oota-llvm.git
7 years agoAMDGPU/SI: Fix extra space when printing v_div_fmas_*
Matt Arsenault [Sun, 28 Jun 2015 18:16:14 +0000 (18:16 +0000)]
AMDGPU/SI: Fix extra space when printing v_div_fmas_*

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240911 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLSR] S's basis must have the same type as S
Jingyue Wu [Sun, 28 Jun 2015 17:45:05 +0000 (17:45 +0000)]
[SLSR] S's basis must have the same type as S

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240910 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[x86][AVX512]
Asaf Badouh [Sun, 28 Jun 2015 14:30:39 +0000 (14:30 +0000)]
[x86][AVX512]
Add vscalef support
include encoding and intrinsics

review:
http://reviews.llvm.org/D10730

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240906 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAVX-512: Added all SKX forms of GATHER instructions.
Elena Demikhovsky [Sun, 28 Jun 2015 10:53:29 +0000 (10:53 +0000)]
AVX-512: Added all SKX forms of GATHER instructions.
Added intrinsics.
Added encoding and tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240905 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Debug Info: One more bitfield bugfix. While yesterday's r240853 fixed"
Adrian Prantl [Sat, 27 Jun 2015 21:55:00 +0000 (21:55 +0000)]
Revert "Debug Info: One more bitfield bugfix. While yesterday's r240853 fixed"

This reverts commit 240890. Breaking the gdb buildbot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240893 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDon't use %llc_dwarf for target-specific tests.
Benjamin Kramer [Sat, 27 Jun 2015 21:11:43 +0000 (21:11 +0000)]
Don't use %llc_dwarf for target-specific tests.

Should fix running them on windows.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240892 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SDAG] Now that we have a way to communicate the exact bit on sdiv use it to simplify...
Benjamin Kramer [Sat, 27 Jun 2015 20:33:26 +0000 (20:33 +0000)]
[SDAG] Now that we have a way to communicate the exact bit on sdiv use it to simplify sdiv by a constant.

We had a hack in SDAGBuilder in place to work around this but now we
can avoid that. Call BuildExactSDIV from BuildSDIV so DAGCombiner can
perform this trick automatically.

The added check in DAGCombiner is necessary to prevent exact sdiv by pow2
from regressing as the target-specific pow2 lowering is not aware of
exact bits yet.

This is mostly covered by existing tests. One side effect is that we
get the better lowering for exact vector sdivs now too :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240891 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDebug Info: One more bitfield bugfix. While yesterday's r240853 fixed
Adrian Prantl [Sat, 27 Jun 2015 20:12:43 +0000 (20:12 +0000)]
Debug Info: One more bitfield bugfix. While yesterday's r240853 fixed
the DW_AT_bit_offset computation, the byte offset is in fact also
endian-dependent as it needs to point to the storage unit containing the
most-significant bit of the the bitfield.
I'm so looking forward to emitting the endian-agnostic DWARF 3 version
instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240890 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] Fold duplicate big-endian disassembler tests together.
Daniel Sanders [Sat, 27 Jun 2015 17:56:44 +0000 (17:56 +0000)]
[mips] Fold duplicate big-endian disassembler tests together.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240887 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] Sort big-endian disassembler tests by opcode.
Daniel Sanders [Sat, 27 Jun 2015 16:13:59 +0000 (16:13 +0000)]
[mips] Sort big-endian disassembler tests by opcode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240885 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] Make little-endian disassembler test filenames consistent.
Daniel Sanders [Sat, 27 Jun 2015 15:42:25 +0000 (15:42 +0000)]
[mips] Make little-endian disassembler test filenames consistent.

Most are named *-el.txt. Renamed the three that were *-le.txt

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240884 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] Add COP0 register class and use it in M[FT]C0/DM[FT]C0.
Daniel Sanders [Sat, 27 Jun 2015 15:39:19 +0000 (15:39 +0000)]
[mips] Add COP0 register class and use it in M[FT]C0/DM[FT]C0.

Summary:
Previously it (incorrectly) used GPR's.

Patch by Simon Dardis. A couple small corrections by myself.

Reviewers: dsanders

Reviewed By: dsanders

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10567

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240883 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[OCaml] Bump ctypes dependency to 0.4.
Peter Zotov [Sat, 27 Jun 2015 14:32:30 +0000 (14:32 +0000)]
[OCaml] Bump ctypes dependency to 0.4.

ctypes 0.3 and earlier contains an interface-definig bug:
its ptr_of_raw_address accepts Int64 and not Nativeint. ctypes 0.4
was not released during the 3.6 cycle, and because of that, LLVM 3.6
was released with ctypes 0.3 as a dependency, which now breaks
the build on modern ctypes.

Unbreak.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240882 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[LoopVectorize] Pointer indicies may be wider than the pointer
David Majnemer [Sat, 27 Jun 2015 08:38:17 +0000 (08:38 +0000)]
[LoopVectorize] Pointer indicies may be wider than the pointer

If we are dealing with a pointer induction variable, isInductionPHI
gives back a step value of Stride / size of pointer.  However, we might
be indexing with a legal type wider than the pointer width.
Handle this by inserting casts where appropriate instead of crashing.

This fixes PR23954.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240877 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PruneEH] A naked, noinline function can return via InlineAsm
David Majnemer [Sat, 27 Jun 2015 07:52:53 +0000 (07:52 +0000)]
[PruneEH] A naked, noinline function can return via InlineAsm

The PruneEH pass tries to annotate functions as 'noreturn' if it doesn't
see a ReturnInst.  However, a naked function containing inline assembly
can contain control flow leaving the function.

