llvm-svn: 14179

getReg() on void value.

llvm-svn: 14178

on a double-word boundary instead of a single-word boundary.

llvm-svn: 14177

non-register, non-immed. arguments to SETHI and OR using %hi() and
%lo() respectively.

llvm-svn: 14176

Support copying long constants to register pairs.
Support copying ConstantPointerNulls and ConstantPointerRefs to registers.

llvm-svn: 14175

llvm-svn: 14172

peepholing, and make it work the same way in the JIT as in LLC.

llvm-svn: 14170

comparisons.  In an 'isunordered' predicate, which looks like this at
the LLVM level:

        %a = call bool %llvm.isnan(double %X)
        %b = call bool %llvm.isnan(double %Y)
        %COM = or bool %a, %b

We used to generate this code:

        fxch %ST(1)
        fucomip %ST(0), %ST(0)
        setp %AL
        fucomip %ST(0), %ST(0)
        setp %AH
        or %AL, %AH

With this patch, we generate this code:

        fucomip %ST(0), %ST(1)
        fstp %ST(0)
        setp %AL

Which should make alkis happy.  Tested as X86/compare_folding.llx:test1

llvm-svn: 14148

llvm-svn: 14146

llvm-svn: 14145

we can get rid of the FpUCOM/FpUCOMi pseudo instructions, which makes stuff simpler
and faster.

llvm-svn: 14144

twoarg cases.

llvm-svn: 14143

testcase

llvm-svn: 14141

llvm-svn: 14140

This makes the code much simpler, and the two cases really do belong apart.
Once we do it, it's pretty obvious how flawed the logic was for A != A case,
so I fixed it (fixing PR369).

This also uses freeStackSlotAfter instead of inserting an fxchg then
popStackAfter'ing in the case where there is a dead result (unlikely, but
possible), producing better code.

llvm-svn: 14139

them into the constant pool.

llvm-svn: 14128

llvm-svn: 14110

llvm-svn: 14103

llvm-svn: 14102

llvm-svn: 14101

that cast to bool.

llvm-svn: 14096

accounted for in the map (at least, in its current format).

llvm-svn: 14075

llvm-svn: 14073

llvm-svn: 14027

llvm-svn: 14005

and sweet.

llvm-svn: 13983

llvm-svn: 13981

get this to link before?

llvm-svn: 13980

llvm-svn: 13979

only concrete implementation.

llvm-svn: 13977

that we REALLY don't want in the CBE code.

With this fix, the CBE passes all of the MultiSource tests on cygwin that
it does on linux.  Yaay!

llvm-svn: 13975

llvm-svn: 13954

llvm-svn: 13953

llvm-svn: 13952

llvm-svn: 13951

llvm-svn: 13949

llvm-svn: 13911

llvm-svn: 13908

state. Also, save the state for the incoming register of each phi
node.

llvm-svn: 13906

Call it at a more appropriate point.

llvm-svn: 13905