[llvm-branch-commits] [llvm-branch] r167838 [1/2] - in /llvm/branches/R600: ./ autoconf/ cmake/modules/ docs/ docs/CommandGuide/ examples/Fibonacci/ include/llvm-c/ include/llvm/ include/llvm/ADT/ include/llvm/Analysis/ include/llvm/CodeGen/ include/llvm/CodeGen/PBQP/ include/llvm/Config/ include/llvm/ExecutionEngine/ include/llvm/MC/ include/llvm/MC/MCParser/ include/llvm/Object/ include/llvm/Support/ include/llvm/TableGen/ include/llvm/Target/ include/llvm/Transforms/ include/llvm/Transforms/IPO/ include/llvm/Transforms/Uti...

Tom Stellard thomas.stellard at amd.com
Tue Nov 13 07:21:50 PST 2012


Author: tstellar
Date: Tue Nov 13 09:21:47 2012
New Revision: 167838

URL: http://llvm.org/viewvc/llvm-project?rev=167838&view=rev
Log:
Merge master branch

Build with clang checkouts:

SVN:  r167547
Git Mirror: b578aee665aad5ed1a46a26217c730fdfbfc8c2e

Added:
    llvm/branches/R600/docs/HowToUseInstrMappings.rst   (with props)
    llvm/branches/R600/docs/MarkedUpDisassembly.rst
    llvm/branches/R600/lib/Analysis/CostModel.cpp
    llvm/branches/R600/test/Analysis/CostModel/
    llvm/branches/R600/test/Analysis/CostModel/X86/
    llvm/branches/R600/test/Analysis/CostModel/X86/arith.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/cast.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/cmp.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/i32.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/insert-extract-at-zero.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/lit.local.cfg
    llvm/branches/R600/test/Analysis/CostModel/X86/loop_v2.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/tiny.ll
    llvm/branches/R600/test/Analysis/CostModel/X86/vectorized-loop.ll
    llvm/branches/R600/test/Analysis/CostModel/lit.local.cfg
      - copied, changed from r167837, llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/lit.local.cfg
    llvm/branches/R600/test/Analysis/CostModel/no_info.ll
    llvm/branches/R600/test/CodeGen/ARM/call-noret-minsize.ll
    llvm/branches/R600/test/CodeGen/ARM/indirectbr-2.ll
    llvm/branches/R600/test/CodeGen/Mips/alloca16.ll
    llvm/branches/R600/test/CodeGen/Mips/atomicops.ll
    llvm/branches/R600/test/CodeGen/Mips/brind.ll
    llvm/branches/R600/test/CodeGen/Mips/check-noat.ll
    llvm/branches/R600/test/CodeGen/Mips/eh-dwarf-cfa.ll
    llvm/branches/R600/test/CodeGen/Mips/i32k.ll
    llvm/branches/R600/test/CodeGen/Mips/llcarry.ll
    llvm/branches/R600/test/CodeGen/Mips/misha.ll
    llvm/branches/R600/test/CodeGen/Mips/remat-immed-load.ll
    llvm/branches/R600/test/CodeGen/Mips/selpat.ll
    llvm/branches/R600/test/CodeGen/Mips/seteq.ll
    llvm/branches/R600/test/CodeGen/Mips/seteqz.ll
    llvm/branches/R600/test/CodeGen/Mips/setge.ll
    llvm/branches/R600/test/CodeGen/Mips/setgek.ll
    llvm/branches/R600/test/CodeGen/Mips/setle.ll
    llvm/branches/R600/test/CodeGen/Mips/setlt.ll
    llvm/branches/R600/test/CodeGen/Mips/setltk.ll
    llvm/branches/R600/test/CodeGen/Mips/setne.ll
    llvm/branches/R600/test/CodeGen/Mips/setuge.ll
    llvm/branches/R600/test/CodeGen/Mips/setugt.ll
    llvm/branches/R600/test/CodeGen/Mips/setule.ll
    llvm/branches/R600/test/CodeGen/Mips/setult.ll
    llvm/branches/R600/test/CodeGen/Mips/setultk.ll
    llvm/branches/R600/test/CodeGen/Mips/stchar.ll
    llvm/branches/R600/test/CodeGen/Mips/tls16.ll
    llvm/branches/R600/test/CodeGen/Mips/tls16_2.ll
    llvm/branches/R600/test/CodeGen/PowerPC/asm-Zy.ll
    llvm/branches/R600/test/CodeGen/PowerPC/emptystruct.ll
    llvm/branches/R600/test/CodeGen/PowerPC/jaggedstructs.ll
    llvm/branches/R600/test/CodeGen/PowerPC/ppc64-abi-extend.ll
    llvm/branches/R600/test/CodeGen/PowerPC/ppc64-align-long-double.ll
    llvm/branches/R600/test/CodeGen/PowerPC/pr12757.ll
    llvm/branches/R600/test/CodeGen/PowerPC/varargs-struct-float.ll
    llvm/branches/R600/test/CodeGen/PowerPC/vec_extload.ll
    llvm/branches/R600/test/CodeGen/PowerPC/vec_sqrt.ll
    llvm/branches/R600/test/CodeGen/X86/atom-shuf.ll
    llvm/branches/R600/test/CodeGen/X86/avx-intel-ocl.ll
    llvm/branches/R600/test/CodeGen/X86/cvtv2f32.ll
    llvm/branches/R600/test/CodeGen/X86/inlineasm-sched-bug.ll
    llvm/branches/R600/test/CodeGen/X86/misched-balance.ll
    llvm/branches/R600/test/CodeGen/X86/pr14161.ll
    llvm/branches/R600/test/CodeGen/X86/pr14204.ll
    llvm/branches/R600/test/CodeGen/X86/sse-intel-ocl.ll
    llvm/branches/R600/test/CodeGen/X86/sse_partial_update.ll
    llvm/branches/R600/test/ExecutionEngine/MCJIT/test-common-symbols-alignment.ll
    llvm/branches/R600/test/ExecutionEngine/MCJIT/test-data-align.ll
    llvm/branches/R600/test/Feature/minsize_attr.ll
    llvm/branches/R600/test/Instrumentation/AddressSanitizer/do-not-instrument-internal-globals.ll
    llvm/branches/R600/test/MC/ARM/thumb2-b.w-encodingT4.s
    llvm/branches/R600/test/MC/Disassembler/ARM/marked-up-thumb.txt
    llvm/branches/R600/test/MC/Disassembler/X86/marked-up.txt
    llvm/branches/R600/test/MC/MachO/gen-dwarf-cpp.s
    llvm/branches/R600/test/MC/MachO/gen-dwarf-macro-cpp.s
    llvm/branches/R600/test/MC/Markup/
    llvm/branches/R600/test/MC/Markup/basic-markup.mc
    llvm/branches/R600/test/MC/Markup/lit.local.cfg
    llvm/branches/R600/test/MC/PowerPC/
    llvm/branches/R600/test/MC/PowerPC/lit.local.cfg
    llvm/branches/R600/test/MC/PowerPC/ppc64-relocs-01.ll
    llvm/branches/R600/test/MC/X86/x86-32-ms-inline-asm.s
    llvm/branches/R600/test/Other/extract-alias.ll
    llvm/branches/R600/test/Other/extract-weak-odr.ll
    llvm/branches/R600/test/Other/link-opts.ll
    llvm/branches/R600/test/Transforms/BBVectorize/X86/
    llvm/branches/R600/test/Transforms/BBVectorize/X86/loop1.ll
    llvm/branches/R600/test/Transforms/BBVectorize/X86/simple-ldstr.ll
    llvm/branches/R600/test/Transforms/BBVectorize/X86/simple.ll
    llvm/branches/R600/test/Transforms/BBVectorize/X86/vs-cast.ll
    llvm/branches/R600/test/Transforms/GVN/pr14166.ll
    llvm/branches/R600/test/Transforms/IndVarSimplify/verify-scev.ll
    llvm/branches/R600/test/Transforms/InstCombine/2012-10-25-vector-of-pointers.ll
    llvm/branches/R600/test/Transforms/InstCombine/stpcpy-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/stpcpy-2.ll
    llvm/branches/R600/test/Transforms/InstCombine/stpcpy_chk-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/stpcpy_chk-2.ll
    llvm/branches/R600/test/Transforms/InstCombine/strlen-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/strlen-2.ll
    llvm/branches/R600/test/Transforms/InstCombine/strncpy-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/strncpy-2.ll
    llvm/branches/R600/test/Transforms/InstCombine/strpbrk-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/strpbrk-2.ll
    llvm/branches/R600/test/Transforms/InstCombine/strto-1.ll
    llvm/branches/R600/test/Transforms/InstCombine/vector_gep2.ll
    llvm/branches/R600/test/Transforms/LoopIdiom/crash.ll
    llvm/branches/R600/test/Transforms/LoopIdiom/scev-invalidation.ll
    llvm/branches/R600/test/Transforms/LoopUnroll/pr14167.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/2012-10-22-isconsec.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/avx1.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/conversion-cost.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/cost-model.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/gcc-examples.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/X86/lit.local.cfg
    llvm/branches/R600/test/Transforms/LoopVectorize/cpp-new-array.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/flags.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/small-loop.ll
      - copied, changed from r167837, llvm/branches/R600/test/Transforms/LoopVectorize/non-const-n.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/start-non-zero.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/write-only.ll
      - copied, changed from r167837, llvm/branches/R600/test/Transforms/LoopVectorize/induction_plus.ll
    llvm/branches/R600/test/Transforms/SimplifyCFG/SPARC/
    llvm/branches/R600/test/Transforms/SimplifyCFG/SPARC/lit.local.cfg
    llvm/branches/R600/test/Transforms/SimplifyCFG/SPARC/switch_to_lookup_table.ll
    llvm/branches/R600/tools/llvm-mcmarkup/
    llvm/branches/R600/tools/llvm-mcmarkup/CMakeLists.txt
    llvm/branches/R600/tools/llvm-mcmarkup/LLVMBuild.txt
      - copied, changed from r167837, llvm/branches/R600/tools/lli/LLVMBuild.txt
    llvm/branches/R600/tools/llvm-mcmarkup/Makefile
      - copied, changed from r167837, llvm/branches/R600/tools/llvm-ar/Makefile
    llvm/branches/R600/tools/llvm-mcmarkup/llvm-mcmarkup.cpp
    llvm/branches/R600/utils/TableGen/CodeGenMapTable.cpp
Removed:
    llvm/branches/R600/include/llvm/Analysis/LoopDependenceAnalysis.h
    llvm/branches/R600/include/llvm/Target/TargetELFWriterInfo.h
    llvm/branches/R600/lib/Analysis/LoopDependenceAnalysis.cpp
    llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.cpp
    llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.h
    llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.cpp
    llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.h
    llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.cpp
    llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.h
    llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.cpp
    llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.h
    llvm/branches/R600/lib/Target/TargetELFWriterInfo.cpp
    llvm/branches/R600/lib/Target/X86/X86ELFWriterInfo.cpp
    llvm/branches/R600/lib/Target/X86/X86ELFWriterInfo.h
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/alias.ll
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/lit.local.cfg
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/siv-strong.ll
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/siv-weak-crossing.ll
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/siv-weak-zero.ll
    llvm/branches/R600/test/Analysis/LoopDependenceAnalysis/ziv.ll
    llvm/branches/R600/test/CodeGen/SPARC/load_to_switch.ll
    llvm/branches/R600/test/Transforms/SimplifyLibCalls/2009-02-12-StrTo.ll
    llvm/branches/R600/test/Transforms/SimplifyLibCalls/StpCpy.ll
    llvm/branches/R600/test/Transforms/SimplifyLibCalls/StrLen.ll
    llvm/branches/R600/test/Transforms/SimplifyLibCalls/StrNCpy.ll
    llvm/branches/R600/test/Transforms/SimplifyLibCalls/StrPBrk.ll
Modified:
    llvm/branches/R600/Makefile
    llvm/branches/R600/autoconf/configure.ac
    llvm/branches/R600/cmake/modules/LLVMProcessSources.cmake
    llvm/branches/R600/configure
    llvm/branches/R600/docs/CodeGenerator.rst
    llvm/branches/R600/docs/CodingStandards.rst
    llvm/branches/R600/docs/CommandGuide/FileCheck.rst
    llvm/branches/R600/docs/HowToBuildOnARM.rst
    llvm/branches/R600/docs/LangRef.html
    llvm/branches/R600/docs/Passes.html
    llvm/branches/R600/docs/Phabricator.rst
    llvm/branches/R600/docs/ReleaseNotes.html
    llvm/branches/R600/docs/TestingGuide.html
    llvm/branches/R600/docs/WritingAnLLVMBackend.html
    llvm/branches/R600/docs/subsystems.rst
    llvm/branches/R600/examples/Fibonacci/fibonacci.cpp
    llvm/branches/R600/include/llvm-c/Disassembler.h
    llvm/branches/R600/include/llvm/ADT/APFloat.h
    llvm/branches/R600/include/llvm/ADT/APInt.h
    llvm/branches/R600/include/llvm/ADT/DenseMap.h
    llvm/branches/R600/include/llvm/ADT/StringSet.h
    llvm/branches/R600/include/llvm/ADT/Triple.h
    llvm/branches/R600/include/llvm/Analysis/DependenceAnalysis.h
    llvm/branches/R600/include/llvm/Analysis/Passes.h
    llvm/branches/R600/include/llvm/Analysis/ProfileDataLoader.h
    llvm/branches/R600/include/llvm/Analysis/ScalarEvolution.h
    llvm/branches/R600/include/llvm/Attributes.h
    llvm/branches/R600/include/llvm/CallingConv.h
    llvm/branches/R600/include/llvm/CodeGen/GCMetadata.h
    llvm/branches/R600/include/llvm/CodeGen/MachineInstr.h
    llvm/branches/R600/include/llvm/CodeGen/MachineOperand.h
    llvm/branches/R600/include/llvm/CodeGen/MachineScheduler.h
    llvm/branches/R600/include/llvm/CodeGen/PBQP/Graph.h
    llvm/branches/R600/include/llvm/CodeGen/RegisterPressure.h
    llvm/branches/R600/include/llvm/CodeGen/ScheduleDAG.h
    llvm/branches/R600/include/llvm/CodeGen/ScheduleDAGInstrs.h
    llvm/branches/R600/include/llvm/CodeGen/SelectionDAGNodes.h
    llvm/branches/R600/include/llvm/CodeGen/TargetSchedule.h
    llvm/branches/R600/include/llvm/Config/config.h.cmake
    llvm/branches/R600/include/llvm/Constants.h
    llvm/branches/R600/include/llvm/DataLayout.h
    llvm/branches/R600/include/llvm/ExecutionEngine/ExecutionEngine.h
    llvm/branches/R600/include/llvm/ExecutionEngine/JITEventListener.h
    llvm/branches/R600/include/llvm/ExecutionEngine/JITMemoryManager.h
    llvm/branches/R600/include/llvm/ExecutionEngine/RuntimeDyld.h
    llvm/branches/R600/include/llvm/IRBuilder.h
    llvm/branches/R600/include/llvm/InitializePasses.h
    llvm/branches/R600/include/llvm/InlineAsm.h
    llvm/branches/R600/include/llvm/InstrTypes.h
    llvm/branches/R600/include/llvm/Instructions.h
    llvm/branches/R600/include/llvm/IntrinsicInst.h
    llvm/branches/R600/include/llvm/LinkAllPasses.h
    llvm/branches/R600/include/llvm/MC/MCELFObjectWriter.h
    llvm/branches/R600/include/llvm/MC/MCInstPrinter.h
    llvm/branches/R600/include/llvm/MC/MCParser/MCAsmParser.h
    llvm/branches/R600/include/llvm/MC/MCParser/MCParsedAsmOperand.h
    llvm/branches/R600/include/llvm/MC/MCRegisterInfo.h
    llvm/branches/R600/include/llvm/MC/MCSchedule.h
    llvm/branches/R600/include/llvm/MC/MCTargetAsmParser.h
    llvm/branches/R600/include/llvm/Object/COFF.h
    llvm/branches/R600/include/llvm/Object/ELF.h
    llvm/branches/R600/include/llvm/Object/MachO.h
    llvm/branches/R600/include/llvm/Object/ObjectFile.h
    llvm/branches/R600/include/llvm/Operator.h
    llvm/branches/R600/include/llvm/Support/AlignOf.h
    llvm/branches/R600/include/llvm/Support/CommandLine.h
    llvm/branches/R600/include/llvm/Support/ELF.h
    llvm/branches/R600/include/llvm/Support/InstVisitor.h
    llvm/branches/R600/include/llvm/Support/IntegersSubset.h
    llvm/branches/R600/include/llvm/TableGen/Error.h
    llvm/branches/R600/include/llvm/Target/Target.td
    llvm/branches/R600/include/llvm/Target/TargetCallingConv.h
    llvm/branches/R600/include/llvm/Target/TargetLowering.h
    llvm/branches/R600/include/llvm/Target/TargetMachine.h
    llvm/branches/R600/include/llvm/Target/TargetRegisterInfo.h
    llvm/branches/R600/include/llvm/Target/TargetTransformImpl.h
    llvm/branches/R600/include/llvm/TargetTransformInfo.h
    llvm/branches/R600/include/llvm/Transforms/IPO.h
    llvm/branches/R600/include/llvm/Transforms/IPO/PassManagerBuilder.h
    llvm/branches/R600/include/llvm/Transforms/Utils/Local.h
    llvm/branches/R600/include/llvm/Type.h
    llvm/branches/R600/lib/Analysis/Analysis.cpp
    llvm/branches/R600/lib/Analysis/BasicAliasAnalysis.cpp
    llvm/branches/R600/lib/Analysis/CMakeLists.txt
    llvm/branches/R600/lib/Analysis/CodeMetrics.cpp
    llvm/branches/R600/lib/Analysis/ConstantFolding.cpp
    llvm/branches/R600/lib/Analysis/DependenceAnalysis.cpp
    llvm/branches/R600/lib/Analysis/InlineCost.cpp
    llvm/branches/R600/lib/Analysis/InstructionSimplify.cpp
    llvm/branches/R600/lib/Analysis/LazyValueInfo.cpp
    llvm/branches/R600/lib/Analysis/MemoryDependenceAnalysis.cpp
    llvm/branches/R600/lib/Analysis/ProfileDataLoader.cpp
    llvm/branches/R600/lib/Analysis/ScalarEvolution.cpp
    llvm/branches/R600/lib/Analysis/ValueTracking.cpp
    llvm/branches/R600/lib/AsmParser/LLLexer.cpp
    llvm/branches/R600/lib/AsmParser/LLParser.cpp
    llvm/branches/R600/lib/AsmParser/LLToken.h
    llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp
    llvm/branches/R600/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/branches/R600/lib/CodeGen/AsmPrinter/OcamlGCPrinter.cpp
    llvm/branches/R600/lib/CodeGen/EarlyIfConversion.cpp
    llvm/branches/R600/lib/CodeGen/GCStrategy.cpp
    llvm/branches/R600/lib/CodeGen/IntrinsicLowering.cpp
    llvm/branches/R600/lib/CodeGen/MachineBasicBlock.cpp
    llvm/branches/R600/lib/CodeGen/MachineFunction.cpp
    llvm/branches/R600/lib/CodeGen/MachineInstr.cpp
    llvm/branches/R600/lib/CodeGen/MachineScheduler.cpp
    llvm/branches/R600/lib/CodeGen/MachineVerifier.cpp
    llvm/branches/R600/lib/CodeGen/RegAllocFast.cpp
    llvm/branches/R600/lib/CodeGen/RegAllocPBQP.cpp
    llvm/branches/R600/lib/CodeGen/RegisterCoalescer.cpp
    llvm/branches/R600/lib/CodeGen/RegisterPressure.cpp
    llvm/branches/R600/lib/CodeGen/ScheduleDAGInstrs.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGFast.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
    llvm/branches/R600/lib/CodeGen/SelectionDAG/TargetLowering.cpp
    llvm/branches/R600/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/branches/R600/lib/CodeGen/TargetSchedule.cpp
    llvm/branches/R600/lib/CodeGen/TwoAddressInstructionPass.cpp
    llvm/branches/R600/lib/DebugInfo/DWARFDebugInfoEntry.cpp
    llvm/branches/R600/lib/ExecutionEngine/ExecutionEngine.cpp
    llvm/branches/R600/lib/ExecutionEngine/IntelJITEvents/IntelJITEventListener.cpp
    llvm/branches/R600/lib/ExecutionEngine/Interpreter/Execution.cpp
    llvm/branches/R600/lib/ExecutionEngine/Interpreter/ExternalFunctions.cpp
    llvm/branches/R600/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp
    llvm/branches/R600/lib/ExecutionEngine/MCJIT/CMakeLists.txt
    llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.cpp
    llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.h
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp
    llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h
    llvm/branches/R600/lib/MC/ELFObjectWriter.cpp
    llvm/branches/R600/lib/MC/MCDisassembler/Disassembler.cpp
    llvm/branches/R600/lib/MC/MCDisassembler/EDDisassembler.cpp
    llvm/branches/R600/lib/MC/MCELFObjectTargetWriter.cpp
    llvm/branches/R600/lib/MC/MCInstPrinter.cpp
    llvm/branches/R600/lib/MC/MCParser/AsmParser.cpp
    llvm/branches/R600/lib/Object/COFFObjectFile.cpp
    llvm/branches/R600/lib/Object/MachOObjectFile.cpp
    llvm/branches/R600/lib/Support/APFloat.cpp
    llvm/branches/R600/lib/Support/Atomic.cpp
    llvm/branches/R600/lib/Support/CMakeLists.txt
    llvm/branches/R600/lib/Support/Errno.cpp
    llvm/branches/R600/lib/Support/Host.cpp
    llvm/branches/R600/lib/Support/Makefile
    llvm/branches/R600/lib/Support/MemoryBuffer.cpp
    llvm/branches/R600/lib/Support/Triple.cpp
    llvm/branches/R600/lib/TableGen/CMakeLists.txt
    llvm/branches/R600/lib/TableGen/Error.cpp
    llvm/branches/R600/lib/TableGen/Main.cpp
    llvm/branches/R600/lib/TableGen/Makefile
    llvm/branches/R600/lib/TableGen/Record.cpp
    llvm/branches/R600/lib/Target/ARM/ARMBaseInstrInfo.cpp
    llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.cpp
    llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.h
    llvm/branches/R600/lib/Target/ARM/ARMCallingConv.td
    llvm/branches/R600/lib/Target/ARM/ARMExpandPseudoInsts.cpp
    llvm/branches/R600/lib/Target/ARM/ARMFastISel.cpp
    llvm/branches/R600/lib/Target/ARM/ARMFrameLowering.cpp
    llvm/branches/R600/lib/Target/ARM/ARMISelLowering.cpp
    llvm/branches/R600/lib/Target/ARM/ARMInstrInfo.td
    llvm/branches/R600/lib/Target/ARM/ARMInstrNEON.td
    llvm/branches/R600/lib/Target/ARM/ARMInstrThumb.td
    llvm/branches/R600/lib/Target/ARM/ARMInstrThumb2.td
    llvm/branches/R600/lib/Target/ARM/ARMRegisterInfo.td
    llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.cpp
    llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.h
    llvm/branches/R600/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
    llvm/branches/R600/lib/Target/ARM/CMakeLists.txt
    llvm/branches/R600/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
    llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
    llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.h
    llvm/branches/R600/lib/Target/CMakeLists.txt
    llvm/branches/R600/lib/Target/CellSPU/SPUTargetMachine.cpp
    llvm/branches/R600/lib/Target/CppBackend/CPPBackend.cpp
    llvm/branches/R600/lib/Target/Hexagon/HexagonInstrFormats.td
    llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.cpp
    llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.td
    llvm/branches/R600/lib/Target/Hexagon/HexagonMachineScheduler.cpp
    llvm/branches/R600/lib/Target/Hexagon/HexagonTargetMachine.cpp
    llvm/branches/R600/lib/Target/MBlaze/AsmParser/MBlazeAsmParser.cpp
    llvm/branches/R600/lib/Target/MBlaze/CMakeLists.txt
    llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.cpp
    llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.h
    llvm/branches/R600/lib/Target/MSP430/MSP430ISelLowering.cpp
    llvm/branches/R600/lib/Target/MSP430/MSP430TargetMachine.cpp
    llvm/branches/R600/lib/Target/Mips/AsmParser/MipsAsmParser.cpp
    llvm/branches/R600/lib/Target/Mips/CMakeLists.txt
    llvm/branches/R600/lib/Target/Mips/Mips16FrameLowering.cpp
    llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.cpp
    llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.h
    llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.td
    llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.cpp
    llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.h
    llvm/branches/R600/lib/Target/Mips/Mips64InstrInfo.td
    llvm/branches/R600/lib/Target/Mips/MipsAsmPrinter.cpp
    llvm/branches/R600/lib/Target/Mips/MipsCallingConv.td
    llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.cpp
    llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.h
    llvm/branches/R600/lib/Target/Mips/MipsISelDAGToDAG.cpp
    llvm/branches/R600/lib/Target/Mips/MipsISelLowering.cpp
    llvm/branches/R600/lib/Target/Mips/MipsISelLowering.h
    llvm/branches/R600/lib/Target/Mips/MipsInstrFPU.td
    llvm/branches/R600/lib/Target/Mips/MipsInstrInfo.td
    llvm/branches/R600/lib/Target/Mips/MipsLongBranch.cpp
    llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.cpp
    llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.h
    llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.cpp
    llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.td
    llvm/branches/R600/lib/Target/Mips/MipsSEFrameLowering.cpp
    llvm/branches/R600/lib/Target/Mips/MipsSEInstrInfo.cpp
    llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.cpp
    llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.h
    llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.cpp
    llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.h
    llvm/branches/R600/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
    llvm/branches/R600/lib/Target/NVPTX/NVPTXTargetMachine.cpp
    llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp
    llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp
    llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h
    llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCAsmPrinter.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCCallingConv.td
    llvm/branches/R600/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.h
    llvm/branches/R600/lib/Target/PowerPC/PPCInstr64Bit.td
    llvm/branches/R600/lib/Target/PowerPC/PPCRegisterInfo.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.cpp
    llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.h
    llvm/branches/R600/lib/Target/PowerPC/PPCTargetMachine.cpp
    llvm/branches/R600/lib/Target/Sparc/SparcTargetMachine.cpp
    llvm/branches/R600/lib/Target/TargetTransformImpl.cpp
    llvm/branches/R600/lib/Target/X86/AsmParser/X86AsmParser.cpp
    llvm/branches/R600/lib/Target/X86/CMakeLists.txt
    llvm/branches/R600/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
    llvm/branches/R600/lib/Target/X86/InstPrinter/X86ATTInstPrinter.cpp
    llvm/branches/R600/lib/Target/X86/MCTargetDesc/X86AsmBackend.cpp
    llvm/branches/R600/lib/Target/X86/MCTargetDesc/X86ELFObjectWriter.cpp
    llvm/branches/R600/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h
    llvm/branches/R600/lib/Target/X86/X86.td
    llvm/branches/R600/lib/Target/X86/X86AsmPrinter.cpp
    llvm/branches/R600/lib/Target/X86/X86CallingConv.td
    llvm/branches/R600/lib/Target/X86/X86FastISel.cpp
    llvm/branches/R600/lib/Target/X86/X86FrameLowering.cpp
    llvm/branches/R600/lib/Target/X86/X86ISelLowering.cpp
    llvm/branches/R600/lib/Target/X86/X86ISelLowering.h
    llvm/branches/R600/lib/Target/X86/X86InstrFragmentsSIMD.td
    llvm/branches/R600/lib/Target/X86/X86InstrInfo.cpp
    llvm/branches/R600/lib/Target/X86/X86InstrMMX.td
    llvm/branches/R600/lib/Target/X86/X86InstrSSE.td
    llvm/branches/R600/lib/Target/X86/X86RegisterInfo.cpp
    llvm/branches/R600/lib/Target/X86/X86TargetMachine.cpp
    llvm/branches/R600/lib/Target/X86/X86TargetMachine.h
    llvm/branches/R600/lib/Target/XCore/XCoreTargetMachine.cpp
    llvm/branches/R600/lib/Transforms/IPO/ExtractGV.cpp
    llvm/branches/R600/lib/Transforms/IPO/FunctionAttrs.cpp
    llvm/branches/R600/lib/Transforms/IPO/IPO.cpp
    llvm/branches/R600/lib/Transforms/IPO/Internalize.cpp
    llvm/branches/R600/lib/Transforms/IPO/PassManagerBuilder.cpp
    llvm/branches/R600/lib/Transforms/InstCombine/InstCombine.h
    llvm/branches/R600/lib/Transforms/InstCombine/InstCombineCasts.cpp
    llvm/branches/R600/lib/Transforms/InstCombine/InstCombineCompares.cpp
    llvm/branches/R600/lib/Transforms/InstCombine/InstructionCombining.cpp
    llvm/branches/R600/lib/Transforms/Instrumentation/AddressSanitizer.cpp
    llvm/branches/R600/lib/Transforms/Instrumentation/BoundsChecking.cpp
    llvm/branches/R600/lib/Transforms/Scalar/CodeGenPrepare.cpp
    llvm/branches/R600/lib/Transforms/Scalar/GVN.cpp
    llvm/branches/R600/lib/Transforms/Scalar/IndVarSimplify.cpp
    llvm/branches/R600/lib/Transforms/Scalar/MemCpyOptimizer.cpp
    llvm/branches/R600/lib/Transforms/Scalar/SROA.cpp
    llvm/branches/R600/lib/Transforms/Scalar/SimplifyCFGPass.cpp
    llvm/branches/R600/lib/Transforms/Scalar/SimplifyLibCalls.cpp
    llvm/branches/R600/lib/Transforms/Scalar/TailRecursionElimination.cpp
    llvm/branches/R600/lib/Transforms/Utils/LCSSA.cpp
    llvm/branches/R600/lib/Transforms/Utils/Local.cpp
    llvm/branches/R600/lib/Transforms/Utils/LoopSimplify.cpp
    llvm/branches/R600/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
    llvm/branches/R600/lib/Transforms/Utils/SimplifyCFG.cpp
    llvm/branches/R600/lib/Transforms/Utils/SimplifyLibCalls.cpp
    llvm/branches/R600/lib/Transforms/Vectorize/BBVectorize.cpp
    llvm/branches/R600/lib/Transforms/Vectorize/LoopVectorize.cpp
    llvm/branches/R600/lib/VMCore/AsmWriter.cpp
    llvm/branches/R600/lib/VMCore/Attributes.cpp
    llvm/branches/R600/lib/VMCore/CMakeLists.txt
    llvm/branches/R600/lib/VMCore/ConstantFold.cpp
    llvm/branches/R600/lib/VMCore/DIBuilder.cpp
    llvm/branches/R600/lib/VMCore/DataLayout.cpp
    llvm/branches/R600/lib/VMCore/Instructions.cpp
    llvm/branches/R600/lib/VMCore/Makefile
    llvm/branches/R600/lib/VMCore/TargetTransformInfo.cpp
    llvm/branches/R600/lib/VMCore/Type.cpp
    llvm/branches/R600/lib/VMCore/User.cpp
    llvm/branches/R600/lib/VMCore/Verifier.cpp
    llvm/branches/R600/projects/CMakeLists.txt
    llvm/branches/R600/projects/sample/autoconf/configure.ac
    llvm/branches/R600/projects/sample/configure
    llvm/branches/R600/runtime/libprofile/CommonProfiling.c
    llvm/branches/R600/test/Analysis/BasicAA/nocapture.ll
    llvm/branches/R600/test/BugPoint/crash-narrowfunctiontest.ll
    llvm/branches/R600/test/BugPoint/metadata.ll
    llvm/branches/R600/test/BugPoint/remove_arguments_test.ll
    llvm/branches/R600/test/CMakeLists.txt
    llvm/branches/R600/test/CodeGen/ARM/carry.ll
    llvm/branches/R600/test/CodeGen/ARM/coalesce-subregs.ll
    llvm/branches/R600/test/CodeGen/ARM/integer_insertelement.ll
    llvm/branches/R600/test/CodeGen/ARM/vext.ll
    llvm/branches/R600/test/CodeGen/ARM/vget_lane.ll
    llvm/branches/R600/test/CodeGen/Mips/atomic.ll
    llvm/branches/R600/test/CodeGen/Mips/brdelayslot.ll
    llvm/branches/R600/test/CodeGen/Mips/helloworld.ll
    llvm/branches/R600/test/CodeGen/Mips/largeimm1.ll
    llvm/branches/R600/test/CodeGen/Mips/largeimmprinting.ll
    llvm/branches/R600/test/CodeGen/Mips/longbranch.ll
    llvm/branches/R600/test/CodeGen/Mips/mips64-sret.ll
    llvm/branches/R600/test/CodeGen/Mips/null.ll
    llvm/branches/R600/test/CodeGen/Mips/o32_cc_byval.ll
    llvm/branches/R600/test/CodeGen/Mips/tailcall.ll
    llvm/branches/R600/test/CodeGen/PowerPC/coalesce-ext.ll
    llvm/branches/R600/test/CodeGen/PowerPC/int-fp-conv-1.ll
    llvm/branches/R600/test/CodeGen/PowerPC/structsinregs.ll
    llvm/branches/R600/test/CodeGen/PowerPC/vec_cmp.ll
    llvm/branches/R600/test/CodeGen/Thumb2/carry.ll
    llvm/branches/R600/test/CodeGen/Thumb2/thumb2-uxtb.ll
    llvm/branches/R600/test/CodeGen/X86/2012-01-18-vbitcast.ll
    llvm/branches/R600/test/CodeGen/X86/2012-03-15-build_vector_wl.ll
    llvm/branches/R600/test/CodeGen/X86/2012-07-10-extload64.ll
    llvm/branches/R600/test/CodeGen/X86/add-of-carry.ll
    llvm/branches/R600/test/CodeGen/X86/crash.ll
    llvm/branches/R600/test/CodeGen/X86/fast-cc-callee-pops.ll
    llvm/branches/R600/test/CodeGen/X86/fast-cc-merge-stack-adj.ll
    llvm/branches/R600/test/CodeGen/X86/fast-cc-pass-in-regs.ll
    llvm/branches/R600/test/CodeGen/X86/fp-fast.ll
    llvm/branches/R600/test/CodeGen/X86/jump_sign.ll
    llvm/branches/R600/test/CodeGen/X86/mmx-builtins.ll
    llvm/branches/R600/test/CodeGen/X86/ms-inline-asm.ll
    llvm/branches/R600/test/CodeGen/X86/pointer-vector.ll
    llvm/branches/R600/test/CodeGen/X86/promote.ll
    llvm/branches/R600/test/CodeGen/X86/trunc-ext-ld-st.ll
    llvm/branches/R600/test/CodeGen/X86/vec_compare-2.ll
    llvm/branches/R600/test/CodeGen/X86/vec_shuffle-26.ll
    llvm/branches/R600/test/CodeGen/X86/vec_shuffle-30.ll
    llvm/branches/R600/test/CodeGen/X86/widen_cast-1.ll
    llvm/branches/R600/test/CodeGen/X86/widen_load-2.ll
    llvm/branches/R600/test/MC/ARM/basic-thumb-instructions.s
    llvm/branches/R600/test/MC/Disassembler/ARM/thumb-printf.txt
    llvm/branches/R600/test/MC/Disassembler/ARM/thumb-tests.txt
    llvm/branches/R600/test/MC/Disassembler/ARM/thumb1.txt
    llvm/branches/R600/test/MC/Disassembler/ARM/thumb2.txt
    llvm/branches/R600/test/MC/Disassembler/Mips/mips64.txt
    llvm/branches/R600/test/MC/Disassembler/Mips/mips64_le.txt
    llvm/branches/R600/test/MC/Disassembler/Mips/mips64r2.txt
    llvm/branches/R600/test/MC/Disassembler/Mips/mips64r2_le.txt
    llvm/branches/R600/test/MC/Mips/sext_64_32.ll
    llvm/branches/R600/test/Other/extract.ll
    llvm/branches/R600/test/Transforms/BBVectorize/cycle.ll
    llvm/branches/R600/test/Transforms/BBVectorize/lit.local.cfg
    llvm/branches/R600/test/Transforms/BBVectorize/loop1.ll
    llvm/branches/R600/test/Transforms/BBVectorize/search-limit.ll
    llvm/branches/R600/test/Transforms/BBVectorize/simple-int.ll
    llvm/branches/R600/test/Transforms/BBVectorize/simple-ldstr-ptrs.ll
    llvm/branches/R600/test/Transforms/BBVectorize/simple-ldstr.ll
    llvm/branches/R600/test/Transforms/BBVectorize/simple-sel.ll
    llvm/branches/R600/test/Transforms/BBVectorize/simple.ll
    llvm/branches/R600/test/Transforms/GVN/crash.ll
    llvm/branches/R600/test/Transforms/InstCombine/cast.ll
    llvm/branches/R600/test/Transforms/InstCombine/fcmp.ll
    llvm/branches/R600/test/Transforms/InstCombine/strncpy_chk-1.ll
    llvm/branches/R600/test/Transforms/Internalize/2008-05-09-AllButMain.ll
    llvm/branches/R600/test/Transforms/Internalize/2009-01-05-InternalizeAliases.ll
    llvm/branches/R600/test/Transforms/JumpThreading/crash.ll
    llvm/branches/R600/test/Transforms/LICM/2003-12-11-SinkingToPHI.ll
    llvm/branches/R600/test/Transforms/LoopIdiom/basic.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/2012-10-20-infloop.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/gcc-examples.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/increment.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/induction_plus.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/non-const-n.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/read-only.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/reduction.ll
    llvm/branches/R600/test/Transforms/LoopVectorize/scalar-select.ll
    llvm/branches/R600/test/Transforms/SROA/basictest.ll
    llvm/branches/R600/test/Transforms/SROA/big-endian.ll
    llvm/branches/R600/test/Transforms/SROA/vector-promotion.ll
    llvm/branches/R600/test/Transforms/SimplifyCFG/switch_to_lookup_table.ll
    llvm/branches/R600/test/Transforms/TailCallElim/nocapture.ll
    llvm/branches/R600/test/lit.cfg
    llvm/branches/R600/tools/CMakeLists.txt
    llvm/branches/R600/tools/LLVMBuild.txt
    llvm/branches/R600/tools/Makefile
    llvm/branches/R600/tools/gold/gold-plugin.cpp
    llvm/branches/R600/tools/lli/CMakeLists.txt
    llvm/branches/R600/tools/lli/LLVMBuild.txt
    llvm/branches/R600/tools/lli/Makefile
    llvm/branches/R600/tools/lli/RemoteTarget.cpp
    llvm/branches/R600/tools/lli/lli.cpp
    llvm/branches/R600/tools/llvm-ar/CMakeLists.txt
    llvm/branches/R600/tools/llvm-ar/Makefile
    llvm/branches/R600/tools/llvm-ar/llvm-ar.cpp
    llvm/branches/R600/tools/llvm-as/CMakeLists.txt
    llvm/branches/R600/tools/llvm-bcanalyzer/CMakeLists.txt
    llvm/branches/R600/tools/llvm-dis/CMakeLists.txt
    llvm/branches/R600/tools/llvm-extract/llvm-extract.cpp
    llvm/branches/R600/tools/llvm-mc/llvm-mc.cpp
    llvm/branches/R600/tools/llvm-ranlib/CMakeLists.txt
    llvm/branches/R600/tools/llvm-ranlib/Makefile
    llvm/branches/R600/tools/llvm-ranlib/llvm-ranlib.cpp
    llvm/branches/R600/tools/lto/lto.exports
    llvm/branches/R600/tools/opt/opt.cpp
    llvm/branches/R600/unittests/ADT/APFloatTest.cpp
    llvm/branches/R600/unittests/ADT/DenseMapTest.cpp
    llvm/branches/R600/unittests/ExecutionEngine/JIT/JITTest.cpp
    llvm/branches/R600/unittests/ExecutionEngine/JIT/MultiJITTest.cpp
    llvm/branches/R600/unittests/ExecutionEngine/MCJIT/MCJITTest.cpp
    llvm/branches/R600/unittests/ExecutionEngine/MCJIT/SectionMemoryManager.cpp
    llvm/branches/R600/unittests/ExecutionEngine/MCJIT/SectionMemoryManager.h
    llvm/branches/R600/unittests/VMCore/IRBuilderTest.cpp
    llvm/branches/R600/unittests/VMCore/InstructionsTest.cpp
    llvm/branches/R600/utils/TableGen/AsmMatcherEmitter.cpp
    llvm/branches/R600/utils/TableGen/AsmWriterEmitter.cpp
    llvm/branches/R600/utils/TableGen/AsmWriterInst.cpp
    llvm/branches/R600/utils/TableGen/CMakeLists.txt
    llvm/branches/R600/utils/TableGen/CallingConvEmitter.cpp
    llvm/branches/R600/utils/TableGen/CodeGenDAGPatterns.cpp
    llvm/branches/R600/utils/TableGen/CodeGenDAGPatterns.h
    llvm/branches/R600/utils/TableGen/CodeGenInstruction.cpp
    llvm/branches/R600/utils/TableGen/CodeGenInstruction.h
    llvm/branches/R600/utils/TableGen/CodeGenRegisters.cpp
    llvm/branches/R600/utils/TableGen/CodeGenSchedule.cpp
    llvm/branches/R600/utils/TableGen/CodeGenTarget.cpp
    llvm/branches/R600/utils/TableGen/CodeGenTarget.h
    llvm/branches/R600/utils/TableGen/DAGISelMatcherGen.cpp
    llvm/branches/R600/utils/TableGen/DisassemblerEmitter.cpp
    llvm/branches/R600/utils/TableGen/EDEmitter.cpp
    llvm/branches/R600/utils/TableGen/FastISelEmitter.cpp
    llvm/branches/R600/utils/TableGen/FixedLenDecoderEmitter.cpp
    llvm/branches/R600/utils/TableGen/InstrInfoEmitter.cpp
    llvm/branches/R600/utils/TableGen/IntrinsicEmitter.cpp
    llvm/branches/R600/utils/TableGen/Makefile
    llvm/branches/R600/utils/TableGen/PseudoLoweringEmitter.cpp
    llvm/branches/R600/utils/TableGen/RegisterInfoEmitter.cpp
    llvm/branches/R600/utils/TableGen/SetTheory.cpp
    llvm/branches/R600/utils/TableGen/SetTheory.h
    llvm/branches/R600/utils/TableGen/SubtargetEmitter.cpp
    llvm/branches/R600/utils/TableGen/TableGenBackends.h

Modified: llvm/branches/R600/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/Makefile?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/Makefile (original)
+++ llvm/branches/R600/Makefile Tue Nov 13 09:21:47 2012
@@ -68,7 +68,8 @@
 
 ifeq ($(MAKECMDGOALS),install-clang)
   DIRS := tools/clang/tools/driver tools/clang/lib/Headers \
-          tools/clang/tools/libclang tools/clang/tools/c-index-test \
+          tools/clang/tools/libclang \
+          tools/clang/tools/c-index-test \
           tools/clang/include/clang-c \
           tools/clang/runtime tools/clang/docs \
           tools/lto runtime

Modified: llvm/branches/R600/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/autoconf/configure.ac?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/autoconf/configure.ac (original)
+++ llvm/branches/R600/autoconf/configure.ac Tue Nov 13 09:21:47 2012
@@ -363,8 +363,8 @@
   sparc*-*)               llvm_cv_target_arch="Sparc" ;;
   powerpc*-*)             llvm_cv_target_arch="PowerPC" ;;
   arm*-*)                 llvm_cv_target_arch="ARM" ;;
-  mips-*)                 llvm_cv_target_arch="Mips" ;;
-  mipsel-*)               llvm_cv_target_arch="Mips" ;;
+  mips-* | mips64-*)      llvm_cv_target_arch="Mips" ;;
+  mipsel-* | mips64el-*)  llvm_cv_target_arch="Mips" ;;
   xcore-*)                llvm_cv_target_arch="XCore" ;;
   msp430-*)               llvm_cv_target_arch="MSP430" ;;
   hexagon-*)              llvm_cv_target_arch="Hexagon" ;;
@@ -396,8 +396,8 @@
   sparc*-*)               host_arch="Sparc" ;;
   powerpc*-*)             host_arch="PowerPC" ;;
   arm*-*)                 host_arch="ARM" ;;
-  mips-*)                 host_arch="Mips" ;;
-  mipsel-*)               host_arch="Mips" ;;
+  mips-* | mips64-*)      host_arch="Mips" ;;
+  mipsel-* | mips64el-*)  host_arch="Mips" ;;
   xcore-*)                host_arch="XCore" ;;
   msp430-*)               host_arch="MSP430" ;;
   hexagon-*)              host_arch="Hexagon" ;;
@@ -714,6 +714,8 @@
         arm)      TARGETS_TO_BUILD="ARM $TARGETS_TO_BUILD" ;;
         mips)     TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
         mipsel)   TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
+        mips64)   TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
+        mips64el) TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
         spu)      TARGETS_TO_BUILD="CellSPU $TARGETS_TO_BUILD" ;;
         xcore)    TARGETS_TO_BUILD="XCore $TARGETS_TO_BUILD" ;;
         msp430)   TARGETS_TO_BUILD="MSP430 $TARGETS_TO_BUILD" ;;

Modified: llvm/branches/R600/cmake/modules/LLVMProcessSources.cmake
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/cmake/modules/LLVMProcessSources.cmake?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/cmake/modules/LLVMProcessSources.cmake (original)
+++ llvm/branches/R600/cmake/modules/LLVMProcessSources.cmake Tue Nov 13 09:21:47 2012
@@ -48,7 +48,7 @@
     set( f ${CMAKE_CURRENT_SOURCE_DIR}/${s} )
     add_file_dependencies( ${f} ${TABLEGEN_OUTPUT} )
   endforeach(s)
-  if( MSVC_IDE )
+  if( MSVC_IDE OR XCODE )
     # This adds .td and .h files to the Visual Studio solution:
     # FIXME: Shall we handle *.def here?
     add_td_sources(sources)

Modified: llvm/branches/R600/configure
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/configure?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/configure (original)
+++ llvm/branches/R600/configure Tue Nov 13 09:21:47 2012
@@ -3903,8 +3903,8 @@
   sparc*-*)               llvm_cv_target_arch="Sparc" ;;
   powerpc*-*)             llvm_cv_target_arch="PowerPC" ;;
   arm*-*)                 llvm_cv_target_arch="ARM" ;;
-  mips-*)                 llvm_cv_target_arch="Mips" ;;
-  mipsel-*)               llvm_cv_target_arch="Mips" ;;
+  mips-* | mips64-*)      llvm_cv_target_arch="Mips" ;;
+  mipsel-* | mips64el-*)  llvm_cv_target_arch="Mips" ;;
   xcore-*)                llvm_cv_target_arch="XCore" ;;
   msp430-*)               llvm_cv_target_arch="MSP430" ;;
   hexagon-*)              llvm_cv_target_arch="Hexagon" ;;
@@ -3936,8 +3936,8 @@
   sparc*-*)               host_arch="Sparc" ;;
   powerpc*-*)             host_arch="PowerPC" ;;
   arm*-*)                 host_arch="ARM" ;;
-  mips-*)                 host_arch="Mips" ;;
-  mipsel-*)               host_arch="Mips" ;;
+  mips-* | mips64-*)      host_arch="Mips" ;;
+  mipsel-* | mips64el-*)  host_arch="Mips" ;;
   xcore-*)                host_arch="XCore" ;;
   msp430-*)               host_arch="MSP430" ;;
   hexagon-*)              host_arch="Hexagon" ;;
@@ -5428,6 +5428,8 @@
         arm)      TARGETS_TO_BUILD="ARM $TARGETS_TO_BUILD" ;;
         mips)     TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
         mipsel)   TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
+        mips64)   TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
+        mips64el) TARGETS_TO_BUILD="Mips $TARGETS_TO_BUILD" ;;
         spu)      TARGETS_TO_BUILD="CellSPU $TARGETS_TO_BUILD" ;;
         xcore)    TARGETS_TO_BUILD="XCore $TARGETS_TO_BUILD" ;;
         msp430)   TARGETS_TO_BUILD="MSP430 $TARGETS_TO_BUILD" ;;
@@ -10313,7 +10315,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 10316 "configure"
+#line 10318 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H

Modified: llvm/branches/R600/docs/CodeGenerator.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/CodeGenerator.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/CodeGenerator.rst (original)
+++ llvm/branches/R600/docs/CodeGenerator.rst Tue Nov 13 09:21:47 2012
@@ -224,7 +224,7 @@
 ------------------------
 
 The ``DataLayout`` class is the only required target description class, and it
-is the only class that is not extensible (you cannot derived a new class from
+is the only class that is not extensible (you cannot derive a new class from
 it).  ``DataLayout`` specifies information about how the target lays out memory
 for structures, the alignment requirements for various data types, the size of
 pointers in the target, and whether the target is little-endian or
@@ -248,7 +248,7 @@
 * the type to use for shift amounts, and
 
 * various high-level characteristics, like whether it is profitable to turn
-  division by a constant into a multiplication sequence
+  division by a constant into a multiplication sequence.
 
 The ``TargetRegisterInfo`` class
 --------------------------------
@@ -256,10 +256,10 @@
 The ``TargetRegisterInfo`` class is used to describe the register file of the
 target and any interactions between the registers.
 
-Registers in the code generator are represented in the code generator by
-unsigned integers.  Physical registers (those that actually exist in the target
-description) are unique small numbers, and virtual registers are generally
-large.  Note that register ``#0`` is reserved as a flag value.
+Registers are represented in the code generator by unsigned integers.  Physical
+registers (those that actually exist in the target description) are unique
+small numbers, and virtual registers are generally large.  Note that
+register ``#0`` is reserved as a flag value.
 
 Each register in the processor description has an associated
 ``TargetRegisterDesc`` entry, which provides a textual name for the register
@@ -838,8 +838,7 @@
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 The initial SelectionDAG is na\ :raw-html:`ï`\ vely peephole expanded from
-the LLVM input by the ``SelectionDAGLowering`` class in the
-``lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp`` file.  The intent of this pass
+the LLVM input by the ``SelectionDAGBuilder`` class.  The intent of this pass
 is to expose as much low-level, target-specific details to the SelectionDAG as
 possible.  This pass is mostly hard-coded (e.g. an LLVM ``add`` turns into an
 ``SDNode add`` while a ``getelementptr`` is expanded into the obvious

Modified: llvm/branches/R600/docs/CodingStandards.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/CodingStandards.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/CodingStandards.rst (original)
+++ llvm/branches/R600/docs/CodingStandards.rst Tue Nov 13 09:21:47 2012
@@ -862,23 +862,28 @@
 
 You get the idea.
 
-Please be aware that, when adding assert statements, not all compilers are aware
-of the semantics of the assert.  In some places, asserts are used to indicate a
-piece of code that should not be reached.  These are typically of the form:
+In the past, asserts were used to indicate a piece of code that should not be
+reached.  These were typically of the form:
 
 .. code-block:: c++
 
-  assert(0 && "Some helpful error message");
+  assert(0 && "Invalid radix for integer literal");
 
-When used in a function that returns a value, they should be followed with a
-return statement and a comment indicating that this line is never reached.  This
-will prevent a compiler which is unable to deduce that the assert statement
-never returns from generating a warning.
+This has a few issues, the main one being that some compilers might not
+understand the assertion, or warn about a missing return in builds where
+assertions are compiled out.
+
+Today, we have something much better: ``llvm_unreachable``:
 
 .. code-block:: c++
 
-  assert(0 && "Some helpful error message");
-  return 0;
+  llvm_unreachable("Invalid radix for integer literal");
+
+When assertions are enabled, this will print the message if it's ever reached
+and then exit the program. When assertions are disabled (i.e. in release
+builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
+code for this branch. If the compiler does not support this, it will fall back
+to the "abort" implementation.
 
 Another issue is that values used only by assertions will produce an "unused
 value" warning when assertions are disabled.  For example, this code will warn:

Modified: llvm/branches/R600/docs/CommandGuide/FileCheck.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/CommandGuide/FileCheck.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/CommandGuide/FileCheck.rst (original)
+++ llvm/branches/R600/docs/CommandGuide/FileCheck.rst Tue Nov 13 09:21:47 2012
@@ -45,6 +45,11 @@
 
 
 
+**--input-file** *filename*
+
+  File to check (defaults to stdin).
+
+
 **--strict-whitespace**
 
  By default, FileCheck canonicalizes input horizontal whitespace (spaces and
@@ -271,8 +276,9 @@
 The first check line matches a regex (**%[a-z]+**) and captures it into
 the variable "REGISTER".  The second line verifies that whatever is in REGISTER
 occurs later in the file after an "andw".  FileCheck variable references are
-always contained in **[[ ]]** pairs, are named, and their names can be
-name, then it is a definition of the variable, if not, it is a use.
+always contained in **[[ ]]** pairs, and their names can be formed with the
+regex **[a-zA-Z][a-zA-Z0-9]***.  If a colon follows the name, then it is a
+definition of the variable; otherwise, it is a use.
 
 FileCheck variables can be defined multiple times, and uses always get the
 latest value.  Note that variables are all read at the start of a "CHECK" line

Modified: llvm/branches/R600/docs/HowToBuildOnARM.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/HowToBuildOnARM.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/HowToBuildOnARM.rst (original)
+++ llvm/branches/R600/docs/HowToBuildOnARM.rst Tue Nov 13 09:21:47 2012
@@ -27,8 +27,21 @@
 
 #. If you want to run ``make
    check-all`` after building LLVM/Clang, to avoid false alarms (eg, ARCMT
-   failure) please use the following configuration:
+   failure) please use at least the following configuration:
 
    .. code-block:: bash
 
-     $ ../$LLVM_SRC_DIR/configure --with-abi=aapcs
+     $ ../$LLVM_SRC_DIR/configure --with-abi=aapcs-vfp
+
+#. The most popular linaro/ubuntu OS's for ARM boards, eg, the
+   Pandaboard, have become hard-float platforms. The following set
+   of configuration options appears to be a good choice for this
+   platform:
+
+   .. code-block:: bash
+
+     ./configure --build=armv7l-unknown-linux-gnueabihf
+     --host=armv7l-unknown-linux-gnueabihf
+     --target=armv7l-unknown-linux-gnueabihf --with-cpu=cortex-a9
+     --with-float=hard --with-abi=aapcs-vfp --with-fpu=neon
+     --enable-targets=arm --disable-optimized --enable-assertions

Added: llvm/branches/R600/docs/HowToUseInstrMappings.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/HowToUseInstrMappings.rst?rev=167838&view=auto
==============================================================================
--- llvm/branches/R600/docs/HowToUseInstrMappings.rst (added)
+++ llvm/branches/R600/docs/HowToUseInstrMappings.rst Tue Nov 13 09:21:47 2012
@@ -0,0 +1,179 @@
+.. _how_to_use_instruction_mappings:
+
+===============================
+How To Use Instruction Mappings
+===============================
+
+.. sectionauthor:: Jyotsna Verma <jverma at codeaurora.org>
+
+.. contents::
+   :local:
+
+Introduction
+============
+
+This document contains information about adding instruction mapping support
+for a target. The motivation behind this feature comes from the need to switch
+between different instruction formats during various optimizations. One approach
+could be to use switch cases which list all the instructions along with formats
+they can transition to. However, it has large maintenance overhead
+because of the hardcoded instruction names. Also, whenever a new instruction is
+added in the .td files, all the relevant switch cases should be modified
+accordingly. Instead, the same functionality could be achieved with TableGen and
+some support from the .td files for a fraction of maintenance cost.
+
+``InstrMapping`` Class Overview
+===============================
+
+TableGen uses relationship models to map instructions with each other. These
+models are described using ``InstrMapping`` class as a base. Each model sets
+various fields of the ``InstrMapping`` class such that they can uniquely
+describe all the instructions using that model. TableGen parses all the relation
+models and uses the information to construct relation tables which relate
+instructions with each other. These tables are emitted in the
+``XXXInstrInfo.inc`` file along with the functions to query them. Following
+is the definition of ``InstrMapping`` class definied in Target.td file:
+
+.. code-block:: llvm
+
+  class InstrMapping {
+    // Used to reduce search space only to the instructions using this
+    // relation model.
+    string FilterClass;
+
+    // List of fields/attributes that should be same for all the instructions in
+    // a row of the relation table. Think of this as a set of properties shared
+    // by all the instructions related by this relationship.
+    list<string> RowFields = [];
+
+    // List of fields/attributes that are same for all the instructions
+    // in a column of the relation table.
+    list<string> ColFields = [];
+
+    // Values for the fields/attributes listed in 'ColFields' corresponding to
+    // the key instruction. This is the instruction that will be transformed
+    // using this relation model.
+    list<string> KeyCol = [];
+
+    // List of values for the fields/attributes listed in 'ColFields', one for
+    // each column in the relation table. These are the instructions a key
+    // instruction will be transformed into.
+    list<list<string> > ValueCols = [];
+  }
+
+Sample Example
+--------------
+
+Let's say that we want to have a function
+``int getPredOpcode(uint16_t Opcode, enum PredSense inPredSense)`` which
+takes a non-predicated instruction and returns its predicated true or false form
+depending on some input flag, ``inPredSense``. The first step in the process is
+to define a relationship model that relates predicated instructions to their
+non-predicated form by assigning appropriate values to the ``InstrMapping``
+fields. For this relationship, non-predicated instructions are treated as key
+instruction since they are the one used to query the interface function.
+
+.. code-block:: llvm
+
+  def getPredOpcode : InstrMapping {
+    // Choose a FilterClass that is used as a base class for all the
+    // instructions modeling this relationship. This is done to reduce the
+    // search space only to these set of instructions.
+    let FilterClass = "PredRel";
+
+    // Instructions with same values for all the fields in RowFields form a
+    // row in the resulting relation table.
+    // For example, if we want to relate 'ADD' (non-predicated) with 'Add_pt'
+    // (predicated true) and 'Add_pf' (predicated false), then all 3
+    // instructions need to have same value for BaseOpcode field. It can be any
+    // unique value (Ex: XYZ) and should not be shared with any other
+    // instruction not related to 'add'.
+    let RowFields = ["BaseOpcode"];
+
+    // List of attributes that can be used to define key and column instructions
+    // for a relation. Key instruction is passed as an argument
+    // to the function used for querying relation tables. Column instructions
+    // are the instructions they (key) can transform into.
+    //
+    // Here, we choose 'PredSense' as ColFields since this is the unique
+    // attribute of the key (non-predicated) and column (true/false)
+    // instructions involved in this relationship model.
+    let ColFields = ["PredSense"];
+
+    // The key column contains non-predicated instructions.
+    let KeyCol = ["none"];
+
+    // Two value columns - first column contains instructions with
+    // PredSense=true while second column has instructions with PredSense=false.
+    let ValueCols = [["true"], ["false"]];
+  }
+
+TableGen uses the above relationship model to emit relation table that maps
+non-predicated instructions with their predicated forms. It also outputs the
+interface function
+``int getPredOpcode(uint16_t Opcode, enum PredSense inPredSense)`` to query
+the table. Here, Function ``getPredOpcode`` takes two arguments, opcode of the
+current instruction and PredSense of the desired instruction, and returns
+predicated form of the instruction, if found in the relation table.
+In order for an instruction to be added into the relation table, it needs
+to include relevant information in its definition. For example, consider
+following to be the current definitions of ADD, ADD_pt (true) and ADD_pf (false)
+instructions:
+
+.. code-block::llvm
+
+  def ADD : ALU32_rr<(outs IntRegs:$dst), (ins IntRegs:$a, IntRegs:$b),
+              "$dst = add($a, $b)",
+              [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$a),
+                                             (i32 IntRegs:$b)))]>;
+
+  def ADD_Pt : ALU32_rr<(outs IntRegs:$dst),
+                         (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),
+              "if ($p) $dst = add($a, $b)",
+              []>;
+
+  def ADD_Pf : ALU32_rr<(outs IntRegs:$dst),
+                         (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),
+              "if (!$p) $dst = add($a, $b)",
+              []>;
+
+In this step, we modify these instructions to include the information
+required by the relationship model, <tt>getPredOpcode</tt>, so that they can
+be related.
+
+.. code-block::llvm
+
+  def ADD : PredRel, ALU32_rr<(outs IntRegs:$dst), (ins IntRegs:$a, IntRegs:$b),
+              "$dst = add($a, $b)",
+              [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$a),
+                                             (i32 IntRegs:$b)))]> {
+    let BaseOpcode = "ADD";
+    let PredSense = "none";
+  }
+
+  def ADD_Pt : PredRel, ALU32_rr<(outs IntRegs:$dst),
+                         (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),
+              "if ($p) $dst = add($a, $b)",
+              []> {
+    let BaseOpcode = "ADD";
+    let PredSense = "true";
+  }
+
+  def ADD_Pf : PredRel, ALU32_rr<(outs IntRegs:$dst),
+                         (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),
+              "if (!$p) $dst = add($a, $b)",
+              []> {
+    let BaseOpcode = "ADD";
+    let PredSense = "false";
+  }
+
+Please note that all the above instructions use ``PredRel`` as a base class.
+This is extremely important since TableGen uses it as a filter for selecting
+instructions for ``getPredOpcode`` model. Any instruction not derived from
+``PredRel`` is excluded from the analysis. ``BaseOpcode`` is another important
+field. Since it's selected as a ``RowFields`` of the model, it is required
+to have the same value for all 3 instructions in order to be related. Next,
+``PredSense`` is used to determine their column positions by comparing its value
+with ``KeyCol`` and ``ValueCols``. If an instruction sets its ``PredSense``
+value to something not used in the relation model, it will not be assigned
+a column in the relation table.

Propchange: llvm/branches/R600/docs/HowToUseInstrMappings.rst
------------------------------------------------------------------------------
    svn:executable = *

Modified: llvm/branches/R600/docs/LangRef.html
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/LangRef.html?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/LangRef.html (original)
+++ llvm/branches/R600/docs/LangRef.html Tue Nov 13 09:21:47 2012
@@ -5060,7 +5060,7 @@
 
 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
    operation (that is, the alignment of the memory address). A value of 0 or an
-   omitted <tt>align</tt> argument means that the operation has the preferential
+   omitted <tt>align</tt> argument means that the operation has the abi
    alignment for the target. It is the responsibility of the code emitter to
    ensure that the alignment information is correct. Overestimating the
    alignment results in undefined behavior. Underestimating the alignment may
@@ -5141,7 +5141,7 @@
 
 <p>The optional constant "align" argument specifies the alignment of the
    operation (that is, the alignment of the memory address). A value of 0 or an
-   omitted "align" argument means that the operation has the preferential
+   omitted "align" argument means that the operation has the abi
    alignment for the target. It is the responsibility of the code emitter to
    ensure that the alignment information is correct. Overestimating the
    alignment results in an undefined behavior. Underestimating the alignment may

Added: llvm/branches/R600/docs/MarkedUpDisassembly.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/MarkedUpDisassembly.rst?rev=167838&view=auto
==============================================================================
--- llvm/branches/R600/docs/MarkedUpDisassembly.rst (added)
+++ llvm/branches/R600/docs/MarkedUpDisassembly.rst Tue Nov 13 09:21:47 2012
@@ -0,0 +1,88 @@
+.. _marked_up_disassembly:
+
+=======================================
+LLVM's Optional Rich Disassembly Output
+=======================================
+
+.. contents::
+   :local:
+
+Introduction
+============
+
+LLVM's default disassembly output is raw text. To allow consumers more ability
+to introspect the instructions' textual representation or to reformat for a more
+user friendly display there is an optional rich disassembly output.
+
+This optional output is sufficient to reference into individual portions of the
+instruction text. This is intended for clients like disassemblers, list file
+generators, and pretty-printers, which need more than the raw instructions and
+the ability to print them.
+
+To provide this functionality the assembly text is marked up with annotations.
+The markup is simple enough in syntax to be robust even in the case of version
+mismatches between consumers and producers. That is, the syntax generally does
+not carry semantics beyond "this text has an annotation," so consumers can
+simply ignore annotations they do not understand or do not care about.
+
+After calling ``LLVMCreateDisasm()`` to create a disassembler context the
+optional output is enable with this call:
+
+.. code-block:: c
+
+    LLVMSetDisasmOptions(DC, LLVMDisassembler_Option_UseMarkup);
+
+Then subsequent calls to ``LLVMDisasmInstruction()`` will return output strings
+with the marked up annotations.
+
+Instruction Annotations
+=======================
+
+.. _contextual markups:
+
+Contextual markups
+------------------
+
+Annoated assembly display will supply contextual markup to help clients more
+efficiently implement things like pretty printers. Most markup will be target
+independent, so clients can effectively provide good display without any target
+specific knowledge.
+
+Annotated assembly goes through the normal instruction printer, but optionally
+includes contextual tags on portions of the instruction string. An annotation
+is any '<' '>' delimited section of text(1).
+
+.. code-block:: bat
+
+    annotation: '<' tag-name tag-modifier-list ':' annotated-text '>'
+    tag-name: identifier
+    tag-modifier-list: comma delimited identifier list
+
+The tag-name is an identifier which gives the type of the annotation. For the
+first pass, this will be very simple, with memory references, registers, and
+immediates having the tag names "mem", "reg", and "imm", respectively.
+
+The tag-modifier-list is typically additional target-specific context, such as
+register class.
+
+Clients should accept and ignore any tag-names or tag-modifiers they do not
+understand, allowing the annotations to grow in richness without breaking older
+clients.
+
+For example, a possible annotation of an ARM load of a stack-relative location
+might be annotated as:
+
+.. code-block:: nasm
+
+   ldr <reg gpr:r0>, <mem regoffset:[<reg gpr:sp>, <imm:#4>]>
+
+
+1: For assembly dialects in which '<' and/or '>' are legal tokens, a literal token is escaped by following immediately with a repeat of the character.  For example, a literal '<' character is output as '<<' in an annotated assembly string.
+
+C API Details
+-------------
+
+The intended consumers of this information use the C API, therefore the new C
+API function for the disassembler will be added to provide an option to produce
+disassembled instructions with annotations, ``LLVMSetDisasmOptions()`` and the
+``LLVMDisassembler_Option_UseMarkup`` option (see above).

Modified: llvm/branches/R600/docs/Passes.html
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/Passes.html?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/Passes.html (original)
+++ llvm/branches/R600/docs/Passes.html Tue Nov 13 09:21:47 2012
@@ -77,6 +77,7 @@
 <tr><td><a href="#basicaa">-basicaa</a></td><td>Basic Alias Analysis (stateless AA impl)</td></tr>
 <tr><td><a href="#basiccg">-basiccg</a></td><td>Basic CallGraph Construction</td></tr>
 <tr><td><a href="#count-aa">-count-aa</a></td><td>Count Alias Analysis Query Responses</td></tr>
+<tr><td><a href="#da">-da</a></td><td>Dependence Analysis</td></tr>
 <tr><td><a href="#debug-aa">-debug-aa</a></td><td>AA use debugger</td></tr>
 <tr><td><a href="#domfrontier">-domfrontier</a></td><td>Dominance Frontier Construction</td></tr>
 <tr><td><a href="#domtree">-domtree</a></td><td>Dominator Tree Construction</td></tr>
@@ -92,7 +93,6 @@
 <tr><td><a href="#intervals">-intervals</a></td><td>Interval Partition Construction</td></tr>
 <tr><td><a href="#iv-users">-iv-users</a></td><td>Induction Variable Users</td></tr>
 <tr><td><a href="#lazy-value-info">-lazy-value-info</a></td><td>Lazy Value Information Analysis</td></tr>
-<tr><td><a href="#lda">-lda</a></td><td>Loop Dependence Analysis</td></tr>
 <tr><td><a href="#libcall-aa">-libcall-aa</a></td><td>LibCall Alias Analysis</td></tr>
 <tr><td><a href="#lint">-lint</a></td><td>Statically lint-checks LLVM IR</td></tr>
 <tr><td><a href="#loops">-loops</a></td><td>Natural Loop Information</td></tr>
@@ -182,7 +182,6 @@
 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
-<tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr>
 
 
 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
@@ -251,6 +250,15 @@
 
 <!-------------------------------------------------------------------------- -->
 <h3>
+  <a name="da">-da: Dependence Analysis</a>
+</h3>
+<div>
+  <p>Dependence analysis framework, which is used to detect dependences in
+  memory accesses.</p>
+</div>
+
+<!-------------------------------------------------------------------------- -->
+<h3>
   <a name="debug-aa">-debug-aa: AA use debugger</a>
 </h3>
 <div>
@@ -433,15 +441,6 @@
 
 <!-------------------------------------------------------------------------- -->
 <h3>
-  <a name="lda">-lda: Loop Dependence Analysis</a>
-</h3>
-<div>
-  <p>Loop dependence analysis framework, which is used to detect dependences in
-  memory accesses in loops.</p>
-</div>
-
-<!-------------------------------------------------------------------------- -->
-<h3>
   <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
 </h3>
 <div>
@@ -1862,22 +1861,6 @@
   </ul>
 </div>
 
-<!-------------------------------------------------------------------------- -->
-<h3>
-  <a name="tailduplicate">-tailduplicate: Tail Duplication</a>
-</h3>
-<div>
-  <p>
-  This pass performs a limited form of tail duplication, intended to simplify
-  CFGs by removing some unconditional branches.  This pass is necessary to
-  straighten out loops created by the C front-end, but also is capable of
-  making other code nicer.  After this pass is run, the CFG simplify pass
-  should be run to clean up the mess.
-  </p>
-</div>
-
-</div>
-
 <!-- ======================================================================= -->
 <h2><a name="utilities">Utility Passes</a></h2>
 <div>

Modified: llvm/branches/R600/docs/Phabricator.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/Phabricator.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/Phabricator.rst (original)
+++ llvm/branches/R600/docs/Phabricator.rst Tue Nov 13 09:21:47 2012
@@ -12,10 +12,16 @@
 Sign up
 -------
 
-Sign up with one of the supported OAuth account types. If
-you use your Subversion user name as Phabricator user name,
-Phabricator will automatically connect your submits to your
-Phabricator user in the `Code Repository Browser`_.
+There are two options to get an account on Phabricator. You can sign up
+immediately with one of the supported OAuth account types if you're comfortable
+with OAuth, but you can also email chandlerc at gmail.com to request an account to
+be created manually without using OAuth. We're working to get support in
+Phabricator to directly create new accounts, but currently this is a manual
+process.
+
+Note that if you use your Subversion user name as Phabricator user name,
+Phabricator will automatically connect your submits to your Phabricator user in
+the `Code Repository Browser`_.
 
 
 Requesting a review via the command line
@@ -44,8 +50,8 @@
 To get a full diff, use one of the following commands (or just use Arcanist
 to upload your patch):
 
-* git diff -U999999 other-branch
-* svn diff --diff-cmd=diff -x -U999999
+* ``git diff -U999999 other-branch``
+* ``svn diff --diff-cmd=diff -x -U999999``
 
 To upload a new patch:
 

Modified: llvm/branches/R600/docs/ReleaseNotes.html
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/ReleaseNotes.html?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/ReleaseNotes.html (original)
+++ llvm/branches/R600/docs/ReleaseNotes.html Tue Nov 13 09:21:47 2012
@@ -468,7 +468,11 @@
 
 <p> Loop Vectorizer - We've added a loop vectorizer and we are now able to
     vectorize small loops. The loop vectorizer is disabled by default and
-    can be enabled using the <b>-mllvm -vectorize</b> flag. <br/>
+    can be enabled using the <b>-mllvm -vectorize-loops</b> flag.
+    The SIMD vector width can be specified using the flag
+    <b>-mllvm -force-vector-width=4</b>.
+    The default value is <b>0</b> which means auto-select.
+    <br/>
     We can now vectorize this code:
 
     <pre class="doc_code">
@@ -478,9 +482,13 @@
     }
     </pre>
 
- </p>
+</p>
+
+<p>SROA - We've re-written SROA to be significantly more powerful.
+<!-- FIXME: Add more text here... --></p>
 
 <ul>
+  <li>Branch weight metadata is preseved through more of the optimizer.</li>
   <li>...</li>
 </ul>
 
@@ -666,6 +674,9 @@
   "TargetTransformInfo" provides a number of low-level interfaces.
   LSR and LowerInvoke already use the new interface. </p>
 
+<p> The TargetData structure has been renamed to DataLayout and moved to VMCore
+to remove a dependency on Target. </p>
+
 <ul>
   <li>...</li>
 </ul>

Modified: llvm/branches/R600/docs/TestingGuide.html
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/TestingGuide.html?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/TestingGuide.html (original)
+++ llvm/branches/R600/docs/TestingGuide.html Tue Nov 13 09:21:47 2012
@@ -218,11 +218,11 @@
 
 <p>To run individual tests or subsets of tests, you can use the 'llvm-lit'
 script which is built as part of LLVM. For example, to run the
-'Integer/BitCast.ll' test by itself you can run:</p>
+'Integer/BitPacked.ll' test by itself you can run:</p>
 
 <div class="doc_code">
 <pre>
-% llvm-lit ~/llvm/test/Integer/BitCast.ll 
+% llvm-lit ~/llvm/test/Integer/BitPacked.ll 
 </pre>
 </div>
 

Modified: llvm/branches/R600/docs/WritingAnLLVMBackend.html
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/WritingAnLLVMBackend.html?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/WritingAnLLVMBackend.html (original)
+++ llvm/branches/R600/docs/WritingAnLLVMBackend.html Tue Nov 13 09:21:47 2012
@@ -32,6 +32,7 @@
   <li><a href="#InstructionSet">Instruction Set</a>
   <ul>  
     <li><a href="#operandMapping">Instruction Operand Mapping</a></li>
+    <li><a href="#relationMapping">Instruction Relation Mapping</a></li>
     <li><a href="#implementInstr">Implement a subclass of TargetInstrInfo</a></li>
     <li><a href="#branchFolding">Branch Folding and If Conversion</a></li>
   </ul></li>
@@ -1259,6 +1260,29 @@
 
 <!-- ======================================================================= -->
 <h3>
+  <a name="relationMapping">Instruction Relation Mapping</a>
+</h3>
+
+<div>
+
+<p>
+This TableGen feature is used to relate instructions with each other. It is
+particularly useful when you have multiple instruction formats and need to
+switch between them after instruction selection. This entire feature is driven
+by relation models which can be defined in <tt>XXXInstrInfo.td</tt> files
+according to the target-specific instruction set. Relation models are defined
+using <tt>InstrMapping</tt> class as a base. TableGen parses all the models
+and generates instruction relation maps using the specified information.
+Relation maps are emitted as tables in the <tt>XXXGenInstrInfo.inc</tt> file
+along with the functions to query them. For the detailed information on how to
+use this feature, please refer to
+<a href="HowToUseInstrMappings.html">How to add Instruction Mappings</a>
+document.
+</p>
+</div>
+
+<!-- ======================================================================= -->
+<h3>
   <a name="implementInstr">Implement a subclass of </a>
   <a href="CodeGenerator.html#targetinstrinfo">TargetInstrInfo</a>
 </h3>

Modified: llvm/branches/R600/docs/subsystems.rst
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/docs/subsystems.rst?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/docs/subsystems.rst (original)
+++ llvm/branches/R600/docs/subsystems.rst Tue Nov 13 09:21:47 2012
@@ -17,6 +17,7 @@
    TableGenFundamentals
    DebuggingJITedCode
    GoldPlugin
+   MarkedUpDisassembly
 
 * `Writing an LLVM Pass <WritingAnLLVMPass.html>`_
     
@@ -98,3 +99,8 @@
    architecture.
 
 .. _`Howto: Implementing LLVM Integrated Assembler`: http://www.embecosm.com/download/ean10.html
+
+* :ref:`marked_up_disassembly`
+
+   This document describes the optional rich disassembly output syntax.
+

Modified: llvm/branches/R600/examples/Fibonacci/fibonacci.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/examples/Fibonacci/fibonacci.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/examples/Fibonacci/fibonacci.cpp (original)
+++ llvm/branches/R600/examples/Fibonacci/fibonacci.cpp Tue Nov 13 09:21:47 2012
@@ -37,7 +37,7 @@
 using namespace llvm;
 
 static Function *CreateFibFunction(Module *M, LLVMContext &Context) {
-  // Create the fib function and insert it into module M.  This function is said
+  // Create the fib function and insert it into module M. This function is said
   // to return an int and take an int parameter.
   Function *FibF =
     cast<Function>(M->getOrInsertFunction("fib", Type::getInt32Ty(Context),

Modified: llvm/branches/R600/include/llvm-c/Disassembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm-c/Disassembler.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm-c/Disassembler.h (original)
+++ llvm/branches/R600/include/llvm-c/Disassembler.h Tue Nov 13 09:21:47 2012
@@ -146,6 +146,15 @@
                                       LLVMSymbolLookupCallback SymbolLookUp);
 
 /**
+ * Set the disassembler's options.  Returns 1 if it can set the Options and 0
+ * otherwise.
+ */
+int LLVMSetDisasmOptions(LLVMDisasmContextRef DC, uint64_t Options);
+
+/* The option to produce marked up assembly. */
+#define LLVMDisassembler_Option_UseMarkup 1
+
+/**
  * Dispose of a disassembler context.
  */
 void LLVMDisasmDispose(LLVMDisasmContextRef DC);

Modified: llvm/branches/R600/include/llvm/ADT/APFloat.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ADT/APFloat.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ADT/APFloat.h (original)
+++ llvm/branches/R600/include/llvm/ADT/APFloat.h Tue Nov 13 09:21:47 2012
@@ -455,14 +455,11 @@
 
     /* The sign bit of this number.  */
     unsigned int sign: 1;
-
-    /* For PPCDoubleDouble, we have a second exponent and sign (the second
-       significand is appended to the first one, although it would be wrong to
-       regard these as a single number for arithmetic purposes).  These fields
-       are not meaningful for any other type. */
-    exponent_t exponent2 : 11;
-    unsigned int sign2: 1;
   };
+
+  // See friend declaration above. This additional declaration is required in
+  // order to compile LLVM with IBM xlC compiler.
+  hash_code hash_value(const APFloat &Arg);
 } /* namespace llvm */
 
 #endif /* LLVM_FLOAT_H */

Modified: llvm/branches/R600/include/llvm/ADT/APInt.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ADT/APInt.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ADT/APInt.h (original)
+++ llvm/branches/R600/include/llvm/ADT/APInt.h Tue Nov 13 09:21:47 2012
@@ -760,7 +760,7 @@
   APInt shl(unsigned shiftAmt) const {
     assert(shiftAmt <= BitWidth && "Invalid shift amount");
     if (isSingleWord()) {
-      if (shiftAmt == BitWidth)
+      if (shiftAmt >= BitWidth)
         return APInt(BitWidth, 0); // avoid undefined shift results
       return APInt(BitWidth, VAL << shiftAmt);
     }
@@ -1780,6 +1780,9 @@
 
 } // End of APIntOps namespace
 
+  // See friend declaration above. This additional declaration is required in
+  // order to compile LLVM with IBM xlC compiler.
+  hash_code hash_value(const APInt &Arg);
 } // End of llvm namespace
 
 #endif

Modified: llvm/branches/R600/include/llvm/ADT/DenseMap.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ADT/DenseMap.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ADT/DenseMap.h (original)
+++ llvm/branches/R600/include/llvm/ADT/DenseMap.h Tue Nov 13 09:21:47 2012
@@ -420,7 +420,7 @@
       NumBuckets = getNumBuckets();
     }
     if (NumBuckets-(NewNumEntries+getNumTombstones()) <= NumBuckets/8) {
-      this->grow(NumBuckets);
+      this->grow(NumBuckets * 2);
       LookupBucketFor(Key, TheBucket);
     }
     assert(TheBucket);
@@ -600,7 +600,7 @@
     unsigned OldNumBuckets = NumBuckets;
     BucketT *OldBuckets = Buckets;
 
-    allocateBuckets(std::max<unsigned>(64, NextPowerOf2(AtLeast)));
+    allocateBuckets(std::max<unsigned>(64, NextPowerOf2(AtLeast-1)));
     assert(Buckets);
     if (!OldBuckets) {
       this->BaseT::initEmpty();
@@ -826,11 +826,11 @@
   }
 
   void grow(unsigned AtLeast) {
-    if (AtLeast > InlineBuckets)
-      AtLeast = std::max<unsigned>(64, NextPowerOf2(AtLeast));
+    if (AtLeast >= InlineBuckets)
+      AtLeast = std::max<unsigned>(64, NextPowerOf2(AtLeast-1));
 
     if (Small) {
-      if (AtLeast <= InlineBuckets)
+      if (AtLeast < InlineBuckets)
         return; // Nothing to do.
 
