 Apr 2015 23:16:22 +0000 (23:16 +0000)]
[PowerPC] Support register name prefixes for vector registers

Match binutils by supporting the optional register name prefix for new vector
registers ("vs" for VSX registers and "q" for QPX registers).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235665 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Use sync inst alias when printing
Hal Finkel [Thu, 23 Apr 2015 23:05:08 +0000 (23:05 +0000)]
[PowerPC] Use sync inst alias when printing

So long as the choice between printing msync and sync is not ambiguous, we can
print 'sync 0' and just 'sync'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235663 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600: Correctly lower CONCAT_VECTOR nodes with more than 2 operands
Tom Stellard [Thu, 23 Apr 2015 22:59:24 +0000 (22:59 +0000)]
R600: Correctly lower CONCAT_VECTOR nodes with more than 2 operands

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235662 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix comment for NoCommonBits.
Michael Zolotukhin [Thu, 23 Apr 2015 22:55:48 +0000 (22:55 +0000)]
Fix comment for NoCommonBits.

Maybe there is a better wording, but at least it should be technically
correct now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235660 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Add asm/disasm support for dcbt with hint
Hal Finkel [Thu, 23 Apr 2015 22:47:57 +0000 (22:47 +0000)]
[PowerPC] Add asm/disasm support for dcbt with hint

Add assembler/disassembler support for dcbt/dcbtst (and aliases) with the hint
field specified (non-zero). Unforunately, the syntax for this instruction is
special in that it differs for server vs. embedded cores:
   dcbt ra, rb, th [server]
   dcbt th, ra, rb [embedded]
where th can be omitted when it is 0. dcbtst is the same. Thus we need to play
games in the parser and the printer to flip the operands around on the embedded
cores. We'll use the server syntax as the default (binutils currently uses the
embedded form by default, but IBM is changing that).

We also stop marking dcbtst as having unmodeled side effects (this is not
necessary, it is just a hint like dcbt -- noticed by inspection, so no separate
test case).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235657 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Ignore filter clauses while mapping landing pad blocks.
Andrew Kaylor [Thu, 23 Apr 2015 22:38:36 +0000 (22:38 +0000)]
[WinEH] Ignore filter clauses while mapping landing pad blocks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235656 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove trivial assert to fix NDEBUG Werror builds
Reid Kleckner [Thu, 23 Apr 2015 21:36:32 +0000 (21:36 +0000)]
Remove trivial assert to fix NDEBUG Werror builds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235652 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRecommit r235458: [opaque pointer type] Avoid using PointerType::getElementType for...
David Blaikie [Thu, 23 Apr 2015 21:36:23 +0000 (21:36 +0000)]
Recommit r235458: [opaque pointer type] Avoid using PointerType::getElementType for a few cases of CallInst

(reverted in r235533)

Original commit message:

"Calls to llvm::Value::mutateType are becoming extra-sensitive now that
instructions have extra type information that will not be derived from
operands or result type (alloca, gep, load, call/invoke, etc... ). The
special-handling for mutateType will get more complicated as this work
continues - it might be worth making mutateType virtual & pushing the
complexity down into the classes that need special handling. But with
only two significant uses of mutateType (vectorization and linking) this
seems OK for now.

Totally open to ideas/suggestions/improvements, of course.

With this, and a bunch of exceptions, we can roundtrip an indirect call
site through bitcode and IR. (a direct call site is actually trickier...
I haven't figured out how to deal with the IR deserializer's lazy
construction of Function/GlobalVariable decl's based on the type of the
entity which means looking through the "pointer to T" type referring to
the global)"

The remapping done in ValueMapper for LTO was insufficient as the types
weren't correctly mapped (though I was using the post-mapped operands,
some of those operands might not have been mapped yet so the type
wouldn't be post-mapped yet). Instead use the pre-mapped type and
explicitly map all the types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235651 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Replace more lpad value uses with undef
Reid Kleckner [Thu, 23 Apr 2015 21:22:30 +0000 (21:22 +0000)]
[WinEH] Replace more lpad value uses with undef

We were asserting on code like this:
  extern "C" unsigned long _exception_code();
  void might_crash(unsigned long);
  void foo() {
    __try {
      might_crash(0);
    } __except(1) {
      might_crash(_exception_code());
    }
  }

