use can't be dominated, saving one domtree lookup.

llvm-svn: 142066

Based on Horspool's simplified version of Boyer-Moore. We use a constant-sized table of
uint8_ts to keep cache thrashing low, needles bigger than 255 bytes are uncommon anyways.

The worst case is still O(n*m) but we do a lot better on the average case now.

llvm-svn: 142061

cases we need the second type-legalization pass in order to support all cases.

llvm-svn: 142060

Just because we're dealing with a GEP doesn't mean we can assert the
SCEV has a pointer type. The fix is simply to ignore the SCEV pointer
type, which we really didn't need.
Fixes PR11138 webkit crash.

llvm-svn: 142058

It extremely breaks builds when optimization is enabled. Twine should not hold temporary objects.

By the way, I might feel sad if I saw "i786-" "i1586-" or "iF86-".

llvm-svn: 142055

from config.h, it discovers the triple based on the execution environment.

Patch by Aaron Ballman <aaron@aaronballman.com>

llvm-svn: 142046

Windows/Memory.inc: Support the ability to allocate memory "near" another block of memory on Win32. It has fixed FIXME.

Thanks to Aaron Ballman!

llvm-svn: 142039

I rewrote the algorithm a while back so it doesn't require map lookup,
but neglected to change the data structure. This was caught by
llvm-gcc self host, not because there's anything special about
llvm-gcc, but because it is the only test for nondeterminism we
currently have. Unit tests don't work well for everything; we should
always try to have a nondeterminism stress test running.

Fixes PR11133: llvm-gcc self host .o mismatch after enable-iv-rewrite=false

llvm-svn: 142036

This isn't put into the 'clear()' method because the information needs to stick
around (at least for a little bit) after the selection DAG is built.

llvm-svn: 142032

It really doesn't, but when r141929 removed the hasSideEffects flag from
this instruction, it caused miscompilations.  I am guessing that it got
moved across a stack pointer update.

Also clear isRematerializable after checking that this instruction is
in fact never rematerialized in the nightly test suite.

llvm-svn: 142030

rdar://10288916 is tracking this fix.

In the past, instcombine and other passes were promoting alloca alignment past
the natural alignment, resulting in dynamic stack realignment.  Lang's work now
prevents this from happening (LLVM commit r141599).  Now that this really 
shouldn't happen report a fatal error rather than silently generate bad code.

llvm-svn: 142028

llvm-svn: 142027

llvm-svn: 142022

llvm-svn: 142021

The callee-saved registers cannot be live across an invoke call because the
control flow may continue along the exceptional edge. When this happens, all of
the callee-saved registers are no longer valid.

llvm-svn: 142018

llvm-svn: 142013

llvm-svn: 142010

    assert("bad SymbolicOp.VariantKind");
To:
    assert(0 && "bad SymbolicOp.VariantKind");

llvm-svn: 142000

llvm-svn: 141990

llvm-svn: 141989

llvm-svn: 141988

Disable code/data region symbols on ELF targets, where different mapping symbols are used for ARM/Thumb mode code. This should only be re-enabled once we have a solution to properly distinguish these.

llvm-svn: 141984

llvm-svn: 141981

llvm-svn: 141978

llvm-svn: 141972

Fixes PR10814. Patch by Jan Sjödin!

llvm-svn: 141960

llvm-svn: 141959

Pointed out by George Russell.

llvm-svn: 141956

could be arguments, for example.

No testcase because this is a bug-fix broken out of a larger optimization patch.

llvm-svn: 141951

llvm-svn: 141947

Add X86 TZCNT instruction and patterns to select it. Also added core-avx2 processor which is gcc's name for Haswell.

llvm-svn: 141939

Patch by Jack Carter and Reed Kotler at Mips. 

llvm-svn: 141938

Patch by Jack Carter and Reed Kotler at Mips. 

llvm-svn: 141937

Patch by Jack Carter and Reed Kotler at Mips. 

llvm-svn: 141936

Patch by Jack Carter at Mips. 

llvm-svn: 141934

Patch by Jack Carter at Mips. 

llvm-svn: 141932

TableGen infers unmodeled side effects on instructions without a
pattern.  Fix some instruction definitions where that was overlooked.

Also raise an error if a rematerializable instruction has unmodeled side
effects. That doen't make any sense.

llvm-svn: 141929

TableGen will mark any pattern-less instruction as having unmodeled side
effects. This is extra bad for V_SET0 which gets rematerialized a lot.

This was part of the cause for PR11125, but the real bug was fixed
in r141923.

llvm-svn: 141924

When spilling around an instruction with a dead def, remember to add a
value number for the def.

The missing value number wouldn't normally create problems since there
would be an incoming live range as well.  However, due to another bug
we could spill a dead V_SET0 instruction which doesn't read any values.

The missing value number caused an empty live range to be created which
is dangerous since it doesn't interfere with anything.

This fixes part of PR11125.

llvm-svn: 141923

Someone more familiar with LSR should double-check that the extra cast is actually doing the right thing in the overflow cases; I'm not completely confident that's that case. 

llvm-svn: 141916