llvm-svn: 109876

llvm-svn: 109875

llvm-svn: 109872

the llvm.memset() intrinsic family.
No content changes.

llvm-svn: 109863

llvm-svn: 109859

llvm-svn: 109858

check the range of the constant when optimizing a comparison between a
constant and a sign_extend_inreg node.

llvm-svn: 109854

llvm-svn: 109852

handles with a pointer to the containing map.  When a map is copied, these
pointers need to be corrected to point to the new map.  If not, then consider
the case of a map M1 which maps a value V to something.  Create a copy M2 of
M1.  At this point there are two value handles on V, one representing V as a
key in M1, the other representing V as a key in M2.  But both value handles
point to M1 as the containing map.  Now delete V.  The value handles remove
themselves from their containing map (which destroys them), but only the first
value handle is successful: the second one cannot remove itself from M1 as
(once the first one has removed itself) there is nothing there to remove; it
is therefore not destroyed.  This causes an assertion failure "All references
to V were not removed?".

llvm-svn: 109851

have 4 bits per register in the operand encoding), but have undefined
behavior when the operand value is 13 or 15 (SP and PC, respectively).
The trivial coalescer in linear scan sometimes will merge a copy from
SP into a subsequent instruction which uses the copy, and if that
instruction cannot legally reference SP, we get bad code such as:
  mls r0,r9,r0,sp
instead of:
  mov r2, sp
  mls r0, r9, r0, r2

This patch adds a new register class for use by Thumb2 that excludes
the problematic registers (SP and PC) and is used instead of GPR
for those operands which cannot legally reference PC or SP. The
trivial coalescer explicitly requires that the register class
of the destination for the COPY instruction contain the source
register for the COPY to be considered for coalescing. This prevents
errant instructions like that above.

PR7499

llvm-svn: 109842

llvm-svn: 109821

llvm-svn: 109820

llvm-svn: 109818

completed a successful build.

llvm-svn: 109815

llvm-svn: 109813

llvm-svn: 109810

llvm-svn: 109809

llvm-svn: 109801

transformations.

llvm-svn: 109800

integers with mov + vdup.  8003375.  This is
currently disabled by default because LICM will
not hoist a VDUP, so it pessimizes the code if
the construct occurs inside a loop (8248029).

llvm-svn: 109799

PR7745.

llvm-svn: 109788

llvm-svn: 109787

llvm-svn: 109775

llvm-svn: 109774

if CExpr is a ConstantInt.

llvm-svn: 109773

llvm-svn: 109772

Add intrinsics __builtin_arm_qadd & __builtin_arm_qsub to allow access to the QADD & QSUB instructions.
Behave identically to __qadd & __qsub RealView instruction intrinsics.

llvm-svn: 109770

message is properly contaminated with nonsense about timing passes
that doesn't apply at all to this utility. :)

llvm-svn: 109769

Plug the remaining MC leaks by giving MCObjectStreamer/MCAsmStreamer ownership of the TargetAsmBackend and the MCCodeEmitter.

llvm-svn: 109767

llvm-svn: 109765

We do sometimes load from a too small stack slot when dealing with x86 arguments
(varargs and smaller-than-32-bit args). It looks like we know what we are doing
in those cases, so I am going to remove the assert instead of artifically
enlarging stack slot sizes.

The assert in storeRegToStackSlot stays in. We don't want to write beyond the
bounds of a stack slot.

llvm-svn: 109764

llvm-svn: 109752

memory when one of the original BB is destroyed.

llvm-svn: 109747

llvm-svn: 109746

llvm-svn: 109745

it establishes a context and does a complaining diff.  Also make sure we
unify the prelude and postlude of a diff after a block-diff call.

llvm-svn: 109744

structurally identical.

llvm-svn: 109743

any differences we see.  This should only happen if there are "non-structural"
differences between the instructions, i.e. differences which wouldn't cause
diff to return true.

llvm-svn: 109742

in despite not ever incrementing any path costs, so that the only nonzero costs
arose from the all-left path in the first column.  Anyway.  Perform the diff
starting from the beginning of the block to avoid capturing (say) loads of
allocas.

Vastly improves diff results on code that hasn't been mem2reg'ed.

llvm-svn: 109741

llvm-svn: 109740