Extend the builder interface to use the new instruction positioning code.

This adds support for instruction iterators, as well as rewriting the
builder code to use these new functions. This lets us eliminate the C
bindings for moving around the builder.

Patch by Erick Tryzelaar!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48774 91177308-0d34-0410-b5e6-96231b3b80d8

Fix PR2062: Don't build Intrinsics.gen in the source directory. Do it in the
object directory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48766 91177308-0d34-0410-b5e6-96231b3b80d8

Disable building llvm-upgrade.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48764 91177308-0d34-0410-b5e6-96231b3b80d8

Make a note of the fact that EmitOwnedPtr() has nothing to do with the
OwningPtr<> class.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48763 91177308-0d34-0410-b5e6-96231b3b80d8

Byebye llvm-upgrade!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48762 91177308-0d34-0410-b5e6-96231b3b80d8

lastRegisterUse() should ignore identity copies. Those will be erased.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48759 91177308-0d34-0410-b5e6-96231b3b80d8

check struct layout

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48758 91177308-0d34-0410-b5e6-96231b3b80d8

Remove an unneeded test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48755 91177308-0d34-0410-b5e6-96231b3b80d8

Add \t after .set. Fix by Jay Freeman.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48753 91177308-0d34-0410-b5e6-96231b3b80d8

If the coalescer commuted a def MI to allow coalescing, it can changed a previously coalesced copy into an non-identity copy.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48752 91177308-0d34-0410-b5e6-96231b3b80d8

Add an assertion to catch register of illegal class.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48751 91177308-0d34-0410-b5e6-96231b3b80d8

Use the bit size of the operand instead of the hard-coded 32 to generate the
mask.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48750 91177308-0d34-0410-b5e6-96231b3b80d8

With debug info, there are nameless constant global values.  do not crash when we hit one

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48749 91177308-0d34-0410-b5e6-96231b3b80d8

Add file to project.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48747 91177308-0d34-0410-b5e6-96231b3b80d8

- SSE4.1 extractfps extracts a f32 into a gr32 register. Very useful! Not. Fix the instruction specification and teaches lowering code to use it only when the only use is a store instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48746 91177308-0d34-0410-b5e6-96231b3b80d8

Oops.  I wanted the compile flags for C++, not the C preprocessor flags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48744 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r48676.  I had plans for using it, but now it's just dead code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48743 91177308-0d34-0410-b5e6-96231b3b80d8

void type is also a valid function return type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48740 91177308-0d34-0410-b5e6-96231b3b80d8

Remove #include<iostream>, which I was using for debugging.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48739 91177308-0d34-0410-b5e6-96231b3b80d8

Add incoming value from header only if phi node has any use inside the loop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48738 91177308-0d34-0410-b5e6-96231b3b80d8

Added Passes.html to the documentation list.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48736 91177308-0d34-0410-b5e6-96231b3b80d8

Fix typo. Clarify underlying message.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48734 91177308-0d34-0410-b5e6-96231b3b80d8

Fix test name.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48733 91177308-0d34-0410-b5e6-96231b3b80d8

apparently tclsh doesn't lex like bash.  Weird.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48732 91177308-0d34-0410-b5e6-96231b3b80d8

pass the option so this test tests the right thing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48731 91177308-0d34-0410-b5e6-96231b3b80d8

Add new test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48730 91177308-0d34-0410-b5e6-96231b3b80d8

Remove an unnecessary #include.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48729 91177308-0d34-0410-b5e6-96231b3b80d8

Remove incorrect comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48728 91177308-0d34-0410-b5e6-96231b3b80d8

Shrink the size of AllocationInst by using its SubclassData
field to store the alignment value instead of haing a
separate field.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48727 91177308-0d34-0410-b5e6-96231b3b80d8

APIntify SelectionDAG's EXTRACT_ELEMENT code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48726 91177308-0d34-0410-b5e6-96231b3b80d8

Increasing the inline limit from (overly conservative) 200 to 300. Given each BB costs 20 and each instruction costs 5, 200 means a 4 BB function + 24 instructions (actually less because caller's size also contributes to it).

Furthermore, double the limit when more than 10% of the callee instructions are vector instructions. Multimedia kernels tend to love inlining.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48725 91177308-0d34-0410-b5e6-96231b3b80d8

Small format/grammar changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48723 91177308-0d34-0410-b5e6-96231b3b80d8

Grammar Nazi strikes! :-)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48722 91177308-0d34-0410-b5e6-96231b3b80d8

Update Function type documentation to clarify how multiple return values are supported.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48721 91177308-0d34-0410-b5e6-96231b3b80d8

Temporarily disabling memset forming optimization. Add an option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48720 91177308-0d34-0410-b5e6-96231b3b80d8

Be sure to remove intervals after we've joined them.  Also, remove some duplicated code.

