and redo as linked list walk.  Logic moved into RA.
Per review feedback.

llvm-svn: 56326

Fixes PR 2805

llvm-svn: 56321

catches a fair number of common cases. Note that this currently
causes Fast-ISel to leave behind lots of dead instructions.
Those will be dealt with in subsequent commits.

llvm-svn: 56320

Thanks to Ji Young Park for the patch!

llvm-svn: 56316

llvm-svn: 56315

llvm-svn: 56314

llvm-svn: 56311

defs to be necessarily live.

llvm-svn: 56310

llvm-svn: 56308

copy of the BURRList scheduler, but with several parts ripped
out, such as backtracking, online topological sort maintenance
(needed by backtracking), the priority queue, and Sethi-Ullman
number computation and maintenance (needed by the priority
queue). As a result of all this, it generates somewhat lower
quality code, but that's its tradeoff for running about 30%
faster than list-burr in -fast mode in many cases.

This is somewhat experimental. Moving forward, major pieces of
this can be refactored with pieces in common with
ScheduleDAGRRList.cpp.

llvm-svn: 56307

Preliminary support for systems which require changing JIT memory regions privilege from read / write to read / executable.

llvm-svn: 56303

llvm-svn: 56301

llvm-svn: 56300

llvm-svn: 56299

over having it in a register. And wait until after checking type
legality before requesting that the callee address be placed in a
register. Also, fix support for calls with void return type.

This speeds up fast-isel isel time by about 15% and reduces
instruction counts by about 3% overall on certain testcases. It also
changes many indirect calls to direct calls.

llvm-svn: 56292

with an earlyclobber operand elsewhere.  Propagate
this bit and the earlyclobber bit through SDISel.
Change linear-scan RA not to allocate regs in a way 
that conflicts with an earlyclobber.  See also comments.

llvm-svn: 56290

llvm-svn: 56287

llvm-svn: 56286

This one slipped through cracks very well.

llvm-svn: 56284

llvm-svn: 56282

llvm-svn: 56281

llvm-svn: 56277

llvm-svn: 56276

llvm-svn: 56275

up some new ascii art to illustrate what it does. This change
currently has no effect on generated code.

llvm-svn: 56270

be used with fast-isel.

llvm-svn: 56268

function with appropriate parameters. This allows us to support blocks on PPC.

llvm-svn: 56267

llvm-svn: 56265

When converting a CopyFromReg to a copy instruction, use the register class of its uses to determine the right destination register class of the copy. This is important for targets where a physical register may belong to multiple register classes.

llvm-svn: 56258

llvm-svn: 56256

llvm-svn: 56255

ConstantPoolSDNode, using the target's preferred alignment for the
constant type.

In LegalizeDAG, when performing loads from the constant pool, the
ConstantPoolSDNode's alignment is used in the calls to getLoad and
getExtLoad.

This change prevents SelectionDAG::getLoad/getExtLoad from incorrectly
choosing the ABI alignment for constant pool loads when Alignment == 0.
The incorrect alignment is only a performance issue when ABI alignment
does not equal preferred alignment (i.e., on x86 it was generating
MOVUPS instead of MOVAPS for v4f32 constant loads when the default ABI
alignment for 128bit vectors is forced to 1 byte.)

Patch by Paul Redmond!

llvm-svn: 56253

Apologies for the thrashing.

llvm-svn: 56251

llvm-svn: 56250

- Add linkage to SymbolSDNode (default to external).
- Change ISD::ExternalSymbol to ISD::Symbol.
- Change ISD::TargetExternalSymbol to ISD::TargetSymbol

These changes pave the way to allowing SymbolSDNodes with non-external linkage.

llvm-svn: 56249

MachineConstantPool::getConstantPoolIndex actually expects
a log2-encoded alignment.

llvm-svn: 56248

of r56230, r56232, and r56246.

llvm-svn: 56247

 - Recognize expressions like "x > -1 ? x : 0" as min/max and turn them
   into expressions like "x < 0 ? 0 : x", which is easily recognizable
   as a min/max operation.
 - Refrain from folding expression like "y/2 < 1" to "y < 2" when the
   comparison is being used as part of a min or max idiom, like
   "y/2 < 1 ? 1 : y/2". In that case, the division has another use, so
   folding doesn't eliminate it, and obfuscates the min/max, making it
   harder to recognize as a min/max operation.

These benefit ScalarEvolution, CodeGen, and anything else that wants to
recognize integer min and max.

llvm-svn: 56246

llvm-svn: 56244

llvm-svn: 56243