oota-llvm.git
10 years agoKill the LLVM global lock.
Zachary Turner [Mon, 16 Jun 2014 22:40:42 +0000 (22:40 +0000)]
Kill the LLVM global lock.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211069 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove some code churn.
Zachary Turner [Mon, 16 Jun 2014 22:40:29 +0000 (22:40 +0000)]
Remove some code churn.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211068 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove some more code out into a separate CL.
Zachary Turner [Mon, 16 Jun 2014 22:40:17 +0000 (22:40 +0000)]
Remove some more code out into a separate CL.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211067 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUsers of the llvm global mutex must now acquire it manually.
Zachary Turner [Mon, 16 Jun 2014 22:39:38 +0000 (22:39 +0000)]
Users of the llvm global mutex must now acquire it manually.

This allows the mutex to be acquired in a guarded, RAII fashion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211066 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd load/store functionality
Reed Kotler [Mon, 16 Jun 2014 22:05:47 +0000 (22:05 +0000)]
Add load/store functionality

Summary:
This patches allows non conversions like i1=i2; where both are global ints.
In addition, arithmetic and other things start to work since fast-isel will use
existing patterns for non fast-isel from tablegen files where applicable.

In addition i8, i16 will work in this limited context for assignment without the need
for sign extension (zero or signed). It does not matter how i8 or i16 are loaded (zero or sign extended)
since only the 8 or 16 relevant bits are used and clang will ask for sign extension before using them in
arithmetic. This is all made more complete in forthcoming patches.

for example:
  int i, j=1, k=3;

  void foo() {
    i = j + k;
  }

Keep in mind that this pass is not enabled right now and is an experimental pass
It can only be enabled with a hidden option to llvm of -mips-fast-isel.

Test Plan: Run test-suite, loadstore2.ll and I will run some executable tests.

Reviewers: dsanders

Subscribers: mcrosier

Differential Revision: http://reviews.llvm.org/D3856

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211061 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64: Add backend intrinsic for rbit.
Jim Grosbach [Mon, 16 Jun 2014 21:55:35 +0000 (21:55 +0000)]
AArch64: Add backend intrinsic for rbit.

Define an intrinsic for the frontend to use and pattern match it to
the RBIT instruction.

rdar://9283021

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211058 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: intrinsic support for rbit.
Jim Grosbach [Mon, 16 Jun 2014 21:55:30 +0000 (21:55 +0000)]
ARM: intrinsic support for rbit.

We already have an ARMISD node. Create an intrinsic to map to it so we can
add support for the frontend __rbit() intrinsic.

rdar://9283021

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211057 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PPC64] Fix PR19893 - improve code generation for local function addresses
Bill Schmidt [Mon, 16 Jun 2014 21:36:02 +0000 (21:36 +0000)]
[PPC64] Fix PR19893 - improve code generation for local function addresses

Rafael opened http://llvm.org/bugs/show_bug.cgi?id=19893 to track non-optimal
code generation for forming a function address that is local to the compile
unit.  The existing code was treating both local and non-local functions
identically.

This patch fixes the problem by properly identifying local functions and
generating the proper addis/addi code.  I also noticed that Rafael's earlier
changes to correct the surrounding code in PPCISelLowering.cpp were also
needed for fast instruction selection in PPCFastISel.cpp, so this patch
fixes that code as well.

The existing test/CodeGen/PowerPC/func-addr.ll is modified to test the new
code generation.  I've added a -O0 run line to test the fast-isel code as
well.

Tested on powerpc64[le]-unknown-linux-gnu with no regressions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211056 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSince the DataLayout is always found off of the subtarget go ahead
Eric Christopher [Mon, 16 Jun 2014 21:18:27 +0000 (21:18 +0000)]
Since the DataLayout is always found off of the subtarget go ahead
and query the base target machine implementation for it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211055 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoClean up some unnecessary mutex guards.
Zachary Turner [Mon, 16 Jun 2014 20:54:28 +0000 (20:54 +0000)]
Clean up some unnecessary mutex guards.

These were being used as unreferenced parameters to enforce that
the methods must not be called without holding a mutex, but all
of the methods in question were internal, and the methods were
only exposed through an interface whose entire purpose was to
serialize access to these structures, so expecting the methods
to be accessed under a mutex is reasonable enough.

Reviewed by: blaikie

Differential Revision: http://reviews.llvm.org/D4162

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211054 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoImprove comments for r211040
Louis Gerbarg [Mon, 16 Jun 2014 20:31:50 +0000 (20:31 +0000)]
Improve comments for r211040

Added comment to clarify why we r211040 choose to bail out of fast isel instead
of generating a more complicated relocation, and fix mislabelled register in the
comments of the asan test case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211052 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert "lit: warn when passed invalid pathname" (r210597)
Hans Wennborg [Mon, 16 Jun 2014 20:18:41 +0000 (20:18 +0000)]
Revert "lit: warn when passed invalid pathname" (r210597)

It was pointed out that this breaks the "virtual test discovery"
mechanism, which allows for narming tests in the test exec root.

