llvm-svn: 40985

llvm-svn: 40982

llvm-svn: 40981

llvm-svn: 40980

llvm-svn: 40979

llvm-svn: 40978

not split condition constraints.

llvm-svn: 40977

llvm-svn: 40976

llvm-svn: 40975

to handle values bigger than double.  If we assume host==target and host
long double works correctly, this is not too bad, but we don't want to
have that limitation longterm.  I could implement accepting double
constants as long double or something like that, which would lead to
incorrect codegen with no errors; the more I think about that the worse
it seems.  Rather than do such a hack that would be backed out later,
I'm settling for giving reasonable error messages, for now.

llvm-svn: 40974

llvm-svn: 40973

divb / mulb outputs to ah. Under x86-64 it's not legal to read ah if the instruction requires a rex prefix (i.e. outputs to r8b, etc.). So issue shift right by 8 on AX and then truncate it to 8 bits instead.

llvm-svn: 40972

alloca, increase the alignment of the load, turning it into an aligned load.

This allows us to compile:

#include <xmmintrin.h>
__m128i foo(__m128i x){
 static const unsigned int c_0[4] = { 0, 0, 0, 0 };
	  __m128i v_Zero = _mm_loadu_si128((__m128i*)c_0);
  x  = _mm_unpacklo_epi8(x,  v_Zero);
  return x;
}

into:

_foo:
	punpcklbw	_c_0.5944, %xmm0
	ret
	.data
	.lcomm	_c_0.5944,16,4		# c_0.5944

instead of:

_foo:
	movdqu	_c_0.5944, %xmm1
	punpcklbw	%xmm1, %xmm0
	ret
	.data
	.lcomm	_c_0.5944,16,2		# c_0.5944

llvm-svn: 40971

GR16_ sub-register class should be GR8_, not GR8. That is, it should only be 8-bit registers in 32-bit mode. Ditto for GR32_.

llvm-svn: 40970

llvm-svn: 40965

llvm-svn: 40961

llvm-svn: 40960

(constants are still not handled).  Adds ConvertActions
to control fp-to-fp conversions (these are currently
defaulted for all other targets, so no changes there).

llvm-svn: 40958

llvm-svn: 40957

to do so.

llvm-svn: 40955

llvm-svn: 40953

llvm-svn: 40952

llvm-svn: 40950

llvm-svn: 40947

llvm-svn: 40946

llvm-svn: 40944

llvm-svn: 40941

documentation implies and as its uses depend.

llvm-svn: 40939

stdin through llvm-ld and llvm-link.

llvm-svn: 40938

llvm-svn: 40936

and one hack to avoid hitting a bad case when the alias analysis is imprecise.

llvm-svn: 40935

it for potentially undeading pointers.

llvm-svn: 40933

No functionality change.

llvm-svn: 40932

llvm-svn: 40927

llvm-svn: 40926

llvm-svn: 40925

ConstantFoldInstruction on calls) by avoiding Value::getName().  getName() constructs
and returns an std::string, which does heap allocation stuff.  This slightly speeds up
instcombine.

llvm-svn: 40924

1. domtree is a tree, not a graph.  There is no need to avoid revisiting nodes with a set.
2. the worklist can contain the child iterator pointers so we don't get N^2 rescanning of children.

This speeds up updateDFSNumbers significantly, making it basically free.  On the testcase in PR1432,
this speeds up loopsimplify by another 3x, dropping it from the 12th most expensive pass to the to
the 30th. :)  It used to be #1.

llvm-svn: 40923

llvm-svn: 40922

llvm-svn: 40921