llvm-svn: 45059

llvm-svn: 45058

llvm-svn: 45050

llvm-svn: 45041

Fix (mem) <-> low 64-bits of xmm bugs pointed out by David Greene. Mac OS X Leopard assembler recognizes movq.

llvm-svn: 45040

endianness of the target not of the host.  Done by the
simple expedient of reversing bytes for primitive types
if the host and target endianness don't match.  This is
correct for integer and pointer types.  I don't know if
it is correct for floating point types.

llvm-svn: 45039

for parameter passing (only for that, on Darwin).

llvm-svn: 45038

llvm-svn: 45037

llvm-svn: 45036

llvm-svn: 45034

llvm-svn: 45030

Fix ctlz and cttz. llvm definition requires them to return number of bits in of the src type when value is zero.

llvm-svn: 45029

llvm-svn: 45028

All MMX shift instructions took a <2 x i32> vector as the shift amount parameter. Change this to be <1 x i64> instead, which matches the assembler instruction.

llvm-svn: 45027

llvm-svn: 45024

side effects for machine instructions.

llvm-svn: 45022

Let: %q = GEP %p, X, ...
If %p is a GEP, we can chase baseptr further, only if X==0.

llvm-svn: 44999

llvm-svn: 44997

llvm-svn: 44987

Remove ugly and horrible code.  It's not necessary for correctness, and can be added back later if it causes code quality issues.

llvm-svn: 44986

llvm-svn: 44981

llvm-svn: 44970

llvm-svn: 44965

Implicit def instructions, e.g. X86::IMPLICIT_DEF_GR32, are always re-materializable and they should not be spilled.

llvm-svn: 44960

put it in a new header System/Host.h instead.
Instead of getting the endianness from configure,
calculate it directly.

llvm-svn: 44959

SelectionDAG::getConstant, in the same way as vector floating-point
constants. This allows the legalize expansion code for @llvm.ctpop and
friends to be usable with vector types.

llvm-svn: 44954

and readnone for functions with bodies because it
broke llvm-gcc-4.2 bootstrap.  It turns out that,
because of LLVM's array_ref hack, gcc was computing
pure/const attributes wrong (now fixed by turning
off the gcc ipa-pure-const pass).

llvm-svn: 44937

2. Using zero-extended value of Scale and unsigned division is safe provided
   that Scale doesn't have the sign bit set.
   Previously these 2 instructions:
        %p = bitcast [100 x {i8,i8,i8}]* %x to i8*
        %q = getelementptr i8* %p, i32 -4
   were combined into:
        %q = getelementptr [100 x { i8, i8, i8 }]* %x, i32 0,
               i32 1431655764, i32 0
   what was incorrect.

llvm-svn: 44936

llvm-svn: 44934

llvm-svn: 44933

llvm-svn: 44929

Lower a build_vector with all constants into a constpool load unless it can be done with a move to low part.

llvm-svn: 44921

not support it.

llvm-svn: 44914

Make lli interpreter correctly call external functions sin()/cos(),
__cxa_guard_acquire() and __cxa_guard_release().

llvm-svn: 44910

pointing this out and correcting the patch!

llvm-svn: 44907

llvm-svn: 44905

llvm-svn: 44904

interfered with other registers.  Seems like that might be a good thing to do. :-)

llvm-svn: 44902

llvm-svn: 44899

Bug fix. Only safe to perform extension uses optimization if the source of extension is also defined in the same BB as the extension.

llvm-svn: 44896