With this pass, StrongPHIElim can compile very simple testcases correctly.  There's still a ways
to go before it's ready for prime time, though.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48719 91177308-0d34-0410-b5e6-96231b3b80d8

Ignoring per-user file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48718 91177308-0d34-0410-b5e6-96231b3b80d8

Ignoring generated files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48717 91177308-0d34-0410-b5e6-96231b3b80d8

Transform (zext (or (icmp), (icmp))) to (or (zext (cimp), (zext icmp))) if at least one of the (zext icmp) can be transformed to eliminate an icmp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48715 91177308-0d34-0410-b5e6-96231b3b80d8

Remove duplicated entries.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48714 91177308-0d34-0410-b5e6-96231b3b80d8

Another typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48713 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48712 91177308-0d34-0410-b5e6-96231b3b80d8

Objective Caml bindings for basic block, function, global, and arg iterators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48711 91177308-0d34-0410-b5e6-96231b3b80d8

Minor typo fixes. Also add another FIXME.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48710 91177308-0d34-0410-b5e6-96231b3b80d8

Update project to reflect deleted files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48709 91177308-0d34-0410-b5e6-96231b3b80d8

Preserve calling convention during function cloning

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48708 91177308-0d34-0410-b5e6-96231b3b80d8

Add license header

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48707 91177308-0d34-0410-b5e6-96231b3b80d8

Add Win64 compilation callback. This allows easy examples to be JITed on Win64!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48706 91177308-0d34-0410-b5e6-96231b3b80d8

Use C-style comments :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48705 91177308-0d34-0410-b5e6-96231b3b80d8

Provide a JIT selector on win64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48704 91177308-0d34-0410-b5e6-96231b3b80d8

Hack out the PIC mode on Win64 targets. This needs to be investigated later.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48703 91177308-0d34-0410-b5e6-96231b3b80d8

Code cleanup. Provide generic way of selecting JIT pointer bitwidth regardless
of compiler used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48702 91177308-0d34-0410-b5e6-96231b3b80d8

Remove old-standing obsolete code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48701 91177308-0d34-0410-b5e6-96231b3b80d8

Add Mikhail to credits file

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48700 91177308-0d34-0410-b5e6-96231b3b80d8

Add first proof-of-concept universal compiler driver framework based
on ideas mentioned in PR686.
Written by Mikhail Glushenkov and contributed by Codedgers, Inc.

Old llvmc will be removed soon after new one will have all its properties.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48699 91177308-0d34-0410-b5e6-96231b3b80d8

New testcase.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48697 91177308-0d34-0410-b5e6-96231b3b80d8

Don't strip things when we're doing a debug build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48696 91177308-0d34-0410-b5e6-96231b3b80d8

Honour built-in defines on win64 targets for automatically subtarget recognize.
Force stack alignment to 16 bytes on win targets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48695 91177308-0d34-0410-b5e6-96231b3b80d8

Recognize "windows" in target triple, not only "win32"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48694 91177308-0d34-0410-b5e6-96231b3b80d8

Use normal naming convention for test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48693 91177308-0d34-0410-b5e6-96231b3b80d8

Add information about callee-saved registers on Win64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48692 91177308-0d34-0410-b5e6-96231b3b80d8

Add convenient helper for win64 check. Simplify things slightly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48691 91177308-0d34-0410-b5e6-96231b3b80d8

Initial support for Win64 calling conventions. Still in early state.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48690 91177308-0d34-0410-b5e6-96231b3b80d8

Register EH frames emitted in JIT when using gcc unwinding runtime

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48688 91177308-0d34-0410-b5e6-96231b3b80d8

Fix merge error

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48687 91177308-0d34-0410-b5e6-96231b3b80d8

Add testcase for prev. commit. Minor fixes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48686 91177308-0d34-0410-b5e6-96231b3b80d8

Aliasing chains cleanups: update langref, add check into verifier

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48685 91177308-0d34-0410-b5e6-96231b3b80d8

Support chained aliases for LLVM IR printing. This fixes PR2145

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48684 91177308-0d34-0410-b5e6-96231b3b80d8

Another comments fixing

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48683 91177308-0d34-0410-b5e6-96231b3b80d8

Slightly increase default set size. It's cheap and won't hurt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48682 91177308-0d34-0410-b5e6-96231b3b80d8

Typo fixes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48681 91177308-0d34-0410-b5e6-96231b3b80d8

implement an initial hack at a straight-line store -> memset optimization.
This fires dozens of times across spec and multisource, but I don't know
if it actually speeds stuff up.  Hopefully the testers will show something
nice :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48680 91177308-0d34-0410-b5e6-96231b3b80d8

implement the logic for memset insertion and store deletion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48679 91177308-0d34-0410-b5e6-96231b3b80d8

