llvm-svn: 47671

operand of a VECTOR_SHUFFLE.  The mask is a
vector of constant integers.  The code in
LegalizeDAG doesn't bother to legalize the
mask, since it's basically just storage for
a bunch of constants, however LegalizeTypes
is more picky.  The problem is that there may
not exist any legal vector-of-integers type
with a legal element type, so it is impossible
to create a legal mask!  Unless of course you
cheat by creating a BUILD_VECTOR where the
operands have a different type to the element
type of the vector being built...  This is
pretty ugly but works - all relevant tests in
the testsuite pass, and produce the same
assembler with and without LegalizeTypes.

llvm-svn: 47670

llvm-svn: 47669

Don't track max alignment during stack object allocations since they can be deleted later. Let PEI compute it.

llvm-svn: 47668

llvm-svn: 47667

successor of blocks created above it.

llvm-svn: 47666

In such cases, handle them just like Visit().

llvm-svn: 47665

globals/parameters at the beginning of the analysis.

llvm-svn: 47664

llvm-svn: 47663

llvm-svn: 47662

llvm-svn: 47661

stack slot and store if the  SINT_TO_FP is actually legal.  This allows
us to compile:

double a(double b) {return (unsigned)b;}

to:

_a:
	cvttsd2siq	%xmm0, %rax
	movl	%eax, %eax
	cvtsi2sdq	%rax, %xmm0
	ret

instead of:

_a:
	subq	$8, %rsp
	cvttsd2siq	%xmm0, %rax
	movl	%eax, %eax
	cvtsi2sdq	%rax, %xmm0
	addq	$8, %rsp
	ret

crazy.

llvm-svn: 47660

llvm-svn: 47659

_test:
	movl	%edi, %eax
	ret

instead of:

_test:
        movl    $4294967295, %ecx
        movq    %rdi, %rax
        andq    %rcx, %rax
        ret

It would be great to write this as a Pat pattern that used subregs 
instead of a 'pseudo' instruction, but I don't know how to do that
in td files.

llvm-svn: 47658

llvm-svn: 47657

llvm-svn: 47656

ComputeMaskedBits to use the APInt form, and remove the
non-APInt form.

llvm-svn: 47654

llvm-svn: 47653

llvm-svn: 47652

ExplodedNode for the Callee subexpression we would not evaluate the CallExpr
transfer function.

llvm-svn: 47651

sometimes skip evaluating all the arguments when some arguments would not
create new ExplodedNodes.

llvm-svn: 47650

multiple basic blocks (which can happen when they contain '&&', '||', '?'). The
bug was that the loop backedge when to the last block in the loop condition, not
the first.

llvm-svn: 47649

Change several cases in SimplifyDemandedMask that don't ever do any
simplifying to reuse the logic in ComputeMaskedBits instead of
duplicating it.

llvm-svn: 47648

llvm-svn: 47646

Fix bug when processing '?' operator: invalidate the old "Uninitialized" value of the block-level expression for ?.

llvm-svn: 47645

llvm-svn: 47644

llvm-svn: 47642

llvm-svn: 47641

llvm-svn: 47640

in an invalid transformation.

llvm-svn: 47639

GRSimpleVals performs the following action: invalidate all values passed-by-reference.

llvm-svn: 47638

llvm-svn: 47637

uninitialized) value. At this point we're just too imprecise.

llvm-svn: 47636

llvm-svn: 47635

llvm-svn: 47634

llvm-svn: 47633

instead of init'ing it maximally to zeros on entry.  getFreePhysReg
is pretty hot and only a few elements are typically used.  This speeds
up linscan by 5% on 176.gcc.

llvm-svn: 47631

Before it was the negation of hasAutoStorage(), which is incorrect. 
"Register" storage for Decls do not indicate global variables.

llvm-svn: 47630

llvm-svn: 47629

llvm-svn: 47628