Gtest and many other libraries get the exception code from the __except
block. What's supposed to happen here is that EAX is live into the
__except block, and it contains the exception code. Eventually we'll
represent that as a use of the landingpad ehptr value, but for now we
can replace it with undef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235649 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[MachineCopyPropagation] Handle undef flags conservatively so that we do not
Quentin Colombet [Thu, 23 Apr 2015 21:17:39 +0000 (21:17 +0000)]
[MachineCopyPropagation] Handle undef flags conservatively so that we do not
remove copies that are useful after breaking some hardware dependencies.
In other words, handle this kind of situations conservatively by assuming reg2
is redefined by the undef flag.
reg1 = copy reg2
= inst reg2<undef>
reg2 = copy reg1
Copy propagation used to remove the last copy.
This is incorrect because the undef flag on reg2 in inst, allows next
passes to put whatever trashed value in reg2 that may help.
In practice we end up with this code:
reg1 = copy reg2
reg2 = 0
= inst reg2<undef>
reg2 = copy reg1

This fixes PR21743.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235647 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUnbreak build
Krzysztof Parzyszek [Thu, 23 Apr 2015 20:57:39 +0000 (20:57 +0000)]
Unbreak build

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235646 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] Minor cleanup in HexagonFrameLowering
Krzysztof Parzyszek [Thu, 23 Apr 2015 20:42:20 +0000 (20:42 +0000)]
[Hexagon] Minor cleanup in HexagonFrameLowering

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235645 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix indirect addressing with a negative constant offset
Tom Stellard [Thu, 23 Apr 2015 20:32:01 +0000 (20:32 +0000)]
R600/SI: Fix indirect addressing with a negative constant offset

When the base register index of the vector plus the constant offset
was less than zero, we were passing the wrong base register to the indirect
addressing instruction.

In this case, we need to set the base register to v0 and then add
the computed (negative) index to m0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235641 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThumb2: When applying branch optimizations, visit branches in reverse order.
Peter Collingbourne [Thu, 23 Apr 2015 20:31:35 +0000 (20:31 +0000)]
Thumb2: When applying branch optimizations, visit branches in reverse order.

The order in which branches appear in ImmBranches is approximately their
order within the function body. By visiting later branches first, we reduce
the distance between earlier forward branches and their targets, making it
more likely that the cbn?z optimization, which can only apply to forward
branches, will succeed for those earlier branches.

Differential Revision: http://reviews.llvm.org/D9185

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235640 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM: When re-creating a branch via InsertBranch, preserve CPSR flags.
Peter Collingbourne [Thu, 23 Apr 2015 20:31:32 +0000 (20:31 +0000)]
ARM: When re-creating a branch via InsertBranch, preserve CPSR flags.

In particular, this preserves the kill flag, which allows the Thumb2 cbn?z
optimization to be applied in cases where a branch has been re-created after
the live variables analysis pass, e.g. by the machine block placement pass.

This appears to be low risk; a number of other targets seem to already be
doing something similar, e.g. AArch64, PowerPC.

Differential Revision: http://reviews.llvm.org/D9184

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235639 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThumb2: When optimizing for size, do not if-convert branches involving comparisons...
Peter Collingbourne [Thu, 23 Apr 2015 20:31:30 +0000 (20:31 +0000)]
Thumb2: When optimizing for size, do not if-convert branches involving comparisons with zero.

This allows the constant island pass to lower these branches to cbn?z
instructions, resulting in a shorter instruction sequence.

Differential Revision: http://reviews.llvm.org/D9183

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235638 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM: When spilling extra registers for alignment, prefer low registers on all Thumb...
Peter Collingbourne [Thu, 23 Apr 2015 20:31:26 +0000 (20:31 +0000)]
ARM: When spilling extra registers for alignment, prefer low registers on all Thumb targets.

This makes it more likely that we can use the 16-bit push and pop instructions
on Thumb-2, saving around 4 bytes per function.

Differential Revision: http://reviews.llvm.org/D9165

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235637 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoARM: Only enforce 4-byte alignment on Thumb-2 functions with constant pools.
Peter Collingbourne [Thu, 23 Apr 2015 20:31:22 +0000 (20:31 +0000)]
ARM: Only enforce 4-byte alignment on Thumb-2 functions with constant pools.

This appears to have been introduced back in r76698 as part of an unrelated
change. I can find no official ARM documentation stating that Thumb-2 functions
require 4-byte alignment; in fact, ARM documentation appears to contradict
this (see, e.g., ARM Architecture Reference Manual Thumb-2 Supplement,
section 2.6.1: "Thumb-2 enforces 16-bit alignment on all instructions.").

Also remove code that sets alignment for ARM functions, which is redundant
with code in the MachineFunction constructor, and remove the hidden
-arm-align-constant-islands flag, which has been enabled by default since
r146739 (Dec 2011) and has probably received sufficient testing by now.