Reverting until I can figure out how to fix this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211048 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: implement correct atomic operations on v7M
Tim Northover [Mon, 16 Jun 2014 18:49:36 +0000 (18:49 +0000)]
ARM: implement correct atomic operations on v7M

ARM v7M has ldrex/strex but not ldrexd/strexd. This means 32-bit
operations should work as normal, but 64-bit ones are almost certainly
doomed.

Patch by Phoebe Buckheister.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211042 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix illegal relocations in X86FastISel
Louis Gerbarg [Mon, 16 Jun 2014 17:35:40 +0000 (17:35 +0000)]
Fix illegal relocations in X86FastISel

On x86_86  the lea instruction can only use a 32 bit immediate value. When
the code is compiled statically the RIP register is not used, meaning the
immediate is all that can be used for the relocation, which is not sufficient
in the case of targets more than +/- 2GB away. This patch bails out of fast
isel in those cases and reverts to DAG which does the right thing.

Test case included.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211040 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoLowerSwitch: track bounding range for the condition tree.
Jim Grosbach [Mon, 16 Jun 2014 16:55:20 +0000 (16:55 +0000)]
LowerSwitch: track bounding range for the condition tree.

When LowerSwitch transforms a switch instruction into a tree of ifs it
is actually performing a binary search into the various case ranges, to
see if the current value falls into one cases range of values.

So, if we have a program with something like this:

switch (a) {
case 0:
  do0();
  break;
case 1:
  do1();
  break;
case 2:
  do2();
  break;
default:
  break;
}

the code produced is something like this:

  if (a < 1) {
    if (a == 0) {
      do0();
    }
  } else {
    if (a < 2) {
      if (a == 1) {
        do1();
      }
    } else {
      if (a == 2) {
        do2();
      }
    }
  }

This code is inefficient because the check (a == 1) to execute do1() is
not needed.

The reason is that because we already checked that (a >= 1) initially by
checking that also  (a < 2) we basically already inferred that (a == 1)
without the need of an extra basic block spawned to check if actually (a
== 1).

The patch addresses this problem by keeping track of already
checked bounds in the LowerSwitch algorithm, so that when the time
arrives to produce a Leaf Block that checks the equality with the case
value / range the algorithm can decide if that block is really needed
depending on the already checked bounds .

For example, the above with "a = 1" would work like this:

the bounds start as LB: NONE , UB: NONE
as (a < 1) is emitted the bounds for the else path become LB: 1 UB:
NONE. This happens because by failing the test (a < 1) we know that the
value "a" cannot be smaller than 1 if we enter the else branch.
After the emitting the check (a < 2) the bounds in the if branch become
LB: 1 UB: 1. This is because by checking that "a" is smaller than 2 then
the upper bound becomes 2 - 1 = 1.

When it is time to emit the leaf block for "case 1:" we notice that 1
can be squeezed exactly in between the LB and UB, which means that if we
arrived to that block there is no need to emit a block that checks if (a
== 1).

Patch by: Marcello Maggioni <hayarms@gmail.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211038 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRefactor the disabling of Thumb-1 LDM/STM generation
James Molloy [Mon, 16 Jun 2014 16:42:53 +0000 (16:42 +0000)]
Refactor the disabling of Thumb-1 LDM/STM generation

Originally I switched the LD/ST optimizer off in TargetMachine as it was previously, but Eric has suggested he'd prefer that it be short-circuited in the pass itself.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211037 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix pr17056.
Rafael Espindola [Mon, 16 Jun 2014 16:41:00 +0000 (16:41 +0000)]
Fix pr17056.

This makes llvm-nm ignore members that are not sufficiently aligned for
lib/Object to handle.

These archives are invalid. GNU AR is able to handle this, but in general
just warns about broken archive members.

We should probably start warning too, but for now just make sure llvm-nm
exits with an 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211036 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert the Archive API to use ErrorOr.
Rafael Espindola [Mon, 16 Jun 2014 16:08:36 +0000 (16:08 +0000)]
Convert the Archive API to use ErrorOr.

Now that we have c++11, even things like ErrorOr<std::unique_ptr<...>> are
easy to use.

No intended functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211033 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64] Remove dead code.
Tilmann Scheller [Mon, 16 Jun 2014 15:15:41 +0000 (15:15 +0000)]
[AArch64] Remove dead code.

Both function declarations lack a callee and an implementation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211029 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoHook up vector int_ctlz for AVX512.
Cameron McInally [Mon, 16 Jun 2014 14:12:28 +0000 (14:12 +0000)]
Hook up vector int_ctlz for AVX512.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211024 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] ssnop is deprecated on MIPS32r6/MIPS64r6
Daniel Sanders [Mon, 16 Jun 2014 13:25:35 +0000 (13:25 +0000)]
[mips][mips64r6] ssnop is deprecated on MIPS32r6/MIPS64r6

Summary: Depends on D4120

Reviewers: jkolek, zoran.jovanovic, vmedic

Reviewed By: zoran.jovanovic, vmedic

Differential Revision: http://reviews.llvm.org/D4121

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211021 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] cl[oz], and dcl[oz] are re-encoded in MIPS32r6/MIPS64r6
Daniel Sanders [Mon, 16 Jun 2014 13:18:59 +0000 (13:18 +0000)]
[mips][mips64r6] cl[oz], and dcl[oz] are re-encoded in MIPS32r6/MIPS64r6

Summary:
There is no change to the restrictions, just the result register is stored
once in the encoding rather than twice. The rt field is zero in
MIPS32r6/MIPS64r6.

