LegalizeTypes support for legalizing the mask
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.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47670 91177308-0d34-0410-b5e6-96231b3b80d8

LegalizeTypes support for INSERT_VECTOR_ELT.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47669 91177308-0d34-0410-b5e6-96231b3b80d8

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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47668 91177308-0d34-0410-b5e6-96231b3b80d8

Support for legalizing MEMBARRIER.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47667 91177308-0d34-0410-b5e6-96231b3b80d8

Final de-tabification.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47663 91177308-0d34-0410-b5e6-96231b3b80d8

Add path separator support, patch by Sam Bishop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47662 91177308-0d34-0410-b5e6-96231b3b80d8

Make X86TargetLowering::LowerSINT_TO_FP return without creating a dead
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.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47660 91177308-0d34-0410-b5e6-96231b3b80d8

this code is correct but strange looking ;-)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47659 91177308-0d34-0410-b5e6-96231b3b80d8

Compile x86-64-and-mask.ll into:

_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.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47658 91177308-0d34-0410-b5e6-96231b3b80d8

Spiller now remove unused spill slots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47657 91177308-0d34-0410-b5e6-96231b3b80d8

Teach Legalize how to expand an EXTRACT_ELEMENT.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47656 91177308-0d34-0410-b5e6-96231b3b80d8

Convert the last remaining users of the non-APInt form of
ComputeMaskedBits to use the APInt form, and remove the
non-APInt form.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47654 91177308-0d34-0410-b5e6-96231b3b80d8

Add comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47653 91177308-0d34-0410-b5e6-96231b3b80d8

add a note

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47652 91177308-0d34-0410-b5e6-96231b3b80d8

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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47648 91177308-0d34-0410-b5e6-96231b3b80d8

Upgrade testcase.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47644 91177308-0d34-0410-b5e6-96231b3b80d8

regenerate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47642 91177308-0d34-0410-b5e6-96231b3b80d8

Add assert.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47641 91177308-0d34-0410-b5e6-96231b3b80d8

Add PR number to testcase.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47640 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an issue where GVN had the sizes of the two memcpy's reverse, resulting
in an invalid transformation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47639 91177308-0d34-0410-b5e6-96231b3b80d8

Add assert to check return type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47637 91177308-0d34-0410-b5e6-96231b3b80d8

Refactor according to Evan's and Anton's suggestions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47635 91177308-0d34-0410-b5e6-96231b3b80d8

regenerate

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47634 91177308-0d34-0410-b5e6-96231b3b80d8

Remove unncessary ReturnInst constructors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47633 91177308-0d34-0410-b5e6-96231b3b80d8

Use a smallvector for inactiveCounts and initialize it lazily
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.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47631 91177308-0d34-0410-b5e6-96231b3b80d8

Rename PrintableName to Name.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47629 91177308-0d34-0410-b5e6-96231b3b80d8

Update per review. Patch by Mikhail Glushenkov!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47628 91177308-0d34-0410-b5e6-96231b3b80d8

Change "Name" to "AsmName" in the target register info. Gee, a refactoring tool
would have been a Godsend here!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47625 91177308-0d34-0410-b5e6-96231b3b80d8

Enable -coalescer-commute-instrs by default.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47623 91177308-0d34-0410-b5e6-96231b3b80d8

Do not treat Lexer.cpp as a generated file any more, let svn show it with a question mark if it is around.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47622 91177308-0d34-0410-b5e6-96231b3b80d8

first commit of new LTO system.  It is not hooked up in the llvm/tools/Makefile, so no one will build it be default yet

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47621 91177308-0d34-0410-b5e6-96231b3b80d8

Use SmallVector while constructing ReturnInst.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47619 91177308-0d34-0410-b5e6-96231b3b80d8

Avoid const_casts

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47616 91177308-0d34-0410-b5e6-96231b3b80d8

fix http://llvm.org/bugs/show_bug.cgi?id=2097

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47615 91177308-0d34-0410-b5e6-96231b3b80d8

Remove unnecessary getOperand/setOperand overriders.
Simplify getReturnValue()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47614 91177308-0d34-0410-b5e6-96231b3b80d8

Avoid aborting on invalid shift counts.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47612 91177308-0d34-0410-b5e6-96231b3b80d8

Unify to ReturnInst::init() member functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47611 91177308-0d34-0410-b5e6-96231b3b80d8

Fix for pr2093: direct operands aren't necessarily addresses, so don't
try to simplify them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47610 91177308-0d34-0410-b5e6-96231b3b80d8

fix this test so that the fn name doesn't match the regex

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47608 91177308-0d34-0410-b5e6-96231b3b80d8

Optimize most common case by using single RetVal in ReturnInst.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47607 91177308-0d34-0410-b5e6-96231b3b80d8

Correct function comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47606 91177308-0d34-0410-b5e6-96231b3b80d8

Fix PR2096, a regression introduced with my patch last night.  This
also fixes cfrac, flops, and 175.vpr

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47605 91177308-0d34-0410-b5e6-96231b3b80d8

Really feed llvm-as with the testcase, do not let it read from stdin. This fixes the hangs seen on solaris10.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47604 91177308-0d34-0410-b5e6-96231b3b80d8

unbreak check-one
by supplying a dummy
"verbose" procedure

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47603 91177308-0d34-0410-b5e6-96231b3b80d8

While hunting for two hanging tests,

(on solaris10, which are:
  CodeGen/PowerPC/frounds.ll
  Transforms/InstCombine/2008-02-23-MulSub.ll)

I needed a tool to figure out which one is the guilty.

