delete the MDNode that changed, rather than the other MDNode.
This is less work, because it doesn't require the changed node
to be re-inserted into the uniquing map and it doesn't require
the is-function-local flag to be recomputed. Also, it avoids
trouble when the existing node is part of a complicated
data structure.

llvm-svn: 114996

cost modeling for if-conversion.  Now if only we had a way to estimate the misprediction probability.

Adjsut CodeGen/ARM/ifcvt10.ll.  The pipeline on Cortex-A8 is long enough that it is still profitable
to predicate an ldm, but the shorter pipeline on Cortex-A9 makes it unprofitable.

llvm-svn: 114995

llvm-svn: 114994

  ARMTargetMachine.cpp:53: error: control reaches end of non-void function

llvm-svn: 114992

Patch by Evzen Muller!

llvm-svn: 114991

it's not used afterwards.

llvm-svn: 114986

estimates.

llvm-svn: 114981

llvm-svn: 114974

Rather than having arbitrary cutoffs, actually try to cost model the conversion.

For now, the constants are tuned to more or less match our existing behavior, but these will be
changed to reflect realistic values as this work proceeds.

llvm-svn: 114973

also fixes PR8250.

llvm-svn: 114972

llvm-svn: 114969

This should make the arm-linux self-host buildbot happy again.

llvm-svn: 114964

llvm-svn: 114958

llvm-svn: 114953

llvm-svn: 114952

llvm-svn: 114949

llvm-svn: 114948

llvm-svn: 114947

llvm-svn: 114943

accesses for ARM targets that would otherwise allow it.  Radar 8465431.

llvm-svn: 114941

now takes a libcall operand, sets up the arguments correctly and
handles stack adjustments.

llvm-svn: 114934

llvm-svn: 114931

llvm-svn: 114930

llvm-svn: 114920

register pressure and thus excess spills, which we don't currently recover from well.  This should
be re-evaluated in the future if our ability to generate good spills/splits improves.

Partial fix for <rdar://problem/7635585>.

llvm-svn: 114919

enable it for real. Leaving the CL option in place to it's easy to disable it
again if (when) testers find something I've missed.

llvm-svn: 114915

makes files easier to diff.

llvm-svn: 114898

llvm-svn: 114896

llvm-svn: 114895

Usually we wouldn't do this anyway because llvm_fenv_testexcept would return an
exception, but we have seen some cases where neither errno nor fenv detect an
exception on arm-linux.

llvm-svn: 114893

enable it for real. Leaving the CL option in place to it's easy to disable it
again if (when) testers find something I've missed.

llvm-svn: 114892

llvm-svn: 114891

llvm-svn: 114888

llvm-svn: 114862

llvm-svn: 114861

This reverts revision 114633. It was breaking llvm-gcc-i386-linux-selfhost.

It seems there is a downstream bug that is exposed by
-cgp-critical-edge-splitting=0. When that bug is fixed, this patch can go back
in.

Note that the changes to tailcallfp2.ll are not reverted. They were good are
required.

llvm-svn: 114859

llc now recognizes the "intent" to support MC/obj emission for ARM, but
given that they are all stubs, it asserts on --filetype=obj --march=arm

Patch by Jason Kim.

llvm-svn: 114856

llvm-svn: 114852

llvm-svn: 114847

llvm-svn: 114844