Depends on D4119

Reviewers: zoran.jovanovic, jkolek, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4120

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211019 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] ll, sc, lld, and scd are re-encoded on MIPS32r6/MIPS64r6.
Daniel Sanders [Mon, 16 Jun 2014 13:13:03 +0000 (13:13 +0000)]
[mips][mips64r6] ll, sc, lld, and scd are re-encoded on MIPS32r6/MIPS64r6.

Summary:
The linked-load, store-conditional operations have been re-encoded such
that have a 9-bit offset instead of the 16-bit offset they have prior to
MIPS32r6/MIPS64r6.

While implementing this, I noticed that the atomic load/store pseudos always
emit a sign extension using sll and sra. I have improved this to use seb/seh
when they are available (MIPS32r2/MIPS64r2 and above).

Depends on D4118

Reviewers: jkolek, zoran.jovanovic, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4119

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211018 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSupport/ConvertUTF: restore compatibility with MSVC, which only implements C89
Dmitri Gribenko [Mon, 16 Jun 2014 11:22:33 +0000 (11:22 +0000)]
Support/ConvertUTF: restore compatibility with MSVC, which only implements C89

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211016 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSupport/ConvertUTF: implement U+FFFD insertion according to the recommendation
Dmitri Gribenko [Mon, 16 Jun 2014 11:09:46 +0000 (11:09 +0000)]
Support/ConvertUTF: implement U+FFFD insertion according to the recommendation
given in the Unicode spec

That is, replace every maximal subpart of an ill-formed subsequence with one
U+FFFD.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211015 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64] Fix a fencepost error in lowering for llvm.aarch64.neon.uqshl.
James Molloy [Mon, 16 Jun 2014 10:39:21 +0000 (10:39 +0000)]
[AArch64] Fix a fencepost error in lowering for llvm.aarch64.neon.uqshl.

Patch by Jiangning Liu!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211014 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Merge most of the big/little endian checks in atomic.ll
Daniel Sanders [Mon, 16 Jun 2014 10:25:17 +0000 (10:25 +0000)]
[mips] Merge most of the big/little endian checks in atomic.ll

Summary:
There is very little difference between the big and little endian cases in
test/CodeGen/Mips/atomic.ll. Merge them together using multiple
FileCheck prefixes.

Depends on D4117

Reviewers: jkolek, zoran.jovanovic, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4118

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211013 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] [ls][wd]c2 were re-encoded with 11-bit signed immediates rather...
Daniel Sanders [Mon, 16 Jun 2014 10:00:45 +0000 (10:00 +0000)]
[mips][mips64r6] [ls][wd]c2 were re-encoded with 11-bit signed immediates rather than 16-bit in MIPS32r6/MIPS64r6

Summary:
The error message for the invalid.s cases isn't very helpful. It happens because
there is an instruction with a wider immediate that would have matched if the
NotMips32r6 predicate were true. I have some WIP to improve the message but it
affects most error messages for removed/re-encoded instructions on
MIPS32r6/MIPS64r6 and should therefore be a separate commit.

Depens on D4115

Reviewers: zoran.jovanovic, jkolek, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4117

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211012 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARMEB: Fix trunc store for vector types
Christian Pirker [Mon, 16 Jun 2014 09:17:30 +0000 (09:17 +0000)]
ARMEB: Fix trunc store for vector types

Reviewed at http://reviews.llvm.org/D4135

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211010 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCanonicalize addrspacecast ConstExpr between different pointer types
Jingyue Wu [Sun, 15 Jun 2014 21:40:57 +0000 (21:40 +0000)]
Canonicalize addrspacecast ConstExpr between different pointer types

As a follow-up to r210375 which canonicalizes addrspacecast
instructions, this patch canonicalizes addrspacecast constant
expressions.

Given clang uses ConstantExpr::getAddrSpaceCast to emit addrspacecast
cosntant expressions, this patch is also a step towards having the
frontend emit canonicalized addrspacecasts.

