llvm-svn: 28148

llvm-svn: 28147

sign_extend_inreg operations.  Though ComputeNumSignBits is still rudimentary,
this is enough to compile this:

short test(short X, short x) {
  int Y = X+x;
  return (Y >> 1);
}
short test2(short X, short x) {
  int Y = (short)(X+x);
  return Y >> 1;
}

into:

_test:
        add r2, r3, r4
        srawi r3, r2, 1
        blr
_test2:
        add r2, r3, r4
        extsh r2, r2
        srawi r3, r2, 1
        blr

instead of:

_test:
        add r2, r3, r4
        srawi r2, r2, 1
        extsh r3, r2
        blr
_test2:
        add r2, r3, r4
        extsh r2, r2
        srawi r2, r2, 1
        extsh r3, r2
        blr

llvm-svn: 28146

This will certainly be enhanced in the future.

llvm-svn: 28145

llvm-svn: 28144

a cast immediately before a PHI node.

This fixes Regression/CodeGen/Generic/2006-05-06-GEP-Cast-Sink-Crash.ll

llvm-svn: 28143

llvm-svn: 28142

Make the "fold (and (cast A), (cast B)) -> (cast (and A, B))" transformation
only apply when both casts really will cause code to be generated.  If one or
both doesn't, then this xform doesn't remove a cast.

This fixes Transforms/InstCombine/2006-05-06-Infloop.ll

llvm-svn: 28141

llvm-svn: 28140

llvm-svn: 28139

llvm-svn: 28138

27,28c27
<       movzwl %di, %edi
<       movl %edi, %ebx
---
>       movw %di, %bx

llvm-svn: 28137

llvm-svn: 28136

llvm-svn: 28135

using them.

llvm-svn: 28134

llvm-svn: 28133

llvm-svn: 28132

llvm-svn: 28131

llvm-svn: 28130

generated:

        movl 8(%esp), %eax
        movl %eax, %edx
        addl $4316, %edx
        cmpb $1, %cl
        ja LBB1_2       #cond_false
LBB1_1: #cond_true
        movl L_QuantizationTables720$non_lazy_ptr, %ecx
        movl %ecx, (%edx)
        movl L_QNOtoQuantTableShift720$non_lazy_ptr, %edx
        movl %edx, 4460(%eax)
        ret
...

Now we generate:

        movl 8(%esp), %eax
        cmpb $1, %cl
        ja LBB1_2       #cond_false
LBB1_1: #cond_true
        movl L_QuantizationTables720$non_lazy_ptr, %ecx
        movl %ecx, 4316(%eax)
        movl L_QNOtoQuantTableShift720$non_lazy_ptr, %ecx
        movl %ecx, 4460(%eax)
        ret

... which uses one fewer register.

llvm-svn: 28129

llvm-svn: 28128

llvm-svn: 28127

llvm-svn: 28126

llvm-svn: 28125

llvm-svn: 28124

  // fold (and (sext x), (sext y)) -> (sext (and x, y))
  // fold (or  (sext x), (sext y)) -> (sext (or  x, y))
  // fold (xor (sext x), (sext y)) -> (sext (xor x, y))
  // fold (and (aext x), (aext y)) -> (aext (and x, y))
  // fold (or  (aext x), (aext y)) -> (aext (or  x, y))
  // fold (xor (aext x), (aext y)) -> (aext (xor x, y))

llvm-svn: 28123

        mov EAX, DWORD PTR [ESP + 4]
        mov ECX, DWORD PTR [EAX]
        mov EDX, ECX
        add EDX, EDX
        or EDX, ECX
        and EDX, -2147483648
        and ECX, 2147483647
        or EDX, ECX
        mov DWORD PTR [EAX], EDX
        ret

instead of:

        sub ESP, 4
        mov DWORD PTR [ESP], ESI
        mov EAX, DWORD PTR [ESP + 8]
        mov ECX, DWORD PTR [EAX]
        mov EDX, ECX
        add EDX, EDX
        mov ESI, ECX
        and ESI, -2147483648
        and EDX, -2147483648
        or EDX, ESI
        and ECX, 2147483647
        or EDX, ECX
        mov DWORD PTR [EAX], EDX
        mov ESI, DWORD PTR [ESP]
        add ESP, 4
        ret

llvm-svn: 28122

llvm-svn: 28121

  // fold (and (trunc x), (trunc y)) -> (trunc (and x, y))
  // fold (or  (trunc x), (trunc y)) -> (trunc (or  x, y))
  // fold (xor (trunc x), (trunc y)) -> (trunc (xor x, y))

llvm-svn: 28120

that gets emitted as movl (for r32 to i16, i8) or a movw (for r16 to i8). And
if the destination gets allocated a subregister of the source operand, then
the instruction will not be emitted at all.

llvm-svn: 28119

llvm-svn: 28118

llvm-svn: 28117

llvm-svn: 28116

of cross-block live ranges, and allows the bb-at-a-time selector to always
coallesce these away, at isel time.

This reduces the load on the coallescer and register allocator.  For example
on a codec on X86, we went from:

   1643 asm-printer           - Number of machine instrs printed
    419 liveintervals         - Number of loads/stores folded into instructions
   1144 liveintervals         - Number of identity moves eliminated after coalescing
   1022 liveintervals         - Number of interval joins performed
    282 liveintervals         - Number of intervals after coalescing
   1304 liveintervals         - Number of original intervals
     86 regalloc              - Number of times we had to backtrack
1.90232 regalloc              - Ratio of intervals processed over total intervals
     40 spiller               - Number of values reused
    182 spiller               - Number of loads added
    121 spiller               - Number of stores added
    132 spiller               - Number of register spills
      6 twoaddressinstruction - Number of instructions commuted to coalesce
    360 twoaddressinstruction - Number of two-address instructions

to:

   1636 asm-printer           - Number of machine instrs printed
    403 liveintervals         - Number of loads/stores folded into instructions
   1155 liveintervals         - Number of identity moves eliminated after coalescing
   1033 liveintervals         - Number of interval joins performed
    279 liveintervals         - Number of intervals after coalescing
   1312 liveintervals         - Number of original intervals
     76 regalloc              - Number of times we had to backtrack
1.88998 regalloc              - Ratio of intervals processed over total intervals
      1 spiller               - Number of copies elided
     41 spiller               - Number of values reused
    191 spiller               - Number of loads added
    114 spiller               - Number of stores added
    128 spiller               - Number of register spills
      4 twoaddressinstruction - Number of instructions commuted to coalesce
    356 twoaddressinstruction - Number of two-address instructions

On this testcase, this change provides a modest reduction in spill code,
regalloc iterations, and total instructions emitted.  It increases the number
of register coallesces.

llvm-svn: 28115

llvm-svn: 28114

llvm-svn: 28113

llvm-svn: 28112

llvm-svn: 28111

llvm-svn: 28110

scheduler can go into a "vertical mode" (i.e. traversing up the two-address
chain, etc.) when the register pressure is low.
This does seem to reduce the number of spills in the cases I've looked at. But
with x86, it's no guarantee the performance of the code improves.
It can be turned on with -sched-vertically option.

llvm-svn: 28108