llvm-svn: 170440

LoopVectorize: Emit reductions as log2(vectorsize) shuffles + vector ops instead of scalar operations.

For example on x86 with SSE4.2 a <8 x i8> add reduction becomes
	movdqa	%xmm0, %xmm1
	movhlps	%xmm1, %xmm1            ## xmm1 = xmm1[1,1]
	paddw	%xmm0, %xmm1
	pshufd	$1, %xmm1, %xmm0        ## xmm0 = xmm1[1,0,0,0]
	paddw	%xmm1, %xmm0
	phaddw	%xmm0, %xmm0
	pextrb	$0, %xmm0, %edx

instead of
	pextrb	$2, %xmm0, %esi
	pextrb	$0, %xmm0, %edx
	addb	%sil, %dl
	pextrb	$4, %xmm0, %esi
	addb	%dl, %sil
	pextrb	$6, %xmm0, %edx
	addb	%sil, %dl
	pextrb	$8, %xmm0, %esi
	addb	%dl, %sil
	pextrb	$10, %xmm0, %edi
	pextrb	$14, %xmm0, %edx
	addb	%sil, %dil
	pextrb	$12, %xmm0, %esi
	addb	%dil, %sil
	addb	%sil, %dl

llvm-svn: 170439

llvm-svn: 170438

The normal insert() function takes an MBB::iterator position, and
inserts a stand-alone MachineInstr as before.

The insert() function that takes an MBB::instr_iterator position can
insert instructions inside a bundle, and will now update the bundle
flags correctly when that happens.

When the insert position is between two bundles, it is unclear whether
the instruction should be appended to the previous bundle, prepended to
the next bundle, or stand on its own. The MBB::insert() function doesn't
bundle the instruction in that case, use the MIBundleBuilder class for
that.

llvm-svn: 170437

A register can be associated with several distinct register classes.
For example, on PPC, the floating point registers are each associated with
both F4RC (which holds f32) and F8RC (which holds f64). As a result, this code
would fail when provided with a floating point register and an f64 operand
because it would happen to find the register in the F4RC class first and
return that. From the F4RC class, SDAG would extract f32 as the register
type and then assert because of the invalid implied conversion between
the f64 value and the f32 register.

Instead, search all register classes. If a register class containing the
the requested register has the requested type, then return that register
class. Otherwise, as before, return the first register class found that
contains the requested register.

llvm-svn: 170436

llvm-svn: 170435

llvm-svn: 170434

llvm-svn: 170433

Add OpenCL images as clang builtin types.

llvm-svn: 170432

Revert changes from r170428, as I accidentally changed the line endings of these files to Windows style.

llvm-svn: 170431

llvm-svn: 170430

llvm-svn: 170429

llvm-svn: 170428

llvm-svn: 170427

The runtime skips atexit sleep if there are no threads now,
so it must be fast w/o it.
Allows to specify own TSAN_OPTIONS for the tests.

llvm-svn: 170426

llvm-svn: 170425

[Sanitizer] Expose StackTrace::GetPreviousInstructionPc() to get PC of call instruction from return address

llvm-svn: 170424

has got us to the wrong offset within an object.

llvm-svn: 170423

llvm-svn: 170422

Original patch by Tobias Grosser, slightly modified by Sebastian Pop.

llvm-svn: 170420

isStride now takes a partial schedule as input.

Patch from Tobias Grosser <tobias@grosser.es>.

llvm-svn: 170419

llvm-svn: 170418

llvm-svn: 170417

llvm-svn: 170416

pointing at the bad location and a snippet of nearby memory values. This is
strictly best-effort; reading these bytes to display the note could lead to a
seg fault, and that's explicitly OK.

llvm-svn: 170415

into the same file in the future.

llvm-svn: 170414

llvm-svn: 170413

getScalarSizeInBits could not handle vectors of pointers.

llvm-svn: 170412

I don't want to suppress this to be excluded with the feature "dev-fd-fs". This should work also on freebsd.

llvm-svn: 170411

llvm-svn: 170410

This can be used by lldb to ask for information
about the process debugserver is attached to/launched.
Particularly useful on a 64-bit x86 Mac system which
can run  32-bit or 64-bit user-land processes.

llvm-svn: 170409

llvm-svn: 170408

one out from the return address. Currently, we can only resolve this address to
a file and line number if we have an external symbolizer.

llvm-svn: 170407

This fixes the storage class of extern decls that are merged with file level
statics. The patch also fixes the linkage computation so that they are
considered internal.

llvm-svn: 170406

but remove unneeded options and fix the comment.

llvm-svn: 170405

llvm-svn: 170404

llvm-svn: 170403

(This change only affects ObjC.)

<rdar://problem/12857117>.

llvm-svn: 170402

As per Ted's advice. It can be brought back from version control if
needed.

This also fixes a Sphinx warning.

llvm-svn: 170401

<rdar://problem/11597849>

llvm-svn: 170400