Differential Revision: http://reviews.llvm.org/D9138

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235636 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] Fix compiler warnings in release build
Krzysztof Parzyszek [Thu, 23 Apr 2015 20:26:21 +0000 (20:26 +0000)]
[Hexagon] Fix compiler warnings in release build

Patch by Aditya Nandakumar.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235635 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[getUnderlyingOjbects] Analyze loop PHIs further to remove false positives
Adam Nemet [Thu, 23 Apr 2015 20:09:20 +0000 (20:09 +0000)]
[getUnderlyingOjbects] Analyze loop PHIs further to remove false positives

Specifically, if a pointer accesses different underlying objects in each
iteration, don't look through the phi node defining the pointer.

The motivating case is the underlyling-objects-2.ll testcase.  Consider
the loop nest:

  int **A;
  for (i)
    for (j)
       A[i][j] = A[i-1][j] * B[j]

This loop is transformed by Load-PRE to stash away A[i] for the next
iteration of the outer loop:

  Curr = A[0];          // Prev_0
  for (i: 1..N) {
    Prev = Curr;        // Prev = PHI (Prev_0, Curr)
    Curr = A[i];
    for (j: 0..N)
       Curr[j] = Prev[j] * B[j]
  }

Since A[i] and A[i-1] are likely to be independent pointers,
getUnderlyingObjects should not assume that Curr and Prev share the same
underlying object in the inner loop.

If it did we would try to dependence-analyze Curr and Prev and the
analysis of the corresponding SCEVs would fail with non-constant
distance.

To fix this, the getUnderlyingObjects API is extended with an optional
LoopInfo parameter.  This is effectively what controls whether we want
the above behavior or the original.  Currently, I only changed to use
this approach for LoopAccessAnalysis.

The other testcase is to guard the opposite case where we do want to
look through the loop PHI.  If we step through an array by incrementing
a pointer, the underlying object is the incoming value of the phi as the
loop is entered.

Fixes rdar://problem/19566729

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235634 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NVPTX] run SeparateConstOffsetFromGEP before SLSR
Jingyue Wu [Thu, 23 Apr 2015 20:00:04 +0000 (20:00 +0000)]
[NVPTX] run SeparateConstOffsetFromGEP before SLSR

Summary:
We pick this order because SeparateConstOffsetFromGEP may create more
opportunities for SLSR.

Test Plan:
reassociate-geps-and-slsr.ll
no performance regression on internal benchmarks

Reviewers: meheff

Subscribers: llvm-commits, jholewinski

Differential Revision: http://reviews.llvm.org/D9230

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235632 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Add missing -mcpu=SI to assembler test
Tom Stellard [Thu, 23 Apr 2015 19:33:55 +0000 (19:33 +0000)]
R600/SI: Add missing -mcpu=SI to assembler test

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235630 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Add assembler support for all CI and VI VOP1 instructions
Tom Stellard [Thu, 23 Apr 2015 19:33:54 +0000 (19:33 +0000)]
R600/SI: Add assembler support for all CI and VI VOP1 instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235629 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: v_mov_fed_b32 does not exist on VI
Tom Stellard [Thu, 23 Apr 2015 19:33:52 +0000 (19:33 +0000)]
R600/SI: v_mov_fed_b32 does not exist on VI

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235628 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Use a better error message for unsupported instructions in the assembler
Tom Stellard [Thu, 23 Apr 2015 19:33:51 +0000 (19:33 +0000)]
R600/SI: Use a better error message for unsupported instructions in the assembler

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235627 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Improve AsmParser support for forced e64 encoding
Tom Stellard [Thu, 23 Apr 2015 19:33:48 +0000 (19:33 +0000)]
R600/SI: Improve AsmParser support for forced e64 encoding

We can now force e64 encoding even when the operands would be legal
for e32 encoding.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235626 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Handle stubs for outlined functions that have only unreached terminators.
Andrew Kaylor [Thu, 23 Apr 2015 18:37:39 +0000 (18:37 +0000)]
[WinEH] Handle stubs for outlined functions that have only unreached terminators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235618 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert "[SEH] Remove the old __C_specific_handler code now that WinEHPrepare works"
Reid Kleckner [Thu, 23 Apr 2015 18:34:01 +0000 (18:34 +0000)]
Revert "[SEH] Remove the old __C_specific_handler code now that WinEHPrepare works"

We still have some "uses remain after removal" issues in -O0 builds.

This reverts commit r235557.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235617 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Enable printing instructions using aliases
Hal Finkel [Thu, 23 Apr 2015 18:30:38 +0000 (18:30 +0000)]
[PowerPC] Enable printing instructions using aliases

TableGen had been nicely generating code to print a number of instructions using
shorter aliases (and PowerPC has plenty of short mnemonics), but we were not
calling it. For some of the aliases we support in the parser, TableGen can't
infer the "inverse" alias relationship, so there is still more to do.

Thus, after some hours of updating test cases...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235616 91177308-0d34-0410-b5e6-96231b3b80d8