llvm-svn: 26536

fold (shl (add x, c1), c2) -> (add (shl x, c2), c1<<c2)

This allows us to compile CodeGen/PowerPC/addi-reassoc.ll into:

_test1:
        slwi r2, r4, 4
        add r2, r2, r3
        lwz r3, 36(r2)
        blr
_test2:
        mulli r2, r4, 5
        add r2, r2, r3
        lbz r2, 11(r2)
        extsb r3, r2
        blr

instead of:

_test1:
        addi r2, r4, 2
        slwi r2, r2, 4
        add r2, r3, r2
        lwz r3, 4(r2)
        blr
_test2:
        addi r2, r4, 2
        mulli r2, r2, 5
        add r2, r3, r2
        lbz r2, 1(r2)
        extsb r3, r2
        blr

llvm-svn: 26535

llvm-svn: 26529

Testcase here: Transforms/Reassociate/mulfactor.ll

llvm-svn: 26524

llvm-svn: 26523

llvm-svn: 26522

llvm-svn: 26521

llvm-svn: 26520

This implements Transforms/InstCombine/add.ll:test31

llvm-svn: 26519

rep/stos and rep/mov if the count is not a constant. We could do
  rep/stosl; and $count, 3; rep/stosb
For now, I will lower them to memset / memcpy calls. We will revisit this after
a little bit experiment.

Also need to take care of the trailing bytes even if the count is a constant.
Since the max. number of trailing bytes are 3, we will simply issue loads /
stores.

llvm-svn: 26517

don't assume that A[1][0] and A[0][i] can't alias.  "i" might be out of
range, or even negative.  This fixes a miscompilation of 188.ammp (which
does bad pointer tricks) with the new CFE.

Testcase here: Analysis/BasicAA/2006-03-03-BadArraySubscript.ll

llvm-svn: 26515

llvm-svn: 26513

llvm-svn: 26512

llvm-svn: 26509

llvm-svn: 26506

llvm-svn: 26505

llvm-svn: 26504

llvm-svn: 26503

llvm-svn: 26502

llvm-svn: 26501

llvm-svn: 26500

llvm-svn: 26498

llvm-svn: 26493

llvm-svn: 26490

llvm-svn: 26484

llvm-svn: 26482

llvm-svn: 26481

llvm-svn: 26479

PR709, and paving the way for future progress.

llvm-svn: 26476

llvm-svn: 26475

PR709, and paving the way for future progress.

Significantly refactor autoupgrading code, to handle the more complex case
(where we upgrade one argument in a function), and fix some bugs in it.

Testcase here: llvm/test/Regression/Bytecode/memcpy.ll

llvm-svn: 26474

llvm-svn: 26472

  a predecessor appearing more than once in the operand list was counted as
  multiple predecessor; priority1 should be updated during scheduling;
  CycleBound was updated after the node is inserted into priority queue; one
  of the tie breaking condition was flipped.
- Take into consideration of two address opcodes. If a predecessor is a def&use
  operand, it should have a higher priority.
- Scheduler should also favor floaters, i.e. nodes that do not have real
  predecessors such as MOV32ri.
- The scheduling fixes / tweaks fixed bug 478:
        .text
        .align  4
        .globl  _f
_f:
        movl 4(%esp), %eax
        movl 8(%esp), %ecx
        movl %eax, %edx
        imull %ecx, %edx
        imull %eax, %eax
        imull %ecx, %ecx
        addl %eax, %ecx
        leal (%ecx,%edx,2), %eax
        ret

  It is also a slight performance win (1% - 3%) for most tests.

llvm-svn: 26470

pointed out: realize the AND can provide factors and look through Casts.

llvm-svn: 26469

llvm-svn: 26466

span multiple lines.

llvm-svn: 26463

llvm-svn: 26462

multi-line comments.

llvm-svn: 26461

in the dag combiner on 176.gcc on x86.

llvm-svn: 26459

debug forms.

llvm-svn: 26455