Also fold MOV8r0, MOV16r0, MOV32r0  + store to MOV8mi, MOV16mi, and MOV32mi.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26817 91177308-0d34-0410-b5e6-96231b3b80d8

Add some missing entries to X86RegisterInfo::foldMemoryOperand(). e.g.
ADD32ri8.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26816 91177308-0d34-0410-b5e6-96231b3b80d8

- Nuke 16-bit SBB instructions. We'll never use them.
- Nuke a bogus comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26815 91177308-0d34-0410-b5e6-96231b3b80d8

Remove BRTWOWAY*
Make the PPC backend not dependent on BRTWOWAY_CC and make the branch
selector smarter about the code it generates, fixing a case in the
readme.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26814 91177308-0d34-0410-b5e6-96231b3b80d8

remove dead variable

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26813 91177308-0d34-0410-b5e6-96231b3b80d8

add an assert to get a slightly better msg about this problem

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26812 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a problem fully scalarizing values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26811 91177308-0d34-0410-b5e6-96231b3b80d8

A new entry.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26810 91177308-0d34-0410-b5e6-96231b3b80d8

Notes on how to kill the eeevil brtwoway, and make ppc branch selector
more target independant, generate better code, and be less conservative.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26809 91177308-0d34-0410-b5e6-96231b3b80d8

Strangely, calls clobber call-clobbered vector regs.  Whodathoughtit?

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26808 91177308-0d34-0410-b5e6-96231b3b80d8

add a note

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26807 91177308-0d34-0410-b5e6-96231b3b80d8

teach the ppc backend how to spill/reload vector regs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26806 91177308-0d34-0410-b5e6-96231b3b80d8

add callee saved vector regs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26805 91177308-0d34-0410-b5e6-96231b3b80d8

Bug fix: condition inverted.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26804 91177308-0d34-0410-b5e6-96231b3b80d8

For each loop, keep track of all the IV expressions inserted indexed by
stride. For a set of uses of the IV of a stride which is a multiple
of another stride, do not insert a new IV expression. Rather, reuse the
previous IV and rewrite the uses as uses of IV expression multiplied by
the factor.

e.g.
x = 0 ...; x ++
y = 0 ...; y += 4
then use of y can be rewritten as use of 4*x for x86.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26803 91177308-0d34-0410-b5e6-96231b3b80d8

Added a way for TargetLowering to specify what values can be used as the
scale component of the target addressing mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26802 91177308-0d34-0410-b5e6-96231b3b80d8

in functions that use a lot of callee saved regs, this can be more than
5 instructions away.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26801 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for copying registers. still needed: spilling and reloading them

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26800 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for CopyFromReg from vector values.  Note: this doesn't support
illegal vector types yet!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26799 91177308-0d34-0410-b5e6-96231b3b80d8

Teach CreateRegForValue how to handle vector types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26798 91177308-0d34-0410-b5e6-96231b3b80d8

set TransformToType correctly for vector types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26797 91177308-0d34-0410-b5e6-96231b3b80d8

Add markers for the first and last vector value type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26796 91177308-0d34-0410-b5e6-96231b3b80d8

Another case we could do better on.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26795 91177308-0d34-0410-b5e6-96231b3b80d8

New testcase, the new CFE compiles this into insertelement instructions, the
old one crashes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26794 91177308-0d34-0410-b5e6-96231b3b80d8

Save/restore VRSAVE once per function, not once per block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26793 91177308-0d34-0410-b5e6-96231b3b80d8

Typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26792 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a typo Eric van Riet Paap noticed

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26791 91177308-0d34-0410-b5e6-96231b3b80d8

Add a RUN: line so this test doesn't fail.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26790 91177308-0d34-0410-b5e6-96231b3b80d8

add support for the bitconvert node

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26789 91177308-0d34-0410-b5e6-96231b3b80d8

add support for vector->vector casts

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26788 91177308-0d34-0410-b5e6-96231b3b80d8

Add a note, this code should be moved to the dag combiner.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26787 91177308-0d34-0410-b5e6-96231b3b80d8

minor changes so that GCC builtin can be specified before the Intrinsic info,
add some more PPC intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26786 91177308-0d34-0410-b5e6-96231b3b80d8

Teach the strip pass to strip type names in addition to value names.  This
is fallout from the type/value split in the symtab long long ago :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26785 91177308-0d34-0410-b5e6-96231b3b80d8

allow the GCCBuiltinName field to be optional

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26784 91177308-0d34-0410-b5e6-96231b3b80d8

Update doc to reflect changes in subprogram and block descriptors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26783 91177308-0d34-0410-b5e6-96231b3b80d8

Expand subprogram and added block descriptor.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26782 91177308-0d34-0410-b5e6-96231b3b80d8

no heap is happening here

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26781 91177308-0d34-0410-b5e6-96231b3b80d8

better check

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26780 91177308-0d34-0410-b5e6-96231b3b80d8

remove qsort for now

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26779 91177308-0d34-0410-b5e6-96231b3b80d8

