FreshBSD

[mips][msa] Fix suboptimal FrameIndex lowering for ld.[hwd] and st.[hwd]

Summary:
The immediate in these instructions is scaled before use as an offset.
They therefore have a wider reach than ld.b/st.b.

Reviewers: matheusalmeida

Reviewed By: matheusalmeida

Differential Revision: http://llvm-reviews.chandlerc.com/D2338

Remove stray parts of (reverted on trunk) r196205 that are causing test
failures.

Method parseSetAssignment treats every operand with '$' sign as register and the parsing 
is directed to set alias for register. This will result in errors reported when 
expressions containing label references are parsed(for example long jumps)
As we can't make a complete solution now it has been decided to enable .set directive to 
handle long jump expressions. This will cause parser to report errors when parsing integer 
based register assignments, for example:
   .set r3, will be reported as error. Still, the need for expressions is higher priority 
as the integer based register assignments are Mips specific and can be avoided using 
register names.

Merge rest of r196210. Some bits strayed into r196701, turning 3.4 red. This
should fix the issue.
------------------------------------------------------------------------
r196210 | haoliu | 2013-12-03 06:06:55 +0000 (Tue, 03 Dec 2013) | 3 lines

[AArch64]Add missing floating point convert, round and misc intrinsics.
E.g. int64x1_t vcvt_s64_f64(float64x1_t a) -> FCVTZS Dd, Dn

------------------------------------------------------------------------

[-cxx-abi microsoft] Mangle large integral constants correctly

Testing has revealed that large integral constants (i.e. > INT64_MAX)
are always mangled as-if they are negative, even in places where it
would not make sense for them to be negative (like non-type template
parameters of type unsigned long long).

To address this, we change the way we model number mangling: always
mangle as-if our number is an int64_t.  This should result in correct
results when we have large unsigned numbers.

N.B.  Bizarrely, things that are 32-bit displacements like vbptr offsets
are mangled as-if they are unsigned 32-bit numbers.  This is a pretty
egregious waste of space, it would be a 4x savings if we could mangle it
like a signed 32-bit number.  Instead, we explicitly cast these
displacements to uint32_t and let the mangler proceed.

Merge r196725 (conflicts on same API as before):
------------------------------------------------------------------------
r196725 | tnorthover | 2013-12-08 15:56:50 +0000 (Sun, 08 Dec 2013) |
19 lines

ARM: fix folding of stack-adjustment (yet again).

When trying to eliminate an "sub sp, sp, #N" instruction by folding
it into an existing push/pop using dummy registers, we need to account
for the fact that this might affect precisely how "fp" gets set in the
prologue.

We were attempting this, but assuming that *whenever* we performed a
fold it would make a difference. This is false, for example, in:
    push {r4, r7, lr}
    add fp, sp, #4
    vpush {d8}
    sub sp, sp, #8

we can fold the "sub" into the "vpush", forming "vpush {d7, d8}".
However, in that case the "add fp" instruction mustn't change, which
we were getting wrong before.

Should fix PR18160.
------------------------------------------------------------------------

Merging r196751:
------------------------------------------------------------------------
r196751 | venkatra | 2013-12-08 20:02:15 -0800 (Sun, 08 Dec 2013) | 3 lines

[Sparc]: Implement getSetCCResultType() in SparcTargetLowering so that umulo/smulo can be 
lowered on sparcv9 without an assertion error.


------------------------------------------------------------------------

Versions of GCC newer than 4.8 exist these days =)

Merging r196755:
------------------------------------------------------------------------
r196755 | venkatra | 2013-12-08 21:13:25 -0800 (Sun, 08 Dec 2013) | 2 lines

[SPARCV9]: Adjust the resultant pointer of DYNAMIC_STACKALLOC with the stack BIAS on 
sparcV9.

------------------------------------------------------------------------

Minor cleanup and reorg of C++ status page.