Piggyback a minor refactor in InstCombineCasts.cpp

Update three affected tests in addrspacecast-alias.ll,
access-non-generic.ll and constant-fold-gep.ll and added one new test in
constant-fold-address-space-pointer.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211004 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix copy paste error
Matt Arsenault [Sun, 15 Jun 2014 21:22:52 +0000 (21:22 +0000)]
Fix copy paste error

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211003 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Add a rotr testcase I forgot to add
Matt Arsenault [Sun, 15 Jun 2014 21:09:00 +0000 (21:09 +0000)]
R600: Add a rotr testcase I forgot to add

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211002 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Remove a few more things from AMDILISelLowering
Matt Arsenault [Sun, 15 Jun 2014 21:08:58 +0000 (21:08 +0000)]
R600: Remove a few more things from AMDILISelLowering

Try to keep all the setOperationActions for integer ops
together.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211001 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Fix assert on vector sdiv
Matt Arsenault [Sun, 15 Jun 2014 21:08:54 +0000 (21:08 +0000)]
R600: Fix assert on vector sdiv

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211000 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Move / cleanup more leftover AMDIL stuff.
Matt Arsenault [Sun, 15 Jun 2014 20:23:38 +0000 (20:23 +0000)]
R600: Move / cleanup more leftover AMDIL stuff.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210998 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Move division custom lowering out of AMDILISelLowering
Matt Arsenault [Sun, 15 Jun 2014 20:08:02 +0000 (20:08 +0000)]
R600: Move division custom lowering out of AMDILISelLowering

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210997 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTemporarily revert r210953 in an attempt to bring the ARM buildbots
Eric Christopher [Sun, 15 Jun 2014 19:55:14 +0000 (19:55 +0000)]
Temporarily revert r210953 in an attempt to bring the ARM buildbots
back.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210996 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Report that integer division is expensive.
Matt Arsenault [Sun, 15 Jun 2014 19:48:16 +0000 (19:48 +0000)]
R600: Report that integer division is expensive.

Divides by weird constants now emit much better code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210995 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Remove dead code
Matt Arsenault [Sun, 15 Jun 2014 19:48:13 +0000 (19:48 +0000)]
R600: Remove dead code

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210994 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPR20038: DebugInfo missing DIEs for some concrete variables.
David Blaikie [Sun, 15 Jun 2014 19:34:26 +0000 (19:34 +0000)]
PR20038: DebugInfo missing DIEs for some concrete variables.

I haven't nailed this down entirely, but this is about as small of a
test case as I can seem to construct and adequately demonstrates the
crasher. I'll continue investigating the root cause/fix(es).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210993 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd specialization of FoldingSetTrait for std::pair.
Manuel Klimek [Sun, 15 Jun 2014 14:42:25 +0000 (14:42 +0000)]
Add specialization of FoldingSetTrait for std::pair.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210990 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoLegalizeDAG: make sure cast is unsigned before using FP_TO_UINT.
Tim Northover [Sun, 15 Jun 2014 09:27:20 +0000 (09:27 +0000)]
LegalizeDAG: make sure cast is unsigned before using FP_TO_UINT.

It's valid to use FP_TO_SINT when asking for a smaller type (e.g. all
"unsigned int16" values fit into a "signed int32"), but the reverse
isn't true.

Unfortunately, I'm not actually aware of any architecture with
asymmetric FP_TO_SINT and FP_TO_UINT handling and the logic happens to
work in the symmetric case, so I can't actually write a test for this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210986 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64: improve handling & modelling of FP_TO_XINT nodes.
Tim Northover [Sun, 15 Jun 2014 09:27:15 +0000 (09:27 +0000)]
AArch64: improve handling & modelling of FP_TO_XINT nodes.

There's probably no acatual change in behaviour here, just updating
the LowerFP_TO_INT function to be more similar to the reverse
implementation and updating costs to current CodeGen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210985 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64: improve vector [su]itofp handling.
Tim Northover [Sun, 15 Jun 2014 09:27:06 +0000 (09:27 +0000)]
AArch64: improve vector [su]itofp handling.

This somehow got missed in the AArch64 merge, so should fix a
performance regression since 3.4.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210984 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't expect tests always crashing. Add "REQUIRES:asserts".
NAKAMURA Takumi [Sun, 15 Jun 2014 01:01:11 +0000 (01:01 +0000)]
Don't expect tests always crashing. Add "REQUIRES:asserts".

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210983 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReplacing the private implementations of SwapValue with calls to sys::swapByteOrder()
Artyom Skrobov [Sat, 14 Jun 2014 13:49:57 +0000 (13:49 +0000)]
Replacing the private implementations of SwapValue with calls to sys::swapByteOrder()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210980 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUsing llvm::sys::swapByteOrder() for the common case of byte-swapping a value in...
Artyom Skrobov [Sat, 14 Jun 2014 13:18:07 +0000 (13:18 +0000)]
Using llvm::sys::swapByteOrder() for the common case of byte-swapping a value in place

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210978 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdding llvm::sys::swapByteOrder() for the common use-case of byte-swapping a value...
Artyom Skrobov [Sat, 14 Jun 2014 12:52:55 +0000 (12:52 +0000)]
Adding llvm::sys::swapByteOrder() for the common use-case of byte-swapping a value in place

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210976 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRenaming SwapByteOrder() to getSwappedBytes()
Artyom Skrobov [Sat, 14 Jun 2014 11:36:01 +0000 (11:36 +0000)]
Renaming SwapByteOrder() to getSwappedBytes()

The next commit will add swapByteOrder(), acting in-place

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210973 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Add failing testcases.
Matt Arsenault [Sat, 14 Jun 2014 04:26:09 +0000 (04:26 +0000)]
R600: Add failing testcases.