This fixes PR23971.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240876 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Stackmap] Pre-assemble the stackmap parser test case. (Fix builders).
Lang Hames [Sat, 27 Jun 2015 03:49:25 +0000 (03:49 +0000)]
[Stackmap] Pre-assemble the stackmap parser test case. (Fix builders).

This case had been failing on testers that didn't have x86 support. Rather
than XFAIL it on testers without x86 support, I've just assembled it and used
the raw object as the test input.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240875 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agollvm/test/CodeGen/X86/xor.ll: Appease Win32 targets since r240796.
NAKAMURA Takumi [Sat, 27 Jun 2015 03:46:58 +0000 (03:46 +0000)]
llvm/test/CodeGen/X86/xor.ll: Appease Win32 targets since r240796.

  %struct.ref_s = type { %union.v, i16, i16 }
  %union.v = type { i64 }

It seems %struct.ref_s is incompatible in tail padding.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240874 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[MC] Ensure that pending labels are flushed when -mc-relax-all flag is used
Petr Hosek [Sat, 27 Jun 2015 01:54:17 +0000 (01:54 +0000)]
[MC] Ensure that pending labels are flushed when -mc-relax-all flag is used

Summary:
The current implementation doesn't always flush all pending labels
beforeemitting data which can result in an incorrectly placed labels in
case when when instruction bundling is enabled and -mc-relax-all flag is
being used. To address this issue, we always flush pending labels before
emitting data.

The change was tested by running PNaCl toolchain trybots with
-mc-relax-all flag set.

Fixes https://code.google.com/p/nativeclient/issues/detail?id=4063

Test Plan: Regression test attached

Reviewers: mseaborn

Subscribers: jfb, llvm-commits

Differential Revision: http://reviews.llvm.org/D10325

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240870 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[MC] Align fragments when -mc-relax-all flag is used
Petr Hosek [Sat, 27 Jun 2015 01:49:53 +0000 (01:49 +0000)]
[MC] Align fragments when -mc-relax-all flag is used

Summary:
Ensure that fragments are bundle aligned when instruction bundling
is enabled and the -mc-relax-all flag is set. This is implicitly
assumed by the bundle padding implementation but this assumption
does not hold when custom alignment is being used.

The change was tested by running PNaCl toolchain trybots with
-mc-relax-all flag set.

Fixes https://code.google.com/p/nativeclient/issues/detail?id=4063

Test Plan: Regression test attached

Reviewers: mseaborn

Subscribers: jfb, llvm-commits

Differential Revision: http://reviews.llvm.org/D10044

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240869 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAsmPrinter: Document why DIEValueList uses a linked-list, NFC
Duncan P. N. Exon Smith [Sat, 27 Jun 2015 01:19:17 +0000 (01:19 +0000)]
AsmPrinter: Document why DIEValueList uses a linked-list, NFC

There are two main reasons why a linked-list makes sense for
`DIEValueList`.

 1. We want `DIE` to be on a `BumpPtrAllocator` to improve teardown
    efficiency.  Making `DIEValueList` array-based would make that much
    more complicated.
 2. The singly-linked list is fairly memory efficient.  The histogram
    [1] shows that most DIEs have relatively few values, so we often pay
    less than the 2/3-pointer static overhead of a vector.  Furthermore,
    we don't know ahead of time exactly how many values a `DIE` needs,
    so a vector-like scheme will on average over-allocate by ~50%.  As
    it happens, that's the same memory overhead as the linked list node.

[1]: http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-May/085910.html

The comment I added to the code is a little more succinct, but I think
it's enough to give the idea.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240868 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoIR: Expose ModuleSlotTracker in Value::print()
Duncan P. N. Exon Smith [Sat, 27 Jun 2015 00:38:26 +0000 (00:38 +0000)]
IR: Expose ModuleSlotTracker in Value::print()

Allow callers of `Value::print()` and `Metadata::print()` to pass in a
`ModuleSlotTracker`.  This allows them to pay only once for calculating
module-level slots (such as Metadata).

This is related to PR23865, where there was a huge cost for
`MachineFunction::print()`.  Although I don't have a *particular* user
in mind for this new code, I have hit big slowdowns before when running
`opt -debug`, and I think this will be useful.  Going forward, if
someone hits a big slowdown with `print()` statements, they can create a
`ModuleSlotTracker` and send it through.  Similarly, adding support to
`Value::dump()` and `Metadata::dump()` should be trivial.

I added unit tests to be sure the `print()` functions actually behave
the same way with and without the slot tracker.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240867 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLowerBitSets: Ignore bitset entries that do not directly refer to a global.
Peter Collingbourne [Sat, 27 Jun 2015 00:17:51 +0000 (00:17 +0000)]
LowerBitSets: Ignore bitset entries that do not directly refer to a global.

It is possible for a global to be substituted with another global of a
different type or a different kind (i.e. an alias) at IR link time. One
example of this scenario is when a Microsoft ABI vtable is substituted with
an alias referring to a larger vtable containing an RTTI reference.

This will cause the global to be RAUW'd with a possibly bitcasted reference
to the other global. This will of course also affect any references to the
global in bitset metadata.