       // First move the inline buckets into a temporary storage.

Modified: llvm/branches/R600/include/llvm/ADT/StringSet.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ADT/StringSet.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ADT/StringSet.h (original)
+++ llvm/branches/R600/include/llvm/ADT/StringSet.h Tue Nov 13 09:21:47 2012
@@ -29,8 +29,13 @@
       assert(!InLang.empty());
       const char *KeyStart = InLang.data();
       const char *KeyEnd = KeyStart + InLang.size();
-      return base::insert(llvm::StringMapEntry<char>::
-                          Create(KeyStart, KeyEnd, base::getAllocator(), '+'));
+      llvm::StringMapEntry<char> *Entry = llvm::StringMapEntry<char>::
+                            Create(KeyStart, KeyEnd, base::getAllocator(), '+');
+      if (!base::insert(Entry)) {
+        Entry->Destroy(base::getAllocator());
+        return false;
+      }
+      return true;
     }
   };
 }

Modified: llvm/branches/R600/include/llvm/ADT/Triple.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ADT/Triple.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ADT/Triple.h (original)
+++ llvm/branches/R600/include/llvm/ADT/Triple.h Tue Nov 13 09:21:47 2012
@@ -426,11 +426,6 @@
   /// architecture name (e.g., "x86").
   static ArchType getArchTypeForLLVMName(StringRef Str);
 
-  /// getArchTypeForDarwinArchName - Get the architecture type for a "Darwin"
-  /// architecture name, for example as accepted by "gcc -arch" (see also
-  /// arch(3)).
-  static ArchType getArchTypeForDarwinArchName(StringRef Str);
-
   /// @}
 };
 

Modified: llvm/branches/R600/include/llvm/Analysis/DependenceAnalysis.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Analysis/DependenceAnalysis.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Analysis/DependenceAnalysis.h (original)
+++ llvm/branches/R600/include/llvm/Analysis/DependenceAnalysis.h Tue Nov 13 09:21:47 2012
@@ -30,23 +30,17 @@
 #ifndef LLVM_ANALYSIS_DEPENDENCEANALYSIS_H
 #define LLVM_ANALYSIS_DEPENDENCEANALYSIS_H
 
-#include "llvm/BasicBlock.h"
-#include "llvm/Function.h"
-#include "llvm/Instruction.h"
+#include "llvm/Instructions.h"
 #include "llvm/Pass.h"
 #include "llvm/ADT/SmallBitVector.h"
-#include "llvm/Analysis/ScalarEvolution.h"
-#include "llvm/Analysis/ScalarEvolutionExpressions.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/LoopInfo.h"
-#include "llvm/Support/raw_ostream.h"
-
 
 namespace llvm {
   class AliasAnalysis;
+  class Loop;
+  class LoopInfo;
   class ScalarEvolution;
   class SCEV;
-  class Value;
+  class SCEVConstant;
   class raw_ostream;
 
   /// Dependence - This class represents a dependence between two memory
@@ -55,7 +49,7 @@
   /// determine anything beyond the existence of a dependence; that is, it
   /// represents a confused dependence (see also FullDependence). In most
   /// cases (for output, flow, and anti dependences), the dependence implies
-  /// an ordering, where the source must preceed the destination; in contrast,
+  /// an ordering, where the source must precede the destination; in contrast,
   /// input dependences are unordered.
   class Dependence {
   public:
@@ -132,7 +126,7 @@
     virtual bool isConsistent() const { return false; }
 
     /// getLevels - Returns the number of common loops surrounding the
-    /// souce and destination of the dependence.
+    /// source and destination of the dependence.
     virtual unsigned getLevels() const { return 0; }
 
     /// getDirection - Returns the direction associated with a particular
@@ -175,7 +169,7 @@
   /// able to accurately analyze the interaction of the references; that is,
   /// it is not a confused dependence (see Dependence). In most cases
   /// (for output, flow, and anti dependences), the dependence implies an
-  /// ordering, where the source must preceed the destination; in contrast,
+  /// ordering, where the source must precede the destination; in contrast,
   /// input dependences are unordered.
   class FullDependence : public Dependence {
   public:
@@ -201,7 +195,7 @@
     bool isConsistent() const { return Consistent; }
 
     /// getLevels - Returns the number of common loops surrounding the
-    /// souce and destination of the dependence.
+    /// source and destination of the dependence.
     unsigned getLevels() const { return Levels; }
 
     /// getDirection - Returns the direction associated with a particular
@@ -511,7 +505,7 @@
 
     /// isKnownPredicate - Compare X and Y using the predicate Pred.
     /// Basically a wrapper for SCEV::isKnownPredicate,
-    /// but tries harder, especially in the presense of sign and zero
+    /// but tries harder, especially in the presence of sign and zero
     /// extensions and symbolics.
     bool isKnownPredicate(ICmpInst::Predicate Pred,
                           const SCEV *X,
@@ -679,7 +673,7 @@
     /// where i and j are induction variable, c1 and c2 are loop invariant,
     /// and a and b are constants.
     /// Returns true if any possible dependence is disproved.
-    /// Marks the result as inconsistant.
+    /// Marks the result as inconsistent.
     /// Works in some cases that symbolicRDIVtest doesn't,
     /// and vice versa.
     bool exactRDIVtest(const SCEV *SrcCoeff,
@@ -695,7 +689,7 @@
     /// where i and j are induction variable, c1 and c2 are loop invariant,
     /// and a and b are constants.
     /// Returns true if any possible dependence is disproved.
-    /// Marks the result as inconsistant.
+    /// Marks the result as inconsistent.
     /// Works in some cases that exactRDIVtest doesn't,
     /// and vice versa. Can also be used as a backup for
     /// ordinary SIV tests.
@@ -708,7 +702,7 @@
 
     /// gcdMIVtest - Tests an MIV subscript pair for dependence.
     /// Returns true if any possible dependence is disproved.
-    /// Marks the result as inconsistant.
+    /// Marks the result as inconsistent.
     /// Can sometimes disprove the equal direction for 1 or more loops.
     //  Can handle some symbolics that even the SIV tests don't get,
     /// so we use it as a backup for everything.
@@ -718,7 +712,7 @@
 
     /// banerjeeMIVtest - Tests an MIV subscript pair for dependence.
     /// Returns true if any possible dependence is disproved.
-    /// Marks the result as inconsistant.
+    /// Marks the result as inconsistent.
     /// Computes directions.
     bool banerjeeMIVtest(const SCEV *Src,
                          const SCEV *Dst,

Removed: llvm/branches/R600/include/llvm/Analysis/LoopDependenceAnalysis.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Analysis/LoopDependenceAnalysis.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/include/llvm/Analysis/LoopDependenceAnalysis.h (original)
+++ llvm/branches/R600/include/llvm/Analysis/LoopDependenceAnalysis.h (removed)
@@ -1,124 +0,0 @@
-//===- llvm/Analysis/LoopDependenceAnalysis.h --------------- -*- C++ -*---===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// LoopDependenceAnalysis is an LLVM pass that analyses dependences in memory
-// accesses in loops.
-//
-// Please note that this is work in progress and the interface is subject to
-// change.
-//
-// TODO: adapt as interface progresses
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_ANALYSIS_LOOP_DEPENDENCE_ANALYSIS_H
-#define LLVM_ANALYSIS_LOOP_DEPENDENCE_ANALYSIS_H
-
-#include "llvm/ADT/DenseSet.h"
-#include "llvm/ADT/FoldingSet.h"
-#include "llvm/ADT/SmallVector.h"
-#include "llvm/Analysis/LoopPass.h"
-#include "llvm/Support/Allocator.h"
-
-namespace llvm {
-
-class AliasAnalysis;
-class AnalysisUsage;
-class ScalarEvolution;
-class SCEV;
-class Value;
-class raw_ostream;
-
-class LoopDependenceAnalysis : public LoopPass {
-  AliasAnalysis *AA;
-  ScalarEvolution *SE;
-
-  /// L - The loop we are currently analysing.
-  Loop *L;
-
-  /// TODO: doc
-  enum DependenceResult { Independent = 0, Dependent = 1, Unknown = 2 };
-
-  /// TODO: doc
-  struct Subscript {
-    /// TODO: Add distance, direction, breaking conditions, ...
-  };
-
-  /// DependencePair - Represents a data dependence relation between to memory
-  /// reference instructions.
-  struct DependencePair : public FastFoldingSetNode {
-    Value *A;
-    Value *B;
-    DependenceResult Result;
-    SmallVector<Subscript, 4> Subscripts;
-
-    DependencePair(const FoldingSetNodeID &ID, Value *a, Value *b) :
-        FastFoldingSetNode(ID), A(a), B(b), Result(Unknown), Subscripts() {}
-  };
-
-  /// findOrInsertDependencePair - Return true if a DependencePair for the
-  /// given Values already exists, false if a new DependencePair had to be
-  /// created. The third argument is set to the pair found or created.
-  bool findOrInsertDependencePair(Value*, Value*, DependencePair*&);
-
-  /// getLoops - Collect all loops of the loop nest L in which
-  /// a given SCEV is variant.
-  void getLoops(const SCEV*, DenseSet<const Loop*>*) const;
-
-  /// isLoopInvariant - True if a given SCEV is invariant in all loops of the
-  /// loop nest starting at the innermost loop L.
-  bool isLoopInvariant(const SCEV*) const;
-
-  /// isAffine - An SCEV is affine with respect to the loop nest starting at
-  /// the innermost loop L if it is of the form A+B*X where A, B are invariant
-  /// in the loop nest and X is a induction variable in the loop nest.
-  bool isAffine(const SCEV*) const;
-
-  /// TODO: doc
-  bool isZIVPair(const SCEV*, const SCEV*) const;
-  bool isSIVPair(const SCEV*, const SCEV*) const;
-  DependenceResult analyseZIV(const SCEV*, const SCEV*, Subscript*) const;
-  DependenceResult analyseSIV(const SCEV*, const SCEV*, Subscript*) const;
-  DependenceResult analyseMIV(const SCEV*, const SCEV*, Subscript*) const;
-  DependenceResult analyseSubscript(const SCEV*, const SCEV*, Subscript*) const;
-  DependenceResult analysePair(DependencePair*) const;
-
-public:
-  static char ID; // Class identification, replacement for typeinfo
-  LoopDependenceAnalysis() : LoopPass(ID) {
-    initializeLoopDependenceAnalysisPass(*PassRegistry::getPassRegistry());
-  }
-
-  /// isDependencePair - Check whether two values can possibly give rise to
-  /// a data dependence: that is the case if both are instructions accessing
-  /// memory and at least one of those accesses is a write.
-  bool isDependencePair(const Value*, const Value*) const;
-
-  /// depends - Return a boolean indicating if there is a data dependence
-  /// between two instructions.
-  bool depends(Value*, Value*);
-
-  bool runOnLoop(Loop*, LPPassManager&);
-  virtual void releaseMemory();
-  virtual void getAnalysisUsage(AnalysisUsage&) const;
-  void print(raw_ostream&, const Module* = 0) const;
-
-private:
-  FoldingSet<DependencePair> Pairs;
-  BumpPtrAllocator PairAllocator;
-}; // class LoopDependenceAnalysis
-
-// createLoopDependenceAnalysisPass - This creates an instance of the
-// LoopDependenceAnalysis pass.
-//
-LoopPass *createLoopDependenceAnalysisPass();
-
-} // namespace llvm
-
-#endif /* LLVM_ANALYSIS_LOOP_DEPENDENCE_ANALYSIS_H */

Modified: llvm/branches/R600/include/llvm/Analysis/Passes.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Analysis/Passes.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Analysis/Passes.h (original)
+++ llvm/branches/R600/include/llvm/Analysis/Passes.h Tue Nov 13 09:21:47 2012
@@ -187,10 +187,10 @@
 
   //===--------------------------------------------------------------------===//
   //
-  // createLoopDependenceAnalysisPass - This creates an instance of the
-  // LoopDependenceAnalysis pass.
+  // createCostModelAnalysisPass - This creates an instance of the
+  // CostModelAnalysis pass.
   //
-  LoopPass *createLoopDependenceAnalysisPass();
+  FunctionPass *createCostModelAnalysisPass();
 
   //===--------------------------------------------------------------------===//
   //

Modified: llvm/branches/R600/include/llvm/Analysis/ProfileDataLoader.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Analysis/ProfileDataLoader.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Analysis/ProfileDataLoader.h (original)
+++ llvm/branches/R600/include/llvm/Analysis/ProfileDataLoader.h Tue Nov 13 09:21:47 2012
@@ -115,9 +115,6 @@
   /// been counted yet.
   static const unsigned Uncounted;
 
-  /// The maximum value that can be stored in a profiling counter.
-  static const unsigned MaxCount;
-
   /// getNumExecutions - Return the number of times the target program was run
   /// to generate this profiling data.
   unsigned getNumExecutions() const { return CommandLines.size(); }

Modified: llvm/branches/R600/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Analysis/ScalarEvolution.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/branches/R600/include/llvm/Analysis/ScalarEvolution.h Tue Nov 13 09:21:47 2012
@@ -873,6 +873,7 @@
     virtual void releaseMemory();
     virtual void getAnalysisUsage(AnalysisUsage &AU) const;
     virtual void print(raw_ostream &OS, const Module* = 0) const;
+    virtual void verifyAnalysis() const;
 
   private:
     FoldingSet<SCEV> UniqueSCEVs;

Modified: llvm/branches/R600/include/llvm/Attributes.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Attributes.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Attributes.h (original)
+++ llvm/branches/R600/include/llvm/Attributes.h Tue Nov 13 09:21:47 2012
@@ -51,40 +51,41 @@
 
   enum AttrVal {
     // IR-Level Attributes
-    None            = 0,   ///< No attributes have been set
-    AddressSafety   = 1,   ///< Address safety checking is on.
-    Alignment       = 2,   ///< Alignment of parameter (5 bits)
+    None,                  ///< No attributes have been set
+    AddressSafety,         ///< Address safety checking is on.
+    Alignment,             ///< Alignment of parameter (5 bits)
                            ///< stored as log2 of alignment with +1 bias
                            ///< 0 means unaligned different from align 1
-    AlwaysInline    = 3,   ///< inline=always
-    ByVal           = 4,   ///< Pass structure by value
-    InlineHint      = 5,   ///< Source said inlining was desirable
-    InReg           = 6,   ///< Force argument to be passed in register
-    Naked           = 7,   ///< Naked function
-    Nest            = 8,   ///< Nested function static chain
-    NoAlias         = 9,   ///< Considered to not alias after call
-    NoCapture       = 10,  ///< Function creates no aliases of pointer
-    NoImplicitFloat = 11,  ///< Disable implicit floating point insts
-    NoInline        = 12,  ///< inline=never
-    NonLazyBind     = 13,  ///< Function is called early and/or
+    AlwaysInline,          ///< inline=always
+    ByVal,                 ///< Pass structure by value
+    InlineHint,            ///< Source said inlining was desirable
+    InReg,                 ///< Force argument to be passed in register
+    MinSize,               ///< Function must be optimized for size first
+    Naked,                 ///< Naked function
+    Nest,                  ///< Nested function static chain
+    NoAlias,               ///< Considered to not alias after call
+    NoCapture,             ///< Function creates no aliases of pointer
+    NoImplicitFloat,       ///< Disable implicit floating point insts
+    NoInline,              ///< inline=never
+    NonLazyBind,           ///< Function is called early and/or
                            ///< often, so lazy binding isn't worthwhile
-    NoRedZone       = 14,  ///< Disable redzone
-    NoReturn        = 15,  ///< Mark the function as not returning
-    NoUnwind        = 16,  ///< Function doesn't unwind stack
-    OptimizeForSize = 17,  ///< opt_size
-    ReadNone        = 18,  ///< Function does not access memory
-    ReadOnly        = 19,  ///< Function only reads from memory
-    ReturnsTwice    = 20,  ///< Function can return twice
-    SExt            = 21,  ///< Sign extended before/after call
-    StackAlignment  = 22,  ///< Alignment of stack for function (3 bits)
+    NoRedZone,             ///< Disable redzone
+    NoReturn,              ///< Mark the function as not returning
+    NoUnwind,              ///< Function doesn't unwind stack
+    OptimizeForSize,       ///< opt_size
+    ReadNone,              ///< Function does not access memory
+    ReadOnly,              ///< Function only reads from memory
+    ReturnsTwice,          ///< Function can return twice
+    SExt,                  ///< Sign extended before/after call
+    StackAlignment,        ///< Alignment of stack for function (3 bits)
                            ///< stored as log2 of alignment with +1 bias 0
                            ///< means unaligned (different from
                            ///< alignstack={1))
-    StackProtect    = 23,  ///< Stack protection.
-    StackProtectReq = 24,  ///< Stack protection required.
-    StructRet       = 25,  ///< Hidden pointer to structure to return
-    UWTable         = 26,  ///< Function must be in a unwind table
-    ZExt            = 27   ///< Zero extended before/after call
+    StackProtect,          ///< Stack protection.
+    StackProtectReq,       ///< Stack protection required.
+    StructRet,             ///< Hidden pointer to structure to return
+    UWTable,               ///< Function must be in a unwind table
+    ZExt                   ///< Zero extended before/after call
   };
 private:
   AttributesImpl *Attrs;
@@ -152,7 +153,8 @@
       hasAttribute(Attributes::UWTable) ||
       hasAttribute(Attributes::NonLazyBind) ||
       hasAttribute(Attributes::ReturnsTwice) ||
-      hasAttribute(Attributes::AddressSafety);
+      hasAttribute(Attributes::AddressSafety) ||
+      hasAttribute(Attributes::MinSize);
   }
 
   bool operator==(const Attributes &A) const {
@@ -263,7 +265,8 @@
       .removeAttribute(Attributes::UWTable)
       .removeAttribute(Attributes::NonLazyBind)
       .removeAttribute(Attributes::ReturnsTwice)
-      .removeAttribute(Attributes::AddressSafety);
+      .removeAttribute(Attributes::AddressSafety)
+      .removeAttribute(Attributes::MinSize);
   }
 
   uint64_t Raw() const { return Bits; }

Modified: llvm/branches/R600/include/llvm/CallingConv.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CallingConv.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CallingConv.h (original)
+++ llvm/branches/R600/include/llvm/CallingConv.h Tue Nov 13 09:21:47 2012
@@ -112,7 +112,11 @@
     /// Cannot have variable arguments.
     /// Can also be called by the host.
     /// Is externally visible.
-    SPIR_KERNEL = 76
+    SPIR_KERNEL = 76,
+
+    /// Intel_OCL_BI - Calling conventions for Intel OpenCL built-ins
+    Intel_OCL_BI = 77
+
   };
 } // End CallingConv namespace
 

Modified: llvm/branches/R600/include/llvm/CodeGen/GCMetadata.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/GCMetadata.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/GCMetadata.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/GCMetadata.h Tue Nov 13 09:21:47 2012
@@ -122,6 +122,11 @@
       Roots.push_back(GCRoot(Num, Metadata));
     }
 
+    /// removeStackRoot - Removes a root.
+    roots_iterator removeStackRoot(roots_iterator position) {
+      return Roots.erase(position);
+    }
+
     /// addSafePoint - Notes the existence of a safe point. Num is the ID of the
     /// label just prior to the safe point (if the code generator is using
     /// MachineModuleInfo).

Modified: llvm/branches/R600/include/llvm/CodeGen/MachineInstr.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/MachineInstr.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/MachineInstr.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/MachineInstr.h Tue Nov 13 09:21:47 2012
@@ -445,6 +445,11 @@
   /// Instructions with this flag set are not necessarily simple load
   /// instructions, they may load a value and modify it, for example.
   bool mayLoad(QueryType Type = AnyInBundle) const {
+    if (isInlineAsm()) {
+      unsigned ExtraInfo = getOperand(InlineAsm::MIOp_ExtraInfo).getImm();
+      if (ExtraInfo & InlineAsm::Extra_MayLoad)
+        return true;
+    }
     return hasProperty(MCID::MayLoad, Type);
   }
 
@@ -454,6 +459,11 @@
   /// instructions, they may store a modified value based on their operands, or
   /// may not actually modify anything, for example.
   bool mayStore(QueryType Type = AnyInBundle) const {
+    if (isInlineAsm()) {
+      unsigned ExtraInfo = getOperand(InlineAsm::MIOp_ExtraInfo).getImm();
+      if (ExtraInfo & InlineAsm::Extra_MayStore)
+        return true;
+    }
     return hasProperty(MCID::MayStore, Type);
   }
 

Modified: llvm/branches/R600/include/llvm/CodeGen/MachineOperand.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/MachineOperand.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/MachineOperand.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/MachineOperand.h Tue Nov 13 09:21:47 2012
@@ -687,6 +687,9 @@
   return OS;
 }
 
+  // See friend declaration above. This additional declaration is required in
+  // order to compile LLVM with IBM xlC compiler.
+  hash_code hash_value(const MachineOperand &MO);
 } // End llvm namespace
 
 #endif

Modified: llvm/branches/R600/include/llvm/CodeGen/MachineScheduler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/MachineScheduler.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/MachineScheduler.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/MachineScheduler.h Tue Nov 13 09:21:47 2012
@@ -154,6 +154,8 @@
 
   bool empty() const { return Queue.empty(); }
 
+  void clear() { Queue.clear(); }
+
   unsigned size() const { return Queue.size(); }
 
   typedef std::vector<SUnit*>::iterator iterator;
@@ -171,10 +173,12 @@
     SU->NodeQueueId |= ID;
   }
 
-  void remove(iterator I) {
+  iterator remove(iterator I) {
     (*I)->NodeQueueId &= ~ID;
     *I = Queue.back();
+    unsigned idx = I - Queue.begin();
     Queue.pop_back();
+    return Queue.begin() + idx;
   }
 
 #ifndef NDEBUG
@@ -306,6 +310,9 @@
   /// Reinsert debug_values recorded in ScheduleDAGInstrs::DbgValues.
   void placeDebugValues();
 
+  /// \brief dump the scheduled Sequence.
+  void dumpSchedule() const;
+
   // Lesser helpers...
 
   void initRegPressure();

Modified: llvm/branches/R600/include/llvm/CodeGen/PBQP/Graph.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/PBQP/Graph.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/PBQP/Graph.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/PBQP/Graph.h Tue Nov 13 09:21:47 2012
@@ -19,6 +19,7 @@
 
 #include <list>
 #include <map>
+#include <llvm/ADT/ilist.h>
 
 namespace PBQP {
 
@@ -31,16 +32,16 @@
     class NodeEntry;
     class EdgeEntry;
 
-    typedef std::list<NodeEntry> NodeList;
-    typedef std::list<EdgeEntry> EdgeList;
+    typedef llvm::ilist<NodeEntry> NodeList;
+    typedef llvm::ilist<EdgeEntry> EdgeList;
 
   public:
 
-    typedef NodeList::iterator NodeItr;
-    typedef NodeList::const_iterator ConstNodeItr;
+    typedef NodeEntry* NodeItr;
+    typedef const NodeEntry* ConstNodeItr;
 
-    typedef EdgeList::iterator EdgeItr;
-    typedef EdgeList::const_iterator ConstEdgeItr;
+    typedef EdgeEntry* EdgeItr;
+    typedef const EdgeEntry* ConstEdgeItr;
 
   private:
 
@@ -52,12 +53,14 @@
 
   private:
 
-    class NodeEntry {
+    class NodeEntry : public llvm::ilist_node<NodeEntry> {
+      friend struct llvm::ilist_sentinel_traits<NodeEntry>;
     private:
       Vector costs;      
       AdjEdgeList adjEdges;
       unsigned degree;
       void *data;
+      NodeEntry() : costs(0, 0) {}
     public:
       NodeEntry(const Vector &costs) : costs(costs), degree(0) {}
       Vector& getCosts() { return costs; }
@@ -77,12 +80,14 @@
       void* getData() { return data; }
     };
 
-    class EdgeEntry {
+    class EdgeEntry : public llvm::ilist_node<EdgeEntry> {
+      friend struct llvm::ilist_sentinel_traits<EdgeEntry>;
     private:
       NodeItr node1, node2;
       Matrix costs;
       AdjEdgeItr node1AEItr, node2AEItr;
       void *data;
+      EdgeEntry() : costs(0, 0, 0) {}
     public:
       EdgeEntry(NodeItr node1, NodeItr node2, const Matrix &costs)
         : node1(node1), node2(node2), costs(costs) {}

Modified: llvm/branches/R600/include/llvm/CodeGen/RegisterPressure.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/RegisterPressure.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/RegisterPressure.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/RegisterPressure.h Tue Nov 13 09:21:47 2012
@@ -43,7 +43,7 @@
   /// class. This is only useful to account for spilling or rematerialization.
   void decrease(const TargetRegisterClass *RC, const TargetRegisterInfo *TRI);
 
-  void dump(const TargetRegisterInfo *TRI);
+  void dump(const TargetRegisterInfo *TRI) const;
 };
 
 /// RegisterPressure computed within a region of instructions delimited by
@@ -197,6 +197,7 @@
   /// This result is complete if either advance() or recede() has returned true,
   /// or if closeRegion() was explicitly invoked.
   RegisterPressure &getPressure() { return P; }
+  const RegisterPressure &getPressure() const { return P; }
 
   /// Get the register set pressure at the current position, which may be less
   /// than the pressure across the traversed region.

Modified: llvm/branches/R600/include/llvm/CodeGen/ScheduleDAG.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/ScheduleDAG.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/ScheduleDAG.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/ScheduleDAG.h Tue Nov 13 09:21:47 2012
@@ -31,6 +31,7 @@
   class MachineFunction;
   class MachineRegisterInfo;
   class MachineInstr;
+  struct MCSchedClassDesc;
   class TargetRegisterInfo;
   class ScheduleDAG;
   class SDNode;
@@ -52,6 +53,13 @@
       Order        ///< Any other ordering dependency.
     };
 
+    enum OrderKind {
+      Barrier,      ///< An unknown scheduling barrier.
+      MayAliasMem,  ///< Nonvolatile load/Store instructions that may alias.
+      MustAliasMem, ///< Nonvolatile load/Store instructions that must alias.
+      Artificial    ///< Arbitrary weak DAG edge (no actual dependence).
+    };
+
   private:
     /// Dep - A pointer to the depending/depended-on SUnit, and an enum
     /// indicating the kind of the dependency.
@@ -65,20 +73,7 @@
       unsigned Reg;
 
       /// Order - Additional information about Order dependencies.
-      struct {
-        /// isNormalMemory - True if both sides of the dependence
-        /// access memory in non-volatile and fully modeled ways.
-        bool isNormalMemory : 1;
-
-        /// isMustAlias - True if both sides of the dependence are known to
-        /// access the same memory.
-        bool isMustAlias : 1;
-
-        /// isArtificial - True if this is an artificial dependency, meaning
-        /// it is not necessary for program correctness, and may be safely
-        /// deleted if necessary.
-        bool isArtificial : 1;
-      } Order;
+      unsigned OrdKind; // enum OrderKind
     } Contents;
 
     /// Latency - The time associated with this edge. Often this is just
@@ -86,6 +81,9 @@
     /// models may provide additional information about specific edges.
     unsigned Latency;
     /// Record MinLatency seperately from "expected" Latency.
+    ///
+    /// FIXME: this field is not packed on LP64. Convert to 16-bit DAG edge
+    /// latency after introducing saturating truncation.
     unsigned MinLatency;
 
   public:
@@ -95,28 +93,28 @@
     SDep() : Dep(0, Data) {}
 
     /// SDep - Construct an SDep with the specified values.
-    SDep(SUnit *S, Kind kind, unsigned latency = 1, unsigned Reg = 0,
-         bool isNormalMemory = false, bool isMustAlias = false,
-         bool isArtificial = false)
-      : Dep(S, kind), Contents(), Latency(latency), MinLatency(latency) {
+    SDep(SUnit *S, Kind kind, unsigned Reg)
+      : Dep(S, kind), Contents() {
       switch (kind) {
+      default:
+        llvm_unreachable("Reg given for non-register dependence!");
       case Anti:
       case Output:
         assert(Reg != 0 &&
                "SDep::Anti and SDep::Output must use a non-zero Reg!");
-        // fall through
-      case Data:
-        assert(!isMustAlias && "isMustAlias only applies with SDep::Order!");
-        assert(!isArtificial && "isArtificial only applies with SDep::Order!");
         Contents.Reg = Reg;
+        Latency = 0;
         break;
-      case Order:
-        assert(Reg == 0 && "Reg given for non-register dependence!");
-        Contents.Order.isNormalMemory = isNormalMemory;
-        Contents.Order.isMustAlias = isMustAlias;
-        Contents.Order.isArtificial = isArtificial;
+      case Data:
+        Contents.Reg = Reg;
+        Latency = 1;
         break;
       }
+      MinLatency = Latency;
+    }
+    SDep(SUnit *S, OrderKind kind)
+      : Dep(S, Order), Contents(), Latency(0), MinLatency(0) {
+      Contents.OrdKind = kind;
     }
 
     /// Return true if the specified SDep is equivalent except for latency.
@@ -128,10 +126,7 @@
       case Output:
         return Contents.Reg == Other.Contents.Reg;
       case Order:
-        return Contents.Order.isNormalMemory ==
-                 Other.Contents.Order.isNormalMemory &&
-               Contents.Order.isMustAlias == Other.Contents.Order.isMustAlias &&
-               Contents.Order.isArtificial == Other.Contents.Order.isArtificial;
+        return Contents.OrdKind == Other.Contents.OrdKind;
       }
       llvm_unreachable("Invalid dependency kind!");
     }
@@ -194,20 +189,21 @@
     /// memory accesses where both sides of the dependence access memory
     /// in non-volatile and fully modeled ways.
     bool isNormalMemory() const {
-      return getKind() == Order && Contents.Order.isNormalMemory;
+      return getKind() == Order && (Contents.OrdKind == MayAliasMem
+                                    || Contents.OrdKind == MustAliasMem);
     }
 
     /// isMustAlias - Test if this is an Order dependence that is marked
     /// as "must alias", meaning that the SUnits at either end of the edge
     /// have a memory dependence on a known memory location.
     bool isMustAlias() const {
-      return getKind() == Order && Contents.Order.isMustAlias;
+      return getKind() == Order && Contents.OrdKind == MustAliasMem;
     }
 
     /// isArtificial - Test if this is an Order dependence that is marked
     /// as "artificial", meaning it isn't necessary for correctness.
     bool isArtificial() const {
-      return getKind() == Order && Contents.Order.isArtificial;
+      return getKind() == Order && Contents.OrdKind == Artificial;
     }
 
     /// isAssignedRegDep - Test if this is a Data dependence that is
@@ -254,6 +250,8 @@
                                         // this node was cloned.
                                         // (SD scheduling only)
 
+    const MCSchedClassDesc *SchedClass; // NULL or resolved SchedClass.
+
     // Preds/Succs - The SUnits before/after us in the graph.
     SmallVector<SDep, 4> Preds;  // All sunit predecessors.
     SmallVector<SDep, 4> Succs;  // All sunit successors.
@@ -301,7 +299,7 @@
     /// SUnit - Construct an SUnit for pre-regalloc scheduling to represent
     /// an SDNode and any nodes flagged to it.
     SUnit(SDNode *node, unsigned nodenum)
-      : Node(node), Instr(0), OrigNode(0), NodeNum(nodenum),
+      : Node(node), Instr(0), OrigNode(0), SchedClass(0), NodeNum(nodenum),
         NodeQueueId(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
         NumSuccsLeft(0), NumRegDefsLeft(0), Latency(0),
         isVRegCycle(false), isCall(false), isCallOp(false), isTwoAddress(false),
@@ -315,7 +313,7 @@
     /// SUnit - Construct an SUnit for post-regalloc scheduling to represent
     /// a MachineInstr.
     SUnit(MachineInstr *instr, unsigned nodenum)
-      : Node(0), Instr(instr), OrigNode(0), NodeNum(nodenum),
+      : Node(0), Instr(instr), OrigNode(0), SchedClass(0), NodeNum(nodenum),
         NodeQueueId(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
         NumSuccsLeft(0), NumRegDefsLeft(0), Latency(0),
         isVRegCycle(false), isCall(false), isCallOp(false), isTwoAddress(false),
@@ -328,7 +326,7 @@
 
     /// SUnit - Construct a placeholder SUnit.
     SUnit()
-      : Node(0), Instr(0), OrigNode(0), NodeNum(~0u),
+      : Node(0), Instr(0), OrigNode(0), SchedClass(0), NodeNum(~0u),
         NodeQueueId(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
         NumSuccsLeft(0), NumRegDefsLeft(0), Latency(0),
         isVRegCycle(false), isCall(false), isCallOp(false), isTwoAddress(false),

Modified: llvm/branches/R600/include/llvm/CodeGen/ScheduleDAGInstrs.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/ScheduleDAGInstrs.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/ScheduleDAGInstrs.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/ScheduleDAGInstrs.h Tue Nov 13 09:21:47 2012
@@ -189,6 +189,13 @@
     /// \brief Get the machine model for instruction scheduling.
     const TargetSchedModel *getSchedModel() const { return &SchedModel; }
 
+    /// \brief Resolve and cache a resolved scheduling class for an SUnit.
+    const MCSchedClassDesc *getSchedClass(SUnit *SU) const {
+      if (!SU->SchedClass)
+        SU->SchedClass = SchedModel.resolveSchedClass(SU->getInstr());
+      return SU->SchedClass;
+    }
+
     /// begin - Return an iterator to the top of the current scheduling region.
     MachineBasicBlock::iterator begin() const { return RegionBegin; }
 

Modified: llvm/branches/R600/include/llvm/CodeGen/SelectionDAGNodes.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/SelectionDAGNodes.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/SelectionDAGNodes.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/SelectionDAGNodes.h Tue Nov 13 09:21:47 2012
@@ -1199,9 +1199,6 @@
   /// have to duplicate its logic everywhere it's called.
   bool isExactlyValue(double V) const {
     bool ignored;
-    // convert is not supported on this type
-    if (&Value->getValueAPF().getSemantics() == &APFloat::PPCDoubleDouble)
-      return false;
     APFloat Tmp(V);
     Tmp.convert(Value->getValueAPF().getSemantics(),
                 APFloat::rmNearestTiesToEven, &ignored);

Modified: llvm/branches/R600/include/llvm/CodeGen/TargetSchedule.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/CodeGen/TargetSchedule.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/CodeGen/TargetSchedule.h (original)
+++ llvm/branches/R600/include/llvm/CodeGen/TargetSchedule.h Tue Nov 13 09:21:47 2012
@@ -16,8 +16,10 @@
 #ifndef LLVM_TARGET_TARGETSCHEDMODEL_H
 #define LLVM_TARGET_TARGETSCHEDMODEL_H
 
+#include "llvm/Target/TargetSubtargetInfo.h"
 #include "llvm/MC/MCSchedule.h"
 #include "llvm/MC/MCInstrItineraries.h"
+#include "llvm/ADT/SmallVector.h"
 
 namespace llvm {
 
@@ -34,6 +36,10 @@
   InstrItineraryData InstrItins;
   const TargetSubtargetInfo *STI;
   const TargetInstrInfo *TII;
+
+  SmallVector<unsigned, 16> ResourceFactors;
+  unsigned MicroOpFactor; // Multiply to normalize microops to resource units.
+  unsigned ResourceLCM;   // Resource units per cycle. Latency normalization factor.
 public:
   TargetSchedModel(): STI(0), TII(0) {}
 
@@ -45,6 +51,9 @@
   void init(const MCSchedModel &sm, const TargetSubtargetInfo *sti,
             const TargetInstrInfo *tii);
 
+  /// Return the MCSchedClassDesc for this instruction.
+  const MCSchedClassDesc *resolveSchedClass(const MachineInstr *MI) const;
+
   /// \brief TargetInstrInfo getter.
   const TargetInstrInfo *getInstrInfo() const { return TII; }
 
@@ -76,7 +85,48 @@
   unsigned getIssueWidth() const { return SchedModel.IssueWidth; }
 
   /// \brief Return the number of issue slots required for this MI.
-  unsigned getNumMicroOps(MachineInstr *MI) const;
+  unsigned getNumMicroOps(const MachineInstr *MI,
+                          const MCSchedClassDesc *SC = 0) const;
+
+  /// \brief Get the number of kinds of resources for this target.
+  unsigned getNumProcResourceKinds() const {
+    return SchedModel.getNumProcResourceKinds();
+  }
+
+  /// \brief Get a processor resource by ID for convenience.
+  const MCProcResourceDesc *getProcResource(unsigned PIdx) const {
+    return SchedModel.getProcResource(PIdx);
+  }
+
+  typedef const MCWriteProcResEntry *ProcResIter;
+
+  // \brief Get an iterator into the processor resources consumed by this
+  // scheduling class.
+  ProcResIter getWriteProcResBegin(const MCSchedClassDesc *SC) const {
+    // The subtarget holds a single resource table for all processors.
+    return STI->getWriteProcResBegin(SC);
+  }
+  ProcResIter getWriteProcResEnd(const MCSchedClassDesc *SC) const {
+    return STI->getWriteProcResEnd(SC);
+  }
+
+  /// \brief Multiply the number of units consumed for a resource by this factor
+  /// to normalize it relative to other resources.
+  unsigned getResourceFactor(unsigned ResIdx) const {
+    return ResourceFactors[ResIdx];
+  }
+
+  /// \brief Multiply number of micro-ops by this factor to normalize it
+  /// relative to other resources.
+  unsigned getMicroOpFactor() const {
+    return MicroOpFactor;
+  }
+
+  /// \brief Multiply cycle count by this factor to normalize it relative to
+  /// other resources. This is the number of resource units per cycle.
+  unsigned getLatencyFactor() const {
+    return ResourceLCM;
+  }
 
   /// \brief Compute operand latency based on the available machine model.
   ///
@@ -105,15 +155,11 @@
   unsigned computeOutputLatency(const MachineInstr *DefMI, unsigned DefIdx,
                                 const MachineInstr *DepMI) const;
 
-
 private:
   /// getDefLatency is a helper for computeOperandLatency. Return the
   /// instruction's latency if operand lookup is not required.
   /// Otherwise return -1.
   int getDefLatency(const MachineInstr *DefMI, bool FindMin) const;
-
-  /// Return the MCSchedClassDesc for this instruction.
-  const MCSchedClassDesc *resolveSchedClass(const MachineInstr *MI) const;
 };
 
 } // namespace llvm

Modified: llvm/branches/R600/include/llvm/Config/config.h.cmake
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Config/config.h.cmake?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Config/config.h.cmake (original)
+++ llvm/branches/R600/include/llvm/Config/config.h.cmake Tue Nov 13 09:21:47 2012
@@ -1,6 +1,4 @@
-/**************************************
-** Created by Kevin from config.h.in **
-***************************************/
+/* include/llvm/Config/config.h.cmake corresponding to config.h.in. */
 
 #ifndef CONFIG_H
 #define CONFIG_H

Modified: llvm/branches/R600/include/llvm/Constants.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Constants.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Constants.h (original)
+++ llvm/branches/R600/include/llvm/Constants.h Tue Nov 13 09:21:47 2012
@@ -282,9 +282,6 @@
 
   bool isExactlyValue(double V) const {
     bool ignored;
-    // convert is not supported on this type
-    if (&Val.getSemantics() == &APFloat::PPCDoubleDouble)
-      return false;
     APFloat FV(V);
     FV.convert(Val.getSemantics(), APFloat::rmNearestTiesToEven, &ignored);
     return isExactlyValue(FV);

Modified: llvm/branches/R600/include/llvm/DataLayout.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/DataLayout.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/DataLayout.h (original)
+++ llvm/branches/R600/include/llvm/DataLayout.h Tue Nov 13 09:21:47 2012
@@ -231,7 +231,9 @@
   }
 
   /// Layout pointer alignment
-  unsigned getPointerABIAlignment(unsigned AS)  const {
+  /// FIXME: The defaults need to be removed once all of
+  /// the backends/clients are updated.
+  unsigned getPointerABIAlignment(unsigned AS = 0)  const {
     DenseMap<unsigned, PointerAlignElem>::const_iterator val = Pointers.find(AS);
     if (val == Pointers.end()) {
       val = Pointers.find(0);
@@ -239,7 +241,9 @@
     return val->second.ABIAlign;
   }
   /// Return target's alignment for stack-based pointers
-  unsigned getPointerPrefAlignment(unsigned AS) const {
+  /// FIXME: The defaults need to be removed once all of
+  /// the backends/clients are updated.
+  unsigned getPointerPrefAlignment(unsigned AS = 0) const {
     DenseMap<unsigned, PointerAlignElem>::const_iterator val = Pointers.find(AS);
     if (val == Pointers.end()) {
       val = Pointers.find(0);
@@ -247,7 +251,9 @@
     return val->second.PrefAlign;
   }
   /// Layout pointer size
-  unsigned getPointerSize(unsigned AS)          const {
+  /// FIXME: The defaults need to be removed once all of
+  /// the backends/clients are updated.
+  unsigned getPointerSize(unsigned AS = 0)          const {
     DenseMap<unsigned, PointerAlignElem>::const_iterator val = Pointers.find(AS);
     if (val == Pointers.end()) {
       val = Pointers.find(0);
@@ -255,12 +261,10 @@
     return val->second.TypeBitWidth;
   }
   /// Layout pointer size, in bits
-  unsigned getPointerSizeInBits(unsigned AS)    const {
-    DenseMap<unsigned, PointerAlignElem>::const_iterator val = Pointers.find(AS);
-    if (val == Pointers.end()) {
-      val = Pointers.find(0);
-    }
-    return 8*val->second.TypeBitWidth;
+  /// FIXME: The defaults need to be removed once all of
+  /// the backends/clients are updated.
+  unsigned getPointerSizeInBits(unsigned AS = 0)    const {
+    return getPointerSize(AS) * 8;
   }
   /// Size examples:
   ///
@@ -337,12 +341,15 @@
   ///
   unsigned getPreferredTypeAlignmentShift(Type *Ty) const;
 
-  /// getIntPtrType - Return an unsigned integer type that is the same size or
-  /// greater to the host pointer size.
-  /// FIXME: Need to remove the default argument when the rest of the LLVM code
-  /// base has been updated.
+  /// getIntPtrType - Return an integer type with size at least as big as that
+  /// of a pointer in the given address space.
   IntegerType *getIntPtrType(LLVMContext &C, unsigned AddressSpace = 0) const;
 
+  /// getIntPtrType - Return an integer (vector of integer) type with size at
+  /// least as big as that of a pointer of the given pointer (vector of pointer)
+  /// type.
+  Type *getIntPtrType(Type *) const;
+
   /// getIndexedOffset - return the offset from the beginning of the type for
   /// the specified indices.  This is used to implement getelementptr.
   ///

Modified: llvm/branches/R600/include/llvm/ExecutionEngine/ExecutionEngine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ExecutionEngine/ExecutionEngine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ExecutionEngine/ExecutionEngine.h (original)
+++ llvm/branches/R600/include/llvm/ExecutionEngine/ExecutionEngine.h Tue Nov 13 09:21:47 2012
@@ -249,6 +249,13 @@
                      "EE!");
   }
 
+  // finalizeObject - This method should be called after sections within an
+  // object have been relocated using mapSectionAddress.  When this method is
+  // called the MCJIT execution engine will reapply relocations for a loaded
+  // object.  This method has no effect for the legacy JIT engine or the
+  // interpeter.
+  virtual void finalizeObject() {}
+
   /// runStaticConstructorsDestructors - This method is used to execute all of
   /// the static constructors or destructors for a program.
   ///

Modified: llvm/branches/R600/include/llvm/ExecutionEngine/JITEventListener.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ExecutionEngine/JITEventListener.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ExecutionEngine/JITEventListener.h (original)
+++ llvm/branches/R600/include/llvm/ExecutionEngine/JITEventListener.h Tue Nov 13 09:21:47 2012
@@ -26,6 +26,7 @@
 class MachineFunction;
 class OProfileWrapper;
 class IntelJITEventsWrapper;
+class ObjectImage;
 
 /// JITEvent_EmittedFunctionDetails - Helper struct for containing information
 /// about a generated machine code function.
@@ -76,6 +77,20 @@
   /// matching NotifyFreeingMachineCode call.
   virtual void NotifyFreeingMachineCode(void *) {}
 
+  /// NotifyObjectEmitted - Called after an object has been successfully
+  /// emitted to memory.  NotifyFunctionEmitted will not be called for
+  /// individual functions in the object.
+  ///
+  /// ELF-specific information
+  /// The ObjectImage contains the generated object image
+  /// with section headers updated to reflect the address at which sections
+  /// were loaded and with relocations performed in-place on debug sections.
+  virtual void NotifyObjectEmitted(const ObjectImage &Obj) {}
+
+  /// NotifyFreeingObject - Called just before the memory associated with
+  /// a previously emitted object is released.
+  virtual void NotifyFreeingObject(const ObjectImage &Obj) {}
+
 #if LLVM_USE_INTEL_JITEVENTS
   // Construct an IntelJITEventListener
   static JITEventListener *createIntelJITEventListener();

Modified: llvm/branches/R600/include/llvm/ExecutionEngine/JITMemoryManager.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ExecutionEngine/JITMemoryManager.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ExecutionEngine/JITMemoryManager.h (original)
+++ llvm/branches/R600/include/llvm/ExecutionEngine/JITMemoryManager.h Tue Nov 13 09:21:47 2012
@@ -10,7 +10,9 @@
 #ifndef LLVM_EXECUTION_ENGINE_JIT_MEMMANAGER_H
 #define LLVM_EXECUTION_ENGINE_JIT_MEMMANAGER_H
 
+#include "llvm/ExecutionEngine/RuntimeDyld.h"
 #include "llvm/Support/DataTypes.h"
+
 #include <string>
 
 namespace llvm {
@@ -22,7 +24,7 @@
 /// memory for the code generated by the JIT.  This can be reimplemented by
 /// clients that have a strong desire to control how the layout of JIT'd memory
 /// works.
-class JITMemoryManager {
+class JITMemoryManager : public RTDyldMemoryManager {
 protected:
   bool HasGOT;
 
@@ -47,17 +49,6 @@
   /// debugging, and may be turned on by default in debug mode.
   virtual void setPoisonMemory(bool poison) = 0;
 
-  /// getPointerToNamedFunction - This method returns the address of the
-  /// specified function. As such it is only useful for resolving library
-  /// symbols, not code generated symbols.
-  ///
-  /// If AbortOnFailure is false and no function with the given name is
-  /// found, this function silently returns a null pointer. Otherwise,
-  /// it prints a message to stderr and aborts.
-  ///
-  virtual void *getPointerToNamedFunction(const std::string &Name,
-                                          bool AbortOnFailure = true) = 0;
-
   //===--------------------------------------------------------------------===//
   // Global Offset Table Management
   //===--------------------------------------------------------------------===//
@@ -112,22 +103,6 @@
   virtual void endFunctionBody(const Function *F, uint8_t *FunctionStart,
                                uint8_t *FunctionEnd) = 0;
 
-  /// allocateCodeSection - Allocate a memory block of (at least) the given
-  /// size suitable for executable code. The SectionID is a unique identifier
-  /// assigned by the JIT and passed through to the memory manager for
-  /// the instance class to use if it needs to communicate to the JIT about
-  /// a given section after the fact.
-  virtual uint8_t *allocateCodeSection(uintptr_t Size, unsigned Alignment,
-                                       unsigned SectionID) = 0;
-
-  /// allocateDataSection - Allocate a memory block of (at least) the given
-  /// size suitable for data. The SectionID is a unique identifier
-  /// assigned by the JIT and passed through to the memory manager for
-  /// the instance class to use if it needs to communicate to the JIT about
-  /// a given section after the fact.
-  virtual uint8_t *allocateDataSection(uintptr_t Size, unsigned Alignment,
-                                       unsigned SectionID) = 0;
-
   /// allocateSpace - Allocate a memory block of the given size.  This method
   /// cannot be called between calls to startFunctionBody and endFunctionBody.
   virtual uint8_t *allocateSpace(intptr_t Size, unsigned Alignment) = 0;

Modified: llvm/branches/R600/include/llvm/ExecutionEngine/RuntimeDyld.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/ExecutionEngine/RuntimeDyld.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/ExecutionEngine/RuntimeDyld.h (original)
+++ llvm/branches/R600/include/llvm/ExecutionEngine/RuntimeDyld.h Tue Nov 13 09:21:47 2012
@@ -24,9 +24,9 @@
 class ObjectImage;
 
 // RuntimeDyld clients often want to handle the memory management of
-// what gets placed where. For JIT clients, this is an abstraction layer
-// over the JITMemoryManager, which references objects by their source
-// representations in LLVM IR.
+// what gets placed where. For JIT clients, this is the subset of
+// JITMemoryManager required for dynamic loading of binaries.
+//
 // FIXME: As the RuntimeDyld fills out, additional routines will be needed
 //        for the varying types of objects to be allocated.
 class RTDyldMemoryManager {
@@ -37,15 +37,26 @@
   virtual ~RTDyldMemoryManager();
 
   /// allocateCodeSection - Allocate a memory block of (at least) the given
-  /// size suitable for executable code.
+  /// size suitable for executable code. The SectionID is a unique identifier
+  /// assigned by the JIT engine, and optionally recorded by the memory manager
+  /// to access a loaded section.
   virtual uint8_t *allocateCodeSection(uintptr_t Size, unsigned Alignment,
                                        unsigned SectionID) = 0;
 
   /// allocateDataSection - Allocate a memory block of (at least) the given
-  /// size suitable for data.
+  /// size suitable for data. The SectionID is a unique identifier
+  /// assigned by the JIT engine, and optionally recorded by the memory manager
+  /// to access a loaded section.
   virtual uint8_t *allocateDataSection(uintptr_t Size, unsigned Alignment,
                                        unsigned SectionID) = 0;
 
+  /// getPointerToNamedFunction - This method returns the address of the
+  /// specified function. As such it is only useful for resolving library
+  /// symbols, not code generated symbols.
+  ///
+  /// If AbortOnFailure is false and no function with the given name is
+  /// found, this function returns a null pointer. Otherwise, it prints a
+  /// message to stderr and aborts.
   virtual void *getPointerToNamedFunction(const std::string &Name,
                                           bool AbortOnFailure = true) = 0;
 };

Modified: llvm/branches/R600/include/llvm/IRBuilder.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/IRBuilder.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/IRBuilder.h (original)
+++ llvm/branches/R600/include/llvm/IRBuilder.h Tue Nov 13 09:21:47 2012
@@ -17,6 +17,7 @@
 
 #include "llvm/Instructions.h"
 #include "llvm/BasicBlock.h"
+#include "llvm/DataLayout.h"
 #include "llvm/LLVMContext.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/StringRef.h"
@@ -266,6 +267,10 @@
     return Type::getInt8PtrTy(Context, AddrSpace);
   }
 
+  IntegerType* getIntPtrTy(DataLayout *DL, unsigned AddrSpace = 0) {
+    return DL->getIntPtrType(Context, AddrSpace);
+  }
+
   //===--------------------------------------------------------------------===//
   // Intrinsic creation methods
   //===--------------------------------------------------------------------===//

Modified: llvm/branches/R600/include/llvm/InitializePasses.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/InitializePasses.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/InitializePasses.h (original)
+++ llvm/branches/R600/include/llvm/InitializePasses.h Tue Nov 13 09:21:47 2012
@@ -88,6 +88,7 @@
 void initializeConstantMergePass(PassRegistry&);
 void initializeConstantPropagationPass(PassRegistry&);
 void initializeMachineCopyPropagationPass(PassRegistry&);
+void initializeCostModelAnalysisPass(PassRegistry&);
 void initializeCorrelatedValuePropagationPass(PassRegistry&);
 void initializeDAEPass(PassRegistry&);
 void initializeDAHPass(PassRegistry&);
@@ -147,7 +148,6 @@
 void initializePathProfileLoaderPassPass(PassRegistry&);
 void initializeLocalStackSlotPassPass(PassRegistry&);
 void initializeLoopDeletionPass(PassRegistry&);
-void initializeLoopDependenceAnalysisPass(PassRegistry&);
 void initializeLoopExtractorPass(PassRegistry&);
 void initializeLoopInfoPass(PassRegistry&);
 void initializeLoopInstSimplifyPass(PassRegistry&);

Modified: llvm/branches/R600/include/llvm/InlineAsm.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/InlineAsm.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/InlineAsm.h (original)
+++ llvm/branches/R600/include/llvm/InlineAsm.h Tue Nov 13 09:21:47 2012
@@ -214,6 +214,8 @@
     Extra_HasSideEffects = 1,
     Extra_IsAlignStack = 2,
     Extra_AsmDialect = 4,
+    Extra_MayLoad = 8,
+    Extra_MayStore = 16,
 
     // Inline asm operands map to multiple SDNode / MachineInstr operands.
     // The first operand is an immediate describing the asm operand, the low

Modified: llvm/branches/R600/include/llvm/InstrTypes.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/InstrTypes.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/InstrTypes.h (original)
+++ llvm/branches/R600/include/llvm/InstrTypes.h Tue Nov 13 09:21:47 2012
@@ -588,7 +588,9 @@
     Type *SrcTy, ///< SrcTy of 1st cast
     Type *MidTy, ///< DstTy of 1st cast & SrcTy of 2nd cast
     Type *DstTy, ///< DstTy of 2nd cast
-    Type *IntPtrTy ///< Integer type corresponding to Ptr types, or null
+    Type *SrcIntPtrTy, ///< Integer type corresponding to Ptr SrcTy, or null
+    Type *MidIntPtrTy, ///< Integer type corresponding to Ptr MidTy, or null
+    Type *DstIntPtrTy  ///< Integer type corresponding to Ptr DstTy, or null
   );
 
   /// @brief Return the opcode of this CastInst

Modified: llvm/branches/R600/include/llvm/Instructions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Instructions.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Instructions.h (original)
+++ llvm/branches/R600/include/llvm/Instructions.h Tue Nov 13 09:21:47 2012
@@ -225,8 +225,9 @@
   const Value *getPointerOperand() const { return getOperand(0); }
   static unsigned getPointerOperandIndex() { return 0U; }
 
+  /// \brief Returns the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
-    return cast<PointerType>(getPointerOperand()->getType())->getAddressSpace();
+    return getPointerOperand()->getType()->getPointerAddressSpace();
   }
 
 
@@ -347,17 +348,9 @@
   const Value *getPointerOperand() const { return getOperand(1); }
   static unsigned getPointerOperandIndex() { return 1U; }
 
+  /// \brief Returns the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
-    if (getPointerOperand()->getType()->isPointerTy())
-      return cast<PointerType>(getPointerOperand()->getType())
-        ->getAddressSpace();
-    if (getPointerOperand()->getType()->isVectorTy()
-        && cast<VectorType>(getPointerOperand()->getType())->isPointerTy())
-      return cast<PointerType>(cast<VectorType>(
-            getPointerOperand()->getType())->getElementType())
-        ->getAddressSpace();
-    llvm_unreachable("Only a vector of pointers or pointers can be used!");
-    return 0;
+    return getPointerOperand()->getType()->getPointerAddressSpace();
   }
 
   // Methods for support type inquiry through isa, cast, and dyn_cast:
@@ -526,8 +519,9 @@
   Value *getNewValOperand() { return getOperand(2); }
   const Value *getNewValOperand() const { return getOperand(2); }
   
+  /// \brief Returns the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
-    return cast<PointerType>(getPointerOperand()->getType())->getAddressSpace();
+    return getPointerOperand()->getType()->getPointerAddressSpace();
   }
   
   // Methods for support type inquiry through isa, cast, and dyn_cast:
@@ -669,8 +663,9 @@
   Value *getValOperand() { return getOperand(1); }
   const Value *getValOperand() const { return getOperand(1); }
 
+  /// \brief Returns the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
-    return cast<PointerType>(getPointerOperand()->getType())->getAddressSpace();
+    return getPointerOperand()->getType()->getPointerAddressSpace();
   }
 
   // Methods for support type inquiry through isa, cast, and dyn_cast:
@@ -771,6 +766,13 @@
     return reinterpret_cast<PointerType*>(Instruction::getType());
   }
 
+  /// \brief Returns the address space of this instruction's pointer type.
+  unsigned getAddressSpace() const {
+    // Note that this is always the same as the pointer operand's address space
+    // and that is cheaper to compute, so cheat here.
+    return getPointerAddressSpace();
+  }
+
   /// getIndexedType - Returns the type of the element that would be loaded with
   /// a load instruction with the specified parameters.
   ///
@@ -781,10 +783,6 @@
   static Type *getIndexedType(Type *Ptr, ArrayRef<Constant *> IdxList);
   static Type *getIndexedType(Type *Ptr, ArrayRef<uint64_t> IdxList);
 
-  /// getAddressSpace - Returns the address space used by the GEP pointer.
-  ///
-  static unsigned getAddressSpace(Value *Ptr);
-
   inline op_iterator       idx_begin()       { return op_begin()+1; }
   inline const_op_iterator idx_begin() const { return op_begin()+1; }
   inline op_iterator       idx_end()         { return op_end(); }
@@ -800,22 +798,23 @@
     return 0U;    // get index for modifying correct operand.
   }
 
-  unsigned getPointerAddressSpace() const {
-    return cast<PointerType>(getPointerOperandType())->getAddressSpace();
-  }
-
   /// getPointerOperandType - Method to return the pointer operand as a
   /// PointerType.
   Type *getPointerOperandType() const {
     return getPointerOperand()->getType();
   }
 
+  /// \brief Returns the address space of the pointer operand.
+  unsigned getPointerAddressSpace() const {
+    return getPointerOperandType()->getPointerAddressSpace();
+  }
+
   /// GetGEPReturnType - Returns the pointer type returned by the GEP
   /// instruction, which may be a vector of pointers.
   static Type *getGEPReturnType(Value *Ptr, ArrayRef<Value *> IdxList) {
     Type *PtrTy = PointerType::get(checkGEPType(
                                    getIndexedType(Ptr->getType(), IdxList)),
-                                   getAddressSpace(Ptr));
+                                   Ptr->getType()->getPointerAddressSpace());
     // Vector GEP
     if (Ptr->getType()->isVectorTy()) {
       unsigned NumElem = cast<VectorType>(Ptr->getType())->getNumElements();
@@ -899,13 +898,13 @@
 /// This instruction compares its operands according to the predicate given
 /// to the constructor. It only operates on integers or pointers. The operands
 /// must be identical types.
-/// @brief Represent an integer comparison operator.
+/// \brief Represent an integer comparison operator.
 class ICmpInst: public CmpInst {
 protected:
-  /// @brief Clone an identical ICmpInst
+  /// \brief Clone an identical ICmpInst
   virtual ICmpInst *clone_impl() const;
 public:
-  /// @brief Constructor with insert-before-instruction semantics.
+  /// \brief Constructor with insert-before-instruction semantics.
   ICmpInst(
     Instruction *InsertBefore,  ///< Where to insert
     Predicate pred,  ///< The predicate to use for the comparison
@@ -926,7 +925,7 @@
            "Invalid operand types for ICmp instruction");
   }
 
-  /// @brief Constructor with insert-at-end semantics.
+  /// \brief Constructor with insert-at-end semantics.
   ICmpInst(
     BasicBlock &InsertAtEnd, ///< Block to insert into.
     Predicate pred,  ///< The predicate to use for the comparison
@@ -947,7 +946,7 @@
            "Invalid operand types for ICmp instruction");
   }
 
-  /// @brief Constructor with no-insertion semantics
+  /// \brief Constructor with no-insertion semantics
   ICmpInst(
     Predicate pred, ///< The predicate to use for the comparison
     Value *LHS,     ///< The left-hand-side of the expression
@@ -969,25 +968,25 @@
   /// For example, EQ->EQ, SLE->SLE, UGT->SGT, etc.
   /// @returns the predicate that would be the result if the operand were
   /// regarded as signed.
-  /// @brief Return the signed version of the predicate
+  /// \brief Return the signed version of the predicate
   Predicate getSignedPredicate() const {
     return getSignedPredicate(getPredicate());
   }
 
   /// This is a static version that you can use without an instruction.
-  /// @brief Return the signed version of the predicate.
+  /// \brief Return the signed version of the predicate.
   static Predicate getSignedPredicate(Predicate pred);
 
   /// For example, EQ->EQ, SLE->ULE, UGT->UGT, etc.
   /// @returns the predicate that would be the result if the operand were
   /// regarded as unsigned.
-  /// @brief Return the unsigned version of the predicate
+  /// \brief Return the unsigned version of the predicate
   Predicate getUnsignedPredicate() const {
     return getUnsignedPredicate(getPredicate());
   }
 
   /// This is a static version that you can use without an instruction.
-  /// @brief Return the unsigned version of the predicate.
+  /// \brief Return the unsigned version of the predicate.
   static Predicate getUnsignedPredicate(Predicate pred);
 
   /// isEquality - Return true if this predicate is either EQ or NE.  This also
@@ -1003,7 +1002,7 @@
   }
 
   /// @returns true if the predicate of this ICmpInst is commutative
-  /// @brief Determine if this relation is commutative.
+  /// \brief Determine if this relation is commutative.
   bool isCommutative() const { return isEquality(); }
 
   /// isRelational - Return true if the predicate is relational (not EQ or NE).
@@ -1019,14 +1018,14 @@
   }
 
   /// Initialize a set of values that all satisfy the predicate with C.
-  /// @brief Make a ConstantRange for a relation with a constant value.
+  /// \brief Make a ConstantRange for a relation with a constant value.
   static ConstantRange makeConstantRange(Predicate pred, const APInt &C);
 
   /// Exchange the two operands to this instruction in such a way that it does
   /// not modify the semantics of the instruction. The predicate value may be
   /// changed to retain the same result if the predicate is order dependent
   /// (e.g. ult).
-  /// @brief Swap operands and adjust predicate.
+  /// \brief Swap operands and adjust predicate.
   void swapOperands() {
     setPredicate(getSwappedPredicate());
     Op<0>().swap(Op<1>());
@@ -1049,13 +1048,13 @@
 /// This instruction compares its operands according to the predicate given
 /// to the constructor. It only operates on floating point values or packed
 /// vectors of floating point values. The operands must be identical types.
-/// @brief Represents a floating point comparison operator.
+/// \brief Represents a floating point comparison operator.
 class FCmpInst: public CmpInst {
 protected:
-  /// @brief Clone an identical FCmpInst
+  /// \brief Clone an identical FCmpInst
   virtual FCmpInst *clone_impl() const;
 public:
-  /// @brief Constructor with insert-before-instruction semantics.
+  /// \brief Constructor with insert-before-instruction semantics.
   FCmpInst(
     Instruction *InsertBefore, ///< Where to insert
     Predicate pred,  ///< The predicate to use for the comparison
@@ -1074,7 +1073,7 @@
            "Invalid operand types for FCmp instruction");
   }
 
-  /// @brief Constructor with insert-at-end semantics.
+  /// \brief Constructor with insert-at-end semantics.
   FCmpInst(
     BasicBlock &InsertAtEnd, ///< Block to insert into.
     Predicate pred,  ///< The predicate to use for the comparison
@@ -1093,7 +1092,7 @@
            "Invalid operand types for FCmp instruction");
   }
 
-  /// @brief Constructor with no-insertion semantics
+  /// \brief Constructor with no-insertion semantics
   FCmpInst(
     Predicate pred, ///< The predicate to use for the comparison
     Value *LHS,     ///< The left-hand-side of the expression
@@ -1111,14 +1110,14 @@
   }
 
   /// @returns true if the predicate of this instruction is EQ or NE.
-  /// @brief Determine if this is an equality predicate.
+  /// \brief Determine if this is an equality predicate.
   bool isEquality() const {
     return getPredicate() == FCMP_OEQ || getPredicate() == FCMP_ONE ||
            getPredicate() == FCMP_UEQ || getPredicate() == FCMP_UNE;
   }
 
   /// @returns true if the predicate of this instruction is commutative.
-  /// @brief Determine if this is a commutative predicate.
+  /// \brief Determine if this is a commutative predicate.
   bool isCommutative() const {
     return isEquality() ||
            getPredicate() == FCMP_FALSE ||
@@ -1128,20 +1127,20 @@
   }
 
   /// @returns true if the predicate is relational (not EQ or NE).
-  /// @brief Determine if this a relational predicate.
+  /// \brief Determine if this a relational predicate.
   bool isRelational() const { return !isEquality(); }
 
   /// Exchange the two operands to this instruction in such a way that it does
   /// not modify the semantics of the instruction. The predicate value may be
   /// changed to retain the same result if the predicate is order dependent
   /// (e.g. ult).
-  /// @brief Swap operands and adjust predicate.
+  /// \brief Swap operands and adjust predicate.
   void swapOperands() {
     setPredicate(getSwappedPredicate());
     Op<0>().swap(Op<1>());
   }
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == Instruction::FCmp;
   }
@@ -1163,12 +1162,12 @@
   void init(Value *Func, const Twine &NameStr);
 
   /// Construct a CallInst given a range of arguments.
-  /// @brief Construct a CallInst from a range of arguments
+  /// \brief Construct a CallInst from a range of arguments
   inline CallInst(Value *Func, ArrayRef<Value *> Args,
                   const Twine &NameStr, Instruction *InsertBefore);
 
   /// Construct a CallInst given a range of arguments.
-  /// @brief Construct a CallInst from a range of arguments
+  /// \brief Construct a CallInst from a range of arguments
   inline CallInst(Value *Func, ArrayRef<Value *> Args,
                   const Twine &NameStr, BasicBlock *InsertAtEnd);
 
@@ -1267,25 +1266,25 @@
   /// removeAttribute - removes the attribute from the list of attributes.
   void removeAttribute(unsigned i, Attributes attr);
 
-  /// @brief Determine whether this call has the given attribute.
+  /// \brief Determine whether this call has the given attribute.
   bool hasFnAttr(Attributes::AttrVal A) const;
 
-  /// @brief Determine whether the call or the callee has the given attributes.
+  /// \brief Determine whether the call or the callee has the given attributes.
   bool paramHasAttr(unsigned i, Attributes::AttrVal A) const;
 
-  /// @brief Extract the alignment for a call or parameter (0=unknown).
+  /// \brief Extract the alignment for a call or parameter (0=unknown).
   unsigned getParamAlignment(unsigned i) const {
     return AttributeList.getParamAlignment(i);
   }
 
-  /// @brief Return true if the call should not be inlined.
+  /// \brief Return true if the call should not be inlined.
   bool isNoInline() const { return hasFnAttr(Attributes::NoInline); }
   void setIsNoInline() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoInline));
   }
 
-  /// @brief Return true if the call can return twice
+  /// \brief Return true if the call can return twice
   bool canReturnTwice() const {
     return hasFnAttr(Attributes::ReturnsTwice);
   }
@@ -1294,7 +1293,7 @@
                  Attributes::get(getContext(), Attributes::ReturnsTwice));
   }
 
-  /// @brief Determine if the call does not access memory.
+  /// \brief Determine if the call does not access memory.
   bool doesNotAccessMemory() const {
     return hasFnAttr(Attributes::ReadNone);
   }
@@ -1303,7 +1302,7 @@
                  Attributes::get(getContext(), Attributes::ReadNone));
   }
 
-  /// @brief Determine if the call does not access or only reads memory.
+  /// \brief Determine if the call does not access or only reads memory.
   bool onlyReadsMemory() const {
     return doesNotAccessMemory() || hasFnAttr(Attributes::ReadOnly);
   }
@@ -1312,28 +1311,28 @@
                  Attributes::get(getContext(), Attributes::ReadOnly));
   }
 
-  /// @brief Determine if the call cannot return.
+  /// \brief Determine if the call cannot return.
   bool doesNotReturn() const { return hasFnAttr(Attributes::NoReturn); }
   void setDoesNotReturn() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoReturn));
   }
 
-  /// @brief Determine if the call cannot unwind.
+  /// \brief Determine if the call cannot unwind.
   bool doesNotThrow() const { return hasFnAttr(Attributes::NoUnwind); }
   void setDoesNotThrow() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoUnwind));
   }
 
-  /// @brief Determine if the call returns a structure through first
+  /// \brief Determine if the call returns a structure through first
   /// pointer argument.
   bool hasStructRetAttr() const {
     // Be friendly and also check the callee.
     return paramHasAttr(1, Attributes::StructRet);
   }
 
-  /// @brief Determine if any call argument is an aggregate passed by value.
+  /// \brief Determine if any call argument is an aggregate passed by value.
   bool hasByValArgument() const {
     for (unsigned I = 0, E = AttributeList.getNumAttrs(); I != E; ++I)
       if (AttributeList.getAttributesAtIndex(I).hasAttribute(Attributes::ByVal))
@@ -2950,14 +2949,14 @@
 
   /// Construct an InvokeInst given a range of arguments.
   ///
-  /// @brief Construct an InvokeInst from a range of arguments
+  /// \brief Construct an InvokeInst from a range of arguments
   inline InvokeInst(Value *Func, BasicBlock *IfNormal, BasicBlock *IfException,
                     ArrayRef<Value *> Args, unsigned Values,
                     const Twine &NameStr, Instruction *InsertBefore);
 
   /// Construct an InvokeInst given a range of arguments.
   ///
-  /// @brief Construct an InvokeInst from a range of arguments
+  /// \brief Construct an InvokeInst from a range of arguments
   inline InvokeInst(Value *Func, BasicBlock *IfNormal, BasicBlock *IfException,
                     ArrayRef<Value *> Args, unsigned Values,
                     const Twine &NameStr, BasicBlock *InsertAtEnd);
@@ -3016,25 +3015,25 @@
   /// removeAttribute - removes the attribute from the list of attributes.
   void removeAttribute(unsigned i, Attributes attr);
 
-  /// @brief Determine whether this call has the NoAlias attribute.
+  /// \brief Determine whether this call has the NoAlias attribute.
   bool hasFnAttr(Attributes::AttrVal A) const;
 
-  /// @brief Determine whether the call or the callee has the given attributes.
+  /// \brief Determine whether the call or the callee has the given attributes.
   bool paramHasAttr(unsigned i, Attributes::AttrVal A) const;
 
-  /// @brief Extract the alignment for a call or parameter (0=unknown).
+  /// \brief Extract the alignment for a call or parameter (0=unknown).
   unsigned getParamAlignment(unsigned i) const {
     return AttributeList.getParamAlignment(i);
   }
 
-  /// @brief Return true if the call should not be inlined.
+  /// \brief Return true if the call should not be inlined.
   bool isNoInline() const { return hasFnAttr(Attributes::NoInline); }
   void setIsNoInline() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoInline));
   }
 
-  /// @brief Determine if the call does not access memory.
+  /// \brief Determine if the call does not access memory.
   bool doesNotAccessMemory() const {
     return hasFnAttr(Attributes::ReadNone);
   }
@@ -3043,7 +3042,7 @@
                  Attributes::get(getContext(), Attributes::ReadNone));
   }
 
-  /// @brief Determine if the call does not access or only reads memory.
+  /// \brief Determine if the call does not access or only reads memory.
   bool onlyReadsMemory() const {
     return doesNotAccessMemory() || hasFnAttr(Attributes::ReadOnly);
   }
@@ -3052,28 +3051,28 @@
                  Attributes::get(getContext(), Attributes::ReadOnly));
   }
 
-  /// @brief Determine if the call cannot return.
+  /// \brief Determine if the call cannot return.
   bool doesNotReturn() const { return hasFnAttr(Attributes::NoReturn); }
   void setDoesNotReturn() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoReturn));
   }
 
-  /// @brief Determine if the call cannot unwind.
+  /// \brief Determine if the call cannot unwind.
   bool doesNotThrow() const { return hasFnAttr(Attributes::NoUnwind); }
   void setDoesNotThrow() {
     addAttribute(AttrListPtr::FunctionIndex,
                  Attributes::get(getContext(), Attributes::NoUnwind));
   }
 
-  /// @brief Determine if the call returns a structure through first
+  /// \brief Determine if the call returns a structure through first
   /// pointer argument.
   bool hasStructRetAttr() const {
     // Be friendly and also check the callee.
     return paramHasAttr(1, Attributes::StructRet);
   }
 
-  /// @brief Determine if any call argument is an aggregate passed by value.
+  /// \brief Determine if any call argument is an aggregate passed by value.
   bool hasByValArgument() const {
     for (unsigned I = 0, E = AttributeList.getNumAttrs(); I != E; ++I)
       if (AttributeList.getAttributesAtIndex(I).hasAttribute(Attributes::ByVal))
@@ -3268,14 +3267,14 @@
 //                                 TruncInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a truncation of integer types.
+/// \brief This class represents a truncation of integer types.
 class TruncInst : public CastInst {
 protected:
-  /// @brief Clone an identical TruncInst
+  /// \brief Clone an identical TruncInst
   virtual TruncInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   TruncInst(
     Value *S,                     ///< The value to be truncated
     Type *Ty,               ///< The (smaller) type to truncate to
@@ -3283,7 +3282,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   TruncInst(
     Value *S,                     ///< The value to be truncated
     Type *Ty,               ///< The (smaller) type to truncate to
@@ -3291,7 +3290,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == Trunc;
   }
@@ -3304,14 +3303,14 @@
 //                                 ZExtInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents zero extension of integer types.
+/// \brief This class represents zero extension of integer types.
 class ZExtInst : public CastInst {
 protected:
-  /// @brief Clone an identical ZExtInst
+  /// \brief Clone an identical ZExtInst
   virtual ZExtInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   ZExtInst(
     Value *S,                     ///< The value to be zero extended
     Type *Ty,               ///< The type to zero extend to
@@ -3319,7 +3318,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end semantics.
+  /// \brief Constructor with insert-at-end semantics.
   ZExtInst(
     Value *S,                     ///< The value to be zero extended
     Type *Ty,               ///< The type to zero extend to
@@ -3327,7 +3326,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == ZExt;
   }
@@ -3340,14 +3339,14 @@
 //                                 SExtInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a sign extension of integer types.
+/// \brief This class represents a sign extension of integer types.
 class SExtInst : public CastInst {
 protected:
-  /// @brief Clone an identical SExtInst
+  /// \brief Clone an identical SExtInst
   virtual SExtInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   SExtInst(
     Value *S,                     ///< The value to be sign extended
     Type *Ty,               ///< The type to sign extend to
@@ -3355,7 +3354,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   SExtInst(
     Value *S,                     ///< The value to be sign extended
     Type *Ty,               ///< The type to sign extend to
@@ -3363,7 +3362,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == SExt;
   }
@@ -3376,14 +3375,14 @@
 //                                 FPTruncInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a truncation of floating point types.
+/// \brief This class represents a truncation of floating point types.
 class FPTruncInst : public CastInst {
 protected:
-  /// @brief Clone an identical FPTruncInst
+  /// \brief Clone an identical FPTruncInst
   virtual FPTruncInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   FPTruncInst(
     Value *S,                     ///< The value to be truncated
     Type *Ty,               ///< The type to truncate to
@@ -3391,7 +3390,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   FPTruncInst(
     Value *S,                     ///< The value to be truncated
     Type *Ty,               ///< The type to truncate to
@@ -3399,7 +3398,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == FPTrunc;
   }
@@ -3412,14 +3411,14 @@
 //                                 FPExtInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents an extension of floating point types.
+/// \brief This class represents an extension of floating point types.
 class FPExtInst : public CastInst {
 protected:
-  /// @brief Clone an identical FPExtInst
+  /// \brief Clone an identical FPExtInst
   virtual FPExtInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   FPExtInst(
     Value *S,                     ///< The value to be extended
     Type *Ty,               ///< The type to extend to
@@ -3427,7 +3426,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   FPExtInst(
     Value *S,                     ///< The value to be extended
     Type *Ty,               ///< The type to extend to
@@ -3435,7 +3434,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == FPExt;
   }
@@ -3448,14 +3447,14 @@
 //                                 UIToFPInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast unsigned integer to floating point.
+/// \brief This class represents a cast unsigned integer to floating point.
 class UIToFPInst : public CastInst {
 protected:
-  /// @brief Clone an identical UIToFPInst
+  /// \brief Clone an identical UIToFPInst
   virtual UIToFPInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   UIToFPInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3463,7 +3462,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   UIToFPInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3471,7 +3470,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == UIToFP;
   }
@@ -3484,14 +3483,14 @@
 //                                 SIToFPInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast from signed integer to floating point.
+/// \brief This class represents a cast from signed integer to floating point.
 class SIToFPInst : public CastInst {
 protected:
-  /// @brief Clone an identical SIToFPInst
+  /// \brief Clone an identical SIToFPInst
   virtual SIToFPInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   SIToFPInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3499,7 +3498,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   SIToFPInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3507,7 +3506,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == SIToFP;
   }
@@ -3520,14 +3519,14 @@
 //                                 FPToUIInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast from floating point to unsigned integer
+/// \brief This class represents a cast from floating point to unsigned integer
 class FPToUIInst  : public CastInst {
 protected:
-  /// @brief Clone an identical FPToUIInst
+  /// \brief Clone an identical FPToUIInst
   virtual FPToUIInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   FPToUIInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3535,7 +3534,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   FPToUIInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3543,7 +3542,7 @@
     BasicBlock *InsertAtEnd       ///< Where to insert the new instruction
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == FPToUI;
   }
@@ -3556,14 +3555,14 @@
 //                                 FPToSIInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast from floating point to signed integer.
+/// \brief This class represents a cast from floating point to signed integer.
 class FPToSIInst  : public CastInst {
 protected:
-  /// @brief Clone an identical FPToSIInst
+  /// \brief Clone an identical FPToSIInst
   virtual FPToSIInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   FPToSIInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3571,7 +3570,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   FPToSIInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3579,7 +3578,7 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Methods for support type inquiry through isa, cast, and dyn_cast:
+  /// \brief Methods for support type inquiry through isa, cast, and dyn_cast:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == FPToSI;
   }
@@ -3592,10 +3591,10 @@
 //                                 IntToPtrInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast from an integer to a pointer.
+/// \brief This class represents a cast from an integer to a pointer.
 class IntToPtrInst : public CastInst {
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   IntToPtrInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3603,7 +3602,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   IntToPtrInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3611,20 +3610,12 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief Clone an identical IntToPtrInst
+  /// \brief Clone an identical IntToPtrInst
   virtual IntToPtrInst *clone_impl() const;
 
-  /// @brief return the address space of the pointer.
+  /// \brief Returns the address space of this instruction's pointer type.
   unsigned getAddressSpace() const {
-    if (getType()->isPointerTy()) 
-      return cast<PointerType>(getType())->getAddressSpace();
-    if (getType()->isVectorTy() &&
-        cast<VectorType>(getType())->getElementType()->isPointerTy())
-      return cast<PointerType>(
-          cast<VectorType>(getType())->getElementType())
-        ->getAddressSpace();
-    llvm_unreachable("Must be a pointer or a vector of pointers.");
-    return 0;
+    return getType()->getPointerAddressSpace();
   }
 
   // Methods for support type inquiry through isa, cast, and dyn_cast:
@@ -3640,14 +3631,14 @@
 //                                 PtrToIntInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a cast from a pointer to an integer
+/// \brief This class represents a cast from a pointer to an integer
 class PtrToIntInst : public CastInst {
 protected:
-  /// @brief Clone an identical PtrToIntInst
+  /// \brief Clone an identical PtrToIntInst
   virtual PtrToIntInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   PtrToIntInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3655,7 +3646,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   PtrToIntInst(
     Value *S,                     ///< The value to be converted
     Type *Ty,               ///< The type to convert to
@@ -3663,18 +3654,16 @@
     BasicBlock *InsertAtEnd       ///< The block to insert the instruction into
   );
 
-  /// @brief return the address space of the pointer.
+  /// \brief Gets the pointer operand.
+  Value *getPointerOperand() { return getOperand(0); }
+  /// \brief Gets the pointer operand.
+  const Value *getPointerOperand() const { return getOperand(0); }
+  /// \brief Gets the operand index of the pointer operand.
+  static unsigned getPointerOperandIndex() { return 0U; }
+
+  /// \brief Returns the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
-    Type *Ty = getOperand(0)->getType();
-    if (Ty->isPointerTy())
-      return cast<PointerType>(Ty)->getAddressSpace();
-    if (Ty->isVectorTy()
-        && cast<VectorType>(Ty)->getElementType()->isPointerTy())
-      return cast<PointerType>(
-          cast<VectorType>(Ty)->getElementType())
-        ->getAddressSpace();
-    llvm_unreachable("Must be a pointer or a vector of pointers.");
-    return 0;
+    return getPointerOperand()->getType()->getPointerAddressSpace();
   }
 
   // Methods for support type inquiry through isa, cast, and dyn_cast:
@@ -3690,14 +3679,14 @@
 //                             BitCastInst Class
 //===----------------------------------------------------------------------===//
 
-/// @brief This class represents a no-op cast from one type to another.
+/// \brief This class represents a no-op cast from one type to another.
 class BitCastInst : public CastInst {
 protected:
-  /// @brief Clone an identical BitCastInst
+  /// \brief Clone an identical BitCastInst
   virtual BitCastInst *clone_impl() const;
 
 public:
-  /// @brief Constructor with insert-before-instruction semantics
+  /// \brief Constructor with insert-before-instruction semantics
   BitCastInst(
     Value *S,                     ///< The value to be casted
     Type *Ty,               ///< The type to casted to
@@ -3705,7 +3694,7 @@
     Instruction *InsertBefore = 0 ///< Where to insert the new instruction
   );
 
-  /// @brief Constructor with insert-at-end-of-block semantics
+  /// \brief Constructor with insert-at-end-of-block semantics
   BitCastInst(
     Value *S,                     ///< The value to be casted
     Type *Ty,               ///< The type to casted to

Modified: llvm/branches/R600/include/llvm/IntrinsicInst.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/IntrinsicInst.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/IntrinsicInst.h (original)
+++ llvm/branches/R600/include/llvm/IntrinsicInst.h Tue Nov 13 09:21:47 2012
@@ -268,6 +268,49 @@
     }
   };
 
+  /// VAStartInst - This represents the llvm.va_start intrinsic.
+  ///
+  class VAStartInst : public IntrinsicInst {
+  public:
+    static inline bool classof(const IntrinsicInst *I) {
+      return I->getIntrinsicID() == Intrinsic::vastart;
+    }
+    static inline bool classof(const Value *V) {
+      return isa<IntrinsicInst>(V) && classof(cast<IntrinsicInst>(V));
+    }
+
+    Value *getArgList() const { return const_cast<Value*>(getArgOperand(0)); }
+  };
+
+  /// VAEndInst - This represents the llvm.va_end intrinsic.
+  ///
+  class VAEndInst : public IntrinsicInst {
+  public:
+    static inline bool classof(const IntrinsicInst *I) {
+      return I->getIntrinsicID() == Intrinsic::vaend;
+    }
+    static inline bool classof(const Value *V) {
+      return isa<IntrinsicInst>(V) && classof(cast<IntrinsicInst>(V));
+    }
+
+    Value *getArgList() const { return const_cast<Value*>(getArgOperand(0)); }
+  };
+
+  /// VACopyInst - This represents the llvm.va_copy intrinsic.
+  ///
+  class VACopyInst : public IntrinsicInst {
+  public:
+    static inline bool classof(const IntrinsicInst *I) {
+      return I->getIntrinsicID() == Intrinsic::vacopy;
+    }
+    static inline bool classof(const Value *V) {
+      return isa<IntrinsicInst>(V) && classof(cast<IntrinsicInst>(V));
+    }
+
+    Value *getDest() const { return const_cast<Value*>(getArgOperand(0)); }
+    Value *getSrc() const { return const_cast<Value*>(getArgOperand(1)); }
+  };
+
 }
 
 #endif

Modified: llvm/branches/R600/include/llvm/LinkAllPasses.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/LinkAllPasses.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/LinkAllPasses.h (original)
+++ llvm/branches/R600/include/llvm/LinkAllPasses.h Tue Nov 13 09:21:47 2012
@@ -60,6 +60,7 @@
       (void) llvm::createCFGSimplificationPass();
       (void) llvm::createConstantMergePass();
       (void) llvm::createConstantPropagationPass();
+      (void) llvm::createCostModelAnalysisPass();
       (void) llvm::createDeadArgEliminationPass();
       (void) llvm::createDeadCodeEliminationPass();
       (void) llvm::createDeadInstEliminationPass();
@@ -82,11 +83,10 @@
       (void) llvm::createIPSCCPPass();
       (void) llvm::createIndVarSimplifyPass();
       (void) llvm::createInstructionCombiningPass();
-      (void) llvm::createInternalizePass(false);
+      (void) llvm::createInternalizePass();
       (void) llvm::createLCSSAPass();
       (void) llvm::createLICMPass();
       (void) llvm::createLazyValueInfoPass();
-      (void) llvm::createLoopDependenceAnalysisPass();
       (void) llvm::createLoopExtractorPass();
       (void) llvm::createLoopSimplifyPass();
       (void) llvm::createLoopStrengthReducePass();

Modified: llvm/branches/R600/include/llvm/MC/MCELFObjectWriter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCELFObjectWriter.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCELFObjectWriter.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCELFObjectWriter.h Tue Nov 13 09:21:47 2012
@@ -85,6 +85,9 @@
                                          const MCFragment &F,
                                          const MCFixup &Fixup,
                                          bool IsPCRel) const;
+  virtual const MCSymbol *undefinedExplicitRelSym(const MCValue &Target,
+                                                  const MCFixup &Fixup,
+                                                  bool IsPCRel) const;
   virtual void adjustFixupOffset(const MCFixup &Fixup,
                                  uint64_t &RelocOffset);
 
@@ -93,9 +96,9 @@
 
   /// @name Accessors
   /// @{
-  uint8_t getOSABI() { return OSABI; }
-  uint16_t getEMachine() { return EMachine; }
-  bool hasRelocationAddend() { return HasRelocationAddend; }
+  uint8_t getOSABI() const { return OSABI; }
+  uint16_t getEMachine() const { return EMachine; }
+  bool hasRelocationAddend() const { return HasRelocationAddend; }
   bool is64Bit() const { return Is64Bit; }
   bool isN64() const { return IsN64; }
   /// @}

Modified: llvm/branches/R600/include/llvm/MC/MCInstPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCInstPrinter.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCInstPrinter.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCInstPrinter.h Tue Nov 13 09:21:47 2012
@@ -33,12 +33,16 @@
   /// The current set of available features.
   unsigned AvailableFeatures;
 
+  /// True if we are printing marked up assembly.
+  bool UseMarkup;
+
   /// Utility function for printing annotations.
   void printAnnotation(raw_ostream &OS, StringRef Annot);
 public:
   MCInstPrinter(const MCAsmInfo &mai, const MCInstrInfo &mii,
                 const MCRegisterInfo &mri)
-    : CommentStream(0), MAI(mai), MII(mii), MRI(mri), AvailableFeatures(0) {}
+    : CommentStream(0), MAI(mai), MII(mii), MRI(mri), AvailableFeatures(0),
+      UseMarkup(0) {}
 
   virtual ~MCInstPrinter();
 
@@ -59,6 +63,13 @@
 
   unsigned getAvailableFeatures() const { return AvailableFeatures; }
   void setAvailableFeatures(unsigned Value) { AvailableFeatures = Value; }
+
+  bool getUseMarkup() const { return UseMarkup; }
+  void setUseMarkup(bool Value) { UseMarkup = Value; }
+
+  /// Utility functions to make adding mark ups simpler.
+  StringRef markup(StringRef s) const;
+  StringRef markup(StringRef a, StringRef b) const;
 };
 
 } // namespace llvm

Modified: llvm/branches/R600/include/llvm/MC/MCParser/MCAsmParser.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCParser/MCAsmParser.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCParser/MCAsmParser.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCParser/MCAsmParser.h Tue Nov 13 09:21:47 2012
@@ -37,6 +37,8 @@
   virtual ~MCAsmParserSemaCallback(); 
   virtual void *LookupInlineAsmIdentifier(StringRef Name, void *Loc,
                                           unsigned &Size) = 0;
+  virtual bool LookupInlineAsmField(StringRef Base, StringRef Member,
+                                    unsigned &Offset) = 0;
 };
 
 /// MCAsmParser - Generic assembler parser interface, for use by target specific
@@ -90,16 +92,13 @@
   /// ParseMSInlineAsm - Parse ms-style inline assembly.
   virtual bool ParseMSInlineAsm(void *AsmLoc, std::string &AsmString,
                                 unsigned &NumOutputs, unsigned &NumInputs,
-                                SmallVectorImpl<void *> &OpDecls,
+                                SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
                                 SmallVectorImpl<std::string> &Constraints,
                                 SmallVectorImpl<std::string> &Clobbers,
                                 const MCInstrInfo *MII,
                                 const MCInstPrinter *IP,
                                 MCAsmParserSemaCallback &SI) = 0;
 
-  /// ParseStatement - Parse the next statement.
-  virtual bool ParseStatement() = 0;
-
   /// Warning - Emit a warning at the location \p L, with the message \p Msg.
   ///
   /// \return The return value is true, if warnings are fatal.

