------------------------------------------------------------------------
r276389 | majnemer | 2016-07-21 21:54:44 -0700 (Thu, 21 Jul 2016) | 6 lines

Don't remove side effecting instructions due to ConstantFoldInstruction

Just because we can constant fold the result of an instruction does not
imply that we can delete the instruction.  It may have side effects.

This fixes PR28655.
------------------------------------------------------------------------

llvm-svn: 276660

------------------------------------------------------------------------
r276361 | wolfgangp | 2016-07-21 16:28:18 -0700 (Thu, 21 Jul 2016) | 5 lines

Reverting r275115 which caused PR28634.
When empty (forwarding) basic blocks that are referenced by user labels
are removed, incorrect code may be generated.


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

llvm-svn: 276656

------------------------------------------------------------------------
r276479 | asbirlea | 2016-07-22 15:02:19 -0700 (Fri, 22 Jul 2016) | 11 lines

Add flag to PassManagerBuilder to disable GVN Hoist Pass.

Summary:
Adding a flag to diable GVN Hoisting by default.
Note: The GVN Hoist Pass causes some Halide tests to hang. Halide will disable the pass while investigating.

Reviewers: llvm-commits, chandlerc, spop, dberlin

Subscribers: mehdi_amini

Differential Revision: https://reviews.llvm.org/D22639
------------------------------------------------------------------------

llvm-svn: 276647

------------------------------------------------------------------------
r275967 | dsanders | 2016-07-19 03:49:03 -0700 (Tue, 19 Jul 2016) | 16 lines

[mips] Correct label prefixes for N32 and N64.

Summary:
N32 and N64 follow the standard ELF conventions (.L) whereas O32 uses its own
($).

This fixes the majority of object differences between -fintegrated-as and
-fno-integrated-as.

Reviewers: sdardis

Subscribers: dsanders, sdardis, llvm-commits

Differential Revision: https://reviews.llvm.org/D22412


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

llvm-svn: 276643

------------------------------------------------------------------------
r275968 | dsanders | 2016-07-19 11:58:06 +0100 (Tue, 19 Jul 2016) | 7 lines

[mips][ias] R_MIPS_GOT_(PAGE|OFST) do not need symbols

Reviewers: sdardis

Subscribers: dsanders, llvm-commits, sdardis

Differential Revision: https://reviews.llvm.org/D22458
------------------------------------------------------------------------

llvm-svn: 276562

------------------------------------------------------------------------
r275967 | dsanders | 2016-07-19 11:49:03 +0100 (Tue, 19 Jul 2016) | 16 lines

[mips] Correct label prefixes for N32 and N64.

Summary:
N32 and N64 follow the standard ELF conventions (.L) whereas O32 uses its own
($).

This fixes the majority of object differences between -fintegrated-as and
-fno-integrated-as.

Reviewers: sdardis

Subscribers: dsanders, sdardis, llvm-commits

Differential Revision: https://reviews.llvm.org/D22412


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

llvm-svn: 276561

------------------------------------------------------------------------
r275966 | dsanders | 2016-07-19 11:22:19 +0100 (Tue, 19 Jul 2016) | 11 lines

[mips] Recognise the triple used by Debian stretch for mips64el.

Summary:
The triple used for this distribution is mips64el-linux-gnuabi64.

Reviewers: sdardis

Subscribers: sdardis, llvm-commits

Differential Revision: https://reviews.llvm.org/D22406

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

llvm-svn: 276560

------------------------------------------------------------------------
r276232 | gbiv | 2016-07-20 20:28:13 -0700 (Wed, 20 Jul 2016) | 13 lines

[Sema] Fix PR28623.

In atomic builtins, we assumed that the LValue conversion on the first
argument would succeed. So, we would crash given code like:

```
void ovl(char);
void ovl(int);
__atomic_store_n(ovl, 0, 0);
```

This patch makes us not assume that said conversion is successful. :)

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

llvm-svn: 276489

------------------------------------------------------------------------
r276215 | ericwf | 2016-07-20 17:56:42 -0600 (Wed, 20 Jul 2016) | 9 lines

[libunwind] Properly align _Unwind_Exception.

Summary: _Unwind_Exception is required to be double word aligned. Currently the struct is under aligned.

Reviewers: mclow.lists, compnerd, kledzik, emaste

Subscribers: emaste, cfe-commits

Differential Revision: https://reviews.llvm.org/D22543
------------------------------------------------------------------------

llvm-svn: 276462

llvm-svn: 276424

llvm-svn: 276421

------------------------------------------------------------------------
r276358 | spop | 2016-07-21 16:22:10 -0700 (Thu, 21 Jul 2016) | 6 lines

GVH-hoist: only clone GEPs (PR28606)

Do not clone stored values unless they are GEPs that are special cased to avoid
hoisting them without hoisting their associated ld/st.

