When the greedy register allocator is splitting multiple global live ranges, it
tends to look at the same interference data many times. The InterferenceCache
class caches queries for unaltered LiveIntervalUnions.

llvm-svn: 128764

This is more compact and faster than using DenseMap.

llvm-svn: 128763

registers that arise from argument shuffling with the soft float ABI. These
instructions are particularly slow on Cortex A8. This fixes one half of
<rdar://problem/8674845>.

llvm-svn: 128759

transformations in target-specific DAG combines without causing DAGCombiner to
delete the same node twice. If you know of a better way to avoid this (see my
next patch for an example), please let me know.

llvm-svn: 128758

instead of the second operand in addrmode_imm12.

rdar://problem/9225289

llvm-svn: 128757

llvm-svn: 128751

llvm-svn: 128750

rdar://problem/9224276

llvm-svn: 128749

rdar://problem/9224120

llvm-svn: 128748

all the instruction have:

    let Inst{31-27} = 0b1110; // non-predicated

Before, the ARM decoder was confusing:

> 0x40 0xf3 0xb8 0x80

as:

Opcode=16 Name=ADCSSrs Format=ARM_FORMAT_DPSOREGFRM(5)
 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9  8  7  6  5  4  3  2  1  0 
-------------------------------------------------------------------------------------------------
| 1: 0: 0: 0| 0: 0: 0: 0| 1: 0: 1: 1| 1: 0: 0: 0| 1: 1: 1: 1| 0: 0: 1: 1| 0: 1: 0: 0| 0: 0: 0: 0|
-------------------------------------------------------------------------------------------------

	adcs	pc, r8, r0, asr #6

since the cond field for ADCSSrs is a wild card, and so is ADCrs, with the ADCSSrs having Inst{20} as '1'.

Now, the AR decoder behaves correctly:

> 0x40 0xf3 0xb8 0x80
> END
Executing command: /Volumes/data/lldb/llvm/Debug+Asserts/bin/llvm-mc -disassemble -triple=arm-apple-darwin -debug-only=arm-disassembler mc-input.txt

Opcode=19 Name=ADCrs Format=ARM_FORMAT_DPSOREGFRM(5)
 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9  8  7  6  5  4  3  2  1  0 
-------------------------------------------------------------------------------------------------
| 1: 0: 0: 0| 0: 0: 0: 0| 1: 0: 1: 1| 1: 0: 0: 0| 1: 1: 1: 1| 0: 0: 1: 1| 0: 1: 0: 0| 0: 0: 0: 0|
-------------------------------------------------------------------------------------------------

	adcshi	pc, r8, r0, asr #6
> 

rdar://problem/9223094

llvm-svn: 128746

llvm-svn: 128745

- Localize the check if an icmp has one use to a place where we know we're
  introducing something that's likely more expensive than a sext from i1.
- Add an assert to make sure a case that would lead to a miscompilation is
  folded away earlier.
- Fix a typo.

llvm-svn: 128744

llvm-svn: 128743

Remove redundant code. There are assignments to variables Base and Offset right after the code that is removed.

llvm-svn: 128742

llvm-svn: 128741

llvm-svn: 128740

When the architecture is explicitly armv6 or thumbv6, we need to mark the object file appropriately.

llvm-svn: 128739

llvm-svn: 128736

llvm-svn: 128735

Fix a LDRT/LDRBT decoding bug where for Encoding A2, if Inst{4} != 0, we should reject the instruction
as invalid.

llvm-svn: 128734

llvm-svn: 128733

int test1(unsigned x) { return (x&8) ? 0 : -1; }
int test3(unsigned x) { return (x&8) ? -1 : 0; }

before (x86_64):
_test1:
	andl	$8, %edi
	cmpl	$1, %edi
	sbbl	%eax, %eax
	ret
_test3:
	andl	$8, %edi
	cmpl	$1, %edi
	sbbl	%eax, %eax
	notl	%eax
	ret

after:
_test1:
	shrl	$3, %edi
	andl	$1, %edi
	leal	-1(%rdi), %eax
	ret
_test3:
	shll	$28, %edi
	movl	%edi, %eax
	sarl	$31, %eax
	ret

llvm-svn: 128732

llvm-svn: 128731

llvm-svn: 128730

Assign node order numbers to results of call instruction lowering. This should improve src line debug info when sdisel is used. rdar://9199118

llvm-svn: 128728

with the contents of CMAKE_C(XX)_FLAGS too, else `llvm-config
--c(xx)flags' doesn't tell the absolute truth.

This comes from PR9603 and is based on a patch by Ryuta Suzuki!

llvm-svn: 128727

Modifies MipsAsmPrinter::isBlockOnlyReachableByFallthrough so that it handles delay slots correctly.

llvm-svn: 128724

Fix LDRi12 immediate operand, which was changed to be the second operand in $addrmode_imm12 => (ops GPR:$base, i32imm:$offsimm).

rdar://problem/9219356

llvm-svn: 128722

Patch by arrowdoger.

llvm-svn: 128719

llvm-svn: 128718

llvm-svn: 128709

list of operands. Simplify and rename them accordingly.

llvm-svn: 128708

llvm-svn: 128707

llvm-svn: 128706

had gotten out of sync: isCastable didn't think it was possible to
cast the x86_mmx type to anything, while it did think it possible
to cast an i64 to x86_mmx.

llvm-svn: 128705

Add annotations to tablegen-generated processor itineraries, or replace them with something meaningful. I want to be able to read and debug the generated tables.

llvm-svn: 128703

llvm-svn: 128701

rdar://8911343

llvm-svn: 128696

llvm-svn: 128692

llvm-svn: 128690