Modified: llvm/branches/R600/include/llvm/MC/MCParser/MCParsedAsmOperand.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCParser/MCParsedAsmOperand.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCParser/MCParsedAsmOperand.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCParser/MCParsedAsmOperand.h Tue Nov 13 09:21:47 2012
@@ -64,6 +64,20 @@
   /// getEndLoc - Get the location of the last token of this operand.
   virtual SMLoc getEndLoc() const = 0;
 
+  /// needAsmRewrite - AsmRewrites happen in both the target-independent and
+  /// target-dependent parsers.  The target-independent parser calls this
+  /// function to determine if the target-dependent parser has already taken
+  /// care of the rewrites.  Only valid when parsing MS-style inline assembly.
+  virtual bool needAsmRewrite() const { return true; }
+
+  /// isOffsetOf - Do we need to emit code to get the offset of the variable,
+  /// rather then the value of the variable?   Only valid when parsing MS-style
+  /// inline assembly.
+  virtual bool isOffsetOf() const { return false; }
+
+  /// getOffsetOfLoc - Get the location of the offset operator.
+  virtual SMLoc getOffsetOfLoc() const { return SMLoc(); }
+
   /// needSizeDirective - Do we need to emit a sizing directive for this
   /// operand?  Only valid when parsing MS-style inline assembly.
   virtual bool needSizeDirective() const { return false; }

Modified: llvm/branches/R600/include/llvm/MC/MCRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCRegisterInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCRegisterInfo.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCRegisterInfo.h Tue Nov 13 09:21:47 2012
@@ -370,7 +370,7 @@
 
   /// getRegClass - Returns the register class associated with the enumeration
   /// value.  See class MCOperandInfo.
-  const MCRegisterClass getRegClass(unsigned i) const {
+  const MCRegisterClass& getRegClass(unsigned i) const {
     assert(i < getNumRegClasses() && "Register Class ID out of range");
     return Classes[i];
   }

Modified: llvm/branches/R600/include/llvm/MC/MCSchedule.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCSchedule.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCSchedule.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCSchedule.h Tue Nov 13 09:21:47 2012
@@ -219,6 +219,10 @@
   /// Does this machine model include instruction-level scheduling.
   bool hasInstrSchedModel() const { return SchedClassTable; }
 
+  unsigned getNumProcResourceKinds() const {
+    return NumProcResourceKinds;
+  }
+
   const MCProcResourceDesc *getProcResource(unsigned ProcResourceIdx) const {
     assert(hasInstrSchedModel() && "No scheduling machine model");
 

Modified: llvm/branches/R600/include/llvm/MC/MCTargetAsmParser.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/MC/MCTargetAsmParser.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/MC/MCTargetAsmParser.h (original)
+++ llvm/branches/R600/include/llvm/MC/MCTargetAsmParser.h Tue Nov 13 09:21:47 2012
@@ -21,6 +21,39 @@
 class MCInst;
 template <typename T> class SmallVectorImpl;
 
+enum AsmRewriteKind {
+  AOK_DotOperator,    // Rewrite a dot operator expression as an immediate.
+                      // E.g., [eax].foo.bar -> [eax].8
+  AOK_Emit,           // Rewrite _emit as .byte.
+  AOK_Imm,            // Rewrite as $$N.
+  AOK_ImmPrefix,      // Add $$ before a parsed Imm.
+  AOK_Input,          // Rewrite in terms of $N.
+  AOK_Output,         // Rewrite in terms of $N.
+  AOK_SizeDirective,  // Add a sizing directive (e.g., dword ptr).
+  AOK_Skip            // Skip emission (e.g., offset/type operators).
+};
+
+struct AsmRewrite {
+  AsmRewriteKind Kind;
+  SMLoc Loc;
+  unsigned Len;
+  unsigned Val;
+public:
+  AsmRewrite(AsmRewriteKind kind, SMLoc loc, unsigned len = 0, unsigned val = 0)
+    : Kind(kind), Loc(loc), Len(len), Val(val) {}
+};
+
+struct ParseInstructionInfo {
+
+  SmallVectorImpl<AsmRewrite> *AsmRewrites;
+
+  ParseInstructionInfo() : AsmRewrites(0) {}
+  ParseInstructionInfo(SmallVectorImpl<AsmRewrite> *rewrites)
+    : AsmRewrites(rewrites) {}
+
+  ~ParseInstructionInfo() {}
+};
+
 /// MCTargetAsmParser - Generic interface to target specific assembly parsers.
 class MCTargetAsmParser : public MCAsmParserExtension {
 public:
@@ -77,7 +110,8 @@
   /// \param Operands [out] - The list of parsed operands, this returns
   ///        ownership of them to the caller.
   /// \return True on failure.
-  virtual bool ParseInstruction(StringRef Name, SMLoc NameLoc,
+  virtual bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
+                                SMLoc NameLoc,
                             SmallVectorImpl<MCParsedAsmOperand*> &Operands) = 0;
 
   /// ParseDirective - Parse a target specific assembler directive

Modified: llvm/branches/R600/include/llvm/Object/COFF.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Object/COFF.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Object/COFF.h (original)
+++ llvm/branches/R600/include/llvm/Object/COFF.h Tue Nov 13 09:21:47 2012
@@ -116,6 +116,7 @@
   virtual error_code getSymbolType(DataRefImpl Symb, SymbolRef::Type &Res) const;
   virtual error_code getSymbolSection(DataRefImpl Symb,
                                       section_iterator &Res) const;
+  virtual error_code getSymbolValue(DataRefImpl Symb, uint64_t &Val) const;
 
   virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const;
   virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const;

Modified: llvm/branches/R600/include/llvm/Object/ELF.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Object/ELF.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Object/ELF.h (original)
+++ llvm/branches/R600/include/llvm/Object/ELF.h Tue Nov 13 09:21:47 2012
@@ -620,6 +620,7 @@
   virtual error_code getSymbolType(DataRefImpl Symb, SymbolRef::Type &Res) const;
   virtual error_code getSymbolSection(DataRefImpl Symb,
                                       section_iterator &Res) const;
+  virtual error_code getSymbolValue(DataRefImpl Symb, uint64_t &Val) const;
 
   friend class DynRefImpl<target_endianness, is64Bits>;
   virtual error_code getDynNext(DataRefImpl DynData, DynRef &Result) const;
@@ -1162,6 +1163,16 @@
 
 template<support::endianness target_endianness, bool is64Bits>
 error_code ELFObjectFile<target_endianness, is64Bits>
+                        ::getSymbolValue(DataRefImpl Symb,
+                                         uint64_t &Val) const {
+  validateSymbol(Symb);
+  const Elf_Sym *symb = getSymbol(Symb);
+  Val = symb->st_value;
+  return object_error::success;
+}
+
+template<support::endianness target_endianness, bool is64Bits>
+error_code ELFObjectFile<target_endianness, is64Bits>
                         ::getSectionNext(DataRefImpl Sec, SectionRef &Result) const {
   const uint8_t *sec = reinterpret_cast<const uint8_t *>(Sec.p);
   sec += Header->e_shentsize;

Modified: llvm/branches/R600/include/llvm/Object/MachO.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Object/MachO.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Object/MachO.h (original)
+++ llvm/branches/R600/include/llvm/Object/MachO.h Tue Nov 13 09:21:47 2012
@@ -61,6 +61,7 @@
   virtual error_code getSymbolType(DataRefImpl Symb, SymbolRef::Type &Res) const;
   virtual error_code getSymbolSection(DataRefImpl Symb,
                                       section_iterator &Res) const;
+  virtual error_code getSymbolValue(DataRefImpl Symb, uint64_t &Val) const;
 
   virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const;
   virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const;

Modified: llvm/branches/R600/include/llvm/Object/ObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Object/ObjectFile.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Object/ObjectFile.h (original)
+++ llvm/branches/R600/include/llvm/Object/ObjectFile.h Tue Nov 13 09:21:47 2012
@@ -234,6 +234,9 @@
   /// end_sections() if it is undefined or is an absolute symbol.
   error_code getSection(section_iterator &Result) const;
 
+  /// @brief Get value of the symbol in the symbol table.
+  error_code getValue(uint64_t &Val) const;
+
   DataRefImpl getRawDataRefImpl() const;
 };
 typedef content_iterator<SymbolRef> symbol_iterator;
@@ -300,6 +303,7 @@
                                     uint32_t &Res) const = 0;
   virtual error_code getSymbolSection(DataRefImpl Symb,
                                       section_iterator &Res) const = 0;
+  virtual error_code getSymbolValue(DataRefImpl Symb, uint64_t &Val) const = 0;
 
   // Same as above for SectionRef.
   friend class SectionRef;
@@ -444,6 +448,10 @@
   return OwningObject->getSymbolType(SymbolPimpl, Result);
 }
 
+inline error_code SymbolRef::getValue(uint64_t &Val) const {
+  return OwningObject->getSymbolValue(SymbolPimpl, Val);
+}
+
 inline DataRefImpl SymbolRef::getRawDataRefImpl() const {
   return SymbolPimpl;
 }

Modified: llvm/branches/R600/include/llvm/Operator.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Operator.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Operator.h (original)
+++ llvm/branches/R600/include/llvm/Operator.h Tue Nov 13 09:21:47 2012
@@ -36,8 +36,11 @@
   void *operator new(size_t, unsigned) LLVM_DELETED_FUNCTION;
   void *operator new(size_t s) LLVM_DELETED_FUNCTION;
   Operator() LLVM_DELETED_FUNCTION;
-  // NOTE: cannot use LLVM_DELETED_FUNCTION because it's not legal to delete
-  // an overridden method that's not deleted in the base class.
+
+protected:
+  // NOTE: Cannot use LLVM_DELETED_FUNCTION because it's not legal to delete
+  // an overridden method that's not deleted in the base class. Cannot leave
+  // this unimplemented because that leads to an ODR-violation.
   ~Operator();
 
 public:
@@ -79,8 +82,6 @@
   };
 
 private:
-  ~OverflowingBinaryOperator(); // DO NOT IMPLEMENT
-
   friend class BinaryOperator;
   friend class ConstantExpr;
   void setHasNoUnsignedWrap(bool B) {
@@ -132,8 +133,6 @@
   };
   
 private:
-  ~PossiblyExactOperator(); // DO NOT IMPLEMENT
-
   friend class BinaryOperator;
   friend class ConstantExpr;
   void setIsExact(bool B) {
@@ -168,9 +167,6 @@
 /// FPMathOperator - Utility class for floating point operations which can have
 /// information about relaxed accuracy requirements attached to them.
 class FPMathOperator : public Operator {
-private:
-  ~FPMathOperator(); // DO NOT IMPLEMENT
-
 public:
 
   /// \brief Get the maximum error permitted by this operation in ULPs.  An
@@ -191,7 +187,6 @@
 /// opcodes.
 template<typename SuperClass, unsigned Opc>
 class ConcreteOperator : public SuperClass {
-  ~ConcreteOperator(); // DO NOT IMPLEMENT
 public:
   static inline bool classof(const Instruction *I) {
     return I->getOpcode() == Opc;
@@ -207,45 +202,35 @@
 
 class AddOperator
   : public ConcreteOperator<OverflowingBinaryOperator, Instruction::Add> {
-  ~AddOperator(); // DO NOT IMPLEMENT
 };
 class SubOperator
   : public ConcreteOperator<OverflowingBinaryOperator, Instruction::Sub> {
-  ~SubOperator(); // DO NOT IMPLEMENT
 };
 class MulOperator
   : public ConcreteOperator<OverflowingBinaryOperator, Instruction::Mul> {
-  ~MulOperator(); // DO NOT IMPLEMENT
 };
 class ShlOperator
   : public ConcreteOperator<OverflowingBinaryOperator, Instruction::Shl> {
-  ~ShlOperator(); // DO NOT IMPLEMENT
 };
 
 
 class SDivOperator
   : public ConcreteOperator<PossiblyExactOperator, Instruction::SDiv> {
-  ~SDivOperator(); // DO NOT IMPLEMENT
 };
 class UDivOperator
   : public ConcreteOperator<PossiblyExactOperator, Instruction::UDiv> {
-  ~UDivOperator(); // DO NOT IMPLEMENT
 };
 class AShrOperator
   : public ConcreteOperator<PossiblyExactOperator, Instruction::AShr> {
-  ~AShrOperator(); // DO NOT IMPLEMENT
 };
 class LShrOperator
   : public ConcreteOperator<PossiblyExactOperator, Instruction::LShr> {
-  ~LShrOperator(); // DO NOT IMPLEMENT
 };
 
 
 
 class GEPOperator
   : public ConcreteOperator<Operator, Instruction::GetElementPtr> {
-  ~GEPOperator(); // DO NOT IMPLEMENT
-
   enum {
     IsInBounds = (1 << 0)
   };

Modified: llvm/branches/R600/include/llvm/Support/AlignOf.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Support/AlignOf.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Support/AlignOf.h (original)
+++ llvm/branches/R600/include/llvm/Support/AlignOf.h Tue Nov 13 09:21:47 2012
@@ -78,7 +78,7 @@
   template <> struct AlignedCharArrayImpl<x> { \
     char alignas(x) aligned; \
   }
-#elif defined(__GNUC__)
+#elif defined(__GNUC__) || defined(__IBM_ATTRIBUTES)
 #define LLVM_ALIGNEDCHARARRAY_TEMPLATE_ALIGNMENT(x) \
   template <> struct AlignedCharArrayImpl<x> { \
     char aligned __attribute__((aligned(x))); \

Modified: llvm/branches/R600/include/llvm/Support/CommandLine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Support/CommandLine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Support/CommandLine.h (original)
+++ llvm/branches/R600/include/llvm/Support/CommandLine.h Tue Nov 13 09:21:47 2012
@@ -1507,7 +1507,7 @@
       typename ParserClass::parser_data_type();
     if (Parser.parse(*this, ArgName, Arg, Val))
       return true;  // Parse Error!
-    addValue(Val);
+    this->addValue(Val);
     setPosition(pos);
     Positions.push_back(pos);
     return false;

Modified: llvm/branches/R600/include/llvm/Support/ELF.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Support/ELF.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Support/ELF.h (original)
+++ llvm/branches/R600/include/llvm/Support/ELF.h Tue Nov 13 09:21:47 2012
@@ -689,8 +689,36 @@
   R_MIPS_NUM               = 218
 };
 
+// Hexagon Specific e_flags
+// Release 5 ABI
+enum {
+  // Object processor version flags, bits[3:0]
+  EF_HEXAGON_MACH_V2      = 0x00000001,   // Hexagon V2
+  EF_HEXAGON_MACH_V3      = 0x00000002,   // Hexagon V3
+  EF_HEXAGON_MACH_V4      = 0x00000003,   // Hexagon V4
+  EF_HEXAGON_MACH_V5      = 0x00000004,   // Hexagon V5
+
+  // Highest ISA version flags
+  EF_HEXAGON_ISA_MACH     = 0x00000000,   // Same as specified in bits[3:0]
+                                          // of e_flags
+  EF_HEXAGON_ISA_V2       = 0x00000010,   // Hexagon V2 ISA
+  EF_HEXAGON_ISA_V3       = 0x00000020,   // Hexagon V3 ISA
+  EF_HEXAGON_ISA_V4       = 0x00000030,   // Hexagon V4 ISA
+  EF_HEXAGON_ISA_V5       = 0x00000040    // Hexagon V5 ISA
+};
+
+// Hexagon specific Section indexes for common small data
+// Release 5 ABI 
+enum {
+  SHN_HEXAGON_SCOMMON     = 0xff00,       // Other access sizes
+  SHN_HEXAGON_SCOMMON_1   = 0xff01,       // Byte-sized access
+  SHN_HEXAGON_SCOMMON_2   = 0xff02,       // Half-word-sized access
+  SHN_HEXAGON_SCOMMON_4   = 0xff03,       // Word-sized access
+  SHN_HEXAGON_SCOMMON_8   = 0xff04        // Double-word-size access
+};   
+
 // ELF Relocation types for Hexagon
-// Release 5 ABI - Document: 80-V9418-3 Rev. J
+// Release 5 ABI
 enum {
   R_HEX_NONE              =  0,
   R_HEX_B22_PCREL         =  1,
@@ -1118,6 +1146,9 @@
   PT_PHDR    = 6, // The program header table itself.
   PT_TLS     = 7, // The thread-local storage template.
   PT_LOOS    = 0x60000000, // Lowest operating system-specific pt entry type.
+  PT_HIOS    = 0x6fffffff, // Highest operating system-specific pt entry type.
+  PT_LOPROC  = 0x70000000, // Lowest processor-specific program hdr entry type.
+  PT_HIPROC  = 0x7fffffff, // Highest processor-specific program hdr entry type.
 
   // x86-64 program header types.
   // These all contain stack unwind tables.
@@ -1128,9 +1159,11 @@
   PT_GNU_STACK  = 0x6474e551, // Indicates stack executability.
   PT_GNU_RELRO  = 0x6474e552, // Read-only after relocation.
 
-  PT_HIOS    = 0x6fffffff, // Highest operating system-specific pt entry type.
-  PT_LOPROC  = 0x70000000, // Lowest processor-specific program hdr entry type.
-  PT_HIPROC  = 0x7fffffff  // Highest processor-specific program hdr entry type.
+  // ARM program header types.
+  PT_ARM_ARCHEXT = 0x70000000, // Platform architecture compatibility information
+  // These all contain stack unwind tables.
+  PT_ARM_EXIDX   = 0x70000001,
+  PT_ARM_UNWIND  = 0x70000001
 };
 
 // Segment flag bits.

Modified: llvm/branches/R600/include/llvm/Support/InstVisitor.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Support/InstVisitor.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Support/InstVisitor.h (original)
+++ llvm/branches/R600/include/llvm/Support/InstVisitor.h Tue Nov 13 09:21:47 2012
@@ -209,6 +209,9 @@
   RetTy visitMemMoveInst(MemMoveInst &I)          { DELEGATE(MemTransferInst); }
   RetTy visitMemTransferInst(MemTransferInst &I)  { DELEGATE(MemIntrinsic); }
   RetTy visitMemIntrinsic(MemIntrinsic &I)        { DELEGATE(IntrinsicInst); }
+  RetTy visitVAStartInst(VAStartInst &I)          { DELEGATE(IntrinsicInst); }
+  RetTy visitVAEndInst(VAEndInst &I)              { DELEGATE(IntrinsicInst); }
+  RetTy visitVACopyInst(VACopyInst &I)            { DELEGATE(IntrinsicInst); }
   RetTy visitIntrinsicInst(IntrinsicInst &I)      { DELEGATE(CallInst); }
 
   // Call and Invoke are slightly different as they delegate first through
@@ -262,6 +265,9 @@
       case Intrinsic::memcpy:      DELEGATE(MemCpyInst);
       case Intrinsic::memmove:     DELEGATE(MemMoveInst);
       case Intrinsic::memset:      DELEGATE(MemSetInst);
+      case Intrinsic::vastart:     DELEGATE(VAStartInst);
+      case Intrinsic::vaend:       DELEGATE(VAEndInst);
+      case Intrinsic::vacopy:      DELEGATE(VACopyInst);
       case Intrinsic::not_intrinsic: break;
       }
     }

Modified: llvm/branches/R600/include/llvm/Support/IntegersSubset.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Support/IntegersSubset.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Support/IntegersSubset.h (original)
+++ llvm/branches/R600/include/llvm/Support/IntegersSubset.h Tue Nov 13 09:21:47 2012
@@ -411,8 +411,8 @@
   unsigned getSize() const {
     APInt sz(((const APInt&)getItem(0).getLow()).getBitWidth(), 0);
     for (unsigned i = 0, e = getNumItems(); i != e; ++i) {
-      const APInt &Low = getItem(i).getLow();
-      const APInt &High = getItem(i).getHigh();
+      const APInt Low = getItem(i).getLow();
+      const APInt High = getItem(i).getHigh();
       APInt S = High - Low + 1;
       sz += S;
     }
@@ -426,8 +426,8 @@
   APInt getSingleValue(unsigned idx) const {
     APInt sz(((const APInt&)getItem(0).getLow()).getBitWidth(), 0);
     for (unsigned i = 0, e = getNumItems(); i != e; ++i) {
-      const APInt &Low = getItem(i).getLow();
-      const APInt &High = getItem(i).getHigh();
+      const APInt Low = getItem(i).getLow();
+      const APInt High = getItem(i).getHigh();
       APInt S = High - Low + 1;
       APInt oldSz = sz;
       sz += S;

Modified: llvm/branches/R600/include/llvm/TableGen/Error.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/TableGen/Error.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/TableGen/Error.h (original)
+++ llvm/branches/R600/include/llvm/TableGen/Error.h Tue Nov 13 09:21:47 2012
@@ -19,27 +19,17 @@
 
 namespace llvm {
 
-class TGError {
-  SmallVector<SMLoc, 4> Locs;
-  std::string Message;
-public:
-  TGError(ArrayRef<SMLoc> locs, const std::string &message)
-    : Locs(locs.begin(), locs.end()), Message(message) {}
-
-  ArrayRef<SMLoc> getLoc() const { return Locs; }
-  const std::string &getMessage() const { return Message; }
-};
-
 void PrintWarning(ArrayRef<SMLoc> WarningLoc, const Twine &Msg);
 void PrintWarning(const char *Loc, const Twine &Msg);
 void PrintWarning(const Twine &Msg);
-void PrintWarning(const TGError &Warning);
 
 void PrintError(ArrayRef<SMLoc> ErrorLoc, const Twine &Msg);
 void PrintError(const char *Loc, const Twine &Msg);
 void PrintError(const Twine &Msg);
-void PrintError(const TGError &Error);
 
+LLVM_ATTRIBUTE_NORETURN void PrintFatalError(const std::string &Msg);
+LLVM_ATTRIBUTE_NORETURN void PrintFatalError(ArrayRef<SMLoc> ErrorLoc,
+                                             const std::string &Msg);
 
 extern SourceMgr SrcMgr;
 

Modified: llvm/branches/R600/include/llvm/Target/Target.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/Target.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/Target.td (original)
+++ llvm/branches/R600/include/llvm/Target/Target.td Tue Nov 13 09:21:47 2012
@@ -998,6 +998,55 @@
 }
 
 //===----------------------------------------------------------------------===//
+// InstrMapping - This class is used to create mapping tables to relate
+// instructions with each other based on the values specified in RowFields,
+// ColFields, KeyCol and ValueCols.
+//
+class InstrMapping {
+  // FilterClass - Used to limit search space only to the instructions that
+  // define the relationship modeled by this InstrMapping record.
+  string FilterClass;
+
+  // RowFields - List of fields/attributes that should be same for all the
+  // instructions in a row of the relation table. Think of this as a set of
+  // properties shared by all the instructions related by this relationship
+  // model and is used to categorize instructions into subgroups. For instance,
+  // if we want to define a relation that maps 'Add' instruction to its
+  // predicated forms, we can define RowFields like this:
+  //
+  // let RowFields = BaseOp
+  // All add instruction predicated/non-predicated will have to set their BaseOp
+  // to the same value.
+  //
+  // def Add: { let BaseOp = 'ADD'; let predSense = 'nopred' }
+  // def Add_predtrue: { let BaseOp = 'ADD'; let predSense = 'true' }
+  // def Add_predfalse: { let BaseOp = 'ADD'; let predSense = 'false'  }
+  list<string> RowFields = [];
+
+  // List of fields/attributes that are same for all the instructions
+  // in a column of the relation table.
+  // Ex: let ColFields = 'predSense' -- It means that the columns are arranged
+  // based on the 'predSense' values. All the instruction in a specific
+  // column have the same value and it is fixed for the column according
+  // to the values set in 'ValueCols'.
+  list<string> ColFields = [];
+
+  // Values for the fields/attributes listed in 'ColFields'.
+  // Ex: let KeyCol = 'nopred' -- It means that the key instruction (instruction
+  // that models this relation) should be non-predicated.
+  // In the example above, 'Add' is the key instruction.
+  list<string> KeyCol = [];
+
+  // List of values for the fields/attributes listed in 'ColFields', one for
+  // each column in the relation table.
+  //
+  // Ex: let ValueCols = [['true'],['false']] -- It adds two columns in the
+  // table. First column requires all the instructions to have predSense
+  // set to 'true' and second column requires it to be 'false'.
+  list<list<string> > ValueCols = [];
+}
+
+//===----------------------------------------------------------------------===//
 // Pull in the common support for calling conventions.
 //
 include "llvm/Target/TargetCallingConv.td"

Modified: llvm/branches/R600/include/llvm/Target/TargetCallingConv.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetCallingConv.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetCallingConv.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetCallingConv.h Tue Nov 13 09:21:47 2012
@@ -140,9 +140,19 @@
     /// IsFixed - Is this a "fixed" value, ie not passed through a vararg "...".
     bool IsFixed;
 
+    /// Index original Function's argument.
+    unsigned OrigArgIndex;
+
+    /// Offset in bytes of current output value relative to the beginning of
+    /// original argument. E.g. if argument was splitted into four 32 bit
+    /// registers, we got 4 OutputArgs with PartOffsets 0, 4, 8 and 12.
+    unsigned PartOffset;
+
     OutputArg() : IsFixed(false) {}
-    OutputArg(ArgFlagsTy flags, EVT vt, bool isfixed)
-      : Flags(flags), IsFixed(isfixed) {
+    OutputArg(ArgFlagsTy flags, EVT vt, bool isfixed,
+              unsigned origIdx, unsigned partOffs)
+      : Flags(flags), IsFixed(isfixed), OrigArgIndex(origIdx),
+        PartOffset(partOffs) {
       VT = vt.getSimpleVT();
     }
   };

Removed: llvm/branches/R600/include/llvm/Target/TargetELFWriterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetELFWriterInfo.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetELFWriterInfo.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetELFWriterInfo.h (removed)
@@ -1,121 +0,0 @@
-//===-- llvm/Target/TargetELFWriterInfo.h - ELF Writer Info -----*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file defines the TargetELFWriterInfo class.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_TARGET_TARGETELFWRITERINFO_H
-#define LLVM_TARGET_TARGETELFWRITERINFO_H
-
-namespace llvm {
-
-  //===--------------------------------------------------------------------===//
-  //                          TargetELFWriterInfo
-  //===--------------------------------------------------------------------===//
-
-  class TargetELFWriterInfo {
-  protected:
-    // EMachine - This field is the target specific value to emit as the
-    // e_machine member of the ELF header.
-    unsigned short EMachine;
-    bool is64Bit, isLittleEndian;
-  public:
-
-    // Machine architectures
-    enum MachineType {
-      EM_NONE = 0,     // No machine
-      EM_M32 = 1,      // AT&T WE 32100
-      EM_SPARC = 2,    // SPARC
-      EM_386 = 3,      // Intel 386
-      EM_68K = 4,      // Motorola 68000
-      EM_88K = 5,      // Motorola 88000
-      EM_486 = 6,      // Intel 486 (deprecated)
-      EM_860 = 7,      // Intel 80860
-      EM_MIPS = 8,     // MIPS R3000
-      EM_PPC = 20,     // PowerPC
-      EM_ARM = 40,     // ARM
-      EM_ALPHA = 41,   // DEC Alpha
-      EM_SPARCV9 = 43, // SPARC V9
-      EM_X86_64 = 62,  // AMD64
-      EM_HEXAGON = 164 // Qualcomm Hexagon
-    };
-
-    // ELF File classes
-    enum {
-      ELFCLASS32 = 1, // 32-bit object file
-      ELFCLASS64 = 2  // 64-bit object file
-    };
-
-    // ELF Endianess
-    enum {
-      ELFDATA2LSB = 1, // Little-endian object file
-      ELFDATA2MSB = 2  // Big-endian object file
-    };
-
-    explicit TargetELFWriterInfo(bool is64Bit_, bool isLittleEndian_);
-    virtual ~TargetELFWriterInfo();
-
-    unsigned short getEMachine() const { return EMachine; }
-    unsigned getEFlags() const { return 0; }
-    unsigned getEIClass() const { return is64Bit ? ELFCLASS64 : ELFCLASS32; }
-    unsigned getEIData() const {
-      return isLittleEndian ? ELFDATA2LSB : ELFDATA2MSB;
-    }
-
-    /// ELF Header and ELF Section Header Info
-    unsigned getHdrSize() const { return is64Bit ? 64 : 52; }
-    unsigned getSHdrSize() const { return is64Bit ? 64 : 40; }
-
-    /// Symbol Table Info
-    unsigned getSymTabEntrySize() const { return is64Bit ? 24 : 16; }
-
-    /// getPrefELFAlignment - Returns the preferred alignment for ELF. This
-    /// is used to align some sections.
-    unsigned getPrefELFAlignment() const { return is64Bit ? 8 : 4; }
-
-    /// getRelocationEntrySize - Entry size used in the relocation section
-    unsigned getRelocationEntrySize() const {
-      return is64Bit ? (hasRelocationAddend() ? 24 : 16)
-                     : (hasRelocationAddend() ? 12 : 8);
-    }
-
-    /// getRelocationType - Returns the target specific ELF Relocation type.
-    /// 'MachineRelTy' contains the object code independent relocation type
-    virtual unsigned getRelocationType(unsigned MachineRelTy) const = 0;
-
-    /// hasRelocationAddend - True if the target uses an addend in the
-    /// ELF relocation entry.
-    virtual bool hasRelocationAddend() const = 0;
-
-    /// getDefaultAddendForRelTy - Gets the default addend value for a
-    /// relocation entry based on the target ELF relocation type.
-    virtual long int getDefaultAddendForRelTy(unsigned RelTy,
-                                              long int Modifier = 0) const = 0;
-
-    /// getRelTySize - Returns the size of relocatable field in bits
-    virtual unsigned getRelocationTySize(unsigned RelTy) const = 0;
-
-    /// isPCRelativeRel - True if the relocation type is pc relative
-    virtual bool isPCRelativeRel(unsigned RelTy) const = 0;
-
-    /// getJumpTableRelocationTy - Returns the machine relocation type used
-    /// to reference a jumptable.
-    virtual unsigned getAbsoluteLabelMachineRelTy() const = 0;
-
-    /// computeRelocation - Some relocatable fields could be relocated
-    /// directly, avoiding the relocation symbol emission, compute the
-    /// final relocation value for this symbol.
-    virtual long int computeRelocation(unsigned SymOffset, unsigned RelOffset,
-                                       unsigned RelTy) const = 0;
-  };
-
-} // end llvm namespace
-
-#endif // LLVM_TARGET_TARGETELFWRITERINFO_H

Modified: llvm/branches/R600/include/llvm/Target/TargetLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetLowering.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetLowering.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetLowering.h Tue Nov 13 09:21:47 2012
@@ -103,6 +103,10 @@
     TypeWidenVector      // This vector should be widened into a larger vector.
   };
 
+  /// LegalizeKind holds the legalization kind that needs to happen to EVT
+  /// in order to type-legalize it.
+  typedef std::pair<LegalizeTypeAction, EVT> LegalizeKind;
+
   enum BooleanContent { // How the target represents true/false values.
     UndefinedBooleanContent,    // Only bit 0 counts, the rest can hold garbage.
     ZeroOrOneBooleanContent,        // All bits zero except for bit 0.
@@ -407,6 +411,13 @@
        getOperationAction(Op, VT) == Custom);
   }
 
+  /// isOperationExpand - Return true if the specified operation is illegal on
+  /// this target or unlikely to be made legal with custom lowering. This is
+  /// used to help guide high-level lowering decisions.
+  bool isOperationExpand(unsigned Op, EVT VT) const {
+    return (!isTypeLegal(VT) || getOperationAction(Op, VT) == Expand);
+  }
+
   /// isOperationLegal - Return true if the specified operation is legal on this
   /// target.
   bool isOperationLegal(unsigned Op, EVT VT) const {
@@ -1253,7 +1264,7 @@
 public:
   //===--------------------------------------------------------------------===//
   // Lowering methods - These methods must be implemented by targets so that
-  // the SelectionDAGLowering code knows how to lower these.
+  // the SelectionDAGBuilder code knows how to lower these.
   //
 
   /// LowerFormalArguments - This hook must be implemented to lower the
@@ -1954,8 +1965,7 @@
 
   ValueTypeActionImpl ValueTypeActions;
 
-  typedef std::pair<LegalizeTypeAction, EVT> LegalizeKind;
-
+public:
   LegalizeKind
   getTypeConversion(LLVMContext &Context, EVT VT) const {
     // If this is a simple type, use the ComputeRegisterProp mechanism.
@@ -1970,6 +1980,9 @@
          ValueTypeActions.getTypeAction(NVT.getSimpleVT()) != TypePromoteInteger)
          && "Promote may not follow Expand or Promote");
 
+      if (LA == TypeSplitVector)
+        NVT = EVT::getVectorVT(Context, VT.getVectorElementType(),
+                               VT.getVectorNumElements() / 2);
       return LegalizeKind(LA, NVT);
     }
 
@@ -2072,6 +2085,7 @@
     return LegalizeKind(TypeSplitVector, NVT);
   }
 
+private:
   std::vector<std::pair<EVT, const TargetRegisterClass*> > AvailableRegClasses;
 
   /// TargetDAGCombineArray - Targets can specify ISD nodes that they would

Modified: llvm/branches/R600/include/llvm/Target/TargetMachine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetMachine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetMachine.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetMachine.h Tue Nov 13 09:21:47 2012
@@ -34,7 +34,6 @@
 class PassManagerBase;
 class Target;
 class DataLayout;
-class TargetELFWriterInfo;
 class TargetFrameLowering;
 class TargetInstrInfo;
 class TargetIntrinsicInfo;
@@ -148,11 +147,6 @@
     return 0;
   }
 
-  /// getELFWriterInfo - If this target supports an ELF writer, return
-  /// information for it, otherwise return null.
-  ///
-  virtual const TargetELFWriterInfo *getELFWriterInfo() const { return 0; }
-
   /// hasMCRelaxAll - Check whether all machine code instructions should be
   /// relaxed.
   bool hasMCRelaxAll() const { return MCRelaxAll; }

Modified: llvm/branches/R600/include/llvm/Target/TargetRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetRegisterInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetRegisterInfo.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetRegisterInfo.h Tue Nov 13 09:21:47 2012
@@ -446,18 +446,6 @@
     return MCRegisterInfo::getMatchingSuperReg(Reg, SubIdx, RC->MC);
   }
 
-  /// canCombineSubRegIndices - Given a register class and a list of
-  /// subregister indices, return true if it's possible to combine the
-  /// subregister indices into one that corresponds to a larger
-  /// subregister. Return the new subregister index by reference. Note the
-  /// new index may be zero if the given subregisters can be combined to
-  /// form the whole register.
-  virtual bool canCombineSubRegIndices(const TargetRegisterClass *RC,
-                                       SmallVectorImpl<unsigned> &SubIndices,
-                                       unsigned &NewSubIdx) const {
-    return 0;
-  }
-
   /// getMatchingSuperRegClass - Return a subclass of the specified register
   /// class A so that each register in it has a sub-register of the
   /// specified sub-register index which is in the specified register class B.
@@ -488,6 +476,8 @@
   /// composeSubRegIndices - Return the subregister index you get from composing
   /// two subregister indices.
   ///
+  /// The special null sub-register index composes as the identity.
+  ///
   /// If R:a:b is the same register as R:c, then composeSubRegIndices(a, b)
   /// returns c. Note that composeSubRegIndices does not tell you about illegal
   /// compositions. If R does not have a subreg a, or R:a does not have a subreg
@@ -497,11 +487,19 @@
   /// ssub_0:S0 - ssub_3:S3 subregs.
   /// If you compose subreg indices dsub_1, ssub_0 you get ssub_2.
   ///
-  virtual unsigned composeSubRegIndices(unsigned a, unsigned b) const {
-    // This default implementation is correct for most targets.
-    return b;
+  unsigned composeSubRegIndices(unsigned a, unsigned b) const {
+    if (!a) return b;
+    if (!b) return a;
+    return composeSubRegIndicesImpl(a, b);
   }
 
+protected:
+  /// Overridden by TableGen in targets that have sub-registers.
+  virtual unsigned composeSubRegIndicesImpl(unsigned, unsigned) const {
+    llvm_unreachable("Target has no sub-registers");
+  }
+
+public:
   /// getCommonSuperRegClass - Find a common super-register class if it exists.
   ///
   /// Find a register class, SuperRC and two sub-register indices, PreA and

Modified: llvm/branches/R600/include/llvm/Target/TargetTransformImpl.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Target/TargetTransformImpl.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Target/TargetTransformImpl.h (original)
+++ llvm/branches/R600/include/llvm/Target/TargetTransformImpl.h Tue Nov 13 09:21:47 2012
@@ -16,6 +16,7 @@
 #define LLVM_TARGET_TARGET_TRANSFORMATION_IMPL_H
 
 #include "llvm/TargetTransformInfo.h"
+#include "llvm/CodeGen/ValueTypes.h"
 
 namespace llvm {
 
@@ -45,9 +46,52 @@
   virtual unsigned getJumpBufAlignment() const;
 
   virtual unsigned getJumpBufSize() const;
+
+  virtual bool shouldBuildLookupTables() const;
 };
 
-class VectorTargetTransformImpl : public VectorTargetTransformInfo { };
+class VectorTargetTransformImpl : public VectorTargetTransformInfo {
+protected:
+  const TargetLowering *TLI;
+
+  /// Estimate the cost of type-legalization and the legalized type.
+  std::pair<unsigned, MVT> getTypeLegalizationCost(Type *Ty) const;
+
+  /// Estimate the overhead of scalarizing an instruction. Insert and Extract
+  /// are set if the result needs to be inserted and/or extracted from vectors.
+  unsigned getScalarizationOverhead(Type *Ty, bool Insert, bool Extract) const;
+
+  // Get the ISD node that corresponds to the Instruction class opcode.
+  int InstructionOpcodeToISD(unsigned Opcode) const;
+
+public:
+  explicit VectorTargetTransformImpl(const TargetLowering *TL) : TLI(TL) {}
+
+  virtual ~VectorTargetTransformImpl() {}
+
+  virtual unsigned getInstrCost(unsigned Opcode, Type *Ty1, Type *Ty2) const;
+
+  virtual unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty) const;
+
+  virtual unsigned getBroadcastCost(Type *Tp) const;
+
+  virtual unsigned getCastInstrCost(unsigned Opcode, Type *Dst,
+                                    Type *Src) const;
+
+  virtual unsigned getCFInstrCost(unsigned Opcode) const;
+
+  virtual unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
+                                      Type *CondTy) const;
+
+  virtual unsigned getVectorInstrCost(unsigned Opcode, Type *Val,
+                                      unsigned Index) const;
+
+  virtual unsigned getMemoryOpCost(unsigned Opcode, Type *Src,
+                                   unsigned Alignment,
+                                   unsigned AddressSpace) const;
+
+  virtual unsigned getNumberOfParts(Type *Tp) const;
+};
 
 } // end llvm namespace
 

Modified: llvm/branches/R600/include/llvm/TargetTransformInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/TargetTransformInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/TargetTransformInfo.h (original)
+++ llvm/branches/R600/include/llvm/TargetTransformInfo.h Tue Nov 13 09:21:47 2012
@@ -45,19 +45,19 @@
   /// used.
   TargetTransformInfo();
 
-  explicit TargetTransformInfo(const ScalarTargetTransformInfo* S,
-                               const VectorTargetTransformInfo *V)
-    : ImmutablePass(ID), STTI(S), VTTI(V) {
-      initializeTargetTransformInfoPass(*PassRegistry::getPassRegistry());
-    }
+  TargetTransformInfo(const ScalarTargetTransformInfo* S,
+                      const VectorTargetTransformInfo *V)
+      : ImmutablePass(ID), STTI(S), VTTI(V) {
+    initializeTargetTransformInfoPass(*PassRegistry::getPassRegistry());
+  }
 
   TargetTransformInfo(const TargetTransformInfo &T) :
     ImmutablePass(ID), STTI(T.STTI), VTTI(T.VTTI) { }
 
-  const ScalarTargetTransformInfo* getScalarTargetTransformInfo() {
+  const ScalarTargetTransformInfo* getScalarTargetTransformInfo() const {
     return STTI;
   }
-  const VectorTargetTransformInfo* getVectorTargetTransformInfo() {
+  const VectorTargetTransformInfo* getVectorTargetTransformInfo() const {
     return VTTI;
   }
 
@@ -102,7 +102,7 @@
   /// isTruncateFree - Return true if it's free to truncate a value of
   /// type Ty1 to type Ty2. e.g. On x86 it's free to truncate a i32 value in
   /// register EAX to i16 by referencing its sub-register AX.
-  virtual bool isTruncateFree(Type * /*Ty1*/, Type * /*Ty2*/) const {
+  virtual bool isTruncateFree(Type *Ty1, Type *Ty2) const {
     return false;
   }
   /// Is this type legal.
@@ -117,10 +117,86 @@
   virtual unsigned getJumpBufSize() const {
     return 0;
   }
+  /// shouldBuildLookupTables - Return true if switches should be turned into
+  /// lookup tables for the target.
+  virtual bool shouldBuildLookupTables() const {
+    return true;
+  }
 };
 
+/// VectorTargetTransformInfo - This interface is used by the vectorizers
+/// to estimate the profitability of vectorization for different instructions.
 class VectorTargetTransformInfo {
-  // TODO: define an interface for VectorTargetTransformInfo.
+public:
+  virtual ~VectorTargetTransformInfo() {}
+
+  /// Returns the expected cost of the instruction opcode. The opcode is one of
+  /// the enums like Instruction::Add. The type arguments are the type of the
+  /// operation.
+  /// Most instructions only use the first type and in that case the second
+  /// operand is ignored.
+  ///
+  /// Exceptions:
+  /// * Br instructions do not use any of the types.
+  /// * Select instructions pass the return type as Ty1 and the selector as Ty2.
+  /// * Cast instructions pass the destination as Ty1 and the source as Ty2.
+  /// * Insert/Extract element pass only the vector type as Ty1.
+  /// * ShuffleVector, Load, Store do not use this call.
+  virtual unsigned getInstrCost(unsigned Opcode,
+                                Type *Ty1 = 0,
+                                Type *Ty2 = 0) const {
+    return 1;
+  }
+
+  /// Returns the expected cost of arithmetic ops, such as mul, xor, fsub, etc.
+  virtual unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty) const {
+    return 1;
+  }
+
+  /// Returns the cost of a vector broadcast of a scalar at place zero to a
+  /// vector of type 'Tp'.
+  virtual unsigned getBroadcastCost(Type *Tp) const {
+    return 1;
+  }
+
+  /// Returns the expected cost of cast instructions, such as bitcast, trunc,
+  /// zext, etc.
+  virtual unsigned getCastInstrCost(unsigned Opcode, Type *Dst,
+                                    Type *Src) const {
+    return 1;
+  }
+
+  /// Returns the expected cost of control-flow related instrutctions such as
+  /// Phi, Ret, Br.
+  virtual unsigned getCFInstrCost(unsigned Opcode) const {
+    return 1;
+  }
+
+  /// Returns the expected cost of compare and select instructions.
+  virtual unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
+                                      Type *CondTy = 0) const {
+    return 1;
+  }
+
+  /// Returns the expected cost of vector Insert and Extract.
+  /// Use -1 to indicate that there is no information on the index value.
+  virtual unsigned getVectorInstrCost(unsigned Opcode, Type *Val,
+                                      unsigned Index = -1) const {
+    return 1;
+  }
+
+  /// Returns the cost of Load and Store instructions.
+  virtual unsigned getMemoryOpCost(unsigned Opcode, Type *Src,
+                                   unsigned Alignment,
+                                   unsigned AddressSpace) const {
+    return 1;
+  }
+
+  /// Returns the number of pieces into which the provided type must be
+  /// split during legalization. Zero is returned when the answer is unknown.
+  virtual unsigned getNumberOfParts(Type *Tp) const {
+    return 0;
+  }
 };
 
 } // End llvm namespace

Modified: llvm/branches/R600/include/llvm/Transforms/IPO.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Transforms/IPO.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Transforms/IPO.h (original)
+++ llvm/branches/R600/include/llvm/Transforms/IPO.h Tue Nov 13 09:21:47 2012
@@ -104,23 +104,14 @@
 
 //===----------------------------------------------------------------------===//
 /// createInternalizePass - This pass loops over all of the functions in the
-/// input module, internalizing all globals (functions and variables) not part
-/// of the api.  If a list of symbols is specified with the
-/// -internalize-public-api-* command line options, those symbols are not
-/// internalized and all others are.  Otherwise if AllButMain is set and the
-/// main function is found, all other globals are marked as internal. If no api
-/// is supplied and AllButMain is not set, or no main function is found, nothing
-/// is internalized.
-///
-ModulePass *createInternalizePass(bool AllButMain);
-
-/// createInternalizePass - This pass loops over all of the functions in the
 /// input module, internalizing all globals (functions and variables) not in the
 /// given exportList.
 ///
 /// Note that commandline options that are used with the above function are not
-/// used now! Also, when exportList is empty, nothing is internalized.
+/// used now!
 ModulePass *createInternalizePass(const std::vector<const char *> &exportList);
+/// createInternalizePass - Same as above, but with an empty exportList.
+ModulePass *createInternalizePass();
 
 //===----------------------------------------------------------------------===//
 /// createDeadArgEliminationPass - This pass removes arguments from functions

Modified: llvm/branches/R600/include/llvm/Transforms/IPO/PassManagerBuilder.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Transforms/IPO/PassManagerBuilder.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Transforms/IPO/PassManagerBuilder.h (original)
+++ llvm/branches/R600/include/llvm/Transforms/IPO/PassManagerBuilder.h Tue Nov 13 09:21:47 2012
@@ -104,6 +104,7 @@
   bool DisableUnitAtATime;
   bool DisableUnrollLoops;
   bool Vectorize;
+  bool LoopVectorize;
 
 private:
   /// ExtensionList - This is list of all of the extensions that are registered.

Modified: llvm/branches/R600/include/llvm/Transforms/Utils/Local.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Transforms/Utils/Local.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Transforms/Utils/Local.h (original)
+++ llvm/branches/R600/include/llvm/Transforms/Utils/Local.h Tue Nov 13 09:21:47 2012
@@ -37,6 +37,7 @@
 class ConstantExpr;
 class DataLayout;
 class TargetLibraryInfo;
+class TargetTransformInfo;
 class DIBuilder;
 
 template<typename T> class SmallVectorImpl;
@@ -134,7 +135,8 @@
 /// of the CFG.  It returns true if a modification was made, possibly deleting
 /// the basic block that was pointed to.
 ///
-bool SimplifyCFG(BasicBlock *BB, const DataLayout *TD = 0);
+bool SimplifyCFG(BasicBlock *BB, const DataLayout *TD = 0,
+                 const TargetTransformInfo *TTI = 0);
 
 /// FoldBranchToCommonDest - If this basic block is ONLY a setcc and a branch,
 /// and if a predecessor branches to us and one of our successors, fold the
@@ -186,8 +188,7 @@
   bool isInBounds = cast<GEPOperator>(GEP)->isInBounds() && !NoAssumptions;
 
   // Build a mask for high order bits.
-  unsigned AS = cast<GEPOperator>(GEP)->getPointerAddressSpace();
-  unsigned IntPtrWidth = TD.getPointerSizeInBits(AS);
+  unsigned IntPtrWidth = TD.getPointerSizeInBits();
   uint64_t PtrSizeMask = ~0ULL >> (64-IntPtrWidth);
 
   for (User::op_iterator i = GEP->op_begin() + 1, e = GEP->op_end(); i != e;

Modified: llvm/branches/R600/include/llvm/Type.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/include/llvm/Type.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/include/llvm/Type.h (original)
+++ llvm/branches/R600/include/llvm/Type.h Tue Nov 13 09:21:47 2012
@@ -153,7 +153,7 @@
   /// isPPC_FP128Ty - Return true if this is powerpc long double.
   bool isPPC_FP128Ty() const { return getTypeID() == PPC_FP128TyID; }
 
-  /// isFloatingPointTy - Return true if this is one of the five floating point
+  /// isFloatingPointTy - Return true if this is one of the six floating point
   /// types
   bool isFloatingPointTy() const {
     return getTypeID() == HalfTyID || getTypeID() == FloatTyID ||
@@ -167,7 +167,7 @@
 
   /// isFPOrFPVectorTy - Return true if this is a FP type or a vector of FP.
   ///
-  bool isFPOrFPVectorTy() const;
+  bool isFPOrFPVectorTy() const { return getScalarType()->isFloatingPointTy(); }
  
   /// isLabelTy - Return true if this is 'label'.
   bool isLabelTy() const { return getTypeID() == LabelTyID; }
@@ -185,7 +185,7 @@
   /// isIntOrIntVectorTy - Return true if this is an integer type or a vector of
   /// integer types.
   ///
-  bool isIntOrIntVectorTy() const;
+  bool isIntOrIntVectorTy() const { return getScalarType()->isIntegerTy(); }
   
   /// isFunctionTy - True if this is an instance of FunctionType.
   ///
@@ -203,6 +203,11 @@
   ///
   bool isPointerTy() const { return getTypeID() == PointerTyID; }
 
+  /// isPtrOrPtrVectorTy - Return true if this is a pointer type or a vector of
+  /// pointer types.
+  ///
+  bool isPtrOrPtrVectorTy() const { return getScalarType()->isPointerTy(); }
+ 
   /// isVectorTy - True if this is an instance of VectorType.
   ///
   bool isVectorTy() const { return getTypeID() == VectorTyID; }
@@ -293,6 +298,7 @@
 
   /// getScalarType - If this is a vector type, return the element type,
   /// otherwise return 'this'.
+  const Type *getScalarType() const;
   Type *getScalarType();
 
   //===--------------------------------------------------------------------===//
@@ -340,8 +346,10 @@
   unsigned getVectorNumElements() const;
   Type *getVectorElementType() const { return getSequentialElementType(); }
 
-  unsigned getPointerAddressSpace() const;
   Type *getPointerElementType() const { return getSequentialElementType(); }
+
+  /// \brief Get the address space of this pointer or pointer vector type.
+  unsigned getPointerAddressSpace() const;
   
   //===--------------------------------------------------------------------===//
   // Static members exported by the Type class itself.  Useful for getting

Modified: llvm/branches/R600/lib/Analysis/Analysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/Analysis.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/Analysis.cpp (original)
+++ llvm/branches/R600/lib/Analysis/Analysis.cpp Tue Nov 13 09:21:47 2012
@@ -26,6 +26,7 @@
   initializeBasicAliasAnalysisPass(Registry);
   initializeBlockFrequencyInfoPass(Registry);
   initializeBranchProbabilityInfoPass(Registry);
+  initializeCostModelAnalysisPass(Registry);
   initializeCFGViewerPass(Registry);
   initializeCFGPrinterPass(Registry);
   initializeCFGOnlyViewerPass(Registry);
@@ -47,7 +48,6 @@
   initializeLazyValueInfoPass(Registry);
   initializeLibCallAliasAnalysisPass(Registry);
   initializeLintPass(Registry);
-  initializeLoopDependenceAnalysisPass(Registry);
   initializeLoopInfoPass(Registry);
   initializeMemDepPrinterPass(Registry);
   initializeMemoryDependenceAnalysisPass(Registry);

Modified: llvm/branches/R600/lib/Analysis/BasicAliasAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/BasicAliasAnalysis.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/BasicAliasAnalysis.cpp (original)
+++ llvm/branches/R600/lib/Analysis/BasicAliasAnalysis.cpp Tue Nov 13 09:21:47 2012
@@ -58,12 +58,12 @@
   // then it has not escaped before entering the function.  Check if it escapes
   // inside the function.
   if (const Argument *A = dyn_cast<Argument>(V))
-    if (A->hasByValAttr() || A->hasNoAliasAttr()) {
-      // Don't bother analyzing arguments already known not to escape.
-      if (A->hasNoCaptureAttr())
-        return true;
+    if (A->hasByValAttr() || A->hasNoAliasAttr())
+      // Note even if the argument is marked nocapture we still need to check
+      // for copies made inside the function. The nocapture attribute only
+      // specifies that there are no copies made that outlive the function.
       return !PointerMayBeCaptured(V, false, /*StoreCaptures=*/true);
-    }
+
   return false;
 }
 
@@ -286,8 +286,7 @@
       V = GEPOp->getOperand(0);
       continue;
     }
-
-    unsigned AS = GEPOp->getPointerAddressSpace();
+    
     // Walk the indices of the GEP, accumulating them into BaseOff/VarIndices.
     gep_type_iterator GTI = gep_type_begin(GEPOp);
     for (User::const_op_iterator I = GEPOp->op_begin()+1,
@@ -316,7 +315,7 @@
       // If the integer type is smaller than the pointer size, it is implicitly
       // sign extended to pointer size.
       unsigned Width = cast<IntegerType>(Index->getType())->getBitWidth();
-      if (TD->getPointerSizeInBits(AS) > Width)
+      if (TD->getPointerSizeInBits() > Width)
         Extension = EK_SignExt;
       
       // Use GetLinearExpression to decompose the index into a C1*V+C2 form.
@@ -345,7 +344,7 @@
       
       // Make sure that we have a scale that makes sense for this target's
       // pointer size.
-      if (unsigned ShiftBits = 64-TD->getPointerSizeInBits(AS)) {
+      if (unsigned ShiftBits = 64-TD->getPointerSizeInBits()) {
         Scale <<= ShiftBits;
         Scale = (int64_t)Scale >> ShiftBits;
       }
@@ -1246,6 +1245,7 @@
     std::swap(V1, V2);
     std::swap(V1Size, V2Size);
     std::swap(O1, O2);
+    std::swap(V1TBAAInfo, V2TBAAInfo);
   }
   if (const GEPOperator *GV1 = dyn_cast<GEPOperator>(V1)) {
     AliasResult Result = aliasGEP(GV1, V1Size, V1TBAAInfo, V2, V2Size, V2TBAAInfo, O1, O2);
@@ -1255,6 +1255,7 @@
   if (isa<PHINode>(V2) && !isa<PHINode>(V1)) {
     std::swap(V1, V2);
     std::swap(V1Size, V2Size);
+    std::swap(V1TBAAInfo, V2TBAAInfo);
   }
   if (const PHINode *PN = dyn_cast<PHINode>(V1)) {
     AliasResult Result = aliasPHI(PN, V1Size, V1TBAAInfo,
@@ -1265,6 +1266,7 @@
   if (isa<SelectInst>(V2) && !isa<SelectInst>(V1)) {
     std::swap(V1, V2);
     std::swap(V1Size, V2Size);
+    std::swap(V1TBAAInfo, V2TBAAInfo);
   }
   if (const SelectInst *S1 = dyn_cast<SelectInst>(V1)) {
     AliasResult Result = aliasSelect(S1, V1Size, V1TBAAInfo,

Modified: llvm/branches/R600/lib/Analysis/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Analysis/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -10,6 +10,7 @@
   BranchProbabilityInfo.cpp
   CFGPrinter.cpp
   CaptureTracking.cpp
+  CostModel.cpp
   CodeMetrics.cpp
   ConstantFolding.cpp
   DbgInfoPrinter.cpp
@@ -27,7 +28,6 @@
   LibCallSemantics.cpp
   Lint.cpp
   Loads.cpp
-  LoopDependenceAnalysis.cpp
   LoopInfo.cpp
   LoopPass.cpp
   MemDepPrinter.cpp

Modified: llvm/branches/R600/lib/Analysis/CodeMetrics.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/CodeMetrics.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/CodeMetrics.cpp (original)
+++ llvm/branches/R600/lib/Analysis/CodeMetrics.cpp Tue Nov 13 09:21:47 2012
@@ -91,16 +91,14 @@
     // which doesn't contain values outside the range of a pointer.
     if (isa<IntToPtrInst>(CI) && TD &&
         TD->isLegalInteger(Op->getType()->getScalarSizeInBits()) &&
-        Op->getType()->getScalarSizeInBits() <= TD->getPointerSizeInBits(
-          cast<IntToPtrInst>(CI)->getAddressSpace()))
+        Op->getType()->getScalarSizeInBits() <= TD->getPointerSizeInBits())
       return true;
 
     // A ptrtoint cast is free so long as the result is large enough to store
     // the pointer, and a legal integer type.
     if (isa<PtrToIntInst>(CI) && TD &&
         TD->isLegalInteger(Op->getType()->getScalarSizeInBits()) &&
-        Op->getType()->getScalarSizeInBits() >= TD->getPointerSizeInBits(
-          cast<PtrToIntInst>(CI)->getPointerAddressSpace()))
+        Op->getType()->getScalarSizeInBits() >= TD->getPointerSizeInBits())
       return true;
 
     // trunc to a native type is free (assuming the target has compare and

Modified: llvm/branches/R600/lib/Analysis/ConstantFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/ConstantFolding.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/ConstantFolding.cpp (original)
+++ llvm/branches/R600/lib/Analysis/ConstantFolding.cpp Tue Nov 13 09:21:47 2012
@@ -41,7 +41,7 @@
 // Constant Folding internal helper functions
 //===----------------------------------------------------------------------===//
 
-/// FoldBitCast - Constant fold bitcast, symbolically evaluating it with 
+/// FoldBitCast - Constant fold bitcast, symbolically evaluating it with
 /// DataLayout.  This always returns a non-null constant, but it may be a
 /// ConstantExpr if unfoldable.
 static Constant *FoldBitCast(Constant *C, Type *DestTy,
@@ -59,9 +59,9 @@
       return ConstantExpr::getBitCast(C, DestTy);
 
     unsigned NumSrcElts = CDV->getType()->getNumElements();
-    
+
     Type *SrcEltTy = CDV->getType()->getElementType();
-    
+
     // If the vector is a vector of floating point, convert it to vector of int
     // to simplify things.
     if (SrcEltTy->isFloatingPointTy()) {
@@ -72,7 +72,7 @@
       C = ConstantExpr::getBitCast(C, SrcIVTy);
       CDV = cast<ConstantDataVector>(C);
     }
-    
+
     // Now that we know that the input value is a vector of integers, just shift
     // and insert them into our result.
     unsigned BitShift = TD.getTypeAllocSizeInBits(SrcEltTy);
@@ -84,43 +84,43 @@
       else
         Result |= CDV->getElementAsInteger(i);
     }
-   
+
     return ConstantInt::get(IT, Result);
   }
-  
+
   // The code below only handles casts to vectors currently.
   VectorType *DestVTy = dyn_cast<VectorType>(DestTy);
   if (DestVTy == 0)
     return ConstantExpr::getBitCast(C, DestTy);
-  
+
   // If this is a scalar -> vector cast, convert the input into a <1 x scalar>
   // vector so the code below can handle it uniformly.
   if (isa<ConstantFP>(C) || isa<ConstantInt>(C)) {
     Constant *Ops = C; // don't take the address of C!
     return FoldBitCast(ConstantVector::get(Ops), DestTy, TD);
   }
-  
+
   // If this is a bitcast from constant vector -> vector, fold it.
   if (!isa<ConstantDataVector>(C) && !isa<ConstantVector>(C))
     return ConstantExpr::getBitCast(C, DestTy);
-  
+
   // If the element types match, VMCore can fold it.
   unsigned NumDstElt = DestVTy->getNumElements();
   unsigned NumSrcElt = C->getType()->getVectorNumElements();
   if (NumDstElt == NumSrcElt)
     return ConstantExpr::getBitCast(C, DestTy);
-  
+
   Type *SrcEltTy = C->getType()->getVectorElementType();
   Type *DstEltTy = DestVTy->getElementType();
-  
-  // Otherwise, we're changing the number of elements in a vector, which 
+
+  // Otherwise, we're changing the number of elements in a vector, which
   // requires endianness information to do the right thing.  For example,
   //    bitcast (<2 x i64> <i64 0, i64 1> to <4 x i32>)
   // folds to (little endian):
   //    <4 x i32> <i32 0, i32 0, i32 1, i32 0>
   // and to (big endian):
   //    <4 x i32> <i32 0, i32 0, i32 0, i32 1>
-  
+
   // First thing is first.  We only want to think about integer here, so if
   // we have something in FP form, recast it as integer.
   if (DstEltTy->isFloatingPointTy()) {
@@ -130,11 +130,11 @@
       VectorType::get(IntegerType::get(C->getContext(), FPWidth), NumDstElt);
     // Recursively handle this integer conversion, if possible.
     C = FoldBitCast(C, DestIVTy, TD);
-    
+
     // Finally, VMCore can handle this now that #elts line up.
     return ConstantExpr::getBitCast(C, DestTy);
   }
-  
+
   // Okay, we know the destination is integer, if the input is FP, convert
   // it to integer first.
   if (SrcEltTy->isFloatingPointTy()) {
@@ -148,13 +148,13 @@
         !isa<ConstantDataVector>(C))
       return C;
   }
-  
+
   // Now we know that the input and output vectors are both integer vectors
   // of the same size, and that their #elements is not the same.  Do the
   // conversion here, which depends on whether the input or output has
   // more elements.
   bool isLittleEndian = TD.isLittleEndian();
-  
+
   SmallVector<Constant*, 32> Result;
   if (NumDstElt < NumSrcElt) {
     // Handle: bitcast (<4 x i32> <i32 0, i32 1, i32 2, i32 3> to <2 x i64>)
@@ -170,15 +170,15 @@
         Constant *Src =dyn_cast<ConstantInt>(C->getAggregateElement(SrcElt++));
         if (!Src)  // Reject constantexpr elements.
           return ConstantExpr::getBitCast(C, DestTy);
-        
+
         // Zero extend the element to the right size.
         Src = ConstantExpr::getZExt(Src, Elt->getType());
-        
+
         // Shift it to the right place, depending on endianness.
-        Src = ConstantExpr::getShl(Src, 
+        Src = ConstantExpr::getShl(Src,
                                    ConstantInt::get(Src->getType(), ShiftAmt));
         ShiftAmt += isLittleEndian ? SrcBitSize : -SrcBitSize;
-        
+
         // Mix it in.
         Elt = ConstantExpr::getOr(Elt, Src);
       }
@@ -186,30 +186,30 @@
     }
     return ConstantVector::get(Result);
   }
-  
+
   // Handle: bitcast (<2 x i64> <i64 0, i64 1> to <4 x i32>)
   unsigned Ratio = NumDstElt/NumSrcElt;
   unsigned DstBitSize = DstEltTy->getPrimitiveSizeInBits();
-  
+
   // Loop over each source value, expanding into multiple results.
   for (unsigned i = 0; i != NumSrcElt; ++i) {
     Constant *Src = dyn_cast<ConstantInt>(C->getAggregateElement(i));
     if (!Src)  // Reject constantexpr elements.
       return ConstantExpr::getBitCast(C, DestTy);
-    
+
     unsigned ShiftAmt = isLittleEndian ? 0 : DstBitSize*(Ratio-1);
     for (unsigned j = 0; j != Ratio; ++j) {
       // Shift the piece of the value into the right place, depending on
       // endianness.
-      Constant *Elt = ConstantExpr::getLShr(Src, 
+      Constant *Elt = ConstantExpr::getLShr(Src,
                                   ConstantInt::get(Src->getType(), ShiftAmt));
       ShiftAmt += isLittleEndian ? DstBitSize : -DstBitSize;
-      
+
       // Truncate and remember this piece.
       Result.push_back(ConstantExpr::getTrunc(Elt, DstEltTy));
     }
   }
-  
+
   return ConstantVector::get(Result);
 }
 
@@ -224,28 +224,28 @@
     Offset = 0;
     return true;
   }
-  
+
   // Otherwise, if this isn't a constant expr, bail out.
   ConstantExpr *CE = dyn_cast<ConstantExpr>(C);
   if (!CE) return false;
-  
+
   // Look through ptr->int and ptr->ptr casts.
   if (CE->getOpcode() == Instruction::PtrToInt ||
       CE->getOpcode() == Instruction::BitCast)
     return IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, TD);
-  
-  // i32* getelementptr ([5 x i32]* @a, i32 0, i32 5)    
+
+  // i32* getelementptr ([5 x i32]* @a, i32 0, i32 5)
   if (CE->getOpcode() == Instruction::GetElementPtr) {
     // Cannot compute this if the element type of the pointer is missing size
     // info.
     if (!cast<PointerType>(CE->getOperand(0)->getType())
                  ->getElementType()->isSized())
       return false;
-    
+
     // If the base isn't a global+constant, we aren't either.
     if (!IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, TD))
       return false;
-    
+
     // Otherwise, add any offset that our operands provide.
     gep_type_iterator GTI = gep_type_begin(CE);
     for (User::const_op_iterator i = CE->op_begin() + 1, e = CE->op_end();
@@ -253,7 +253,7 @@
       ConstantInt *CI = dyn_cast<ConstantInt>(*i);
       if (!CI) return false;  // Index isn't a simple constant?
       if (CI->isZero()) continue;  // Not adding anything.
-      
+
       if (StructType *ST = dyn_cast<StructType>(*GTI)) {
         // N = N + Offset
         Offset += TD.getStructLayout(ST)->getElementOffset(CI->getZExtValue());
@@ -264,7 +264,7 @@
     }
     return true;
   }
-  
+
   return false;
 }
 
@@ -277,27 +277,27 @@
                                const DataLayout &TD) {
   assert(ByteOffset <= TD.getTypeAllocSize(C->getType()) &&
          "Out of range access");
-  
+
   // If this element is zero or undefined, we can just return since *CurPtr is
   // zero initialized.
   if (isa<ConstantAggregateZero>(C) || isa<UndefValue>(C))
     return true;
-  
+
   if (ConstantInt *CI = dyn_cast<ConstantInt>(C)) {
     if (CI->getBitWidth() > 64 ||
         (CI->getBitWidth() & 7) != 0)
       return false;
-    
+
     uint64_t Val = CI->getZExtValue();
     unsigned IntBytes = unsigned(CI->getBitWidth()/8);
-    
+
     for (unsigned i = 0; i != BytesLeft && ByteOffset != IntBytes; ++i) {
       CurPtr[i] = (unsigned char)(Val >> (ByteOffset * 8));
       ++ByteOffset;
     }
     return true;
   }
-  
+
   if (ConstantFP *CFP = dyn_cast<ConstantFP>(C)) {
     if (CFP->getType()->isDoubleTy()) {
       C = FoldBitCast(C, Type::getInt64Ty(C->getContext()), TD);
@@ -309,13 +309,13 @@
     }
     return false;
   }
-  
+
   if (ConstantStruct *CS = dyn_cast<ConstantStruct>(C)) {
     const StructLayout *SL = TD.getStructLayout(CS->getType());
     unsigned Index = SL->getElementContainingOffset(ByteOffset);
     uint64_t CurEltOffset = SL->getElementOffset(Index);
     ByteOffset -= CurEltOffset;
-    
+
     while (1) {
       // If the element access is to the element itself and not to tail padding,
       // read the bytes from the element.
@@ -325,9 +325,9 @@
           !ReadDataFromGlobal(CS->getOperand(Index), ByteOffset, CurPtr,
                               BytesLeft, TD))
         return false;
-      
+
       ++Index;
-      
+
       // Check to see if we read from the last struct element, if so we're done.
       if (Index == CS->getType()->getNumElements())
         return true;
@@ -375,11 +375,11 @@
     }
     return true;
   }
-      
+
   if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
     if (CE->getOpcode() == Instruction::IntToPtr &&
-        CE->getOperand(0)->getType() == TD.getIntPtrType(CE->getContext())) 
-      return ReadDataFromGlobal(CE->getOperand(0), ByteOffset, CurPtr, 
+        CE->getOperand(0)->getType() == TD.getIntPtrType(CE->getContext()))
+      return ReadDataFromGlobal(CE->getOperand(0), ByteOffset, CurPtr,
                                 BytesLeft, TD);
   }
 