These are reduced from assert in the
OpenCV CvtColor8u.BGR5652GRAY test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210969 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix typo
Matt Arsenault [Sat, 14 Jun 2014 04:26:07 +0000 (04:26 +0000)]
Fix typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210968 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Fix asserts related to constant initializers
Matt Arsenault [Sat, 14 Jun 2014 04:26:05 +0000 (04:26 +0000)]
R600: Fix asserts related to constant initializers

This would assert if a constant address space was extern
and therefore didn't have an initializer. If the initializer
was undef, it would hit the unreachable unhandled initializer case.

An extern global should never really occur since we don't have
machine linking, but bugpoint likes to remove initializers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210967 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Use address space enum instead of value
Matt Arsenault [Sat, 14 Jun 2014 04:26:01 +0000 (04:26 +0000)]
R600: Use address space enum instead of value

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210966 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove extra whitespace in function declaration. No functionality change.
Nick Lewycky [Sat, 14 Jun 2014 03:48:29 +0000 (03:48 +0000)]
Remove extra whitespace in function declaration. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210965 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Remove some extra handling of abstract variables and instead rely solely...
David Blaikie [Fri, 13 Jun 2014 23:52:55 +0000 (23:52 +0000)]
DebugInfo: Remove some extra handling of abstract variables and instead rely solely on the delayed handling introduced in r210946

Now that we handle finding abstract variables at the end of the module,
remove the upfront handling and just ensure the abstract variable is
built when necessary.

In theory we could have a split implementation, where inlined variables
are immediately constructed referencing the abstract definition, and
concrete variables are delayed - but let's go with one solution for now
unless there's a reason not to.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210961 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove InstrItineraryData off of the TargetMachine - it's already
Eric Christopher [Fri, 13 Jun 2014 23:11:13 +0000 (23:11 +0000)]
Remove InstrItineraryData off of the TargetMachine - it's already
on the subtarget and just forward the accessor.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210955 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove ARMJITInfo off of the TargetMachine and down onto the subtarget.
Eric Christopher [Fri, 13 Jun 2014 23:04:46 +0000 (23:04 +0000)]
Move ARMJITInfo off of the TargetMachine and down onto the subtarget.
This required untangling a mess of headers that included around.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210953 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoComplete remove test/Transforms/GlobalMerge.
Jiangning Liu [Fri, 13 Jun 2014 23:00:36 +0000 (23:00 +0000)]
Complete remove test/Transforms/GlobalMerge.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210952 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove GlobalMerge from Transform to CodeGen.
Jiangning Liu [Fri, 13 Jun 2014 22:57:59 +0000 (22:57 +0000)]
Move GlobalMerge from Transform to CodeGen.

This patch is to move GlobalMerge pass from Transform/Scalar
to CodeGen, because GlobalMerge depends on TargetMachine.
In the mean time, the macro INITIALIZE_TM_PASS is also moved
to CodeGen/Passes.h. With this fix we can avoid making
libScalarOpts depend on libCodeGen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210951 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoThe hazard recognizer only needs a subtarget, not a target machine
Eric Christopher [Fri, 13 Jun 2014 22:38:52 +0000 (22:38 +0000)]
The hazard recognizer only needs a subtarget, not a target machine
so make it take one. Fix up all users accordingly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210948 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix typo.
Eric Christopher [Fri, 13 Jun 2014 22:38:48 +0000 (22:38 +0000)]
Fix typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210947 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Reference abstract definitions from variables in concrete definitions...
David Blaikie [Fri, 13 Jun 2014 22:35:44 +0000 (22:35 +0000)]
DebugInfo: Reference abstract definitions from variables in concrete definitions that preceed their first inline definition.

Rather than relying on abstract variables looked up at the time the
concrete variable is created, look them up at the end of the module to
ensure they're referenced even if they're created after the concrete
definition. This completes/matches the work done in r209677 to handle
this for the subprograms themselves.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210946 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[DWARF parser] Use distinction between DW_AT_ranges_base and DW_AT_GNU_ranges_base...
Alexey Samsonov [Fri, 13 Jun 2014 22:31:03 +0000 (22:31 +0000)]
[DWARF parser] Use distinction between DW_AT_ranges_base and DW_AT_GNU_ranges_base instead of DWARF version

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210945 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDwarfDebug::getExistingAbstractVariable: constify an existing reference parameter...
David Blaikie [Fri, 13 Jun 2014 22:29:31 +0000 (22:29 +0000)]
DwarfDebug::getExistingAbstractVariable: constify an existing reference parameter that didn't need to be mutated.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210944 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Following up to r209677, refactor local variable emission to delay the...
David Blaikie [Fri, 13 Jun 2014 22:18:23 +0000 (22:18 +0000)]
DebugInfo: Following up to r209677, refactor local variable emission to delay the choice between emitting the definition attributes or using DW_AT_abstract_definition

This doesn't fix the abstract variable handling yet, but it introduces a
similar delay mechanism as was added for subprograms, causing
DW_AT_location to be reordered to the beginning of the attribute list
for local variables, and fixes all the test fallout for that.

