Do not mark no-return calls tail calls. It'll screw up special calls like longjmp and it doesn't make much sense for performance reason. If my logic is faulty, please let me know.

llvm-svn: 94937

llvm-svn: 94915

llvm-svn: 94914

Even if they are suported by the core, they can be disabled
(this is just a configuration bit inside some register).

Allow unaligned memops on darwin and conservatively disallow them otherwise.

llvm-svn: 94889

unconditionally.  Besides checking the offset, also check that the underlying
object is aligned as much as the load itself.

llvm-svn: 94875

llvm-svn: 94873

llvm-svn: 94872

llvm-svn: 94870

something totally broken and parsing them as immediates, but the .td file also
had the wrong match class so things sortof worked. Except, that is, that we
would parse
  movl $0, %eax
as
  movl 0, %eax
Feel free to guess how well that worked.

llvm-svn: 94869

llvm-svn: 94866

llvm-svn: 94863

llvm-svn: 94861

register, and use to cleanup a FIXME in X86AsmParser.cpp.

llvm-svn: 94859

This bug was exposed by my inliner cost changes in r94615, and caused failures
of lencod on most architectures when building with LTO.

This patch fixes lencod and 464.h264ref on x86-64 (and likely others).

llvm-svn: 94858

bit (Inst{22}) and the M bit (Inst{5}) should be left unspecified.  For binary
format instructions, Inst{6} and Inst{4} need to specified for proper decodings.

llvm-svn: 94855

them from values that are not actually defined in the module.

llvm-svn: 94854

llvm-svn: 94853

create a testcase where this matters.  The select+load transformation only
occurs when isSafeToLoadUnconditionally is true, and in those situations,
instcombine also changes the underlying objects to be aligned.  This seems
like a good idea regardless, and I've verified that it doesn't pessimize
the subsequent realignment.

llvm-svn: 94850

llvm-svn: 94843

We were not writing bitcode for function-local metadata whose operands have been erased (making it not have any more function-local operands)

llvm-svn: 94842

llvm-svn: 94841

llvm-svn: 94835

indices are safe if the result is known to be within the bounds of the
underlying object.

llvm-svn: 94829

llvm-svn: 94822

Before inserting   llvm.dbg.declare intrinsic at the end of a basic block, check whether the basic block has a terminator or not.
This API is used by clang and the test case is test/CodeGen/debug-info-crash.c in clang module.

llvm-svn: 94820

llvm-svn: 94809

llvm-svn: 94808

llvm-svn: 94807

llvm-svn: 94805

llvm-svn: 94804

(via APInt &RHSKnownZero = KnownZero, etc) seems dangerous and confusing to me: it
is easy not to notice this, and then wonder why KnownZero/RHSKnownZero changed
underneath you when you modified RHSKnownZero/KnownZero etc.  So get rid of this.
No intended functionality change (tested with "make check" + llvm-gcc bootstrap).

llvm-svn: 94802

(fix for PR6165) are needed here too.

llvm-svn: 94801

llvm-svn: 94783

lowering. We'll either figure it out, or not and be lowered by
SelectionDAGBuild.

Add test.

llvm-svn: 94775

llvm-svn: 94771

llvm-svn: 94765

enhanced disassembler, and the necessary makefile
rules to build the table for X86.

llvm-svn: 94764

llvm-svn: 94763

"visit*" method is called, take the newly created nodes, walk them in a DFS
fashion, and if they don't have an ordering set, then give it one.

llvm-svn: 94757

  getelementptr (i8* inttoptr (i64 1 to i8*), i32 -1) 
  to
  inttoptr (i64 0 to i8*)
from the VMCore constant folder. It didn't handle sign-extension properly
in the case where the source integer is smaller than a pointer size. And,
it relied on an assumption about sizeof(i8).

The Analysis constant folder still folds these kinds of things; it has
access to TargetData, so it can do them right.

Add a testcase which tests that the VMCore constant folder doesn't
miscompile this, and that the Analysis folder does fold it.

llvm-svn: 94750