Differential revision: https://reviews.llvm.org/D22652
------------------------------------------------------------------------

------------------------------------------------------------------------
r276364 | spop | 2016-07-21 16:32:39 -0700 (Thu, 21 Jul 2016) | 1 line

GVN-hoist: add missing check for all GEP operands available
------------------------------------------------------------------------

------------------------------------------------------------------------
r276368 | spop | 2016-07-21 17:07:01 -0700 (Thu, 21 Jul 2016) | 1 line

GVN-hoist: move check before mutating the IR
------------------------------------------------------------------------

llvm-svn: 276420

------------------------------------------------------------------------
r276181 | majnemer | 2016-07-20 14:05:01 -0700 (Wed, 20 Jul 2016) | 6 lines

[GVNHoist] Don't hoist PHI nodes

We hoisted PHIs without respecting their special insertion point in the
block, leading to verfier errors.

This fixes PR28626.
------------------------------------------------------------------------

llvm-svn: 276418

The now-removed assertion was really more for initial debugging; it's
perfectly valid (albeit relatively rare) for `Q.Visited.size()` to be
greater than 1,000. A similar patch hasn't been applied to trunk
because the piece of code this assertion was a part of no longer exists
in trunk.

Fix okayed by Danny, merge to 3.9 okayed by Hans.

llvm-svn: 276337

------------------------------------------------------------------------
r276003 | ericwf | 2016-07-19 11:56:20 -0600 (Tue, 19 Jul 2016) | 35 lines

Fix undefined behavior in __tree

Summary:
This patch attempts to fix the undefined behavior in __tree by changing the node pointer types used throughout. The pointer types are changed for raw pointers in the current ABI and for fancy pointers in ABI V2 (since the fancy pointer types may not be ABI compatible).

The UB in `__tree` arises because tree downcasts the embedded end node and then deferences that pointer. Currently there are 3 node types in __tree.

* `__tree_end_node` which contains the `__left_` pointer. This node is embedded within the container.
* `__tree_node_base` which contains `__right_`, `__parent_` and `__is_black`. This node is used throughout the tree rebalancing algorithms.
* `__tree_node` which contains `__value_`.

Currently `__tree` stores the start of the tree, `__begin_node_`, as a pointer to a `__tree_node`. Additionally the iterators store their position as a pointer to a `__tree_node`. In both of these cases the pointee can be the end node. This is fixed by changing them to store `__tree_end_node` pointers instead.

To make this change I introduced an `__iter_pointer` typedef which is defined to be a pointer to either `__tree_end_node` in the new ABI or `__tree_node` in the current one.
Both `__tree::__begin_node_` and iterator pointers are now stored as `__iter_pointers`.

The other situation where `__tree_end_node` is stored as the wrong type is in `__tree_node_base::__parent_`.  Currently `__left_`, `__right_`, and `__parent_` are all `__tree_node_base` pointers. Since the end node will only be stored in `__parent_` the fix is to change `__parent_` to be a pointer to `__tree_end_node`.

To make this change I introduced a `__parent_pointer` typedef which is defined to be a pointer to either `__tree_end_node` in the new ABI or `__tree_node_base` in the current one.

Note that in the new ABI `__iter_pointer` and `__parent_pointer` are the same type (but not in the old one). The confusion between these two types is unfortunate but it was the best solution I could come up with that maintains the ABI.

The typedef changes force a ton of explicit type casts to correct pointer types and to make current code compatible with both the old and new pointer typedefs. This is the bulk of the change and it's really messy. Unfortunately I don't know how to avoid it.

Please let me know what you think.





Reviewers: howard.hinnant, mclow.lists

Subscribers: howard.hinnant, bbannier, cfe-commits

Differential Revision: https://reviews.llvm.org/D20786
------------------------------------------------------------------------

llvm-svn: 276212

------------------------------------------------------------------------
r275935 | arsenm | 2016-07-18 17:35:22 -0700 (Mon, 18 Jul 2016) | 6 lines

AMDGPU/SI: Fix SI scheduler refcount issue

Without this fix, releaseSuccessors when InOrOutBlock is
false could release SUs outside the schedule BasicBlock.