A subsequent commit will remove the abstract variable handling in
DbgVariable and just do the abstract variable lookup at module end to
ensure that abstract variables introduced after their concrete
counterparts are appropriately referenced by the concrete variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210943 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Refactor some tests to allow DW_AT_name to not be the first attribute...
David Blaikie [Fri, 13 Jun 2014 21:52:33 +0000 (21:52 +0000)]
DebugInfo: Refactor some tests to allow DW_AT_name to not be the first attribute in a local variable.

In an effort to fix concrete variables referencing abstract origins
where the concrete variable preceeds the first inlined usage, the
addition of attributes such as name, file, etc will be delayed until the
end of the module (to wait to see if any inlined instances have
occurred, thus necessitating an abstract definition that the concrete
definition should also reference).

These test cases don't actually need to care about this ordering of
attributes, so update them to be more resilient to such changes coming
in the near future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210940 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agotest/DebugInfo/X86/dbg-value-isel.s: correct lexical block descriptor to match schema
David Blaikie [Fri, 13 Jun 2014 21:52:28 +0000 (21:52 +0000)]
test/DebugInfo/X86/dbg-value-isel.s: correct lexical block descriptor to match schema

This silently broke a long time ago when I unified some aspects of the
debug info schema. I'm just cleaning these up if/when they become a
problem.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210939 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake the error-handling functions thread-safe.
Zachary Turner [Fri, 13 Jun 2014 21:20:44 +0000 (21:20 +0000)]
Make the error-handling functions thread-safe.

Prior to this change, error handling functions must be installed
and removed only inside of an llvm_[start/stop]_multithreading
pair.  This change allows error handling functions to be installed
any time, and from any thread.

Reviewed by: chandlerc

Differential Revision: http://reviews.llvm.org/D4140

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210937 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove top-level Clang -fsanitize= flags for optional ASan features.
Alexey Samsonov [Fri, 13 Jun 2014 17:53:44 +0000 (17:53 +0000)]
Remove top-level Clang -fsanitize= flags for optional ASan features.

Init-order and use-after-return modes can currently be enabled
by runtime flags. use-after-scope mode is not really working at the
moment.

The only problem I see is that users won't be able to disable extra
instrumentation for init-order and use-after-scope by a top-level Clang flag.
But this instrumentation was implicitly enabled for quite a while and
we didn't hear from users hurt by it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210924 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86: lower ATOMIC_CMP_SWAP_WITH_SUCCESS directly
Tim Northover [Fri, 13 Jun 2014 17:29:39 +0000 (17:29 +0000)]
X86: lower ATOMIC_CMP_SWAP_WITH_SUCCESS directly

Lowering this new node allows us to fold the almost universal
comparison for success before it's even formed. Instead we can create
a copy from EFLAGS and an X86ISD::SETCC operation since all "cmpxchg"
instructions set the zero-flag to the correct value.

rdar://problem/13201607

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210923 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Cleanup some old AMDIL stuff.
Matt Arsenault [Fri, 13 Jun 2014 17:20:53 +0000 (17:20 +0000)]
R600: Cleanup some old AMDIL stuff.

Move / delete some of the more obviously wrong
setOperationAction calls. Most of these are setting Expand
for types that aren't legal which is the default anyway.
Leave stuff that might require more thought on whether it's
junk or not as it is.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210922 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFinishing touch for the std::error_code transition.
Rafael Espindola [Fri, 13 Jun 2014 17:20:48 +0000 (17:20 +0000)]
Finishing touch for the std::error_code transition.

While std::error_code itself seems to work OK in all platforms, there
are few annoying differences with regards to the std::errc enumeration.

This patch adds a simple llvm enumeration, which will hopefully avoid build
breakages in other platforms and surprises as we get more uses of
std::error_code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210920 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAtomics: make use of the "cmpxchg weak" instruction.
Tim Northover [Fri, 13 Jun 2014 16:45:52 +0000 (16:45 +0000)]
Atomics: make use of the "cmpxchg weak" instruction.

This also simplifies the IR we create slightly: instead of working out
where success & failure should go manually, it turns out we can just
always jump to a success/failure block created for the purpose. Later
phases will sort out the mess without much difficulty.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210917 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAtomics: switch direction of cmpxchg comparison
Tim Northover [Fri, 13 Jun 2014 16:45:36 +0000 (16:45 +0000)]
Atomics: switch direction of cmpxchg comparison

This has two benefits: it makes the result more suitable for direct
insertaion into the struct to emulate the new cmpxchg, and it means
the name we give the instruction matches its actual effect better.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210916 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Remove AMDIL instruction and register definitions
Tom Stellard [Fri, 13 Jun 2014 16:38:59 +0000 (16:38 +0000)]
R600: Remove AMDIL instruction and register definitions

Most of these are no longer used any more.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210915 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoopt: Initialize asm printers
Tobias Grosser [Fri, 13 Jun 2014 16:12:08 +0000 (16:12 +0000)]
opt: Initialize asm printers

Without initializing the assembly printers a shared library build of opt is
linked with these libraries whereas for a static build these libraries are dead
code eliminated. This is unfortunate for plugins in case they want to use them,
as they neither can rely on opt to provide this functionality nor can they link
the printers in themselves as this breaks with a shared object build of opt.