Also requires -mattr=-sse3

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26778 91177308-0d34-0410-b5e6-96231b3b80d8

allow field sensitivity to be a tunable parameter

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26777 91177308-0d34-0410-b5e6-96231b3b80d8

Update scheduling info for vrsave instruction

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26776 91177308-0d34-0410-b5e6-96231b3b80d8

Handle one offset with growth case seen in povray.  Namely, if we have an offset,
and the offset lands at a field boundary in the old type, construct a new type,
copying the fields masked by the offset from the old type, and unify with that.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26775 91177308-0d34-0410-b5e6-96231b3b80d8

improve mem intrinsics and add a few things povray uses

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26774 91177308-0d34-0410-b5e6-96231b3b80d8

Fix VC++ build error.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26773 91177308-0d34-0410-b5e6-96231b3b80d8

remove typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26772 91177308-0d34-0410-b5e6-96231b3b80d8

Autogenerate a table of intrinsic names, so we can map from intrinsic ID to
LLVM intrinsic function name.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26771 91177308-0d34-0410-b5e6-96231b3b80d8

Autogenerate code to map from GCC builtin to LLVM intrinsic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26770 91177308-0d34-0410-b5e6-96231b3b80d8

Set TargetPrefix on target-specific intrinsics.  That way, in theory, different
targets could have different implemenations of a __builtin_foo gcc intrinsic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26769 91177308-0d34-0410-b5e6-96231b3b80d8

Fixed a typo in the vsetfp examples.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26768 91177308-0d34-0410-b5e6-96231b3b80d8

Split the 'vset' instruction into two instructions, 'vsetint' and
'vsetfp', to reflect the fact that the semantics are different for
integer and fp values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26767 91177308-0d34-0410-b5e6-96231b3b80d8

Update this to match the documentation in the GC doc and to match actual
practice.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26766 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the gcread/gcwrite intrinsic specifications, fixing
CodeGen/Generic/GC/alloc_loop.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26765 91177308-0d34-0410-b5e6-96231b3b80d8

Bugfix, unbreaking CodeGen/PowerPC/cttz.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26764 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an et-forest memory leak.  Patch by Daniel Berlin.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26763 91177308-0d34-0410-b5e6-96231b3b80d8

add a note

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26762 91177308-0d34-0410-b5e6-96231b3b80d8

Reflect the fact that empty strings can be expressed as null.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26761 91177308-0d34-0410-b5e6-96231b3b80d8

1. Use null for serialized empty strings.
2. Allow for user defined debug descriptors.
3. Allow for user augmented fields on debug descriptors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26760 91177308-0d34-0410-b5e6-96231b3b80d8

Bring debugging information up to date.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26759 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an off by one error that caused PPC LLC failures last night.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26758 91177308-0d34-0410-b5e6-96231b3b80d8

Implement a FIXME, recusively reassociating
A*A*B + A*A*C   -->   A*(A*B+A*C)   -->   A*(A*(B+C))

This implements Reassociate/mul-factor3.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26757 91177308-0d34-0410-b5e6-96231b3b80d8

new testcase from a FIXME in the code

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26756 91177308-0d34-0410-b5e6-96231b3b80d8

extract some code into a method, no functionality change

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26755 91177308-0d34-0410-b5e6-96231b3b80d8

transformation implemented

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26754 91177308-0d34-0410-b5e6-96231b3b80d8

Promote shifts by a constant to multiplies so that we can reassociate
(x<<1)+(y<<1) -> (X+Y)<<1.  This implements
Transforms/Reassociate/shift-factor.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26753 91177308-0d34-0410-b5e6-96231b3b80d8

Test that we can reassociate (x<<1)+(y<<1) -> (X+Y)<<1.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26752 91177308-0d34-0410-b5e6-96231b3b80d8

Allow a sequence of digits at the end of the directory name when trying to
determine the top directory. This allows the tool to find the correct top
directory when you have something like:

   /x/llvm
   /x/llvm2
   /x/llvm3

Previously the scripts would always find /x/llvm even if you were in the
llvm2 or llvm3 directory because the pattern didn't allow the digits at
the end of the path.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26751 91177308-0d34-0410-b5e6-96231b3b80d8

Change llvm.cs.uiuc.edu -> llvm.org

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26750 91177308-0d34-0410-b5e6-96231b3b80d8

Add a newline at the end to avoid gcc warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26749 91177308-0d34-0410-b5e6-96231b3b80d8

Convert llvm.cs.uiuc.edu -> llvm.org

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26748 91177308-0d34-0410-b5e6-96231b3b80d8

Change docs for llvm.cs.uiuc.edu -> llvm.org

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26747 91177308-0d34-0410-b5e6-96231b3b80d8

Changes docs for llvm.cs.uiuc.edu -> llvm.org

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26746 91177308-0d34-0410-b5e6-96231b3b80d8