The right way to handle such metadata is simply to ignore it. This is sound
because the linked module should contain another copy of the bitset entries as
applied to the new global.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240866 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoPlug a leak introduced by r240848
Duncan P. N. Exon Smith [Sat, 27 Jun 2015 00:15:32 +0000 (00:15 +0000)]
Plug a leak introduced by r240848

Apparently this obvious leak was never exercised before, but r240848
exposed it.  Plug it.

http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/5075

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240865 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd original source code to test case as suggested in review.
Adrian Prantl [Sat, 27 Jun 2015 00:01:33 +0000 (00:01 +0000)]
Add original source code to test case as suggested in review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240863 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[StackMaps] Add a lightweight parser for stackmap version 1 sections.
Lang Hames [Fri, 26 Jun 2015 23:56:53 +0000 (23:56 +0000)]
[StackMaps] Add a lightweight parser for stackmap version 1 sections.

The parser provides a convenient interface for reading llvm stackmap v1 sections
in object files.

This patch also includes a new option for llvm-readobj, '-stackmap', which uses
the parser to pretty-print stackmap sections for debugging/testing purposes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240860 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDebug Info: Fix a bug in the DW_AT_bit_offset calculation that would
Adrian Prantl [Fri, 26 Jun 2015 23:31:27 +0000 (23:31 +0000)]
Debug Info: Fix a bug in the DW_AT_bit_offset calculation that would
result in negative offsets and attempt a better job at documenting
the algorithm.

rdar://21082998

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240853 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCodeGen: Create a proper ModuleSlotTracker for MachineInstr
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 23:18:44 +0000 (23:18 +0000)]
CodeGen: Create a proper ModuleSlotTracker for MachineInstr

Another follow-up related to r240848: try a little harder to share slot
tracking calculations within a single `MachineInstr` dump.  This is
unrelated to `MachineFunction::print()`, since that should be passing
through the function's `ModuleSlotTracker` by now, but could affect the
speed of dumping from a debugger if there is more than one IR-level
operand.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240852 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMIR Serialization: Serialize global address machine operands.
Alex Lorenz [Fri, 26 Jun 2015 22:56:48 +0000 (22:56 +0000)]
MIR Serialization: Serialize global address machine operands.

This commit serializes the global address machine operands.
This commit doesn't serialize the operand's offset and target
flags, it serializes only the global value reference.

Reviewers: Duncan P. N. Exon Smith

Differential Revision: http://reviews.llvm.org/D10671

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240851 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[RewriteStatepointsForGC] Generalized vector phi/select handling for base pointers
Philip Reames [Fri, 26 Jun 2015 22:47:37 +0000 (22:47 +0000)]
[RewriteStatepointsForGC] Generalized vector phi/select handling for base pointers

This change extends the detection of base pointers for vector constructs to handle arbitrary phi and select nodes. The existing non-vector code already handles those, so this is basically just extending the vector special case to be less special cased. It still isn't generalized vector handling since we can't handle arbitrary vector instructions (e.g. shufflevectors), but it's a lot closer.

The general structure of the change is as follows:
 * Extend the base defining value relation over a subset of vector instructions and vector typed phi & select instructions.
 * Move scalarization from before base pointer rewriting to after base pointer rewriting. The extension of the BDV relation is sufficient to find vector base phis for vector inputs.
 * Preserve the existing special case logic for when the base of a vector element is locally obvious. This general idea could be extended to the scalar case as well.

Differential Revision: http://reviews.llvm.org/D10461#inline-84275

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240850 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[NVPTX] noop when kernel pointers are already global
Jingyue Wu [Fri, 26 Jun 2015 22:35:43 +0000 (22:35 +0000)]
[NVPTX] noop when kernel pointers are already global

Summary:
Some front ends make kernel pointers global already. In that case,
handlePointerParams does nothing.

Test Plan: more tests in lower-kernel-ptr-arg.ll

Reviewers: grosser

Subscribers: jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D10779

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240849 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCodeGen: Push the ModuleSlotTracker through Metadata
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 22:28:47 +0000 (22:28 +0000)]
CodeGen: Push the ModuleSlotTracker through Metadata

For another 1% speedup on the testcase in PR23865, push the
`ModuleSlotTracker` through to metadata-related printing in
`MachineBasicBlock::print()`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240848 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMinor style cleanup after 240843 [NFC]
Philip Reames [Fri, 26 Jun 2015 22:21:52 +0000 (22:21 +0000)]
Minor style cleanup after 240843 [NFC]

Use a for-each loop in one case and rename the function to reflect it's new usage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240847 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCodeGen: Push the ModuleSlotTracker through MachineOperands
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 22:06:47 +0000 (22:06 +0000)]
CodeGen: Push the ModuleSlotTracker through MachineOperands

Push `ModuleSlotTracker` through `MachineOperand`s, dropping the time
for `llc -print-machineinstrs` on the testcase in PR23865 from ~13
seconds to ~9 seconds.  Now `SlotTracker::processFunctionMetadata()`
accounts for only 8% of the runtime, which seems reasonable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240845 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Verifier] Follow on to 240836
Philip Reames [Fri, 26 Jun 2015 22:04:34 +0000 (22:04 +0000)]
[Verifier] Follow on to 240836

Address one missed review comment and do the rename I left out of that patch to make it reviewable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240843 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCodeGen: Use a single SlotTracker in MachineFunction::print()
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 22:04:20 +0000 (22:04 +0000)]
CodeGen: Use a single SlotTracker in MachineFunction::print()

Expose enough of the IR-level `SlotTracker` so that
`MachineFunction::print()` can use a single one for printing
`BasicBlock`s.  Next step would be to lift this through a few more APIs
so that we can make other print methods faster.