This patch calls InitializeAllAsmPrinters() from opt, which increases the static
binary size from 50MB -> 52MB on my system (all backends compiled) and causes no
measurable increase in the time needed to run 'make check'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210914 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove unused and odd code.
Rafael Espindola [Fri, 13 Jun 2014 15:36:17 +0000 (15:36 +0000)]
Remove unused and odd code.

This code was never being used and any use of it would look fairly strange.
For example, it would try to map a object_error::parse_failed to
std::errc::invalid_argument.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210912 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove broken include.
Rafael Espindola [Fri, 13 Jun 2014 15:21:50 +0000 (15:21 +0000)]
Remove broken include.

Looks like I got some git merge wrong.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210911 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix KillTheDoctor after r210725.
Rafael Espindola [Fri, 13 Jun 2014 15:01:11 +0000 (15:01 +0000)]
Fix KillTheDoctor after r210725.

We don't map these windows errors to generic ones since errc::timed_out is
not defined on mingw. Just use the raw windows error value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210910 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSCCP: update for cmpxchg returning { iN, i1 } now.
Tim Northover [Fri, 13 Jun 2014 14:54:09 +0000 (14:54 +0000)]
SCCP: update for cmpxchg returning { iN, i1 } now.

I accidentally missed this one since its use looked OK locally.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210909 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] Relocation R_MIPS_PC18_S3
Zoran Jovanovic [Fri, 13 Jun 2014 14:26:47 +0000 (14:26 +0000)]
[mips][mips64r6] Relocation R_MIPS_PC18_S3
Differential Revision: http://reviews.llvm.org/D3890

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210908 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDocs: remove extra {} around result types.
Tim Northover [Fri, 13 Jun 2014 14:24:23 +0000 (14:24 +0000)]
Docs: remove extra {} around result types.

It makes the types look like they're single-element structures. And
when we have instructions that *do* result in a struct, that can get
confusing rather quickly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210905 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDocs: fix grammar error in description
Tim Northover [Fri, 13 Jun 2014 14:24:16 +0000 (14:24 +0000)]
Docs: fix grammar error in description

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210904 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIR: add "cmpxchg weak" variant to support permitted failure.
Tim Northover [Fri, 13 Jun 2014 14:24:07 +0000 (14:24 +0000)]
IR: add "cmpxchg weak" variant to support permitted failure.

This commit adds a weak variant of the cmpxchg operation, as described
in C++11. A cmpxchg instruction with this modifier is permitted to
fail to store, even if the comparison indicated it should.

As a result, cmpxchg instructions must return a flag indicating
success in addition to their original iN value loaded. Thus, for
uniformity *all* cmpxchg instructions now return "{ iN, i1 }". The
second flag is 1 when the store succeeded.

At the DAG level, a new ATOMIC_CMP_SWAP_WITH_SUCCESS node has been
added as the natural representation for the new cmpxchg instructions.
It is a strong cmpxchg.

By default this gets Expanded to the existing ATOMIC_CMP_SWAP during
Legalization, so existing backends should see no change in behaviour.
If they wish to deal with the enhanced node instead, they can call
setOperationAction on it. Beware: as a node with 2 results, it cannot
be selected from TableGen.

Currently, no use is made of the extra information provided in this
patch. Test updates are almost entirely adapting the input IR to the
new scheme.

Summary for out of tree users:
------------------------------

+ Legacy Bitcode files are upgraded during read.
+ Legacy assembly IR files will be invalid.
+ Front-ends must adapt to different type for "cmpxchg".
+ Backends should be unaffected by default.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210903 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix bad copy-and-paste from r210652. AVX512 masked leading zero intrinsics.
Cameron McInally [Fri, 13 Jun 2014 13:20:01 +0000 (13:20 +0000)]
Fix bad copy-and-paste from r210652. AVX512 masked leading zero intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210901 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Add cache and pref instructions
Daniel Sanders [Fri, 13 Jun 2014 13:15:59 +0000 (13:15 +0000)]
[mips] Add cache and pref instructions

Summary:
cache and pref were added in MIPS-III, and MIPS32 but were re-encoded in
MIPS32r6/MIPS64r6 to use a 9-bit offset rather than the 16-bit offset
available to earlier cores.

Resolved the decoding conflict between pref and lwc3.

Depends on D4115

Reviewers: zoran.jovanovic, jkolek, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4116

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210900 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] bc1any[24] are not available on MIPS32r6/MIPS64r6
Daniel Sanders [Fri, 13 Jun 2014 13:08:38 +0000 (13:08 +0000)]
[mips][mips64r6] bc1any[24] are not available on MIPS32r6/MIPS64r6

Summary:
These MIPS-3D instructions have never been implemented in LLVM so we only
add testcases.