@@ -391,7 +391,7 @@
                                                  const DataLayout &TD) {
   Type *LoadTy = cast<PointerType>(C->getType())->getElementType();
   IntegerType *IntType = dyn_cast<IntegerType>(LoadTy);
-  
+
   // If this isn't an integer load we can't fold it directly.
   if (!IntType) {
     // If this is a float/double load, we can try folding it as an int32/64 load
@@ -415,15 +415,15 @@
       return FoldBitCast(Res, LoadTy, TD);
     return 0;
   }
-  
+
   unsigned BytesLoaded = (IntType->getBitWidth() + 7) / 8;
   if (BytesLoaded > 32 || BytesLoaded == 0) return 0;
-  
+
   GlobalValue *GVal;
   int64_t Offset;
   if (!IsConstantOffsetFromGlobal(C, GVal, Offset, TD))
     return 0;
-  
+
   GlobalVariable *GV = dyn_cast<GlobalVariable>(GVal);
   if (!GV || !GV->isConstant() || !GV->hasDefinitiveInitializer() ||
       !GV->getInitializer()->getType()->isSized())
@@ -432,11 +432,11 @@
   // If we're loading off the beginning of the global, some bytes may be valid,
   // but we don't try to handle this.
   if (Offset < 0) return 0;
-  
+
   // If we're not accessing anything in this constant, the result is undefined.
   if (uint64_t(Offset) >= TD.getTypeAllocSize(GV->getInitializer()->getType()))
     return UndefValue::get(IntType);
-  
+
   unsigned char RawBytes[32] = {0};
   if (!ReadDataFromGlobal(GV->getInitializer(), Offset, RawBytes,
                           BytesLoaded, TD))
@@ -464,15 +464,15 @@
   // If the loaded value isn't a constant expr, we can't handle it.
   ConstantExpr *CE = dyn_cast<ConstantExpr>(C);
   if (!CE) return 0;
-  
+
   if (CE->getOpcode() == Instruction::GetElementPtr) {
     if (GlobalVariable *GV = dyn_cast<GlobalVariable>(CE->getOperand(0)))
       if (GV->isConstant() && GV->hasDefinitiveInitializer())
-        if (Constant *V = 
+        if (Constant *V =
              ConstantFoldLoadThroughGEPConstantExpr(GV->getInitializer(), CE))
           return V;
   }
-  
+
   // Instead of loading constant c string, use corresponding integer value
   // directly if string length is small enough.
   StringRef Str;
@@ -500,14 +500,14 @@
         SingleChar = 0;
         StrVal = (StrVal << 8) | SingleChar;
       }
-      
+
       Constant *Res = ConstantInt::get(CE->getContext(), StrVal);
       if (Ty->isFloatingPointTy())
         Res = ConstantExpr::getBitCast(Res, Ty);
       return Res;
     }
   }
-  
+
   // If this load comes from anywhere in a constant global, and if the global
   // is all undef or zero, we know what it loads.
   if (GlobalVariable *GV =
@@ -520,7 +520,7 @@
         return UndefValue::get(ResTy);
     }
   }
-  
+
   // Try hard to fold loads from bitcasted strange and non-type-safe things.  We
   // currently don't do any of this for big endian systems.  It can be
   // generalized in the future if someone is interested.
@@ -531,7 +531,7 @@
 
 static Constant *ConstantFoldLoadInst(const LoadInst *LI, const DataLayout *TD){
   if (LI->isVolatile()) return 0;
-  
+
   if (Constant *C = dyn_cast<Constant>(LI->getOperand(0)))
     return ConstantFoldLoadFromConstPtr(C, TD);
 
@@ -540,23 +540,23 @@
 
 /// SymbolicallyEvaluateBinop - One of Op0/Op1 is a constant expression.
 /// Attempt to symbolically evaluate the result of a binary operator merging
-/// these together.  If target data info is available, it is provided as TD, 
+/// these together.  If target data info is available, it is provided as TD,
 /// otherwise TD is null.
 static Constant *SymbolicallyEvaluateBinop(unsigned Opc, Constant *Op0,
                                            Constant *Op1, const DataLayout *TD){
   // SROA
-  
+
   // Fold (and 0xffffffff00000000, (shl x, 32)) -> shl.
   // Fold (lshr (or X, Y), 32) -> (lshr [X/Y], 32) if one doesn't contribute
   // bits.
-  
-  
+
+
   // If the constant expr is something like &A[123] - &A[4].f, fold this into a
   // constant.  This happens frequently when iterating over a global array.
   if (Opc == Instruction::Sub && TD) {
     GlobalValue *GV1, *GV2;
     int64_t Offs1, Offs2;
-    
+
     if (IsConstantOffsetFromGlobal(Op0, GV1, Offs1, *TD))
       if (IsConstantOffsetFromGlobal(Op1, GV2, Offs2, *TD) &&
           GV1 == GV2) {
@@ -564,7 +564,7 @@
         return ConstantInt::get(Op0->getType(), Offs1-Offs2);
       }
   }
-    
+
   return 0;
 }
 
@@ -628,14 +628,14 @@
   if (!TD || !cast<PointerType>(Ptr->getType())->getElementType()->isSized() ||
       !Ptr->getType()->isPointerTy())
     return 0;
-  
+
   Type *IntPtrTy = TD->getIntPtrType(Ptr->getContext());
 
   // If this is a constant expr gep that is effectively computing an
   // "offsetof", fold it into 'cast int Size to T*' instead of 'gep 0, 0, 12'
   for (unsigned i = 1, e = Ops.size(); i != e; ++i)
     if (!isa<ConstantInt>(Ops[i])) {
-      
+
       // If this is "gep i8* Ptr, (sub 0, V)", fold this as:
       // "inttoptr (sub (ptrtoint Ptr), V)"
       if (Ops.size() == 2 &&
@@ -709,12 +709,12 @@
         // The only pointer indexing we'll do is on the first index of the GEP.
         if (!NewIdxs.empty())
           break;
-       
+
         // Only handle pointers to sized types, not pointers to functions.
         if (!ATy->getElementType()->isSized())
           return 0;
       }
-        
+
       // Determine which element of the array the offset points into.
       APInt ElemSize(BitWidth, TD->getTypeAllocSize(ATy->getElementType()));
       IntegerType *IntPtrTy = TD->getIntPtrType(Ty->getContext());
@@ -837,7 +837,7 @@
   if (const CmpInst *CI = dyn_cast<CmpInst>(I))
     return ConstantFoldCompareInstOperands(CI->getPredicate(), Ops[0], Ops[1],
                                            TD, TLI);
-  
+
   if (const LoadInst *LI = dyn_cast<LoadInst>(I))
     return ConstantFoldLoadInst(LI, TD);
 
@@ -887,19 +887,19 @@
 /// information, due to only being passed an opcode and operands. Constant
 /// folding using this function strips this information.
 ///
-Constant *llvm::ConstantFoldInstOperands(unsigned Opcode, Type *DestTy, 
+Constant *llvm::ConstantFoldInstOperands(unsigned Opcode, Type *DestTy,
                                          ArrayRef<Constant *> Ops,
                                          const DataLayout *TD,
-                                         const TargetLibraryInfo *TLI) {                                         
+                                         const TargetLibraryInfo *TLI) {
   // Handle easy binops first.
   if (Instruction::isBinaryOp(Opcode)) {
     if (isa<ConstantExpr>(Ops[0]) || isa<ConstantExpr>(Ops[1]))
       if (Constant *C = SymbolicallyEvaluateBinop(Opcode, Ops[0], Ops[1], TD))
         return C;
-    
+
     return ConstantExpr::get(Opcode, Ops[0], Ops[1]);
   }
-  
+
   switch (Opcode) {
   default: return 0;
   case Instruction::ICmp:
@@ -916,11 +916,10 @@
       if (TD && CE->getOpcode() == Instruction::IntToPtr) {
         Constant *Input = CE->getOperand(0);
         unsigned InWidth = Input->getType()->getScalarSizeInBits();
-        unsigned AS = cast<PointerType>(CE->getType())->getAddressSpace();
-        if (TD->getPointerSizeInBits(AS) < InWidth) {
-          Constant *Mask = 
+        if (TD->getPointerSizeInBits() < InWidth) {
+          Constant *Mask =
             ConstantInt::get(CE->getContext(), APInt::getLowBitsSet(InWidth,
-                                                  TD->getPointerSizeInBits(AS)));
+                                                  TD->getPointerSizeInBits()));
           Input = ConstantExpr::getAnd(Input, Mask);
         }
         // Do a zext or trunc to get to the dest size.
@@ -933,10 +932,9 @@
     // the int size is >= the ptr size.  This requires knowing the width of a
     // pointer, so it can't be done in ConstantExpr::getCast.
     if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ops[0]))
-      if (TD && CE->getOpcode() == Instruction::PtrToInt &&
-          TD->getPointerSizeInBits(
-            cast<PointerType>(CE->getOperand(0)->getType())->getAddressSpace())
-          <= CE->getType()->getScalarSizeInBits())
+      if (TD &&
+          TD->getPointerSizeInBits() <= CE->getType()->getScalarSizeInBits() &&
+          CE->getOpcode() == Instruction::PtrToInt)
         return FoldBitCast(CE->getOperand(0), DestTy, *TD);
 
     return ConstantExpr::getCast(Opcode, Ops[0], DestTy);
@@ -967,7 +965,7 @@
       return C;
     if (Constant *C = SymbolicallyEvaluateGEP(Ops, DestTy, TD, TLI))
       return C;
-    
+
     return ConstantExpr::getGetElementPtr(Ops[0], Ops.slice(1));
   }
 }
@@ -977,7 +975,7 @@
 /// returns a constant expression of the specified operands.
 ///
 Constant *llvm::ConstantFoldCompareInstOperands(unsigned Predicate,
-                                                Constant *Ops0, Constant *Ops1, 
+                                                Constant *Ops0, Constant *Ops1,
                                                 const DataLayout *TD,
                                                 const TargetLibraryInfo *TLI) {
   // fold: icmp (inttoptr x), null         -> icmp x, 0
@@ -998,17 +996,17 @@
         Constant *Null = Constant::getNullValue(C->getType());
         return ConstantFoldCompareInstOperands(Predicate, C, Null, TD, TLI);
       }
-      
+
       // Only do this transformation if the int is intptrty in size, otherwise
       // there is a truncation or extension that we aren't modeling.
-      if (CE0->getOpcode() == Instruction::PtrToInt && 
+      if (CE0->getOpcode() == Instruction::PtrToInt &&
           CE0->getType() == IntPtrTy) {
         Constant *C = CE0->getOperand(0);
         Constant *Null = Constant::getNullValue(C->getType());
         return ConstantFoldCompareInstOperands(Predicate, C, Null, TD, TLI);
       }
     }
-    
+
     if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(Ops1)) {
       if (TD && CE0->getOpcode() == CE1->getOpcode()) {
         Type *IntPtrTy = TD->getIntPtrType(CE0->getContext());
@@ -1032,24 +1030,24 @@
                                                  CE1->getOperand(0), TD, TLI);
       }
     }
-    
+
     // icmp eq (or x, y), 0 -> (icmp eq x, 0) & (icmp eq y, 0)
     // icmp ne (or x, y), 0 -> (icmp ne x, 0) | (icmp ne y, 0)
     if ((Predicate == ICmpInst::ICMP_EQ || Predicate == ICmpInst::ICMP_NE) &&
         CE0->getOpcode() == Instruction::Or && Ops1->isNullValue()) {
-      Constant *LHS = 
+      Constant *LHS =
         ConstantFoldCompareInstOperands(Predicate, CE0->getOperand(0), Ops1,
                                         TD, TLI);
-      Constant *RHS = 
+      Constant *RHS =
         ConstantFoldCompareInstOperands(Predicate, CE0->getOperand(1), Ops1,
                                         TD, TLI);
-      unsigned OpC = 
+      unsigned OpC =
         Predicate == ICmpInst::ICMP_EQ ? Instruction::And : Instruction::Or;
       Constant *Ops[] = { LHS, RHS };
       return ConstantFoldInstOperands(OpC, LHS->getType(), Ops, TD, TLI);
     }
   }
-  
+
   return ConstantExpr::getCompare(Predicate, Ops0, Ops1);
 }
 
@@ -1057,7 +1055,7 @@
 /// ConstantFoldLoadThroughGEPConstantExpr - Given a constant and a
 /// getelementptr constantexpr, return the constant value being addressed by the
 /// constant expression, or null if something is funny and we can't decide.
-Constant *llvm::ConstantFoldLoadThroughGEPConstantExpr(Constant *C, 
+Constant *llvm::ConstantFoldLoadThroughGEPConstantExpr(Constant *C,
                                                        ConstantExpr *CE) {
   if (!CE->getOperand(1)->isNullValue())
     return 0;  // Do not allow stepping over the value!
@@ -1127,14 +1125,14 @@
 
   if (!F->hasName()) return false;
   StringRef Name = F->getName();
-  
+
   // In these cases, the check of the length is required.  We don't want to
   // return true for a name like "cos\0blah" which strcmp would return equal to
   // "cos", but has length 8.
   switch (Name[0]) {
   default: return false;
   case 'a':
-    return Name == "acos" || Name == "asin" || 
+    return Name == "acos" || Name == "asin" ||
       Name == "atan" || Name == "atan2";
   case 'c':
     return Name == "cos" || Name == "ceil" || Name == "cosf" || Name == "cosh";
@@ -1154,7 +1152,7 @@
   }
 }
 
-static Constant *ConstantFoldFP(double (*NativeFP)(double), double V, 
+static Constant *ConstantFoldFP(double (*NativeFP)(double), double V,
                                 Type *Ty) {
   sys::llvm_fenv_clearexcept();
   V = NativeFP(V);
@@ -1162,7 +1160,7 @@
     sys::llvm_fenv_clearexcept();
     return 0;
   }
-  
+
   if (Ty->isFloatTy())
     return ConstantFP::get(Ty->getContext(), APFloat((float)V));
   if (Ty->isDoubleTy())
@@ -1178,7 +1176,7 @@
     sys::llvm_fenv_clearexcept();
     return 0;
   }
-  
+
   if (Ty->isFloatTy())
     return ConstantFP::get(Ty->getContext(), APFloat((float)V));
   if (Ty->isDoubleTy())
@@ -1272,7 +1270,7 @@
       case 'e':
         if (Name == "exp" && TLI->has(LibFunc::exp))
           return ConstantFoldFP(exp, V, Ty);
-  
+
         if (Name == "exp2" && TLI->has(LibFunc::exp2)) {
           // Constant fold exp2(x) as pow(2,x) in case the host doesn't have a
           // C99 library.
@@ -1348,7 +1346,7 @@
     }
 
     // Support ConstantVector in case we have an Undef in the top.
-    if (isa<ConstantVector>(Operands[0]) || 
+    if (isa<ConstantVector>(Operands[0]) ||
         isa<ConstantDataVector>(Operands[0])) {
       Constant *Op = cast<Constant>(Operands[0]);
       switch (F->getIntrinsicID()) {
@@ -1367,11 +1365,11 @@
       case Intrinsic::x86_sse2_cvttsd2si64:
         if (ConstantFP *FPOp =
               dyn_cast_or_null<ConstantFP>(Op->getAggregateElement(0U)))
-          return ConstantFoldConvertToInt(FPOp->getValueAPF(), 
+          return ConstantFoldConvertToInt(FPOp->getValueAPF(),
                                           /*roundTowardZero=*/true, Ty);
       }
     }
-  
+
     if (isa<UndefValue>(Operands[0])) {
       if (F->getIntrinsicID() == Intrinsic::bswap)
         return Operands[0];
@@ -1385,14 +1383,14 @@
     if (ConstantFP *Op1 = dyn_cast<ConstantFP>(Operands[0])) {
       if (!Ty->isFloatTy() && !Ty->isDoubleTy())
         return 0;
-      double Op1V = Ty->isFloatTy() ? 
+      double Op1V = Ty->isFloatTy() ?
                       (double)Op1->getValueAPF().convertToFloat() :
                       Op1->getValueAPF().convertToDouble();
       if (ConstantFP *Op2 = dyn_cast<ConstantFP>(Operands[1])) {
         if (Op2->getType() != Op1->getType())
           return 0;
 
-        double Op2V = Ty->isFloatTy() ? 
+        double Op2V = Ty->isFloatTy() ?
                       (double)Op2->getValueAPF().convertToFloat():
                       Op2->getValueAPF().convertToDouble();
 
@@ -1419,7 +1417,7 @@
       }
       return 0;
     }
-    
+
     if (ConstantInt *Op1 = dyn_cast<ConstantInt>(Operands[0])) {
       if (ConstantInt *Op2 = dyn_cast<ConstantInt>(Operands[1])) {
         switch (F->getIntrinsicID()) {
@@ -1469,7 +1467,7 @@
           return ConstantInt::get(Ty, Op1->getValue().countLeadingZeros());
         }
       }
-      
+
       return 0;
     }
     return 0;

Added: llvm/branches/R600/lib/Analysis/CostModel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/CostModel.cpp?rev=167838&view=auto
==============================================================================
--- llvm/branches/R600/lib/Analysis/CostModel.cpp (added)
+++ llvm/branches/R600/lib/Analysis/CostModel.cpp Tue Nov 13 09:21:47 2012
@@ -0,0 +1,193 @@
+//===- CostModel.cpp ------ Cost Model Analysis ---------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the cost model analysis. It provides a very basic cost
+// estimation for LLVM-IR. The cost result can be thought of as cycles, but it
+// is really unit-less. The estimated cost is ment to be used for comparing
+// alternatives.
+//
+//===----------------------------------------------------------------------===//
+
+#define CM_NAME "cost-model"
+#define DEBUG_TYPE CM_NAME
+#include "llvm/Analysis/Passes.h"
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/Pass.h"
+#include "llvm/TargetTransformInfo.h"
+#include "llvm/Value.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/raw_ostream.h"
+using namespace llvm;
+
+namespace {
+  class CostModelAnalysis : public FunctionPass {
+
+  public:
+    static char ID; // Class identification, replacement for typeinfo
+    CostModelAnalysis() : FunctionPass(ID), F(0), VTTI(0) {
+      initializeCostModelAnalysisPass(
+        *PassRegistry::getPassRegistry());
+    }
+
+    /// Returns the expected cost of the instruction.
+    /// Returns -1 if the cost is unknown.
+    /// Note, this method does not cache the cost calculation and it
+    /// can be expensive in some cases.
+    unsigned getInstructionCost(Instruction *I) const;
+
+  private:
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const;
+    virtual bool runOnFunction(Function &F);
+    virtual void print(raw_ostream &OS, const Module*) const;
+
+    /// The function that we analyze.
+    Function *F;
+    /// Vector target information.
+    const VectorTargetTransformInfo *VTTI;
+  };
+}  // End of anonymous namespace
+
+// Register this pass.
+char CostModelAnalysis::ID = 0;
+static const char cm_name[] = "Cost Model Analysis";
+INITIALIZE_PASS_BEGIN(CostModelAnalysis, CM_NAME, cm_name, false, true)
+INITIALIZE_PASS_END  (CostModelAnalysis, CM_NAME, cm_name, false, true)
+
+FunctionPass *llvm::createCostModelAnalysisPass() {
+  return new CostModelAnalysis();
+}
+
+void
+CostModelAnalysis::getAnalysisUsage(AnalysisUsage &AU) const {
+  AU.setPreservesAll();
+}
+
+bool
+CostModelAnalysis::runOnFunction(Function &F) {
+ this->F = &F;
+
+ // Target information.
+ TargetTransformInfo *TTI;
+ TTI = getAnalysisIfAvailable<TargetTransformInfo>();
+ if (TTI)
+   VTTI = TTI->getVectorTargetTransformInfo();
+
+ return false;
+}
+
+unsigned CostModelAnalysis::getInstructionCost(Instruction *I) const {
+  if (!VTTI)
+    return -1;
+
+  switch (I->getOpcode()) {
+  case Instruction::Ret:
+  case Instruction::PHI:
+  case Instruction::Br: {
+    return VTTI->getCFInstrCost(I->getOpcode());
+  }
+  case Instruction::Add:
+  case Instruction::FAdd:
+  case Instruction::Sub:
+  case Instruction::FSub:
+  case Instruction::Mul:
+  case Instruction::FMul:
+  case Instruction::UDiv:
+  case Instruction::SDiv:
+  case Instruction::FDiv:
+  case Instruction::URem:
+  case Instruction::SRem:
+  case Instruction::FRem:
+  case Instruction::Shl:
+  case Instruction::LShr:
+  case Instruction::AShr:
+  case Instruction::And:
+  case Instruction::Or:
+  case Instruction::Xor: {
+    return VTTI->getArithmeticInstrCost(I->getOpcode(), I->getType());
+  }
+  case Instruction::Select: {
+    SelectInst *SI = cast<SelectInst>(I);
+    Type *CondTy = SI->getCondition()->getType();
+    return VTTI->getCmpSelInstrCost(I->getOpcode(), I->getType(), CondTy);
+  }
+  case Instruction::ICmp:
+  case Instruction::FCmp: {
+    Type *ValTy = I->getOperand(0)->getType();
+    return VTTI->getCmpSelInstrCost(I->getOpcode(), ValTy);
+  }
+  case Instruction::Store: {
+    StoreInst *SI = cast<StoreInst>(I);
+    Type *ValTy = SI->getValueOperand()->getType();
+    return VTTI->getMemoryOpCost(I->getOpcode(), ValTy,
+                                 SI->getAlignment(),
+                                 SI->getPointerAddressSpace());
+  }
+  case Instruction::Load: {
+    LoadInst *LI = cast<LoadInst>(I);
+    return VTTI->getMemoryOpCost(I->getOpcode(), I->getType(),
+                                 LI->getAlignment(),
+                                 LI->getPointerAddressSpace());
+  }
+  case Instruction::ZExt:
+  case Instruction::SExt:
+  case Instruction::FPToUI:
+  case Instruction::FPToSI:
+  case Instruction::FPExt:
+  case Instruction::PtrToInt:
+  case Instruction::IntToPtr:
+  case Instruction::SIToFP:
+  case Instruction::UIToFP:
+  case Instruction::Trunc:
+  case Instruction::FPTrunc:
+  case Instruction::BitCast: {
+    Type *SrcTy = I->getOperand(0)->getType();
+    return VTTI->getCastInstrCost(I->getOpcode(), I->getType(), SrcTy);
+  }
+  case Instruction::ExtractElement: {
+    ExtractElementInst * EEI = cast<ExtractElementInst>(I);
+    ConstantInt *CI = dyn_cast<ConstantInt>(I->getOperand(1));
+    unsigned Idx = -1;
+    if (CI)
+      Idx = CI->getZExtValue();
+    return VTTI->getVectorInstrCost(I->getOpcode(),
+                                    EEI->getOperand(0)->getType(), Idx);
+  }
+  case Instruction::InsertElement: {
+      InsertElementInst * IE = cast<InsertElementInst>(I);
+      ConstantInt *CI = dyn_cast<ConstantInt>(IE->getOperand(2));
+      unsigned Idx = -1;
+      if (CI)
+        Idx = CI->getZExtValue();
+      return VTTI->getVectorInstrCost(I->getOpcode(),
+                                      IE->getType(), Idx);
+    }
+  default:
+    // We don't have any information on this instruction.
+    return -1;
+  }
+}
+
+void CostModelAnalysis::print(raw_ostream &OS, const Module*) const {
+  if (!F)
+    return;
+
+  for (Function::iterator B = F->begin(), BE = F->end(); B != BE; ++B) {
+    for (BasicBlock::iterator it = B->begin(), e = B->end(); it != e; ++it) {
+      Instruction *Inst = it;
+      unsigned Cost = getInstructionCost(Inst);
+      if (Cost != (unsigned)-1)
+        OS << "Cost Model: Found an estimated cost of " << Cost;
+      else
+        OS << "Cost Model: Unknown cost";
+
+      OS << " for instruction: "<< *Inst << "\n";
+    }
+  }
+}

Modified: llvm/branches/R600/lib/Analysis/DependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/DependenceAnalysis.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/DependenceAnalysis.cpp (original)
+++ llvm/branches/R600/lib/Analysis/DependenceAnalysis.cpp Tue Nov 13 09:21:47 2012
@@ -55,12 +55,16 @@
 
 #include "llvm/Analysis/DependenceAnalysis.h"
 #include "llvm/ADT/Statistic.h"
-#include "llvm/Instructions.h"
 #include "llvm/Operator.h"
+#include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
+#include "llvm/Analysis/ScalarEvolution.h"
+#include "llvm/Analysis/ScalarEvolutionExpressions.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/InstIterator.h"
+#include "llvm/Support/raw_ostream.h"
 
 using namespace llvm;
 
@@ -1769,7 +1773,7 @@
 // where i and j are induction variable, c1 and c2 are loop invariant,
 // and a and b are constants.
 // Returns true if any possible dependence is disproved.
-// Marks the result as inconsistant.
+// Marks the result as inconsistent.
 // Works in some cases that symbolicRDIVtest doesn't, and vice versa.
 bool DependenceAnalysis::exactRDIVtest(const SCEV *SrcCoeff,
                                        const SCEV *DstCoeff,
@@ -2198,7 +2202,7 @@
 // gcdMIVtest -
 // Tests an MIV subscript pair for dependence.
 // Returns true if any possible dependence is disproved.
-// Marks the result as inconsistant.
+// Marks the result as inconsistent.
 // Can sometimes disprove the equal direction for 1 or more loops,
 // as discussed in Michael Wolfe's book,
 // High Performance Compilers for Parallel Computing, page 235.
@@ -2274,11 +2278,12 @@
         assert(!Constant && "Surprised to find multiple constants");
         Constant = cast<SCEVConstant>(Operand);
       }
-      else if (isa<SCEVMulExpr>(Operand)) {
+      else if (const SCEVMulExpr *Product = dyn_cast<SCEVMulExpr>(Operand)) {
         // Search for constant operand to participate in GCD;
         // If none found; return false.
-        const SCEVConstant *ConstOp =
-          getConstantPart(cast<SCEVMulExpr>(Operand));
+        const SCEVConstant *ConstOp = getConstantPart(Product);
+        if (!ConstOp)
+          return false;
         APInt ConstOpValue = ConstOp->getValue()->getValue();
         ExtraGCD = APIntOps::GreatestCommonDivisor(ExtraGCD,
                                                    ConstOpValue.abs());

Modified: llvm/branches/R600/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/InlineCost.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/InlineCost.cpp (original)
+++ llvm/branches/R600/lib/Analysis/InlineCost.cpp Tue Nov 13 09:21:47 2012
@@ -243,8 +243,7 @@
   if (!TD)
     return false;
 
-  unsigned AS = GEP.getPointerAddressSpace();
-  unsigned IntPtrWidth = TD->getPointerSizeInBits(AS);
+  unsigned IntPtrWidth = TD->getPointerSizeInBits();
   assert(IntPtrWidth == Offset.getBitWidth());
 
   for (gep_type_iterator GTI = gep_type_begin(GEP), GTE = gep_type_end(GEP);
@@ -392,8 +391,7 @@
   // Track base/offset pairs when converted to a plain integer provided the
   // integer is large enough to represent the pointer.
   unsigned IntegerSize = I.getType()->getScalarSizeInBits();
-  unsigned AS = I.getPointerAddressSpace();
-  if (TD && IntegerSize >= TD->getPointerSizeInBits(AS)) {
+  if (TD && IntegerSize >= TD->getPointerSizeInBits()) {
     std::pair<Value *, APInt> BaseAndOffset
       = ConstantOffsetPtrs.lookup(I.getOperand(0));
     if (BaseAndOffset.first)
@@ -427,8 +425,7 @@
   // modifications provided the integer is not too large.
   Value *Op = I.getOperand(0);
   unsigned IntegerSize = Op->getType()->getScalarSizeInBits();
-  unsigned AS = I.getAddressSpace();
-  if (TD && IntegerSize <= TD->getPointerSizeInBits(AS)) {
+  if (TD && IntegerSize <= TD->getPointerSizeInBits()) {
     std::pair<Value *, APInt> BaseAndOffset = ConstantOffsetPtrs.lookup(Op);
     if (BaseAndOffset.first)
       ConstantOffsetPtrs[&I] = BaseAndOffset;
@@ -763,8 +760,7 @@
   if (!TD || !V->getType()->isPointerTy())
     return 0;
 
-  unsigned AS = cast<PointerType>(V->getType())->getAddressSpace();;
-  unsigned IntPtrWidth = TD->getPointerSizeInBits(AS);
+  unsigned IntPtrWidth = TD->getPointerSizeInBits();
   APInt Offset = APInt::getNullValue(IntPtrWidth);
 
   // Even though we don't look through PHI nodes, we could be called on an
@@ -828,8 +824,7 @@
         // size of the byval type by the target's pointer size.
         PointerType *PTy = cast<PointerType>(CS.getArgument(I)->getType());
         unsigned TypeSize = TD->getTypeSizeInBits(PTy->getElementType());
-        unsigned AS = PTy->getAddressSpace();
-        unsigned PointerSize = TD->getPointerSizeInBits(AS);
+        unsigned PointerSize = TD->getPointerSizeInBits();
         // Ceiling division.
         unsigned NumStores = (TypeSize + PointerSize - 1) / PointerSize;
 

Modified: llvm/branches/R600/lib/Analysis/InstructionSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/InstructionSimplify.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/InstructionSimplify.cpp (original)
+++ llvm/branches/R600/lib/Analysis/InstructionSimplify.cpp Tue Nov 13 09:21:47 2012
@@ -666,8 +666,7 @@
 /// 'Offset' APInt must be the bitwidth of the target's pointer size.
 static bool accumulateGEPOffset(const DataLayout &TD, GEPOperator *GEP,
                                 APInt &Offset) {
-  unsigned AS = GEP->getPointerAddressSpace();
-  unsigned IntPtrWidth = TD.getPointerSizeInBits(AS);
+  unsigned IntPtrWidth = TD.getPointerSizeInBits();
   assert(IntPtrWidth == Offset.getBitWidth());
 
   gep_type_iterator GTI = gep_type_begin(GEP);
@@ -697,14 +696,12 @@
 /// accumulates the total constant offset applied in the returned constant. It
 /// returns 0 if V is not a pointer, and returns the constant '0' if there are
 /// no constant offsets applied.
-/// FIXME: This function also exists in InlineCost.cpp.
 static Constant *stripAndComputeConstantOffsets(const DataLayout &TD,
                                                 Value *&V) {
   if (!V->getType()->isPointerTy())
     return 0;
 
-  unsigned AS = cast<PointerType>(V->getType())->getAddressSpace();;
-  unsigned IntPtrWidth = TD.getPointerSizeInBits(AS);
+  unsigned IntPtrWidth = TD.getPointerSizeInBits();
   APInt Offset = APInt::getNullValue(IntPtrWidth);
 
   // Even though we don't look through PHI nodes, we could be called on an
@@ -1880,9 +1877,7 @@
     // Turn icmp (ptrtoint x), (ptrtoint/constant) into a compare of the input
     // if the integer type is the same size as the pointer type.
     if (MaxRecurse && Q.TD && isa<PtrToIntInst>(LI) &&
-        Q.TD->getPointerSizeInBits(
-          cast<PtrToIntInst>(LI)->getPointerAddressSpace()) ==
-        DstTy->getPrimitiveSizeInBits()) {
+        Q.TD->getPointerSizeInBits() == DstTy->getPrimitiveSizeInBits()) {
       if (Constant *RHSC = dyn_cast<Constant>(RHS)) {
         // Transfer the cast to the constant.
         if (Value *V = SimplifyICmpInst(Pred, SrcOp,

Modified: llvm/branches/R600/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/LazyValueInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/branches/R600/lib/Analysis/LazyValueInfo.cpp Tue Nov 13 09:21:47 2012
@@ -13,6 +13,7 @@
 //===----------------------------------------------------------------------===//
 
 #define DEBUG_TYPE "lazy-value-info"
+#include "llvm/Analysis/AliasAnalysis.h"
 #include "llvm/Analysis/LazyValueInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/Constants.h"
@@ -294,7 +295,7 @@
 //===----------------------------------------------------------------------===//
 
 namespace {
-  /// LVIValueHandle - A callback value handle update the cache when
+  /// LVIValueHandle - A callback value handle updates the cache when
   /// values are erased.
   class LazyValueInfoCache;
   struct LVIValueHandle : public CallbackVH {
@@ -557,13 +558,11 @@
 static bool InstructionDereferencesPointer(Instruction *I, Value *Ptr) {
   if (LoadInst *L = dyn_cast<LoadInst>(I)) {
     return L->getPointerAddressSpace() == 0 &&
-        GetUnderlyingObject(L->getPointerOperand()) ==
-        GetUnderlyingObject(Ptr);
+        GetUnderlyingObject(L->getPointerOperand()) == Ptr;
   }
   if (StoreInst *S = dyn_cast<StoreInst>(I)) {
     return S->getPointerAddressSpace() == 0 &&
-        GetUnderlyingObject(S->getPointerOperand()) ==
-        GetUnderlyingObject(Ptr);
+        GetUnderlyingObject(S->getPointerOperand()) == Ptr;
   }
   if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(I)) {
     if (MI->isVolatile()) return false;
@@ -573,11 +572,11 @@
     if (!Len || Len->isZero()) return false;
 
     if (MI->getDestAddressSpace() == 0)
-      if (MI->getRawDest() == Ptr || MI->getDest() == Ptr)
+      if (GetUnderlyingObject(MI->getRawDest()) == Ptr)
         return true;
     if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(MI))
       if (MTI->getSourceAddressSpace() == 0)
-        if (MTI->getRawSource() == Ptr || MTI->getSource() == Ptr)
+        if (GetUnderlyingObject(MTI->getRawSource()) == Ptr)
           return true;
   }
   return false;
@@ -591,13 +590,19 @@
   // then we know that the pointer can't be NULL.
   bool NotNull = false;
   if (Val->getType()->isPointerTy()) {
-    if (isa<AllocaInst>(Val)) {
+    if (isKnownNonNull(Val)) {
       NotNull = true;
     } else {
-      for (BasicBlock::iterator BI = BB->begin(), BE = BB->end();BI != BE;++BI){
-        if (InstructionDereferencesPointer(BI, Val)) {
-          NotNull = true;
-          break;
+      Value *UnderlyingVal = GetUnderlyingObject(Val);
+      // If 'GetUnderlyingObject' didn't converge, skip it. It won't converge
+      // inside InstructionDereferencesPointer either.
+      if (UnderlyingVal == GetUnderlyingObject(UnderlyingVal, NULL, 1)) {
+        for (BasicBlock::iterator BI = BB->begin(), BE = BB->end();
+             BI != BE; ++BI) {
+          if (InstructionDereferencesPointer(BI, UnderlyingVal)) {
+            NotNull = true;
+            break;
+          }
         }
       }
     }

Removed: llvm/branches/R600/lib/Analysis/LoopDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/LoopDependenceAnalysis.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Analysis/LoopDependenceAnalysis.cpp (original)
+++ llvm/branches/R600/lib/Analysis/LoopDependenceAnalysis.cpp (removed)
@@ -1,362 +0,0 @@
-//===- LoopDependenceAnalysis.cpp - LDA Implementation ----------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This is the (beginning) of an implementation of a loop dependence analysis
-// framework, which is used to detect dependences in memory accesses in loops.
-//
-// Please note that this is work in progress and the interface is subject to
-// change.
-//
-// TODO: adapt as implementation progresses.
-//
-// TODO: document lingo (pair, subscript, index)
-//
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "lda"
-#include "llvm/ADT/DenseSet.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/LoopDependenceAnalysis.h"
-#include "llvm/Analysis/LoopPass.h"
-#include "llvm/Analysis/ScalarEvolution.h"
-#include "llvm/Analysis/ScalarEvolutionExpressions.h"
-#include "llvm/Analysis/ValueTracking.h"
-#include "llvm/Assembly/Writer.h"
-#include "llvm/Instructions.h"
-#include "llvm/Operator.h"
-#include "llvm/Support/Allocator.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/raw_ostream.h"
-#include "llvm/DataLayout.h"
-using namespace llvm;
-
-STATISTIC(NumAnswered,    "Number of dependence queries answered");
-STATISTIC(NumAnalysed,    "Number of distinct dependence pairs analysed");
-STATISTIC(NumDependent,   "Number of pairs with dependent accesses");
-STATISTIC(NumIndependent, "Number of pairs with independent accesses");
-STATISTIC(NumUnknown,     "Number of pairs with unknown accesses");
-
-LoopPass *llvm::createLoopDependenceAnalysisPass() {
-  return new LoopDependenceAnalysis();
-}
-
-INITIALIZE_PASS_BEGIN(LoopDependenceAnalysis, "lda",
-                "Loop Dependence Analysis", false, true)
-INITIALIZE_PASS_DEPENDENCY(ScalarEvolution)
-INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
-INITIALIZE_PASS_END(LoopDependenceAnalysis, "lda",
-                "Loop Dependence Analysis", false, true)
-char LoopDependenceAnalysis::ID = 0;
-
-//===----------------------------------------------------------------------===//
-//                             Utility Functions
-//===----------------------------------------------------------------------===//
-
-static inline bool IsMemRefInstr(const Value *V) {
-  const Instruction *I = dyn_cast<const Instruction>(V);
-  return I && (I->mayReadFromMemory() || I->mayWriteToMemory());
-}
-
-static void GetMemRefInstrs(const Loop *L,
-                            SmallVectorImpl<Instruction*> &Memrefs) {
-  for (Loop::block_iterator b = L->block_begin(), be = L->block_end();
-       b != be; ++b)
-    for (BasicBlock::iterator i = (*b)->begin(), ie = (*b)->end();
-         i != ie; ++i)
-      if (IsMemRefInstr(i))
-        Memrefs.push_back(i);
-}
-
-static bool IsLoadOrStoreInst(Value *I) {
-  // Returns true if the load or store can be analyzed. Atomic and volatile
-  // operations have properties which this analysis does not understand.
-  if (LoadInst *LI = dyn_cast<LoadInst>(I))
-    return LI->isUnordered();
-  else if (StoreInst *SI = dyn_cast<StoreInst>(I))
-    return SI->isUnordered();
-  return false;
-}
-
-static Value *GetPointerOperand(Value *I) {
-  if (LoadInst *i = dyn_cast<LoadInst>(I))
-    return i->getPointerOperand();
-  if (StoreInst *i = dyn_cast<StoreInst>(I))
-    return i->getPointerOperand();
-  llvm_unreachable("Value is no load or store instruction!");
-}
-
-static AliasAnalysis::AliasResult UnderlyingObjectsAlias(AliasAnalysis *AA,
-                                                         const Value *A,
-                                                         const Value *B) {
-  const Value *aObj = GetUnderlyingObject(A);
-  const Value *bObj = GetUnderlyingObject(B);
-  return AA->alias(aObj, AA->getTypeStoreSize(aObj->getType()),
-                   bObj, AA->getTypeStoreSize(bObj->getType()));
-}
-
-static inline const SCEV *GetZeroSCEV(ScalarEvolution *SE) {
-  return SE->getConstant(Type::getInt32Ty(SE->getContext()), 0L);
-}
-
-//===----------------------------------------------------------------------===//
-//                             Dependence Testing
-//===----------------------------------------------------------------------===//
-
-bool LoopDependenceAnalysis::isDependencePair(const Value *A,
-                                              const Value *B) const {
-  return IsMemRefInstr(A) &&
-         IsMemRefInstr(B) &&
-         (cast<const Instruction>(A)->mayWriteToMemory() ||
-          cast<const Instruction>(B)->mayWriteToMemory());
-}
-
-bool LoopDependenceAnalysis::findOrInsertDependencePair(Value *A,
-                                                        Value *B,
-                                                        DependencePair *&P) {
-  void *insertPos = 0;
-  FoldingSetNodeID id;
-  id.AddPointer(A);
-  id.AddPointer(B);
-
-  P = Pairs.FindNodeOrInsertPos(id, insertPos);
-  if (P) return true;
-
-  P = new (PairAllocator) DependencePair(id, A, B);
-  Pairs.InsertNode(P, insertPos);
-  return false;
-}
-
-void LoopDependenceAnalysis::getLoops(const SCEV *S,
-                                      DenseSet<const Loop*>* Loops) const {
-  // Refactor this into an SCEVVisitor, if efficiency becomes a concern.
-  for (const Loop *L = this->L; L != 0; L = L->getParentLoop())
-    if (!SE->isLoopInvariant(S, L))
-      Loops->insert(L);
-}
-
-bool LoopDependenceAnalysis::isLoopInvariant(const SCEV *S) const {
-  DenseSet<const Loop*> loops;
-  getLoops(S, &loops);
-  return loops.empty();
-}
-
-bool LoopDependenceAnalysis::isAffine(const SCEV *S) const {
-  const SCEVAddRecExpr *rec = dyn_cast<SCEVAddRecExpr>(S);
-  return isLoopInvariant(S) || (rec && rec->isAffine());
-}
-
-bool LoopDependenceAnalysis::isZIVPair(const SCEV *A, const SCEV *B) const {
-  return isLoopInvariant(A) && isLoopInvariant(B);
-}
-
-bool LoopDependenceAnalysis::isSIVPair(const SCEV *A, const SCEV *B) const {
-  DenseSet<const Loop*> loops;
-  getLoops(A, &loops);
-  getLoops(B, &loops);
-  return loops.size() == 1;
-}
-
-LoopDependenceAnalysis::DependenceResult
-LoopDependenceAnalysis::analyseZIV(const SCEV *A,
-                                   const SCEV *B,
-                                   Subscript *S) const {
-  assert(isZIVPair(A, B) && "Attempted to ZIV-test non-ZIV SCEVs!");
-  return A == B ? Dependent : Independent;
-}
-
-LoopDependenceAnalysis::DependenceResult
-LoopDependenceAnalysis::analyseSIV(const SCEV *A,
-                                   const SCEV *B,
-                                   Subscript *S) const {
-  return Unknown; // TODO: Implement.
-}
-
-LoopDependenceAnalysis::DependenceResult
-LoopDependenceAnalysis::analyseMIV(const SCEV *A,
-                                   const SCEV *B,
-                                   Subscript *S) const {
-  return Unknown; // TODO: Implement.
-}
-
-LoopDependenceAnalysis::DependenceResult
-LoopDependenceAnalysis::analyseSubscript(const SCEV *A,
-                                         const SCEV *B,
-                                         Subscript *S) const {
-  DEBUG(dbgs() << "  Testing subscript: " << *A << ", " << *B << "\n");
-
-  if (A == B) {
-    DEBUG(dbgs() << "  -> [D] same SCEV\n");
-    return Dependent;
-  }
-
-  if (!isAffine(A) || !isAffine(B)) {
-    DEBUG(dbgs() << "  -> [?] not affine\n");
-    return Unknown;
-  }
-
-  if (isZIVPair(A, B))
-    return analyseZIV(A, B, S);
-
-  if (isSIVPair(A, B))
-    return analyseSIV(A, B, S);
-
-  return analyseMIV(A, B, S);
-}
-
-LoopDependenceAnalysis::DependenceResult
-LoopDependenceAnalysis::analysePair(DependencePair *P) const {
-  DEBUG(dbgs() << "Analysing:\n" << *P->A << "\n" << *P->B << "\n");
-
-  // We only analyse loads and stores but no possible memory accesses by e.g.
-  // free, call, or invoke instructions.
-  if (!IsLoadOrStoreInst(P->A) || !IsLoadOrStoreInst(P->B)) {
-    DEBUG(dbgs() << "--> [?] no load/store\n");
-    return Unknown;
-  }
-
-  Value *aPtr = GetPointerOperand(P->A);
-  Value *bPtr = GetPointerOperand(P->B);
-
-  switch (UnderlyingObjectsAlias(AA, aPtr, bPtr)) {
-  case AliasAnalysis::MayAlias:
-  case AliasAnalysis::PartialAlias:
-    // We can not analyse objects if we do not know about their aliasing.
-    DEBUG(dbgs() << "---> [?] may alias\n");
-    return Unknown;
-
-  case AliasAnalysis::NoAlias:
-    // If the objects noalias, they are distinct, accesses are independent.
-    DEBUG(dbgs() << "---> [I] no alias\n");
-    return Independent;
-
-  case AliasAnalysis::MustAlias:
-    break; // The underlying objects alias, test accesses for dependence.
-  }
-
-  const GEPOperator *aGEP = dyn_cast<GEPOperator>(aPtr);
-  const GEPOperator *bGEP = dyn_cast<GEPOperator>(bPtr);
-
-  if (!aGEP || !bGEP)
-    return Unknown;
-
-  // FIXME: Is filtering coupled subscripts necessary?
-
-  // Collect GEP operand pairs (FIXME: use GetGEPOperands from BasicAA), adding
-  // trailing zeroes to the smaller GEP, if needed.
-  typedef SmallVector<std::pair<const SCEV*, const SCEV*>, 4> GEPOpdPairsTy;
-  GEPOpdPairsTy opds;
-  for(GEPOperator::const_op_iterator aIdx = aGEP->idx_begin(),
-                                     aEnd = aGEP->idx_end(),
-                                     bIdx = bGEP->idx_begin(),
-                                     bEnd = bGEP->idx_end();
-      aIdx != aEnd && bIdx != bEnd;
-      aIdx += (aIdx != aEnd), bIdx += (bIdx != bEnd)) {
-    const SCEV* aSCEV = (aIdx != aEnd) ? SE->getSCEV(*aIdx) : GetZeroSCEV(SE);
-    const SCEV* bSCEV = (bIdx != bEnd) ? SE->getSCEV(*bIdx) : GetZeroSCEV(SE);
-    opds.push_back(std::make_pair(aSCEV, bSCEV));
-  }
-
-  if (!opds.empty() && opds[0].first != opds[0].second) {
-    // We cannot (yet) handle arbitrary GEP pointer offsets. By limiting
-    //
-    // TODO: this could be relaxed by adding the size of the underlying object
-    // to the first subscript. If we have e.g. (GEP x,0,i; GEP x,2,-i) and we
-    // know that x is a [100 x i8]*, we could modify the first subscript to be
-    // (i, 200-i) instead of (i, -i).
-    return Unknown;
-  }
-
-  // Now analyse the collected operand pairs (skipping the GEP ptr offsets).
-  for (GEPOpdPairsTy::const_iterator i = opds.begin() + 1, end = opds.end();
-       i != end; ++i) {
-    Subscript subscript;
-    DependenceResult result = analyseSubscript(i->first, i->second, &subscript);
-    if (result != Dependent) {
-      // We either proved independence or failed to analyse this subscript.
-      // Further subscripts will not improve the situation, so abort early.
-      return result;
-    }
-    P->Subscripts.push_back(subscript);
-  }
-  // We successfully analysed all subscripts but failed to prove independence.
-  return Dependent;
-}
-
-bool LoopDependenceAnalysis::depends(Value *A, Value *B) {
-  assert(isDependencePair(A, B) && "Values form no dependence pair!");
-  ++NumAnswered;
-
-  DependencePair *p;
-  if (!findOrInsertDependencePair(A, B, p)) {
-    // The pair is not cached, so analyse it.
-    ++NumAnalysed;
-    switch (p->Result = analysePair(p)) {
-    case Dependent:   ++NumDependent;   break;
-    case Independent: ++NumIndependent; break;
-    case Unknown:     ++NumUnknown;     break;
-    }
-  }
-  return p->Result != Independent;
-}
-
-//===----------------------------------------------------------------------===//
-//                   LoopDependenceAnalysis Implementation
-//===----------------------------------------------------------------------===//
-
-bool LoopDependenceAnalysis::runOnLoop(Loop *L, LPPassManager &) {
-  this->L = L;
-  AA = &getAnalysis<AliasAnalysis>();
-  SE = &getAnalysis<ScalarEvolution>();
-  return false;
-}
-
-void LoopDependenceAnalysis::releaseMemory() {
-  Pairs.clear();
-  PairAllocator.Reset();
-}
-
-void LoopDependenceAnalysis::getAnalysisUsage(AnalysisUsage &AU) const {
-  AU.setPreservesAll();
-  AU.addRequiredTransitive<AliasAnalysis>();
-  AU.addRequiredTransitive<ScalarEvolution>();
-}
-
-static void PrintLoopInfo(raw_ostream &OS,
-                          LoopDependenceAnalysis *LDA, const Loop *L) {
-  if (!L->empty()) return; // ignore non-innermost loops
-
-  SmallVector<Instruction*, 8> memrefs;
-  GetMemRefInstrs(L, memrefs);
-
-  OS << "Loop at depth " << L->getLoopDepth() << ", header block: ";
-  WriteAsOperand(OS, L->getHeader(), false);
-  OS << "\n";
-
-  OS << "  Load/store instructions: " << memrefs.size() << "\n";
-  for (SmallVector<Instruction*, 8>::const_iterator x = memrefs.begin(),
-       end = memrefs.end(); x != end; ++x)
-    OS << "\t" << (x - memrefs.begin()) << ": " << **x << "\n";
-
-  OS << "  Pairwise dependence results:\n";
-  for (SmallVector<Instruction*, 8>::const_iterator x = memrefs.begin(),
-       end = memrefs.end(); x != end; ++x)
-    for (SmallVector<Instruction*, 8>::const_iterator y = x + 1;
-         y != end; ++y)
-      if (LDA->isDependencePair(*x, *y))
-        OS << "\t" << (x - memrefs.begin()) << "," << (y - memrefs.begin())
-           << ": " << (LDA->depends(*x, *y) ? "dependent" : "independent")
-           << "\n";
-}
-
-void LoopDependenceAnalysis::print(raw_ostream &OS, const Module*) const {
-  // TODO: doc why const_cast is safe
-  PrintLoopInfo(OS, const_cast<LoopDependenceAnalysis*>(this), this->L);
-}

Modified: llvm/branches/R600/lib/Analysis/MemoryDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/MemoryDependenceAnalysis.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/MemoryDependenceAnalysis.cpp (original)
+++ llvm/branches/R600/lib/Analysis/MemoryDependenceAnalysis.cpp Tue Nov 13 09:21:47 2012
@@ -983,7 +983,7 @@
     for (NonLocalDepInfo::iterator I = Cache->begin(), E = Cache->end();
          I != E; ++I) {
       Visited.insert(std::make_pair(I->getBB(), Addr));
-      if (!I->getResult().isNonLocal())
+      if (!I->getResult().isNonLocal() && DT->isReachableFromEntry(I->getBB()))
         Result.push_back(NonLocalDepResult(I->getBB(), I->getResult(), Addr));
     }
     ++NumCacheCompleteNonLocalPtr;
@@ -1029,7 +1029,7 @@
                                                  NumSortedEntries);
       
       // If we got a Def or Clobber, add this to the list of results.
-      if (!Dep.isNonLocal()) {
+      if (!Dep.isNonLocal() && DT->isReachableFromEntry(BB)) {
         Result.push_back(NonLocalDepResult(BB, Dep, Pointer.getAddr()));
         continue;
       }

Modified: llvm/branches/R600/lib/Analysis/ProfileDataLoader.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/ProfileDataLoader.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/ProfileDataLoader.cpp (original)
+++ llvm/branches/R600/lib/Analysis/ProfileDataLoader.cpp Tue Nov 13 09:21:47 2012
@@ -51,13 +51,7 @@
   if (A == ProfileDataLoader::Uncounted) return B;
   if (B == ProfileDataLoader::Uncounted) return A;
 
-  // Saturate to the maximum storable value.  This could change taken/nottaken
-  // ratios, but is presumably better than wrapping and thus potentially
-  // inverting ratios.
-  uint64_t tmp = (uint64_t)A + (uint64_t)B;
-  if (tmp > (uint64_t)ProfileDataLoader::MaxCount)
-    tmp = ProfileDataLoader::MaxCount;
-  return (unsigned)tmp;
+  return A + B;
 }
 
 /// ReadProfilingData - Load 'NumEntries' items of type 'T' from file 'F'
@@ -120,7 +114,6 @@
 }
 
 const unsigned ProfileDataLoader::Uncounted = ~0U;
-const unsigned ProfileDataLoader::MaxCount = ~0U - 1U;
 
 /// ProfileDataLoader ctor - Read the specified profiling data file, reporting
 /// a fatal error if the file is invalid or broken.

Modified: llvm/branches/R600/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/ScalarEvolution.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/branches/R600/lib/Analysis/ScalarEvolution.cpp Tue Nov 13 09:21:47 2012
@@ -105,6 +105,11 @@
                                  "derived loop"),
                         cl::init(100));
 
+// FIXME: Enable this with XDEBUG when the test suite is clean.
+static cl::opt<bool>
+VerifySCEV("verify-scev",
+           cl::desc("Verify ScalarEvolution's backedge taken counts (slow)"));
+
 INITIALIZE_PASS_BEGIN(ScalarEvolution, "scalar-evolution",
                 "Scalar Evolution Analysis", false, true)
 INITIALIZE_PASS_DEPENDENCY(LoopInfo)
@@ -3980,8 +3985,11 @@
 
   ConstantInt *Result = MulC->getValue();
 
-  // Guard against huge trip counts.
-  if (!Result || Result->getValue().getActiveBits() > 32)
+  // Guard against huge trip counts (this requires checking
+  // for zero to handle the case where the trip count == -1 and the
+  // addition wraps).
+  if (!Result || Result->getValue().getActiveBits() > 32 ||
+      Result->getValue().getActiveBits() == 0)
     return 1;
 
   return (unsigned)Result->getZExtValue();
@@ -6932,3 +6940,87 @@
   UnsignedRanges.erase(S);
   SignedRanges.erase(S);
 }
+
+typedef DenseMap<const Loop *, std::string> VerifyMap;
+
+/// replaceSubString - Replaces all occurences of From in Str with To.
+static void replaceSubString(std::string &Str, StringRef From, StringRef To) {
+  size_t Pos = 0;
+  while ((Pos = Str.find(From, Pos)) != std::string::npos) {
+    Str.replace(Pos, From.size(), To.data(), To.size());
+    Pos += To.size();
+  }
+}
+
+/// getLoopBackedgeTakenCounts - Helper method for verifyAnalysis.
+static void
+getLoopBackedgeTakenCounts(Loop *L, VerifyMap &Map, ScalarEvolution &SE) {
+  for (Loop::reverse_iterator I = L->rbegin(), E = L->rend(); I != E; ++I) {
+    getLoopBackedgeTakenCounts(*I, Map, SE); // recurse.
+
+    std::string &S = Map[L];
+    if (S.empty()) {
+      raw_string_ostream OS(S);
+      SE.getBackedgeTakenCount(L)->print(OS);
+
+      // false and 0 are semantically equivalent. This can happen in dead loops.
+      replaceSubString(OS.str(), "false", "0");
+      // Remove wrap flags, their use in SCEV is highly fragile.
+      // FIXME: Remove this when SCEV gets smarter about them.
+      replaceSubString(OS.str(), "<nw>", "");
+      replaceSubString(OS.str(), "<nsw>", "");
+      replaceSubString(OS.str(), "<nuw>", "");
+    }
+  }
+}
+
+void ScalarEvolution::verifyAnalysis() const {
+  if (!VerifySCEV)
+    return;
+
+  ScalarEvolution &SE = *const_cast<ScalarEvolution *>(this);
+
+  // Gather stringified backedge taken counts for all loops using SCEV's caches.
+  // FIXME: It would be much better to store actual values instead of strings,
+  //        but SCEV pointers will change if we drop the caches.
+  VerifyMap BackedgeDumpsOld, BackedgeDumpsNew;
+  for (LoopInfo::reverse_iterator I = LI->rbegin(), E = LI->rend(); I != E; ++I)
+    getLoopBackedgeTakenCounts(*I, BackedgeDumpsOld, SE);
+
+  // Gather stringified backedge taken counts for all loops without using
+  // SCEV's caches.
+  SE.releaseMemory();
+  for (LoopInfo::reverse_iterator I = LI->rbegin(), E = LI->rend(); I != E; ++I)
+    getLoopBackedgeTakenCounts(*I, BackedgeDumpsNew, SE);
+
+  // Now compare whether they're the same with and without caches. This allows
+  // verifying that no pass changed the cache.
+  assert(BackedgeDumpsOld.size() == BackedgeDumpsNew.size() &&
+         "New loops suddenly appeared!");
+
+  for (VerifyMap::iterator OldI = BackedgeDumpsOld.begin(),
+                           OldE = BackedgeDumpsOld.end(),
+                           NewI = BackedgeDumpsNew.begin();
+       OldI != OldE; ++OldI, ++NewI) {
+    assert(OldI->first == NewI->first && "Loop order changed!");
+
+    // Compare the stringified SCEVs. We don't care if undef backedgetaken count
+    // changes.
+    // FIXME: We currently ignore SCEV changes from/to CouldNotCompute. This
+    // means that a pass is buggy or SCEV has to learn a new pattern but is
+    // usually not harmful.
+    if (OldI->second != NewI->second &&
+        OldI->second.find("undef") == std::string::npos &&
+        NewI->second.find("undef") == std::string::npos &&
+        OldI->second != "***COULDNOTCOMPUTE***" &&
+        NewI->second != "***COULDNOTCOMPUTE***") {
+      dbgs() << "SCEVValidator: SCEV for loop '"
+             << OldI->first->getHeader()->getName()
+             << "' changed from '" << OldI->second
+             << "' to '" << NewI->second << "'!\n";
+      std::abort();
+    }
+  }
+
+  // TODO: Verify more things.
+}

Modified: llvm/branches/R600/lib/Analysis/ValueTracking.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Analysis/ValueTracking.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Analysis/ValueTracking.cpp (original)
+++ llvm/branches/R600/lib/Analysis/ValueTracking.cpp Tue Nov 13 09:21:47 2012
@@ -40,8 +40,7 @@
   if (unsigned BitWidth = Ty->getScalarSizeInBits())
     return BitWidth;
   assert(isa<PointerType>(Ty) && "Expected a pointer type!");
-  return TD ?
-    TD->getPointerSizeInBits(cast<PointerType>(Ty)->getAddressSpace()) : 0;
+  return TD ? TD->getPointerSizeInBits() : 0;
 }
 
 static void ComputeMaskedBitsAddSub(bool Add, Value *Op0, Value *Op1, bool NSW,
@@ -430,15 +429,13 @@
   case Instruction::ZExt:
   case Instruction::Trunc: {
     Type *SrcTy = I->getOperand(0)->getType();
-    
+
     unsigned SrcBitWidth;
     // Note that we handle pointer operands here because of inttoptr/ptrtoint
     // which fall through here.
-    if (SrcTy->isPointerTy())
-      SrcBitWidth = TD->getTypeSizeInBits(SrcTy);
-    else
-      SrcBitWidth = SrcTy->getScalarSizeInBits();
-    
+    SrcBitWidth = TD->getTypeSizeInBits(SrcTy->getScalarType());
+
+    assert(SrcBitWidth && "SrcBitWidth can't be zero");
     KnownZero = KnownZero.zextOrTrunc(SrcBitWidth);
     KnownOne = KnownOne.zextOrTrunc(SrcBitWidth);
     ComputeMaskedBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1);
@@ -1622,8 +1619,7 @@
   
   // Re-sign extend from the pointer size if needed to get overflow edge cases
   // right.
-  unsigned AS = GEP->getPointerAddressSpace();
-  unsigned PtrSize = TD.getPointerSizeInBits(AS);
+  unsigned PtrSize = TD.getPointerSizeInBits();
   if (PtrSize < 64)
     Offset = SignExtend64(Offset, PtrSize);
   

Modified: llvm/branches/R600/lib/AsmParser/LLLexer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/AsmParser/LLLexer.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/AsmParser/LLLexer.cpp (original)
+++ llvm/branches/R600/lib/AsmParser/LLLexer.cpp Tue Nov 13 09:21:47 2012
@@ -527,6 +527,7 @@
   KEYWORD(ptx_device);
   KEYWORD(spir_kernel);
   KEYWORD(spir_func);
+  KEYWORD(intel_ocl_bicc);
 
   KEYWORD(cc);
   KEYWORD(c);
@@ -557,6 +558,7 @@
   KEYWORD(naked);
   KEYWORD(nonlazybind);
   KEYWORD(address_safety);
+  KEYWORD(minsize);
 
   KEYWORD(type);
   KEYWORD(opaque);

Modified: llvm/branches/R600/lib/AsmParser/LLParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/AsmParser/LLParser.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/AsmParser/LLParser.cpp (original)
+++ llvm/branches/R600/lib/AsmParser/LLParser.cpp Tue Nov 13 09:21:47 2012
@@ -953,6 +953,7 @@
     case lltok::kw_naked:           B.addAttribute(Attributes::Naked); break;
     case lltok::kw_nonlazybind:     B.addAttribute(Attributes::NonLazyBind); break;
     case lltok::kw_address_safety:  B.addAttribute(Attributes::AddressSafety); break;
+    case lltok::kw_minsize:         B.addAttribute(Attributes::MinSize); break;
 
     case lltok::kw_alignstack: {
       unsigned Alignment;
@@ -1011,6 +1012,7 @@
     case lltok::kw_nonlazybind:
     case lltok::kw_returns_twice:
     case lltok::kw_address_safety:
+    case lltok::kw_minsize:
       if (AttrKind != 2)
         HaveError |= Error(AttrLoc, "invalid use of function-only attribute");
       break;
@@ -1092,6 +1094,7 @@
 ///   ::= /*empty*/
 ///   ::= 'ccc'
 ///   ::= 'fastcc'
+///   ::= 'kw_intel_ocl_bicc'
 ///   ::= 'coldcc'
 ///   ::= 'x86_stdcallcc'
 ///   ::= 'x86_fastcallcc'
@@ -1123,6 +1126,7 @@
   case lltok::kw_ptx_device:     CC = CallingConv::PTX_Device; break;
   case lltok::kw_spir_kernel:    CC = CallingConv::SPIR_KERNEL; break;
   case lltok::kw_spir_func:      CC = CallingConv::SPIR_FUNC; break;
+  case lltok::kw_intel_ocl_bicc: CC = CallingConv::Intel_OCL_BI; break;
   case lltok::kw_cc: {
       unsigned ArbitraryCC;
       Lex.Lex();

Modified: llvm/branches/R600/lib/AsmParser/LLToken.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/AsmParser/LLToken.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/AsmParser/LLToken.h (original)
+++ llvm/branches/R600/lib/AsmParser/LLToken.h Tue Nov 13 09:21:47 2012
@@ -77,6 +77,7 @@
     kw_c,
 
     kw_cc, kw_ccc, kw_fastcc, kw_coldcc,
+	  kw_intel_ocl_bicc,
     kw_x86_stdcallcc, kw_x86_fastcallcc, kw_x86_thiscallcc,
     kw_arm_apcscc, kw_arm_aapcscc, kw_arm_aapcs_vfpcc,
     kw_msp430_intrcc,
@@ -109,6 +110,7 @@
     kw_naked,
     kw_nonlazybind,
     kw_address_safety,
+    kw_minsize,
 
     kw_type,
     kw_opaque,

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -385,8 +385,7 @@
     //   - __tlv_bootstrap - used to make sure support exists
     //   - spare pointer, used when mapped by the runtime
     //   - pointer to mangled symbol above with initializer
-    unsigned AS = GV->getType()->getAddressSpace();
-    unsigned PtrSize = TD->getPointerSizeInBits(AS)/8;
+    unsigned PtrSize = TD->getPointerSizeInBits()/8;
     OutStreamer.EmitSymbolValue(GetExternalSymbolSymbol("_tlv_bootstrap"),
                           PtrSize, 0);
     OutStreamer.EmitIntValue(0, PtrSize, 0);
@@ -1300,7 +1299,7 @@
 
   // Emit the function pointers in the target-specific order
   const DataLayout *TD = TM.getDataLayout();
-  unsigned Align = Log2_32(TD->getPointerPrefAlignment(0));
+  unsigned Align = Log2_32(TD->getPointerPrefAlignment());
   std::stable_sort(Structors.begin(), Structors.end(), priority_order);
   for (unsigned i = 0, e = Structors.size(); i != e; ++i) {
     const MCSection *OutputSection =
@@ -1481,9 +1480,8 @@
     if (Offset == 0)
       return Base;
 
-    unsigned AS = cast<PointerType>(CE->getType())->getAddressSpace();
     // Truncate/sext the offset to the pointer size.
-    unsigned Width = TD.getPointerSizeInBits(AS);
+    unsigned Width = TD.getPointerSizeInBits();
     if (Width < 64)
       Offset = SignExtend64(Offset, Width);
 

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp Tue Nov 13 09:21:47 2012
@@ -112,7 +112,7 @@
   
   switch (Encoding & 0x07) {
   default: llvm_unreachable("Invalid encoded value.");
-  case dwarf::DW_EH_PE_absptr: return TM.getDataLayout()->getPointerSize(0);
+  case dwarf::DW_EH_PE_absptr: return TM.getDataLayout()->getPointerSize();
   case dwarf::DW_EH_PE_udata2: return 2;
   case dwarf::DW_EH_PE_udata4: return 4;
   case dwarf::DW_EH_PE_udata8: return 8;

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/DIE.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/DIE.cpp Tue Nov 13 09:21:47 2012
@@ -200,7 +200,7 @@
   case dwarf::DW_FORM_udata: Asm->EmitULEB128(Integer); return;
   case dwarf::DW_FORM_sdata: Asm->EmitSLEB128(Integer); return;
   case dwarf::DW_FORM_addr:
-    Size = Asm->getDataLayout().getPointerSize(0); break;
+    Size = Asm->getDataLayout().getPointerSize(); break;
   default: llvm_unreachable("DIE Value form not supported yet");
   }
   Asm->OutStreamer.EmitIntValue(Integer, Size, 0/*addrspace*/);
@@ -222,7 +222,7 @@
   case dwarf::DW_FORM_data8: return sizeof(int64_t);
   case dwarf::DW_FORM_udata: return MCAsmInfo::getULEB128Size(Integer);
   case dwarf::DW_FORM_sdata: return MCAsmInfo::getSLEB128Size(Integer);
-  case dwarf::DW_FORM_addr:  return AP->getDataLayout().getPointerSize(0);
+  case dwarf::DW_FORM_addr:  return AP->getDataLayout().getPointerSize();
   default: llvm_unreachable("DIE Value form not supported yet");
   }
 }
@@ -249,7 +249,7 @@
 unsigned DIELabel::SizeOf(AsmPrinter *AP, unsigned Form) const {
   if (Form == dwarf::DW_FORM_data4) return 4;
   if (Form == dwarf::DW_FORM_strp) return 4;
-  return AP->getDataLayout().getPointerSize(0);
+  return AP->getDataLayout().getPointerSize();
 }
 
 #ifndef NDEBUG
@@ -273,7 +273,7 @@
 unsigned DIEDelta::SizeOf(AsmPrinter *AP, unsigned Form) const {
   if (Form == dwarf::DW_FORM_data4) return 4;
   if (Form == dwarf::DW_FORM_strp) return 4;
-  return AP->getDataLayout().getPointerSize(0);
+  return AP->getDataLayout().getPointerSize();
 }
 
 #ifndef NDEBUG

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Nov 13 09:21:47 2012
@@ -307,47 +307,51 @@
   assert(SPDie && "Unable to find subprogram DIE!");
   DISubprogram SP(SPNode);
 
-  DISubprogram SPDecl = SP.getFunctionDeclaration();
-  if (!SPDecl.isSubprogram()) {
-    // There is not any need to generate specification DIE for a function
-    // defined at compile unit level. If a function is defined inside another
-    // function then gdb prefers the definition at top level and but does not
-    // expect specification DIE in parent function. So avoid creating
-    // specification DIE for a function defined inside a function.
-    if (SP.isDefinition() && !SP.getContext().isCompileUnit() &&
-        !SP.getContext().isFile() &&
-        !isSubprogramContext(SP.getContext())) {
-      SPCU->addFlag(SPDie, dwarf::DW_AT_declaration);
-      
-      // Add arguments.
-      DICompositeType SPTy = SP.getType();
-      DIArray Args = SPTy.getTypeArray();
-      unsigned SPTag = SPTy.getTag();
-      if (SPTag == dwarf::DW_TAG_subroutine_type)
-        for (unsigned i = 1, N = Args.getNumElements(); i < N; ++i) {
-          DIE *Arg = new DIE(dwarf::DW_TAG_formal_parameter);
-          DIType ATy = DIType(Args.getElement(i));
-          SPCU->addType(Arg, ATy);
-          if (ATy.isArtificial())
-            SPCU->addFlag(Arg, dwarf::DW_AT_artificial);
-          if (ATy.isObjectPointer())
-            SPCU->addDIEEntry(SPDie, dwarf::DW_AT_object_pointer,
-                              dwarf::DW_FORM_ref4, Arg);
-          SPDie->addChild(Arg);
-        }
-      DIE *SPDeclDie = SPDie;
-      SPDie = new DIE(dwarf::DW_TAG_subprogram);
-      SPCU->addDIEEntry(SPDie, dwarf::DW_AT_specification, dwarf::DW_FORM_ref4,
-                        SPDeclDie);
-      SPCU->addDie(SPDie);
-    }
-  }
-  // Pick up abstract subprogram DIE.
+  // If we're updating an abstract DIE, then we will be adding the children and
+  // object pointer later on. But what we don't want to do is process the
+  // concrete DIE twice.
   if (DIE *AbsSPDIE = AbstractSPDies.lookup(SPNode)) {
+    // Pick up abstract subprogram DIE.
     SPDie = new DIE(dwarf::DW_TAG_subprogram);
     SPCU->addDIEEntry(SPDie, dwarf::DW_AT_abstract_origin,
                       dwarf::DW_FORM_ref4, AbsSPDIE);
     SPCU->addDie(SPDie);
+  } else {
+    DISubprogram SPDecl = SP.getFunctionDeclaration();
+    if (!SPDecl.isSubprogram()) {
+      // There is not any need to generate specification DIE for a function
+      // defined at compile unit level. If a function is defined inside another
+      // function then gdb prefers the definition at top level and but does not
+      // expect specification DIE in parent function. So avoid creating
+      // specification DIE for a function defined inside a function.
+      if (SP.isDefinition() && !SP.getContext().isCompileUnit() &&
+          !SP.getContext().isFile() &&
+          !isSubprogramContext(SP.getContext())) {
+        SPCU->addFlag(SPDie, dwarf::DW_AT_declaration);
+
+        // Add arguments.
+        DICompositeType SPTy = SP.getType();
+        DIArray Args = SPTy.getTypeArray();
+        unsigned SPTag = SPTy.getTag();
+        if (SPTag == dwarf::DW_TAG_subroutine_type)
+          for (unsigned i = 1, N = Args.getNumElements(); i < N; ++i) {
+            DIE *Arg = new DIE(dwarf::DW_TAG_formal_parameter);
+            DIType ATy = DIType(Args.getElement(i));
+            SPCU->addType(Arg, ATy);
+            if (ATy.isArtificial())
+              SPCU->addFlag(Arg, dwarf::DW_AT_artificial);
+            if (ATy.isObjectPointer())
+              SPCU->addDIEEntry(SPDie, dwarf::DW_AT_object_pointer,
+                                dwarf::DW_FORM_ref4, Arg);
+            SPDie->addChild(Arg);
+          }
+        DIE *SPDeclDie = SPDie;
+        SPDie = new DIE(dwarf::DW_TAG_subprogram);
+        SPCU->addDIEEntry(SPDie, dwarf::DW_AT_specification, dwarf::DW_FORM_ref4,
+                          SPDeclDie);
+        SPCU->addDie(SPDie);
+      }
+    }
   }
 
   SPCU->addLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
@@ -384,7 +388,7 @@
     // DW_AT_ranges appropriately.
     TheCU->addUInt(ScopeDIE, dwarf::DW_AT_ranges, dwarf::DW_FORM_data4,
                    DebugRangeSymbols.size() 
-                   * Asm->getDataLayout().getPointerSize(0));
+                   * Asm->getDataLayout().getPointerSize());
     for (SmallVector<InsnRange, 4>::const_iterator RI = Ranges.begin(),
          RE = Ranges.end(); RI != RE; ++RI) {
       DebugRangeSymbols.push_back(getLabelBeforeInsn(RI->first));
@@ -424,7 +428,7 @@
   DISubprogram InlinedSP = getDISubprogram(DS);
   DIE *OriginDIE = TheCU->getDIE(InlinedSP);
   if (!OriginDIE) {
-    DEBUG(dbgs() << "Unable to find original DIE for inlined subprogram.");
+    DEBUG(dbgs() << "Unable to find original DIE for an inlined subprogram.");
     return NULL;
   }
 
@@ -433,7 +437,7 @@
   const MCSymbol *EndLabel = getLabelAfterInsn(RI->second);
 
   if (StartLabel == 0 || EndLabel == 0) {
-    llvm_unreachable("Unexpected Start and End labels for a inlined scope!");
+    llvm_unreachable("Unexpected Start and End labels for an inlined scope!");
   }
   assert(StartLabel->isDefined() &&
          "Invalid starting label for an inlined scope!");
@@ -450,7 +454,7 @@
     // DW_AT_ranges appropriately.
     TheCU->addUInt(ScopeDIE, dwarf::DW_AT_ranges, dwarf::DW_FORM_data4,
                    DebugRangeSymbols.size() 
-                   * Asm->getDataLayout().getPointerSize(0));
+                   * Asm->getDataLayout().getPointerSize());
     for (SmallVector<InsnRange, 4>::const_iterator RI = Ranges.begin(),
          RE = Ranges.end(); RI != RE; ++RI) {
       DebugRangeSymbols.push_back(getLabelBeforeInsn(RI->first));
@@ -803,7 +807,7 @@
         LexicalScope *Scope = 
           new LexicalScope(NULL, DIDescriptor(SP), NULL, false);
         DeadFnScopeMap[SP] = Scope;
-        
+
         // Construct subprogram DIE and add variables DIEs.
         CompileUnit *SPCU = CUMap.lookup(TheCU);
         assert(SPCU && "Unable to find Compile Unit!");
@@ -850,9 +854,9 @@
   Asm->OutStreamer.EmitLabel(Asm->GetTempSymbol("data_end"));
 
   // End text sections.
-  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
-    Asm->OutStreamer.SwitchSection(SectionMap[i]);
-    Asm->OutStreamer.EmitLabel(Asm->GetTempSymbol("section_end", i));
+  for (unsigned I = 0, E = SectionMap.size(); I != E; ++I) {
+    Asm->OutStreamer.SwitchSection(SectionMap[I]);
+    Asm->OutStreamer.EmitLabel(Asm->GetTempSymbol("section_end", I+1));
   }
 
   // Compute DIE offsets and sizes.
@@ -1765,7 +1769,7 @@
     Asm->EmitSectionOffset(Asm->GetTempSymbol("abbrev_begin"),
                            DwarfAbbrevSectionSym);
     Asm->OutStreamer.AddComment("Address Size (in bytes)");
-    Asm->EmitInt8(Asm->getDataLayout().getPointerSize(0));
+    Asm->EmitInt8(Asm->getDataLayout().getPointerSize());
 
     emitDIE(Die);
     Asm->OutStreamer.EmitLabel(Asm->GetTempSymbol("info_end", TheCU->getID()));
@@ -1811,14 +1815,14 @@
   Asm->EmitInt8(0);
 
   Asm->OutStreamer.AddComment("Op size");
-  Asm->EmitInt8(Asm->getDataLayout().getPointerSize(0) + 1);
+  Asm->EmitInt8(Asm->getDataLayout().getPointerSize() + 1);
   Asm->OutStreamer.AddComment("DW_LNE_set_address");
   Asm->EmitInt8(dwarf::DW_LNE_set_address);
 
   Asm->OutStreamer.AddComment("Section end label");
 
   Asm->OutStreamer.EmitSymbolValue(Asm->GetTempSymbol("section_end",SectionEnd),
-                                   Asm->getDataLayout().getPointerSize(0),
+                                   Asm->getDataLayout().getPointerSize(),
                                    0/*AddrSpace*/);
 
   // Mark end of matrix.
@@ -2047,7 +2051,7 @@
   // Start the dwarf loc section.
   Asm->OutStreamer.SwitchSection(
     Asm->getObjFileLowering().getDwarfLocSection());
-  unsigned char Size = Asm->getDataLayout().getPointerSize(0);
+  unsigned char Size = Asm->getDataLayout().getPointerSize();
   Asm->OutStreamer.EmitLabel(Asm->GetTempSymbol("debug_loc", 0));
   unsigned index = 1;
   for (SmallVector<DotDebugLocEntry, 4>::iterator
@@ -2144,7 +2148,7 @@
   // Start the dwarf ranges section.
   Asm->OutStreamer.SwitchSection(
     Asm->getObjFileLowering().getDwarfRangesSection());
-  unsigned char Size = Asm->getDataLayout().getPointerSize(0);
+  unsigned char Size = Asm->getDataLayout().getPointerSize();
   for (SmallVector<const MCSymbol *, 8>::iterator
          I = DebugRangeSymbols.begin(), E = DebugRangeSymbols.end();
        I != E; ++I) {
@@ -2202,7 +2206,7 @@
   Asm->OutStreamer.AddComment("Dwarf Version");
   Asm->EmitInt16(dwarf::DWARF_VERSION);
   Asm->OutStreamer.AddComment("Address Size (in bytes)");
-  Asm->EmitInt8(Asm->getDataLayout().getPointerSize(0));
+  Asm->EmitInt8(Asm->getDataLayout().getPointerSize());
 
   for (SmallVector<const MDNode *, 4>::iterator I = InlinedSPNodes.begin(),
          E = InlinedSPNodes.end(); I != E; ++I) {
@@ -2233,7 +2237,7 @@
 
       if (Asm->isVerbose()) Asm->OutStreamer.AddComment("low_pc");
       Asm->OutStreamer.EmitSymbolValue(LI->first,
-                                       Asm->getDataLayout().getPointerSize(0),0);
+                                       Asm->getDataLayout().getPointerSize(),0);
     }
   }
 

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfDebug.h Tue Nov 13 09:21:47 2012
@@ -21,9 +21,9 @@
 #include "llvm/MC/MachineLocation.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/FoldingSet.h"
+#include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/StringMap.h"
-#include "llvm/ADT/UniqueVector.h"
 #include "llvm/Support/Allocator.h"
 #include "llvm/Support/DebugLoc.h"
 
@@ -231,7 +231,7 @@
   
   /// SectionMap - Provides a unique id per text section.
   ///
-  UniqueVector<const MCSection*> SectionMap;
+  SetVector<const MCSection*> SectionMap;
 
   /// CurrentFnArguments - List of Arguments (DbgValues) for current function.
   SmallVector<DbgVariable *, 8> CurrentFnArguments;

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Nov 13 09:21:47 2012
@@ -417,7 +417,7 @@
     // that we're omitting that bit.
     TTypeEncoding = dwarf::DW_EH_PE_omit;
     // dwarf::DW_EH_PE_absptr
-    TypeFormatSize = Asm->getDataLayout().getPointerSize(0);
+    TypeFormatSize = Asm->getDataLayout().getPointerSize();
   } else {
     // Okay, we have actual filters or typeinfos to emit.  As such, we need to
     // pick a type encoding for them.  We're about to emit a list of pointers to

Modified: llvm/branches/R600/lib/CodeGen/AsmPrinter/OcamlGCPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/AsmPrinter/OcamlGCPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/AsmPrinter/OcamlGCPrinter.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/AsmPrinter/OcamlGCPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -91,7 +91,7 @@
 /// either condition is detected in a function which uses the GC.
 ///
 void OcamlGCMetadataPrinter::finishAssembly(AsmPrinter &AP) {
-  unsigned IntPtrSize = AP.TM.getDataLayout()->getPointerSize(0);
+  unsigned IntPtrSize = AP.TM.getDataLayout()->getPointerSize();
 
   AP.OutStreamer.SwitchSection(AP.getObjFileLowering().getTextSection());
   EmitCamlGlobal(getModule(), AP, "code_end");

Modified: llvm/branches/R600/lib/CodeGen/EarlyIfConversion.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/EarlyIfConversion.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/EarlyIfConversion.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/EarlyIfConversion.cpp Tue Nov 13 09:21:47 2012
@@ -797,6 +797,5 @@
     if (tryConvertIf(I->getBlock()))
       Changed = true;
 