Fixes PR23865, changing the runtime of `llc -print-machineinstrs` from
many minutes (killed after 3 minutes, but it wasn't very close) to
13 seconds for a 502185 line dump.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240842 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDPGU/SI: Use correct resource descriptors for VI on HSA
Tom Stellard [Fri, 26 Jun 2015 21:58:42 +0000 (21:58 +0000)]
AMDPGU/SI: Use correct resource descriptors for VI on HSA

Summary: We need to set MTYPE = 2 for VI shaders when targeting the HSA runtime.

Reviewers: arsenm

Differential Revision: http://reviews.llvm.org/D10777

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240841 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Update amd_kernel_code_t definition and add assembler support
Tom Stellard [Fri, 26 Jun 2015 21:58:31 +0000 (21:58 +0000)]
AMDGPU/SI: Update amd_kernel_code_t definition and add assembler support

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10772

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240839 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Remove unused variable
Tom Stellard [Fri, 26 Jun 2015 21:58:26 +0000 (21:58 +0000)]
AMDGPU/SI: Remove unused variable

This should fix some bots that were broken by r240831.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240838 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Verifier] Verify invokes of intrinsics
Philip Reames [Fri, 26 Jun 2015 21:39:44 +0000 (21:39 +0000)]
[Verifier] Verify invokes of intrinsics

We support invoking a subset of llvm's intrinsics, but the verifier didn't account for this.  We had previously added a special case to verify invokes of statepoints.  By generalizing the code in terms of CallSite, we can verify invokes of other intrinsics as well.  Interestingly, this found one test case which was invalid.

Note: I'm deliberately leaving the naming change from CI to CS to a follow up change.  That will happen shortly, I just wanted to reduce the diff to make it clear what was happening with this one.

Differential Revision: http://reviews.llvm.org/D10118

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240836 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDebug Info: Clarify the documentation for bitfields emission.
Adrian Prantl [Fri, 26 Jun 2015 21:27:30 +0000 (21:27 +0000)]
Debug Info: Clarify the documentation for bitfields emission.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240835 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDebug info: Add more test coverage for bitfields.
Adrian Prantl [Fri, 26 Jun 2015 21:27:16 +0000 (21:27 +0000)]
Debug info: Add more test coverage for bitfields.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240834 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Set ELF OS/ABI to ELFOSABI_AMDGPU_HSA
Tom Stellard [Fri, 26 Jun 2015 21:15:11 +0000 (21:15 +0000)]
AMDGPU/SI: Set ELF OS/ABI to ELFOSABI_AMDGPU_HSA

Reviewers: arsenm, rafael

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10708

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240832 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Add hsa code object directives
Tom Stellard [Fri, 26 Jun 2015 21:15:07 +0000 (21:15 +0000)]
AMDGPU/SI: Add hsa code object directives

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10757

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240831 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: There are no implicit kernel args in the amdhsa ABI
Tom Stellard [Fri, 26 Jun 2015 21:15:03 +0000 (21:15 +0000)]
AMDGPU/SI: There are no implicit kernel args in the amdhsa ABI

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10706

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240830 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Emit amd_kernel_code_t in EmitFunctionBodyStart()
Tom Stellard [Fri, 26 Jun 2015 21:14:58 +0000 (21:14 +0000)]
AMDGPU/SI: Emit amd_kernel_code_t in EmitFunctionBodyStart()

Summary:
This way the function symbol points to the start of amd_kernel_code_t
rather than the start of the function.

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10705

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240829 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoTeach InlineCost to account for a null check which can be folded away
Philip Reames [Fri, 26 Jun 2015 20:51:17 +0000 (20:51 +0000)]
Teach InlineCost to account for a null check which can be folded away

If we have a caller that knows a particular argument can never be null, we can exploit this fact while simplifying values in the inline cost analysis. This has the effect of reducing the cost for inlining when a null check is present in the callee, but the value is known non null in the caller. In particular, any dependent control flow can be discounted from the cost estimate.

Note that we use the parameter attributes at the call site to memoize the analysis within the caller's code.  The setting of this attribute is done in InstCombine, the inline cost analysis just consumes it.  This is intentional and important because we want the inline cost analysis results to be easily cachable themselves.  We're not currently doing so, but initial results on LTO indicate this will quickly become important.

Differential Revision: http://reviews.llvm.org/D9129

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240828 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDataLayout now returns a const ref to its member string representation
Mehdi Amini [Fri, 26 Jun 2015 20:44:16 +0000 (20:44 +0000)]
DataLayout now returns a const ref to its member string representation

There was no particular reason to return by value in the first place.

From: Mehdi Amini <mehdi.amini@apple.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240826 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU: really don't commute REV opcodes if the target variant doesn't exist
Marek Olsak [Fri, 26 Jun 2015 20:29:10 +0000 (20:29 +0000)]
AMDGPU: really don't commute REV opcodes if the target variant doesn't exist

If pseudoToMCOpcode failed, we would return the original opcode, so operands
would be swapped, but the instruction would remain the same.
It resulted in LSHLREV a, b ---> LSHLREV b, a.

This fixes Glamor text rendering and
piglit/arb_sample_shading-builtin-gl-sample-mask on VI.