Merging r195494:
------------------------------------------------------------------------
r195494 | mren | 2013-11-22 11:41:59 -0800 (Fri, 22 Nov 2013) | 4 lines

Debug Info: add a constant for debug info version number.

This will be used to output the debug info version number as a module flag.

------------------------------------------------------------------------

Merging r196157:
------------------------------------------------------------------------
r196157 | mren | 2013-12-02 13:26:02 -0800 (Mon, 02 Dec 2013) | 4 lines

Debug info metadata without a version or with an outdated version will be
dropped. Add a version here to avoid that.


------------------------------------------------------------------------

Update compatibility page to list some GCC language extensions that Clang does
not support as a possible reason for choosing GCC instead of Clang (and vice
versa). Weaken some of the claimed advantages of Clang in light of GCC
improvements.

Avoid extra error messages if method definition is inside function.

[SPARCV9]: Adjust the resultant pointer of DYNAMIC_STACKALLOC with the stack BIAS on 
sparcV9.

Avoid adding some decls to DeferredDeclsToEmit.

Before this patch GetOrCreateLLVMFunction would add a decl to
DeferredDeclsToEmit even when it was being called by the function trying to
emit that decl.

[-cxx-abi microsoft] Properly mangle enums

While testing our ability to mangle large constants (PR18175), I
incidentally discovered that we did not properly mangle enums correctly.

Previously, we would append the width of the enum in bytes after the
type-tag differentiator.

This would mean "enum : short" would be mangled as 'W2' while "enum :
char" would be mangled as 'W1'.  Upon testing this with several versions
of MSVC, I found that this did not match their behavior: they always use
'W4'.

N.B.  Quick testing uncovered that undname allows different numbers to
follow the 'W' in the following way:

'W0' -> "enum char"
'W1' -> "enum unsigned char"
'W2' -> "enum short"
'W3' -> "enum unsigned short"
'W4' -> "enum"
'W5' -> "enum unsigned int"
'W6' -> "enum long"
'W7' -> "enum unsigned long"


    [2 lines not shown]

[Sparc]: Implement getSetCCResultType() in SparcTargetLowering so that umulo/smulo can be 
lowered on sparcv9 without an assertion error.

[AArch64]Add missing pair intrinsics such as:
    int32_t vminv_s32(int32x2_t a) 
which should be compiled into SMINP Vd.2S,Vn.2S,Vm.2S

[AArch64]Add missing pair intrinsics such as:
    int32_t vminv_s32(int32x2_t a)
which should be compiled into SMINP Vd.2S,Vn.2S,Vm.2S

[AArch64]Pattern match failures for truncate store and extend load

Merging r196720:
------------------------------------------------------------------------
r196720 | joerg | 2013-12-08 05:54:58 -0800 (Sun, 08 Dec 2013) | 3 lines

Extend assembler handling for NetBSD/MIPS to pass down the correct ABI,
architecture and PIC flag.

------------------------------------------------------------------------

Merging r196724:
------------------------------------------------------------------------
r196724 | tnorthover | 2013-12-08 07:24:55 -0800 (Sun, 08 Dec 2013) | 5 lines

ARM: teach Sema that "r" can match 64-bit values

We already support using "r" on 64-bit values (a GPRPair is
allocated), but Sema doesn't know this yet so issues a warning. This
should fix it.
------------------------------------------------------------------------

Merging r196712:
------------------------------------------------------------------------
r196712 | rafael | 2013-12-07 17:13:22 -0800 (Sat, 07 Dec 2013) | 12 lines

Fix pr18174.

Clang outputs LLVM one top level decl at a time. This combined with the
visibility computation code looking for the newest NamespaceDecl would cause
it to produce different results for nested namespaces.

The two options for producing consistent results are
* Delay codegen of anything inside a namespace until the end of the file.
* Don't look for the newest NamespaceDecl.

