llvm-svn: 155366

llvm-svn: 155365

llvm-svn: 155364

llvm-svn: 155363

Original commit message:

Defer some shl transforms to DAGCombine.

The shl instruction is used to represent multiplication by a constant
power of two as well as bitwise left shifts. Some InstCombine
transformations would turn an shl instruction into a bit mask operation,
making it difficult for later analysis passes to recognize the
constsnt multiplication.

Disable those shl transformations, deferring them to DAGCombine time.
An 'shl X, C' instruction is now treated mostly the same was as 'mul X, C'.

These transformations are deferred:

  (X >>? C) << C   --> X & (-1 << C)  (When X >> C has multiple uses)
  (X >>? C1) << C2 --> X << (C2-C1) & (-1 << C2)   (When C2 > C1)
  (X >>? C1) << C2 --> X >>? (C1-C2) & (-1 << C2)  (When C1 > C2)

The corresponding exact transformations are preserved, just like
div-exact + mul:

  (X >>?,exact C) << C   --> X
  (X >>?,exact C1) << C2 --> X << (C2-C1)
  (X >>?,exact C1) << C2 --> X >>?,exact (C1-C2)

The disabled transformations could also prevent the instruction selector
from recognizing rotate patterns in hash functions and cryptographic
primitives. I have a test case for that, but it is too fragile.

llvm-svn: 155362

interpreter.

llvm-svn: 155360

existentially quantified lock expressions.

llvm-svn: 155357

<rdar://problem/11284902>.

llvm-svn: 155356

llvm-svn: 155355

The problem is that the struct file_status on UNIX systems has two
members called st_dev and st_ino; those are also members of the
struct stat, and they are reserved identifiers which can also be
provided as #define (and this is the case for st_dev on Hurd).
The solution (attached) is to rename them, for example adding a
"fs_" prefix (= file status) to them.

Patch by Pino Toscano

llvm-svn: 155354

llvm-svn: 155353

llvm-svn: 155352

llvm-svn: 155351

llvm-svn: 155350

 reference, so that we do not need to type an extra '&' operator when calling the function.

llvm-svn: 155349

2. Include the helper function and the helper class in the LoopGenerator.h into the polly namespace.

llvm-svn: 155348

Fix issue 67 by checking that the interface functions weren't redefined in the compiled source file.

llvm-svn: 155346

llvm-svn: 155345

llvm-svn: 155344

llvm-svn: 155343

Do not use stdint.h, driver might provide invalid location for it. Instead, provide the types directly.
This should fix PR12628

llvm-svn: 155342

llvm-svn: 155341

llvm-svn: 155340

Use MVT instead of EVT through all of LowerVECTOR_SHUFFLEtoBlend and not just the switch. Saves a little bit of binary size.

llvm-svn: 155339

llvm-svn: 155338

Make getZeroVector and getOnesVector more alike as far as how they detect 128-bit versus 256-bit vectors. Be explicit about both sizes and use llvm_unreachable. Similar changes to getLegalSplat.

llvm-svn: 155337

llvm-svn: 155336

Tidy up spacing in LowerVECTOR_SHUFFLEtoBlend. Remove code that checks if shuffle operand has a different type than the the shuffle result since it can never happen.

llvm-svn: 155333

llvm-svn: 155332

llvm-svn: 155331

llvm-svn: 155330

fixes an assert reading "1239123123123123" when the result is already 64-bit.

llvm-svn: 155329

llvm-svn: 155328

intructions are processed. So there's no need to look at them if they're used as
operands of other instructions.

llvm-svn: 155327

by a chrome build on OS X. Thanks to Nico Weber for testing a patch and
providing the .ii file.

llvm-svn: 155326

Make Extract128BitVector and Insert128BitVector take an unsigned instead of an ConstantNode SDValue. getConstant was almost always called just before only to have the functions take it apart and build a new ConstantSDNode.

llvm-svn: 155325

Found by valgrind.

llvm-svn: 155324

Remove unnecessary StringRef->char*->StringRef conversion, which read uninitialized memory if the input wasn't 0-terminated.

Found by valgrind.

llvm-svn: 155323

llvm-svn: 155321

Make calls to getVectorShuffle more consistent. Use shuffle VT for calls to getUNDEF instead of requerying. Use &Mask[0] instead of Mask.data().

llvm-svn: 155320