This is a candidate for stable branches.

v2: the test was simplified by Tom Stellard

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240824 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd Value.def to the list of textual includes, excluding it from the modules build.
Benjamin Kramer [Fri, 26 Jun 2015 20:16:44 +0000 (20:16 +0000)]
Add Value.def to the list of textual includes, excluding it from the modules build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240823 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoConvert a bunch of loops to foreach. NFC.
Pete Cooper [Fri, 26 Jun 2015 19:37:02 +0000 (19:37 +0000)]
Convert a bunch of loops to foreach.  NFC.

This uses the new SDNode::op_values() iterator range committed in r240805.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240822 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd missing builtins to the PPC back end for ABI compliance (vol. 1)
Nemanja Ivanovic [Fri, 26 Jun 2015 19:26:53 +0000 (19:26 +0000)]
Add missing builtins to the PPC back end for ABI compliance (vol. 1)

This patch corresponds to review:
http://reviews.llvm.org/D10638

This is the back end portion of patch
http://reviews.llvm.org/D10637
It just adds the code gen and intrinsic functions necessary to support that patch to the back end.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240820 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoWrap assert loops in #ifndef NDEBUG
Pete Cooper [Fri, 26 Jun 2015 19:23:20 +0000 (19:23 +0000)]
Wrap assert loops in #ifndef NDEBUG

The body of the loops here only contained asserts.  This triggered an unused variable
warning on release builds and -Werror on the bots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240819 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoConvert a bunch of loops to foreach. NFC.
Pete Cooper [Fri, 26 Jun 2015 19:18:49 +0000 (19:18 +0000)]
Convert a bunch of loops to foreach.  NFC.

This uses the new SDNode::op_values() iterator range committed in r240805.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240817 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoConvert a bunch of loops to foreach. NFC.
Pete Cooper [Fri, 26 Jun 2015 19:08:33 +0000 (19:08 +0000)]
Convert a bunch of loops to foreach.  NFC.

This uses the new SDNode::op_values() iterator range committed in r240805.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240815 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMake header parse standalone. NFC.
Benjamin Kramer [Fri, 26 Jun 2015 19:04:11 +0000 (19:04 +0000)]
Make header parse standalone. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240814 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoShow invariant loads in MMO dumping
Matt Arsenault [Fri, 26 Jun 2015 19:00:11 +0000 (19:00 +0000)]
Show invariant loads in MMO dumping

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240813 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Revert r240762 "[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant""
David Majnemer [Fri, 26 Jun 2015 18:55:48 +0000 (18:55 +0000)]
Revert "Revert r240762 "[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant""

This reverts commit r240793 while fixing how we handle array constant
pool entries.

This fixes PR23966.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240811 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix error handling in getString and simplify callers.
Rafael Espindola [Fri, 26 Jun 2015 18:42:17 +0000 (18:42 +0000)]
Fix error handling in getString and simplify callers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240810 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoConvert a bunch of loops to foreach. NFC.
Pete Cooper [Fri, 26 Jun 2015 18:41:54 +0000 (18:41 +0000)]
Convert a bunch of loops to foreach.  NFC.

This uses the new SDNode::op_values() iterator range committed in r240805.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240809 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDelete dead code. NFC.
Rafael Espindola [Fri, 26 Jun 2015 18:32:53 +0000 (18:32 +0000)]
Delete dead code. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240807 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd op_values() to iterate over the SDValue operands of an SDNode.
Pete Cooper [Fri, 26 Jun 2015 18:17:36 +0000 (18:17 +0000)]
Add op_values() to iterate over the SDValue operands of an SDNode.

SDNode already had ops() which would iterate over the operands and return
SDUse*.  This version instead gets the SDValue's out of the SDUse's so that
we can use foreach in more places.

Reviewed by David Blaikie.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240805 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMove VectorUtils from Transforms to Analysis to correct layering violation
David Blaikie [Fri, 26 Jun 2015 18:02:52 +0000 (18:02 +0000)]
Move VectorUtils from Transforms to Analysis to correct layering violation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240804 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ARM] Cortex-R5 is not VFPOnlySP
Javed Absar [Fri, 26 Jun 2015 17:42:37 +0000 (17:42 +0000)]
[ARM] Cortex-R5 is not VFPOnlySP

This patch fixes the error in ARM.td which stated that Cortex-R5
floating point unit can do only single precision, when it can do double as well.

Reviewers: rengolin

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10769

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240799 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[LAA] Try to prove non-wrapping of pointers if SCEV cannot
Adam Nemet [Fri, 26 Jun 2015 17:25:43 +0000 (17:25 +0000)]
[LAA] Try to prove non-wrapping of pointers if SCEV cannot

Summary:
Scalar evolution does not propagate the non-wrapping flags to values
that are derived from a non-wrapping induction variable because
the non-wrapping property could be flow-sensitive.

This change is a first attempt to establish the non-wrapping property in
some simple cases.  The main idea is to look through the operations
defining the pointer.  As long as we arrive to a non-wrapping AddRec via
a small chain of non-wrapping instruction, the pointer should not wrap
either.

I believe that this essentially is what Andy described in
http://article.gmane.org/gmane.comp.compilers.llvm.cvs/220731 as the way
forward.

Reviewers: aschwaighofer, nadav, sanjoy, atrick

Reviewed By: atrick

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10472

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240798 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix unused variable from r240792.
Alex Lorenz [Fri, 26 Jun 2015 17:07:27 +0000 (17:07 +0000)]
Fix unused variable from r240792.