This patch implements the second option.
This matches the gcc behavior too.
------------------------------------------------------------------------

Merging r196735:
------------------------------------------------------------------------
r196735 | venkatra | 2013-12-08 14:06:07 -0800 (Sun, 08 Dec 2013) | 3 lines

[SparcV9]: Expand MULHU/MULHS:i64 and UMUL_LOHI/SMUL_LOHI:i64 on sparcv9.
  This fixes PR18150.

------------------------------------------------------------------------

Fix Debian GNU/kFreeBSD build

Use the same LaunchProcessPosixSpawn as on FreeBSD and Linux.

Rename a variable that I missed in the previous refactoring.

Eliminate the last remaining header NDEBUG

VerifyDiagnosticConsumer is long-lived so the two additional members shouldn't
have any impact on release builds.

The clang headers are now free of NDEBUG conditionals. Let's keep it that way!

Note that they're not yet structurally stable, pending a few fixes in the LLVM
core headers.

Remove an old stdio.h include from the invalid-array test

This should get it up and running on win and other builders without system
headers.

Fix three tests that weren't checking anything

Add -verify and update the test directives to match current expectations.

Also add a FIXME to an ObjC test that has expected-* directives but no -verify.

Re-enable verification of test and update outdated diag checks

Going by PR6913 it looks like this one can no longer reach CodeGen so remove
the redundant -emit-llvm case and treat it as an ordinary Sema test.

[SparcV9]: Expand MULHU/MULHS:i64 and UMUL_LOHI/SMUL_LOHI:i64 on sparcv9.
  This fixes PR18150.

Re-enable ms inline asm parser test.

LLVM r196044 should make it pass.

Remove all DISABLE lines from tests

There's no evidence that a 'DISABLE' directive ever existed.

Let's see if anything breaks..

Revert 196544 due to internal bot failures.

Make sure we mark these registers as defined. Previously was done
in the td file.

There's no such thing as %clang_cc1_only

These were being substituted into approximately the following:

  clang -cc1 -internal-isystem ../lib/clang/3.5/include_only

Fix a test that hasn't worked since 2007

Due to a missing -verify, 2007-10-01-BuildArrayRef.c was a no-op.

The message was changed 5 years ago so also update the test to reflect the new wording.

Remove duplicated -cc1 in tests

GlobalAlias::isDeclaration is always false. Remove dead code.

Cleaning up of prologue/epilogue code for Mips16. First step
here is to make save/restore into variable number of argument instructions.

ARM: fix folding of stack-adjustment (yet again).

When trying to eliminate an "sub sp, sp, #N" instruction by folding
it into an existing push/pop using dummy registers, we need to account
for the fact that this might affect precisely how "fp" gets set in the
prologue.

We were attempting this, but assuming that *whenever* we performed a
fold it would make a difference. This is false, for example, in:
    push {r4, r7, lr}
    add fp, sp, #4
    vpush {d8}
    sub sp, sp, #8

we can fold the "sub" into the "vpush", forming "vpush {d7, d8}".
However, in that case the "add fp" instruction mustn't change, which
we were getting wrong before.

Should fix PR18160.

ARM: teach Sema that "r" can match 64-bit values

We already support using "r" on 64-bit values (a GPRPair is
allocated), but Sema doesn't know this yet so issues a warning. This
should fix it.

Move a generic lambda test into the more logical test file.

Fix the message to go along with the assertion that was just fixed.

argh!

Fix an assertion introduced by my previous refactoring.

Add back the test that was triggering the assertion (which I removed mistakenly thinking 
it was triggering just a warning and not an assertion).  My error was brought to my 
attention by Rafael (Thanks!).

Extend assembler handling for NetBSD/MIPS to pass down the correct ABI,
architecture and PIC flag.

Fixed CRLF

Ensure bitcode encoding of visibility styles stays stable. Patch by Boaz Ouriel.

Merging r196493. Simple conflict due to change API of updated
function.