Add a comment, and fix a bug where AllocateRW recurred to AllocateRWX instead of itself.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48677 91177308-0d34-0410-b5e6-96231b3b80d8

Add an AllocateRW to match AllocateRWX.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48676 91177308-0d34-0410-b5e6-96231b3b80d8

Teach DAG combiner to commute commutable binary nodes in order to achieve sdisel CSE.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48673 91177308-0d34-0410-b5e6-96231b3b80d8

This is a partially implemented and currently disabled start of a store
merging optimization.  Nothing to see here, hopefully more later :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48670 91177308-0d34-0410-b5e6-96231b3b80d8

Don't include <map> in Pass.h, which doesn't need it. This requires
adding <map> to many files that actually do need it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48667 91177308-0d34-0410-b5e6-96231b3b80d8

Avoid calling array_endof in a static initializer, as it incurs
dynamic initialization.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48666 91177308-0d34-0410-b5e6-96231b3b80d8

Specialize FORCE_DEFINING_FILE_TO_BE_LINKED using a GCC trick
to avoid using constructor calls for static objects. This reduces
the number of objects requiring static constructors in a typical
LLVM build by around 20%.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48665 91177308-0d34-0410-b5e6-96231b3b80d8

Fix -view-sunit-dags to support cross-rc-copy nodes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48664 91177308-0d34-0410-b5e6-96231b3b80d8

a happier version of happiness

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48663 91177308-0d34-0410-b5e6-96231b3b80d8

the size of a smallvector shouldn't be part of the interface to these methods.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48662 91177308-0d34-0410-b5e6-96231b3b80d8

make gvn marginally faster by reallocating the lastSeenLoad map for
each basic block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48660 91177308-0d34-0410-b5e6-96231b3b80d8

Minor cleanups and shrinkification.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48658 91177308-0d34-0410-b5e6-96231b3b80d8

Handle getresult instructions in different basic blocks
from their aggregate operands by moving the getresult
instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48657 91177308-0d34-0410-b5e6-96231b3b80d8

Restore this assert now that the livevar bug is fixed.
This verifies kill info for "ret" fp operands is right.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48656 91177308-0d34-0410-b5e6-96231b3b80d8

Testcase for PR2160.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48655 91177308-0d34-0410-b5e6-96231b3b80d8

A couple of kill marker maintainence bug.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48653 91177308-0d34-0410-b5e6-96231b3b80d8

clean up and fix the call section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48651 91177308-0d34-0410-b5e6-96231b3b80d8

cleanups for the getresult section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48650 91177308-0d34-0410-b5e6-96231b3b80d8

FunctionExtractorPass has been superceded by GVExtractorPass

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48648 91177308-0d34-0410-b5e6-96231b3b80d8

Make it possible to get an empty struct using
the new StructType::get method.  The second NULL
is to pacify the gcc warning mechanism.  This
patch compiles but is otherwise untested.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48645 91177308-0d34-0410-b5e6-96231b3b80d8

Introduce a new node for holding call argument
flags.  This is needed by the new legalize types
infrastructure which wants to expand the 64 bit
constants previously used to hold the flags on
32 bit machines.  There are two functional changes:
(1) in LowerArguments, if a parameter has the zext
attribute set then that is marked in the flags;
before it was being ignored; (2) PPC had some bogus
code for handling two word arguments when using the
ELF 32 ABI, which was hard to convert because of
the bogusness.  As suggested by the original author
(Nicolas Geoffray), I've disabled it for the moment.
Tested with "make check" and the Ada ACATS testsuite.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48640 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the build for gcc-4.2.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48639 91177308-0d34-0410-b5e6-96231b3b80d8

remove Evan's "ugly hack" that sorta attempted to get
x86-64 return conventions correct, but was never enabled.
We can now do the "right thing" with multiple return values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48635 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for calls that return two FP values in
ST(0)/ST(1).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48634 91177308-0d34-0410-b5e6-96231b3b80d8

disable a bogus assertion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48633 91177308-0d34-0410-b5e6-96231b3b80d8

Enable support for returning two long-double values in ST(0)/ST(1).
This allows us to compile fp-stack-2results.ll into:

_test:
fldz
fld1
ret

which returns 1 in ST(0) and 0 in ST(1).  This is needed for x86-64
_Complex long double.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48632 91177308-0d34-0410-b5e6-96231b3b80d8

Teach masked value is zero about add and sub, and use MVIZ to
simplify things like (X & 4) >> 1 == 2  --> (X & 4) == 4.

since it is obvious that the shift doesn't remove any bits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48631 91177308-0d34-0410-b5e6-96231b3b80d8