The variable 'I' wasn't used when assertions were disabled.
This commit ensures that 'I' is used outside of an assert.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240797 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[DAGCombine] Fix demanded bits computation for exact shifts.
Benjamin Kramer [Fri, 26 Jun 2015 16:59:31 +0000 (16:59 +0000)]
[DAGCombine] Fix demanded bits computation for exact shifts.

Fixes a miscompilation of MultiSource/Benchmarks/MallocBench/gs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240796 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[X86]: Correctly sign-extend 16-bit immediate in CALL instruction.
Douglas Katzman [Fri, 26 Jun 2015 16:58:59 +0000 (16:58 +0000)]
[X86]: Correctly sign-extend 16-bit immediate in CALL instruction.

Patch by Matthew Barney. Thanks!

Differential Revision: http://reviews.llvm.org/D9514

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240795 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix ODR violation waiting to happen by making static function definitions in VectorUt...
David Blaikie [Fri, 26 Jun 2015 16:57:30 +0000 (16:57 +0000)]
Fix ODR violation waiting to happen by making static function definitions in VectorUtils.h non-static and defined out of line

Patch by Ashutosh Nema

Differential Revision: http://reviews.llvm.org/D10682

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240794 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert r240762 "[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant"
Hans Wennborg [Fri, 26 Jun 2015 16:48:02 +0000 (16:48 +0000)]
Revert r240762 "[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant"

It seems to have caused PR23966: "UNREACHABLE executed at ..\lib\Target\X86\X86TargetObjectFile.cpp:148"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240793 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMIR Serialization: Serialize machine basic block operands.
Alex Lorenz [Fri, 26 Jun 2015 16:46:11 +0000 (16:46 +0000)]
MIR Serialization: Serialize machine basic block operands.

This commit serializes machine basic block operands. The
machine basic block operands use the following syntax:

  %bb.<id>[.<name>]

This commit also modifies the YAML representation for the
machine basic blocks - a new, required field 'id' is added
to the MBB YAML mapping.

The id is used to resolve the MBB references to the
actual MBBs. And while the name of the MBB can be
included in a MBB reference, this name isn't used to
resolve MBB references - as it's possible that multiple
MBBs will reference the same BB and thus they will have the
same name. If the name is specified, the parser will verify
that it is equal to the name of the MBB with the specified id.

Reviewers: Duncan P. N. Exon Smith

Differential Revision: http://reviews.llvm.org/D10608

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240792 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoELF: Simplify the rel/rela implementation.
Rafael Espindola [Fri, 26 Jun 2015 15:27:04 +0000 (15:27 +0000)]
ELF: Simplify the rel/rela implementation.

Now the rela class inherits from rel and just adds the addend.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240790 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[DAGCombiner] Preserve the exact bit when simplifying SRA to SRL.
Benjamin Kramer [Fri, 26 Jun 2015 14:51:49 +0000 (14:51 +0000)]
[DAGCombiner] Preserve the exact bit when simplifying SRA to SRL.

Allows more aggressive folding of ashr/shl pairs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240788 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[DAGCombine] fold (X >>?,exact C1) << C2 --> X << (C2-C1)
Benjamin Kramer [Fri, 26 Jun 2015 14:51:36 +0000 (14:51 +0000)]
[DAGCombine] fold (X >>?,exact C1) << C2 --> X << (C2-C1)

Instcombine also does this but many opportunities only become visible
after GEPs are lowered.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240787 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSilencing spurious MSVC C4189 warnings regarding local variables that are initialized...
Aaron Ballman [Fri, 26 Jun 2015 14:51:22 +0000 (14:51 +0000)]
Silencing spurious MSVC C4189 warnings regarding local variables that are initialized but not used; NFC. This bug has been reported to Microsoft (https://connect.microsoft.com/VisualStudio/feedback/details/1475983).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240786 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRename getObjectFile to getObject for consistency.
Rafael Espindola [Fri, 26 Jun 2015 14:51:16 +0000 (14:51 +0000)]
Rename getObjectFile to getObject for consistency.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240785 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSimplify isSymbolList64Bit. NFC.
Rafael Espindola [Fri, 26 Jun 2015 14:11:54 +0000 (14:11 +0000)]
Simplify isSymbolList64Bit. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240784 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSimplify isObject. NFC.
Rafael Espindola [Fri, 26 Jun 2015 13:24:23 +0000 (13:24 +0000)]
Simplify isObject. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240783 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] [IAS] Add partial support for the ULW pseudo-instruction.
Toma Tabacu [Fri, 26 Jun 2015 13:20:17 +0000 (13:20 +0000)]
[mips] [IAS] Add partial support for the ULW pseudo-instruction.

Summary:
This only adds support for ULW of an immediate address with/without a source register.
It does not include support for ULW of the address of a symbol.

Reviewers: dsanders

Reviewed By: dsanders

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D9663

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240782 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoImplement elf_section_iterator and getELFType().
Rafael Espindola [Fri, 26 Jun 2015 13:11:15 +0000 (13:11 +0000)]
Implement elf_section_iterator and getELFType().