To this end I have added a verbosity
option to the test/Makefile.

It can be invoked thus:

gmake check TESTSUITE=CodeGen/PowerPC VERBOSE="-v -v"

(The number of "-v"s specifies the verbosity level.
 Instead of "-v" other aliases can be specified,
 please consult the dejagnu docs for info.)

At level >= 2 following line is logged for each
test, before running it:

ABOUT TO RUN: <test>.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47602 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a nasty bug in LegalizeTypes (spotted in
CodeGen/PowerPC/illegal-element-type.ll): suppose
a node X is processed, and processing maps it to
a node Y.  Then X continues to exist in the DAG,
but with no users.  While processing some other
node, a new node may be created that happens to
be equal to X, and thus X will be reused rather
than a truly new node.  This can cause X to
"magically reappear", and since it is in the
Processed state in will not be reprocessed, so
at the end of type legalization the illegal node
X can still be present.  The solution is to replace
X with Y whenever X gets resurrected like this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47601 91177308-0d34-0410-b5e6-96231b3b80d8

De-tabify

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47600 91177308-0d34-0410-b5e6-96231b3b80d8

De-tabify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47599 91177308-0d34-0410-b5e6-96231b3b80d8

De-tabify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47598 91177308-0d34-0410-b5e6-96231b3b80d8

Detabify

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47597 91177308-0d34-0410-b5e6-96231b3b80d8

Detabify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47596 91177308-0d34-0410-b5e6-96231b3b80d8

De-tabify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47595 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for intermodule tail calls on x86/32bit with
GOT-style position independent code. Before only tail calls to
protected/hidden functions within the same module were optimized.
Now all function calls are tail call optimized.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47594 91177308-0d34-0410-b5e6-96231b3b80d8

Change the lowering of arguments for tail call optimized
calls. Before arguments that could overwrite each other were
explicitly lowered to a stack slot, not giving the register allocator
a chance to optimize. Now a sequence of copyto/copyfrom virtual
registers ensures that arguments are loaded in (virtual) registers
before they are lowered to the stack slot (and might overwrite each
other). Also parameter stack slots are marked mutable for
(potentially) tail calling functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47593 91177308-0d34-0410-b5e6-96231b3b80d8

This is possible:
vr1 = extract_subreg vr2, 3
...
vr3 = extract_subreg vr1, 2
The end result is vr3 is equal to vr2 with subidx 2.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47592 91177308-0d34-0410-b5e6-96231b3b80d8

Fix isNegatibleForFree to not return true for ConstantFP nodes
after legalize.  Just because a constant is legal (e.g. 0.0 in SSE)
doesn't mean that its negated value is legal (-0.0).  We could make
this stronger by checking to see if the negated constant is actually
legal post negation, but it doesn't seem like a big deal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47591 91177308-0d34-0410-b5e6-96231b3b80d8

Fix PR2076. CodeGenPrepare now sinks address computation for inline asm memory
operands into inline asm block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47589 91177308-0d34-0410-b5e6-96231b3b80d8

Refactor inline asm constraint matching code out of SDIsel into TargetLowering.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47587 91177308-0d34-0410-b5e6-96231b3b80d8

Remove debugging help.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47585 91177308-0d34-0410-b5e6-96231b3b80d8

test svn commit capability

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47584 91177308-0d34-0410-b5e6-96231b3b80d8

Update bitcode reader and writer to handle multiple return values.
Take 2.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47583 91177308-0d34-0410-b5e6-96231b3b80d8

Pass const vectors by reference.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47577 91177308-0d34-0410-b5e6-96231b3b80d8

Add missing include (for ptrdiff_t).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47576 91177308-0d34-0410-b5e6-96231b3b80d8

Revert the assert for MUL_LOHI with an unused high result; Chris
pointed out that this isn't correct at -O0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47575 91177308-0d34-0410-b5e6-96231b3b80d8

when making bytecode modules, link as library

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47574 91177308-0d34-0410-b5e6-96231b3b80d8

Revise previous patch per review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47573 91177308-0d34-0410-b5e6-96231b3b80d8

Factor the assert for indexed loads/stores out of LoadSDNode
and StoreSDNode into LSBaseSDNode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47570 91177308-0d34-0410-b5e6-96231b3b80d8

Add an assert to verify that we don't see an
{S,U}MUL_LOHI with an unused high value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47569 91177308-0d34-0410-b5e6-96231b3b80d8

Remove the hack that turned an {S,U}MUL_LOHI with an unused high
result into a MUL late in the X86 codegen process. ISD::MUL is
once again Legal on X86, so this is no longer needed. And, the
hack was suboptimal; see PR1874 for details.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47567 91177308-0d34-0410-b5e6-96231b3b80d8

