properly account for the *global* probability of the edge being taken.
This manifested as a very large number of unconditional branches to
blocks being merged against the CFG even though they weren't
particularly hot within the CFG.

The fix is to check whether the edge being merged is both locally hot
relative to other successors for the source block, and globally hot
compared to other (unmerged) predecessors of the destination block.

This introduces a new crasher on GCC single-source, but it's currently
behind a flag, and Ben has offered to work on the reduction. =]

llvm-svn: 145010

setFlags doesn't modify its arguments.

llvm-svn: 145007

Add code for lowering v32i8 shifts by a splat to AVX2 immediate shift instructions. Remove 256-bit splat handling from LowerShift as it was already handled by PerformShiftCombine.

llvm-svn: 145005

llvm-svn: 145004

Remove some of the special classes that worked around an old tablegen limitation of not being able to remove redundant bitconverts from patterns.

llvm-svn: 145003

llvm-svn: 144999

formation phase and into the initial walk of the basic blocks. We
essentially pre-merge all blocks where unanalyzable fallthrough exists,
as we won't be able to update the terminators effectively after any
reorderings. This is quite a bit more principled as there may be CFGs
where the second half of the unanalyzable pair has some analyzable
predecessor that gets placed first. Then it may get placed next,
implicitly breaking the unanalyzable branch even though we never even
looked at the part that isn't analyzable. I've included a test case that
triggers this (thanks Benjamin yet again!), and I'm hoping to synthesize
some more general ones as I dig into related issues.

Also, to make this new scheme work we have to be able to handle branches
into the middle of a chain, so add this check. We always fallback on the
incoming ordering.

Finally, this starts to really underscore a known limitation of the
current implementation -- we don't consider broken predecessors when
merging successors. This can caused major missed opportunities, and is
something I'm planning on looking at next (modulo more bug reports).

llvm-svn: 144994

Synthesize SSSE3/AVX 128-bit horizontal integer add/sub instructions from add/sub of appropriate shuffle vectors.

llvm-svn: 144989

llvm-svn: 144988

llvm-svn: 144987

llvm-svn: 144985

The loop tree's inclusive block lists are painful and expensive to
update. (I have no idea why they're inclusive). The design was
supposed to handle this case but the implementation missed it and my
unit tests weren't thorough enough.

Fixes PR11335: loop unroll update.

llvm-svn: 144970

llvm-svn: 144967

[asan] workaround for reg alloc bug 11395: don't instrument functions with large chunks of inline assembler

llvm-svn: 144962

llvm-svn: 144959

DISubrange supports unsigned lower/upper array bounds, so let's not fake it in the end while emitting DWARF. If a FE needs to encode signed lower/upper array bounds then we need to extend DISubrange or ad DISignedSubrange. 

llvm-svn: 144937

quick fix: remove GlobalVariable::GlobalVariable mistakenly commited at r144933. For some reason this compiles on linux 

llvm-svn: 144936

The right way to check for a binary operation is
cast<BinaryOperator>. The original check: cast<Instruction> &&
numOperands() == 2 would match phi "instructions", leading to an
infinite loop in extreme corner case: a useless phi with operands
[self, constant] that prior optimization passes failed to remove,
being used in the loop by another useless phi, in turn being used by an
lshr or udiv.

Fixes PR11350: runaway iteration assertion.

llvm-svn: 144935

fall back to explicit list of allowed linkages when instrumenting globals in asan; add a test check that asan does not touch linkonce_odr

llvm-svn: 144933

llvm-svn: 144920

Fix SSE/AVX integer comparison patterns to understand that all integer vector loads are promoted to i64 vector loads so patterns need a bitconvert. Also slightly simplify the AVX2 variable shift patterns by using the predefined bitconvert pattern fragments.

llvm-svn: 144896

llvm-svn: 144888

ADDs.  MaxOffs is used as a threshold to limit the size of the offset. Tradeoffs
being: (1) If we can't materialize the large constant then we'll cause fast-isel
to bail. (2) Too large of an offset can't be directly encoded in the ADD
resulting in a MOV+ADD.  Generally not a bad thing because otherwise we would
have had ADD+ADD, but on Thumb this turns into a MOVS+MOVT+ADD. Working on a fix
for that. (3) Conversely, too low of a threshold we'll miss opportunities to 
coalesce ADDs.
rdar://10412592

llvm-svn: 144886

llvm-svn: 144885

Add support for custom names for library functions in TargetLibraryInfo.  Add a custom name for fwrite and fputs on x86-32 OSX.  Make SimplifyLibCalls honor the custom
names for fwrite and fputs.

Fixes <rdar://problem/9815881>.

llvm-svn: 144876

rdar://10456186

llvm-svn: 144872

Turn on vzeroupper insertion on call boundaries for AVX; it works as far as I know, and I'd like to see wider testing.

llvm-svn: 144867

Make sure to replace the chain properly when DAGCombining a LOAD+EXTRACT_VECTOR_ELT into a single LOAD.  Fixes PR10747/PR11393.

llvm-svn: 144863

llvm-svn: 144861

We don't (yet) have the granularity in the fixups to be specific about which
bitranges are affected. That's a future cleanup, but we're not there yet.

llvm-svn: 144852

llvm-svn: 144849

llvm-svn: 144847

llvm-svn: 144842

nodes.

llvm-svn: 144841

llvm-svn: 144840

llvm-svn: 144839

llvm-svn: 144837

llvm-svn: 144836

target-independent selector or the target-specific selector.

llvm-svn: 144833

for a single miss and not all predecessor instructions that get selected by
the selection DAG instruction selector.  This is still not exact (e.g., over
states misses when folded/dead instructions are present), but it is a step in
the right direction.

llvm-svn: 144832