-  MF.verify(this, "After early if-conversion");
   return Changed;
 }

Modified: llvm/branches/R600/lib/CodeGen/GCStrategy.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/GCStrategy.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/GCStrategy.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/GCStrategy.cpp Tue Nov 13 09:21:47 2012
@@ -20,6 +20,7 @@
 #include "llvm/IntrinsicInst.h"
 #include "llvm/Module.h"
 #include "llvm/Analysis/Dominators.h"
+#include "llvm/Analysis/DominatorInternals.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
@@ -387,9 +388,16 @@
   const TargetFrameLowering *TFI = TM->getFrameLowering();
   assert(TFI && "TargetRegisterInfo not available!");
 
-  for (GCFunctionInfo::roots_iterator RI = FI->roots_begin(),
-                                      RE = FI->roots_end(); RI != RE; ++RI)
-    RI->StackOffset = TFI->getFrameIndexOffset(MF, RI->Num);
+  for (GCFunctionInfo::roots_iterator RI = FI->roots_begin();
+       RI != FI->roots_end();) {
+    // If the root references a dead object, no need to keep it.
+    if (MF.getFrameInfo()->isDeadObjectIndex(RI->Num)) {
+      RI = FI->removeStackRoot(RI);
+    } else {
+      RI->StackOffset = TFI->getFrameIndexOffset(MF, RI->Num);
+      ++RI;
+    }
+  }
 }
 
 bool GCMachineCodeAnalysis::runOnMachineFunction(MachineFunction &MF) {

Modified: llvm/branches/R600/lib/CodeGen/IntrinsicLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/IntrinsicLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/IntrinsicLowering.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/IntrinsicLowering.cpp Tue Nov 13 09:21:47 2012
@@ -457,7 +457,7 @@
     break;   // Strip out annotate intrinsic
     
   case Intrinsic::memcpy: {
-    IntegerType *IntPtr = TD.getIntPtrType(Context);
+    Type *IntPtr = TD.getIntPtrType(Context);
     Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
                                         /* isSigned */ false);
     Value *Ops[3];
@@ -468,7 +468,7 @@
     break;
   }
   case Intrinsic::memmove: {
-    IntegerType *IntPtr = TD.getIntPtrType(Context);
+    Type *IntPtr = TD.getIntPtrType(Context);
     Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
                                         /* isSigned */ false);
     Value *Ops[3];
@@ -479,7 +479,7 @@
     break;
   }
   case Intrinsic::memset: {
-    IntegerType *IntPtr = TD.getIntPtrType(Context);
+    Type *IntPtr = TD.getIntPtrType(Context);
     Value *Size = Builder.CreateIntCast(CI->getArgOperand(2), IntPtr,
                                         /* isSigned */ false);
     Value *Ops[3];

Modified: llvm/branches/R600/lib/CodeGen/MachineBasicBlock.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/MachineBasicBlock.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/MachineBasicBlock.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/MachineBasicBlock.cpp Tue Nov 13 09:21:47 2012
@@ -145,7 +145,8 @@
   instr_iterator I = instr_begin(), E = instr_end();
   while (I != E && I->isPHI())
     ++I;
-  assert(!I->isInsideBundle() && "First non-phi MI cannot be inside a bundle!");
+  assert((I == E || !I->isInsideBundle()) &&
+         "First non-phi MI cannot be inside a bundle!");
   return I;
 }
 
@@ -156,7 +157,7 @@
     ++I;
   // FIXME: This needs to change if we wish to bundle labels / dbg_values
   // inside the bundle.
-  assert(!I->isInsideBundle() &&
+  assert((I == E || !I->isInsideBundle()) &&
          "First non-phi / non-label instruction is inside a bundle!");
   return I;
 }

Modified: llvm/branches/R600/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/MachineFunction.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/MachineFunction.cpp Tue Nov 13 09:21:47 2012
@@ -550,7 +550,7 @@
   // address of a block, in which case it is the pointer size.
   switch (getEntryKind()) {
   case MachineJumpTableInfo::EK_BlockAddress:
-    return TD.getPointerSize(0);
+    return TD.getPointerSize();
   case MachineJumpTableInfo::EK_GPRel64BlockAddress:
     return 8;
   case MachineJumpTableInfo::EK_GPRel32BlockAddress:
@@ -570,7 +570,7 @@
   // alignment.
   switch (getEntryKind()) {
   case MachineJumpTableInfo::EK_BlockAddress:
-    return TD.getPointerABIAlignment(0);
+    return TD.getPointerABIAlignment();
   case MachineJumpTableInfo::EK_GPRel64BlockAddress:
     return TD.getABIIntegerTypeAlignment(64);
   case MachineJumpTableInfo::EK_GPRel32BlockAddress:

Modified: llvm/branches/R600/lib/CodeGen/MachineInstr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/MachineInstr.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/MachineInstr.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/MachineInstr.cpp Tue Nov 13 09:21:47 2012
@@ -1015,9 +1015,10 @@
 unsigned MachineInstr::getBundleSize() const {
   assert(isBundle() && "Expecting a bundle");
 
-  MachineBasicBlock::const_instr_iterator I = *this;
+  const MachineBasicBlock *MBB = getParent();
+  MachineBasicBlock::const_instr_iterator I = *this, E = MBB->instr_end();
   unsigned Size = 0;
-  while ((++I)->isInsideBundle()) {
+  while ((++I != E) && I->isInsideBundle()) {
     ++Size;
   }
   assert(Size > 1 && "Malformed bundle");

Modified: llvm/branches/R600/lib/CodeGen/MachineScheduler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/MachineScheduler.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/MachineScheduler.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/MachineScheduler.cpp Tue Nov 13 09:21:47 2012
@@ -49,6 +49,15 @@
 static bool ViewMISchedDAGs = false;
 #endif // NDEBUG
 
+// Threshold to very roughly model an out-of-order processor's instruction
+// buffers. If the actual value of this threshold matters much in practice, then
+// it can be specified by the machine model. For now, it's an experimental
+// tuning knob to determine when and if it matters.
+static cl::opt<unsigned> ILPWindow("ilp-window", cl::Hidden,
+  cl::desc("Allow expected latency to exceed the critical path by N cycles "
+           "before attempting to balance ILP"),
+  cl::init(10U));
+
 //===----------------------------------------------------------------------===//
 // Machine Instruction Scheduling Pass and Registry
 //===----------------------------------------------------------------------===//
@@ -220,7 +229,7 @@
     // The Scheduler may insert instructions during either schedule() or
     // exitRegion(), even for empty regions. So the local iterators 'I' and
     // 'RegionEnd' are invalid across these calls.
-    unsigned RemainingCount = MBB->size();
+    unsigned RemainingInstrs = MBB->size();
     for(MachineBasicBlock::iterator RegionEnd = MBB->end();
         RegionEnd != MBB->begin(); RegionEnd = Scheduler->begin()) {
 
@@ -229,19 +238,19 @@
           || TII->isSchedulingBoundary(llvm::prior(RegionEnd), MBB, *MF)) {
         --RegionEnd;
         // Count the boundary instruction.
-        --RemainingCount;
+        --RemainingInstrs;
       }
 
       // The next region starts above the previous region. Look backward in the
       // instruction stream until we find the nearest boundary.
       MachineBasicBlock::iterator I = RegionEnd;
-      for(;I != MBB->begin(); --I, --RemainingCount) {
+      for(;I != MBB->begin(); --I, --RemainingInstrs) {
         if (TII->isSchedulingBoundary(llvm::prior(I), MBB, *MF))
           break;
       }
       // Notify the scheduler of the region, even if we may skip scheduling
       // it. Perhaps it still needs to be bundled.
-      Scheduler->enterRegion(MBB, I, RegionEnd, RemainingCount);
+      Scheduler->enterRegion(MBB, I, RegionEnd, RemainingInstrs);
 
       // Skip empty scheduling regions (0 or 1 schedulable instructions).
       if (I == RegionEnd || I == llvm::prior(RegionEnd)) {
@@ -255,7 +264,7 @@
             << ":BB#" << MBB->getNumber() << "\n  From: " << *I << "    To: ";
             if (RegionEnd != MBB->end()) dbgs() << *RegionEnd;
             else dbgs() << "End";
-            dbgs() << " Remaining: " << RemainingCount << "\n");
+            dbgs() << " Remaining: " << RemainingInstrs << "\n");
 
       // Schedule a region: possibly reorder instructions.
       // This invalidates 'RegionEnd' and 'I'.
@@ -268,7 +277,7 @@
       // scheduler for the top of it's scheduled region.
       RegionEnd = Scheduler->begin();
     }
-    assert(RemainingCount == 0 && "Instruction count mismatch!");
+    assert(RemainingInstrs == 0 && "Instruction count mismatch!");
     Scheduler->finishBlock();
   }
   Scheduler->finalizeSchedule();
@@ -487,6 +496,13 @@
   assert(CurrentTop == CurrentBottom && "Nonempty unscheduled zone.");
 
   placeDebugValues();
+
+  DEBUG({
+      unsigned BBNum = top()->getParent()->getNumber();
+      dbgs() << "*** Final schedule for BB#" << BBNum << " ***\n";
+      dumpSchedule();
+      dbgs() << '\n';
+    });
 }
 
 /// Build the DAG and setup three register pressure trackers.
@@ -627,6 +643,17 @@
   FirstDbgValue = NULL;
 }
 
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+void ScheduleDAGMI::dumpSchedule() const {
+  for (MachineBasicBlock::iterator MI = begin(), ME = end(); MI != ME; ++MI) {
+    if (SUnit *SU = getSUnit(&(*MI)))
+      SU->dump(this);
+    else
+      dbgs() << "Missing SUnit\n";
+  }
+}
+#endif
+
 //===----------------------------------------------------------------------===//
 // ConvergingScheduler - Implementation of the standard MachineSchedStrategy.
 //===----------------------------------------------------------------------===//
@@ -635,33 +662,127 @@
 /// ConvergingScheduler shrinks the unscheduled zone using heuristics to balance
 /// the schedule.
 class ConvergingScheduler : public MachineSchedStrategy {
+public:
+  /// Represent the type of SchedCandidate found within a single queue.
+  /// pickNodeBidirectional depends on these listed by decreasing priority.
+  enum CandReason {
+    NoCand, SingleExcess, SingleCritical, ResourceReduce, ResourceDemand,
+    BotHeightReduce, BotPathReduce, TopDepthReduce, TopPathReduce,
+    SingleMax, MultiPressure, NextDefUse, NodeOrder};
+
+#ifndef NDEBUG
+  static const char *getReasonStr(ConvergingScheduler::CandReason Reason);
+#endif
+
+  /// Policy for scheduling the next instruction in the candidate's zone.
+  struct CandPolicy {
+    bool ReduceLatency;
+    unsigned ReduceResIdx;
+    unsigned DemandResIdx;
+
+    CandPolicy(): ReduceLatency(false), ReduceResIdx(0), DemandResIdx(0) {}
+  };
+
+  /// Status of an instruction's critical resource consumption.
+  struct SchedResourceDelta {
+    // Count critical resources in the scheduled region required by SU.
+    unsigned CritResources;
+
+    // Count critical resources from another region consumed by SU.
+    unsigned DemandedResources;
+
+    SchedResourceDelta(): CritResources(0), DemandedResources(0) {}
+
+    bool operator==(const SchedResourceDelta &RHS) const {
+      return CritResources == RHS.CritResources
+        && DemandedResources == RHS.DemandedResources;
+    }
+    bool operator!=(const SchedResourceDelta &RHS) const {
+      return !operator==(RHS);
+    }
+  };
 
   /// Store the state used by ConvergingScheduler heuristics, required for the
   /// lifetime of one invocation of pickNode().
   struct SchedCandidate {
+    CandPolicy Policy;
+
     // The best SUnit candidate.
     SUnit *SU;
 
+    // The reason for this candidate.
+    CandReason Reason;
+
     // Register pressure values for the best candidate.
     RegPressureDelta RPDelta;
 
-    SchedCandidate(): SU(NULL) {}
+    // Critical resource consumption of the best candidate.
+    SchedResourceDelta ResDelta;
+
+    SchedCandidate(const CandPolicy &policy)
+    : Policy(policy), SU(NULL), Reason(NoCand) {}
+
+    bool isValid() const { return SU; }
+
+    // Copy the status of another candidate without changing policy.
+    void setBest(SchedCandidate &Best) {
+      assert(Best.Reason != NoCand && "uninitialized Sched candidate");
+      SU = Best.SU;
+      Reason = Best.Reason;
+      RPDelta = Best.RPDelta;
+      ResDelta = Best.ResDelta;
+    }
+
+    void initResourceDelta(const ScheduleDAGMI *DAG,
+                           const TargetSchedModel *SchedModel);
+  };
+
+  /// Summarize the unscheduled region.
+  struct SchedRemainder {
+    // Critical path through the DAG in expected latency.
+    unsigned CriticalPath;
+
+    // Unscheduled resources
+    SmallVector<unsigned, 16> RemainingCounts;
+    // Critical resource for the unscheduled zone.
+    unsigned CritResIdx;
+    // Number of micro-ops left to schedule.
+    unsigned RemainingMicroOps;
+    // Is the unscheduled zone resource limited.
+    bool IsResourceLimited;
+
+    unsigned MaxRemainingCount;
+
+    void reset() {
+      CriticalPath = 0;
+      RemainingCounts.clear();
+      CritResIdx = 0;
+      RemainingMicroOps = 0;
+      IsResourceLimited = false;
+      MaxRemainingCount = 0;
+    }
+
+    SchedRemainder() { reset(); }
+
+    void init(ScheduleDAGMI *DAG, const TargetSchedModel *SchedModel);
   };
-  /// Represent the type of SchedCandidate found within a single queue.
-  enum CandResult {
-    NoCand, NodeOrder, SingleExcess, SingleCritical, SingleMax, MultiPressure };
 
   /// Each Scheduling boundary is associated with ready queues. It tracks the
-  /// current cycle in whichever direction at has moved, and maintains the state
+  /// current cycle in the direction of movement, and maintains the state
   /// of "hazards" and other interlocks at the current cycle.
   struct SchedBoundary {
     ScheduleDAGMI *DAG;
     const TargetSchedModel *SchedModel;
+    SchedRemainder *Rem;
 
     ReadyQueue Available;
     ReadyQueue Pending;
     bool CheckPending;
 
+    // For heuristics, keep a list of the nodes that immediately depend on the
+    // most recently scheduled node.
+    SmallPtrSet<const SUnit*, 8> NextSUs;
+
     ScheduleHazardRecognizer *HazardRec;
 
     unsigned CurrCycle;
@@ -670,34 +791,88 @@
     /// MinReadyCycle - Cycle of the soonest available instruction.
     unsigned MinReadyCycle;
 
+    // The expected latency of the critical path in this scheduled zone.
+    unsigned ExpectedLatency;
+
+    // Resources used in the scheduled zone beyond this boundary.
+    SmallVector<unsigned, 16> ResourceCounts;
+
+    // Cache the critical resources ID in this scheduled zone.
+    unsigned CritResIdx;
+
+    // Is the scheduled region resource limited vs. latency limited.
+    bool IsResourceLimited;
+
+    unsigned ExpectedCount;
+
+    // Policy flag: attempt to find ILP until expected latency is covered.
+    bool ShouldIncreaseILP;
+
+#ifndef NDEBUG
     // Remember the greatest min operand latency.
     unsigned MaxMinLatency;
+#endif
+
+    void reset() {
+      Available.clear();
+      Pending.clear();
+      CheckPending = false;
+      NextSUs.clear();
+      HazardRec = 0;
+      CurrCycle = 0;
+      IssueCount = 0;
+      MinReadyCycle = UINT_MAX;
+      ExpectedLatency = 0;
+      ResourceCounts.resize(1);
+      assert(!ResourceCounts[0] && "nonzero count for bad resource");
+      CritResIdx = 0;
+      IsResourceLimited = false;
+      ExpectedCount = 0;
+      ShouldIncreaseILP = false;
+#ifndef NDEBUG
+      MaxMinLatency = 0;
+#endif
+      // Reserve a zero-count for invalid CritResIdx.
+      ResourceCounts.resize(1);
+    }
 
     /// Pending queues extend the ready queues with the same ID and the
     /// PendingFlag set.
     SchedBoundary(unsigned ID, const Twine &Name):
-      DAG(0), SchedModel(0), Available(ID, Name+".A"),
-      Pending(ID << ConvergingScheduler::LogMaxQID, Name+".P"),
-      CheckPending(false), HazardRec(0), CurrCycle(0), IssueCount(0),
-      MinReadyCycle(UINT_MAX), MaxMinLatency(0) {}
+      DAG(0), SchedModel(0), Rem(0), Available(ID, Name+".A"),
+      Pending(ID << ConvergingScheduler::LogMaxQID, Name+".P") {
+      reset();
+    }
 
     ~SchedBoundary() { delete HazardRec; }
 
-    void init(ScheduleDAGMI *dag, const TargetSchedModel *smodel) {
-      DAG = dag;
-      SchedModel = smodel;
-    }
+    void init(ScheduleDAGMI *dag, const TargetSchedModel *smodel,
+              SchedRemainder *rem);
 
     bool isTop() const {
       return Available.getID() == ConvergingScheduler::TopQID;
     }
 
+    unsigned getUnscheduledLatency(SUnit *SU) const {
+      if (isTop())
+        return SU->getHeight();
+      return SU->getDepth();
+    }
+
+    unsigned getCriticalCount() const {
+      return ResourceCounts[CritResIdx];
+    }
+
     bool checkHazard(SUnit *SU);
 
+    void checkILPPolicy();
+
     void releaseNode(SUnit *SU, unsigned ReadyCycle);
 
     void bumpCycle();
 
+    void countResource(unsigned PIdx, unsigned Cycles);
+
     void bumpNode(SUnit *SU);
 
     void releasePending();
@@ -707,11 +882,13 @@
     SUnit *pickOnlyChoice();
   };
 
+private:
   ScheduleDAGMI *DAG;
   const TargetSchedModel *SchedModel;
   const TargetRegisterInfo *TRI;
 
   // State of the top and bottom scheduled instruction boundaries.
+  SchedRemainder Rem;
   SchedBoundary Top;
   SchedBoundary Bot;
 
@@ -736,25 +913,75 @@
 
   virtual void releaseBottomNode(SUnit *SU);
 
+  virtual void registerRoots();
+
 protected:
-  SUnit *pickNodeBidrectional(bool &IsTopNode);
+  void balanceZones(
+    ConvergingScheduler::SchedBoundary &CriticalZone,
+    ConvergingScheduler::SchedCandidate &CriticalCand,
+    ConvergingScheduler::SchedBoundary &OppositeZone,
+    ConvergingScheduler::SchedCandidate &OppositeCand);
+
+  void checkResourceLimits(ConvergingScheduler::SchedCandidate &TopCand,
+                           ConvergingScheduler::SchedCandidate &BotCand);
+
+  void tryCandidate(SchedCandidate &Cand,
+                    SchedCandidate &TryCand,
+                    SchedBoundary &Zone,
+                    const RegPressureTracker &RPTracker,
+                    RegPressureTracker &TempTracker);
+
+  SUnit *pickNodeBidirectional(bool &IsTopNode);
+
+  void pickNodeFromQueue(SchedBoundary &Zone,
+                         const RegPressureTracker &RPTracker,
+                         SchedCandidate &Candidate);
 
-  CandResult pickNodeFromQueue(ReadyQueue &Q,
-                               const RegPressureTracker &RPTracker,
-                               SchedCandidate &Candidate);
 #ifndef NDEBUG
-  void traceCandidate(const char *Label, const ReadyQueue &Q, SUnit *SU,
-                      PressureElement P = PressureElement());
+  void traceCandidate(const SchedCandidate &Cand, const SchedBoundary &Zone);
 #endif
 };
 } // namespace
 
+void ConvergingScheduler::SchedRemainder::
+init(ScheduleDAGMI *DAG, const TargetSchedModel *SchedModel) {
+  reset();
+  if (!SchedModel->hasInstrSchedModel())
+    return;
+  RemainingCounts.resize(SchedModel->getNumProcResourceKinds());
+  for (std::vector<SUnit>::iterator
+         I = DAG->SUnits.begin(), E = DAG->SUnits.end(); I != E; ++I) {
+    const MCSchedClassDesc *SC = DAG->getSchedClass(&*I);
+    RemainingMicroOps += SchedModel->getNumMicroOps(I->getInstr(), SC);
+    for (TargetSchedModel::ProcResIter
+           PI = SchedModel->getWriteProcResBegin(SC),
+           PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
+      unsigned PIdx = PI->ProcResourceIdx;
+      unsigned Factor = SchedModel->getResourceFactor(PIdx);
+      RemainingCounts[PIdx] += (Factor * PI->Cycles);
+    }
+  }
+}
+
+void ConvergingScheduler::SchedBoundary::
+init(ScheduleDAGMI *dag, const TargetSchedModel *smodel, SchedRemainder *rem) {
+  reset();
+  DAG = dag;
+  SchedModel = smodel;
+  Rem = rem;
+  if (SchedModel->hasInstrSchedModel())
+    ResourceCounts.resize(SchedModel->getNumProcResourceKinds());
+}
+
 void ConvergingScheduler::initialize(ScheduleDAGMI *dag) {
   DAG = dag;
   SchedModel = DAG->getSchedModel();
   TRI = DAG->TRI;
-  Top.init(DAG, SchedModel);
-  Bot.init(DAG, SchedModel);
+  Rem.init(DAG, SchedModel);
+  Top.init(DAG, SchedModel, &Rem);
+  Bot.init(DAG, SchedModel, &Rem);
+
+  // Initialize resource counts.
 
   // Initialize the HazardRecognizers. If itineraries don't exist, are empty, or
   // are disabled, then these HazardRecs will be disabled.
@@ -803,6 +1030,17 @@
   Bot.releaseNode(SU, SU->BotReadyCycle);
 }
 
+void ConvergingScheduler::registerRoots() {
+  Rem.CriticalPath = DAG->ExitSU.getDepth();
+  // Some roots may not feed into ExitSU. Check all of them in case.
+  for (std::vector<SUnit*>::const_iterator
+         I = Bot.Available.begin(), E = Bot.Available.end(); I != E; ++I) {
+    if ((*I)->getDepth() > Rem.CriticalPath)
+      Rem.CriticalPath = (*I)->getDepth();
+  }
+  DEBUG(dbgs() << "Critical Path: " << Rem.CriticalPath << '\n');
+}
+
 /// Does this SU have a hazard within the current instruction group.
 ///
 /// The scheduler supports two modes of hazard recognition. The first is the
@@ -821,14 +1059,26 @@
     return HazardRec->getHazardType(SU) != ScheduleHazardRecognizer::NoHazard;
 
   unsigned uops = SchedModel->getNumMicroOps(SU->getInstr());
-  if (IssueCount + uops > SchedModel->getIssueWidth())
+  if ((IssueCount > 0) && (IssueCount + uops > SchedModel->getIssueWidth())) {
+    DEBUG(dbgs() << "  SU(" << SU->NodeNum << ") uops="
+          << SchedModel->getNumMicroOps(SU->getInstr()) << '\n');
     return true;
-
+  }
   return false;
 }
 
+/// If expected latency is covered, disable ILP policy.
+void ConvergingScheduler::SchedBoundary::checkILPPolicy() {
+  if (ShouldIncreaseILP
+      && (IsResourceLimited || ExpectedLatency <= CurrCycle)) {
+    ShouldIncreaseILP = false;
+    DEBUG(dbgs() << "Disable ILP: " << Available.getName() << '\n');
+  }
+}
+
 void ConvergingScheduler::SchedBoundary::releaseNode(SUnit *SU,
                                                      unsigned ReadyCycle) {
+
   if (ReadyCycle < MinReadyCycle)
     MinReadyCycle = ReadyCycle;
 
@@ -838,6 +1088,18 @@
     Pending.push(SU);
   else
     Available.push(SU);
+
+  // Record this node as an immediate dependent of the scheduled node.
+  NextSUs.insert(SU);
+
+  // If CriticalPath has been computed, then check if the unscheduled nodes
+  // exceed the ILP window. Before registerRoots, CriticalPath==0.
+  if (Rem->CriticalPath && (ExpectedLatency + getUnscheduledLatency(SU)
+                            > Rem->CriticalPath + ILPWindow)) {
+    ShouldIncreaseILP = true;
+    DEBUG(dbgs() << "Increase ILP: " << Available.getName() << " "
+          << ExpectedLatency << " + " << getUnscheduledLatency(SU) << '\n');
+  }
 }
 
 /// Move the boundary of scheduled code by one cycle.
@@ -845,8 +1107,12 @@
   unsigned Width = SchedModel->getIssueWidth();
   IssueCount = (IssueCount <= Width) ? 0 : IssueCount - Width;
 
+  unsigned NextCycle = CurrCycle + 1;
   assert(MinReadyCycle < UINT_MAX && "MinReadyCycle uninitialized");
-  unsigned NextCycle = std::max(CurrCycle + 1, MinReadyCycle);
+  if (MinReadyCycle > NextCycle) {
+    IssueCount = 0;
+    NextCycle = MinReadyCycle;
+  }
 
   if (!HazardRec->isEnabled()) {
     // Bypass HazardRec virtual calls.
@@ -862,11 +1128,39 @@
     }
   }
   CheckPending = true;
+  IsResourceLimited = getCriticalCount() > std::max(ExpectedLatency, CurrCycle);
 
-  DEBUG(dbgs() << "*** " << Available.getName() << " cycle "
+  DEBUG(dbgs() << "  *** " << Available.getName() << " cycle "
         << CurrCycle << '\n');
 }
 
+/// Add the given processor resource to this scheduled zone.
+void ConvergingScheduler::SchedBoundary::countResource(unsigned PIdx,
+                                                       unsigned Cycles) {
+  unsigned Factor = SchedModel->getResourceFactor(PIdx);
+  DEBUG(dbgs() << "  " << SchedModel->getProcResource(PIdx)->Name
+        << " +(" << Cycles << "x" << Factor
+        << ") / " << SchedModel->getLatencyFactor() << '\n');
+
+  unsigned Count = Factor * Cycles;
+  ResourceCounts[PIdx] += Count;
+  assert(Rem->RemainingCounts[PIdx] >= Count && "resource double counted");
+  Rem->RemainingCounts[PIdx] -= Count;
+
+  // Reset MaxRemainingCount for sanity.
+  Rem->MaxRemainingCount = 0;
+
+  // Check if this resource exceeds the current critical resource by a full
+  // cycle. If so, it becomes the critical resource.
+  if ((int)(ResourceCounts[PIdx] - ResourceCounts[CritResIdx])
+      >= (int)SchedModel->getLatencyFactor()) {
+    CritResIdx = PIdx;
+    DEBUG(dbgs() << "  *** Critical resource "
+          << SchedModel->getProcResource(PIdx)->Name << " x"
+          << ResourceCounts[PIdx] << '\n');
+  }
+}
+
 /// Move the boundary of scheduled code by one SUnit.
 void ConvergingScheduler::SchedBoundary::bumpNode(SUnit *SU) {
   // Update the reservation table.
@@ -878,11 +1172,38 @@
     }
     HazardRec->EmitInstruction(SU);
   }
+  // Update resource counts and critical resource.
+  if (SchedModel->hasInstrSchedModel()) {
+    const MCSchedClassDesc *SC = DAG->getSchedClass(SU);
+    Rem->RemainingMicroOps -= SchedModel->getNumMicroOps(SU->getInstr(), SC);
+    for (TargetSchedModel::ProcResIter
+           PI = SchedModel->getWriteProcResBegin(SC),
+           PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
+      countResource(PI->ProcResourceIdx, PI->Cycles);
+    }
+  }
+  if (isTop()) {
+    if (SU->getDepth() > ExpectedLatency)
+      ExpectedLatency = SU->getDepth();
+  }
+  else {
+    if (SU->getHeight() > ExpectedLatency)
+      ExpectedLatency = SU->getHeight();
+  }
+
+  IsResourceLimited = getCriticalCount() > std::max(ExpectedLatency, CurrCycle);
+
   // Check the instruction group dispatch limit.
   // TODO: Check if this SU must end a dispatch group.
   IssueCount += SchedModel->getNumMicroOps(SU->getInstr());
+
+  // checkHazard prevents scheduling multiple instructions per cycle that exceed
+  // issue width. However, we commonly reach the maximum. In this case
+  // opportunistically bump the cycle to avoid uselessly checking everything in
+  // the readyQ. Furthermore, a single instruction may produce more than one
+  // cycle's worth of micro-ops.
   if (IssueCount >= SchedModel->getIssueWidth()) {
-    DEBUG(dbgs() << "*** Max instrs at cycle " << CurrCycle << '\n');
+    DEBUG(dbgs() << "  *** Max instrs at cycle " << CurrCycle << '\n');
     bumpCycle();
   }
 }
@@ -913,6 +1234,7 @@
     Pending.remove(Pending.begin()+i);
     --i; --e;
   }
+  DEBUG(if (!Pending.empty()) Pending.dump());
   CheckPending = false;
 }
 
@@ -927,12 +1249,23 @@
 }
 
 /// If this queue only has one ready candidate, return it. As a side effect,
-/// advance the cycle until at least one node is ready. If multiple instructions
-/// are ready, return NULL.
+/// defer any nodes that now hit a hazard, and advance the cycle until at least
+/// one node is ready. If multiple instructions are ready, return NULL.
 SUnit *ConvergingScheduler::SchedBoundary::pickOnlyChoice() {
   if (CheckPending)
     releasePending();
 
+  if (IssueCount > 0) {
+    // Defer any ready instrs that now have a hazard.
+    for (ReadyQueue::iterator I = Available.begin(); I != Available.end();) {
+      if (checkHazard(*I)) {
+        Pending.push(*I);
+        I = Available.remove(I);
+        continue;
+      }
+      ++I;
+    }
+  }
   for (unsigned i = 0; Available.empty(); ++i) {
     assert(i <= (HazardRec->getMaxLookAhead() + MaxMinLatency) &&
            "permanent hazard"); (void)i;
@@ -944,18 +1277,262 @@
   return NULL;
 }
 
-#ifndef NDEBUG
-void ConvergingScheduler::traceCandidate(const char *Label, const ReadyQueue &Q,
-                                         SUnit *SU, PressureElement P) {
-  dbgs() << Label << " " << Q.getName() << " ";
-  if (P.isValid())
-    dbgs() << TRI->getRegPressureSetName(P.PSetID) << ":" << P.UnitIncrease
-           << " ";
-  else
-    dbgs() << "     ";
-  SU->dump(DAG);
+/// Record the candidate policy for opposite zones with different critical
+/// resources.
+///
+/// If the CriticalZone is latency limited, don't force a policy for the
+/// candidates here. Instead, When releasing each candidate, releaseNode
+/// compares the region's critical path to the candidate's height or depth and
+/// the scheduled zone's expected latency then sets ShouldIncreaseILP.
+void ConvergingScheduler::balanceZones(
+  ConvergingScheduler::SchedBoundary &CriticalZone,
+  ConvergingScheduler::SchedCandidate &CriticalCand,
+  ConvergingScheduler::SchedBoundary &OppositeZone,
+  ConvergingScheduler::SchedCandidate &OppositeCand) {
+
+  if (!CriticalZone.IsResourceLimited)
+    return;
+
+  SchedRemainder *Rem = CriticalZone.Rem;
+
+  // If the critical zone is overconsuming a resource relative to the
+  // remainder, try to reduce it.
+  unsigned RemainingCritCount =
+    Rem->RemainingCounts[CriticalZone.CritResIdx];
+  if ((int)(Rem->MaxRemainingCount - RemainingCritCount)
+      > (int)SchedModel->getLatencyFactor()) {
+    CriticalCand.Policy.ReduceResIdx = CriticalZone.CritResIdx;
+    DEBUG(dbgs() << "Balance " << CriticalZone.Available.getName() << " reduce "
+          << SchedModel->getProcResource(CriticalZone.CritResIdx)->Name
+          << '\n');
+  }
+  // If the other zone is underconsuming a resource relative to the full zone,
+  // try to increase it.
+  unsigned OppositeCount =
+    OppositeZone.ResourceCounts[CriticalZone.CritResIdx];
+  if ((int)(OppositeZone.ExpectedCount - OppositeCount)
+      > (int)SchedModel->getLatencyFactor()) {
+    OppositeCand.Policy.DemandResIdx = CriticalZone.CritResIdx;
+    DEBUG(dbgs() << "Balance " << OppositeZone.Available.getName() << " demand "
+          << SchedModel->getProcResource(OppositeZone.CritResIdx)->Name
+          << '\n');
+  }
+}
+
+/// Determine if the scheduled zones exceed resource limits or critical path and
+/// set each candidate's ReduceHeight policy accordingly.
+void ConvergingScheduler::checkResourceLimits(
+  ConvergingScheduler::SchedCandidate &TopCand,
+  ConvergingScheduler::SchedCandidate &BotCand) {
+
+  Bot.checkILPPolicy();
+  Top.checkILPPolicy();
+  if (Bot.ShouldIncreaseILP)
+    BotCand.Policy.ReduceLatency = true;
+  if (Top.ShouldIncreaseILP)
+    TopCand.Policy.ReduceLatency = true;
+
+  // Handle resource-limited regions.
+  if (Top.IsResourceLimited && Bot.IsResourceLimited
+      && Top.CritResIdx == Bot.CritResIdx) {
+    // If the scheduled critical resource in both zones is no longer the
+    // critical remaining resource, attempt to reduce resource height both ways.
+    if (Top.CritResIdx != Rem.CritResIdx) {
+      TopCand.Policy.ReduceResIdx = Top.CritResIdx;
+      BotCand.Policy.ReduceResIdx = Bot.CritResIdx;
+      DEBUG(dbgs() << "Reduce scheduled "
+            << SchedModel->getProcResource(Top.CritResIdx)->Name << '\n');
+    }
+    return;
+  }
+  // Handle latency-limited regions.
+  if (!Top.IsResourceLimited && !Bot.IsResourceLimited) {
+    // If the total scheduled expected latency exceeds the region's critical
+    // path then reduce latency both ways.
+    //
+    // Just because a zone is not resource limited does not mean it is latency
+    // limited. Unbuffered resource, such as max micro-ops may cause CurrCycle
+    // to exceed expected latency.
+    if ((Top.ExpectedLatency + Bot.ExpectedLatency >= Rem.CriticalPath)
+        && (Rem.CriticalPath > Top.CurrCycle + Bot.CurrCycle)) {
+      TopCand.Policy.ReduceLatency = true;
+      BotCand.Policy.ReduceLatency = true;
+      DEBUG(dbgs() << "Reduce scheduled latency " << Top.ExpectedLatency
+            << " + " << Bot.ExpectedLatency << '\n');
+    }
+    return;
+  }
+  // The critical resource is different in each zone, so request balancing.
+
+  // Compute the cost of each zone.
+  Rem.MaxRemainingCount = std::max(
+    Rem.RemainingMicroOps * SchedModel->getMicroOpFactor(),
+    Rem.RemainingCounts[Rem.CritResIdx]);
+  Top.ExpectedCount = std::max(Top.ExpectedLatency, Top.CurrCycle);
+  Top.ExpectedCount = std::max(
+    Top.getCriticalCount(),
+    Top.ExpectedCount * SchedModel->getLatencyFactor());
+  Bot.ExpectedCount = std::max(Bot.ExpectedLatency, Bot.CurrCycle);
+  Bot.ExpectedCount = std::max(
+    Bot.getCriticalCount(),
+    Bot.ExpectedCount * SchedModel->getLatencyFactor());
+
+  balanceZones(Top, TopCand, Bot, BotCand);
+  balanceZones(Bot, BotCand, Top, TopCand);
+}
+
+void ConvergingScheduler::SchedCandidate::
+initResourceDelta(const ScheduleDAGMI *DAG,
+                  const TargetSchedModel *SchedModel) {
+  if (!Policy.ReduceResIdx && !Policy.DemandResIdx)
+    return;
+
+  const MCSchedClassDesc *SC = DAG->getSchedClass(SU);
+  for (TargetSchedModel::ProcResIter
+         PI = SchedModel->getWriteProcResBegin(SC),
+         PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
+    if (PI->ProcResourceIdx == Policy.ReduceResIdx)
+      ResDelta.CritResources += PI->Cycles;
+    if (PI->ProcResourceIdx == Policy.DemandResIdx)
+      ResDelta.DemandedResources += PI->Cycles;
+  }
+}
+
+/// Return true if this heuristic determines order.
+static bool tryLess(unsigned TryVal, unsigned CandVal,
+                    ConvergingScheduler::SchedCandidate &TryCand,
+                    ConvergingScheduler::SchedCandidate &Cand,
+                    ConvergingScheduler::CandReason Reason) {
+  if (TryVal < CandVal) {
+    TryCand.Reason = Reason;
+    return true;
+  }
+  if (TryVal > CandVal) {
+    if (Cand.Reason > Reason)
+      Cand.Reason = Reason;
+    return true;
+  }
+  return false;
+}
+static bool tryGreater(unsigned TryVal, unsigned CandVal,
+                       ConvergingScheduler::SchedCandidate &TryCand,
+                       ConvergingScheduler::SchedCandidate &Cand,
+                       ConvergingScheduler::CandReason Reason) {
+  if (TryVal > CandVal) {
+    TryCand.Reason = Reason;
+    return true;
+  }
+  if (TryVal < CandVal) {
+    if (Cand.Reason > Reason)
+      Cand.Reason = Reason;
+    return true;
+  }
+  return false;
+}
+
+/// Apply a set of heursitics to a new candidate. Heuristics are currently
+/// hierarchical. This may be more efficient than a graduated cost model because
+/// we don't need to evaluate all aspects of the model for each node in the
+/// queue. But it's really done to make the heuristics easier to debug and
+/// statistically analyze.
+///
+/// \param Cand provides the policy and current best candidate.
+/// \param TryCand refers to the next SUnit candidate, otherwise uninitialized.
+/// \param Zone describes the scheduled zone that we are extending.
+/// \param RPTracker describes reg pressure within the scheduled zone.
+/// \param TempTracker is a scratch pressure tracker to reuse in queries.
+void ConvergingScheduler::tryCandidate(SchedCandidate &Cand,
+                                       SchedCandidate &TryCand,
+                                       SchedBoundary &Zone,
+                                       const RegPressureTracker &RPTracker,
+                                       RegPressureTracker &TempTracker) {
+
+  // Always initialize TryCand's RPDelta.
+  TempTracker.getMaxPressureDelta(TryCand.SU->getInstr(), TryCand.RPDelta,
+                                  DAG->getRegionCriticalPSets(),
+                                  DAG->getRegPressure().MaxSetPressure);
+
+  // Initialize the candidate if needed.
+  if (!Cand.isValid()) {
+    TryCand.Reason = NodeOrder;
+    return;
+  }
+  // Avoid exceeding the target's limit.
+  if (tryLess(TryCand.RPDelta.Excess.UnitIncrease,
+              Cand.RPDelta.Excess.UnitIncrease, TryCand, Cand, SingleExcess))
+    return;
+  if (Cand.Reason == SingleExcess)
+    Cand.Reason = MultiPressure;
+
+  // Avoid increasing the max critical pressure in the scheduled region.
+  if (tryLess(TryCand.RPDelta.CriticalMax.UnitIncrease,
+              Cand.RPDelta.CriticalMax.UnitIncrease,
+              TryCand, Cand, SingleCritical))
+    return;
+  if (Cand.Reason == SingleCritical)
+    Cand.Reason = MultiPressure;
+
+  // Avoid critical resource consumption and balance the schedule.
+  TryCand.initResourceDelta(DAG, SchedModel);
+  if (tryLess(TryCand.ResDelta.CritResources, Cand.ResDelta.CritResources,
+              TryCand, Cand, ResourceReduce))
+    return;
+  if (tryGreater(TryCand.ResDelta.DemandedResources,
+                 Cand.ResDelta.DemandedResources,
+                 TryCand, Cand, ResourceDemand))
+    return;
+
+  // Avoid serializing long latency dependence chains.
+  if (Cand.Policy.ReduceLatency) {
+    if (Zone.isTop()) {
+      if (Cand.SU->getDepth() * SchedModel->getLatencyFactor()
+          > Zone.ExpectedCount) {
+        if (tryLess(TryCand.SU->getDepth(), Cand.SU->getDepth(),
+                    TryCand, Cand, TopDepthReduce))
+          return;
+      }
+      if (tryGreater(TryCand.SU->getHeight(), Cand.SU->getHeight(),
+                     TryCand, Cand, TopPathReduce))
+        return;
+    }
+    else {
+      if (Cand.SU->getHeight() * SchedModel->getLatencyFactor()
+          > Zone.ExpectedCount) {
+        if (tryLess(TryCand.SU->getHeight(), Cand.SU->getHeight(),
+                    TryCand, Cand, BotHeightReduce))
+          return;
+      }
+      if (tryGreater(TryCand.SU->getDepth(), Cand.SU->getDepth(),
+                     TryCand, Cand, BotPathReduce))
+        return;
+    }
+  }
+
+  // Avoid increasing the max pressure of the entire region.
+  if (tryLess(TryCand.RPDelta.CurrentMax.UnitIncrease,
+              Cand.RPDelta.CurrentMax.UnitIncrease, TryCand, Cand, SingleMax))
+    return;
+  if (Cand.Reason == SingleMax)
+    Cand.Reason = MultiPressure;
+
+  // Prefer immediate defs/users of the last scheduled instruction. This is a
+  // nice pressure avoidance strategy that also conserves the processor's
+  // register renaming resources and keeps the machine code readable.
+  if (Zone.NextSUs.count(TryCand.SU) && !Zone.NextSUs.count(Cand.SU)) {
+    TryCand.Reason = NextDefUse;
+    return;
+  }
+  if (!Zone.NextSUs.count(TryCand.SU) && Zone.NextSUs.count(Cand.SU)) {
+    if (Cand.Reason > NextDefUse)
+      Cand.Reason = NextDefUse;
+    return;
+  }
+  // Fall through to original instruction order.
+  if ((Zone.isTop() && TryCand.SU->NodeNum < Cand.SU->NodeNum)
+      || (!Zone.isTop() && TryCand.SU->NodeNum > Cand.SU->NodeNum)) {
+    TryCand.Reason = NodeOrder;
+  }
 }
-#endif
 
 /// pickNodeFromQueue helper that returns true if the LHS reg pressure effect is
 /// more desirable than RHS from scheduling standpoint.
@@ -966,109 +1543,143 @@
   // have UnitIncrease==0, so are neutral.
 
   // Avoid increasing the max critical pressure in the scheduled region.
-  if (LHS.Excess.UnitIncrease != RHS.Excess.UnitIncrease)
+  if (LHS.Excess.UnitIncrease != RHS.Excess.UnitIncrease) {
+    DEBUG(dbgs() << "RP excess top - bot: "
+          << (LHS.Excess.UnitIncrease - RHS.Excess.UnitIncrease) << '\n');
     return LHS.Excess.UnitIncrease < RHS.Excess.UnitIncrease;
-
+  }
   // Avoid increasing the max critical pressure in the scheduled region.
-  if (LHS.CriticalMax.UnitIncrease != RHS.CriticalMax.UnitIncrease)
+  if (LHS.CriticalMax.UnitIncrease != RHS.CriticalMax.UnitIncrease) {
+    DEBUG(dbgs() << "RP critical top - bot: "
+          << (LHS.CriticalMax.UnitIncrease - RHS.CriticalMax.UnitIncrease)
+          << '\n');
     return LHS.CriticalMax.UnitIncrease < RHS.CriticalMax.UnitIncrease;
-
+  }
   // Avoid increasing the max pressure of the entire region.
-  if (LHS.CurrentMax.UnitIncrease != RHS.CurrentMax.UnitIncrease)
+  if (LHS.CurrentMax.UnitIncrease != RHS.CurrentMax.UnitIncrease) {
+    DEBUG(dbgs() << "RP current top - bot: "
+          << (LHS.CurrentMax.UnitIncrease - RHS.CurrentMax.UnitIncrease)
+          << '\n');
     return LHS.CurrentMax.UnitIncrease < RHS.CurrentMax.UnitIncrease;
-
+  }
   return false;
 }
 
+#ifndef NDEBUG
+const char *ConvergingScheduler::getReasonStr(
+  ConvergingScheduler::CandReason Reason) {
+  switch (Reason) {
+  case NoCand:         return "NOCAND    ";
+  case SingleExcess:   return "REG-EXCESS";
+  case SingleCritical: return "REG-CRIT  ";
+  case SingleMax:      return "REG-MAX   ";
+  case MultiPressure:  return "REG-MULTI ";
+  case ResourceReduce: return "RES-REDUCE";
+  case ResourceDemand: return "RES-DEMAND";
+  case TopDepthReduce: return "TOP-DEPTH ";
+  case TopPathReduce:  return "TOP-PATH  ";
+  case BotHeightReduce:return "BOT-HEIGHT";
+  case BotPathReduce:  return "BOT-PATH  ";
+  case NextDefUse:     return "DEF-USE   ";
+  case NodeOrder:      return "ORDER     ";
+  };
+}
+
+void ConvergingScheduler::traceCandidate(const SchedCandidate &Cand,
+                                         const SchedBoundary &Zone) {
+  const char *Label = getReasonStr(Cand.Reason);
+  PressureElement P;
+  unsigned ResIdx = 0;
+  unsigned Latency = 0;
+  switch (Cand.Reason) {
+  default:
+    break;
+  case SingleExcess:
+    P = Cand.RPDelta.Excess;
+    break;
+  case SingleCritical:
+    P = Cand.RPDelta.CriticalMax;
+    break;
+  case SingleMax:
+    P = Cand.RPDelta.CurrentMax;
+    break;
+  case ResourceReduce:
+    ResIdx = Cand.Policy.ReduceResIdx;
+    break;
+  case ResourceDemand:
+    ResIdx = Cand.Policy.DemandResIdx;
+    break;
+  case TopDepthReduce:
+    Latency = Cand.SU->getDepth();
+    break;
+  case TopPathReduce:
+    Latency = Cand.SU->getHeight();
+    break;
+  case BotHeightReduce:
+    Latency = Cand.SU->getHeight();
+    break;
+  case BotPathReduce:
+    Latency = Cand.SU->getDepth();
+    break;
+  }
+  dbgs() << Label << " " << Zone.Available.getName() << " ";
+  if (P.isValid())
+    dbgs() << TRI->getRegPressureSetName(P.PSetID) << ":" << P.UnitIncrease
+           << " ";
+  else
+    dbgs() << "     ";
+  if (ResIdx)
+    dbgs() << SchedModel->getProcResource(ResIdx)->Name << " ";
+  else
+    dbgs() << "        ";
+  if (Latency)
+    dbgs() << Latency << " cycles ";
+  else
+    dbgs() << "         ";
+  Cand.SU->dump(DAG);
+}
+#endif
+
 /// Pick the best candidate from the top queue.
 ///
 /// TODO: getMaxPressureDelta results can be mostly cached for each SUnit during
 /// DAG building. To adjust for the current scheduling location we need to
 /// maintain the number of vreg uses remaining to be top-scheduled.
-ConvergingScheduler::CandResult ConvergingScheduler::
-pickNodeFromQueue(ReadyQueue &Q, const RegPressureTracker &RPTracker,
-                  SchedCandidate &Candidate) {
+void ConvergingScheduler::pickNodeFromQueue(SchedBoundary &Zone,
+                                            const RegPressureTracker &RPTracker,
+                                            SchedCandidate &Cand) {
+  ReadyQueue &Q = Zone.Available;
+
   DEBUG(Q.dump());
 
   // getMaxPressureDelta temporarily modifies the tracker.
   RegPressureTracker &TempTracker = const_cast<RegPressureTracker&>(RPTracker);
 
-  // BestSU remains NULL if no top candidates beat the best existing candidate.
-  CandResult FoundCandidate = NoCand;
   for (ReadyQueue::iterator I = Q.begin(), E = Q.end(); I != E; ++I) {
-    RegPressureDelta RPDelta;
-    TempTracker.getMaxPressureDelta((*I)->getInstr(), RPDelta,
-                                    DAG->getRegionCriticalPSets(),
-                                    DAG->getRegPressure().MaxSetPressure);
-
-    // Initialize the candidate if needed.
-    if (!Candidate.SU) {
-      Candidate.SU = *I;
-      Candidate.RPDelta = RPDelta;
-      FoundCandidate = NodeOrder;
-      continue;
-    }
-    // Avoid exceeding the target's limit.
-    if (RPDelta.Excess.UnitIncrease < Candidate.RPDelta.Excess.UnitIncrease) {
-      DEBUG(traceCandidate("ECAND", Q, *I, RPDelta.Excess));
-      Candidate.SU = *I;
-      Candidate.RPDelta = RPDelta;
-      FoundCandidate = SingleExcess;
-      continue;
-    }
-    if (RPDelta.Excess.UnitIncrease > Candidate.RPDelta.Excess.UnitIncrease)
-      continue;
-    if (FoundCandidate == SingleExcess)
-      FoundCandidate = MultiPressure;
-
-    // Avoid increasing the max critical pressure in the scheduled region.
-    if (RPDelta.CriticalMax.UnitIncrease
-        < Candidate.RPDelta.CriticalMax.UnitIncrease) {
-      DEBUG(traceCandidate("PCAND", Q, *I, RPDelta.CriticalMax));
-      Candidate.SU = *I;
-      Candidate.RPDelta = RPDelta;
-      FoundCandidate = SingleCritical;
-      continue;
-    }
-    if (RPDelta.CriticalMax.UnitIncrease
-        > Candidate.RPDelta.CriticalMax.UnitIncrease)
-      continue;
-    if (FoundCandidate == SingleCritical)
-      FoundCandidate = MultiPressure;
-
-    // Avoid increasing the max pressure of the entire region.
-    if (RPDelta.CurrentMax.UnitIncrease
-        < Candidate.RPDelta.CurrentMax.UnitIncrease) {
-      DEBUG(traceCandidate("MCAND", Q, *I, RPDelta.CurrentMax));
-      Candidate.SU = *I;
-      Candidate.RPDelta = RPDelta;
-      FoundCandidate = SingleMax;
-      continue;
-    }
-    if (RPDelta.CurrentMax.UnitIncrease
-        > Candidate.RPDelta.CurrentMax.UnitIncrease)
-      continue;
-    if (FoundCandidate == SingleMax)
-      FoundCandidate = MultiPressure;
-
-    // Fall through to original instruction order.
-    // Only consider node order if Candidate was chosen from this Q.
-    if (FoundCandidate == NoCand)
-      continue;
 
-    if ((Q.getID() == TopQID && (*I)->NodeNum < Candidate.SU->NodeNum)
-        || (Q.getID() == BotQID && (*I)->NodeNum > Candidate.SU->NodeNum)) {
-      DEBUG(traceCandidate("NCAND", Q, *I));
-      Candidate.SU = *I;
-      Candidate.RPDelta = RPDelta;
-      FoundCandidate = NodeOrder;
+    SchedCandidate TryCand(Cand.Policy);
+    TryCand.SU = *I;
+    tryCandidate(Cand, TryCand, Zone, RPTracker, TempTracker);
+    if (TryCand.Reason != NoCand) {
+      // Initialize resource delta if needed in case future heuristics query it.
+      if (TryCand.ResDelta == SchedResourceDelta())
+        TryCand.initResourceDelta(DAG, SchedModel);
+      Cand.setBest(TryCand);
+      DEBUG(traceCandidate(Cand, Zone));
     }
+    TryCand.SU = *I;
   }
-  return FoundCandidate;
+}
+
+static void tracePick(const ConvergingScheduler::SchedCandidate &Cand,
+                      bool IsTop) {
+  DEBUG(dbgs() << "Pick " << (IsTop ? "top" : "bot")
+        << " SU(" << Cand.SU->NodeNum << ") "
+        << ConvergingScheduler::getReasonStr(Cand.Reason) << '\n');
 }
 
 /// Pick the best candidate node from either the top or bottom queue.
-SUnit *ConvergingScheduler::pickNodeBidrectional(bool &IsTopNode) {
+SUnit *ConvergingScheduler::pickNodeBidirectional(bool &IsTopNode) {
   // Schedule as far as possible in the direction of no choice. This is most
   // efficient, but also provides the best heuristics for CriticalPSets.
   if (SUnit *SU = Bot.pickOnlyChoice()) {
@@ -1079,11 +1690,14 @@
     IsTopNode = true;
     return SU;
   }
-  SchedCandidate BotCand;
+  CandPolicy NoPolicy;
+  SchedCandidate BotCand(NoPolicy);
+  SchedCandidate TopCand(NoPolicy);
+  checkResourceLimits(TopCand, BotCand);
+
   // Prefer bottom scheduling when heuristics are silent.
-  CandResult BotResult = pickNodeFromQueue(Bot.Available,
-                                           DAG->getBotRPTracker(), BotCand);
-  assert(BotResult != NoCand && "failed to find the first candidate");
+  pickNodeFromQueue(Bot, DAG->getBotRPTracker(), BotCand);
+  assert(BotCand.Reason != NoCand && "failed to find the first candidate");
 
   // If either Q has a single candidate that provides the least increase in
   // Excess pressure, we can immediately schedule from that Q.
@@ -1092,37 +1706,41 @@
   // affects picking from either Q. If scheduling in one direction must
   // increase pressure for one of the excess PSets, then schedule in that
   // direction first to provide more freedom in the other direction.
-  if (BotResult == SingleExcess || BotResult == SingleCritical) {
+  if (BotCand.Reason == SingleExcess || BotCand.Reason == SingleCritical) {
     IsTopNode = false;
+    tracePick(BotCand, IsTopNode);
     return BotCand.SU;
   }
   // Check if the top Q has a better candidate.
-  SchedCandidate TopCand;
-  CandResult TopResult = pickNodeFromQueue(Top.Available,
-                                           DAG->getTopRPTracker(), TopCand);
-  assert(TopResult != NoCand && "failed to find the first candidate");
+  pickNodeFromQueue(Top, DAG->getTopRPTracker(), TopCand);
+  assert(TopCand.Reason != NoCand && "failed to find the first candidate");
 
-  if (TopResult == SingleExcess || TopResult == SingleCritical) {
-    IsTopNode = true;
-    return TopCand.SU;
-  }
   // If either Q has a single candidate that minimizes pressure above the
   // original region's pressure pick it.
-  if (BotResult == SingleMax) {
+  if (TopCand.Reason <= SingleMax || BotCand.Reason <= SingleMax) {
+    if (TopCand.Reason < BotCand.Reason) {
+      IsTopNode = true;
+      tracePick(TopCand, IsTopNode);
+      return TopCand.SU;
+    }
     IsTopNode = false;
+    tracePick(BotCand, IsTopNode);
     return BotCand.SU;
   }
-  if (TopResult == SingleMax) {
+  // Check for a salient pressure difference and pick the best from either side.
+  if (compareRPDelta(TopCand.RPDelta, BotCand.RPDelta)) {
     IsTopNode = true;
+    tracePick(TopCand, IsTopNode);
     return TopCand.SU;
   }
-  // Check for a salient pressure difference and pick the best from either side.
-  if (compareRPDelta(TopCand.RPDelta, BotCand.RPDelta)) {
+  // Otherwise prefer the bottom candidate, in node order if all else failed.
+  if (TopCand.Reason < BotCand.Reason) {
     IsTopNode = true;
+    tracePick(TopCand, IsTopNode);
     return TopCand.SU;
   }
-  // Otherwise prefer the bottom candidate in node order.
   IsTopNode = false;
+  tracePick(BotCand, IsTopNode);
   return BotCand.SU;
 }
 
@@ -1138,11 +1756,10 @@
     if (ForceTopDown) {
       SU = Top.pickOnlyChoice();
       if (!SU) {
-        SchedCandidate TopCand;
-        CandResult TopResult =
-          pickNodeFromQueue(Top.Available, DAG->getTopRPTracker(), TopCand);
-        assert(TopResult != NoCand && "failed to find the first candidate");
-        (void)TopResult;
+        CandPolicy NoPolicy;
+        SchedCandidate TopCand(NoPolicy);
+        pickNodeFromQueue(Top, DAG->getTopRPTracker(), TopCand);
+        assert(TopCand.Reason != NoCand && "failed to find the first candidate");
         SU = TopCand.SU;
       }
       IsTopNode = true;
@@ -1150,17 +1767,16 @@
     else if (ForceBottomUp) {
       SU = Bot.pickOnlyChoice();
       if (!SU) {
-        SchedCandidate BotCand;
-        CandResult BotResult =
-          pickNodeFromQueue(Bot.Available, DAG->getBotRPTracker(), BotCand);
-        assert(BotResult != NoCand && "failed to find the first candidate");
-        (void)BotResult;
+        CandPolicy NoPolicy;
+        SchedCandidate BotCand(NoPolicy);
+        pickNodeFromQueue(Bot, DAG->getBotRPTracker(), BotCand);
+        assert(BotCand.Reason != NoCand && "failed to find the first candidate");
         SU = BotCand.SU;
       }
       IsTopNode = false;
     }
     else {
-      SU = pickNodeBidrectional(IsTopNode);
+      SU = pickNodeBidirectional(IsTopNode);
     }
   } while (SU->isScheduled);
 

Modified: llvm/branches/R600/lib/CodeGen/MachineVerifier.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/MachineVerifier.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/MachineVerifier.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/MachineVerifier.cpp Tue Nov 13 09:21:47 2012
@@ -707,8 +707,9 @@
     report("Asm string must be an external symbol", MI);
   if (!MI->getOperand(1).isImm())
     report("Asm flags must be an immediate", MI);
-  // Allowed flags are Extra_HasSideEffects = 1, and Extra_IsAlignStack = 2.
-  if (!isUInt<2>(MI->getOperand(1).getImm()))
+  // Allowed flags are Extra_HasSideEffects = 1, Extra_IsAlignStack = 2,
+  // Extra_AsmDialect = 4, Extra_MayLoad = 8, and Extra_MayStore = 16.
+  if (!isUInt<5>(MI->getOperand(1).getImm()))
     report("Unknown asm flags", &MI->getOperand(1), 1);
 
   assert(InlineAsm::MIOp_FirstOperand == 2 && "Asm format changed");

Modified: llvm/branches/R600/lib/CodeGen/RegAllocFast.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/RegAllocFast.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/RegAllocFast.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/RegAllocFast.cpp Tue Nov 13 09:21:47 2012
@@ -175,7 +175,7 @@
                                        unsigned VirtReg, unsigned Hint);
     LiveRegMap::iterator reloadVirtReg(MachineInstr *MI, unsigned OpNum,
                                        unsigned VirtReg, unsigned Hint);
-    void spillAll(MachineInstr *MI);
+    void spillAll(MachineBasicBlock::iterator MI);
     bool setPhysReg(MachineInstr *MI, unsigned OpNum, unsigned PhysReg);
     void addRetOperands(MachineBasicBlock *MBB);
   };
@@ -314,7 +314,7 @@
 }
 
 /// spillAll - Spill all dirty virtregs without killing them.
-void RAFast::spillAll(MachineInstr *MI) {
+void RAFast::spillAll(MachineBasicBlock::iterator MI) {
   if (LiveVirtRegs.empty()) return;
   isBulkSpilling = true;
   // The LiveRegMap is keyed by an unsigned (the virtreg number), so the order
@@ -956,6 +956,11 @@
     bool hasPhysDefs = false;
     for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
       MachineOperand &MO = MI->getOperand(i);
+      // Make sure MRI knows about registers clobbered by regmasks.
+      if (MO.isRegMask()) {
+        MRI->addPhysRegsUsedFromRegMask(MO.getRegMask());
+        continue;
+      }
       if (!MO.isReg()) continue;
       unsigned Reg = MO.getReg();
       if (!Reg) continue;

Modified: llvm/branches/R600/lib/CodeGen/RegAllocPBQP.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/RegAllocPBQP.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/RegAllocPBQP.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/RegAllocPBQP.cpp Tue Nov 13 09:21:47 2012
@@ -118,7 +118,6 @@
   typedef std::vector<AllowedSet> AllowedSetMap;
   typedef std::pair<unsigned, unsigned> RegPair;
   typedef std::map<RegPair, PBQP::PBQPNum> CoalesceMap;
-  typedef std::vector<PBQP::Graph::NodeItr> NodeVector;
   typedef std::set<unsigned> RegSet;
 
 

Modified: llvm/branches/R600/lib/CodeGen/RegisterCoalescer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/RegisterCoalescer.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/RegisterCoalescer.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/RegisterCoalescer.cpp Tue Nov 13 09:21:47 2012
@@ -198,12 +198,6 @@
 
 char RegisterCoalescer::ID = 0;
 
-static unsigned compose(const TargetRegisterInfo &tri, unsigned a, unsigned b) {
-  if (!a) return b;
-  if (!b) return a;
-  return tri.composeSubRegIndices(a, b);
-}
-
 static bool isMoveInstr(const TargetRegisterInfo &tri, const MachineInstr *MI,
                         unsigned &Src, unsigned &Dst,
                         unsigned &SrcSub, unsigned &DstSub) {
@@ -214,8 +208,8 @@
     SrcSub = MI->getOperand(1).getSubReg();
   } else if (MI->isSubregToReg()) {
     Dst = MI->getOperand(0).getReg();
-    DstSub = compose(tri, MI->getOperand(0).getSubReg(),
-                     MI->getOperand(3).getImm());
+    DstSub = tri.composeSubRegIndices(MI->getOperand(0).getSubReg(),
+                                      MI->getOperand(3).getImm());
     Src = MI->getOperand(2).getReg();
     SrcSub = MI->getOperand(2).getSubReg();
   } else
@@ -354,7 +348,8 @@
     if (DstReg != Dst)
       return false;
     // Registers match, do the subregisters line up?
-    return compose(TRI, SrcIdx, SrcSub) == compose(TRI, DstIdx, DstSub);
+    return TRI.composeSubRegIndices(SrcIdx, SrcSub) ==
+           TRI.composeSubRegIndices(DstIdx, DstSub);
   }
 }
 
@@ -430,7 +425,8 @@
   // If AValNo is defined as a copy from IntB, we can potentially process this.
   // Get the instruction that defines this value number.
   MachineInstr *ACopyMI = LIS->getInstructionFromIndex(AValNo->def);
-  if (!CP.isCoalescable(ACopyMI))
+  // Don't allow any partial copies, even if isCoalescable() allows them.
+  if (!CP.isCoalescable(ACopyMI) || !ACopyMI->isFullCopy())
     return false;
 
   // Get the LiveRange in IntB that this value number starts with.
@@ -1314,7 +1310,8 @@
   for (ConstMIOperands MO(DefMI); MO.isValid(); ++MO) {
     if (!MO->isReg() || MO->getReg() != LI.reg || !MO->isDef())
       continue;
-    L |= TRI->getSubRegIndexLaneMask(compose(*TRI, SubIdx, MO->getSubReg()));
+    L |= TRI->getSubRegIndexLaneMask(
+           TRI->composeSubRegIndices(SubIdx, MO->getSubReg()));
     if (MO->readsReg())
       Redef = true;
   }
@@ -1492,6 +1489,20 @@
   if ((V.WriteLanes & OtherV.ValidLanes) == 0)
     return CR_Replace;
 
+  // If the other live range is killed by DefMI and the live ranges are still
+  // overlapping, it must be because we're looking at an early clobber def:
+  //
+  //   %dst<def,early-clobber> = ASM %src<kill>
+  //
+  // In this case, it is illegal to merge the two live ranges since the early
+  // clobber def would clobber %src before it was read.
+  if (OtherLRQ.isKill()) {
+    // This case where the def doesn't overlap the kill is handled above.
+    assert(VNI->def.isEarlyClobber() &&
+           "Only early clobber defs can overlap a kill");
+    return CR_Impossible;
+  }
+
   // VNI is clobbering live lanes in OtherVNI, but there is still the
   // possibility that no instructions actually read the clobbered lanes.
   // If we're clobbering all the lanes in OtherVNI, at least one must be read.
@@ -1632,8 +1643,8 @@
       continue;
     if (!MO->readsReg())
       continue;
-    if (Lanes &
-        TRI->getSubRegIndexLaneMask(compose(*TRI, SubIdx, MO->getSubReg())))
+    if (Lanes & TRI->getSubRegIndexLaneMask(
+                  TRI->composeSubRegIndices(SubIdx, MO->getSubReg())))
       return true;
   }
   return false;

Modified: llvm/branches/R600/lib/CodeGen/RegisterPressure.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/RegisterPressure.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/RegisterPressure.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/RegisterPressure.cpp Tue Nov 13 09:21:47 2012
@@ -64,7 +64,7 @@
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-void RegisterPressure::dump(const TargetRegisterInfo *TRI) {
+void RegisterPressure::dump(const TargetRegisterInfo *TRI) const {
   dbgs() << "Live In: ";
   for (unsigned i = 0, e = LiveInRegs.size(); i < e; ++i)
     dbgs() << PrintReg(LiveInRegs[i], TRI) << " ";
@@ -322,10 +322,8 @@
         if (findReg(MO.getReg(), isVReg, DeadDefs, TRI) == DeadDefs.end())
           DeadDefs.push_back(MO.getReg());
       }
-      else {
-        if (findReg(MO.getReg(), isVReg, Defs, TRI) == Defs.end())
-          Defs.push_back(MO.getReg());
-      }
+      else if (findReg(MO.getReg(), isVReg, Defs, TRI) == Defs.end())
+        Defs.push_back(MO.getReg());
     }
   }
 };
@@ -676,9 +674,16 @@
   decreaseVirtRegPressure(VirtRegOpers.DeadDefs);
 
   // Kill liveness at live defs.
-  decreasePhysRegPressure(PhysRegOpers.Defs);
-  decreaseVirtRegPressure(VirtRegOpers.Defs);
-
+  for (unsigned i = 0, e = PhysRegOpers.Defs.size(); i < e; ++i) {
+    unsigned Reg = PhysRegOpers.Defs[i];
+    if (!findReg(Reg, false, PhysRegOpers.Uses, TRI))
+      decreasePhysRegPressure(PhysRegOpers.Defs);
+  }
+  for (unsigned i = 0, e = VirtRegOpers.Defs.size(); i < e; ++i) {
+    unsigned Reg = VirtRegOpers.Defs[i];
+    if (!findReg(Reg, true, VirtRegOpers.Uses, TRI))
+      decreaseVirtRegPressure(VirtRegOpers.Defs);
+  }
   // Generate liveness for uses.
   for (unsigned i = 0, e = PhysRegOpers.Uses.size(); i < e; ++i) {
     unsigned Reg = PhysRegOpers.Uses[i];

Modified: llvm/branches/R600/lib/CodeGen/ScheduleDAGInstrs.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/ScheduleDAGInstrs.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/ScheduleDAGInstrs.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/ScheduleDAGInstrs.cpp Tue Nov 13 09:21:47 2012
@@ -71,7 +71,7 @@
       // object. We don't have to worry about the case where the
       // object address is somehow being computed by the multiply,
       // because our callers only care when the result is an
-      // identifibale object.
+      // identifiable object.
       if (U->getOpcode() != Instruction::Add ||
           (!isa<ConstantInt>(U->getOperand(1)) &&
            Operator::getOpcode(U->getOperand(1)) != Instruction::Mul))
@@ -245,7 +245,7 @@
       if (UseSU == SU)
         continue;
 
-      SDep dep(SU, SDep::Data, 1, *Alias);
+      SDep dep(SU, SDep::Data, *Alias);
 
       // Adjust the dependence latency using operand def/use information,
       // then allow the target to perform its own adjustments.
@@ -291,11 +291,14 @@
           (Kind != SDep::Output || !MO.isDead() ||
            !DefSU->getInstr()->registerDefIsDead(*Alias))) {
         if (Kind == SDep::Anti)
-          DefSU->addPred(SDep(SU, Kind, 0, /*Reg=*/*Alias));
+          DefSU->addPred(SDep(SU, Kind, /*Reg=*/*Alias));
         else {
-          unsigned AOLat =
+          SDep Dep(SU, Kind, /*Reg=*/*Alias);
+          unsigned OutLatency =
             SchedModel.computeOutputLatency(MI, OperIdx, DefSU->getInstr());
-          DefSU->addPred(SDep(SU, Kind, AOLat, /*Reg=*/*Alias));
+          Dep.setMinLatency(OutLatency);
+          Dep.setLatency(OutLatency);
+          DefSU->addPred(Dep);
         }
       }
     }
@@ -364,9 +367,12 @@
   else {
     SUnit *DefSU = DefI->SU;
     if (DefSU != SU && DefSU != &ExitSU) {
+      SDep Dep(SU, SDep::Output, Reg);
       unsigned OutLatency =
         SchedModel.computeOutputLatency(MI, OperIdx, DefSU->getInstr());
-      DefSU->addPred(SDep(SU, SDep::Output, OutLatency, Reg));
+      Dep.setMinLatency(OutLatency);
+      Dep.setLatency(OutLatency);
+      DefSU->addPred(Dep);
     }
     DefI->SU = SU;
   }
@@ -396,7 +402,7 @@
     if (DefSU) {
       // The reaching Def lives within this scheduling region.
       // Create a data dependence.
-      SDep dep(DefSU, SDep::Data, 1, Reg);
+      SDep dep(DefSU, SDep::Data, Reg);
       // Adjust the dependence latency using operand def/use information, then
       // allow the target to perform its own adjustments.
       int DefOp = Def->findRegisterDefOperandIdx(Reg);
@@ -414,7 +420,7 @@
   // Add antidependence to the following def of the vreg it uses.
   VReg2SUnitMap::iterator DefI = VRegDefs.find(Reg);
   if (DefI != VRegDefs.end() && DefI->SU != SU)
-    DefI->SU->addPred(SDep(SU, SDep::Anti, 0, Reg));
+    DefI->SU->addPred(SDep(SU, SDep::Anti, Reg));
 }
 
 /// Return true if MI is an instruction we are unable to reason about
@@ -554,8 +560,7 @@
   // and stop descending.
   if (*Depth > 200 ||
       MIsNeedChainEdge(AA, MFI, SUa->getInstr(), SUb->getInstr())) {
-    SUb->addPred(SDep(SUa, SDep::Order, /*Latency=*/0, /*Reg=*/0,
-                      /*isNormalMemory=*/true));
+    SUb->addPred(SDep(SUa, SDep::MayAliasMem));
     return *Depth;
   }
   // Track current depth.
@@ -586,9 +591,9 @@
     if (SU == *I)
       continue;
     if (MIsNeedChainEdge(AA, MFI, SU->getInstr(), (*I)->getInstr())) {
-      unsigned Latency = ((*I)->getInstr()->mayLoad()) ? LatencyToLoad : 0;
-      (*I)->addPred(SDep(SU, SDep::Order, Latency, /*Reg=*/0,
-                         /*isNormalMemory=*/true));
+      SDep Dep(SU, SDep::MayAliasMem);
+      Dep.setLatency(((*I)->getInstr()->mayLoad()) ? LatencyToLoad : 0);
+      (*I)->addPred(Dep);
     }
     // Now go through all the chain successors and iterate from them.
     // Keep track of visited nodes.
@@ -611,9 +616,11 @@
   // If this is a false dependency,
   // do not add the edge, but rememeber the rejected node.
   if (!EnableAASchedMI ||
-      MIsNeedChainEdge(AA, MFI, SUa->getInstr(), SUb->getInstr()))
-    SUb->addPred(SDep(SUa, SDep::Order, TrueMemOrderLatency, /*Reg=*/0,
-                      isNormalMemory));
+      MIsNeedChainEdge(AA, MFI, SUa->getInstr(), SUb->getInstr())) {
+    SDep Dep(SUa, isNormalMemory ? SDep::MayAliasMem : SDep::Barrier);
+    Dep.setLatency(TrueMemOrderLatency);
+    SUb->addPred(Dep);
+  }
   else {
     // Duplicate entries should be ignored.
     RejectList.insert(SUb);
@@ -755,16 +762,19 @@
       // references, even those that are known to not alias.
       for (std::map<const Value *, SUnit *>::iterator I =
              NonAliasMemDefs.begin(), E = NonAliasMemDefs.end(); I != E; ++I) {
-        I->second->addPred(SDep(SU, SDep::Order, /*Latency=*/0));
+        I->second->addPred(SDep(SU, SDep::Barrier));
       }
       for (std::map<const Value *, std::vector<SUnit *> >::iterator I =
              NonAliasMemUses.begin(), E = NonAliasMemUses.end(); I != E; ++I) {
-        for (unsigned i = 0, e = I->second.size(); i != e; ++i)
-          I->second[i]->addPred(SDep(SU, SDep::Order, TrueMemOrderLatency));
+        for (unsigned i = 0, e = I->second.size(); i != e; ++i) {
+          SDep Dep(SU, SDep::Barrier);
+          Dep.setLatency(TrueMemOrderLatency);
+          I->second[i]->addPred(Dep);
+        }
       }
       // Add SU to the barrier chain.
       if (BarrierChain)
-        BarrierChain->addPred(SDep(SU, SDep::Order, /*Latency=*/0));
+        BarrierChain->addPred(SDep(SU, SDep::Barrier));
       BarrierChain = SU;
       // This is a barrier event that acts as a pivotal node in the DAG,
       // so it is safe to clear list of exposed nodes.
@@ -852,7 +862,7 @@
         // SU and barrier _could_ be reordered, they should not. In addition,
         // we have lost all RejectMemNodes below barrier.
         if (BarrierChain)
-          BarrierChain->addPred(SDep(SU, SDep::Order, /*Latency=*/0));
+          BarrierChain->addPred(SDep(SU, SDep::Barrier));
       } else {
         // Treat all other stores conservatively.
         goto new_alias_chain;
@@ -861,10 +871,7 @@
       if (!ExitSU.isPred(SU))
         // Push store's up a bit to avoid them getting in between cmp
         // and branches.
-        ExitSU.addPred(SDep(SU, SDep::Order, 0,
-                            /*Reg=*/0, /*isNormalMemory=*/false,
-                            /*isMustAlias=*/false,
-                            /*isArtificial=*/true));
+        ExitSU.addPred(SDep(SU, SDep::Artificial));
     } else if (MI->mayLoad()) {
       bool MayAlias = true;
       if (MI->isInvariantLoad(AA)) {
@@ -899,7 +906,7 @@
         if (MayAlias && AliasChain)
           addChainDependency(AA, MFI, SU, AliasChain, RejectMemNodes);
         if (BarrierChain)
-          BarrierChain->addPred(SDep(SU, SDep::Order, /*Latency=*/0));
+          BarrierChain->addPred(SDep(SU, SDep::Barrier));
       }
     }
   }

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Tue Nov 13 09:21:47 2012
@@ -270,6 +270,8 @@
     SDValue ReduceLoadWidth(SDNode *N);
     SDValue ReduceLoadOpStoreWidth(SDNode *N);
     SDValue TransformFPLoadStorePair(SDNode *N);
+    SDValue reduceBuildVecExtToExtBuildVec(SDNode *N);
+    SDValue reduceBuildVecConvertToConvertBuildVec(SDNode *N);
 
     SDValue GetDemandedBits(SDValue V, const APInt &Mask);
 
@@ -391,10 +393,6 @@
                                const TargetLowering &TLI,
                                const TargetOptions *Options,
                                unsigned Depth = 0) {
-  // No compile time optimizations on this type.
-  if (Op.getValueType() == MVT::ppcf128)
-    return 0;
-
   // fneg is removable even if it has multiple uses.
   if (Op.getOpcode() == ISD::FNEG) return 2;
 
@@ -5703,7 +5701,7 @@
   }
 
   // fold (fadd c1, c2) -> c1 + c2
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)
+  if (N0CFP && N1CFP)
     return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0, N1);
   // canonicalize constant to RHS
   if (N0CFP && !N1CFP)
@@ -5731,6 +5729,18 @@
                        DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                    N0.getOperand(1), N1));
 
+  // If allow, fold (fadd (fneg x), x) -> 0.0
+  if (DAG.getTarget().Options.UnsafeFPMath &&
+      N0.getOpcode() == ISD::FNEG && N0.getOperand(0) == N1) {
+    return DAG.getConstantFP(0.0, VT);
+  }
+
+    // If allow, fold (fadd x, (fneg x)) -> 0.0
+  if (DAG.getTarget().Options.UnsafeFPMath &&
+      N1.getOpcode() == ISD::FNEG && N1.getOperand(0) == N0) {
+    return DAG.getConstantFP(0.0, VT);
+  }
+
   // In unsafe math mode, we can fold chains of FADD's of the same value
   // into multiplications.  This transform is not safe in general because
   // we are reducing the number of rounding steps.
@@ -5890,7 +5900,7 @@
   }
 
   // fold (fsub c1, c2) -> c1-c2
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)
+  if (N0CFP && N1CFP)
     return DAG.getNode(ISD::FSUB, N->getDebugLoc(), VT, N0, N1);
   // fold (fsub A, 0) -> A
   if (DAG.getTarget().Options.UnsafeFPMath &&
@@ -5982,7 +5992,7 @@
   }
 
   // fold (fmul c1, c2) -> c1*c2
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)
+  if (N0CFP && N1CFP)
     return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT, N0, N1);
   // canonicalize constant to RHS
   if (N0CFP && !N1CFP)
@@ -6040,6 +6050,12 @@
   EVT VT = N->getValueType(0);
   DebugLoc dl = N->getDebugLoc();
 
+  if (DAG.getTarget().Options.UnsafeFPMath) {
+    if (N0CFP && N0CFP->isZero())
+      return N2;
+    if (N1CFP && N1CFP->isZero())
+      return N2;
+  }
   if (N0CFP && N0CFP->isExactlyValue(1.0))
     return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N1, N2);
   if (N1CFP && N1CFP->isExactlyValue(1.0))
@@ -6119,11 +6135,11 @@
   }
 
   // fold (fdiv c1, c2) -> c1/c2
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)
+  if (N0CFP && N1CFP)
     return DAG.getNode(ISD::FDIV, N->getDebugLoc(), VT, N0, N1);
 
   // fold (fdiv X, c2) -> fmul X, 1/c2 if losing precision is acceptable.
-  if (N1CFP && VT != MVT::ppcf128 && DAG.getTarget().Options.UnsafeFPMath) {
+  if (N1CFP && DAG.getTarget().Options.UnsafeFPMath) {
     // Compute the reciprocal 1.0 / c2.
     APFloat N1APF = N1CFP->getValueAPF();
     APFloat Recip(N1APF.getSemantics(), 1); // 1.0
@@ -6166,7 +6182,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (frem c1, c2) -> fmod(c1,c2)
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)
+  if (N0CFP && N1CFP)
     return DAG.getNode(ISD::FREM, N->getDebugLoc(), VT, N0, N1);
 
   return SDValue();
@@ -6179,7 +6195,7 @@
   ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
   EVT VT = N->getValueType(0);
 
-  if (N0CFP && N1CFP && VT != MVT::ppcf128)  // Constant fold
+  if (N0CFP && N1CFP)  // Constant fold
     return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT, N0, N1);
 
   if (N1CFP) {
@@ -6229,7 +6245,7 @@
   EVT OpVT = N0.getValueType();
 
   // fold (sint_to_fp c1) -> c1fp
-  if (N0C && OpVT != MVT::ppcf128 &&
+  if (N0C &&
       // ...but only if the target supports immediate floating-point values
       (!LegalOperations ||
        TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT)))
@@ -6286,7 +6302,7 @@
   EVT OpVT = N0.getValueType();
 
   // fold (uint_to_fp c1) -> c1fp
-  if (N0C && OpVT != MVT::ppcf128 &&
+  if (N0C &&
       // ...but only if the target supports immediate floating-point values
       (!LegalOperations ||
        TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT)))
@@ -6341,7 +6357,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (fp_to_uint c1fp) -> c1
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FP_TO_UINT, N->getDebugLoc(), VT, N0);
 
   return SDValue();
@@ -6354,7 +6370,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (fp_round c1fp) -> c1fp
-  if (N0CFP && N0.getValueType() != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FP_ROUND, N->getDebugLoc(), VT, N0, N1);
 