1. Handle removal of all arguments for a morphed intrinsic.
2. Remove the declaration of llvm.dbg.declare.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26745 91177308-0d34-0410-b5e6-96231b3b80d8

Remove the use of llvm.dbg.declare.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26744 91177308-0d34-0410-b5e6-96231b3b80d8

PPC LSR pass should use target lowering hooks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26743 91177308-0d34-0410-b5e6-96231b3b80d8

Added getTargetLowering() to TargetMachine. Refactored targets to support this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26742 91177308-0d34-0410-b5e6-96231b3b80d8

Update

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26741 91177308-0d34-0410-b5e6-96231b3b80d8

Add LSR hooks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26740 91177308-0d34-0410-b5e6-96231b3b80d8

Added getTargetLowering() - returns DAG lowering info.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26739 91177308-0d34-0410-b5e6-96231b3b80d8

Added target lowering hooks which LSR consults to make more intelligent
transformation decisions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26738 91177308-0d34-0410-b5e6-96231b3b80d8

Handle builtins that directly correspond to GCC builtins.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26737 91177308-0d34-0410-b5e6-96231b3b80d8

emit a mapping from LLVM intrinsic -> GCC builtins.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26736 91177308-0d34-0410-b5e6-96231b3b80d8

Verify that packed type operands have the right size and base type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26735 91177308-0d34-0410-b5e6-96231b3b80d8

Add a first ppc altivec intrinsic.  Add packed type support.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26734 91177308-0d34-0410-b5e6-96231b3b80d8

For functions that use vector registers, save VRSAVE, mark used
registers, and update it on entry to each function, then restore it on exit.

This compiles:

void func(vfloat *a, vfloat *b, vfloat *c) {
        *a = *b * *c + *c;
}

to this:

_func:
        mfspr r2, 256
        oris r6, r2, 49152
        mtspr 256, r6
        lvx v0, 0, r5
        lvx v1, 0, r4
        vmaddfp v0, v1, v0, v0
        stvx v0, 0, r3
        mtspr 256, r2
        blr

GCC produces this (which has additional stack accesses):

_func:
        mfspr r0,256
        stw r0,-4(r1)
        oris r0,r0,0xc000
        mtspr 256,r0
        lvx v0,0,r5
        lvx v1,0,r4
        lwz r12,-4(r1)
        vmaddfp v0,v0,v1,v0
        stvx v0,0,r3
        mtspr 256,r12
        blr

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26733 91177308-0d34-0410-b5e6-96231b3b80d8

Optimizing to calloc() to malloc() + memset() can be bad on some platforms.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26732 91177308-0d34-0410-b5e6-96231b3b80d8

make sure dead token factor nodes are removed by the dag combiner.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26731 91177308-0d34-0410-b5e6-96231b3b80d8

Bring makefile back into compliance with standard by using
$(Echo) instead of @echo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26730 91177308-0d34-0410-b5e6-96231b3b80d8

Handle the removal of the debug chain.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26729 91177308-0d34-0410-b5e6-96231b3b80d8

remove two implemented items

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26728 91177308-0d34-0410-b5e6-96231b3b80d8

Fold X+Y -> X|Y when safe.  This implements:
Regression/CodeGen/PowerPC/and_add.ll

a case that occurs with dynamic allocas of constant size.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26727 91177308-0d34-0410-b5e6-96231b3b80d8

new testcase

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26726 91177308-0d34-0410-b5e6-96231b3b80d8

I can't convince myself that this is safe, remove the recursive call.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26725 91177308-0d34-0410-b5e6-96231b3b80d8

add a couple of missing folds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26724 91177308-0d34-0410-b5e6-96231b3b80d8

For targets with FABS/FNEG support, lower copysign to an integer load,
a select and FABS/FNEG.

This speeds up a trivial (aka stupid) copysign benchmark I wrote from 6.73s
to 2.64s, woo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26723 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a couple of bugs that broke the alpha tester build

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26722 91177308-0d34-0410-b5e6-96231b3b80d8

Handle cracked instructions in dispatch group formation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26721 91177308-0d34-0410-b5e6-96231b3b80d8

Mark instructions that are cracked by the PPC970 decoder as such.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26720 91177308-0d34-0410-b5e6-96231b3b80d8

Several big changes:
1. Use flags on the instructions in the .td file to indicate the PPC970 unit
   type instead of a table in the .cpp file.  Much cleaner.
2. Change the hazard recognizer to build d-groups according to the actual
   algorithm used, not my flawed understanding of it.
3. Model "must be in the first slot" and "must be the only instr in a group"
   accurately.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26719 91177308-0d34-0410-b5e6-96231b3b80d8

Don't advance the hazard recognizer when there are no hazards and no instructions
to be emitted.

Don't add one to the latency of a completed instruction if the latency of the
op is 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@26718 91177308-0d34-0410-b5e6-96231b3b80d8