And with those, simplify getSymbolNMTypeChar.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240780 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoExpose getFlags via ELFSectionRef.
Rafael Espindola [Fri, 26 Jun 2015 12:44:10 +0000 (12:44 +0000)]
Expose getFlags via ELFSectionRef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240779 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd a ELFSectionRef class and use it to expose getSectionType.
Rafael Espindola [Fri, 26 Jun 2015 12:33:37 +0000 (12:33 +0000)]
Add a ELFSectionRef class and use it to expose getSectionType.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240778 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSimplify getSymbolType.
Rafael Espindola [Fri, 26 Jun 2015 12:18:49 +0000 (12:18 +0000)]
Simplify getSymbolType.

This is still a really odd function. Most calls are in object format specific
contexts and should probably be replaced with a more direct query, but at least
now this is not too obnoxious to use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240777 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ARM] Cortex-R4F is not VFPOnlySP
Javed Absar [Fri, 26 Jun 2015 12:14:56 +0000 (12:14 +0000)]
[ARM] Cortex-R4F is not VFPOnlySP

Cortex-R4F TRM states that fpu supports both single and double precision.
This patch corrects the information in ARM.td file and corresponding test.

Reviewers: rengolin

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10763

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240776 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMake getOther ELF only.
Rafael Espindola [Fri, 26 Jun 2015 11:39:57 +0000 (11:39 +0000)]
Make getOther ELF only.

No other format has this field.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240774 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoOptimize the creation of mapping symbols.
Rafael Espindola [Fri, 26 Jun 2015 11:31:13 +0000 (11:31 +0000)]
Optimize the creation of mapping symbols.

No need to create two symbols just to assign one to the other.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240773 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDo not ignore projects/LLVMBuild.txt in git repo
Sergey Dmitrouk [Fri, 26 Jun 2015 10:13:56 +0000 (10:13 +0000)]
Do not ignore projects/LLVMBuild.txt in git repo

Without explicit exception for the path, it matches projects/* rule.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240771 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant
David Majnemer [Fri, 26 Jun 2015 07:03:12 +0000 (07:03 +0000)]
[X86] Cleanup X86WindowsTargetObjectFile::getSectionForConstant

No functionality changed, just keeping things clean.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240762 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InterleavedAccess] Fix failures "undefined type 'llvm::raw_ostream'" on windows.
Hao Liu [Fri, 26 Jun 2015 04:38:21 +0000 (04:38 +0000)]
[InterleavedAccess] Fix failures "undefined type 'llvm::raw_ostream'" on windows.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240760 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ARM] Lower interleaved memory accesses to vldN/vstN intrinsics.
Hao Liu [Fri, 26 Jun 2015 02:45:36 +0000 (02:45 +0000)]
[ARM] Lower interleaved memory accesses to vldN/vstN intrinsics.
This patch also adds a function to calculate the cost of interleaved memory accesses.

E.g. Lower an interleaved load:
        %wide.vec = load <8 x i32>, <8 x i32>* %ptr, align 4
        %v0 = shuffle %wide.vec, undef, <0, 2, 4, 6>
        %v1 = shuffle %wide.vec, undef, <1, 3, 5, 7>
     into:
        %vld2 = { <4 x i32>, <4 x i32> } call llvm.arm.neon.vld2(%ptr, 4)
        %vec0 = extractelement { <4 x i32>, <4 x i32> } %vld2, i32 0
        %vec1 = extractelement { <4 x i32>, <4 x i32> } %vld2, i32 1

E.g. Lower an interleaved store:
        %i.vec = shuffle <8 x i32> %v0, <8 x i32> %v1, <0, 4, 8, 1, 5, 9, 2, 6, 10, 3, 7, 11>
        store <12 x i32> %i.vec, <12 x i32>* %ptr, align 4
     into:
        %sub.v0 = shuffle <8 x i32> %v0, <8 x i32> v1, <0, 1, 2, 3>
        %sub.v1 = shuffle <8 x i32> %v0, <8 x i32> v1, <4, 5, 6, 7>
        %sub.v2 = shuffle <8 x i32> %v0, <8 x i32> v1, <8, 9, 10, 11>
        call void llvm.arm.neon.vst3(%ptr, %sub.v0, %sub.v1, %sub.v2, 4)

Differential Revision: http://reviews.llvm.org/D10533

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240755 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64] Lower interleaved memory accesses to ldN/stN intrinsics. This patch also...
Hao Liu [Fri, 26 Jun 2015 02:32:07 +0000 (02:32 +0000)]
[AArch64] Lower interleaved memory accesses to ldN/stN intrinsics. This patch also adds a function to calculate the cost of interleaved memory accesses.

E.g. Lower an interleaved load:
        %wide.vec = load <8 x i32>, <8 x i32>* %ptr
        %v0 = shuffle %wide.vec, undef, <0, 2, 4, 6>
        %v1 = shuffle %wide.vec, undef, <1, 3, 5, 7>
     into:
        %ld2 = { <4 x i32>, <4 x i32> } call llvm.aarch64.neon.ld2(%ptr)
        %vec0 = extractelement { <4 x i32>, <4 x i32> } %ld2, i32 0
        %vec1 = extractelement { <4 x i32>, <4 x i32> } %ld2, i32 1

E.g. Lower an interleaved store:
        %i.vec = shuffle <8 x i32> %v0, <8 x i32> %v1, <0, 4, 8, 1, 5, 9, 2, 6, 10, 3, 7, 11>
        store <12 x i32> %i.vec, <12 x i32>* %ptr
     into:
        %sub.v0 = shuffle <8 x i32> %v0, <8 x i32> v1, <0, 1, 2, 3>
        %sub.v1 = shuffle <8 x i32> %v0, <8 x i32> v1, <4, 5, 6, 7>
        %sub.v2 = shuffle <8 x i32> %v0, <8 x i32> v1, <8, 9, 10, 11>
        call void llvm.aarch64.neon.st3(%sub.v0, %sub.v1, %sub.v2, %ptr)

Differential Revision: http://reviews.llvm.org/D10533

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240754 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InterleavedAccess] Add a pass InterleavedAccess to identify interleaved memory acces...
Hao Liu [Fri, 26 Jun 2015 02:10:27 +0000 (02:10 +0000)]
[InterleavedAccess] Add a pass InterleavedAccess to identify interleaved memory accesses and transform into target specific intrinsics.