   // fold (fp_round (fp_extend x)) -> x
@@ -6408,7 +6424,7 @@
     return SDValue();
 
   // fold (fp_extend c1fp) -> c1fp
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FP_EXTEND, N->getDebugLoc(), VT, N0);
 
   // Turn fp_extend(fp_round(X, 1)) -> x since the fp_round doesn't affect the
@@ -6495,7 +6511,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (fceil c1) -> fceil(c1)
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FCEIL, N->getDebugLoc(), VT, N0);
 
   return SDValue();
@@ -6507,7 +6523,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (ftrunc c1) -> ftrunc(c1)
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FTRUNC, N->getDebugLoc(), VT, N0);
 
   return SDValue();
@@ -6519,7 +6535,7 @@
   EVT VT = N->getValueType(0);
 
   // fold (ffloor c1) -> ffloor(c1)
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FFLOOR, N->getDebugLoc(), VT, N0);
 
   return SDValue();
@@ -6536,7 +6552,7 @@
   }
 
   // fold (fabs c1) -> fabs(c1)
-  if (N0CFP && VT != MVT::ppcf128)
+  if (N0CFP)
     return DAG.getNode(ISD::FABS, N->getDebugLoc(), VT, N0);
   // fold (fabs (fabs x)) -> (fabs x)
   if (N0.getOpcode() == ISD::FABS)
@@ -8356,15 +8372,21 @@
   return SDValue();
 }
 
-SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
+// Simplify (build_vec (ext )) to (bitcast (build_vec ))
+SDValue DAGCombiner::reduceBuildVecExtToExtBuildVec(SDNode *N) {
+  // We perform this optimization post type-legalization because
+  // the type-legalizer often scalarizes integer-promoted vectors.
+  // Performing this optimization before may create bit-casts which
+  // will be type-legalized to complex code sequences.
+  // We perform this optimization only before the operation legalizer because we
+  // may introduce illegal operations.
+  if (Level != AfterLegalizeVectorOps && Level != AfterLegalizeTypes)
+    return SDValue();
+
   unsigned NumInScalars = N->getNumOperands();
   DebugLoc dl = N->getDebugLoc();
   EVT VT = N->getValueType(0);
 
-  // A vector built entirely of undefs is undef.
-  if (ISD::allOperandsUndef(N))
-    return DAG.getUNDEF(VT);
-
   // Check to see if this is a BUILD_VECTOR of a bunch of values
   // which come from any_extend or zero_extend nodes. If so, we can create
   // a new BUILD_VECTOR using bit-casts which may enable other BUILD_VECTOR
@@ -8407,65 +8429,142 @@
   // In order to have valid types, all of the inputs must be extended from the
   // same source type and all of the inputs must be any or zero extend.
   // Scalar sizes must be a power of two.
-  EVT OutScalarTy = N->getValueType(0).getScalarType();
+  EVT OutScalarTy = VT.getScalarType();
   bool ValidTypes = SourceType != MVT::Other &&
                  isPowerOf2_32(OutScalarTy.getSizeInBits()) &&
                  isPowerOf2_32(SourceType.getSizeInBits());
 
-  // We perform this optimization post type-legalization because
-  // the type-legalizer often scalarizes integer-promoted vectors.
-  // Performing this optimization before may create bit-casts which
-  // will be type-legalized to complex code sequences.
-  // We perform this optimization only before the operation legalizer because we
-  // may introduce illegal operations.
   // Create a new simpler BUILD_VECTOR sequence which other optimizations can
   // turn into a single shuffle instruction.
-  if ((Level == AfterLegalizeVectorOps || Level == AfterLegalizeTypes) &&
-      ValidTypes) {
-    bool isLE = TLI.isLittleEndian();
-    unsigned ElemRatio = OutScalarTy.getSizeInBits()/SourceType.getSizeInBits();
-    assert(ElemRatio > 1 && "Invalid element size ratio");
-    SDValue Filler = AllAnyExt ? DAG.getUNDEF(SourceType):
-                                 DAG.getConstant(0, SourceType);
-
-    unsigned NewBVElems = ElemRatio * N->getValueType(0).getVectorNumElements();
-    SmallVector<SDValue, 8> Ops(NewBVElems, Filler);
-
-    // Populate the new build_vector
-    for (unsigned i=0; i < N->getNumOperands(); ++i) {
-      SDValue Cast = N->getOperand(i);
-      assert((Cast.getOpcode() == ISD::ANY_EXTEND ||
-              Cast.getOpcode() == ISD::ZERO_EXTEND ||
-              Cast.getOpcode() == ISD::UNDEF) && "Invalid cast opcode");
-      SDValue In;
-      if (Cast.getOpcode() == ISD::UNDEF)
-        In = DAG.getUNDEF(SourceType);
-      else
-        In = Cast->getOperand(0);
-      unsigned Index = isLE ? (i * ElemRatio) :
-                              (i * ElemRatio + (ElemRatio - 1));
-
-      assert(Index < Ops.size() && "Invalid index");
-      Ops[Index] = In;
-    }
-
-    // The type of the new BUILD_VECTOR node.
-    EVT VecVT = EVT::getVectorVT(*DAG.getContext(), SourceType, NewBVElems);
-    assert(VecVT.getSizeInBits() == N->getValueType(0).getSizeInBits() &&
-           "Invalid vector size");
-    // Check if the new vector type is legal.
-    if (!isTypeLegal(VecVT)) return SDValue();
-
-    // Make the new BUILD_VECTOR.
-    SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(),
-                                 VecVT, &Ops[0], Ops.size());
-
-    // The new BUILD_VECTOR node has the potential to be further optimized.
-    AddToWorkList(BV.getNode());
-    // Bitcast to the desired type.
-    return DAG.getNode(ISD::BITCAST, dl, N->getValueType(0), BV);
+  if (!ValidTypes)
+    return SDValue();
+
+  bool isLE = TLI.isLittleEndian();
+  unsigned ElemRatio = OutScalarTy.getSizeInBits()/SourceType.getSizeInBits();
+  assert(ElemRatio > 1 && "Invalid element size ratio");
+  SDValue Filler = AllAnyExt ? DAG.getUNDEF(SourceType):
+                               DAG.getConstant(0, SourceType);
+
+  unsigned NewBVElems = ElemRatio * VT.getVectorNumElements();
+  SmallVector<SDValue, 8> Ops(NewBVElems, Filler);
+
+  // Populate the new build_vector
+  for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) {
+    SDValue Cast = N->getOperand(i);
+    assert((Cast.getOpcode() == ISD::ANY_EXTEND ||
+            Cast.getOpcode() == ISD::ZERO_EXTEND ||
+            Cast.getOpcode() == ISD::UNDEF) && "Invalid cast opcode");
+    SDValue In;
+    if (Cast.getOpcode() == ISD::UNDEF)
+      In = DAG.getUNDEF(SourceType);
+    else
+      In = Cast->getOperand(0);
+    unsigned Index = isLE ? (i * ElemRatio) :
+                            (i * ElemRatio + (ElemRatio - 1));
+
+    assert(Index < Ops.size() && "Invalid index");
+    Ops[Index] = In;
+  }
+
+  // The type of the new BUILD_VECTOR node.
+  EVT VecVT = EVT::getVectorVT(*DAG.getContext(), SourceType, NewBVElems);
+  assert(VecVT.getSizeInBits() == VT.getSizeInBits() &&
+         "Invalid vector size");
+  // Check if the new vector type is legal.
+  if (!isTypeLegal(VecVT)) return SDValue();
+
+  // Make the new BUILD_VECTOR.
+  SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, dl, VecVT, &Ops[0], Ops.size());
+
+  // The new BUILD_VECTOR node has the potential to be further optimized.
+  AddToWorkList(BV.getNode());
+  // Bitcast to the desired type.
+  return DAG.getNode(ISD::BITCAST, dl, VT, BV);
+}
+
+SDValue DAGCombiner::reduceBuildVecConvertToConvertBuildVec(SDNode *N) {
+  EVT VT = N->getValueType(0);
+
+  unsigned NumInScalars = N->getNumOperands();
+  DebugLoc dl = N->getDebugLoc();
+
+  EVT SrcVT = MVT::Other;
+  unsigned Opcode = ISD::DELETED_NODE;
+  unsigned NumDefs = 0;
+
+  for (unsigned i = 0; i != NumInScalars; ++i) {
+    SDValue In = N->getOperand(i);
+    unsigned Opc = In.getOpcode();
+
+    if (Opc == ISD::UNDEF)
+      continue;
+
+    // If all scalar values are floats and converted from integers.
+    if (Opcode == ISD::DELETED_NODE &&
+        (Opc == ISD::UINT_TO_FP || Opc == ISD::SINT_TO_FP)) {
+      Opcode = Opc;
+      // If not supported by target, bail out.
+      if (TLI.getOperationAction(Opcode, VT) != TargetLowering::Legal &&
+          TLI.getOperationAction(Opcode, VT) != TargetLowering::Custom)
+        return SDValue();
+    }
+    if (Opc != Opcode)
+      return SDValue();
+
+    EVT InVT = In.getOperand(0).getValueType();
+
+    // If all scalar values are typed differently, bail out. It's chosen to
+    // simplify BUILD_VECTOR of integer types.
+    if (SrcVT == MVT::Other)
+      SrcVT = InVT;
+    if (SrcVT != InVT)
+      return SDValue();
+    NumDefs++;
   }
 
+  // If the vector has just one element defined, it's not worth to fold it into
+  // a vectorized one.
+  if (NumDefs < 2)
+    return SDValue();
+
+  assert((Opcode == ISD::UINT_TO_FP || Opcode == ISD::SINT_TO_FP)
+         && "Should only handle conversion from integer to float.");
+  assert(SrcVT != MVT::Other && "Cannot determine source type!");
+
+  EVT NVT = EVT::getVectorVT(*DAG.getContext(), SrcVT, NumInScalars);
+  SmallVector<SDValue, 8> Opnds;
+  for (unsigned i = 0; i != NumInScalars; ++i) {
+    SDValue In = N->getOperand(i);
+
+    if (In.getOpcode() == ISD::UNDEF)
+      Opnds.push_back(DAG.getUNDEF(SrcVT));
+    else
+      Opnds.push_back(In.getOperand(0));
+  }
+  SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, dl, NVT,
+                           &Opnds[0], Opnds.size());
+  AddToWorkList(BV.getNode());
+
+  return DAG.getNode(Opcode, dl, VT, BV);
+}
+
+SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
+  unsigned NumInScalars = N->getNumOperands();
+  DebugLoc dl = N->getDebugLoc();
+  EVT VT = N->getValueType(0);
+
+  // A vector built entirely of undefs is undef.
+  if (ISD::allOperandsUndef(N))
+    return DAG.getUNDEF(VT);
+
+  SDValue V = reduceBuildVecExtToExtBuildVec(N);
+  if (V.getNode())
+    return V;
+
+  V = reduceBuildVecConvertToConvertBuildVec(N);
+  if (V.getNode())
+    return V;
+
   // Check to see if this is a BUILD_VECTOR of a bunch of EXTRACT_VECTOR_ELT
   // operations.  If so, and if the EXTRACT_VECTOR_ELT vector inputs come from
   // at most two distinct vectors, turn this into a shuffle node.
@@ -8549,7 +8648,7 @@
         return SDValue();
 
       // Widen the input vector by adding undef values.
-      VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, N->getDebugLoc(), VT,
+      VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, dl, VT,
                            VecIn1, DAG.getUNDEF(VecIn1.getValueType()));
     }
 
@@ -8570,7 +8669,7 @@
     SDValue Ops[2];
     Ops[0] = VecIn1;
     Ops[1] = VecIn2;
-    return DAG.getVectorShuffle(VT, N->getDebugLoc(), Ops[0], Ops[1], &Mask[0]);
+    return DAG.getVectorShuffle(VT, dl, Ops[0], Ops[1], &Mask[0]);
   }
 
   return SDValue();
@@ -9313,34 +9412,38 @@
       return SDValue();
 
     // Get a SetCC of the condition
-    // FIXME: Should probably make sure that setcc is legal if we ever have a
-    // target where it isn't.
-    SDValue Temp, SCC;
-    // cast from setcc result type to select result type
-    if (LegalTypes) {
-      SCC  = DAG.getSetCC(DL, TLI.getSetCCResultType(N0.getValueType()),
-                          N0, N1, CC);
-      if (N2.getValueType().bitsLT(SCC.getValueType()))
-        Temp = DAG.getZeroExtendInReg(SCC, N2.getDebugLoc(), N2.getValueType());
-      else
+    // NOTE: Don't create a SETCC if it's not legal on this target.
+    if (!LegalOperations ||
+        TLI.isOperationLegal(ISD::SETCC,
+          LegalTypes ? TLI.getSetCCResultType(N0.getValueType()) : MVT::i1)) {
+      SDValue Temp, SCC;
+      // cast from setcc result type to select result type
+      if (LegalTypes) {
+        SCC  = DAG.getSetCC(DL, TLI.getSetCCResultType(N0.getValueType()),
+                            N0, N1, CC);
+        if (N2.getValueType().bitsLT(SCC.getValueType()))
+          Temp = DAG.getZeroExtendInReg(SCC, N2.getDebugLoc(),
+                                        N2.getValueType());
+        else
+          Temp = DAG.getNode(ISD::ZERO_EXTEND, N2.getDebugLoc(),
+                             N2.getValueType(), SCC);
+      } else {
+        SCC  = DAG.getSetCC(N0.getDebugLoc(), MVT::i1, N0, N1, CC);
         Temp = DAG.getNode(ISD::ZERO_EXTEND, N2.getDebugLoc(),
                            N2.getValueType(), SCC);
-    } else {
-      SCC  = DAG.getSetCC(N0.getDebugLoc(), MVT::i1, N0, N1, CC);
-      Temp = DAG.getNode(ISD::ZERO_EXTEND, N2.getDebugLoc(),
-                         N2.getValueType(), SCC);
-    }
+      }
 
-    AddToWorkList(SCC.getNode());
-    AddToWorkList(Temp.getNode());
+      AddToWorkList(SCC.getNode());
+      AddToWorkList(Temp.getNode());
 
-    if (N2C->getAPIntValue() == 1)
-      return Temp;
+      if (N2C->getAPIntValue() == 1)
+        return Temp;
 
-    // shl setcc result by log2 n2c
-    return DAG.getNode(ISD::SHL, DL, N2.getValueType(), Temp,
-                       DAG.getConstant(N2C->getAPIntValue().logBase2(),
-                                       getShiftAmountTy(Temp.getValueType())));
+      // shl setcc result by log2 n2c
+      return DAG.getNode(ISD::SHL, DL, N2.getValueType(), Temp,
+                         DAG.getConstant(N2C->getAPIntValue().logBase2(),
+                                         getShiftAmountTy(Temp.getValueType())));
+    }
   }
 
   // Check to see if this is the equivalent of setcc

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/InstrEmitter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/InstrEmitter.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/InstrEmitter.cpp Tue Nov 13 09:21:47 2012
@@ -897,7 +897,8 @@
     const char *AsmStr = cast<ExternalSymbolSDNode>(AsmStrV)->getSymbol();
     MI->addOperand(MachineOperand::CreateES(AsmStr));
 
-    // Add the HasSideEffect and isAlignStack bits.
+    // Add the HasSideEffect, isAlignStack, AsmDialect, MayLoad and MayStore
+    // bits.
     int64_t ExtraInfo =
       cast<ConstantSDNode>(Node->getOperand(InlineAsm::Op_ExtraInfo))->
                           getZExtValue();

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGFast.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGFast.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGFast.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGFast.cpp Tue Nov 13 09:21:47 2012
@@ -336,7 +336,9 @@
       }
     }
     if (isNewLoad) {
-      AddPred(NewSU, SDep(LoadSU, SDep::Order, LoadSU->Latency));
+      SDep D(LoadSU, SDep::Barrier);
+      D.setLatency(LoadSU->Latency);
+      AddPred(NewSU, D);
     }
 
     ++NumUnfolds;
@@ -412,9 +414,12 @@
   for (unsigned i = 0, e = DelDeps.size(); i != e; ++i) {
     RemovePred(DelDeps[i].first, DelDeps[i].second);
   }
-
-  AddPred(CopyFromSU, SDep(SU, SDep::Data, SU->Latency, Reg));
-  AddPred(CopyToSU, SDep(CopyFromSU, SDep::Data, CopyFromSU->Latency, 0));
+  SDep FromDep(SU, SDep::Data, Reg);
+  FromDep.setLatency(SU->Latency);
+  AddPred(CopyFromSU, FromDep);
+  SDep ToDep(CopyFromSU, SDep::Data, 0);
+  ToDep.setLatency(CopyFromSU->Latency);
+  AddPred(CopyToSU, ToDep);
 
   Copies.push_back(CopyFromSU);
   Copies.push_back(CopyToSU);
@@ -591,18 +596,14 @@
           InsertCopiesAndMoveSuccs(LRDef, Reg, DestRC, RC, Copies);
           DEBUG(dbgs() << "Adding an edge from SU # " << TrySU->NodeNum
                        << " to SU #" << Copies.front()->NodeNum << "\n");
-          AddPred(TrySU, SDep(Copies.front(), SDep::Order, /*Latency=*/1,
-                              /*Reg=*/0, /*isNormalMemory=*/false,
-                              /*isMustAlias=*/false, /*isArtificial=*/true));
+          AddPred(TrySU, SDep(Copies.front(), SDep::Artificial));
           NewDef = Copies.back();
         }
 
         DEBUG(dbgs() << "Adding an edge from SU # " << NewDef->NodeNum
                      << " to SU #" << TrySU->NodeNum << "\n");
         LiveRegDefs[Reg] = NewDef;
-        AddPred(NewDef, SDep(TrySU, SDep::Order, /*Latency=*/1,
-                             /*Reg=*/0, /*isNormalMemory=*/false,
-                             /*isMustAlias=*/false, /*isArtificial=*/true));
+        AddPred(NewDef, SDep(TrySU, SDep::Artificial));
         TrySU->isAvailable = false;
         CurSU = NewDef;
       }

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp Tue Nov 13 09:21:47 2012
@@ -1058,7 +1058,9 @@
 
     // Add a data dependency to reflect that NewSU reads the value defined
     // by LoadSU.
-    AddPred(NewSU, SDep(LoadSU, SDep::Data, LoadSU->Latency));
+    SDep D(LoadSU, SDep::Data, 0);
+    D.setLatency(LoadSU->Latency);
+    AddPred(NewSU, D);
 
     if (isNewLoad)
       AvailableQueue->addNode(LoadSU);
@@ -1140,17 +1142,18 @@
       // Avoid scheduling the def-side copy before other successors. Otherwise
       // we could introduce another physreg interference on the copy and
       // continue inserting copies indefinitely.
-      SDep D(CopyFromSU, SDep::Order, /*Latency=*/0,
-             /*Reg=*/0, /*isNormalMemory=*/false,
-             /*isMustAlias=*/false, /*isArtificial=*/true);
-      AddPred(SuccSU, D);
+      AddPred(SuccSU, SDep(CopyFromSU, SDep::Artificial));
     }
   }
   for (unsigned i = 0, e = DelDeps.size(); i != e; ++i)
     RemovePred(DelDeps[i].first, DelDeps[i].second);
 
-  AddPred(CopyFromSU, SDep(SU, SDep::Data, SU->Latency, Reg));
-  AddPred(CopyToSU, SDep(CopyFromSU, SDep::Data, CopyFromSU->Latency, 0));
+  SDep FromDep(SU, SDep::Data, Reg);
+  FromDep.setLatency(SU->Latency);
+  AddPred(CopyFromSU, FromDep);
+  SDep ToDep(CopyFromSU, SDep::Data, 0);
+  ToDep.setLatency(CopyFromSU->Latency);
+  AddPred(CopyToSU, ToDep);
 
   AvailableQueue->updateNode(SU);
   AvailableQueue->addNode(CopyFromSU);
@@ -1359,9 +1362,7 @@
         if (!BtSU->isPending)
           AvailableQueue->remove(BtSU);
       }
-      AddPred(TrySU, SDep(BtSU, SDep::Order, /*Latency=*/1,
-                          /*Reg=*/0, /*isNormalMemory=*/false,
-                          /*isMustAlias=*/false, /*isArtificial=*/true));
+      AddPred(TrySU, SDep(BtSU, SDep::Artificial));
 
       // If one or more successors has been unscheduled, then the current
       // node is no longer avaialable. Schedule a successor that's now
@@ -1413,20 +1414,14 @@
       InsertCopiesAndMoveSuccs(LRDef, Reg, DestRC, RC, Copies);
       DEBUG(dbgs() << "    Adding an edge from SU #" << TrySU->NodeNum
             << " to SU #" << Copies.front()->NodeNum << "\n");
-      AddPred(TrySU, SDep(Copies.front(), SDep::Order, /*Latency=*/1,
-                          /*Reg=*/0, /*isNormalMemory=*/false,
-                          /*isMustAlias=*/false,
-                          /*isArtificial=*/true));
+      AddPred(TrySU, SDep(Copies.front(), SDep::Artificial));
       NewDef = Copies.back();
     }
 
     DEBUG(dbgs() << "    Adding an edge from SU #" << NewDef->NodeNum
           << " to SU #" << TrySU->NodeNum << "\n");
     LiveRegDefs[Reg] = NewDef;
-    AddPred(NewDef, SDep(TrySU, SDep::Order, /*Latency=*/1,
-                         /*Reg=*/0, /*isNormalMemory=*/false,
-                         /*isMustAlias=*/false,
-                         /*isArtificial=*/true));
+    AddPred(NewDef, SDep(TrySU, SDep::Artificial));
     TrySU->isAvailable = false;
     CurSU = NewDef;
   }
@@ -2936,10 +2931,7 @@
             !scheduleDAG->IsReachable(SuccSU, SU)) {
           DEBUG(dbgs() << "    Adding a pseudo-two-addr edge from SU #"
                        << SU->NodeNum << " to SU #" << SuccSU->NodeNum << "\n");
-          scheduleDAG->AddPred(SU, SDep(SuccSU, SDep::Order, /*Latency=*/0,
-                                        /*Reg=*/0, /*isNormalMemory=*/false,
-                                        /*isMustAlias=*/false,
-                                        /*isArtificial=*/true));
+          scheduleDAG->AddPred(SU, SDep(SuccSU, SDep::Artificial));
         }
       }
     }

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp Tue Nov 13 09:21:47 2012
@@ -485,14 +485,15 @@
         if(isChain && OpN->getOpcode() == ISD::TokenFactor)
           OpLatency = 0;
 
-        const SDep &dep = SDep(OpSU, isChain ? SDep::Order : SDep::Data,
-                               OpLatency, PhysReg);
+        SDep Dep = isChain ? SDep(OpSU, SDep::Barrier)
+          : SDep(OpSU, SDep::Data, PhysReg);
+        Dep.setLatency(OpLatency);
         if (!isChain && !UnitLatencies) {
-          computeOperandLatency(OpN, N, i, const_cast<SDep &>(dep));
-          ST.adjustSchedDependency(OpSU, SU, const_cast<SDep &>(dep));
+          computeOperandLatency(OpN, N, i, Dep);
+          ST.adjustSchedDependency(OpSU, SU, Dep);
         }
 
-        if (!SU->addPred(dep) && !dep.isCtrl() && OpSU->NumRegDefsLeft > 1) {
+        if (!SU->addPred(Dep) && !Dep.isCtrl() && OpSU->NumRegDefsLeft > 1) {
           // Multiple register uses are combined in the same SUnit. For example,
           // we could have a set of glued nodes with all their defs consumed by
           // another set of glued nodes. Register pressure tracking sees this as

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Tue Nov 13 09:21:47 2012
@@ -91,11 +91,6 @@
                                            const APFloat& Val) {
   assert(VT.isFloatingPoint() && "Can only convert between FP types");
 
-  // PPC long double cannot be converted to any other type.
-  if (VT == MVT::ppcf128 ||
-      &Val.getSemantics() == &APFloat::PPCDoubleDouble)
-    return false;
-
   // convert modifies in place, so make a copy.
   APFloat Val2 = APFloat(Val);
   bool losesInfo;
@@ -1612,10 +1607,6 @@
   }
   if (ConstantFPSDNode *N1C = dyn_cast<ConstantFPSDNode>(N1.getNode())) {
     if (ConstantFPSDNode *N2C = dyn_cast<ConstantFPSDNode>(N2.getNode())) {
-      // No compile time operations on this type yet.
-      if (N1C->getValueType(0) == MVT::ppcf128)
-        return SDValue();
-
       APFloat::cmpResult R = N1C->getValueAPF().compare(N2C->getValueAPF());
       switch (Cond) {
       default: break;
@@ -2447,8 +2438,6 @@
       return getConstant(Val.zextOrTrunc(VT.getSizeInBits()), VT);
     case ISD::UINT_TO_FP:
     case ISD::SINT_TO_FP: {
-      // No compile time operations on ppcf128.
-      if (VT == MVT::ppcf128) break;
       APFloat apf(APInt::getNullValue(VT.getSizeInBits()));
       (void)apf.convertFromAPInt(Val,
                                  Opcode==ISD::SINT_TO_FP,
@@ -2477,61 +2466,59 @@
   // Constant fold unary operations with a floating point constant operand.
   if (ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(Operand.getNode())) {
     APFloat V = C->getValueAPF();    // make copy
-    if (VT != MVT::ppcf128 && Operand.getValueType() != MVT::ppcf128) {
-      switch (Opcode) {
-      case ISD::FNEG:
-        V.changeSign();
+    switch (Opcode) {
+    case ISD::FNEG:
+      V.changeSign();
+      return getConstantFP(V, VT);
+    case ISD::FABS:
+      V.clearSign();
+      return getConstantFP(V, VT);
+    case ISD::FCEIL: {
+      APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardPositive);
+      if (fs == APFloat::opOK || fs == APFloat::opInexact)
         return getConstantFP(V, VT);
-      case ISD::FABS:
-        V.clearSign();
+      break;
+    }
+    case ISD::FTRUNC: {
+      APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardZero);
+      if (fs == APFloat::opOK || fs == APFloat::opInexact)
         return getConstantFP(V, VT);
-      case ISD::FCEIL: {
-        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardPositive);
-        if (fs == APFloat::opOK || fs == APFloat::opInexact)
-          return getConstantFP(V, VT);
-        break;
-      }
-      case ISD::FTRUNC: {
-        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardZero);
-        if (fs == APFloat::opOK || fs == APFloat::opInexact)
-          return getConstantFP(V, VT);
-        break;
-      }
-      case ISD::FFLOOR: {
-        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardNegative);
-        if (fs == APFloat::opOK || fs == APFloat::opInexact)
-          return getConstantFP(V, VT);
-        break;
-      }
-      case ISD::FP_EXTEND: {
-        bool ignored;
-        // This can return overflow, underflow, or inexact; we don't care.
-        // FIXME need to be more flexible about rounding mode.
-        (void)V.convert(*EVTToAPFloatSemantics(VT),
-                        APFloat::rmNearestTiesToEven, &ignored);
+      break;
+    }
+    case ISD::FFLOOR: {
+      APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardNegative);
+      if (fs == APFloat::opOK || fs == APFloat::opInexact)
         return getConstantFP(V, VT);
-      }
-      case ISD::FP_TO_SINT:
-      case ISD::FP_TO_UINT: {
-        integerPart x[2];
-        bool ignored;
-        assert(integerPartWidth >= 64);
-        // FIXME need to be more flexible about rounding mode.
-        APFloat::opStatus s = V.convertToInteger(x, VT.getSizeInBits(),
-                              Opcode==ISD::FP_TO_SINT,
-                              APFloat::rmTowardZero, &ignored);
-        if (s==APFloat::opInvalidOp)     // inexact is OK, in fact usual
-          break;
-        APInt api(VT.getSizeInBits(), x);
-        return getConstant(api, VT);
-      }
-      case ISD::BITCAST:
-        if (VT == MVT::i32 && C->getValueType(0) == MVT::f32)
-          return getConstant((uint32_t)V.bitcastToAPInt().getZExtValue(), VT);
-        else if (VT == MVT::i64 && C->getValueType(0) == MVT::f64)
-          return getConstant(V.bitcastToAPInt().getZExtValue(), VT);
+      break;
+    }
+    case ISD::FP_EXTEND: {
+      bool ignored;
+      // This can return overflow, underflow, or inexact; we don't care.
+      // FIXME need to be more flexible about rounding mode.
+      (void)V.convert(*EVTToAPFloatSemantics(VT),
+                      APFloat::rmNearestTiesToEven, &ignored);
+      return getConstantFP(V, VT);
+    }
+    case ISD::FP_TO_SINT:
+    case ISD::FP_TO_UINT: {
+      integerPart x[2];
+      bool ignored;
+      assert(integerPartWidth >= 64);
+      // FIXME need to be more flexible about rounding mode.
+      APFloat::opStatus s = V.convertToInteger(x, VT.getSizeInBits(),
+                            Opcode==ISD::FP_TO_SINT,
+                            APFloat::rmTowardZero, &ignored);
+      if (s==APFloat::opInvalidOp)     // inexact is OK, in fact usual
         break;
-      }
+      APInt api(VT.getSizeInBits(), x);
+      return getConstant(api, VT);
+    }
+    case ISD::BITCAST:
+      if (VT == MVT::i32 && C->getValueType(0) == MVT::f32)
+        return getConstant((uint32_t)V.bitcastToAPInt().getZExtValue(), VT);
+      else if (VT == MVT::i64 && C->getValueType(0) == MVT::f64)
+        return getConstant(V.bitcastToAPInt().getZExtValue(), VT);
+      break;
     }
   }
 
@@ -3052,7 +3039,7 @@
       // Cannonicalize constant to RHS if commutative
       std::swap(N1CFP, N2CFP);
       std::swap(N1, N2);
-    } else if (N2CFP && VT != MVT::ppcf128) {
+    } else if (N2CFP) {
       APFloat V1 = N1CFP->getValueAPF(), V2 = N2CFP->getValueAPF();
       APFloat::opStatus s;
       switch (Opcode) {
@@ -3449,12 +3436,9 @@
   EVT VT = TLI.getOptimalMemOpType(Size, DstAlign, SrcAlign,
                                    IsZeroVal, MemcpyStrSrc,
                                    DAG.getMachineFunction());
-  Type *vtType = VT.isExtended() ? VT.getTypeForEVT(*DAG.getContext()) : NULL;
-  unsigned AS = (vtType && vtType->isPointerTy()) ?
-    cast<PointerType>(vtType)->getAddressSpace() : 0;
 
   if (VT == MVT::Other) {
-    if (DstAlign >= TLI.getDataLayout()->getPointerPrefAlignment(AS) ||
+    if (DstAlign >= TLI.getDataLayout()->getPointerPrefAlignment() ||
         TLI.allowsUnalignedMemoryAccesses(VT)) {
       VT = TLI.getPointerTy();
     } else {

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Tue Nov 13 09:21:47 2012
@@ -1255,7 +1255,7 @@
 
         for (unsigned i = 0; i < NumParts; ++i) {
           Outs.push_back(ISD::OutputArg(Flags, Parts[i].getValueType(),
-                                        /*isfixed=*/true));
+                                        /*isfixed=*/true, 0, 0));
           OutVals.push_back(Parts[i]);
         }
       }
@@ -2604,14 +2604,14 @@
   MachineBasicBlock *IndirectBrMBB = FuncInfo.MBB;
 
   // Update machine-CFG edges with unique successors.
-  SmallVector<BasicBlock*, 32> succs;
-  succs.reserve(I.getNumSuccessors());
-  for (unsigned i = 0, e = I.getNumSuccessors(); i != e; ++i)
-    succs.push_back(I.getSuccessor(i));
-  array_pod_sort(succs.begin(), succs.end());
-  succs.erase(std::unique(succs.begin(), succs.end()), succs.end());
-  for (unsigned i = 0, e = succs.size(); i != e; ++i) {
-    MachineBasicBlock *Succ = FuncInfo.MBBMap[succs[i]];
+  SmallSet<BasicBlock*, 32> Done;
+  for (unsigned i = 0, e = I.getNumSuccessors(); i != e; ++i) {
+    BasicBlock *BB = I.getSuccessor(i);
+    bool Inserted = Done.insert(BB);
+    if (!Inserted)
+        continue;
+
+    MachineBasicBlock *Succ = FuncInfo.MBBMap[BB];
     addSuccessorWithWeight(IndirectBrMBB, Succ);
   }
 
@@ -6128,7 +6128,8 @@
   const MDNode *SrcLoc = CS.getInstruction()->getMetadata("srcloc");
   AsmNodeOperands.push_back(DAG.getMDNode(SrcLoc));
 
-  // Remember the HasSideEffect, AlignStack and AsmDialect bits as operand 3.
+  // Remember the HasSideEffect, AlignStack, AsmDialect, MayLoad and MayStore
+  // bits as operand 3.
   unsigned ExtraInfo = 0;
   if (IA->hasSideEffects())
     ExtraInfo |= InlineAsm::Extra_HasSideEffects;
@@ -6136,6 +6137,27 @@
     ExtraInfo |= InlineAsm::Extra_IsAlignStack;
   // Set the asm dialect.
   ExtraInfo |= IA->getDialect() * InlineAsm::Extra_AsmDialect;
+
+  // Determine if this InlineAsm MayLoad or MayStore based on the constraints.
+  for (unsigned i = 0, e = TargetConstraints.size(); i != e; ++i) {
+    TargetLowering::AsmOperandInfo &OpInfo = TargetConstraints[i];
+
+    // Compute the constraint code and ConstraintType to use.
+    TLI.ComputeConstraintToUse(OpInfo, SDValue());
+
+    // Ideally, we would only check against memory constraints.  However, the
+    // meaning of an other constraint can be target-specific and we can't easily
+    // reason about it.  Therefore, be conservative and set MayLoad/MayStore
+    // for other constriants as well.
+    if (OpInfo.ConstraintType == TargetLowering::C_Memory ||
+        OpInfo.ConstraintType == TargetLowering::C_Other) {
+      if (OpInfo.Type == InlineAsm::isInput)
+        ExtraInfo |= InlineAsm::Extra_MayLoad;
+      else if (OpInfo.Type == InlineAsm::isOutput)
+        ExtraInfo |= InlineAsm::Extra_MayStore;
+    }
+  }
+
   AsmNodeOperands.push_back(DAG.getTargetConstant(ExtraInfo,
                                                   TLI.getPointerTy()));
 
@@ -6518,7 +6540,8 @@
       for (unsigned j = 0; j != NumParts; ++j) {
         // if it isn't first piece, alignment must be 1
         ISD::OutputArg MyFlags(Flags, Parts[j].getValueType(),
-                               i < CLI.NumFixedArgs);
+                               i < CLI.NumFixedArgs,
+                               i, j*Parts[j].getValueType().getStoreSize());
         if (NumParts > 1 && j == 0)
           MyFlags.Flags.setSplit();
         else if (j != 0)

Modified: llvm/branches/R600/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/SelectionDAG/TargetLowering.cpp Tue Nov 13 09:21:47 2012
@@ -1032,7 +1032,7 @@
       Flags.setZExt();
 
     for (unsigned i = 0; i < NumParts; ++i)
-      Outs.push_back(ISD::OutputArg(Flags, PartVT, /*isFixed=*/true));
+      Outs.push_back(ISD::OutputArg(Flags, PartVT, /*isFixed=*/true, 0, 0));
   }
 }
 

Modified: llvm/branches/R600/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Tue Nov 13 09:21:47 2012
@@ -77,9 +77,9 @@
                                                     Flags,
                                                     SectionKind::getDataRel(),
                                                     0, Label->getName());
-  unsigned Size = TM.getDataLayout()->getPointerSize(0);
+  unsigned Size = TM.getDataLayout()->getPointerSize();
   Streamer.SwitchSection(Sec);
-  Streamer.EmitValueToAlignment(TM.getDataLayout()->getPointerABIAlignment(0));
+  Streamer.EmitValueToAlignment(TM.getDataLayout()->getPointerABIAlignment());
   Streamer.EmitSymbolAttribute(Label, MCSA_ELF_TypeObject);
   const MCExpr *E = MCConstantExpr::Create(Size, getContext());
   Streamer.EmitELFSize(Label, E);

Modified: llvm/branches/R600/lib/CodeGen/TargetSchedule.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/TargetSchedule.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/TargetSchedule.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/TargetSchedule.cpp Tue Nov 13 09:21:47 2012
@@ -36,6 +36,21 @@
   return EnableSchedItins && !InstrItins.isEmpty();
 }
 
+static unsigned gcd(unsigned Dividend, unsigned Divisor) {
+  // Dividend and Divisor will be naturally swapped as needed.
+  while(Divisor) {
+    unsigned Rem = Dividend % Divisor;
+    Dividend = Divisor;
+    Divisor = Rem;
+  };
+  return Dividend;
+}
+static unsigned lcm(unsigned A, unsigned B) {
+  unsigned LCM = (uint64_t(A) * B) / gcd(A, B);
+  assert((LCM >= A && LCM >= B) && "LCM overflow");
+  return LCM;
+}
+
 void TargetSchedModel::init(const MCSchedModel &sm,
                             const TargetSubtargetInfo *sti,
                             const TargetInstrInfo *tii) {
@@ -43,17 +58,33 @@
   STI = sti;
   TII = tii;
   STI->initInstrItins(InstrItins);
+
+  unsigned NumRes = SchedModel.getNumProcResourceKinds();
+  ResourceFactors.resize(NumRes);
+  ResourceLCM = SchedModel.IssueWidth;
+  for (unsigned Idx = 0; Idx < NumRes; ++Idx) {
+    unsigned NumUnits = SchedModel.getProcResource(Idx)->NumUnits;
+    if (NumUnits > 0)
+      ResourceLCM = lcm(ResourceLCM, NumUnits);
+  }
+  MicroOpFactor = ResourceLCM / SchedModel.IssueWidth;
+  for (unsigned Idx = 0; Idx < NumRes; ++Idx) {
+    unsigned NumUnits = SchedModel.getProcResource(Idx)->NumUnits;
+    ResourceFactors[Idx] = NumUnits ? (ResourceLCM / NumUnits) : 0;
+  }
 }
 
-unsigned TargetSchedModel::getNumMicroOps(MachineInstr *MI) const {
+unsigned TargetSchedModel::getNumMicroOps(const MachineInstr *MI,
+                                          const MCSchedClassDesc *SC) const {
   if (hasInstrItineraries()) {
     int UOps = InstrItins.getNumMicroOps(MI->getDesc().getSchedClass());
     return (UOps >= 0) ? UOps : TII->getNumMicroOps(&InstrItins, MI);
   }
   if (hasInstrSchedModel()) {
-    const MCSchedClassDesc *SCDesc = resolveSchedClass(MI);
-    if (SCDesc->isValid())
-      return SCDesc->NumMicroOps;
+    if (!SC)
+      SC = resolveSchedClass(MI);
+    if (SC->isValid())
+      return SC->NumMicroOps;
   }
   return MI->isTransient() ? 0 : 1;
 }

Modified: llvm/branches/R600/lib/CodeGen/TwoAddressInstructionPass.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/CodeGen/TwoAddressInstructionPass.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/CodeGen/TwoAddressInstructionPass.cpp (original)
+++ llvm/branches/R600/lib/CodeGen/TwoAddressInstructionPass.cpp Tue Nov 13 09:21:47 2012
@@ -60,116 +60,108 @@
 STATISTIC(NumReSchedDowns,     "Number of instructions re-scheduled down");
 
 namespace {
-  class TwoAddressInstructionPass : public MachineFunctionPass {
-    MachineFunction *MF;
-    const TargetInstrInfo *TII;
-    const TargetRegisterInfo *TRI;
-    const InstrItineraryData *InstrItins;
-    MachineRegisterInfo *MRI;
-    LiveVariables *LV;
-    SlotIndexes *Indexes;
-    LiveIntervals *LIS;
-    AliasAnalysis *AA;
-    CodeGenOpt::Level OptLevel;
-
-    // DistanceMap - Keep track the distance of a MI from the start of the
-    // current basic block.
-    DenseMap<MachineInstr*, unsigned> DistanceMap;
-
-    // SrcRegMap - A map from virtual registers to physical registers which
-    // are likely targets to be coalesced to due to copies from physical
-    // registers to virtual registers. e.g. v1024 = move r0.
-    DenseMap<unsigned, unsigned> SrcRegMap;
-
-    // DstRegMap - A map from virtual registers to physical registers which
-    // are likely targets to be coalesced to due to copies to physical
-    // registers from virtual registers. e.g. r1 = move v1024.
-    DenseMap<unsigned, unsigned> DstRegMap;
-
-    /// RegSequences - Keep track the list of REG_SEQUENCE instructions seen
-    /// during the initial walk of the machine function.
-    SmallVector<MachineInstr*, 16> RegSequences;
-
-    bool Sink3AddrInstruction(MachineBasicBlock *MBB, MachineInstr *MI,
-                              unsigned Reg,
-                              MachineBasicBlock::iterator OldPos);
+class TwoAddressInstructionPass : public MachineFunctionPass {
+  MachineFunction *MF;
+  const TargetInstrInfo *TII;
+  const TargetRegisterInfo *TRI;
+  const InstrItineraryData *InstrItins;
+  MachineRegisterInfo *MRI;
+  LiveVariables *LV;
+  SlotIndexes *Indexes;
+  LiveIntervals *LIS;
+  AliasAnalysis *AA;
+  CodeGenOpt::Level OptLevel;
+
+  // The current basic block being processed.
+  MachineBasicBlock *MBB;
+
+  // DistanceMap - Keep track the distance of a MI from the start of the
+  // current basic block.
+  DenseMap<MachineInstr*, unsigned> DistanceMap;
 
-    bool NoUseAfterLastDef(unsigned Reg, MachineBasicBlock *MBB, unsigned Dist,
-                           unsigned &LastDef);
+  // Set of already processed instructions in the current block.
+  SmallPtrSet<MachineInstr*, 8> Processed;
 
-    bool isProfitableToCommute(unsigned regA, unsigned regB, unsigned regC,
-                               MachineInstr *MI, MachineBasicBlock *MBB,
-                               unsigned Dist);
+  // SrcRegMap - A map from virtual registers to physical registers which are
+  // likely targets to be coalesced to due to copies from physical registers to
+  // virtual registers. e.g. v1024 = move r0.
+  DenseMap<unsigned, unsigned> SrcRegMap;
 
-    bool CommuteInstruction(MachineBasicBlock::iterator &mi,
-                            MachineFunction::iterator &mbbi,
-                            unsigned RegB, unsigned RegC, unsigned Dist);
-
-    bool isProfitableToConv3Addr(unsigned RegA, unsigned RegB);
-
-    bool ConvertInstTo3Addr(MachineBasicBlock::iterator &mi,
-                            MachineBasicBlock::iterator &nmi,
-                            MachineFunction::iterator &mbbi,
-                            unsigned RegA, unsigned RegB, unsigned Dist);
+  // DstRegMap - A map from virtual registers to physical registers which are
+  // likely targets to be coalesced to due to copies to physical registers from
+  // virtual registers. e.g. r1 = move v1024.
+  DenseMap<unsigned, unsigned> DstRegMap;
 
-    bool isDefTooClose(unsigned Reg, unsigned Dist,
-                       MachineInstr *MI, MachineBasicBlock *MBB);
+  /// RegSequences - Keep track the list of REG_SEQUENCE instructions seen
+  /// during the initial walk of the machine function.
+  SmallVector<MachineInstr*, 16> RegSequences;
 
-    bool RescheduleMIBelowKill(MachineBasicBlock *MBB,
-                               MachineBasicBlock::iterator &mi,
-                               MachineBasicBlock::iterator &nmi,
-                               unsigned Reg);
-    bool RescheduleKillAboveMI(MachineBasicBlock *MBB,
-                               MachineBasicBlock::iterator &mi,
+  bool sink3AddrInstruction(MachineInstr *MI, unsigned Reg,
+                            MachineBasicBlock::iterator OldPos);
+
+  bool noUseAfterLastDef(unsigned Reg, unsigned Dist, unsigned &LastDef);
+
+  bool isProfitableToCommute(unsigned regA, unsigned regB, unsigned regC,
+                             MachineInstr *MI, unsigned Dist);
+
+  bool commuteInstruction(MachineBasicBlock::iterator &mi,
+                          unsigned RegB, unsigned RegC, unsigned Dist);
+
+  bool isProfitableToConv3Addr(unsigned RegA, unsigned RegB);
+
+  bool convertInstTo3Addr(MachineBasicBlock::iterator &mi,
+                          MachineBasicBlock::iterator &nmi,
+                          unsigned RegA, unsigned RegB, unsigned Dist);
+
+  bool isDefTooClose(unsigned Reg, unsigned Dist, MachineInstr *MI);
+
+  bool rescheduleMIBelowKill(MachineBasicBlock::iterator &mi,
+                             MachineBasicBlock::iterator &nmi,
+                             unsigned Reg);
+  bool rescheduleKillAboveMI(MachineBasicBlock::iterator &mi,
+                             MachineBasicBlock::iterator &nmi,
+                             unsigned Reg);
+
+  bool tryInstructionTransform(MachineBasicBlock::iterator &mi,
                                MachineBasicBlock::iterator &nmi,
-                               unsigned Reg);
+                               unsigned SrcIdx, unsigned DstIdx,
+                               unsigned Dist);
 
-    bool TryInstructionTransform(MachineBasicBlock::iterator &mi,
-                                 MachineBasicBlock::iterator &nmi,
-                                 MachineFunction::iterator &mbbi,
-                                 unsigned SrcIdx, unsigned DstIdx,
-                                 unsigned Dist,
-                                 SmallPtrSet<MachineInstr*, 8> &Processed);
-
-    void ScanUses(unsigned DstReg, MachineBasicBlock *MBB,
-                  SmallPtrSet<MachineInstr*, 8> &Processed);
-
-    void ProcessCopy(MachineInstr *MI, MachineBasicBlock *MBB,
-                     SmallPtrSet<MachineInstr*, 8> &Processed);
-
-    typedef SmallVector<std::pair<unsigned, unsigned>, 4> TiedPairList;
-    typedef SmallDenseMap<unsigned, TiedPairList> TiedOperandMap;
-    bool collectTiedOperands(MachineInstr *MI, TiedOperandMap&);
-    void processTiedPairs(MachineInstr *MI, TiedPairList&, unsigned &Dist);
-
-    void CoalesceExtSubRegs(SmallVector<unsigned,4> &Srcs, unsigned DstReg);
-
-    /// EliminateRegSequences - Eliminate REG_SEQUENCE instructions as part
-    /// of the de-ssa process. This replaces sources of REG_SEQUENCE as
-    /// sub-register references of the register defined by REG_SEQUENCE.
-    bool EliminateRegSequences();
-
-  public:
-    static char ID; // Pass identification, replacement for typeid
-    TwoAddressInstructionPass() : MachineFunctionPass(ID) {
-      initializeTwoAddressInstructionPassPass(*PassRegistry::getPassRegistry());
-    }
-
-    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      AU.setPreservesCFG();
-      AU.addRequired<AliasAnalysis>();
-      AU.addPreserved<LiveVariables>();
-      AU.addPreserved<SlotIndexes>();
-      AU.addPreserved<LiveIntervals>();
-      AU.addPreservedID(MachineLoopInfoID);
-      AU.addPreservedID(MachineDominatorsID);
-      MachineFunctionPass::getAnalysisUsage(AU);
-    }
-
-    /// runOnMachineFunction - Pass entry point.
-    bool runOnMachineFunction(MachineFunction&);
-  };
-}
+  void scanUses(unsigned DstReg);
+
+  void processCopy(MachineInstr *MI);
+
+  typedef SmallVector<std::pair<unsigned, unsigned>, 4> TiedPairList;
+  typedef SmallDenseMap<unsigned, TiedPairList> TiedOperandMap;
+  bool collectTiedOperands(MachineInstr *MI, TiedOperandMap&);
+  void processTiedPairs(MachineInstr *MI, TiedPairList&, unsigned &Dist);
+
+  /// eliminateRegSequences - Eliminate REG_SEQUENCE instructions as part of
+  /// the de-ssa process. This replaces sources of REG_SEQUENCE as sub-register
+  /// references of the register defined by REG_SEQUENCE.
+  bool eliminateRegSequences();
+
+public:
+  static char ID; // Pass identification, replacement for typeid
+  TwoAddressInstructionPass() : MachineFunctionPass(ID) {
+    initializeTwoAddressInstructionPassPass(*PassRegistry::getPassRegistry());
+  }
+
+  virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+    AU.setPreservesCFG();
+    AU.addRequired<AliasAnalysis>();
+    AU.addPreserved<LiveVariables>();
+    AU.addPreserved<SlotIndexes>();
+    AU.addPreserved<LiveIntervals>();
+    AU.addPreservedID(MachineLoopInfoID);
+    AU.addPreservedID(MachineDominatorsID);
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
+  /// runOnMachineFunction - Pass entry point.
+  bool runOnMachineFunction(MachineFunction&);
+};
+} // end anonymous namespace
 
 char TwoAddressInstructionPass::ID = 0;
 INITIALIZE_PASS_BEGIN(TwoAddressInstructionPass, "twoaddressinstruction",
@@ -180,13 +172,13 @@
 
 char &llvm::TwoAddressInstructionPassID = TwoAddressInstructionPass::ID;
 
-/// Sink3AddrInstruction - A two-address instruction has been converted to a
+/// sink3AddrInstruction - A two-address instruction has been converted to a
 /// three-address instruction to avoid clobbering a register. Try to sink it
 /// past the instruction that would kill the above mentioned register to reduce
 /// register pressure.
-bool TwoAddressInstructionPass::Sink3AddrInstruction(MachineBasicBlock *MBB,
-                                           MachineInstr *MI, unsigned SavedReg,
-                                           MachineBasicBlock::iterator OldPos) {
+bool TwoAddressInstructionPass::
+sink3AddrInstruction(MachineInstr *MI, unsigned SavedReg,
+                     MachineBasicBlock::iterator OldPos) {
   // FIXME: Shouldn't we be trying to do this before we three-addressify the
   // instruction?  After this transformation is done, we no longer need
   // the instruction to be in three-address form.
@@ -299,13 +291,12 @@
   return true;
 }
 
-/// NoUseAfterLastDef - Return true if there are no intervening uses between the
+/// noUseAfterLastDef - Return true if there are no intervening uses between the
 /// last instruction in the MBB that defines the specified register and the
 /// two-address instruction which is being processed. It also returns the last
 /// def location by reference
-bool TwoAddressInstructionPass::NoUseAfterLastDef(unsigned Reg,
-                                           MachineBasicBlock *MBB, unsigned Dist,
-                                           unsigned &LastDef) {
+bool TwoAddressInstructionPass::noUseAfterLastDef(unsigned Reg, unsigned Dist,
+                                                  unsigned &LastDef) {
   LastDef = 0;
   unsigned LastUse = Dist;
   for (MachineRegisterInfo::reg_iterator I = MRI->reg_begin(Reg),
@@ -465,10 +456,9 @@
 /// isProfitableToCommute - Return true if it's potentially profitable to commute
 /// the two-address instruction that's being processed.
 bool
-TwoAddressInstructionPass::isProfitableToCommute(unsigned regA, unsigned regB,
-                                       unsigned regC,
-                                       MachineInstr *MI, MachineBasicBlock *MBB,
-                                       unsigned Dist) {
+TwoAddressInstructionPass::
+isProfitableToCommute(unsigned regA, unsigned regB, unsigned regC,
+                      MachineInstr *MI, unsigned Dist) {
   if (OptLevel == CodeGenOpt::None)
     return false;
 
@@ -516,13 +506,13 @@
   // If there is a use of regC between its last def (could be livein) and this
   // instruction, then bail.
   unsigned LastDefC = 0;
-  if (!NoUseAfterLastDef(regC, MBB, Dist, LastDefC))
+  if (!noUseAfterLastDef(regC, Dist, LastDefC))
     return false;
 
   // If there is a use of regB between its last def (could be livein) and this
   // instruction, then go ahead and make this transformation.
   unsigned LastDefB = 0;
-  if (!NoUseAfterLastDef(regB, MBB, Dist, LastDefB))
+  if (!noUseAfterLastDef(regB, Dist, LastDefB))
     return true;
 
   // Since there are no intervening uses for both registers, then commute
@@ -530,13 +520,12 @@
   return LastDefB && LastDefC && LastDefC > LastDefB;
 }
 
-/// CommuteInstruction - Commute a two-address instruction and update the basic
+/// commuteInstruction - Commute a two-address instruction and update the basic
 /// block, distance map, and live variables if needed. Return true if it is
 /// successful.
-bool
-TwoAddressInstructionPass::CommuteInstruction(MachineBasicBlock::iterator &mi,
-                               MachineFunction::iterator &mbbi,
-                               unsigned RegB, unsigned RegC, unsigned Dist) {
+bool TwoAddressInstructionPass::
+commuteInstruction(MachineBasicBlock::iterator &mi,
+                   unsigned RegB, unsigned RegC, unsigned Dist) {
   MachineInstr *MI = mi;
   DEBUG(dbgs() << "2addr: COMMUTING  : " << *MI);
   MachineInstr *NewMI = TII->commuteInstruction(MI);
@@ -555,8 +544,8 @@
     if (Indexes)
       Indexes->replaceMachineInstrInMaps(MI, NewMI);
 
-    mbbi->insert(mi, NewMI);           // Insert the new inst
-    mbbi->erase(mi);                   // Nuke the old inst.
+    MBB->insert(mi, NewMI);           // Insert the new inst
+    MBB->erase(mi);                   // Nuke the old inst.
     mi = NewMI;
     DistanceMap.insert(std::make_pair(NewMI, Dist));
   }
@@ -588,51 +577,51 @@
   return (ToRegA && !regsAreCompatible(FromRegB, ToRegA, TRI));
 }
 
-/// ConvertInstTo3Addr - Convert the specified two-address instruction into a
+/// convertInstTo3Addr - Convert the specified two-address instruction into a
 /// three address one. Return true if this transformation was successful.
 bool
-TwoAddressInstructionPass::ConvertInstTo3Addr(MachineBasicBlock::iterator &mi,
+TwoAddressInstructionPass::convertInstTo3Addr(MachineBasicBlock::iterator &mi,
                                               MachineBasicBlock::iterator &nmi,
-                                              MachineFunction::iterator &mbbi,
                                               unsigned RegA, unsigned RegB,
                                               unsigned Dist) {
-  MachineInstr *NewMI = TII->convertToThreeAddress(mbbi, mi, LV);
-  if (NewMI) {
-    DEBUG(dbgs() << "2addr: CONVERTING 2-ADDR: " << *mi);
-    DEBUG(dbgs() << "2addr:         TO 3-ADDR: " << *NewMI);
-    bool Sunk = false;
+  // FIXME: Why does convertToThreeAddress() need an iterator reference?
+  MachineFunction::iterator MFI = MBB;
+  MachineInstr *NewMI = TII->convertToThreeAddress(MFI, mi, LV);
+  assert(MBB == MFI && "convertToThreeAddress changed iterator reference");
+  if (!NewMI)
+    return false;
 
-    if (Indexes)
-      Indexes->replaceMachineInstrInMaps(mi, NewMI);
+  DEBUG(dbgs() << "2addr: CONVERTING 2-ADDR: " << *mi);
+  DEBUG(dbgs() << "2addr:         TO 3-ADDR: " << *NewMI);
+  bool Sunk = false;
 
-    if (NewMI->findRegisterUseOperand(RegB, false, TRI))
-      // FIXME: Temporary workaround. If the new instruction doesn't
-      // uses RegB, convertToThreeAddress must have created more
-      // then one instruction.
-      Sunk = Sink3AddrInstruction(mbbi, NewMI, RegB, mi);
+  if (Indexes)
+    Indexes->replaceMachineInstrInMaps(mi, NewMI);
 
-    mbbi->erase(mi); // Nuke the old inst.
+  if (NewMI->findRegisterUseOperand(RegB, false, TRI))
+    // FIXME: Temporary workaround. If the new instruction doesn't
+    // uses RegB, convertToThreeAddress must have created more
+    // then one instruction.
+    Sunk = sink3AddrInstruction(NewMI, RegB, mi);
 
-    if (!Sunk) {
-      DistanceMap.insert(std::make_pair(NewMI, Dist));
-      mi = NewMI;
-      nmi = llvm::next(mi);
-    }
+  MBB->erase(mi); // Nuke the old inst.
 
-    // Update source and destination register maps.
-    SrcRegMap.erase(RegA);
-    DstRegMap.erase(RegB);
-    return true;
+  if (!Sunk) {
+    DistanceMap.insert(std::make_pair(NewMI, Dist));
+    mi = NewMI;
+    nmi = llvm::next(mi);
   }
 
-  return false;
+  // Update source and destination register maps.
+  SrcRegMap.erase(RegA);
+  DstRegMap.erase(RegB);
+  return true;
 }
 
-/// ScanUses - Scan forward recursively for only uses, update maps if the use
+/// scanUses - Scan forward recursively for only uses, update maps if the use
 /// is a copy or a two-address instruction.
 void
-TwoAddressInstructionPass::ScanUses(unsigned DstReg, MachineBasicBlock *MBB,
-                                    SmallPtrSet<MachineInstr*, 8> &Processed) {
+TwoAddressInstructionPass::scanUses(unsigned DstReg) {
   SmallVector<unsigned, 4> VirtRegPairs;
   bool IsDstPhys;
   bool IsCopy = false;
@@ -676,7 +665,7 @@
   }
 }
 
-/// ProcessCopy - If the specified instruction is not yet processed, process it
+/// processCopy - If the specified instruction is not yet processed, process it
 /// if it's a copy. For a copy instruction, we find the physical registers the
 /// source and destination registers might be mapped to. These are kept in
 /// point-to maps used to determine future optimizations. e.g.
@@ -688,9 +677,7 @@
 /// coalesced to r0 (from the input side). v1025 is mapped to r1. v1026 is
 /// potentially joined with r1 on the output side. It's worthwhile to commute
 /// 'add' to eliminate a copy.
-void TwoAddressInstructionPass::ProcessCopy(MachineInstr *MI,
-                                     MachineBasicBlock *MBB,
-                                     SmallPtrSet<MachineInstr*, 8> &Processed) {
+void TwoAddressInstructionPass::processCopy(MachineInstr *MI) {
   if (Processed.count(MI))
     return;
 
@@ -707,21 +694,20 @@
       assert(SrcRegMap[DstReg] == SrcReg &&
              "Can't map to two src physical registers!");
 
-    ScanUses(DstReg, MBB, Processed);
+    scanUses(DstReg);
   }
 
   Processed.insert(MI);
   return;
 }
 
-/// RescheduleMIBelowKill - If there is one more local instruction that reads
+/// rescheduleMIBelowKill - If there is one more local instruction that reads
 /// 'Reg' and it kills 'Reg, consider moving the instruction below the kill
 /// instruction in order to eliminate the need for the copy.
-bool
-TwoAddressInstructionPass::RescheduleMIBelowKill(MachineBasicBlock *MBB,
-                                     MachineBasicBlock::iterator &mi,
-                                     MachineBasicBlock::iterator &nmi,
-                                     unsigned Reg) {
+bool TwoAddressInstructionPass::
+rescheduleMIBelowKill(MachineBasicBlock::iterator &mi,
+                      MachineBasicBlock::iterator &nmi,
+                      unsigned Reg) {
   // Bail immediately if we don't have LV available. We use it to find kills
   // efficiently.
   if (!LV)
@@ -853,8 +839,7 @@
 /// isDefTooClose - Return true if the re-scheduling will put the given
 /// instruction too close to the defs of its register dependencies.
 bool TwoAddressInstructionPass::isDefTooClose(unsigned Reg, unsigned Dist,
-                                              MachineInstr *MI,
-                                              MachineBasicBlock *MBB) {
+                                              MachineInstr *MI) {
   for (MachineRegisterInfo::def_iterator DI = MRI->def_begin(Reg),
          DE = MRI->def_end(); DI != DE; ++DI) {
     MachineInstr *DefMI = &*DI;
@@ -873,15 +858,14 @@
   return false;
 }
 
-/// RescheduleKillAboveMI - If there is one more local instruction that reads
+/// rescheduleKillAboveMI - If there is one more local instruction that reads
 /// 'Reg' and it kills 'Reg, consider moving the kill instruction above the
 /// current two-address instruction in order to eliminate the need for the
 /// copy.
-bool
-TwoAddressInstructionPass::RescheduleKillAboveMI(MachineBasicBlock *MBB,
-                                     MachineBasicBlock::iterator &mi,
-                                     MachineBasicBlock::iterator &nmi,
-                                     unsigned Reg) {
+bool TwoAddressInstructionPass::
+rescheduleKillAboveMI(MachineBasicBlock::iterator &mi,
+                      MachineBasicBlock::iterator &nmi,
+                      unsigned Reg) {
   // Bail immediately if we don't have LV available. We use it to find kills
   // efficiently.
   if (!LV)
@@ -918,7 +902,7 @@
     if (MO.isUse()) {
       if (!MOReg)
         continue;
-      if (isDefTooClose(MOReg, DI->second, MI, MBB))
+      if (isDefTooClose(MOReg, DI->second, MI))
         return false;
       if (MOReg == Reg && !MO.isKill())
         return false;
@@ -1006,18 +990,16 @@
   return true;
 }
 
-/// TryInstructionTransform - For the case where an instruction has a single
+/// tryInstructionTransform - For the case where an instruction has a single
 /// pair of tied register operands, attempt some transformations that may
 /// either eliminate the tied operands or improve the opportunities for
 /// coalescing away the register copy.  Returns true if no copy needs to be
 /// inserted to untie mi's operands (either because they were untied, or
 /// because mi was rescheduled, and will be visited again later).
 bool TwoAddressInstructionPass::
-TryInstructionTransform(MachineBasicBlock::iterator &mi,
+tryInstructionTransform(MachineBasicBlock::iterator &mi,
                         MachineBasicBlock::iterator &nmi,
-                        MachineFunction::iterator &mbbi,
-                        unsigned SrcIdx, unsigned DstIdx, unsigned Dist,
-                        SmallPtrSet<MachineInstr*, 8> &Processed) {
+                        unsigned SrcIdx, unsigned DstIdx, unsigned Dist) {
   if (OptLevel == CodeGenOpt::None)
     return false;
 
@@ -1030,7 +1012,7 @@
   bool regBKilled = isKilled(MI, regB, MRI, TII);
 
   if (TargetRegisterInfo::isVirtualRegister(regA))
-    ScanUses(regA, &*mbbi, Processed);
+    scanUses(regA);
 
   // Check if it is profitable to commute the operands.
   unsigned SrcOp1, SrcOp2;
@@ -1051,7 +1033,7 @@
         // If C dies but B does not, swap the B and C operands.
         // This makes the live ranges of A and C joinable.
         TryCommute = true;
-      else if (isProfitableToCommute(regA, regB, regC, &MI, mbbi, Dist)) {
+      else if (isProfitableToCommute(regA, regB, regC, &MI, Dist)) {
         TryCommute = true;
         AggressiveCommute = true;
       }
@@ -1059,7 +1041,7 @@
   }
 
   // If it's profitable to commute, try to do so.
-  if (TryCommute && CommuteInstruction(mi, mbbi, regB, regC, Dist)) {
+  if (TryCommute && commuteInstruction(mi, regB, regC, Dist)) {
     ++NumCommuted;
     if (AggressiveCommute)
       ++NumAggrCommuted;
@@ -1068,7 +1050,7 @@
 
   // If there is one more use of regB later in the same MBB, consider
   // re-schedule this MI below it.
-  if (RescheduleMIBelowKill(mbbi, mi, nmi, regB)) {
+  if (rescheduleMIBelowKill(mi, nmi, regB)) {
     ++NumReSchedDowns;
     return true;
   }
@@ -1078,7 +1060,7 @@
     // three-address instruction.  Check if it is profitable.
     if (!regBKilled || isProfitableToConv3Addr(regA, regB)) {
       // Try to convert it.
-      if (ConvertInstTo3Addr(mi, nmi, mbbi, regA, regB, Dist)) {
+      if (convertInstTo3Addr(mi, nmi, regA, regB, Dist)) {
         ++NumConvertedTo3Addr;
         return true; // Done with this instruction.
       }
@@ -1087,7 +1069,7 @@
 
   // If there is one more use of regB later in the same MBB, consider
   // re-schedule it before this MI if it's legal.
-  if (RescheduleKillAboveMI(mbbi, mi, nmi, regB)) {
+  if (rescheduleKillAboveMI(mi, nmi, regB)) {
     ++NumReSchedUps;
     return true;
   }
@@ -1131,8 +1113,8 @@
 
         // Tentatively insert the instructions into the block so that they
         // look "normal" to the transformation logic.
-        mbbi->insert(mi, NewMIs[0]);
-        mbbi->insert(mi, NewMIs[1]);
+        MBB->insert(mi, NewMIs[0]);
+        MBB->insert(mi, NewMIs[1]);
 
         DEBUG(dbgs() << "2addr:    NEW LOAD: " << *NewMIs[0]
                      << "2addr:    NEW INST: " << *NewMIs[1]);
@@ -1142,8 +1124,7 @@
         unsigned NewSrcIdx = NewMIs[1]->findRegisterUseOperandIdx(regB);
         MachineBasicBlock::iterator NewMI = NewMIs[1];
         bool TransformSuccess =
-          TryInstructionTransform(NewMI, mi, mbbi,
-                                  NewSrcIdx, NewDstIdx, Dist, Processed);
+          tryInstructionTransform(NewMI, mi, NewSrcIdx, NewDstIdx, Dist);
         if (TransformSuccess ||
             NewMIs[1]->getOperand(NewSrcIdx).isKill()) {
           // Success, or at least we made an improvement. Keep the unfolded
@@ -1378,16 +1359,15 @@
   MRI->leaveSSA();
 
   TiedOperandMap TiedOperands;
-
-  SmallPtrSet<MachineInstr*, 8> Processed;
-  for (MachineFunction::iterator mbbi = MF->begin(), mbbe = MF->end();
-       mbbi != mbbe; ++mbbi) {
+  for (MachineFunction::iterator MBBI = MF->begin(), MBBE = MF->end();
+       MBBI != MBBE; ++MBBI) {
+    MBB = MBBI;
     unsigned Dist = 0;
     DistanceMap.clear();
     SrcRegMap.clear();
     DstRegMap.clear();
     Processed.clear();
-    for (MachineBasicBlock::iterator mi = mbbi->begin(), me = mbbi->end();
+    for (MachineBasicBlock::iterator mi = MBB->begin(), me = MBB->end();
          mi != me; ) {
       MachineBasicBlock::iterator nmi = llvm::next(mi);
       if (mi->isDebugValue()) {
@@ -1401,7 +1381,7 @@
 
       DistanceMap.insert(std::make_pair(mi, ++Dist));
 
-      ProcessCopy(&*mi, &*mbbi, Processed);
+      processCopy(&*mi);
 
       // First scan through all the tied register uses in this instruction
       // and record a list of pairs of tied operands for each register.
@@ -1426,8 +1406,7 @@
           unsigned SrcReg = mi->getOperand(SrcIdx).getReg();
           unsigned DstReg = mi->getOperand(DstIdx).getReg();
           if (SrcReg != DstReg &&
-              TryInstructionTransform(mi, nmi, mbbi, SrcIdx, DstIdx, Dist,
-                                      Processed)) {
+              tryInstructionTransform(mi, nmi, SrcIdx, DstIdx, Dist)) {
             // The tied operands have been eliminated or shifted further down the
             // block to ease elimination. Continue processing with 'nmi'.
             TiedOperands.clear();
@@ -1467,7 +1446,7 @@
 
   // Eliminate REG_SEQUENCE instructions. Their whole purpose was to preseve
   // SSA form. It's now safe to de-SSA.
-  MadeChange |= EliminateRegSequences();
+  MadeChange |= eliminateRegSequences();
 
   return MadeChange;
 }
@@ -1514,127 +1493,6 @@
   return First;
 }
 
-/// CoalesceExtSubRegs - If a number of sources of the REG_SEQUENCE are
-/// EXTRACT_SUBREG from the same register and to the same virtual register
-/// with different sub-register indices, attempt to combine the
-/// EXTRACT_SUBREGs and pre-coalesce them. e.g.
-/// %reg1026<def> = VLDMQ %reg1025<kill>, 260, pred:14, pred:%reg0
-/// %reg1029:6<def> = EXTRACT_SUBREG %reg1026, 6
-/// %reg1029:5<def> = EXTRACT_SUBREG %reg1026<kill>, 5
-/// Since D subregs 5, 6 can combine to a Q register, we can coalesce
-/// reg1026 to reg1029.
-void
-TwoAddressInstructionPass::CoalesceExtSubRegs(SmallVector<unsigned,4> &Srcs,
-                                              unsigned DstReg) {
-  SmallSet<unsigned, 4> Seen;
-  for (unsigned i = 0, e = Srcs.size(); i != e; ++i) {
-    unsigned SrcReg = Srcs[i];
-    if (!Seen.insert(SrcReg))
-      continue;
-
-    // Check that the instructions are all in the same basic block.
-    MachineInstr *SrcDefMI = MRI->getUniqueVRegDef(SrcReg);
-    MachineInstr *DstDefMI = MRI->getUniqueVRegDef(DstReg);
-    if (!SrcDefMI || !DstDefMI ||
-        SrcDefMI->getParent() != DstDefMI->getParent())
-      continue;
-
-    // If there are no other uses than copies which feed into
-    // the reg_sequence, then we might be able to coalesce them.
-    bool CanCoalesce = true;
-    SmallVector<unsigned, 4> SrcSubIndices, DstSubIndices;
-    for (MachineRegisterInfo::use_nodbg_iterator
-           UI = MRI->use_nodbg_begin(SrcReg),
-           UE = MRI->use_nodbg_end(); UI != UE; ++UI) {
-      MachineInstr *UseMI = &*UI;
-      if (!UseMI->isCopy() || UseMI->getOperand(0).getReg() != DstReg) {
-        CanCoalesce = false;
-        break;
-      }
-      SrcSubIndices.push_back(UseMI->getOperand(1).getSubReg());
-      DstSubIndices.push_back(UseMI->getOperand(0).getSubReg());
-    }
-
-    if (!CanCoalesce || SrcSubIndices.size() < 2)
-      continue;
-
-    // Check that the source subregisters can be combined.
-    std::sort(SrcSubIndices.begin(), SrcSubIndices.end());
-    unsigned NewSrcSubIdx = 0;
-    if (!TRI->canCombineSubRegIndices(MRI->getRegClass(SrcReg), SrcSubIndices,
-                                      NewSrcSubIdx))
-      continue;
-
-    // Check that the destination subregisters can also be combined.
-    std::sort(DstSubIndices.begin(), DstSubIndices.end());
-    unsigned NewDstSubIdx = 0;
-    if (!TRI->canCombineSubRegIndices(MRI->getRegClass(DstReg), DstSubIndices,
-                                      NewDstSubIdx))
-      continue;
-
-    // If neither source nor destination can be combined to the full register,
-    // just give up.  This could be improved if it ever matters.
-    if (NewSrcSubIdx != 0 && NewDstSubIdx != 0)
-      continue;
-
-    // Now that we know that all the uses are extract_subregs and that those
-    // subregs can somehow be combined, scan all the extract_subregs again to
-    // make sure the subregs are in the right order and can be composed.
-    MachineInstr *SomeMI = 0;
-    CanCoalesce = true;
-    for (MachineRegisterInfo::use_nodbg_iterator
-           UI = MRI->use_nodbg_begin(SrcReg),
-           UE = MRI->use_nodbg_end(); UI != UE; ++UI) {
-      MachineInstr *UseMI = &*UI;
-      assert(UseMI->isCopy());
-      unsigned DstSubIdx = UseMI->getOperand(0).getSubReg();
-      unsigned SrcSubIdx = UseMI->getOperand(1).getSubReg();
-      assert(DstSubIdx != 0 && "missing subreg from RegSequence elimination");
-      if ((NewDstSubIdx == 0 &&
-           TRI->composeSubRegIndices(NewSrcSubIdx, DstSubIdx) != SrcSubIdx) ||
-          (NewSrcSubIdx == 0 &&
-           TRI->composeSubRegIndices(NewDstSubIdx, SrcSubIdx) != DstSubIdx)) {
-        CanCoalesce = false;
-        break;
-      }
-      // Keep track of one of the uses.  Preferably the first one which has a
-      // <def,undef> flag.
-      if (!SomeMI || UseMI->getOperand(0).isUndef())
-        SomeMI = UseMI;
-    }
-    if (!CanCoalesce)
-      continue;
-
-    // Insert a copy to replace the original.
-    MachineInstr *CopyMI = BuildMI(*SomeMI->getParent(), SomeMI,
-                                   SomeMI->getDebugLoc(),
-                                   TII->get(TargetOpcode::COPY))
-      .addReg(DstReg, RegState::Define |
-                      getUndefRegState(SomeMI->getOperand(0).isUndef()),
-              NewDstSubIdx)
-      .addReg(SrcReg, 0, NewSrcSubIdx);
-
-    // Remove all the old extract instructions.
-    for (MachineRegisterInfo::use_nodbg_iterator
-           UI = MRI->use_nodbg_begin(SrcReg),
-           UE = MRI->use_nodbg_end(); UI != UE; ) {
-      MachineInstr *UseMI = &*UI;
-      ++UI;
-      if (UseMI == CopyMI)
-        continue;
-      assert(UseMI->isCopy());
-      // Move any kills to the new copy or extract instruction.
-      if (UseMI->getOperand(1).isKill()) {
-        CopyMI->getOperand(1).setIsKill();
-        if (LV)
-          // Update live variables
-          LV->replaceKillInstruction(SrcReg, UseMI, &*CopyMI);
-      }
-      UseMI->eraseFromParent();
-    }
-  }
-}
-
 static bool HasOtherRegSequenceUses(unsigned Reg, MachineInstr *RegSeq,
                                     MachineRegisterInfo *MRI) {
   for (MachineRegisterInfo::use_iterator UI = MRI->use_begin(Reg),
@@ -1646,7 +1504,7 @@
   return false;
 }
 
-/// EliminateRegSequences - Eliminate REG_SEQUENCE instructions as part
+/// eliminateRegSequences - Eliminate REG_SEQUENCE instructions as part
 /// of the de-ssa process. This replaces sources of REG_SEQUENCE as
 /// sub-register references of the register defined by REG_SEQUENCE. e.g.
 ///
@@ -1654,7 +1512,7 @@
 /// %reg1031<def> = REG_SEQUENCE %reg1029<kill>, 5, %reg1030<kill>, 6
 /// =>
 /// %reg1031:5<def>, %reg1031:6<def> = VLD1q16 %reg1024<kill>, ...
-bool TwoAddressInstructionPass::EliminateRegSequences() {
+bool TwoAddressInstructionPass::eliminateRegSequences() {
   if (RegSequences.empty())
     return false;
 
@@ -1770,12 +1628,6 @@
       DEBUG(dbgs() << "Eliminated: " << *MI);
       MI->eraseFromParent();
     }
-
-    // Try coalescing some EXTRACT_SUBREG instructions. This can create
-    // INSERT_SUBREG instructions that must have <undef> flags added by
-    // LiveIntervalAnalysis, so only run it when LiveVariables is available.
-    if (LV)
-      CoalesceExtSubRegs(RealSrcs, DstReg);
   }
 
   RegSequences.clear();

Modified: llvm/branches/R600/lib/DebugInfo/DWARFDebugInfoEntry.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/DebugInfo/DWARFDebugInfoEntry.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/DebugInfo/DWARFDebugInfoEntry.cpp (original)
+++ llvm/branches/R600/lib/DebugInfo/DWARFDebugInfoEntry.cpp Tue Nov 13 09:21:47 2012
@@ -341,7 +341,7 @@
                 else
                   debug_info_data.getU64(offset_ptr);
                 break;
-                
+
               default:
                 *offset_ptr = offset;
                 return false;
@@ -411,9 +411,10 @@
 
 const char*
 DWARFDebugInfoEntryMinimal::getAttributeValueAsString(
-    const DWARFCompileUnit* cu,
-    const uint16_t attr,
-    const char* fail_value) const {
+                                                     const DWARFCompileUnit* cu,
+                                                     const uint16_t attr,
+                                                     const char* fail_value)
+                                                     const {
   DWARFFormValue form_value;
   if (getAttributeValue(cu, attr, form_value)) {
     DataExtractor stringExtractor(cu->getContext().getStringSection(),
@@ -425,9 +426,9 @@
 
 uint64_t
 DWARFDebugInfoEntryMinimal::getAttributeValueAsUnsigned(
-    const DWARFCompileUnit* cu,
-    const uint16_t attr,
-    uint64_t fail_value) const {
+                                                    const DWARFCompileUnit* cu,
+                                                    const uint16_t attr,
+                                                    uint64_t fail_value) const {
   DWARFFormValue form_value;
   if (getAttributeValue(cu, attr, form_value))
       return form_value.getUnsigned();
@@ -436,9 +437,9 @@
 
 int64_t
 DWARFDebugInfoEntryMinimal::getAttributeValueAsSigned(
-    const DWARFCompileUnit* cu,
-    const uint16_t attr,
-    int64_t fail_value) const {
+                                                     const DWARFCompileUnit* cu,
+                                                     const uint16_t attr,
+                                                     int64_t fail_value) const {
   DWARFFormValue form_value;
   if (getAttributeValue(cu, attr, form_value))
       return form_value.getSigned();
@@ -447,9 +448,10 @@
 
 uint64_t
 DWARFDebugInfoEntryMinimal::getAttributeValueAsReference(
-                                                  const DWARFCompileUnit* cu,
-                                                  const uint16_t attr,
-                                                  uint64_t fail_value) const {
+                                                     const DWARFCompileUnit* cu,
+                                                     const uint16_t attr,
+                                                     uint64_t fail_value)
+                                                     const {
   DWARFFormValue form_value;
   if (getAttributeValue(cu, attr, form_value))
       return form_value.getReference(cu);
@@ -457,7 +459,8 @@
 }
 
 bool DWARFDebugInfoEntryMinimal::getLowAndHighPC(const DWARFCompileUnit *CU,
-    uint64_t &LowPC, uint64_t &HighPC) const {
+                                                 uint64_t &LowPC,
+                                                 uint64_t &HighPC) const {
   HighPC = -1ULL;
   LowPC = getAttributeValueAsUnsigned(CU, DW_AT_low_pc, -1ULL);
   if (LowPC != -1ULL)
@@ -488,7 +491,9 @@
 
 bool
 DWARFDebugInfoEntryMinimal::addressRangeContainsAddress(
-    const DWARFCompileUnit *CU, const uint64_t Address) const {
+                                                     const DWARFCompileUnit *CU,
+                                                     const uint64_t Address)
+                                                     const {
   if (isNULL())
     return false;
   uint64_t LowPC, HighPC;
@@ -505,8 +510,8 @@
 }
 
 const char*
-DWARFDebugInfoEntryMinimal::getSubroutineName(
-    const DWARFCompileUnit *CU) const {
+DWARFDebugInfoEntryMinimal::getSubroutineName(const DWARFCompileUnit *CU)
+                                                                         const {
   if (!isSubroutineDIE())
     return 0;
   // Try to get mangled name if possible.
@@ -540,9 +545,10 @@
   return 0;
 }
 
-void DWARFDebugInfoEntryMinimal::getCallerFrame(
-    const DWARFCompileUnit *CU, uint32_t &CallFile, uint32_t &CallLine,
-    uint32_t &CallColumn) const {
+void DWARFDebugInfoEntryMinimal::getCallerFrame(const DWARFCompileUnit *CU,
+                                                uint32_t &CallFile,
+                                                uint32_t &CallLine,
+                                                uint32_t &CallColumn) const {
   CallFile = getAttributeValueAsUnsigned(CU, DW_AT_call_file, 0);
   CallLine = getAttributeValueAsUnsigned(CU, DW_AT_call_line, 0);
   CallColumn = getAttributeValueAsUnsigned(CU, DW_AT_call_column, 0);
@@ -550,7 +556,9 @@
 
 DWARFDebugInfoEntryMinimal::InlinedChain
 DWARFDebugInfoEntryMinimal::getInlinedChainForAddress(
-    const DWARFCompileUnit *CU, const uint64_t Address) const {
+                                                     const DWARFCompileUnit *CU,
+                                                     const uint64_t Address)
+                                                     const {
   DWARFDebugInfoEntryMinimal::InlinedChain InlinedChain;
   if (isNULL())
     return InlinedChain;

Modified: llvm/branches/R600/lib/ExecutionEngine/ExecutionEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/ExecutionEngine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/ExecutionEngine.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/ExecutionEngine.cpp Tue Nov 13 09:21:47 2012
@@ -17,7 +17,6 @@
 
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
-#include "llvm/Instructions.h"
 #include "llvm/Module.h"
 #include "llvm/ExecutionEngine/GenericValue.h"
 #include "llvm/ADT/SmallString.h"
@@ -268,7 +267,7 @@
 void *ArgvArray::reset(LLVMContext &C, ExecutionEngine *EE,
                        const std::vector<std::string> &InputArgv) {
   clear();  // Free the old contents.
-  unsigned PtrSize = EE->getDataLayout()->getPointerSize(0);
+  unsigned PtrSize = EE->getDataLayout()->getPointerSize();
   Array = new char[(InputArgv.size()+1)*PtrSize];
 
   DEBUG(dbgs() << "JIT: ARGV = " << (void*)Array << "\n");
@@ -343,7 +342,7 @@
 #ifndef NDEBUG
 /// isTargetNullPtr - Return whether the target pointer stored at Loc is null.
 static bool isTargetNullPtr(ExecutionEngine *EE, void *Loc) {
-  unsigned PtrSize = EE->getDataLayout()->getPointerSize(0);
+  unsigned PtrSize = EE->getDataLayout()->getPointerSize();
   for (unsigned i = 0; i < PtrSize; ++i)
     if (*(i + (uint8_t*)Loc))
       return false;
@@ -645,18 +644,17 @@
     }
     case Instruction::PtrToInt: {
       GenericValue GV = getConstantValue(Op0);
-      unsigned AS = cast<PointerType>(CE->getOperand(1)->getType())
-        ->getAddressSpace();
-      uint32_t PtrWidth = TD->getPointerSizeInBits(AS);
+      uint32_t PtrWidth = TD->getTypeSizeInBits(Op0->getType());
+      assert(PtrWidth <= 64 && "Bad pointer width");
       GV.IntVal = APInt(PtrWidth, uintptr_t(GV.PointerVal));
+      uint32_t IntWidth = TD->getTypeSizeInBits(CE->getType());
+      GV.IntVal = GV.IntVal.zextOrTrunc(IntWidth);
       return GV;
     }
     case Instruction::IntToPtr: {
       GenericValue GV = getConstantValue(Op0);
-      unsigned AS = cast<PointerType>(CE->getType())->getAddressSpace();
-      uint32_t PtrWidth = TD->getPointerSizeInBits(AS);
-      if (PtrWidth != GV.IntVal.getBitWidth())
-        GV.IntVal = GV.IntVal.zextOrTrunc(PtrWidth);
+      uint32_t PtrWidth = TD->getTypeSizeInBits(CE->getType());
+      GV.IntVal = GV.IntVal.zextOrTrunc(PtrWidth);
       assert(GV.IntVal.getBitWidth() <= 64 && "Bad pointer width");
       GV.PointerVal = PointerTy(uintptr_t(GV.IntVal.getZExtValue()));
       return GV;

Modified: llvm/branches/R600/lib/ExecutionEngine/IntelJITEvents/IntelJITEventListener.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/IntelJITEvents/IntelJITEventListener.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/IntelJITEvents/IntelJITEventListener.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/IntelJITEvents/IntelJITEventListener.cpp Tue Nov 13 09:21:47 2012
@@ -54,6 +54,10 @@
                                      const EmittedFunctionDetails &Details);
 
   virtual void NotifyFreeingMachineCode(void *OldPtr);
+
+  virtual void NotifyObjectEmitted(const ObjectImage &Obj);
+
+  virtual void NotifyFreeingObject(const ObjectImage &Obj);
 };
 
 static LineNumberInfo LineStartToIntelJITFormat(
@@ -164,6 +168,12 @@
   }
 }
 
+void IntelJITEventListener::NotifyObjectEmitted(const ObjectImage &Obj) {
+}
+
+void IntelJITEventListener::NotifyFreeingObject(const ObjectImage &Obj) {
+}
+
 }  // anonymous namespace.
 
 namespace llvm {

Modified: llvm/branches/R600/lib/ExecutionEngine/Interpreter/Execution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/Interpreter/Execution.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/Interpreter/Execution.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/Interpreter/Execution.cpp Tue Nov 13 09:21:47 2012
@@ -1054,8 +1054,7 @@
   GenericValue Dest, Src = getOperandValue(SrcVal, SF);
   assert(DstTy->isPointerTy() && "Invalid PtrToInt instruction");
 
-  unsigned AS = cast<PointerType>(DstTy)->getAddressSpace();
-  uint32_t PtrSize = TD.getPointerSizeInBits(AS);
+  uint32_t PtrSize = TD.getPointerSizeInBits();
   if (PtrSize != Src.IntVal.getBitWidth())
     Src.IntVal = Src.IntVal.zextOrTrunc(PtrSize);
 

Modified: llvm/branches/R600/lib/ExecutionEngine/Interpreter/ExternalFunctions.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/Interpreter/ExternalFunctions.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/Interpreter/ExternalFunctions.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/Interpreter/ExternalFunctions.cpp Tue Nov 13 09:21:47 2012
@@ -376,7 +376,7 @@
       case 'x': case 'X':
         if (HowLong >= 1) {
           if (HowLong == 1 &&
-              TheInterpreter->getDataLayout()->getPointerSizeInBits(0) == 64 &&
+              TheInterpreter->getDataLayout()->getPointerSizeInBits() == 64 &&
               sizeof(long) < sizeof(int64_t)) {
             // Make sure we use %lld with a 64 bit argument because we might be
             // compiling LLI on a 32 bit compiler.

