llvm-svn: 45667

llvm-svn: 45638

the target independent legalizer.

llvm-svn: 45631

unifying the copied algorithms and saving over 500 LOC. There should
be no functionality change, but please test on your favorite x86
target.

llvm-svn: 45627

checking that there was a from a global instead of a load from the stub
for a global, which is the one that's safe to hoist.

Consider this program:

volatile char G[100];
int B(char *F, int N) {
  for (; N > 0; --N)
    F[N] = G[N];
}

In static mode, we shouldn't be hoisting the load from G:

$ llc -relocation-model=static -o - a.bc -march=x86 -machine-licm

LBB1_1: # bb.preheader
        leal    -1(%eax), %edx
        testl   %edx, %edx
        movl    $1, %edx
        cmovns  %eax, %edx
        xorl    %esi, %esi
LBB1_2: # bb
        movb    _G(%eax), %bl
        movb    %bl, (%ecx,%eax)

llvm-svn: 45626

for remat, but can't due to an RA bug.

llvm-svn: 45623

llvm-svn: 45622

llvm-svn: 45621

isReallySideEffectFree and isReallyTriviallyReMaterializable.  Why is a load from
a global considered side-effect-free but not rematable?

llvm-svn: 45620

llvm-svn: 45618

llvm-svn: 45616

llvm-svn: 45605

llvm-svn: 45603

llvm-svn: 45576

llvm-svn: 45575

llvm-svn: 45536

for non-function GV relocations that require function address stubs (e.g. Mac OS X in non-static mode).

llvm-svn: 45527

llvm-svn: 45515

check that register isn't 0 before going further.

llvm-svn: 45498

llvm-svn: 45494

llvm-svn: 45484

Fix a bug in my previous patch: refer to the impl not the pure virtual version.  It's unclear why gcc would ever compile this...

llvm-svn: 45476

a header file from libcodegen.  This violates a layering order: codegen
depends on target, not the other way around.  The fix to this is to 
split TII into two classes, TII and TargetInstrInfoImpl, which defines
stuff that depends on libcodegen.  It is defined in libcodegen, where 
the base is not.

llvm-svn: 45475

Machine-level API cleanup instigated by Chris.

llvm-svn: 45470

that "machine" classes are used to represent the current state of
the code being compiled.  Given this expanded name, we can start 
moving other stuff into it.  For now, move the UsedPhysRegs and
LiveIn/LoveOuts vectors from MachineFunction into it.

Update all the clients to match.

This also reduces some needless #includes, such as MachineModuleInfo
from MachineFunction.

llvm-svn: 45467

e.g. MO.isMBB() instead of MO.isMachineBasicBlock().  I don't plan on 
switching everything over, so new clients should just start using the 
shorter names.

Remove old long accessors, switching everything over to use the short
accessor: getMachineBasicBlock() -> getMBB(), 
getConstantPoolIndex() -> getIndex(), setMachineBasicBlock -> setMBB(), etc.

llvm-svn: 45464

  - Eliminate the static "print" method for operands, moving it
    into MachineOperand::print.
  - Change various set* methods for register flags to take a bool
    for the value to set it to.  Remove unset* methods.
  - Group methods more logically by operand flavor in MachineOperand.h

llvm-svn: 45461

Use MachineOperand::getImm instead of MachineOperand::getImmedValue.  Likewise setImmedValue -> setImm

llvm-svn: 45453

function, then go ahead and hoist it out of the loop. This is the result:

$ cat a.c
volatile int G;

int A(int N) {
  for (; N > 0; --N)
    G++;
}
$ llc -o - -relocation-model=pic
_A:
...
LBB1_2: # bb
        movl    L_G$non_lazy_ptr-"L1$pb"(%eax), %esi
        incl    (%esi)
        incl    %edx
        cmpl    %ecx, %edx
        jne     LBB1_2  # bb
...
$ llc -o - -relocation-model=pic -machine-licm
_A:
...
        movl    L_G$non_lazy_ptr-"L1$pb"(%eax), %eax
LBB1_2: # bb
        incl    (%eax)
        incl    %edx
        cmpl    %ecx, %edx
        jne     LBB1_2  # bb
...

I'm limiting this to the MOV32rm x86 instruction for now.

llvm-svn: 45444

llvm-svn: 45433

llvm-svn: 45418

llvm-svn: 45415

eliminating the llvm.x86.sse2.loadl.pd intrinsic?), one shuffle optzn
may be done (if shufps is better than pinsw, Evan, please review), and
we already know about LICM of simple instructions.

llvm-svn: 45407

comparisons with a constant.  This allows us to compile isnan to:

_foo:
	fcmpu cr7, f1, f1
	mfcr r2
	rlwinm r3, r2, 0, 31, 31
	blr 

instead of:

LCPI1_0:					;  float
	.space	4
_foo:
	lis r2, ha16(LCPI1_0)
	lfs f0, lo16(LCPI1_0)(r2)
	fcmpu cr7, f1, f0
	mfcr r2
	rlwinm r3, r2, 0, 31, 31
	blr 

llvm-svn: 45405

llvm-svn: 45404

as:

_bar:
	pushl	%esi
	subl	$8, %esp
	movl	16(%esp), %esi
	call	L_foo$stub
	fstps	(%esi)
	addl	$8, %esp
	popl	%esi
	#FP_REG_KILL
	ret

instead of:

_bar:
	pushl	%esi
	subl	$8, %esp
	movl	16(%esp), %esi
	call	L_foo$stub
	fstpl	(%esi)
	cvtsd2ss	(%esi), %xmm0
	movss	%xmm0, (%esi)
	addl	$8, %esp
	popl	%esi
	#FP_REG_KILL
	ret

llvm-svn: 45401

if we are just going to store it back anyway.  This improves things 
like:
double foo();
void bar(double *P) { *P = foo(); }

llvm-svn: 45399

llvm-svn: 45397

llvm-svn: 45387

llvm-svn: 45343