expand to multiple basic blocks, in which case fast-isel
needs to informed of which block to use as it resumes
inserting instructions.

llvm-svn: 57040

was setting kill flags on tied uses in two-address instructions.
The kill flags were causing the allocator to think it could
allocate the use and its tied def in different registers.

llvm-svn: 57039

Make it all work in non-pic mode.

llvm-svn: 57034

applicable.

llvm-svn: 57033

Thanks Duncan!

llvm-svn: 57029

incidentally making the case where the memop is a
pointer deref work.  Fix cmp-and-swap regression.

llvm-svn: 57027

While hosting instruction check PHI node.

llvm-svn: 57025

llvm-svn: 57021

llvm-svn: 57020

llvm-svn: 57018

llvm-svn: 57017

Instead, just create one, and make sure everything that needs
it can access it. Previously most of the SelectionDAGISel
subclasses all had their own TargetLowering object, which was
redundant with the TargetLowering object in the TargetMachine
subclasses, except on Sparc, where SparcTargetMachine
didn't have a TargetLowering object. Change Sparc to work
more like the other targets here.

llvm-svn: 57016

llvm-svn: 57014

this handling to work properly for modifying stub functions, relocations
back to entry points after JIT compilation, etc..

llvm-svn: 57013

reg->reg moves.

llvm-svn: 57011

NeedStub/DoesntNeedStub logic was reversed, leading to not using a stub
for global relocations that do need them (libc calls, for example).

llvm-svn: 57010

llvm-svn: 57009

llvm-svn: 57008

llvm-svn: 57007

isReg, etc., from isRegister, etc.

llvm-svn: 57006

llvm-svn: 57005

than the type an i1 is promoted to (eg: i8).  Account
for this.  Noticed by Tilmann Scheller on CellSPU; he
will hopefully take care of fixing this in LegalizeDAG
and adding a testcase!

llvm-svn: 56997

Why?  Because it was there!

llvm-svn: 56996

the lock when calling a method which may materialize the llvm::Function.

llvm-svn: 56995

llvm-svn: 56994

It turns out that this is a fairly common operation,
and it's easy enough to handle.

llvm-svn: 56990

getting inserted into the ValueMap. This avoids infinite
recursion in some rare cases.

llvm-svn: 56989

sequences like this:
       sete    %al
       testb   %al, %al
       jne     LBB11_1
with this:
       je      LBB11_1

llvm-svn: 56969

can be used when deciding if a block can transfer control to another
via a fall-through instead of a branch.

llvm-svn: 56968

llvm-svn: 56965

Use a multimap rather than a map for holding the list of copies to insert, so we don't lose copies when two of them have
the same source.  I don't know what I was thinking when I wrote this originally.
Note: There's probably a more efficient way to do this, but I need to think about it some more, and about what determinism
guarantees need to be present.

llvm-svn: 56964

llvm-svn: 56963

llvm-svn: 56960

Contributed by Lang Hames.

llvm-svn: 56959

llvm-svn: 56957

volatile memory references are safe to move.

llvm-svn: 56948

local register allocator's physreg liveness doesn't recognize subregs,
so it doesn't know that defs of %ecx that are immediately followed by
uses of %cl aren't dead. This comes up due to the way fast-isel emits
shift instructions.

This is a temporary workaround. Arguably, local regalloc should
handle subreg references correctly. On the other hand, perhaps
fast-isel should use INSERT_SUBREG instead of just assigning to the
most convenient super-register of %cl when lowering shifts.

This fixes MultiSource/Benchmarks/MallocBench/espresso,
MultiSource/Applications/hexxagon, and others, under -fast.

llvm-svn: 56947

a constant vector ("{0x123, 0x456}" syntax).  The fix is to simplify the
_mm_srli_si128 macro, and  move the "* 8" from the macro into the compiler
back-end.  I can't change the existing __builtins because so many people are
using them :-(."
Patch by Stuart Hastings!

llvm-svn: 56944

& version as strings.
 - Win32 code is untested.

llvm-svn: 56942

llvm-svn: 56939