Modified: llvm/branches/R600/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp Tue Nov 13 09:21:47 2012
@@ -14,9 +14,7 @@
 
 #include "JIT.h"
 #include "JITDwarfEmitter.h"
-#include "llvm/DerivedTypes.h"
 #include "llvm/Function.h"
-#include "llvm/GlobalVariable.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/CodeGen/JITCodeEmitter.h"
 #include "llvm/CodeGen/MachineFunction.h"
@@ -68,7 +66,7 @@
 void
 JITDwarfEmitter::EmitFrameMoves(intptr_t BaseLabelPtr,
                                 const std::vector<MachineMove> &Moves) const {
-  unsigned PointerSize = TD->getPointerSize(0);
+  unsigned PointerSize = TD->getPointerSize();
   int stackGrowth = stackGrowthDirection == TargetFrameLowering::StackGrowsUp ?
           PointerSize : -PointerSize;
   MCSymbol *BaseLabel = 0;
@@ -380,7 +378,7 @@
   for (unsigned i = 0, e = CallSites.size(); i < e; ++i)
     SizeSites += MCAsmInfo::getULEB128Size(CallSites[i].Action);
 
-  unsigned SizeTypes = TypeInfos.size() * TD->getPointerSize(0);
+  unsigned SizeTypes = TypeInfos.size() * TD->getPointerSize();
 
   unsigned TypeOffset = sizeof(int8_t) + // Call site format
                         // Call-site table length
@@ -456,12 +454,12 @@
     const GlobalVariable *GV = TypeInfos[M - 1];
 
     if (GV) {
-      if (TD->getPointerSize(GV->getType()->getAddressSpace()) == sizeof(int32_t))
+      if (TD->getPointerSize() == sizeof(int32_t))
         JCE->emitInt32((intptr_t)Jit.getOrEmitGlobalVariable(GV));
       else
         JCE->emitInt64((intptr_t)Jit.getOrEmitGlobalVariable(GV));
     } else {
-      if (TD->getPointerSize(0) == sizeof(int32_t))
+      if (TD->getPointerSize() == sizeof(int32_t))
         JCE->emitInt32(0);
       else
         JCE->emitInt64(0);
@@ -483,7 +481,7 @@
 
 unsigned char*
 JITDwarfEmitter::EmitCommonEHFrame(const Function* Personality) const {
-  unsigned PointerSize = TD->getPointerSize(0);
+  unsigned PointerSize = TD->getPointerSize();
   int stackGrowth = stackGrowthDirection == TargetFrameLowering::StackGrowsUp ?
           PointerSize : -PointerSize;
 
@@ -543,7 +541,7 @@
                              unsigned char* StartFunction,
                              unsigned char* EndFunction,
                              unsigned char* ExceptionTable) const {
-  unsigned PointerSize = TD->getPointerSize(0);
+  unsigned PointerSize = TD->getPointerSize();
 
   // EH frame header.
   unsigned char* StartEHPtr = (unsigned char*)JCE->getCurrentPCValue();

Modified: llvm/branches/R600/lib/ExecutionEngine/MCJIT/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/MCJIT/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/MCJIT/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/ExecutionEngine/MCJIT/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -1,4 +1,3 @@
 add_llvm_library(LLVMMCJIT
   MCJIT.cpp
-  MCJITMemoryManager.cpp
   )

Modified: llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.cpp Tue Nov 13 09:21:47 2012
@@ -8,10 +8,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "MCJIT.h"
-#include "MCJITMemoryManager.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/Function.h"
 #include "llvm/ExecutionEngine/GenericValue.h"
+#include "llvm/ExecutionEngine/JITEventListener.h"
 #include "llvm/ExecutionEngine/JITMemoryManager.h"
 #include "llvm/ExecutionEngine/MCJIT.h"
 #include "llvm/ExecutionEngine/ObjectBuffer.h"
@@ -46,7 +46,7 @@
   // FIXME: Don't do this here.
   sys::DynamicLibrary::LoadLibraryPermanently(0, NULL);
 
-  return new MCJIT(M, TM, new MCJITMemoryManager(JMM), GVsWithCode);
+  return new MCJIT(M, TM, JMM, GVsWithCode);
 }
 
 MCJIT::MCJIT(Module *m, TargetMachine *tm, RTDyldMemoryManager *MM,
@@ -58,6 +58,8 @@
 }
 
 MCJIT::~MCJIT() {
+  if (LoadedObject)
+    NotifyFreeingObject(*LoadedObject.get());
   delete MemMgr;
   delete TM;
 }
@@ -108,10 +110,27 @@
   // FIXME: Make this optional, maybe even move it to a JIT event listener
   LoadedObject->registerWithDebugger();
 
+  NotifyObjectEmitted(*LoadedObject);
+
   // FIXME: Add support for per-module compilation state
   isCompiled = true;
 }
 
+// FIXME: Add a parameter to identify which object is being finalized when
+// MCJIT supports multiple modules.
+void MCJIT::finalizeObject() {
+  // If the module hasn't been compiled, just do that.
+  if (!isCompiled) {
+    // If the call to Dyld.resolveRelocations() is removed from emitObject()
+    // we'll need to do that here.
+    emitObject(M);
+    return;
+  }
+
+  // Resolve any relocations.
+  Dyld.resolveRelocations();
+}
+
 void *MCJIT::getPointerToBasicBlock(BasicBlock *BB) {
   report_fatal_error("not yet implemented");
 }
@@ -276,3 +295,33 @@
   }
   return 0;
 }
+
+void MCJIT::RegisterJITEventListener(JITEventListener *L) {
+  if (L == NULL)
+    return;
+  MutexGuard locked(lock);
+  EventListeners.push_back(L);
+}
+void MCJIT::UnregisterJITEventListener(JITEventListener *L) {
+  if (L == NULL)
+    return;
+  MutexGuard locked(lock);
+  SmallVector<JITEventListener*, 2>::reverse_iterator I=
+      std::find(EventListeners.rbegin(), EventListeners.rend(), L);
+  if (I != EventListeners.rend()) {
+    std::swap(*I, EventListeners.back());
+    EventListeners.pop_back();
+  }
+}
+void MCJIT::NotifyObjectEmitted(const ObjectImage& Obj) {
+  MutexGuard locked(lock);
+  for (unsigned I = 0, S = EventListeners.size(); I < S; ++I) {
+    EventListeners[I]->NotifyObjectEmitted(Obj);
+  }
+}
+void MCJIT::NotifyFreeingObject(const ObjectImage& Obj) {
+  MutexGuard locked(lock);
+  for (unsigned I = 0, S = EventListeners.size(); I < S; ++I) {
+    EventListeners[I]->NotifyFreeingObject(Obj);
+  }
+}

Modified: llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.h (original)
+++ llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJIT.h Tue Nov 13 09:21:47 2012
@@ -11,6 +11,7 @@
 #define LLVM_LIB_EXECUTIONENGINE_MCJIT_H
 
 #include "llvm/PassManager.h"
+#include "llvm/ADT/SmallVector.h"
 #include "llvm/ExecutionEngine/ExecutionEngine.h"
 #include "llvm/ExecutionEngine/RuntimeDyld.h"
 
@@ -30,6 +31,7 @@
   MCContext *Ctx;
   RTDyldMemoryManager *MemMgr;
   RuntimeDyld Dyld;
+  SmallVector<JITEventListener*, 2> EventListeners;
 
   // FIXME: Add support for multiple modules
   bool isCompiled;
@@ -42,6 +44,8 @@
   /// @name ExecutionEngine interface implementation
   /// @{
 
+  virtual void finalizeObject();
+
   virtual void *getPointerToBasicBlock(BasicBlock *BB);
 
   virtual void *getPointerToFunction(Function *F);
@@ -73,6 +77,9 @@
     Dyld.mapSectionAddress(LocalAddress, TargetAddress);
   }
 
+  virtual void RegisterJITEventListener(JITEventListener *L);
+  virtual void UnregisterJITEventListener(JITEventListener *L);
+
   /// @}
   /// @name (Private) Registration Interfaces
   /// @{
@@ -96,6 +103,9 @@
   /// is passed as a parameter here to prepare for multiple module support in 
   /// the future.
   void emitObject(Module *M);
+
+  void NotifyObjectEmitted(const ObjectImage& Obj);
+  void NotifyFreeingObject(const ObjectImage& Obj);
 };
 
 } // End llvm namespace

Removed: llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.cpp (removed)
@@ -1,14 +0,0 @@
-//==-- MCJITMemoryManager.cpp - Definition for the Memory Manager -*-C++ -*-==//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-
-#include "MCJITMemoryManager.h"
-
-using namespace llvm;
-
-void MCJITMemoryManager::anchor() { }

Removed: llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.h (original)
+++ llvm/branches/R600/lib/ExecutionEngine/MCJIT/MCJITMemoryManager.h (removed)
@@ -1,50 +0,0 @@
-//===-- MCJITMemoryManager.h - Definition for the Memory Manager ---C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_LIB_EXECUTIONENGINE_MCJITMEMORYMANAGER_H
-#define LLVM_LIB_EXECUTIONENGINE_MCJITMEMORYMANAGER_H
-
-#include "llvm/Module.h"
-#include "llvm/ExecutionEngine/JITMemoryManager.h"
-#include "llvm/ExecutionEngine/RuntimeDyld.h"
-#include <assert.h>
-
-namespace llvm {
-
-// The MCJIT memory manager is a layer between the standard JITMemoryManager
-// and the RuntimeDyld interface that maps objects, by name, onto their
-// matching LLVM IR counterparts in the module(s) being compiled.
-class MCJITMemoryManager : public RTDyldMemoryManager {
-  virtual void anchor();
-  OwningPtr<JITMemoryManager> JMM;
-
-public:
-  MCJITMemoryManager(JITMemoryManager *jmm) :
-    JMM(jmm?jmm:JITMemoryManager::CreateDefaultMemManager()) {}
-
-  uint8_t *allocateDataSection(uintptr_t Size, unsigned Alignment,
-                               unsigned SectionID) {
-    return JMM->allocateDataSection(Size, Alignment, SectionID);
-  }
-
-  uint8_t *allocateCodeSection(uintptr_t Size, unsigned Alignment,
-                               unsigned SectionID) {
-    return JMM->allocateCodeSection(Size, Alignment, SectionID);
-  }
-
-  virtual void *getPointerToNamedFunction(const std::string &Name,
-                                          bool AbortOnFailure = true) {
-    return JMM->getPointerToNamedFunction(Name, AbortOnFailure);
-  }
-
-};
-
-} // End llvm namespace
-
-#endif

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp Tue Nov 13 09:21:47 2012
@@ -17,6 +17,7 @@
 #include "RuntimeDyldELF.h"
 #include "RuntimeDyldMachO.h"
 #include "llvm/Support/Path.h"
+#include "llvm/Support/MathExtras.h"
 
 using namespace llvm;
 using namespace llvm::object;
@@ -27,16 +28,6 @@
 
 namespace llvm {
 
-namespace {
-  // Helper for extensive error checking in debug builds.
-  error_code Check(error_code Err) {
-    if (Err) {
-      report_fatal_error(Err.message());
-    }
-    return Err;
-  }
-} // end anonymous namespace
-
 // Resolve the relocations for all symbols we currently know about.
 void RuntimeDyldImpl::resolveRelocations() {
   // First, resolve relocations associated with external symbols.
@@ -45,7 +36,11 @@
   // Just iterate over the sections we have and resolve all the relocations
   // in them. Gross overkill, but it gets the job done.
   for (int i = 0, e = Sections.size(); i != e; ++i) {
-    reassignSectionAddress(i, Sections[i].LoadAddress);
+    uint64_t Addr = Sections[i].LoadAddress;
+    DEBUG(dbgs() << "Resolving relocations Section #" << i
+            << "\t" << format("%p", (uint8_t *)Addr)
+            << "\n");
+    resolveRelocationList(Relocations[i], Addr);
   }
 }
 
@@ -78,9 +73,9 @@
   // Used sections from the object file
   ObjSectionToIDMap LocalSections;
 
-  // Common symbols requiring allocation, and the total size required to
-  // allocate all common symbols.
+  // Common symbols requiring allocation, with their sizes and alignments
   CommonSymbolMap CommonSymbols;
+  // Maximum required total memory to allocate all common symbols
   uint64_t CommonSize = 0;
 
   error_code err;
@@ -100,10 +95,11 @@
     bool isCommon = flags & SymbolRef::SF_Common;
     if (isCommon) {
       // Add the common symbols to a list.  We'll allocate them all below.
+      uint64_t Align = getCommonSymbolAlignment(*i);
       uint64_t Size = 0;
       Check(i->getSize(Size));
-      CommonSize += Size;
-      CommonSymbols[*i] = Size;
+      CommonSize += Size + Align;
+      CommonSymbols[*i] = CommonSymbolInfo(Size, Align);
     } else {
       if (SymType == object::SymbolRef::ST_Function ||
           SymType == object::SymbolRef::ST_Data ||
@@ -190,7 +186,7 @@
   if (!Addr)
     report_fatal_error("Unable to allocate memory for common symbols!");
   uint64_t Offset = 0;
-  Sections.push_back(SectionEntry(Addr, TotalSize, TotalSize, 0));
+  Sections.push_back(SectionEntry(StringRef(), Addr, TotalSize, TotalSize, 0));
   memset(Addr, 0, TotalSize);
 
   DEBUG(dbgs() << "emitCommonSection SectionID: " << SectionID
@@ -201,11 +197,20 @@
   // Assign the address of each symbol
   for (CommonSymbolMap::const_iterator it = CommonSymbols.begin(),
        itEnd = CommonSymbols.end(); it != itEnd; it++) {
+    uint64_t Size = it->second.first;
+    uint64_t Align = it->second.second;
     StringRef Name;
     it->first.getName(Name);
+    if (Align) {
+      // This symbol has an alignment requirement.
+      uint64_t AlignOffset = OffsetToAlignment((uint64_t)Addr, Align);
+      Addr += AlignOffset;
+      Offset += AlignOffset;
+      DEBUG(dbgs() << "Allocating common symbol " << Name << " address " <<
+                      format("%p\n", Addr));
+    }
     Obj.updateSymbolAddress(it->first, (uint64_t)Addr);
     SymbolTable[Name.data()] = SymbolLoc(SectionID, Offset);
-    uint64_t Size = it->second;
     Offset += Size;
     Addr += Size;
   }
@@ -233,10 +238,12 @@
   bool IsVirtual;
   bool IsZeroInit;
   uint64_t DataSize;
+  StringRef Name;
   Check(Section.isRequiredForExecution(IsRequired));
   Check(Section.isVirtual(IsVirtual));
   Check(Section.isZeroInit(IsZeroInit));
   Check(Section.getSize(DataSize));
+  Check(Section.getName(Name));
 
   unsigned Allocate;
   unsigned SectionID = Sections.size();
@@ -264,6 +271,7 @@
       memcpy(Addr, pData, DataSize);
 
     DEBUG(dbgs() << "emitSection SectionID: " << SectionID
+                 << " Name: " << Name
                  << " obj addr: " << format("%p", pData)
                  << " new addr: " << format("%p", Addr)
                  << " DataSize: " << DataSize
@@ -279,6 +287,7 @@
     Allocate = 0;
     Addr = 0;
     DEBUG(dbgs() << "emitSection SectionID: " << SectionID
+                 << " Name: " << Name
                  << " obj addr: " << format("%p", data.data())
                  << " new addr: 0"
                  << " DataSize: " << DataSize
@@ -287,7 +296,8 @@
                  << "\n");
   }
 
-  Sections.push_back(SectionEntry(Addr, Allocate, DataSize,(uintptr_t)pData));
+  Sections.push_back(SectionEntry(Name, Addr, Allocate, DataSize,
+				  (uintptr_t)pData));
   return SectionID;
 }
 
@@ -353,6 +363,24 @@
     StubAddr++;
     *StubAddr = NopInstr;
     return Addr;
+  } else if (Arch == Triple::ppc64) {
+    // PowerPC64 stub: the address points to a function descriptor
+    // instead of the function itself. Load the function address
+    // on r11 and sets it to control register. Also loads the function
+    // TOC in r2 and environment pointer to r11.
+    writeInt32BE(Addr,    0x3D800000); // lis   r12, highest(addr)
+    writeInt32BE(Addr+4,  0x618C0000); // ori   r12, higher(addr)
+    writeInt32BE(Addr+8,  0x798C07C6); // sldi  r12, r12, 32
+    writeInt32BE(Addr+12, 0x658C0000); // oris  r12, r12, h(addr)
+    writeInt32BE(Addr+16, 0x618C0000); // ori   r12, r12, l(addr)
+    writeInt32BE(Addr+20, 0xF8410028); // std   r2,  40(r1)
+    writeInt32BE(Addr+24, 0xE96C0000); // ld    r11, 0(r12)
+    writeInt32BE(Addr+28, 0xE84C0008); // ld    r2,  0(r12)
+    writeInt32BE(Addr+32, 0x7D6903A6); // mtctr r11
+    writeInt32BE(Addr+36, 0xE96C0010); // ld    r11, 16(r2)
+    writeInt32BE(Addr+40, 0x4E800420); // bctr
+
+    return Addr;
   }
   return Addr;
 }
@@ -363,31 +391,29 @@
                                              uint64_t Addr) {
   // The address to use for relocation resolution is not
   // the address of the local section buffer. We must be doing
-  // a remote execution environment of some sort. Re-apply any
-  // relocations referencing this section with the given address.
+  // a remote execution environment of some sort. Relocations can't
+  // be applied until all the sections have been moved.  The client must
+  // trigger this with a call to MCJIT::finalize() or
+  // RuntimeDyld::resolveRelocations().
   //
   // Addr is a uint64_t because we can't assume the pointer width
   // of the target is the same as that of the host. Just use a generic
   // "big enough" type.
   Sections[SectionID].LoadAddress = Addr;
-  DEBUG(dbgs() << "Resolving relocations Section #" << SectionID
-          << "\t" << format("%p", (uint8_t *)Addr)
-          << "\n");
-  resolveRelocationList(Relocations[SectionID], Addr);
 }
 
 void RuntimeDyldImpl::resolveRelocationEntry(const RelocationEntry &RE,
                                              uint64_t Value) {
   // Ignore relocations for sections that were not loaded
   if (Sections[RE.SectionID].Address != 0) {
-    uint8_t *Target = Sections[RE.SectionID].Address + RE.Offset;
     DEBUG(dbgs() << "\tSectionID: " << RE.SectionID
-          << " + " << RE.Offset << " (" << format("%p", Target) << ")"
+          << " + " << RE.Offset << " ("
+          << format("%p", Sections[RE.SectionID].Address + RE.Offset) << ")"
           << " RelType: " << RE.RelType
           << " Addend: " << RE.Addend
           << "\n");
 
-    resolveRelocation(Target, Sections[RE.SectionID].LoadAddress + RE.Offset,
+    resolveRelocation(Sections[RE.SectionID], RE.Offset,
                       Value, RE.RelType, RE.Addend);
   }
 }

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp Tue Nov 13 09:21:47 2012
@@ -30,6 +30,14 @@
 
 namespace {
 
+static inline
+error_code check(error_code Err) {
+  if (Err) {
+    report_fatal_error(Err.message());
+  }
+  return Err;
+}
+
 template<support::endianness target_endianness, bool is64Bits>
 class DyldELFObject : public ELFObjectFile<target_endianness, is64Bits> {
   LLVM_ELF_IMPORT_TYPES(target_endianness, is64Bits)
@@ -179,8 +187,8 @@
 RuntimeDyldELF::~RuntimeDyldELF() {
 }
 
-void RuntimeDyldELF::resolveX86_64Relocation(uint8_t *LocalAddress,
-                                             uint64_t FinalAddress,
+void RuntimeDyldELF::resolveX86_64Relocation(const SectionEntry &Section,
+                                             uint64_t Offset,
                                              uint64_t Value,
                                              uint32_t Type,
                                              int64_t Addend) {
@@ -189,8 +197,10 @@
     llvm_unreachable("Relocation type not implemented yet!");
   break;
   case ELF::R_X86_64_64: {
-    uint64_t *Target = (uint64_t*)(LocalAddress);
+    uint64_t *Target = reinterpret_cast<uint64_t*>(Section.Address + Offset);
     *Target = Value + Addend;
+    DEBUG(dbgs() << "Writing " << format("%p", (Value + Addend))
+                 << " at " << format("%p\n",Target));
     break;
   }
   case ELF::R_X86_64_32:
@@ -200,37 +210,52 @@
            (Type == ELF::R_X86_64_32S && 
              ((int64_t)Value <= INT32_MAX && (int64_t)Value >= INT32_MIN)));
     uint32_t TruncatedAddr = (Value & 0xFFFFFFFF);
-    uint32_t *Target = reinterpret_cast<uint32_t*>(LocalAddress);
+    uint32_t *Target = reinterpret_cast<uint32_t*>(Section.Address + Offset);
     *Target = TruncatedAddr;
+    DEBUG(dbgs() << "Writing " << format("%p", TruncatedAddr)
+                 << " at " << format("%p\n",Target));
     break;
   }
   case ELF::R_X86_64_PC32: {
-    uint32_t *Placeholder = reinterpret_cast<uint32_t*>(LocalAddress);
+    // Get the placeholder value from the generated object since
+    // a previous relocation attempt may have overwritten the loaded version
+    uint32_t *Placeholder = reinterpret_cast<uint32_t*>(Section.ObjAddress
+                                                                   + Offset);
+    uint32_t *Target = reinterpret_cast<uint32_t*>(Section.Address + Offset);
+    uint64_t  FinalAddress = Section.LoadAddress + Offset;
     int64_t RealOffset = *Placeholder + Value + Addend - FinalAddress;
     assert(RealOffset <= INT32_MAX && RealOffset >= INT32_MIN);
     int32_t TruncOffset = (RealOffset & 0xFFFFFFFF);
-    *Placeholder = TruncOffset;
+    *Target = TruncOffset;
     break;
   }
   }
 }
 
-void RuntimeDyldELF::resolveX86Relocation(uint8_t *LocalAddress,
-                                          uint32_t FinalAddress,
+void RuntimeDyldELF::resolveX86Relocation(const SectionEntry &Section,
+                                          uint64_t Offset,
                                           uint32_t Value,
                                           uint32_t Type,
                                           int32_t Addend) {
   switch (Type) {
   case ELF::R_386_32: {
-    uint32_t *Target = (uint32_t*)(LocalAddress);
-    uint32_t Placeholder = *Target;
-    *Target = Placeholder + Value + Addend;
+    // Get the placeholder value from the generated object since
+    // a previous relocation attempt may have overwritten the loaded version
+    uint32_t *Placeholder = reinterpret_cast<uint32_t*>(Section.ObjAddress
+                                                                   + Offset);
+    uint32_t *Target = reinterpret_cast<uint32_t*>(Section.Address + Offset);
+    *Target = *Placeholder + Value + Addend;
     break;
   }
   case ELF::R_386_PC32: {
-    uint32_t *Placeholder = reinterpret_cast<uint32_t*>(LocalAddress);
+    // Get the placeholder value from the generated object since
+    // a previous relocation attempt may have overwritten the loaded version
+    uint32_t *Placeholder = reinterpret_cast<uint32_t*>(Section.ObjAddress
+                                                                   + Offset);
+    uint32_t *Target = reinterpret_cast<uint32_t*>(Section.Address + Offset);
+    uint32_t  FinalAddress = ((Section.LoadAddress + Offset) & 0xFFFFFFFF);
     uint32_t RealOffset = *Placeholder + Value + Addend - FinalAddress;
-    *Placeholder = RealOffset;
+    *Target = RealOffset;
     break;
     }
     default:
@@ -241,16 +266,18 @@
   }
 }
 
-void RuntimeDyldELF::resolveARMRelocation(uint8_t *LocalAddress,
-                                          uint32_t FinalAddress,
+void RuntimeDyldELF::resolveARMRelocation(const SectionEntry &Section,
+                                          uint64_t Offset,
                                           uint32_t Value,
                                           uint32_t Type,
                                           int32_t Addend) {
   // TODO: Add Thumb relocations.
-  uint32_t* TargetPtr = (uint32_t*)LocalAddress;
+  uint32_t* TargetPtr = (uint32_t*)(Section.Address + Offset);
+  uint32_t FinalAddress = ((Section.LoadAddress + Offset) & 0xFFFFFFFF);
   Value += Addend;
 
-  DEBUG(dbgs() << "resolveARMRelocation, LocalAddress: " << LocalAddress
+  DEBUG(dbgs() << "resolveARMRelocation, LocalAddress: "
+               << Section.Address + Offset
                << " FinalAddress: " << format("%p",FinalAddress)
                << " Value: " << format("%x",Value)
                << " Type: " << format("%x",Type)
@@ -302,16 +329,18 @@
   }
 }
 
-void RuntimeDyldELF::resolveMIPSRelocation(uint8_t *LocalAddress,
-                                           uint32_t FinalAddress,
+void RuntimeDyldELF::resolveMIPSRelocation(const SectionEntry &Section,
+                                           uint64_t Offset,
                                            uint32_t Value,
                                            uint32_t Type,
                                            int32_t Addend) {
-  uint32_t* TargetPtr = (uint32_t*)LocalAddress;
+  uint32_t* TargetPtr = (uint32_t*)(Section.Address + Offset);
   Value += Addend;
 
-  DEBUG(dbgs() << "resolveMipselocation, LocalAddress: " << LocalAddress
-               << " FinalAddress: " << format("%p",FinalAddress)
+  DEBUG(dbgs() << "resolveMipselocation, LocalAddress: "
+               << Section.Address + Offset
+               << " FinalAddress: "
+               << format("%p",Section.LoadAddress + Offset)
                << " Value: " << format("%x",Value)
                << " Type: " << format("%x",Type)
                << " Addend: " << format("%x",Addend)
@@ -340,32 +369,210 @@
    }
 }
 
-void RuntimeDyldELF::resolveRelocation(uint8_t *LocalAddress,
-                                       uint64_t FinalAddress,
+// Return the .TOC. section address to R_PPC64_TOC relocations.
+uint64_t RuntimeDyldELF::findPPC64TOC() const {
+  // The TOC consists of sections .got, .toc, .tocbss, .plt in that
+  // order. The TOC starts where the first of these sections starts.
+  SectionList::const_iterator it = Sections.begin();
+  SectionList::const_iterator ite = Sections.end();
+  for (; it != ite; ++it) {
+    if (it->Name == ".got" ||
+        it->Name == ".toc" ||
+        it->Name == ".tocbss" ||
+        it->Name == ".plt")
+      break;
+  }
+  if (it == ite) {
+    // This may happen for
+    // * references to TOC base base (sym at toc, .odp relocation) without
+    // a .toc directive.
+    // In this case just use the first section (which is usually
+    // the .odp) since the code won't reference the .toc base
+    // directly.
+    it = Sections.begin();
+  }
+  assert (it != ite);
+  // Per the ppc64-elf-linux ABI, The TOC base is TOC value plus 0x8000
+  // thus permitting a full 64 Kbytes segment.
+  return it->LoadAddress + 0x8000;
+}
+
+// Returns the sections and offset associated with the ODP entry referenced
+// by Symbol.
+void RuntimeDyldELF::findOPDEntrySection(ObjectImage &Obj,
+                                         ObjSectionToIDMap &LocalSections,
+                                         RelocationValueRef &Rel) {
+  // Get the ELF symbol value (st_value) to compare with Relocation offset in
+  // .opd entries
+
+  error_code err;
+  for (section_iterator si = Obj.begin_sections(),
+     se = Obj.end_sections(); si != se; si.increment(err)) {
+    StringRef SectionName;
+    check(si->getName(SectionName));
+    if (SectionName != ".opd")
+      continue;
+
+    for (relocation_iterator i = si->begin_relocations(),
+         e = si->end_relocations(); i != e;) {
+      check(err);
+
+      // The R_PPC64_ADDR64 relocation indicates the first field
+      // of a .opd entry
+      uint64_t TypeFunc;
+      check(i->getType(TypeFunc));
+      if (TypeFunc != ELF::R_PPC64_ADDR64) {
+        i.increment(err);
+        continue;
+      }
+
+      SymbolRef TargetSymbol;
+      uint64_t TargetSymbolOffset;
+      int64_t TargetAdditionalInfo;
+      check(i->getSymbol(TargetSymbol));
+      check(i->getOffset(TargetSymbolOffset));
+      check(i->getAdditionalInfo(TargetAdditionalInfo));
+
+      i = i.increment(err);
+      if (i == e)
+        break;
+      check(err);
+
+      // Just check if following relocation is a R_PPC64_TOC
+      uint64_t TypeTOC;
+      check(i->getType(TypeTOC));
+      if (TypeTOC != ELF::R_PPC64_TOC)
+        continue;
+
+      // Finally compares the Symbol value and the target symbol offset
+      // to check if this .opd entry refers to the symbol the relocation
+      // points to.
+      if (Rel.Addend != (intptr_t)TargetSymbolOffset)
+        continue;
+
+      section_iterator tsi(Obj.end_sections());
+      check(TargetSymbol.getSection(tsi));
+      Rel.SectionID = findOrEmitSection(Obj, (*tsi), true, LocalSections);
+      Rel.Addend = (intptr_t)TargetAdditionalInfo;
+      return;
+    }
+  }
+  llvm_unreachable("Attempting to get address of ODP entry!");
+}
+
+// Relocation masks following the #lo(value), #hi(value), #higher(value),
+// and #highest(value) macros defined in section 4.5.1. Relocation Types
+// in PPC-elf64abi document.
+//
+static inline
+uint16_t applyPPClo (uint64_t value)
+{
+  return value & 0xffff;
+}
+
+static inline
+uint16_t applyPPChi (uint64_t value)
+{
+  return (value >> 16) & 0xffff;
+}
+
+static inline
+uint16_t applyPPChigher (uint64_t value)
+{
+  return (value >> 32) & 0xffff;
+}
+
+static inline
+uint16_t applyPPChighest (uint64_t value)
+{
+  return (value >> 48) & 0xffff;
+}
+
+void RuntimeDyldELF::resolvePPC64Relocation(const SectionEntry &Section,
+                                            uint64_t Offset,
+                                            uint64_t Value,
+                                            uint32_t Type,
+                                            int64_t Addend) {
+  uint8_t* LocalAddress = Section.Address + Offset;
+  switch (Type) {
+  default:
+    llvm_unreachable("Relocation type not implemented yet!");
+  break;
+  case ELF::R_PPC64_ADDR16_LO :
+    writeInt16BE(LocalAddress, applyPPClo (Value + Addend));
+    break;
+  case ELF::R_PPC64_ADDR16_HI :
+    writeInt16BE(LocalAddress, applyPPChi (Value + Addend));
+    break;
+  case ELF::R_PPC64_ADDR16_HIGHER :
+    writeInt16BE(LocalAddress, applyPPChigher (Value + Addend));
+    break;
+  case ELF::R_PPC64_ADDR16_HIGHEST :
+    writeInt16BE(LocalAddress, applyPPChighest (Value + Addend));
+    break;
+  case ELF::R_PPC64_ADDR14 : {
+    assert(((Value + Addend) & 3) == 0);
+    // Preserve the AA/LK bits in the branch instruction
+    uint8_t aalk = *(LocalAddress+3);
+    writeInt16BE(LocalAddress + 2, (aalk & 3) | ((Value + Addend) & 0xfffc));
+  } break;
+  case ELF::R_PPC64_REL24 : {
+    uint64_t FinalAddress = (Section.LoadAddress + Offset);
+    int32_t delta = static_cast<int32_t>(Value - FinalAddress + Addend);
+    if (SignExtend32<24>(delta) != delta)
+      llvm_unreachable("Relocation R_PPC64_REL24 overflow");
+    // Generates a 'bl <address>' instruction
+    writeInt32BE(LocalAddress, 0x48000001 | (delta & 0x03FFFFFC));
+  } break;
+  case ELF::R_PPC64_ADDR64 :
+    writeInt64BE(LocalAddress, Value + Addend);
+    break;
+  case ELF::R_PPC64_TOC :
+    writeInt64BE(LocalAddress, findPPC64TOC());
+    break;
+  case ELF::R_PPC64_TOC16 : {
+    uint64_t TOCStart = findPPC64TOC();
+    Value = applyPPClo((Value + Addend) - TOCStart);
+    writeInt16BE(LocalAddress, applyPPClo(Value));
+  } break;
+  case ELF::R_PPC64_TOC16_DS : {
+    uint64_t TOCStart = findPPC64TOC();
+    Value = ((Value + Addend) - TOCStart);
+    writeInt16BE(LocalAddress, applyPPClo(Value));
+  } break;
+  }
+}
+
+
+void RuntimeDyldELF::resolveRelocation(const SectionEntry &Section,
+                                       uint64_t Offset,
                                        uint64_t Value,
                                        uint32_t Type,
                                        int64_t Addend) {
   switch (Arch) {
   case Triple::x86_64:
-    resolveX86_64Relocation(LocalAddress, FinalAddress, Value, Type, Addend);
+    resolveX86_64Relocation(Section, Offset, Value, Type, Addend);
     break;
   case Triple::x86:
-    resolveX86Relocation(LocalAddress, (uint32_t)(FinalAddress & 0xffffffffL),
+    resolveX86Relocation(Section, Offset,
                          (uint32_t)(Value & 0xffffffffL), Type,
                          (uint32_t)(Addend & 0xffffffffL));
     break;
   case Triple::arm:    // Fall through.
   case Triple::thumb:
-    resolveARMRelocation(LocalAddress, (uint32_t)(FinalAddress & 0xffffffffL),
+    resolveARMRelocation(Section, Offset,
                          (uint32_t)(Value & 0xffffffffL), Type,
                          (uint32_t)(Addend & 0xffffffffL));
     break;
   case Triple::mips:    // Fall through.
   case Triple::mipsel:
-    resolveMIPSRelocation(LocalAddress, (uint32_t)(FinalAddress & 0xffffffffL),
+    resolveMIPSRelocation(Section, Offset,
                           (uint32_t)(Value & 0xffffffffL), Type,
                           (uint32_t)(Addend & 0xffffffffL));
     break;
+  case Triple::ppc64:
+    resolvePPC64Relocation(Section, Offset, Value, Type, Addend);
+    break;
   default: llvm_unreachable("Unsupported CPU type!");
   }
 }
@@ -390,6 +597,8 @@
   RelocationValueRef Value;
   // First search for the symbol in the local symbol table
   SymbolTableMap::const_iterator lsi = Symbols.find(TargetName.data());
+  SymbolRef::Type SymType;
+  Symbol.getType(SymType);
   if (lsi != Symbols.end()) {
     Value.SectionID = lsi->second.first;
     Value.Addend = lsi->second.second;
@@ -401,8 +610,6 @@
       Value.SectionID = gsi->second.first;
       Value.Addend = gsi->second.second;
     } else {
-      SymbolRef::Type SymType;
-      Symbol.getType(SymType);
       switch (SymType) {
         case SymbolRef::ST_Debug: {
           // TODO: Now ELF SymbolRef::ST_Debug = STT_SECTION, it's not obviously
@@ -444,13 +651,12 @@
     // This is an ARM branch relocation, need to use a stub function.
     DEBUG(dbgs() << "\t\tThis is an ARM branch relocation.");
     SectionEntry &Section = Sections[Rel.SectionID];
-    uint8_t *Target = Section.Address + Rel.Offset;
 
-    //  Look up for existing stub.
+    // Look for an existing stub.
     StubMap::const_iterator i = Stubs.find(Value);
     if (i != Stubs.end()) {
-      resolveRelocation(Target, (uint64_t)Target, (uint64_t)Section.Address +
-                        i->second, RelType, 0);
+        resolveRelocation(Section, Rel.Offset,
+                          (uint64_t)Section.Address + i->second, RelType, 0);
       DEBUG(dbgs() << " Stub function found\n");
     } else {
       // Create a new stub function.
@@ -465,8 +671,9 @@
       else
         addRelocationForSection(RE, Value.SectionID);
 
-      resolveRelocation(Target, (uint64_t)Target, (uint64_t)Section.Address +
-                        Section.StubOffset, RelType, 0);
+      resolveRelocation(Section, Rel.Offset,
+                        (uint64_t)Section.Address + Section.StubOffset,
+                        RelType, 0);
       Section.StubOffset += getMaxStubSize();
     }
   } else if (Arch == Triple::mipsel && RelType == ELF::R_MIPS_26) {
@@ -484,9 +691,8 @@
     //  Look up for existing stub.
     StubMap::const_iterator i = Stubs.find(Value);
     if (i != Stubs.end()) {
-      resolveRelocation(Target, (uint64_t)Target,
-                        (uint64_t)Section.Address +
-                        i->second, RelType, 0);
+      resolveRelocation(Section, Rel.Offset,
+                        (uint64_t)Section.Address + i->second, RelType, 0);
       DEBUG(dbgs() << " Stub function found\n");
     } else {
       // Create a new stub function.
@@ -511,11 +717,99 @@
         addRelocationForSection(RELo, Value.SectionID);
       }
 
-      resolveRelocation(Target, (uint64_t)Target,
-                        (uint64_t)Section.Address +
-                        Section.StubOffset, RelType, 0);
+      resolveRelocation(Section, Rel.Offset,
+                        (uint64_t)Section.Address + Section.StubOffset,
+                        RelType, 0);
       Section.StubOffset += getMaxStubSize();
     }
+  } else if (Arch == Triple::ppc64) {
+    if (RelType == ELF::R_PPC64_REL24) {
+      // A PPC branch relocation will need a stub function if the target is
+      // an external symbol (Symbol::ST_Unknown) or if the target address
+      // is not within the signed 24-bits branch address.
+      SectionEntry &Section = Sections[Rel.SectionID];
+      uint8_t *Target = Section.Address + Rel.Offset;
+      bool RangeOverflow = false;
+      if (SymType != SymbolRef::ST_Unknown) {
+        // A function call may points to the .opd entry, so the final symbol value
+        // in calculated based in the relocation values in .opd section.
+        findOPDEntrySection(Obj, ObjSectionToID, Value);
+        uint8_t *RelocTarget = Sections[Value.SectionID].Address + Value.Addend;
+        int32_t delta = static_cast<int32_t>(Target - RelocTarget);
+        // If it is within 24-bits branch range, just set the branch target
+        if (SignExtend32<24>(delta) == delta) {
+          RelocationEntry RE(Rel.SectionID, Rel.Offset, RelType, Value.Addend);
+          if (Value.SymbolName)
+            addRelocationForSymbol(RE, Value.SymbolName);
+          else
+            addRelocationForSection(RE, Value.SectionID);
+        } else {
+          RangeOverflow = true;
+        }
+      }
+      if (SymType == SymbolRef::ST_Unknown || RangeOverflow == true) {
+        // It is an external symbol (SymbolRef::ST_Unknown) or within a range
+        // larger than 24-bits.
+        StubMap::const_iterator i = Stubs.find(Value);
+        if (i != Stubs.end()) {
+          // Symbol function stub already created, just relocate to it
+          resolveRelocation(Section, Rel.Offset,
+                            (uint64_t)Section.Address + i->second, RelType, 0);
+          DEBUG(dbgs() << " Stub function found\n");
+        } else {
+          // Create a new stub function.
+          DEBUG(dbgs() << " Create a new stub function\n");
+          Stubs[Value] = Section.StubOffset;
+          uint8_t *StubTargetAddr = createStubFunction(Section.Address +
+                                                       Section.StubOffset);
+          RelocationEntry RE(Rel.SectionID, StubTargetAddr - Section.Address,
+                             ELF::R_PPC64_ADDR64, Value.Addend);
+
+          // Generates the 64-bits address loads as exemplified in section
+          // 4.5.1 in PPC64 ELF ABI.
+          RelocationEntry REhst(Rel.SectionID,
+                                StubTargetAddr - Section.Address + 2,
+                                ELF::R_PPC64_ADDR16_HIGHEST, Value.Addend);
+          RelocationEntry REhr(Rel.SectionID,
+                               StubTargetAddr - Section.Address + 6,
+                               ELF::R_PPC64_ADDR16_HIGHER, Value.Addend);
+          RelocationEntry REh(Rel.SectionID,
+                              StubTargetAddr - Section.Address + 14,
+                              ELF::R_PPC64_ADDR16_HI, Value.Addend);
+          RelocationEntry REl(Rel.SectionID,
+                              StubTargetAddr - Section.Address + 18,
+                              ELF::R_PPC64_ADDR16_LO, Value.Addend);
+
+          if (Value.SymbolName) {
+            addRelocationForSymbol(REhst, Value.SymbolName);
+            addRelocationForSymbol(REhr,  Value.SymbolName);
+            addRelocationForSymbol(REh,   Value.SymbolName);
+            addRelocationForSymbol(REl,   Value.SymbolName);
+          } else {
+            addRelocationForSection(REhst, Value.SectionID);
+            addRelocationForSection(REhr,  Value.SectionID);
+            addRelocationForSection(REh,   Value.SectionID);
+            addRelocationForSection(REl,   Value.SectionID);
+          }
+
+          resolveRelocation(Section, Rel.Offset,
+                            (uint64_t)Section.Address + Section.StubOffset,
+                            RelType, 0);
+          if (SymType == SymbolRef::ST_Unknown)
+            // Restore the TOC for external calls
+            writeInt32BE(Target+4, 0xE8410028); // ld r2,40(r1)
+          Section.StubOffset += getMaxStubSize();
+        }
+      }
+    } else {
+      RelocationEntry RE(Rel.SectionID, Rel.Offset, RelType, Value.Addend);
+      // Extra check to avoid relocation againt empty symbols (usually
+      // the R_PPC64_TOC).
+      if (Value.SymbolName && !TargetName.empty())
+        addRelocationForSymbol(RE, Value.SymbolName);
+      else
+        addRelocationForSection(RE, Value.SectionID);
+    }
   } else {
     RelocationEntry RE(Rel.SectionID, Rel.Offset, RelType, Value.Addend);
     if (Value.SymbolName)
@@ -525,6 +819,13 @@
   }
 }
 
+unsigned RuntimeDyldELF::getCommonSymbolAlignment(const SymbolRef &Sym) {
+  // In ELF, the value of an SHN_COMMON symbol is its alignment requirement.
+  uint64_t Align;
+  Check(Sym.getValue(Align));
+  return Align;
+}
+
 bool RuntimeDyldELF::isCompatibleFormat(const ObjectBuffer *Buffer) const {
   if (Buffer->getBufferSize() < strlen(ELF::ElfMagic))
     return false;

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h Tue Nov 13 09:21:47 2012
@@ -18,36 +18,52 @@
 
 using namespace llvm;
 
-
 namespace llvm {
+
+namespace {
+  // Helper for extensive error checking in debug builds.
+  error_code Check(error_code Err) {
+    if (Err) {
+      report_fatal_error(Err.message());
+    }
+    return Err;
+  }
+} // end anonymous namespace
+
 class RuntimeDyldELF : public RuntimeDyldImpl {
 protected:
-  void resolveX86_64Relocation(uint8_t *LocalAddress,
-                               uint64_t FinalAddress,
+  void resolveX86_64Relocation(const SectionEntry &Section,
+                               uint64_t Offset,
                                uint64_t Value,
                                uint32_t Type,
                                int64_t Addend);
 
-  void resolveX86Relocation(uint8_t *LocalAddress,
-                            uint32_t FinalAddress,
+  void resolveX86Relocation(const SectionEntry &Section,
+                            uint64_t Offset,
                             uint32_t Value,
                             uint32_t Type,
                             int32_t Addend);
 
-  void resolveARMRelocation(uint8_t *LocalAddress,
-                            uint32_t FinalAddress,
+  void resolveARMRelocation(const SectionEntry &Section,
+                            uint64_t Offset,
                             uint32_t Value,
                             uint32_t Type,
                             int32_t Addend);
 
-  void resolveMIPSRelocation(uint8_t *LocalAddress,
-                             uint32_t FinalAddress,
+  void resolveMIPSRelocation(const SectionEntry &Section,
+                             uint64_t Offset,
                              uint32_t Value,
                              uint32_t Type,
                              int32_t Addend);
 
-  virtual void resolveRelocation(uint8_t *LocalAddress,
-                                 uint64_t FinalAddress,
+  void resolvePPC64Relocation(const SectionEntry &Section,
+                              uint64_t Offset,
+                              uint64_t Value,
+                              uint32_t Type,
+                              int64_t Addend);
+
+  virtual void resolveRelocation(const SectionEntry &Section,
+                                 uint64_t Offset,
                                  uint64_t Value,
                                  uint32_t Type,
                                  int64_t Addend);
@@ -58,8 +74,15 @@
                                     const SymbolTableMap &Symbols,
                                     StubMap &Stubs);
 
+  unsigned getCommonSymbolAlignment(const SymbolRef &Sym);
+
   virtual ObjectImage *createObjectImage(ObjectBuffer *InputBuffer);
 
+  uint64_t findPPC64TOC() const;
+  void findOPDEntrySection(ObjectImage &Obj,
+                           ObjSectionToIDMap &LocalSections,
+                           RelocationValueRef &Rel);
+
 public:
   RuntimeDyldELF(RTDyldMemoryManager *mm)
       : RuntimeDyldImpl(mm) {}

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h Tue Nov 13 09:21:47 2012
@@ -24,6 +24,8 @@
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/Format.h"
+#include "llvm/Support/Host.h"
+#include "llvm/Support/SwapByteOrder.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Support/system_error.h"
 #include <map>
@@ -41,6 +43,9 @@
 /// linker.
 class SectionEntry {
 public:
+  /// Name - section name.
+  StringRef Name;
+
   /// Address - address in the linker's memory where the section resides.
   uint8_t *Address;
 
@@ -61,9 +66,9 @@
   /// for calculating relocations in some object formats (like MachO).
   uintptr_t ObjAddress;
 
-  SectionEntry(uint8_t *address, size_t size, uintptr_t stubOffset,
-               uintptr_t objAddress)
-    : Address(address), Size(size), LoadAddress((uintptr_t)address),
+  SectionEntry(StringRef name, uint8_t *address, size_t size,
+	       uintptr_t stubOffset, uintptr_t objAddress)
+    : Name(name), Address(address), Size(size), LoadAddress((uintptr_t)address),
       StubOffset(stubOffset), ObjAddress(objAddress) {}
 };
 
@@ -135,8 +140,10 @@
   typedef StringMap<SymbolLoc> SymbolTableMap;
   SymbolTableMap GlobalSymbolTable;
 
-  // Keep a map of common symbols to their sizes
-  typedef std::map<SymbolRef, unsigned> CommonSymbolMap;
+  // Pair representing the size and alignment requirement for a common symbol.
+  typedef std::pair<unsigned, unsigned> CommonSymbolInfo;
+  // Keep a map of common symbols to their info pairs
+  typedef std::map<SymbolRef, CommonSymbolInfo> CommonSymbolMap;
 
   // For each symbol, keep a list of relocations based on it. Anytime
   // its address is reassigned (the JIT re-compiled the function, e.g.),
@@ -163,6 +170,8 @@
       return 8; // 32-bit instruction and 32-bit address
     else if (Arch == Triple::mipsel)
       return 16;
+    else if (Arch == Triple::ppc64)
+      return 44;
     else
       return 0;
   }
@@ -185,6 +194,42 @@
     return (uint8_t*)Sections[SectionID].Address;
   }
 
+  // Subclasses can override this method to get the alignment requirement of
+  // a common symbol. Returns no alignment requirement if not implemented.
+  virtual unsigned getCommonSymbolAlignment(const SymbolRef &Sym) {
+    return 0;
+  }
+
+
+  void writeInt16BE(uint8_t *Addr, uint16_t Value) {
+    if (sys::isLittleEndianHost())
+      Value = sys::SwapByteOrder(Value);
+    *Addr     = (Value >> 8) & 0xFF;
+    *(Addr+1) = Value & 0xFF;
+  }
+
+  void writeInt32BE(uint8_t *Addr, uint32_t Value) {
+    if (sys::isLittleEndianHost())
+      Value = sys::SwapByteOrder(Value);
+    *Addr     = (Value >> 24) & 0xFF;
+    *(Addr+1) = (Value >> 16) & 0xFF;
+    *(Addr+2) = (Value >> 8) & 0xFF;
+    *(Addr+3) = Value & 0xFF;
+  }
+
+  void writeInt64BE(uint8_t *Addr, uint64_t Value) {
+    if (sys::isLittleEndianHost())
+      Value = sys::SwapByteOrder(Value);
+    *Addr     = (Value >> 56) & 0xFF;
+    *(Addr+1) = (Value >> 48) & 0xFF;
+    *(Addr+2) = (Value >> 40) & 0xFF;
+    *(Addr+3) = (Value >> 32) & 0xFF;
+    *(Addr+4) = (Value >> 24) & 0xFF;
+    *(Addr+5) = (Value >> 16) & 0xFF;
+    *(Addr+6) = (Value >> 8) & 0xFF;
+    *(Addr+7) = Value & 0xFF;
+  }
+
   /// \brief Given the common symbols discovered in the object file, emit a
   /// new section for them and update the symbol mappings in the object and
   /// symbol table.
@@ -227,16 +272,14 @@
   void resolveRelocationEntry(const RelocationEntry &RE, uint64_t Value);
 
   /// \brief A object file specific relocation resolver
-  /// \param LocalAddress The address to apply the relocation action
-  /// \param FinalAddress If the linker prepare code for remote executon then
-  ///                     FinalAddress has the remote address to apply the
-  ///                     relocation action, otherwise is same as LocalAddress
+  /// \param Section The section where the relocation is being applied
+  /// \param Offset The offset into the section for this relocation
   /// \param Value Target symbol address to apply the relocation action
   /// \param Type object file specific relocation type
   /// \param Addend A constant addend used to compute the value to be stored
   ///        into the relocatable field
-  virtual void resolveRelocation(uint8_t *LocalAddress,
-                                 uint64_t FinalAddress,
+  virtual void resolveRelocation(const SectionEntry &Section,
+                                 uint64_t Offset,
                                  uint64_t Value,
                                  uint32_t Type,
                                  int64_t Addend) = 0;

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp Tue Nov 13 09:21:47 2012
@@ -21,11 +21,13 @@
 
 namespace llvm {
 
-void RuntimeDyldMachO::resolveRelocation(uint8_t *LocalAddress,
-                                         uint64_t FinalAddress,
+void RuntimeDyldMachO::resolveRelocation(const SectionEntry &Section,
+                                         uint64_t Offset,
                                          uint64_t Value,
                                          uint32_t Type,
                                          int64_t Addend) {
+  uint8_t *LocalAddress = Section.Address + Offset;
+  uint64_t FinalAddress = Section.LoadAddress + Offset;
   bool isPCRel = (Type >> 24) & 1;
   unsigned MachoType = (Type >> 28) & 0xf;
   unsigned Size = 1 << ((Type >> 25) & 3);
@@ -211,7 +213,6 @@
   uint32_t RelType = (uint32_t) (Rel.Type & 0xffffffffL);
   RelocationValueRef Value;
   SectionEntry &Section = Sections[Rel.SectionID];
-  uint8_t *Target = Section.Address + Rel.Offset;
 
   bool isExtern = (RelType >> 27) & 1;
   if (isExtern) {
@@ -265,7 +266,7 @@
     //  Look up for existing stub.
     StubMap::const_iterator i = Stubs.find(Value);
     if (i != Stubs.end())
-      resolveRelocation(Target, (uint64_t)Target,
+      resolveRelocation(Section, Rel.Offset,
                         (uint64_t)Section.Address + i->second,
                         RelType, 0);
     else {
@@ -279,7 +280,7 @@
         addRelocationForSymbol(RE, Value.SymbolName);
       else
         addRelocationForSection(RE, Value.SectionID);
-      resolveRelocation(Target, (uint64_t)Target,
+      resolveRelocation(Section, Rel.Offset,
                         (uint64_t)Section.Address + Section.StubOffset,
                         RelType, 0);
       Section.StubOffset += getMaxStubSize();

Modified: llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h (original)
+++ llvm/branches/R600/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h Tue Nov 13 09:21:47 2012
@@ -55,8 +55,8 @@
                                     StubMap &Stubs);
 
 public:
-  virtual void resolveRelocation(uint8_t *LocalAddress,
-                                 uint64_t FinalAddress,
+  virtual void resolveRelocation(const SectionEntry &Section,
+                                 uint64_t Offset,
                                  uint64_t Value,
                                  uint32_t Type,
                                  int64_t Addend);

Modified: llvm/branches/R600/lib/MC/ELFObjectWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/ELFObjectWriter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/ELFObjectWriter.cpp (original)
+++ llvm/branches/R600/lib/MC/ELFObjectWriter.cpp Tue Nov 13 09:21:47 2012
@@ -133,6 +133,11 @@
                                    bool IsPCRel) const {
       return TargetObjectWriter->ExplicitRelSym(Asm, Target, F, Fixup, IsPCRel);
     }
+    const MCSymbol *undefinedExplicitRelSym(const MCValue &Target,
+                                            const MCFixup &Fixup,
+                                            bool IsPCRel) const {
+      return TargetObjectWriter->undefinedExplicitRelSym(Target, Fixup, IsPCRel);
+    }
 
     bool is64Bit() const { return TargetObjectWriter->is64Bit(); }
     bool hasRelocationAddend() const {
@@ -639,7 +644,7 @@
   if (ASymbol.isUndefined()) {
     if (Renamed)
       return Renamed;
-    return &ASymbol;
+    return undefinedExplicitRelSym(Target, Fixup, IsPCRel);
   }
 
   if (SD.isExternal()) {
@@ -721,10 +726,13 @@
       MCSymbolData &SD = Asm.getSymbolData(ASymbol);
       MCFragment *F = SD.getFragment();
 
-      Index = F->getParent()->getOrdinal() + 1;
-
-      // Offset of the symbol in the section
-      Value += Layout.getSymbolOffset(&SD);
+      if (F) {
+        Index = F->getParent()->getOrdinal() + 1;
+        // Offset of the symbol in the section
+        Value += Layout.getSymbolOffset(&SD);
+      } else {
+        Index = 0;
+      }
     } else {
       if (Asm.getSymbolData(Symbol).getFlags() & ELF_Other_Weakref)
         WeakrefUsedInReloc.insert(RelocSymbol);

Modified: llvm/branches/R600/lib/MC/MCDisassembler/Disassembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/MCDisassembler/Disassembler.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/MCDisassembler/Disassembler.cpp (original)
+++ llvm/branches/R600/lib/MC/MCDisassembler/Disassembler.cpp Tue Nov 13 09:21:47 2012
@@ -184,3 +184,17 @@
   }
   llvm_unreachable("Invalid DecodeStatus!");
 }
+
+//
+// LLVMSetDisasmOptions() sets the disassembler's options.  It returns 1 if it
+// can set all the Options and 0 otherwise.
+//
+int LLVMSetDisasmOptions(LLVMDisasmContextRef DCR, uint64_t Options){
+  if (Options & LLVMDisassembler_Option_UseMarkup){
+      LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
+      MCInstPrinter *IP = DC->getIP();
+      IP->setUseMarkup(1);
+      Options &= ~LLVMDisassembler_Option_UseMarkup;
+  }
+  return (Options == 0);
+}

Modified: llvm/branches/R600/lib/MC/MCDisassembler/EDDisassembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/MCDisassembler/EDDisassembler.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/MCDisassembler/EDDisassembler.cpp (original)
+++ llvm/branches/R600/lib/MC/MCDisassembler/EDDisassembler.cpp Tue Nov 13 09:21:47 2012
@@ -366,8 +366,9 @@
     instName = OpcodeToken.getString();
     instLoc = OpcodeToken.getLoc();
     
+    ParseInstructionInfo Info;
     if (NextToken.isNot(AsmToken::Eof) &&
-        TargetParser->ParseInstruction(instName, instLoc, operands))
+        TargetParser->ParseInstruction(Info, instName, instLoc, operands))
       ret = -1;
   } else {
     ret = -1;

Modified: llvm/branches/R600/lib/MC/MCELFObjectTargetWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/MCELFObjectTargetWriter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/MCELFObjectTargetWriter.cpp (original)
+++ llvm/branches/R600/lib/MC/MCELFObjectTargetWriter.cpp Tue Nov 13 09:21:47 2012
@@ -9,6 +9,8 @@
 
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/MC/MCELFObjectWriter.h"
+#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCValue.h"
 
 using namespace llvm;
 
@@ -35,6 +37,12 @@
   return NULL;
 }
 
+const MCSymbol *MCELFObjectTargetWriter::undefinedExplicitRelSym(const MCValue &Target,
+                                                                 const MCFixup &Fixup,
+                                                                 bool IsPCRel) const {
+  const MCSymbol &Symbol = Target.getSymA()->getSymbol();
+  return &Symbol.AliasedSymbol();
+}
 
 void MCELFObjectTargetWriter::adjustFixupOffset(const MCFixup &Fixup,
                                                 uint64_t &RelocOffset) {

Modified: llvm/branches/R600/lib/MC/MCInstPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/MCInstPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/MCInstPrinter.cpp (original)
+++ llvm/branches/R600/lib/MC/MCInstPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -36,3 +36,17 @@
       OS << " " << MAI.getCommentString() << " " << Annot;
   }
 }
+
+/// Utility functions to make adding mark ups simpler.
+StringRef MCInstPrinter::markup(StringRef s) const {
+  if (getUseMarkup())
+    return s;
+  else
+    return "";
+}
+StringRef MCInstPrinter::markup(StringRef a, StringRef b) const {
+  if (getUseMarkup())
+    return a;
+  else
+    return b;
+}

Modified: llvm/branches/R600/lib/MC/MCParser/AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/MC/MCParser/AsmParser.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/MC/MCParser/AsmParser.cpp (original)
+++ llvm/branches/R600/lib/MC/MCParser/AsmParser.cpp Tue Nov 13 09:21:47 2012
@@ -86,6 +86,28 @@
                      MemoryBuffer *I);
 };
 
+//struct AsmRewrite;
+struct ParseStatementInfo {
+  /// ParsedOperands - The parsed operands from the last parsed statement.
+  SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
+
+  /// Opcode - The opcode from the last parsed instruction.
+  unsigned Opcode;
+
+  SmallVectorImpl<AsmRewrite> *AsmRewrites;
+
+  ParseStatementInfo() : Opcode(~0U), AsmRewrites(0) {}
+  ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
+    : Opcode(~0), AsmRewrites(rewrites) {}
+
+  ~ParseStatementInfo() {
+    // Free any parsed operands.
+    for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
+      delete ParsedOperands[i];
+    ParsedOperands.clear();
+  }
+};
+
 /// \brief The concrete assembly parser instance.
 class AsmParser : public MCAsmParser {
   friend class GenericAsmParser;
@@ -132,6 +154,7 @@
   StringRef CppHashFilename;
   int64_t CppHashLineNumber;
   SMLoc CppHashLoc;
+  int CppHashBuf;
 
   /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
   unsigned AssemblerDialect;
@@ -142,13 +165,6 @@
   /// ParsingInlineAsm - Are we parsing ms-style inline assembly?
   bool ParsingInlineAsm;
 
-  /// ParsedOperands - The parsed operands from the last parsed statement.
-  SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
-
-  /// Opcode - The opcode from the last parsed instruction.  This is MS-style
-  /// inline asm specific.
-  unsigned Opcode;
-
 public:
   AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
             const MCAsmInfo &MAI);
@@ -192,7 +208,7 @@
 
   bool ParseMSInlineAsm(void *AsmLoc, std::string &AsmString,
                         unsigned &NumOutputs, unsigned &NumInputs,
-                        SmallVectorImpl<void *> &OpDecls,
+                        SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
                         SmallVectorImpl<std::string> &Constraints,
                         SmallVectorImpl<std::string> &Clobbers,
                         const MCInstrInfo *MII,
@@ -209,7 +225,7 @@
 private:
   void CheckForValidSection();
 
-  bool ParseStatement();
+  bool ParseStatement(ParseStatementInfo &Info);
   void EatToEndOfLine();
   bool ParseCppHashLineFilenameComment(const SMLoc &L);
 
@@ -316,10 +332,8 @@
   bool ParseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
   bool ParseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
 
-  // MS-style inline assembly parsing.
-  bool isInstruction() { return Opcode != (unsigned)~0x0; }
-  unsigned getOpcode() { return Opcode; }
-  void setOpcode(unsigned Value) { Opcode = Value; }
+  // "_emit"
+  bool ParseDirectiveEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info);
 };
 
 /// \brief Generic implementations of directive handling, etc. which is shared
@@ -445,8 +459,7 @@
   : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM),
     GenericParser(new GenericAsmParser), PlatformParser(0),
     CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0),
-    AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false),
-    Opcode(~0x0) {
+    AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false) {
   // Save the old handler.
   SavedDiagHandler = SrcMgr.getDiagHandler();
   SavedDiagContext = SrcMgr.getDiagContext();
@@ -585,7 +598,8 @@
 
   // While we have input, parse each statement.
   while (Lexer.isNot(AsmToken::Eof)) {
-    if (!ParseStatement()) continue;
+    ParseStatementInfo Info;
+    if (!ParseStatement(Info)) continue;
 
     // We had an error, validate that one was emitted and recover by skipping to
     // the next line.
@@ -1068,7 +1082,7 @@
 ///   ::= EndOfStatement
 ///   ::= Label* Directive ...Operands... EndOfStatement
 ///   ::= Label* Identifier OperandList* EndOfStatement
-bool AsmParser::ParseStatement() {
+bool AsmParser::ParseStatement(ParseStatementInfo &Info) {
   if (Lexer.is(AsmToken::EndOfStatement)) {
     Out.AddBlankLine();
     Lex();
@@ -1187,7 +1201,7 @@
         return false;
     }
 
-    return ParseStatement();
+    return false;
   }
 
   case AsmToken::Equal:
@@ -1341,6 +1355,10 @@
     return Error(IDLoc, "unknown directive");
   }
 
+  // _emit
+  if (ParsingInlineAsm && IDVal == "_emit")
+    return ParseDirectiveEmit(IDLoc, Info);
+
   CheckForValidSection();
 
   // Canonicalize the opcode to lower case.
@@ -1348,18 +1366,19 @@
   for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
     OpcodeStr.push_back(tolower(IDVal[i]));
 
-  bool HadError = getTargetParser().ParseInstruction(OpcodeStr.str(), IDLoc,
-                                                     ParsedOperands);
+  ParseInstructionInfo IInfo(Info.AsmRewrites);
+  bool HadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr.str(),
+                                                     IDLoc,Info.ParsedOperands);
 
   // Dump the parsed representation, if requested.
   if (getShowParsedOperands()) {
     SmallString<256> Str;
     raw_svector_ostream OS(Str);
     OS << "parsed instruction: [";
-    for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
+    for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
       if (i != 0)
         OS << ", ";
-      ParsedOperands[i]->print(OS);
+      Info.ParsedOperands[i]->print(OS);
     }
     OS << "]";
 
@@ -1371,9 +1390,26 @@
   // the instruction.
   if (!HadError && getContext().getGenDwarfForAssembly() &&
       getContext().getGenDwarfSection() == getStreamer().getCurrentSection() ) {
+
+     unsigned Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
+
+     // If we previously parsed a cpp hash file line comment then make sure the
+     // current Dwarf File is for the CppHashFilename if not then emit the
+     // Dwarf File table for it and adjust the line number for the .loc.
+     const std::vector<MCDwarfFile *> &MCDwarfFiles =
+       getContext().getMCDwarfFiles();
+     if (CppHashFilename.size() != 0) {
+       if(MCDwarfFiles[getContext().getGenDwarfFileNumber()]->getName() !=
+          CppHashFilename)
+	 getStreamer().EmitDwarfFileDirective(
+	   getContext().nextGenDwarfFileNumber(), StringRef(), CppHashFilename);
+
+       unsigned CppHashLocLineNo = SrcMgr.FindLineNumber(CppHashLoc,CppHashBuf);
+       Line = CppHashLineNumber - 1 + (Line - CppHashLocLineNo);
+     }
+
     getStreamer().EmitDwarfLocDirective(getContext().getGenDwarfFileNumber(),
-                                        SrcMgr.FindLineNumber(IDLoc, CurBuffer),
-                                        0, DWARF2_LINE_DEFAULT_IS_STMT ?
+                                        Line, 0, DWARF2_LINE_DEFAULT_IS_STMT ?
                                         DWARF2_FLAG_IS_STMT : 0, 0, 0,
                                         StringRef());
   }
@@ -1381,20 +1417,12 @@
   // If parsing succeeded, match the instruction.
   if (!HadError) {
     unsigned ErrorInfo;
-    HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, Opcode,
-                                                         ParsedOperands, Out,
-                                                         ErrorInfo,
+    HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
+                                                         Info.ParsedOperands,
+                                                         Out, ErrorInfo,
                                                          ParsingInlineAsm);
   }
 
-  // Free any parsed operands.  If parsing ms-style inline assembly the operands
-  // will be freed by the ParseMSInlineAsm() function.
-  if (!ParsingInlineAsm) {
-    for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
-      delete ParsedOperands[i];
-    ParsedOperands.clear();
-  }
-
   // Don't skip the rest of the line, the instruction parser is responsible for
   // that.
   return false;
@@ -1438,6 +1466,7 @@
   CppHashLoc = L;
   CppHashFilename = Filename;
   CppHashLineNumber = LineNumber;
+  CppHashBuf = CurBuffer;
 
   // Ignore any trailing characters, they're just comment.
   EatToEndOfLine();
@@ -3574,36 +3603,34 @@
   return false;
 }
 
-namespace {
-enum AsmRewriteKind {
-   AOK_Imm,
-   AOK_Input,
-   AOK_Output,
-   AOK_SizeDirective,
-   AOK_Skip
-};
+bool AsmParser::ParseDirectiveEmit(SMLoc IDLoc, ParseStatementInfo &Info) {
+  const MCExpr *Value;
+  SMLoc ExprLoc = getLexer().getLoc();
+  if (ParseExpression(Value))
+    return true;
+  const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
+  if (!MCE)
+    return Error(ExprLoc, "unexpected expression in _emit");
+  uint64_t IntValue = MCE->getValue();
+  if (!isUIntN(8, IntValue) && !isIntN(8, IntValue))
+    return Error(ExprLoc, "literal value out of range for directive");
 
-struct AsmRewrite {
-  AsmRewriteKind Kind;
-  SMLoc Loc;
-  unsigned Len;
-  unsigned Size;
-public:
-  AsmRewrite(AsmRewriteKind kind, SMLoc loc, unsigned len, unsigned size = 0)
-    : Kind(kind), Loc(loc), Len(len), Size(size) { }
-};
+  Info.AsmRewrites->push_back(AsmRewrite(AOK_Emit, IDLoc, 5));
+  return false;
 }
 
 bool AsmParser::ParseMSInlineAsm(void *AsmLoc, std::string &AsmString,
                                  unsigned &NumOutputs, unsigned &NumInputs,
-                                 SmallVectorImpl<void *> &OpDecls,
+                                 SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
                                  SmallVectorImpl<std::string> &Constraints,
                                  SmallVectorImpl<std::string> &Clobbers,
                                  const MCInstrInfo *MII,
                                  const MCInstPrinter *IP,
                                  MCAsmParserSemaCallback &SI) {
-  SmallVector<void*, 4> InputDecls;
-  SmallVector<void*, 4> OutputDecls;
+  SmallVector<void *, 4> InputDecls;
+  SmallVector<void *, 4> OutputDecls;
+  SmallVector<bool, 4> InputDeclsOffsetOf;
+  SmallVector<bool, 4> OutputDeclsOffsetOf;
   SmallVector<std::string, 4> InputConstraints;
   SmallVector<std::string, 4> OutputConstraints;
   std::set<std::string> ClobberRegs;
@@ -3617,44 +3644,27 @@
   unsigned InputIdx = 0;
   unsigned OutputIdx = 0;
   while (getLexer().isNot(AsmToken::Eof)) {
-    // Clear the opcode.
-    setOpcode(~0x0);
-
-    // Save the conditional ignore state of the parser prior to parsing the statement.
-    bool PreParseCondStateIgnore = TheCondState.Ignore;
-
-    // Save the starting point of this statement in case we need to skip it.
-    SMLoc Start = getLexer().getLoc();
-
-    if (ParseStatement())
+    ParseStatementInfo Info(&AsmStrRewrites);
+    if (ParseStatement(Info))
       return true;
 
-    // If PreParseCondStateIgnore is false, but TheCondState.Ignore is true, then we
-    // just parsed a directive that changed the state to ignore.  Don't skip 
-    // emitting this directive.
-    if (PreParseCondStateIgnore && TheCondState.Ignore) {
-      unsigned Len = getLexer().getLoc().getPointer() - Start.getPointer();
-      AsmStrRewrites.push_back(AsmRewrite(AOK_Skip, Start, Len));
-      continue;
-    }
-
-    if (isInstruction()) {
-      const MCInstrDesc &Desc = MII->get(getOpcode());
+    if (Info.Opcode != ~0U) {
+      const MCInstrDesc &Desc = MII->get(Info.Opcode);
 
       // Build the list of clobbers, outputs and inputs.
-      for (unsigned i = 1, e = ParsedOperands.size(); i != e; ++i) {
-        MCParsedAsmOperand *Operand = ParsedOperands[i];
+      for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
+        MCParsedAsmOperand *Operand = Info.ParsedOperands[i];
 
         // Immediate.
         if (Operand->isImm()) {
-          AsmStrRewrites.push_back(AsmRewrite(AOK_Imm,
-                                                Operand->getStartLoc(),
-                                                Operand->getNameLen()));
+          if (Operand->needAsmRewrite())
+            AsmStrRewrites.push_back(AsmRewrite(AOK_ImmPrefix,
+                                                Operand->getStartLoc()));
           continue;
         }
 
         // Register operand.
-        if (Operand->isReg()) {
+        if (Operand->isReg() && !Operand->isOffsetOf()) {
           unsigned NumDefs = Desc.getNumDefs();
           // Clobber.
           if (NumDefs && Operand->getMCOperandNum() < NumDefs) {
@@ -3672,33 +3682,31 @@
                                                     Size);
         if (OpDecl) {
           bool isOutput = (i == 1) && Desc.mayStore();
-          if (Operand->needSizeDirective())
+          if (!Operand->isOffsetOf() && Operand->needSizeDirective())
             AsmStrRewrites.push_back(AsmRewrite(AOK_SizeDirective,
-                                                  Operand->getStartLoc(), 0,
-                                                  Operand->getMemSize()));
-          
+                                                Operand->getStartLoc(),
+                                                /*Len*/0,
+                                                Operand->getMemSize()));
           if (isOutput) {
             std::string Constraint = "=";
             ++InputIdx;
             OutputDecls.push_back(OpDecl);
+            OutputDeclsOffsetOf.push_back(Operand->isOffsetOf());
             Constraint += Operand->getConstraint().str();
             OutputConstraints.push_back(Constraint);
             AsmStrRewrites.push_back(AsmRewrite(AOK_Output,
-                                                  Operand->getStartLoc(),
-                                                  Operand->getNameLen()));
+                                                Operand->getStartLoc(),
+                                                Operand->getNameLen()));
           } else {
             InputDecls.push_back(OpDecl);
+            InputDeclsOffsetOf.push_back(Operand->isOffsetOf());
             InputConstraints.push_back(Operand->getConstraint().str());
             AsmStrRewrites.push_back(AsmRewrite(AOK_Input,
-                                                  Operand->getStartLoc(),
-                                                  Operand->getNameLen()));
+                                                Operand->getStartLoc(),
+                                                Operand->getNameLen()));
           }
         }
       }
-      // Free any parsed operands.
-      for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
-        delete ParsedOperands[i];
-      ParsedOperands.clear();
     }
   }
 
@@ -3716,13 +3724,15 @@
     unsigned NumExprs = NumOutputs + NumInputs;
     OpDecls.resize(NumExprs);
     Constraints.resize(NumExprs);
+    // FIXME: Constraints are hard coded to 'm', but we need an 'r'
+    // constraint for offsetof.  This needs to be cleaned up!
     for (unsigned i = 0; i < NumOutputs; ++i) {
-      OpDecls[i] = OutputDecls[i];
-      Constraints[i] = OutputConstraints[i];
+      OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsOffsetOf[i]);
+      Constraints[i] = OutputDeclsOffsetOf[i] ? "=r" : OutputConstraints[i];
     }
     for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
-      OpDecls[j] = InputDecls[i];
-      Constraints[j] = InputConstraints[i];
+      OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsOffsetOf[i]);
+      Constraints[j] = InputDeclsOffsetOf[i] ? "r" : InputConstraints[i];
     }
   }
 
@@ -3753,7 +3763,11 @@
     switch (Kind) {
     default: break;
     case AOK_Imm:
-      OS << Twine("$$") + StringRef(Loc, (*I).Len);
+      OS << Twine("$$");
+      OS << (*I).Val;
+      break;
+    case AOK_ImmPrefix:
+      OS << Twine("$$");
       break;
     case AOK_Input:
       OS << '$';
@@ -3764,7 +3778,7 @@
       OS << OutputIdx++;
       break;
     case AOK_SizeDirective:
-      switch((*I).Size) {
+      switch((*I).Val) {
       default: break;
       case 8:  OS << "byte ptr "; break;
       case 16: OS << "word ptr "; break;
@@ -3774,6 +3788,13 @@
       case 128: OS << "xmmword ptr "; break;
       case 256: OS << "ymmword ptr "; break;
       }
+      break;
+    case AOK_Emit:
+      OS << ".byte";
+      break;
+    case AOK_DotOperator:
+      OS << (*I).Val;
+      break;
     }
 
     // Skip the original expression.