E.g. An interleaved load (Factor = 2):
        %wide.vec = load <8 x i32>, <8 x i32>* %ptr
        %v0 = shuffle <8 x i32> %wide.vec, <8 x i32> undef, <0, 2, 4, 6>
        %v1 = shuffle <8 x i32> %wide.vec, <8 x i32> undef, <1, 3, 5, 7>
It can be transformed into a ld2 intrinsic in AArch64 backend or a vld2 intrinsic in ARM backend.

E.g. An interleaved store (Factor = 3):
        %i.vec = shuffle <8 x i32> %v0, <8 x i32> %v1, <0, 4, 8, 1, 5, 9, 2, 6, 10, 3, 7, 11>
        store <12 x i32> %i.vec, <12 x i32>* %ptr
It can be transformed into a st3 intrinsic in AArch64 backend or a vst3 intrinsic in ARM backend.

Differential Revision: http://reviews.llvm.org/D10533

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240751 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAsmPrinter: More explicitly scope iterator for MSVC
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 00:53:44 +0000 (00:53 +0000)]
AsmPrinter: More explicitly scope iterator for MSVC

r240748 seems to be on the right path.  Be more explicit.

http://lab.llvm.org:8011/builders/clang-x64-ninja-win7/builds/1961/

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240750 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAsmPrinter: Explicitly scope iterator for MSVC
Duncan P. N. Exon Smith [Fri, 26 Jun 2015 00:41:53 +0000 (00:41 +0000)]
AsmPrinter: Explicitly scope iterator for MSVC

Try to placate bots by explicitly scoping a conversion constructor from
`iterator` to `const_iterator`.

http://lab.llvm.org:8011/builders/sanitizer-windows/builds/5931/

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240748 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "X86: Reject register operands with obvious type mismatches."
Matthias Braun [Fri, 26 Jun 2015 00:26:49 +0000 (00:26 +0000)]
Revert "X86: Reject register operands with obvious type mismatches."

Revert until http://llvm.org/PR23955 is investigated.

This reverts commit r239309.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240746 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix mismatched architectures in test
Matthias Braun [Fri, 26 Jun 2015 00:26:46 +0000 (00:26 +0000)]
Fix mismatched architectures in test

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240745 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoaad/fix labels in test/CodeGen/X86/StackColoring.ll
Matthias Braun [Fri, 26 Jun 2015 00:26:44 +0000 (00:26 +0000)]
aad/fix labels in test/CodeGen/X86/StackColoring.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240744 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ASan] Use llvm::getDISubprogram() to get function entry debug location.
Alexey Samsonov [Fri, 26 Jun 2015 00:00:47 +0000 (00:00 +0000)]
[ASan] Use llvm::getDISubprogram() to get function entry debug location.

It can be more robust than copying debug info from first non-alloca
instruction in the entry basic block. We use the same strategy in
coverage instrumentation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240738 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAsmPrinter: Use an intrusively linked list for DIE::Children
Duncan P. N. Exon Smith [Thu, 25 Jun 2015 23:52:10 +0000 (23:52 +0000)]
AsmPrinter: Use an intrusively linked list for DIE::Children

Replace the `std::vector<>` for `DIE::Children` with an intrusively
linked list.  This is a strict memory improvement: it requires no
auxiliary storage, and reduces `sizeof(DIE)` by one pointer.  It also
factors out the DIE-related malloc traffic.

This drops llc memory usage from 735 MB down to 718 MB, or ~2.3%.

(I'm looking at `llc` memory usage on `verify-uselistorder.lto.opt.bc`;
see r236629 for details.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240736 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAsmPrinter: Convert DIE::Values to a linked list
Duncan P. N. Exon Smith [Thu, 25 Jun 2015 23:46:41 +0000 (23:46 +0000)]
AsmPrinter: Convert DIE::Values to a linked list

Change `DIE::Values` to a singly linked list, where each node is
allocated on a `BumpPtrAllocator`.  In order to support `push_back()`,
the list is circular, and points at the tail element instead of the
head.  I abstracted the core list logic out to `IntrusiveBackList` so
that it can be reused for `DIE::Children`, which also cares about
`push_back()`.

This drops llc memory usage from 799 MB down to 735 MB, about 8%.

(I'm looking at `llc` memory usage on `verify-uselistorder.lto.opt.bc`;
see r236629 for details.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240733 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ELF] Move ELF{32,64}{L,B}E typedefs to llvm.
Michael J. Spencer [Thu, 25 Jun 2015 23:41:23 +0000 (23:41 +0000)]
[ELF] Move ELF{32,64}{L,B}E typedefs to llvm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240731 91177308-0d34-0410-b5e6-96231b3b80d8