Reviewers: zoran.jovanovic, jkolek, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4115

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210899 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] b(ge|lt)zal are not available on MIPS32r6/MIPS64r6 and bal is a...
Daniel Sanders [Fri, 13 Jun 2014 13:02:52 +0000 (13:02 +0000)]
[mips][mips64r6] b(ge|lt)zal are not available on MIPS32r6/MIPS64r6 and bal is a normal instruction

Summary:
b(ge|lt)zal have been removed in MIPS32r6/MIPS64r6. However, bal (an alias
for 'bgezal $zero, $offset') still remains with the same encoding it had
prior to MIPS32r6/MIPS64r6.

Updated the MipsNaCLELFStreamer, and MipsLongBranch to correctly handle the
MIPS32r6/MIPS64r6 BAL instruction in addition to the existing BAL_BR pseudo.

No changes were required to the CodeGen test that looks for BAL
(test/CodeGen/Mips/longbranch.ll) since the new instruction has the same
syntax.

Depends on D4113

Reviewers: jkolek, zoran.jovanovic, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4114

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210898 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][mips64r6] daddi is not available on MIPS64r6
Daniel Sanders [Fri, 13 Jun 2014 12:49:06 +0000 (12:49 +0000)]
[mips][mips64r6] daddi is not available on MIPS64r6

Summary:
It's not emitted by the code generator so we only need assembler tests.

Also added missing daddi aliases from dsub mnemonics, and removed a couple
duplicate dsub tests.

Depends on D4112

Reviewers: zoran.jovanovic, jkolek, vmedic

Reviewed By: vmedic

Differential Revision: http://reviews.llvm.org/D4113

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210897 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Win32] Let utils/not aware of abort(), aka llvm_unreachable(), in msvcrt.
NAKAMURA Takumi [Fri, 13 Jun 2014 12:23:56 +0000 (12:23 +0000)]
[Win32] Let utils/not aware of abort(), aka llvm_unreachable(), in msvcrt.

It has exit code as 3. abort(), aka unreachable, may be handled as crash.

FIXME: Could we move this into Win32/Program.inc?

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210895 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm/test/CodeGen/X86/fast-isel-args-fail2.ll: Don't expect to fail with -Asserts...
NAKAMURA Takumi [Fri, 13 Jun 2014 12:05:06 +0000 (12:05 +0000)]
llvm/test/CodeGen/X86/fast-isel-args-fail2.ll: Don't expect to fail with -Asserts. It might or might not crash.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210894 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd HasCDI predicate to AVX512 VPBROADCASTM*.
Cameron McInally [Fri, 13 Jun 2014 11:40:31 +0000 (11:40 +0000)]
Add HasCDI predicate to AVX512 VPBROADCASTM*.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210892 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCPP backend: set volatile property on atomic instructions.
Tim Northover [Fri, 13 Jun 2014 09:14:50 +0000 (09:14 +0000)]
CPP backend: set volatile property on atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210890 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: Fix fastcc calling convention for Thumb1
Oliver Stannard [Fri, 13 Jun 2014 08:33:03 +0000 (08:33 +0000)]
ARM: Fix fastcc calling convention for Thumb1

When targetting Thumb1 on a processor which has a VFP unit (which
is not accessible from Thumb1), we were converting the fastcc calling
convention to AAPCS-VFP, which is not possible.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210889 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Don't call setOperationAction with things that aren't opcodes.
Matt Arsenault [Fri, 13 Jun 2014 07:44:38 +0000 (07:44 +0000)]
R600: Don't call setOperationAction with things that aren't opcodes.

CondCode actions are set with setCondCodeAction.
This should have been a harmless bug since the values seem to only
collide only with nodes that don't need to be handled, and these are
already correctly setup elsewhere.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210888 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Fix selection error on i64 rotl / rotr.
Matt Arsenault [Fri, 13 Jun 2014 04:00:30 +0000 (04:00 +0000)]
R600/SI: Fix selection error on i64 rotl / rotr.

Evergreen is still broken due to missing shl_parts.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210885 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove the last uses of 'using std::error_code'
Rafael Espindola [Fri, 13 Jun 2014 03:20:08 +0000 (03:20 +0000)]
Remove the last uses of 'using std::error_code'

This finishes the transition to std::error_code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210877 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove 'using std::error_code' from tools.
Rafael Espindola [Fri, 13 Jun 2014 03:07:50 +0000 (03:07 +0000)]
Remove 'using std::error_code' from tools.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210876 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix build on windows.
Rafael Espindola [Fri, 13 Jun 2014 02:36:09 +0000 (02:36 +0000)]
Fix build on windows.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210873 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove 'using std::errro_code' from lib.
Rafael Espindola [Fri, 13 Jun 2014 02:24:39 +0000 (02:24 +0000)]
Remove 'using std::errro_code' from lib.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210871 91177308-0d34-0410-b5e6-96231b3b80d8