Make some static variables const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47566 91177308-0d34-0410-b5e6-96231b3b80d8

add an assertion to catch a null PATypeHolder, patch by Erick Tryzelaar

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47563 91177308-0d34-0410-b5e6-96231b3b80d8

Convert MaskedValueIsZero and all its users to use APInt. Also add
a SignBitIsZero function to simplify a common use case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47561 91177308-0d34-0410-b5e6-96231b3b80d8

add LSBaseSDNode::getOffset at Dan's request.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47558 91177308-0d34-0410-b5e6-96231b3b80d8

All remat'ed loads cannot be folded into two-address code. Not just argument loads. This change doesn't really have any impact on codegen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47557 91177308-0d34-0410-b5e6-96231b3b80d8

prune #includes, fit in 80 cols.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47556 91177308-0d34-0410-b5e6-96231b3b80d8

make this self contained.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47555 91177308-0d34-0410-b5e6-96231b3b80d8

Expand removal of MMX memory copies to allow 1 level
of TokenFactor underneath chain (seems to be enough)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47554 91177308-0d34-0410-b5e6-96231b3b80d8

Add a new method to SDNode/SDOperand named getValueSizeInBits
that combines getValueType and MVT::getSizeInBits, since this
occurrs frequently.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47552 91177308-0d34-0410-b5e6-96231b3b80d8

In debug builds check that the key property holds: all
result and operand types are legal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47546 91177308-0d34-0410-b5e6-96231b3b80d8

Correctly determine whether a argument load can be folded into its uses.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47545 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an issue where GVN was performing the return slot optimization when it was
not safe.  This is fixed by more aggressively checking that the return slot is
not used elsewhere in the function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47544 91177308-0d34-0410-b5e6-96231b3b80d8

Temporarily reverting 46959.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47542 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an issue where GVN would try to use an instruction before its definition when performing return slot optimization.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47541 91177308-0d34-0410-b5e6-96231b3b80d8

Add support to LegalizeTypes for building legal vectors
out of illegal elements (BUILD_VECTOR).  Uses and beefs
up BUILD_PAIR, though it didn't really have to.  Like
most of LegalizeTypes, does not support soft-float.
This cures all "make check" vector building failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47537 91177308-0d34-0410-b5e6-96231b3b80d8

Some platforms use the same name for 32-bit and 64-bit registers (like
%r3 on PPC) in their ASM files. However, it's hard for humans to read
during debugging. Adding a new field to the register data that lets you
specify a different name to be printed than the one that goes into the
ASM file -- %x3 instead of %r3, for instance.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47534 91177308-0d34-0410-b5e6-96231b3b80d8

Backing out r47521 for now. This has broken a number of tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47533 91177308-0d34-0410-b5e6-96231b3b80d8

Merge current work back to tree to minimize diffs and drift. Major highlights
for CellSPU modifications:

- SPUInstrInfo.td refactoring: "multiclass" really is _your_ friend.
- Other improvements based on refactoring effort in SPUISelLowering.cpp,
  esp. in SPUISelLowering::PerformDAGCombine(), where zero amount shifts and
  rotates are now eliminiated, other scalar-to-vector-to-scalar silliness
  is also eliminated.
- 64-bit operations are being implemented, _muldi3.c gcc runtime now
  compiles and generates the right code. More work still needs to be done.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47532 91177308-0d34-0410-b5e6-96231b3b80d8

Testcase for Revision 47478.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47531 91177308-0d34-0410-b5e6-96231b3b80d8

Rematerialization logic was overly conservative when it comes to loads from fixed stack slots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47529 91177308-0d34-0410-b5e6-96231b3b80d8

Update test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47527 91177308-0d34-0410-b5e6-96231b3b80d8

If remating a machine instr with virtual register operand, make sure the vr is avaliable at all uses regardless of whether it would be folded.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47526 91177308-0d34-0410-b5e6-96231b3b80d8

Remat of pic loads are now on by default.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47525 91177308-0d34-0410-b5e6-96231b3b80d8

Turning on remat of pic loads.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47524 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a bug that caused opt and other tools to silently ignore
invalid command-line options.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47523 91177308-0d34-0410-b5e6-96231b3b80d8

No need recognize load from a fixed argument slot as re-materializable. LiveIntervalAnalysis already handles it as a special case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47522 91177308-0d34-0410-b5e6-96231b3b80d8

Properly read and write bitcodes for multiple return values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47521 91177308-0d34-0410-b5e6-96231b3b80d8

Recognize loads of arguments as re-materializable first. Therefore if isReallyTriviallyReMaterializable() returns true it doesn't confuse it as a "normal" re-materializable instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47520 91177308-0d34-0410-b5e6-96231b3b80d8

Regenerate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47519 91177308-0d34-0410-b5e6-96231b3b80d8

Use isa check instead of getTypeID() check.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47518 91177308-0d34-0410-b5e6-96231b3b80d8

Fixed buffer overflow reported by Argiris Kirtzidis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47517 91177308-0d34-0410-b5e6-96231b3b80d8