Patch by Axel Davy
------------------------------------------------------------------------

llvm-svn: 276116

------------------------------------------------------------------------
r275943 | compnerd | 2016-07-18 19:13:08 -0700 (Mon, 18 Jul 2016) | 1 line

clang-rename: fix referenced variable in vim-script
------------------------------------------------------------------------

llvm-svn: 276115

------------------------------------------------------------------------
r275946 | bruening | 2016-07-18 22:03:38 -0700 (Mon, 18 Jul 2016) | 5 lines

[esan] Fix sideline thread flaky assert

Fixes an esan sideline thread CHECK that failed to account for the sideline
thread reaching its code before the internal_clone() return value was
assigned in the parent.
------------------------------------------------------------------------

------------------------------------------------------------------------
r275948 | bruening | 2016-07-18 22:06:48 -0700 (Mon, 18 Jul 2016) | 4 lines

[esan|wset] Fix flaky sampling tests

Adds a new esan public interface routine __esan_get_sample_count() and uses
it to ensure that tests of sampling receive the minimum number of samples.
------------------------------------------------------------------------

llvm-svn: 276114

Patch by Dmitry Polukhin <dmitry.polukhin@gmail.com>!

Differential Revision: https://reviews.llvm.org/D21970

llvm-svn: 276113

------------------------------------------------------------------------
r275866 | tnorthover | 2016-07-18 11:28:52 -0700 (Mon, 18 Jul 2016) | 6 lines

CodeGenPrep: use correct function to determine Global's alignment.

Elsewhere (particularly computeKnownBits) we assume that a global will be
aligned to the value returned by Value::getPointerAlignment. This is used to
boost the alignment on memcpy/memset, so any target-specific request can only
increase that value.
------------------------------------------------------------------------

llvm-svn: 275918

------------------------------------------------------------------------
r275879 | rnk | 2016-07-18 11:53:50 -0700 (Mon, 18 Jul 2016) | 1 line

Fix -Wmicrosoft-enum-value in GVNHoist.cpp
------------------------------------------------------------------------

llvm-svn: 275910

------------------------------------------------------------------------
r275880 | dcoughlin | 2016-07-18 11:57:50 -0700 (Mon, 18 Jul 2016) | 3 lines

Revert "[analyzer] Add checker modeling potential C++ self-assignment"

This reverts commit r275820. It is failing on the bots.
------------------------------------------------------------------------

llvm-svn: 275902

------------------------------------------------------------------------
r275898 | hans | 2016-07-18 13:06:27 -0700 (Mon, 18 Jul 2016) | 8 lines

Revert r273099 "If the revision number starts with r, drop it. It will get added back"

This doesn't seem to work with Bash:

$ /work/llvm/utils/release/merge.sh --proj llvm --rev r275870
/work/llvm/utils/release/merge.sh: line 34: ${$1#r}: bad substitution

I get the same error with and without a leading 'r'.
------------------------------------------------------------------------

llvm-svn: 275900

------------------------------------------------------------------------
r275870 | arsenm | 2016-07-18 11:34:59 -0700 (Mon, 18 Jul 2016) | 1 line

AMDGPU/R600: Replace barrier intrinsics
------------------------------------------------------------------------

llvm-svn: 275896

llvm-svn: 275851

llvm-svn: 275850

llvm-svn: 275841
llvm-svn: 275839
llvm-svn: 275838
llvm-svn: 275836
llvm-svn: 275835
llvm-svn: 275834
llvm-svn: 275833
llvm-svn: 275832
llvm-svn: 275831
llvm-svn: 275829
llvm-svn: 275828

llvm-svn: 275826

This is not compliant with the official ABI, but allows experimentation
with calling conventions.

llvm-svn: 275825

llvm-svn: 275824

Currently we only decode broadcasts from a vector of the same size.

llvm-svn: 275823

This is compliant with the official ABI, but allows experimentation with
calling conventions.

llvm-svn: 275822

llvm-svn: 275821

This checker checks copy and move assignment operators whether they are
protected against self-assignment. Since C++ core guidelines discourages
explicit checking for `&rhs==this` in general we take a different approach: in
top-frame analysis we branch the exploded graph for two cases, where &rhs==this
and &rhs!=this and let existing checkers (e.g. unix.Malloc) do the rest of the
work. It is important that we check all copy and move assignment operator in top
frame even if we checked them already since self-assignments may happen
undetected even in the same translation unit (e.g. using random indices for an
array what may or may not be the same).

A patch by Ádám Balogh!

Differential Revision: https://reviews.llvm.org/D19311

llvm-svn: 275820

Don't use pthread initializers in constexpr constructors. Patch by elram. Reviewed at https://reviews.llvm.org/D21637.

llvm-svn: 275819

DAGTypeLegalizer::CanSkipSoftenFloatOperand should allow
SELECT op code for x86_64 fp128 type for MME targets,
so SoftenFloatOperand does not abort on SELECT op code.

Differential Revision: http://reviews.llvm.org/D21758

llvm-svn: 275818

llvm-svn: 275817

It is breaking build 32-bit build bots.

llvm-svn: 275816

We negated a value with a signed type which invited problems when that
value was the most negative signed number.  Use an unsigned type
for the value instead.  It will compute the same twos complement
result without the UB.

llvm-svn: 275815

llvm-svn: 275814