Modified: llvm/branches/R600/lib/Object/COFFObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Object/COFFObjectFile.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Object/COFFObjectFile.cpp (original)
+++ llvm/branches/R600/lib/Object/COFFObjectFile.cpp Tue Nov 13 09:21:47 2012
@@ -288,6 +288,11 @@
   return object_error::success;
 }
 
+error_code COFFObjectFile::getSymbolValue(DataRefImpl Symb,
+                                          uint64_t &Val) const {
+  report_fatal_error("getSymbolValue unimplemented in COFFObjectFile");
+}
+
 error_code COFFObjectFile::getSectionNext(DataRefImpl Sec,
                                           SectionRef &Result) const {
   const coff_section *sec = toSec(Sec);

Modified: llvm/branches/R600/lib/Object/MachOObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Object/MachOObjectFile.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Object/MachOObjectFile.cpp (original)
+++ llvm/branches/R600/lib/Object/MachOObjectFile.cpp Tue Nov 13 09:21:47 2012
@@ -363,6 +363,10 @@
   return object_error::success;
 }
 
+error_code MachOObjectFile::getSymbolValue(DataRefImpl Symb,
+                                           uint64_t &Val) const {
+  report_fatal_error("getSymbolValue unimplemented in MachOObjectFile");
+}
 
 symbol_iterator MachOObjectFile::begin_symbols() const {
   // DRI.d.a = segment number; DRI.d.b = symbol index.

Modified: llvm/branches/R600/lib/Support/APFloat.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/APFloat.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/APFloat.cpp (original)
+++ llvm/branches/R600/lib/Support/APFloat.cpp Tue Nov 13 09:21:47 2012
@@ -46,22 +46,27 @@
     /* Number of bits in the significand.  This includes the integer
        bit.  */
     unsigned int precision;
-
-    /* True if arithmetic is supported.  */
-    unsigned int arithmeticOK;
   };
 
-  const fltSemantics APFloat::IEEEhalf = { 15, -14, 11, true };
-  const fltSemantics APFloat::IEEEsingle = { 127, -126, 24, true };
-  const fltSemantics APFloat::IEEEdouble = { 1023, -1022, 53, true };
-  const fltSemantics APFloat::IEEEquad = { 16383, -16382, 113, true };
-  const fltSemantics APFloat::x87DoubleExtended = { 16383, -16382, 64, true };
-  const fltSemantics APFloat::Bogus = { 0, 0, 0, true };
-
-  // The PowerPC format consists of two doubles.  It does not map cleanly
-  // onto the usual format above.  For now only storage of constants of
-  // this type is supported, no arithmetic.
-  const fltSemantics APFloat::PPCDoubleDouble = { 1023, -1022, 106, false };
+  const fltSemantics APFloat::IEEEhalf = { 15, -14, 11 };
+  const fltSemantics APFloat::IEEEsingle = { 127, -126, 24 };
+  const fltSemantics APFloat::IEEEdouble = { 1023, -1022, 53 };
+  const fltSemantics APFloat::IEEEquad = { 16383, -16382, 113 };
+  const fltSemantics APFloat::x87DoubleExtended = { 16383, -16382, 64 };
+  const fltSemantics APFloat::Bogus = { 0, 0, 0 };
+
+  /* The PowerPC format consists of two doubles.  It does not map cleanly
+     onto the usual format above.  It is approximated using twice the
+     mantissa bits.  Note that for exponents near the double minimum,
+     we no longer can represent the full 106 mantissa bits, so those
+     will be treated as denormal numbers.
+
+     FIXME: While this approximation is equivalent to what GCC uses for
+     compile-time arithmetic on PPC double-double numbers, it is not able
+     to represent all possible values held by a PPC double-double number,
+     for example: (long double) 1.0 + (long double) 0x1p-106
+     Should this be replaced by a full emulation of PPC double-double?  */
+  const fltSemantics APFloat::PPCDoubleDouble = { 1023, -1022 + 53, 53 + 53 };
 
   /* A tight upper bound on number of parts required to hold the value
      pow(5, power) is
@@ -116,12 +121,6 @@
   return -1U;
 }
 
-static inline void
-assertArithmeticOK(const llvm::fltSemantics &semantics) {
-  assert(semantics.arithmeticOK &&
-         "Compile-time arithmetic does not support these semantics");
-}
-
 /* Return the value of a decimal exponent of the form
    [+-]ddddddd.
 
@@ -612,8 +611,6 @@
   sign = rhs.sign;
   category = rhs.category;
   exponent = rhs.exponent;
-  sign2 = rhs.sign2;
-  exponent2 = rhs.exponent2;
   if (category == fcNormal || category == fcNaN)
     copySignificand(rhs);
 }
@@ -707,16 +704,10 @@
       category != rhs.category ||
       sign != rhs.sign)
     return false;
-  if (semantics==(const llvm::fltSemantics*)&PPCDoubleDouble &&
-      sign2 != rhs.sign2)
-    return false;
   if (category==fcZero || category==fcInfinity)
     return true;
   else if (category==fcNormal && exponent!=rhs.exponent)
     return false;
-  else if (semantics==(const llvm::fltSemantics*)&PPCDoubleDouble &&
-           exponent2!=rhs.exponent2)
-    return false;
   else {
     int i= partCount();
     const integerPart* p=significandParts();
@@ -729,9 +720,7 @@
   }
 }
 
-APFloat::APFloat(const fltSemantics &ourSemantics, integerPart value)
-  : exponent2(0), sign2(0) {
-  assertArithmeticOK(ourSemantics);
+APFloat::APFloat(const fltSemantics &ourSemantics, integerPart value) {
   initialize(&ourSemantics);
   sign = 0;
   zeroSignificand();
@@ -740,24 +729,19 @@
   normalize(rmNearestTiesToEven, lfExactlyZero);
 }
 
-APFloat::APFloat(const fltSemantics &ourSemantics) : exponent2(0), sign2(0) {
-  assertArithmeticOK(ourSemantics);
+APFloat::APFloat(const fltSemantics &ourSemantics) {
   initialize(&ourSemantics);
   category = fcZero;
   sign = false;
 }
 
-APFloat::APFloat(const fltSemantics &ourSemantics, uninitializedTag tag)
-  : exponent2(0), sign2(0) {
-  assertArithmeticOK(ourSemantics);
+APFloat::APFloat(const fltSemantics &ourSemantics, uninitializedTag tag) {
   // Allocates storage if necessary but does not initialize it.
   initialize(&ourSemantics);
 }
 
 APFloat::APFloat(const fltSemantics &ourSemantics,
-                 fltCategory ourCategory, bool negative)
-  : exponent2(0), sign2(0) {
-  assertArithmeticOK(ourSemantics);
+                 fltCategory ourCategory, bool negative) {
   initialize(&ourSemantics);
   category = ourCategory;
   sign = negative;
@@ -767,14 +751,12 @@
     makeNaN();
 }
 
-APFloat::APFloat(const fltSemantics &ourSemantics, StringRef text)
-  : exponent2(0), sign2(0) {
-  assertArithmeticOK(ourSemantics);
+APFloat::APFloat(const fltSemantics &ourSemantics, StringRef text) {
   initialize(&ourSemantics);
   convertFromString(text, rmNearestTiesToEven);
 }
 
-APFloat::APFloat(const APFloat &rhs) : exponent2(0), sign2(0) {
+APFloat::APFloat(const APFloat &rhs) {
   initialize(rhs.semantics);
   assign(rhs);
 }
@@ -1561,8 +1543,6 @@
 {
   opStatus fs;
 
-  assertArithmeticOK(*semantics);
-
   fs = addOrSubtractSpecials(rhs, subtract);
 
   /* This return code means it was not a simple case.  */
@@ -1607,7 +1587,6 @@
 {
   opStatus fs;
 
-  assertArithmeticOK(*semantics);
   sign ^= rhs.sign;
   fs = multiplySpecials(rhs);
 
@@ -1627,7 +1606,6 @@
 {
   opStatus fs;
 
-  assertArithmeticOK(*semantics);
   sign ^= rhs.sign;
   fs = divideSpecials(rhs);
 
@@ -1649,7 +1627,6 @@
   APFloat V = *this;
   unsigned int origSign = sign;
 
-  assertArithmeticOK(*semantics);
   fs = V.divide(rhs, rmNearestTiesToEven);
   if (fs == opDivByZero)
     return fs;
@@ -1684,7 +1661,6 @@
 APFloat::mod(const APFloat &rhs, roundingMode rounding_mode)
 {
   opStatus fs;
-  assertArithmeticOK(*semantics);
   fs = modSpecials(rhs);
 
   if (category == fcNormal && rhs.category == fcNormal) {
@@ -1728,8 +1704,6 @@
 {
   opStatus fs;
 
-  assertArithmeticOK(*semantics);
-
   /* Post-multiplication sign, before addition.  */
   sign ^= multiplicand.sign;
 
@@ -1770,7 +1744,6 @@
 /* Rounding-mode corrrect round to integral value.  */
 APFloat::opStatus APFloat::roundToIntegral(roundingMode rounding_mode) {
   opStatus fs;
-  assertArithmeticOK(*semantics);
 
   // If the exponent is large enough, we know that this value is already
   // integral, and the arithmetic below would potentially cause it to saturate
@@ -1817,7 +1790,6 @@
 {
   cmpResult result;
 
-  assertArithmeticOK(*semantics);
   assert(semantics == rhs.semantics);
 
   switch (convolve(category, rhs.category)) {
@@ -1902,8 +1874,6 @@
   int shift;
   const fltSemantics &fromSemantics = *semantics;
 
-  assertArithmeticOK(fromSemantics);
-  assertArithmeticOK(toSemantics);
   lostFraction = lfExactlyZero;
   newPartCount = partCountForBits(toSemantics.precision + 1);
   oldPartCount = partCount();
@@ -1988,8 +1958,6 @@
   const integerPart *src;
   unsigned int dstPartsCount, truncatedBits;
 
-  assertArithmeticOK(*semantics);
-
   *isExact = false;
 
   /* Handle the three special cases first.  */
@@ -2151,7 +2119,6 @@
   integerPart *dst;
   lostFraction lost_fraction;
 
-  assertArithmeticOK(*semantics);
   category = fcNormal;
   omsb = APInt::tcMSB(src, srcCount) + 1;
   dst = significandParts();
@@ -2202,7 +2169,6 @@
 {
   opStatus status;
 
-  assertArithmeticOK(*semantics);
   if (isSigned &&
       APInt::tcExtractBit(src, srcCount * integerPartWidth - 1)) {
     integerPart *copy;
@@ -2336,7 +2302,7 @@
                                       roundingMode rounding_mode)
 {
   unsigned int parts, pow5PartCount;
-  fltSemantics calcSemantics = { 32767, -32767, 0, true };
+  fltSemantics calcSemantics = { 32767, -32767, 0 };
   integerPart pow5Parts[maxPowerOfFiveParts];
   bool isNearest;
 
@@ -2528,7 +2494,6 @@
 APFloat::opStatus
 APFloat::convertFromString(StringRef str, roundingMode rounding_mode)
 {
-  assertArithmeticOK(*semantics);
   assert(!str.empty() && "Invalid string length");
 
   /* Handle a leading minus sign.  */
@@ -2580,8 +2545,6 @@
 {
   char *p;
 
-  assertArithmeticOK(*semantics);
-
   p = dst;
   if (sign)
     *dst++ = '-';
@@ -2790,42 +2753,46 @@
   assert(semantics == (const llvm::fltSemantics*)&PPCDoubleDouble);
   assert(partCount()==2);
 
-  uint64_t myexponent, mysignificand, myexponent2, mysignificand2;
+  uint64_t words[2];
+  opStatus fs;
+  bool losesInfo;
 
-  if (category==fcNormal) {
-    myexponent = exponent + 1023; //bias
-    myexponent2 = exponent2 + 1023;
-    mysignificand = significandParts()[0];
-    mysignificand2 = significandParts()[1];
-    if (myexponent==1 && !(mysignificand & 0x10000000000000LL))
-      myexponent = 0;   // denormal
-    if (myexponent2==1 && !(mysignificand2 & 0x10000000000000LL))
-      myexponent2 = 0;   // denormal
-  } else if (category==fcZero) {
-    myexponent = 0;
-    mysignificand = 0;
-    myexponent2 = 0;
-    mysignificand2 = 0;
-  } else if (category==fcInfinity) {
-    myexponent = 0x7ff;
-    myexponent2 = 0;
-    mysignificand = 0;
-    mysignificand2 = 0;
+  // Convert number to double.  To avoid spurious underflows, we re-
+  // normalize against the "double" minExponent first, and only *then*
+  // truncate the mantissa.  The result of that second conversion
+  // may be inexact, but should never underflow.
+  APFloat extended(*this);
+  fltSemantics extendedSemantics = *semantics;
+  extendedSemantics.minExponent = IEEEdouble.minExponent;
+  fs = extended.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo);
+  assert(fs == opOK && !losesInfo);
+  (void)fs;
+
+  APFloat u(extended);
+  fs = u.convert(IEEEdouble, rmNearestTiesToEven, &losesInfo);
+  assert(fs == opOK || fs == opInexact);
+  (void)fs;
+  words[0] = *u.convertDoubleAPFloatToAPInt().getRawData();
+
+  // If conversion was exact or resulted in a special case, we're done;
+  // just set the second double to zero.  Otherwise, re-convert back to
+  // the extended format and compute the difference.  This now should
+  // convert exactly to double.
+  if (u.category == fcNormal && losesInfo) {
+    fs = u.convert(extendedSemantics, rmNearestTiesToEven, &losesInfo);
+    assert(fs == opOK && !losesInfo);
+    (void)fs;
+
+    APFloat v(extended);
+    v.subtract(u, rmNearestTiesToEven);
+    fs = v.convert(IEEEdouble, rmNearestTiesToEven, &losesInfo);
+    assert(fs == opOK && !losesInfo);
+    (void)fs;
+    words[1] = *v.convertDoubleAPFloatToAPInt().getRawData();
   } else {
-    assert(category == fcNaN && "Unknown category");
-    myexponent = 0x7ff;
-    mysignificand = significandParts()[0];
-    myexponent2 = exponent2;
-    mysignificand2 = significandParts()[1];
+    words[1] = 0;
   }
 
-  uint64_t words[2];
-  words[0] =  ((uint64_t)(sign & 1) << 63) |
-              ((myexponent & 0x7ff) <<  52) |
-              (mysignificand & 0xfffffffffffffLL);
-  words[1] =  ((uint64_t)(sign2 & 1) << 63) |
-              ((myexponent2 & 0x7ff) <<  52) |
-              (mysignificand2 & 0xfffffffffffffLL);
   return APInt(128, words);
 }
 
@@ -3045,47 +3012,23 @@
   assert(api.getBitWidth()==128);
   uint64_t i1 = api.getRawData()[0];
   uint64_t i2 = api.getRawData()[1];
-  uint64_t myexponent = (i1 >> 52) & 0x7ff;
-  uint64_t mysignificand = i1 & 0xfffffffffffffLL;
-  uint64_t myexponent2 = (i2 >> 52) & 0x7ff;
-  uint64_t mysignificand2 = i2 & 0xfffffffffffffLL;
+  opStatus fs;
+  bool losesInfo;
 
-  initialize(&APFloat::PPCDoubleDouble);
-  assert(partCount()==2);
+  // Get the first double and convert to our format.
+  initFromDoubleAPInt(APInt(64, i1));
+  fs = convert(PPCDoubleDouble, rmNearestTiesToEven, &losesInfo);
+  assert(fs == opOK && !losesInfo);
+  (void)fs;
 
-  sign = static_cast<unsigned int>(i1>>63);
-  sign2 = static_cast<unsigned int>(i2>>63);
-  if (myexponent==0 && mysignificand==0) {
-    // exponent, significand meaningless
-    // exponent2 and significand2 are required to be 0; we don't check
-    category = fcZero;
-  } else if (myexponent==0x7ff && mysignificand==0) {
-    // exponent, significand meaningless
-    // exponent2 and significand2 are required to be 0; we don't check
-    category = fcInfinity;
-  } else if (myexponent==0x7ff && mysignificand!=0) {
-    // exponent meaningless.  So is the whole second word, but keep it
-    // for determinism.
-    category = fcNaN;
-    exponent2 = myexponent2;
-    significandParts()[0] = mysignificand;
-    significandParts()[1] = mysignificand2;
-  } else {
-    category = fcNormal;
-    // Note there is no category2; the second word is treated as if it is
-    // fcNormal, although it might be something else considered by itself.
-    exponent = myexponent - 1023;
-    exponent2 = myexponent2 - 1023;
-    significandParts()[0] = mysignificand;
-    significandParts()[1] = mysignificand2;
-    if (myexponent==0)          // denormal
-      exponent = -1022;
-    else
-      significandParts()[0] |= 0x10000000000000LL;  // integer bit
-    if (myexponent2==0)
-      exponent2 = -1022;
-    else
-      significandParts()[1] |= 0x10000000000000LL;  // integer bit
+  // Unless we have a special case, add in second double.
+  if (category == fcNormal) {
+    APFloat v(APInt(64, i2));
+    fs = v.convert(PPCDoubleDouble, rmNearestTiesToEven, &losesInfo);
+    assert(fs == opOK && !losesInfo);
+    (void)fs;
+
+    add(v, rmNearestTiesToEven);
   }
 }
 
@@ -3311,15 +3254,15 @@
   return Val;
 }
 
-APFloat::APFloat(const APInt& api, bool isIEEE) : exponent2(0), sign2(0) {
+APFloat::APFloat(const APInt& api, bool isIEEE) {
   initFromAPInt(api, isIEEE);
 }
 
-APFloat::APFloat(float f) : exponent2(0), sign2(0) {
+APFloat::APFloat(float f) {
   initFromAPInt(APInt::floatToBits(f));
 }
 
-APFloat::APFloat(double d) : exponent2(0), sign2(0) {
+APFloat::APFloat(double d) {
   initFromAPInt(APInt::doubleToBits(d));
 }
 

Modified: llvm/branches/R600/lib/Support/Atomic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/Atomic.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/Atomic.cpp (original)
+++ llvm/branches/R600/lib/Support/Atomic.cpp Tue Nov 13 09:21:47 2012
@@ -21,11 +21,15 @@
 #undef MemoryFence
 #endif
 
+#if defined(__GNUC__) || (defined(__IBMCPP__) && __IBMCPP__ >= 1210)
+#define GNU_ATOMICS
+#endif
+
 void sys::MemoryFence() {
 #if LLVM_HAS_ATOMICS == 0
   return;
 #else
-#  if defined(__GNUC__)
+#  if defined(GNU_ATOMICS)
   __sync_synchronize();
 #  elif defined(_MSC_VER)
   MemoryBarrier();
@@ -43,7 +47,7 @@
   if (result == old_value)
     *ptr = new_value;
   return result;
-#elif defined(__GNUC__)
+#elif defined(GNU_ATOMICS)
   return __sync_val_compare_and_swap(ptr, old_value, new_value);
 #elif defined(_MSC_VER)
   return InterlockedCompareExchange(ptr, new_value, old_value);
@@ -56,7 +60,7 @@
 #if LLVM_HAS_ATOMICS == 0
   ++(*ptr);
   return *ptr;
-#elif defined(__GNUC__)
+#elif defined(GNU_ATOMICS)
   return __sync_add_and_fetch(ptr, 1);
 #elif defined(_MSC_VER)
   return InterlockedIncrement(ptr);
@@ -69,7 +73,7 @@
 #if LLVM_HAS_ATOMICS == 0
   --(*ptr);
   return *ptr;
-#elif defined(__GNUC__)
+#elif defined(GNU_ATOMICS)
   return __sync_sub_and_fetch(ptr, 1);
 #elif defined(_MSC_VER)
   return InterlockedDecrement(ptr);
@@ -82,7 +86,7 @@
 #if LLVM_HAS_ATOMICS == 0
   *ptr += val;
   return *ptr;
-#elif defined(__GNUC__)
+#elif defined(GNU_ATOMICS)
   return __sync_add_and_fetch(ptr, val);
 #elif defined(_MSC_VER)
   return InterlockedExchangeAdd(ptr, val) + val;

Modified: llvm/branches/R600/lib/Support/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Support/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -1,9 +1,3 @@
-## FIXME: This only requires RTTI because tblgen uses it.  Fix that.
-set(LLVM_REQUIRES_RTTI 1)
-if( MINGW )
-  set(LLVM_REQUIRES_EH 1)
-endif()
-
 add_llvm_library(LLVMSupport
   APFloat.cpp
   APInt.cpp

Modified: llvm/branches/R600/lib/Support/Errno.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/Errno.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/Errno.cpp (original)
+++ llvm/branches/R600/lib/Support/Errno.cpp Tue Nov 13 09:21:47 2012
@@ -53,8 +53,10 @@
     str = buffer;
 # endif
 #elif HAVE_DECL_STRERROR_S // "Windows Secure API"
-    if (errnum)
+    if (errnum) {
       strerror_s(buffer, MaxErrStrLen - 1, errnum);
+      str = buffer;
+    }
 #elif defined(HAVE_STRERROR)
   // Copy the thread un-safe result of strerror into
   // the buffer as fast as possible to minimize impact

Modified: llvm/branches/R600/lib/Support/Host.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/Host.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/Host.cpp (original)
+++ llvm/branches/R600/lib/Support/Host.cpp Tue Nov 13 09:21:47 2012
@@ -503,6 +503,7 @@
           .Case("0xb76", "arm1176jz-s")
           .Case("0xc08", "cortex-a8")
           .Case("0xc09", "cortex-a9")
+          .Case("0xc0f", "cortex-a15")
           .Case("0xc20", "cortex-m0")
           .Case("0xc23", "cortex-m3")
           .Case("0xc24", "cortex-m4")

Modified: llvm/branches/R600/lib/Support/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/Makefile?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/Makefile (original)
+++ llvm/branches/R600/lib/Support/Makefile Tue Nov 13 09:21:47 2012
@@ -11,9 +11,6 @@
 LIBRARYNAME = LLVMSupport
 BUILD_ARCHIVE = 1
 
-## FIXME: This only requires RTTI because tblgen uses it.  Fix that.
-REQUIRES_RTTI = 1
-
 EXTRA_DIST = Unix Win32 README.txt
 
 include $(LEVEL)/Makefile.common

Modified: llvm/branches/R600/lib/Support/MemoryBuffer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/MemoryBuffer.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/MemoryBuffer.cpp (original)
+++ llvm/branches/R600/lib/Support/MemoryBuffer.cpp Tue Nov 13 09:21:47 2012
@@ -33,6 +33,9 @@
 #include <unistd.h>
 #else
 #include <io.h>
+#ifndef S_ISFIFO
+#define S_ISFIFO(x) (0)
+#endif
 #endif
 #include <fcntl.h>
 using namespace llvm;
@@ -201,6 +204,27 @@
 };
 }
 
+static error_code getMemoryBufferForStream(int FD, 
+                                           StringRef BufferName,
+                                           OwningPtr<MemoryBuffer> &result) {
+  const ssize_t ChunkSize = 4096*4;
+  SmallString<ChunkSize> Buffer;
+  ssize_t ReadBytes;
+  // Read into Buffer until we hit EOF.
+  do {
+    Buffer.reserve(Buffer.size() + ChunkSize);
+    ReadBytes = read(FD, Buffer.end(), ChunkSize);
+    if (ReadBytes == -1) {
+      if (errno == EINTR) continue;
+      return error_code(errno, posix_category());
+    }
+    Buffer.set_size(Buffer.size() + ReadBytes);
+  } while (ReadBytes != 0);
+
+  result.reset(MemoryBuffer::getMemBufferCopy(Buffer, BufferName));
+  return error_code::success();
+}
+
 error_code MemoryBuffer::getFile(StringRef Filename,
                                  OwningPtr<MemoryBuffer> &result,
                                  int64_t FileSize,
@@ -297,6 +321,13 @@
       if (fstat(FD, &FileInfo) == -1) {
         return error_code(errno, posix_category());
       }
+
+      // If this is a named pipe, we can't trust the size. Create the memory
+      // buffer by copying off the stream.
+      if (S_ISFIFO(FileInfo.st_mode)) {
+        return getMemoryBufferForStream(FD, Filename, result);
+      }
+
       FileSize = FileInfo.st_size;
     }
     MapSize = FileSize;
@@ -370,20 +401,5 @@
   // fallback if it fails.
   sys::Program::ChangeStdinToBinary();
 
-  const ssize_t ChunkSize = 4096*4;
-  SmallString<ChunkSize> Buffer;
-  ssize_t ReadBytes;
-  // Read into Buffer until we hit EOF.
-  do {
-    Buffer.reserve(Buffer.size() + ChunkSize);
-    ReadBytes = read(0, Buffer.end(), ChunkSize);
-    if (ReadBytes == -1) {
-      if (errno == EINTR) continue;
-      return error_code(errno, posix_category());
-    }
-    Buffer.set_size(Buffer.size() + ReadBytes);
-  } while (ReadBytes != 0);
-
-  result.reset(getMemBufferCopy(Buffer, "<stdin>"));
-  return error_code::success();
+  return getMemoryBufferForStream(0, "<stdin>", result);
 }

Modified: llvm/branches/R600/lib/Support/Triple.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Support/Triple.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Support/Triple.cpp (original)
+++ llvm/branches/R600/lib/Support/Triple.cpp Tue Nov 13 09:21:47 2012
@@ -180,38 +180,6 @@
     .Default(UnknownArch);
 }
 
-Triple::ArchType Triple::getArchTypeForDarwinArchName(StringRef Str) {
-  // See arch(3) and llvm-gcc's driver-driver.c. We don't implement support for
-  // archs which Darwin doesn't use.
-
-  // The matching this routine does is fairly pointless, since it is neither the
-  // complete architecture list, nor a reasonable subset. The problem is that
-  // historically the driver driver accepts this and also ties its -march=
-  // handling to the architecture name, so we need to be careful before removing
-  // support for it.
-
-  // This code must be kept in sync with Clang's Darwin specific argument
-  // translation.
-
-  return StringSwitch<ArchType>(Str)
-    .Cases("ppc", "ppc601", "ppc603", "ppc604", "ppc604e", Triple::ppc)
-    .Cases("ppc750", "ppc7400", "ppc7450", "ppc970", Triple::ppc)
-    .Case("ppc64", Triple::ppc64)
-    .Cases("i386", "i486", "i486SX", "i586", "i686", Triple::x86)
-    .Cases("pentium", "pentpro", "pentIIm3", "pentIIm5", "pentium4",
-           Triple::x86)
-    .Case("x86_64", Triple::x86_64)
-    // This is derived from the driver driver.
-    .Cases("arm", "armv4t", "armv5", "armv6", Triple::arm)
-    .Cases("armv7", "armv7f", "armv7k", "armv7s", "xscale", Triple::arm)
-    .Case("r600", Triple::r600)
-    .Case("nvptx", Triple::nvptx)
-    .Case("nvptx64", Triple::nvptx64)
-    .Case("amdil", Triple::amdil)
-    .Case("spir", Triple::spir)
-    .Default(Triple::UnknownArch);
-}
-
 // Returns architecture name that is understood by the target assembler.
 const char *Triple::getArchNameForAssembler() {
   if (!isOSDarwin() && getVendor() != Triple::Apple)

Modified: llvm/branches/R600/lib/TableGen/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/TableGen/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/TableGen/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/TableGen/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -1,5 +1,3 @@
-set(LLVM_REQUIRES_EH 1)
-
 add_llvm_library(LLVMTableGen
   Error.cpp
   Main.cpp

Modified: llvm/branches/R600/lib/TableGen/Error.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/TableGen/Error.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/TableGen/Error.cpp (original)
+++ llvm/branches/R600/lib/TableGen/Error.cpp Tue Nov 13 09:21:47 2012
@@ -16,6 +16,8 @@
 #include "llvm/ADT/Twine.h"
 #include "llvm/Support/raw_ostream.h"
 
+#include <cstdlib>
+
 namespace llvm {
 
 SourceMgr SrcMgr;
@@ -43,10 +45,6 @@
   errs() << "warning:" << Msg << "\n";
 }
 
-void PrintWarning(const TGError &Warning) {
-  PrintWarning(Warning.getLoc(), Warning.getMessage());
-}
-
 void PrintError(ArrayRef<SMLoc> ErrorLoc, const Twine &Msg) {
   PrintMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
 }
@@ -59,8 +57,14 @@
   errs() << "error:" << Msg << "\n";
 }
 
-void PrintError(const TGError &Error) {
-  PrintError(Error.getLoc(), Error.getMessage());
+void PrintFatalError(const std::string &Msg) {
+  PrintError(Twine(Msg));
+  std::exit(1);
+}
+
+void PrintFatalError(ArrayRef<SMLoc> ErrorLoc, const std::string &Msg) {
+  PrintError(ErrorLoc, Msg);
+  std::exit(1);
 }
 
 } // end namespace llvm

Modified: llvm/branches/R600/lib/TableGen/Main.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/TableGen/Main.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/TableGen/Main.cpp (original)
+++ llvm/branches/R600/lib/TableGen/Main.cpp Tue Nov 13 09:21:47 2012
@@ -80,56 +80,46 @@
 int TableGenMain(char *argv0, TableGenMainFn *MainFn) {
   RecordKeeper Records;
 
-  try {
-    // Parse the input file.
-    OwningPtr<MemoryBuffer> File;
-    if (error_code ec =
-          MemoryBuffer::getFileOrSTDIN(InputFilename.c_str(), File)) {
-      errs() << "Could not open input file '" << InputFilename << "': "
-             << ec.message() <<"\n";
-      return 1;
-    }
-    MemoryBuffer *F = File.take();
-
-    // Tell SrcMgr about this buffer, which is what TGParser will pick up.
-    SrcMgr.AddNewSourceBuffer(F, SMLoc());
-
-    // Record the location of the include directory so that the lexer can find
-    // it later.
-    SrcMgr.setIncludeDirs(IncludeDirs);
-
-    TGParser Parser(SrcMgr, Records);
-
-    if (Parser.ParseFile())
-      return 1;
-
-    std::string Error;
-    tool_output_file Out(OutputFilename.c_str(), Error);
-    if (!Error.empty()) {
-      errs() << argv0 << ": error opening " << OutputFilename
-        << ":" << Error << "\n";
-      return 1;
-    }
-    if (!DependFilename.empty())
-      if (int Ret = createDependencyFile(Parser, argv0))
-        return Ret;
-
-    if (MainFn(Out.os(), Records))
-      return 1;
-
-    // Declare success.
-    Out.keep();
-    return 0;
-
-  } catch (const TGError &Error) {
-    PrintError(Error);
-  } catch (const std::string &Error) {
-    PrintError(Error);
-  } catch (const char *Error) {
-    PrintError(Error);
-  } catch (...) {
-    errs() << argv0 << ": Unknown unexpected exception occurred.\n";
+  // Parse the input file.
+  OwningPtr<MemoryBuffer> File;
+  if (error_code ec =
+        MemoryBuffer::getFileOrSTDIN(InputFilename.c_str(), File)) {
+    errs() << "Could not open input file '" << InputFilename << "': "
+           << ec.message() <<"\n";
+    return 1;
   }
+  MemoryBuffer *F = File.take();
+
+  // Tell SrcMgr about this buffer, which is what TGParser will pick up.
+  SrcMgr.AddNewSourceBuffer(F, SMLoc());
+
+  // Record the location of the include directory so that the lexer can find
+  // it later.
+  SrcMgr.setIncludeDirs(IncludeDirs);
+
+  TGParser Parser(SrcMgr, Records);
+
+  if (Parser.ParseFile())
+    return 1;
+
+  std::string Error;
+  tool_output_file Out(OutputFilename.c_str(), Error);
+  if (!Error.empty()) {
+    errs() << argv0 << ": error opening " << OutputFilename
+      << ":" << Error << "\n";
+    return 1;
+  }
+  if (!DependFilename.empty()) {
+    if (int Ret = createDependencyFile(Parser, argv0))
+      return Ret;
+  }
+
+  if (MainFn(Out.os(), Records))
+    return 1;
+
+  // Declare success.
+  Out.keep();
+  return 0;
 
   return 1;
 }

Modified: llvm/branches/R600/lib/TableGen/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/TableGen/Makefile?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/TableGen/Makefile (original)
+++ llvm/branches/R600/lib/TableGen/Makefile Tue Nov 13 09:21:47 2012
@@ -11,6 +11,4 @@
 LIBRARYNAME = LLVMTableGen
 BUILD_ARCHIVE = 1
 
-REQUIRES_EH = 1
-
 include $(LEVEL)/Makefile.common

Modified: llvm/branches/R600/lib/TableGen/Record.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/TableGen/Record.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/TableGen/Record.cpp (original)
+++ llvm/branches/R600/lib/TableGen/Record.cpp Tue Nov 13 09:21:47 2012
@@ -616,7 +616,8 @@
 Record *ListInit::getElementAsRecord(unsigned i) const {
   assert(i < Values.size() && "List element index out of range!");
   DefInit *DI = dyn_cast<DefInit>(Values[i]);
-  if (DI == 0) throw "Expected record in list!";
+  if (DI == 0)
+    PrintFatalError("Expected record in list!");
   return DI->getDef();
 }
 
@@ -725,7 +726,7 @@
         if (CurRec) {
           if (const RecordVal *RV = CurRec->getValue(Name)) {
             if (RV->getType() != getType())
-              throw "type mismatch in cast";
+              PrintFatalError("type mismatch in cast");
             return VarInit::get(Name, RV->getType());
           }
 
@@ -737,7 +738,7 @@
             assert(RV && "Template arg doesn't exist??");
 
             if (RV->getType() != getType())
-              throw "type mismatch in cast";
+              PrintFatalError("type mismatch in cast");
 
             return VarInit::get(TemplateArgName, RV->getType());
           }
@@ -751,7 +752,7 @@
             assert(RV && "Template arg doesn't exist??");
 
             if (RV->getType() != getType())
-              throw "type mismatch in cast";
+              PrintFatalError("type mismatch in cast");
 
             return VarInit::get(MCName, RV->getType());
           }
@@ -760,7 +761,8 @@
         if (Record *D = (CurRec->getRecords()).getDef(Name))
           return DefInit::get(D);
 
-        throw TGError(CurRec->getLoc(), "Undefined reference:'" + Name + "'\n");
+        PrintFatalError(CurRec->getLoc(),
+                        "Undefined reference:'" + Name + "'\n");
       }
     }
     break;
@@ -860,7 +862,7 @@
       DefInit *LOp = dyn_cast<DefInit>(LHSs->getOperator());
       DefInit *ROp = dyn_cast<DefInit>(RHSs->getOperator());
       if (LOp == 0 || ROp == 0 || LOp->getDef() != ROp->getDef())
-        throw "Concated Dag operators do not match!";
+        PrintFatalError("Concated Dag operators do not match!");
       std::vector<Init*> Args;
       std::vector<std::string> ArgNames;
       for (unsigned i = 0, e = LHSs->getNumArgs(); i != e; ++i) {
@@ -1027,14 +1029,13 @@
   OpInit *RHSo = dyn_cast<OpInit>(RHS);
 
   if (!RHSo) {
-    throw TGError(CurRec->getLoc(), "!foreach requires an operator\n");
+    PrintFatalError(CurRec->getLoc(), "!foreach requires an operator\n");
   }
 
   TypedInit *LHSt = dyn_cast<TypedInit>(LHS);
 
-  if (!LHSt) {
-    throw TGError(CurRec->getLoc(), "!foreach requires typed variable\n");
-  }
+  if (!LHSt)
+    PrintFatalError(CurRec->getLoc(), "!foreach requires typed variable\n");
 
   if ((MHSd && isa<DagRecTy>(Type)) || (MHSl && isa<ListRecTy>(Type))) {
     if (MHSd) {
@@ -1632,7 +1633,7 @@
   assert(TypedName && "Record name is not typed!");
   RecTy *Type = TypedName->getType();
   if (!isa<StringRecTy>(Type))
-    throw TGError(getLoc(), "Record name is not a string!");
+    PrintFatalError(getLoc(), "Record name is not a string!");
 }
 
 DefInit *Record::getDefInit() {
@@ -1683,7 +1684,7 @@
       continue;
     if (Init *V = Values[i].getValue())
       if (Values[i].setValue(V->resolveReferences(*this, RV)))
-        throw TGError(getLoc(), "Invalid value is found when setting '"
+        PrintFatalError(getLoc(), "Invalid value is found when setting '"
                       + Values[i].getNameInitAsString()
                       + "' after resolving references"
                       + (RV ? " against '" + RV->getNameInitAsString()
@@ -1738,68 +1739,68 @@
 }
 
 /// getValueInit - Return the initializer for a value with the specified name,
-/// or throw an exception if the field does not exist.
+/// or abort if the field does not exist.
 ///
 Init *Record::getValueInit(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-      FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
   return R->getValue();
 }
 
 
 /// getValueAsString - This method looks up the specified field and returns its
-/// value as a string, throwing an exception if the field does not exist or if
+/// value as a string, aborts if the field does not exist or if
 /// the value is not a string.
 ///
 std::string Record::getValueAsString(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-          FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (StringInit *SI = dyn_cast<StringInit>(R->getValue()))
     return SI->getValue();
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a string initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a string initializer!");
 }
 
 /// getValueAsBitsInit - This method looks up the specified field and returns
-/// its value as a BitsInit, throwing an exception if the field does not exist
-/// or if the value is not the right type.
+/// its value as a BitsInit, aborts if the field does not exist or if
+/// the value is not the right type.
 ///
 BitsInit *Record::getValueAsBitsInit(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-          FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (BitsInit *BI = dyn_cast<BitsInit>(R->getValue()))
     return BI;
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a BitsInit initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a BitsInit initializer!");
 }
 
 /// getValueAsListInit - This method looks up the specified field and returns
-/// its value as a ListInit, throwing an exception if the field does not exist
-/// or if the value is not the right type.
+/// its value as a ListInit, aborting if the field does not exist or if
+/// the value is not the right type.
 ///
 ListInit *Record::getValueAsListInit(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-          FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (ListInit *LI = dyn_cast<ListInit>(R->getValue()))
     return LI;
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a list initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a list initializer!");
 }
 
 /// getValueAsListOfDefs - This method looks up the specified field and returns
-/// its value as a vector of records, throwing an exception if the field does
-/// not exist or if the value is not the right type.
+/// its value as a vector of records, aborting if the field does not exist
+/// or if the value is not the right type.
 ///
 std::vector<Record*>
 Record::getValueAsListOfDefs(StringRef FieldName) const {
@@ -1809,32 +1810,32 @@
     if (DefInit *DI = dyn_cast<DefInit>(List->getElement(i))) {
       Defs.push_back(DI->getDef());
     } else {
-      throw "Record `" + getName() + "', field `" + FieldName.str() +
-            "' list is not entirely DefInit!";
+      PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+        FieldName.str() + "' list is not entirely DefInit!");
     }
   }
   return Defs;
 }
 
 /// getValueAsInt - This method looks up the specified field and returns its
-/// value as an int64_t, throwing an exception if the field does not exist or if
-/// the value is not the right type.
+/// value as an int64_t, aborting if the field does not exist or if the value
+/// is not the right type.
 ///
 int64_t Record::getValueAsInt(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-          FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (IntInit *II = dyn_cast<IntInit>(R->getValue()))
     return II->getValue();
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have an int initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have an int initializer!");
 }
 
 /// getValueAsListOfInts - This method looks up the specified field and returns
-/// its value as a vector of integers, throwing an exception if the field does
-/// not exist or if the value is not the right type.
+/// its value as a vector of integers, aborting if the field does not exist or
+/// if the value is not the right type.
 ///
 std::vector<int64_t>
 Record::getValueAsListOfInts(StringRef FieldName) const {
@@ -1844,16 +1845,16 @@
     if (IntInit *II = dyn_cast<IntInit>(List->getElement(i))) {
       Ints.push_back(II->getValue());
     } else {
-      throw "Record `" + getName() + "', field `" + FieldName.str() +
-            "' does not have a list of ints initializer!";
+      PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+        FieldName.str() + "' does not have a list of ints initializer!");
     }
   }
   return Ints;
 }
 
 /// getValueAsListOfStrings - This method looks up the specified field and
-/// returns its value as a vector of strings, throwing an exception if the
-/// field does not exist or if the value is not the right type.
+/// returns its value as a vector of strings, aborting if the field does not
+/// exist or if the value is not the right type.
 ///
 std::vector<std::string>
 Record::getValueAsListOfStrings(StringRef FieldName) const {
@@ -1863,50 +1864,50 @@
     if (StringInit *II = dyn_cast<StringInit>(List->getElement(i))) {
       Strings.push_back(II->getValue());
     } else {
-      throw "Record `" + getName() + "', field `" + FieldName.str() +
-            "' does not have a list of strings initializer!";
+      PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+        FieldName.str() + "' does not have a list of strings initializer!");
     }
   }
   return Strings;
 }
 
 /// getValueAsDef - This method looks up the specified field and returns its
-/// value as a Record, throwing an exception if the field does not exist or if
-/// the value is not the right type.
+/// value as a Record, aborting if the field does not exist or if the value
+/// is not the right type.
 ///
 Record *Record::getValueAsDef(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-      FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (DefInit *DI = dyn_cast<DefInit>(R->getValue()))
     return DI->getDef();
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a def initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a def initializer!");
 }
 
 /// getValueAsBit - This method looks up the specified field and returns its
-/// value as a bit, throwing an exception if the field does not exist or if
-/// the value is not the right type.
+/// value as a bit, aborting if the field does not exist or if the value is
+/// not the right type.
 ///
 bool Record::getValueAsBit(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-      FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (BitInit *BI = dyn_cast<BitInit>(R->getValue()))
     return BI->getValue();
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a bit initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a bit initializer!");
 }
 
 bool Record::getValueAsBitOrUnset(StringRef FieldName, bool &Unset) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-      FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (R->getValue() == UnsetInit::get()) {
     Unset = true;
@@ -1915,24 +1916,24 @@
   Unset = false;
   if (BitInit *BI = dyn_cast<BitInit>(R->getValue()))
     return BI->getValue();
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a bit initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a bit initializer!");
 }
 
 /// getValueAsDag - This method looks up the specified field and returns its
-/// value as an Dag, throwing an exception if the field does not exist or if
-/// the value is not the right type.
+/// value as an Dag, aborting if the field does not exist or if the value is
+/// not the right type.
 ///
 DagInit *Record::getValueAsDag(StringRef FieldName) const {
   const RecordVal *R = getValue(FieldName);
   if (R == 0 || R->getValue() == 0)
-    throw "Record `" + getName() + "' does not have a field named `" +
-      FieldName.str() + "'!\n";
+    PrintFatalError(getLoc(), "Record `" + getName() +
+      "' does not have a field named `" + FieldName.str() + "'!\n");
 
   if (DagInit *DI = dyn_cast<DagInit>(R->getValue()))
     return DI;
-  throw "Record `" + getName() + "', field `" + FieldName.str() +
-        "' does not have a dag initializer!";
+  PrintFatalError(getLoc(), "Record `" + getName() + "', field `" +
+    FieldName.str() + "' does not have a dag initializer!");
 }
 
 
@@ -1975,7 +1976,7 @@
 RecordKeeper::getAllDerivedDefinitions(const std::string &ClassName) const {
   Record *Class = getClass(ClassName);
   if (!Class)
-    throw "ERROR: Couldn't find the `" + ClassName + "' class!\n";
+    PrintFatalError("ERROR: Couldn't find the `" + ClassName + "' class!\n");
 
   std::vector<Record*> Defs;
   for (std::map<std::string, Record*>::const_iterator I = getDefs().begin(),

Modified: llvm/branches/R600/lib/Target/ARM/ARMBaseInstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMBaseInstrInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMBaseInstrInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMBaseInstrInfo.cpp Tue Nov 13 09:21:47 2012
@@ -702,6 +702,8 @@
     Opc = ARM::VMOVD, BeginIdx = ARM::dsub_0, SubRegs = 3;
   else if (ARM::DQuadRegClass.contains(DestReg, SrcReg))
     Opc = ARM::VMOVD, BeginIdx = ARM::dsub_0, SubRegs = 4;
+  else if (ARM::GPRPairRegClass.contains(DestReg, SrcReg))
+    Opc = ARM::MOVr, BeginIdx = ARM::gsub_0, SubRegs = 2;
 
   else if (ARM::DPairSpcRegClass.contains(DestReg, SrcReg))
     Opc = ARM::VMOVD, BeginIdx = ARM::dsub_0, SubRegs = 2, Spacing = 2;
@@ -791,6 +793,13 @@
         AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::VSTRD))
                    .addReg(SrcReg, getKillRegState(isKill))
                    .addFrameIndex(FI).addImm(0).addMemOperand(MMO));
+      } else if (ARM::GPRPairRegClass.hasSubClassEq(RC)) {
+        MachineInstrBuilder MIB =
+          AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::STMIA))
+                       .addFrameIndex(FI))
+                       .addMemOperand(MMO);
+          MIB = AddDReg(MIB, SrcReg, ARM::gsub_0, getKillRegState(isKill), TRI);
+                AddDReg(MIB, SrcReg, ARM::gsub_1, 0, TRI);
       } else
         llvm_unreachable("Unknown reg class!");
       break;
@@ -938,6 +947,7 @@
   DebugLoc DL;
   if (I != MBB.end()) DL = I->getDebugLoc();
   MachineFunction &MF = *MBB.getParent();
+  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
   MachineFrameInfo &MFI = *MF.getFrameInfo();
   unsigned Align = MFI.getObjectAlignment(FI);
   MachineMemOperand *MMO =
@@ -963,6 +973,15 @@
     if (ARM::DPRRegClass.hasSubClassEq(RC)) {
       AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::VLDRD), DestReg)
                    .addFrameIndex(FI).addImm(0).addMemOperand(MMO));
+    } else if (ARM::GPRPairRegClass.hasSubClassEq(RC)) {
+      unsigned LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA : ARM::LDMIA;
+      MachineInstrBuilder MIB =
+        AddDefaultPred(BuildMI(MBB, I, DL, get(LdmOpc))
+                    .addFrameIndex(FI).addImm(0).addMemOperand(MMO));
+      MIB = AddDReg(MIB, DestReg, ARM::gsub_0, RegState::DefineNoRead, TRI);
+      MIB = AddDReg(MIB, DestReg, ARM::gsub_1, RegState::DefineNoRead, TRI);
+      if (TargetRegisterInfo::isPhysicalRegister(DestReg))
+        MIB.addReg(DestReg, RegState::ImplicitDefine);
     } else
       llvm_unreachable("Unknown reg class!");
     break;

Modified: llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.cpp Tue Nov 13 09:21:47 2012
@@ -84,6 +84,11 @@
     ? CSR_iOS_RegMask : CSR_AAPCS_RegMask;
 }
 
+const uint32_t*
+ARMBaseRegisterInfo::getNoPreservedMask() const {
+  return CSR_NoRegs_RegMask;
+}
+
 BitVector ARMBaseRegisterInfo::
 getReservedRegs(const MachineFunction &MF) const {
   const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
@@ -106,148 +111,12 @@
     for (unsigned i = 0; i != 16; ++i)
       Reserved.set(ARM::D16 + i);
   }
-  return Reserved;
-}
-
-bool ARMBaseRegisterInfo::isReservedReg(const MachineFunction &MF,
-                                        unsigned Reg) const {
-  const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
-
-  switch (Reg) {
-  default: break;
-  case ARM::SP:
-  case ARM::PC:
-    return true;
-  case ARM::R6:
-    if (hasBasePointer(MF))
-      return true;
-    break;
-  case ARM::R7:
-  case ARM::R11:
-    if (FramePtr == Reg && TFI->hasFP(MF))
-      return true;
-    break;
-  case ARM::R9:
-    return STI.isR9Reserved();
-  }
-
-  return false;
-}
+  const TargetRegisterClass *RC  = &ARM::GPRPairRegClass;
+  for(TargetRegisterClass::iterator I = RC->begin(), E = RC->end(); I!=E; ++I)
+    for (MCSubRegIterator SI(*I, this); SI.isValid(); ++SI)
+      if (Reserved.test(*SI)) Reserved.set(*I);
 
-bool
-ARMBaseRegisterInfo::canCombineSubRegIndices(const TargetRegisterClass *RC,
-                                          SmallVectorImpl<unsigned> &SubIndices,
-                                          unsigned &NewSubIdx) const {
-
-  unsigned Size = RC->getSize() * 8;
-  if (Size < 6)
-    return 0;
-
-  NewSubIdx = 0;  // Whole register.
-  unsigned NumRegs = SubIndices.size();
-  if (NumRegs == 8) {
-    // 8 D registers -> 1 QQQQ register.
-    return (Size == 512 &&
-            SubIndices[0] == ARM::dsub_0 &&
-            SubIndices[1] == ARM::dsub_1 &&
-            SubIndices[2] == ARM::dsub_2 &&
-            SubIndices[3] == ARM::dsub_3 &&
-            SubIndices[4] == ARM::dsub_4 &&
-            SubIndices[5] == ARM::dsub_5 &&
-            SubIndices[6] == ARM::dsub_6 &&
-            SubIndices[7] == ARM::dsub_7);
-  } else if (NumRegs == 4) {
-    if (SubIndices[0] == ARM::qsub_0) {
-      // 4 Q registers -> 1 QQQQ register.
-      return (Size == 512 &&
-              SubIndices[1] == ARM::qsub_1 &&
-              SubIndices[2] == ARM::qsub_2 &&
-              SubIndices[3] == ARM::qsub_3);
-    } else if (SubIndices[0] == ARM::dsub_0) {
-      // 4 D registers -> 1 QQ register.
-      if (Size >= 256 &&
-          SubIndices[1] == ARM::dsub_1 &&
-          SubIndices[2] == ARM::dsub_2 &&
-          SubIndices[3] == ARM::dsub_3) {
-        if (Size == 512)
-          NewSubIdx = ARM::qqsub_0;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::dsub_4) {
-      // 4 D registers -> 1 QQ register (2nd).
-      if (Size == 512 &&
-          SubIndices[1] == ARM::dsub_5 &&
-          SubIndices[2] == ARM::dsub_6 &&
-          SubIndices[3] == ARM::dsub_7) {
-        NewSubIdx = ARM::qqsub_1;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::ssub_0) {
-      // 4 S registers -> 1 Q register.
-      if (Size >= 128 &&
-          SubIndices[1] == ARM::ssub_1 &&
-          SubIndices[2] == ARM::ssub_2 &&
-          SubIndices[3] == ARM::ssub_3) {
-        if (Size >= 256)
-          NewSubIdx = ARM::qsub_0;
-        return true;
-      }
-    }
-  } else if (NumRegs == 2) {
-    if (SubIndices[0] == ARM::qsub_0) {
-      // 2 Q registers -> 1 QQ register.
-      if (Size >= 256 && SubIndices[1] == ARM::qsub_1) {
-        if (Size == 512)
-          NewSubIdx = ARM::qqsub_0;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::qsub_2) {
-      // 2 Q registers -> 1 QQ register (2nd).
-      if (Size == 512 && SubIndices[1] == ARM::qsub_3) {
-        NewSubIdx = ARM::qqsub_1;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::dsub_0) {
-      // 2 D registers -> 1 Q register.
-      if (Size >= 128 && SubIndices[1] == ARM::dsub_1) {
-        if (Size >= 256)
-          NewSubIdx = ARM::qsub_0;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::dsub_2) {
-      // 2 D registers -> 1 Q register (2nd).
-      if (Size >= 256 && SubIndices[1] == ARM::dsub_3) {
-        NewSubIdx = ARM::qsub_1;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::dsub_4) {
-      // 2 D registers -> 1 Q register (3rd).
-      if (Size == 512 && SubIndices[1] == ARM::dsub_5) {
-        NewSubIdx = ARM::qsub_2;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::dsub_6) {
-      // 2 D registers -> 1 Q register (3rd).
-      if (Size == 512 && SubIndices[1] == ARM::dsub_7) {
-        NewSubIdx = ARM::qsub_3;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::ssub_0) {
-      // 2 S registers -> 1 D register.
-      if (SubIndices[1] == ARM::ssub_1) {
-        if (Size >= 128)
-          NewSubIdx = ARM::dsub_0;
-        return true;
-      }
-    } else if (SubIndices[0] == ARM::ssub_2) {
-      // 2 S registers -> 1 D register (2nd).
-      if (Size >= 128 && SubIndices[1] == ARM::ssub_3) {
-        NewSubIdx = ARM::dsub_1;
-        return true;
-      }
-    }
-  }
-  return false;
+  return Reserved;
 }
 
 const TargetRegisterClass*
@@ -263,6 +132,7 @@
     case ARM::QPRRegClassID:
     case ARM::QQPRRegClassID:
     case ARM::QQQQPRRegClassID:
+    case ARM::GPRPairRegClassID:
       return Super;
     }
     Super = *I++;
@@ -596,6 +466,7 @@
 
 unsigned ARMBaseRegisterInfo::getRegisterPairEven(unsigned Reg,
                                               const MachineFunction &MF) const {
+  const MachineRegisterInfo &MRI = MF.getRegInfo();
   switch (Reg) {
   default: break;
   // Return 0 if either register of the pair is a special register.
@@ -604,10 +475,10 @@
   case ARM::R3: return ARM::R2;
   case ARM::R5: return ARM::R4;
   case ARM::R7:
-    return (isReservedReg(MF, ARM::R7) || isReservedReg(MF, ARM::R6))
+    return (MRI.isReserved(ARM::R7) || MRI.isReserved(ARM::R6))
       ? 0 : ARM::R6;
-  case ARM::R9: return isReservedReg(MF, ARM::R9)  ? 0 :ARM::R8;
-  case ARM::R11: return isReservedReg(MF, ARM::R11) ? 0 : ARM::R10;
+  case ARM::R9: return MRI.isReserved(ARM::R9)  ? 0 :ARM::R8;
+  case ARM::R11: return MRI.isReserved(ARM::R11) ? 0 : ARM::R10;
 
   case ARM::S1: return ARM::S0;
   case ARM::S3: return ARM::S2;
@@ -649,6 +520,7 @@
 
 unsigned ARMBaseRegisterInfo::getRegisterPairOdd(unsigned Reg,
                                              const MachineFunction &MF) const {
+  const MachineRegisterInfo &MRI = MF.getRegInfo();
   switch (Reg) {
   default: break;
   // Return 0 if either register of the pair is a special register.
@@ -657,10 +529,10 @@
   case ARM::R2: return ARM::R3;
   case ARM::R4: return ARM::R5;
   case ARM::R6:
-    return (isReservedReg(MF, ARM::R7) || isReservedReg(MF, ARM::R6))
+    return (MRI.isReserved(ARM::R7) || MRI.isReserved(ARM::R6))
       ? 0 : ARM::R7;
-  case ARM::R8: return isReservedReg(MF, ARM::R9)  ? 0 :ARM::R9;
-  case ARM::R10: return isReservedReg(MF, ARM::R11) ? 0 : ARM::R11;
+  case ARM::R8: return MRI.isReserved(ARM::R9)  ? 0 :ARM::R9;
+  case ARM::R10: return MRI.isReserved(ARM::R11) ? 0 : ARM::R11;
 
   case ARM::S0: return ARM::S1;
   case ARM::S2: return ARM::S3;

Modified: llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.h (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMBaseRegisterInfo.h Tue Nov 13 09:21:47 2012
@@ -96,19 +96,10 @@
   /// Code Generation virtual methods...
   const uint16_t *getCalleeSavedRegs(const MachineFunction *MF = 0) const;
   const uint32_t *getCallPreservedMask(CallingConv::ID) const;
+  const uint32_t *getNoPreservedMask() const;
 
   BitVector getReservedRegs(const MachineFunction &MF) const;
 
-  /// canCombineSubRegIndices - Given a register class and a list of
-  /// subregister indices, return true if it's possible to combine the
-  /// subregister indices into one that corresponds to a larger
-  /// subregister. Return the new subregister index by reference. Note the
-  /// new index may be zero if the given subregisters can be combined to
-  /// form the whole register.
-  virtual bool canCombineSubRegIndices(const TargetRegisterClass *RC,
-                                       SmallVectorImpl<unsigned> &SubIndices,
-                                       unsigned &NewSubIdx) const;
-
   const TargetRegisterClass*
   getPointerRegClass(const MachineFunction &MF, unsigned Kind = 0) const;
   const TargetRegisterClass*
@@ -170,8 +161,6 @@
                                  unsigned MIFlags = MachineInstr::NoFlags)const;
 
   /// Code Generation virtual methods...
-  virtual bool isReservedReg(const MachineFunction &MF, unsigned Reg) const;
-
   virtual bool requiresRegisterScavenging(const MachineFunction &MF) const;
 
   virtual bool trackLivenessAfterRegAlloc(const MachineFunction &MF) const;

Modified: llvm/branches/R600/lib/Target/ARM/ARMCallingConv.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMCallingConv.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMCallingConv.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMCallingConv.td Tue Nov 13 09:21:47 2012
@@ -190,6 +190,8 @@
 // Callee-saved register lists.
 //===----------------------------------------------------------------------===//
 
+def CSR_NoRegs : CalleeSavedRegs<(add)>;
+
 def CSR_AAPCS : CalleeSavedRegs<(add LR, R11, R10, R9, R8, R7, R6, R5, R4,
                                      (sequence "D%u", 15, 8))>;
 

Removed: llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.cpp (removed)
@@ -1,78 +0,0 @@
-//===-- ARMELFWriterInfo.cpp - ELF Writer Info for the ARM backend --------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the ARM backend.
-//
-//===----------------------------------------------------------------------===//
-
-#include "ARMELFWriterInfo.h"
-#include "ARMRelocations.h"
-#include "llvm/Function.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/DataLayout.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Support/ELF.h"
-
-using namespace llvm;
-
-//===----------------------------------------------------------------------===//
-//  Implementation of the ARMELFWriterInfo class
-//===----------------------------------------------------------------------===//
-
-ARMELFWriterInfo::ARMELFWriterInfo(TargetMachine &TM)
-  : TargetELFWriterInfo(TM.getDataLayout()->getPointerSizeInBits(0) == 64,
-                        TM.getDataLayout()->isLittleEndian()) {
-}
-
-ARMELFWriterInfo::~ARMELFWriterInfo() {}
-
-unsigned ARMELFWriterInfo::getRelocationType(unsigned MachineRelTy) const {
-  switch (MachineRelTy) {
-  case ARM::reloc_arm_absolute:
-  case ARM::reloc_arm_relative:
-  case ARM::reloc_arm_cp_entry:
-  case ARM::reloc_arm_vfp_cp_entry:
-  case ARM::reloc_arm_machine_cp_entry:
-  case ARM::reloc_arm_jt_base:
-  case ARM::reloc_arm_pic_jt:
-    llvm_unreachable("unsupported ARM relocation type");
-
-  case ARM::reloc_arm_branch: return ELF::R_ARM_CALL;
-  case ARM::reloc_arm_movt:   return ELF::R_ARM_MOVT_ABS;
-  case ARM::reloc_arm_movw:   return ELF::R_ARM_MOVW_ABS_NC;
-  default:
-    llvm_unreachable("unknown ARM relocation type");
-  }
-}
-
-long int ARMELFWriterInfo::getDefaultAddendForRelTy(unsigned RelTy,
-                                                    long int Modifier) const {
-  llvm_unreachable("ARMELFWriterInfo::getDefaultAddendForRelTy() not "
-                   "implemented");
-}
-
-unsigned ARMELFWriterInfo::getRelocationTySize(unsigned RelTy) const {
-  llvm_unreachable("ARMELFWriterInfo::getRelocationTySize() not implemented");
-}
-
-bool ARMELFWriterInfo::isPCRelativeRel(unsigned RelTy) const {
-  llvm_unreachable("ARMELFWriterInfo::isPCRelativeRel() not implemented");
-}
-
-unsigned ARMELFWriterInfo::getAbsoluteLabelMachineRelTy() const {
-  llvm_unreachable("ARMELFWriterInfo::getAbsoluteLabelMachineRelTy() not "
-                   "implemented");
-}
-
-long int ARMELFWriterInfo::computeRelocation(unsigned SymOffset,
-                                             unsigned RelOffset,
-                                             unsigned RelTy) const {
-  llvm_unreachable("ARMELFWriterInfo::getAbsoluteLabelMachineRelTy() not "
-                   "implemented");
-}

Removed: llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.h (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMELFWriterInfo.h (removed)
@@ -1,59 +0,0 @@
-//===-- ARMELFWriterInfo.h - ELF Writer Info for ARM ------------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the ARM backend.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef ARM_ELF_WRITER_INFO_H
-#define ARM_ELF_WRITER_INFO_H
-
-#include "llvm/Target/TargetELFWriterInfo.h"
-
-namespace llvm {
-  class TargetMachine;
-
-  class ARMELFWriterInfo : public TargetELFWriterInfo {
-  public:
-    ARMELFWriterInfo(TargetMachine &TM);
-    virtual ~ARMELFWriterInfo();
-
-    /// getRelocationType - Returns the target specific ELF Relocation type.
-    /// 'MachineRelTy' contains the object code independent relocation type
-    virtual unsigned getRelocationType(unsigned MachineRelTy) const;
-
-    /// hasRelocationAddend - True if the target uses an addend in the
-    /// ELF relocation entry.
-    virtual bool hasRelocationAddend() const { return false; }
-
-    /// getDefaultAddendForRelTy - Gets the default addend value for a
-    /// relocation entry based on the target ELF relocation type.
-    virtual long int getDefaultAddendForRelTy(unsigned RelTy,
-                                              long int Modifier = 0) const;
-
-    /// getRelTySize - Returns the size of relocatable field in bits
-    virtual unsigned getRelocationTySize(unsigned RelTy) const;
-
-    /// isPCRelativeRel - True if the relocation type is pc relative
-    virtual bool isPCRelativeRel(unsigned RelTy) const;
-
-    /// getJumpTableRelocationTy - Returns the machine relocation type used
-    /// to reference a jumptable.
-    virtual unsigned getAbsoluteLabelMachineRelTy() const;
-
-    /// computeRelocation - Some relocatable fields could be relocated
-    /// directly, avoiding the relocation symbol emission, compute the
-    /// final relocation value for this symbol.
-    virtual long int computeRelocation(unsigned SymOffset, unsigned RelOffset,
-                                       unsigned RelTy) const;
-  };
-
-} // end llvm namespace
-
-#endif // ARM_ELF_WRITER_INFO_H

Modified: llvm/branches/R600/lib/Target/ARM/ARMExpandPseudoInsts.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMExpandPseudoInsts.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMExpandPseudoInsts.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMExpandPseudoInsts.cpp Tue Nov 13 09:21:47 2012
@@ -777,9 +777,7 @@
       MI.eraseFromParent();
       return true;
     }
-    case ARM::Int_eh_sjlj_dispatchsetup:
-    case ARM::Int_eh_sjlj_dispatchsetup_nofp:
-    case ARM::tInt_eh_sjlj_dispatchsetup: {
+    case ARM::Int_eh_sjlj_dispatchsetup: {
       MachineFunction &MF = *MI.getParent()->getParent();
       const ARMBaseInstrInfo *AII =
         static_cast<const ARMBaseInstrInfo*>(TII);
@@ -1208,57 +1206,6 @@
       ExpandLaneOp(MBBI);
       return true;
 
-    case ARM::VSETLNi8Q:
-    case ARM::VSETLNi16Q: {
-      // Expand VSETLNs acting on a Q register to equivalent VSETLNs acting
-      // on the respective D register.
-
-      unsigned QReg  = MI.getOperand(1).getReg();
-      unsigned QLane = MI.getOperand(3).getImm();
-
-      unsigned NewOpcode, DLane, DSubReg;
-      switch (Opcode) {
-      default: llvm_unreachable("Invalid opcode!");
-      case ARM::VSETLNi8Q:
-        // 4 possible 8-bit lanes per DPR:
-        NewOpcode = ARM::VSETLNi8;
-        DLane = QLane % 8;
-        DSubReg  = (QLane / 8) ? ARM::dsub_1 : ARM::dsub_0;
-        break;
-      case ARM::VSETLNi16Q:
-        // 4 possible 16-bit lanes per DPR.
-        NewOpcode = ARM::VSETLNi16;
-        DLane = QLane % 4;
-        DSubReg  = (QLane / 4) ? ARM::dsub_1 : ARM::dsub_0;
-        break;
-      }
-
-      MachineInstrBuilder MIB =
-        BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(NewOpcode));
-
-      unsigned DReg = TRI->getSubReg(QReg, DSubReg);
-
-      MIB.addReg(DReg, RegState::Define); // Output DPR
-      MIB.addReg(DReg);                   // Input DPR
-      MIB.addOperand(MI.getOperand(2));   // Input GPR
-      MIB.addImm(DLane);                  // Lane
-
-      // Add the predicate operands.
-      MIB.addOperand(MI.getOperand(4));
-      MIB.addOperand(MI.getOperand(5));
-
-      if (MI.getOperand(1).isKill()) // Add an implicit kill for the Q register.
-        MIB->addRegisterKilled(QReg, TRI, true);
-      // And an implicit def of the output register (which should always be the
-      // same as the input register).
-      MIB->addRegisterDefined(QReg, TRI);
-
-      TransferImpOps(MI, MIB, MIB);
-
-      MI.eraseFromParent();
-      return true;
-    }
-
     case ARM::VTBL3Pseudo: ExpandVTBL(MBBI, ARM::VTBL3, false); return true;
     case ARM::VTBL4Pseudo: ExpandVTBL(MBBI, ARM::VTBL4, false); return true;
     case ARM::VTBX3Pseudo: ExpandVTBL(MBBI, ARM::VTBX3, true); return true;

Modified: llvm/branches/R600/lib/Target/ARM/ARMFastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMFastISel.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMFastISel.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMFastISel.cpp Tue Nov 13 09:21:47 2012
@@ -619,7 +619,10 @@
 
   Reloc::Model RelocM = TM.getRelocationModel();
   bool IsIndirect = Subtarget->GVIsIndirectSymbol(GV, RelocM);
-  unsigned DestReg = createResultReg(TLI.getRegClassFor(VT));
+  const TargetRegisterClass *RC = isThumb2 ?
+    (const TargetRegisterClass*)&ARM::rGPRRegClass :
+    (const TargetRegisterClass*)&ARM::GPRRegClass;
+  unsigned DestReg = createResultReg(RC);
 
   // Use movw+movt when possible, it avoids constant pool entries.
   // Darwin targets don't support movt with Reloc::Static, see
@@ -1388,6 +1391,11 @@
   unsigned Opc = isThumb2 ? ARM::tBRIND : ARM::BX;
   AddOptionalDefs(BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc))
                   .addReg(AddrReg));
+
+  const IndirectBrInst *IB = cast<IndirectBrInst>(I);
+  for (unsigned i = 0, e = IB->getNumSuccessors(); i != e; ++i)
+    FuncInfo.MBB->addSuccessor(FuncInfo.MBBMap[IB->getSuccessor(i)]);
+
   return true;
 }
 

Modified: llvm/branches/R600/lib/Target/ARM/ARMFrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMFrameLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMFrameLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMFrameLowering.cpp Tue Nov 13 09:21:47 2012
@@ -153,7 +153,8 @@
   int FramePtrSpillFI = 0;
   int D8SpillFI = 0;
 
-  // All calls are tail calls in GHC calling conv, and functions have no prologue/epilogue.
+  // All calls are tail calls in GHC calling conv, and functions have no
+  // prologue/epilogue.
   if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
     return;
 
@@ -360,7 +361,8 @@
   int NumBytes = (int)MFI->getStackSize();
   unsigned FramePtr = RegInfo->getFrameRegister(MF);
 
-  // All calls are tail calls in GHC calling conv, and functions have no prologue/epilogue.
+  // All calls are tail calls in GHC calling conv, and functions have no
+  // prologue/epilogue.
   if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
     return;
 
@@ -1209,6 +1211,7 @@
     *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
   ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
   MachineFrameInfo *MFI = MF.getFrameInfo();
+  MachineRegisterInfo &MRI = MF.getRegInfo();
   unsigned FramePtr = RegInfo->getFrameRegister(MF);
 
   // Spill R4 if Thumb2 function requires stack realignment - it will be used as
@@ -1218,12 +1221,12 @@
   // FIXME: It will be better just to find spare register here.
   if (AFI->isThumb2Function() &&
       (MFI->hasVarSizedObjects() || RegInfo->needsStackRealignment(MF)))
-    MF.getRegInfo().setPhysRegUsed(ARM::R4);
+    MRI.setPhysRegUsed(ARM::R4);
 
   if (AFI->isThumb1OnlyFunction()) {
     // Spill LR if Thumb1 function uses variable length argument lists.
     if (AFI->getVarArgsRegSaveSize() > 0)
-      MF.getRegInfo().setPhysRegUsed(ARM::LR);
+      MRI.setPhysRegUsed(ARM::LR);
 
     // Spill R4 if Thumb1 epilogue has to restore SP from FP. We don't know
     // for sure what the stack size will be, but for this, an estimate is good
@@ -1233,7 +1236,7 @@
     // FIXME: It will be better just to find spare register here.
     unsigned StackSize = estimateStackSize(MF);
     if (MFI->hasVarSizedObjects() || StackSize > 508)
-      MF.getRegInfo().setPhysRegUsed(ARM::R4);
+      MRI.setPhysRegUsed(ARM::R4);
   }
 
   // See if we can spill vector registers to aligned stack.
@@ -1241,7 +1244,7 @@
 
   // Spill the BasePtr if it's used.
   if (RegInfo->hasBasePointer(MF))
-    MF.getRegInfo().setPhysRegUsed(RegInfo->getBaseRegister());
+    MRI.setPhysRegUsed(RegInfo->getBaseRegister());
 
   // Don't spill FP if the frame can be eliminated. This is determined
   // by scanning the callee-save registers to see if any is used.
@@ -1249,7 +1252,7 @@
   for (unsigned i = 0; CSRegs[i]; ++i) {
     unsigned Reg = CSRegs[i];
     bool Spilled = false;
-    if (MF.getRegInfo().isPhysRegUsed(Reg)) {
+    if (MRI.isPhysRegUsed(Reg)) {
       Spilled = true;
       CanEliminateFrame = false;
     }
@@ -1338,7 +1341,7 @@
     // If LR is not spilled, but at least one of R4, R5, R6, and R7 is spilled.
     // Spill LR as well so we can fold BX_RET to the registers restore (LDM).
     if (!LRSpilled && CS1Spilled) {
-      MF.getRegInfo().setPhysRegUsed(ARM::LR);
+      MRI.setPhysRegUsed(ARM::LR);
       NumGPRSpills++;
       UnspilledCS1GPRs.erase(std::find(UnspilledCS1GPRs.begin(),
                                     UnspilledCS1GPRs.end(), (unsigned)ARM::LR));
@@ -1347,7 +1350,7 @@
     }
 
     if (hasFP(MF)) {
-      MF.getRegInfo().setPhysRegUsed(FramePtr);
+      MRI.setPhysRegUsed(FramePtr);
       NumGPRSpills++;
     }
 
@@ -1362,16 +1365,16 @@
           // Don't spill high register if the function is thumb1
           if (!AFI->isThumb1OnlyFunction() ||
               isARMLowRegister(Reg) || Reg == ARM::LR) {
-            MF.getRegInfo().setPhysRegUsed(Reg);
-            if (!RegInfo->isReservedReg(MF, Reg))
+            MRI.setPhysRegUsed(Reg);
+            if (!MRI.isReserved(Reg))
               ExtraCSSpill = true;
             break;
           }
         }
       } else if (!UnspilledCS2GPRs.empty() && !AFI->isThumb1OnlyFunction()) {
         unsigned Reg = UnspilledCS2GPRs.front();
-        MF.getRegInfo().setPhysRegUsed(Reg);
-        if (!RegInfo->isReservedReg(MF, Reg))
+        MRI.setPhysRegUsed(Reg);
+        if (!MRI.isReserved(Reg))
           ExtraCSSpill = true;
       }
     }
@@ -1389,7 +1392,7 @@
       while (NumExtras && !UnspilledCS1GPRs.empty()) {
         unsigned Reg = UnspilledCS1GPRs.back();
         UnspilledCS1GPRs.pop_back();
-        if (!RegInfo->isReservedReg(MF, Reg) &&
+        if (!MRI.isReserved(Reg) &&
             (!AFI->isThumb1OnlyFunction() || isARMLowRegister(Reg) ||
              Reg == ARM::LR)) {
           Extras.push_back(Reg);
@@ -1401,7 +1404,7 @@
         while (NumExtras && !UnspilledCS2GPRs.empty()) {
           unsigned Reg = UnspilledCS2GPRs.back();
           UnspilledCS2GPRs.pop_back();
-          if (!RegInfo->isReservedReg(MF, Reg)) {
+          if (!MRI.isReserved(Reg)) {
             Extras.push_back(Reg);
             NumExtras--;
           }
@@ -1409,7 +1412,7 @@
       }
       if (Extras.size() && NumExtras == 0) {
         for (unsigned i = 0, e = Extras.size(); i != e; ++i) {
-          MF.getRegInfo().setPhysRegUsed(Extras[i]);
+          MRI.setPhysRegUsed(Extras[i]);
         }
       } else if (!AFI->isThumb1OnlyFunction()) {
         // note: Thumb1 functions spill to R12, not the stack.  Reserve a slot
@@ -1423,7 +1426,7 @@
   }
 
   if (ForceLRSpill) {
-    MF.getRegInfo().setPhysRegUsed(ARM::LR);
+    MRI.setPhysRegUsed(ARM::LR);
     AFI->setLRIsSpilledForFarJump(true);
   }
 }

Modified: llvm/branches/R600/lib/Target/ARM/ARMISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMISelLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMISelLowering.cpp Tue Nov 13 09:21:47 2012
@@ -1594,11 +1594,15 @@
 
   // FIXME: handle tail calls differently.
   unsigned CallOpc;
+  bool HasMinSizeAttr = MF.getFunction()->getFnAttributes().
+    hasAttribute(Attributes::MinSize);
   if (Subtarget->isThumb()) {
     if ((!isDirect || isARMFunc) && !Subtarget->hasV5TOps())
       CallOpc = ARMISD::CALL_NOLINK;
     else if (doesNotRet && isDirect && !isARMFunc &&
-             Subtarget->hasRAS() && !Subtarget->isThumb1Only())
+             Subtarget->hasRAS() && !Subtarget->isThumb1Only() &&
+             // Emit regular call when code size is the priority
+             !HasMinSizeAttr)
       // "mov lr, pc; b _foo" to avoid confusing the RSP
       CallOpc = ARMISD::CALL_NOLINK;
     else
@@ -1606,7 +1610,9 @@
   } else {
     if (!isDirect && !Subtarget->hasV5TOps()) {
       CallOpc = ARMISD::CALL_NOLINK;
-    } else if (doesNotRet && isDirect && Subtarget->hasRAS())
+    } else if (doesNotRet && isDirect && Subtarget->hasRAS() &&
+               // Emit regular call when code size is the priority
+               !HasMinSizeAttr)
       // "mov lr, pc; b _foo" to avoid confusing the RSP
       CallOpc = ARMISD::CALL_NOLINK;
     else
@@ -3923,6 +3929,36 @@
   return SDValue();
 }
 
+// check if an VEXT instruction can handle the shuffle mask when the
+// vector sources of the shuffle are the same.
+static bool isSingletonVEXTMask(ArrayRef<int> M, EVT VT, unsigned &Imm) {
+  unsigned NumElts = VT.getVectorNumElements();
+
+  // Assume that the first shuffle index is not UNDEF.  Fail if it is.
+  if (M[0] < 0)
+    return false;
+
+  Imm = M[0];
+
+  // If this is a VEXT shuffle, the immediate value is the index of the first
+  // element.  The other shuffle indices must be the successive elements after
+  // the first one.
+  unsigned ExpectedElt = Imm;
+  for (unsigned i = 1; i < NumElts; ++i) {
+    // Increment the expected index.  If it wraps around, just follow it
+    // back to index zero and keep going.
+    ++ExpectedElt;
+    if (ExpectedElt == NumElts)
+      ExpectedElt = 0;
+
+    if (M[i] < 0) continue; // ignore UNDEF indices
+    if (ExpectedElt != static_cast<unsigned>(M[i]))
+      return false;
+  }
+
+  return true;
+}
+
 
 static bool isVEXTMask(ArrayRef<int> M, EVT VT,
                        bool &ReverseVEXT, unsigned &Imm) {
@@ -4682,6 +4718,12 @@
     if (isVREVMask(ShuffleMask, VT, 16))
       return DAG.getNode(ARMISD::VREV16, dl, VT, V1);
 
+    if (V2->getOpcode() == ISD::UNDEF &&
+        isSingletonVEXTMask(ShuffleMask, VT, Imm)) {
+      return DAG.getNode(ARMISD::VEXT, dl, VT, V1, V1,
+                         DAG.getConstant(Imm, MVT::i32));
+    }
+
     // Check for Neon shuffles that modify both input vectors in place.
     // If both results are used, i.e., if there are two shuffles with the same
     // source operands and with masks corresponding to both results of one of
@@ -6035,12 +6077,15 @@
                              MachineMemOperand::MOLoad |
                              MachineMemOperand::MOVolatile, 4, 4);
 
-  if (AFI->isThumb1OnlyFunction())
-    BuildMI(DispatchBB, dl, TII->get(ARM::tInt_eh_sjlj_dispatchsetup));
-  else if (!Subtarget->hasVFP2())
-    BuildMI(DispatchBB, dl, TII->get(ARM::Int_eh_sjlj_dispatchsetup_nofp));
-  else
-    BuildMI(DispatchBB, dl, TII->get(ARM::Int_eh_sjlj_dispatchsetup));
+  MachineInstrBuilder MIB;
+  MIB = BuildMI(DispatchBB, dl, TII->get(ARM::Int_eh_sjlj_dispatchsetup));
+
+  const ARMBaseInstrInfo *AII = static_cast<const ARMBaseInstrInfo*>(TII);
+  const ARMBaseRegisterInfo &RI = AII->getRegisterInfo();
+
+  // Add a register mask with no preserved registers.  This results in all
+  // registers being marked as clobbered.
+  MIB.addRegMask(RI.getNoPreservedMask());
 
   unsigned NumLPads = LPadList.size();
   if (Subtarget->isThumb2()) {
@@ -6259,8 +6304,6 @@
   }
 
   // N.B. the order the invoke BBs are processed in doesn't matter here.
-  const ARMBaseInstrInfo *AII = static_cast<const ARMBaseInstrInfo*>(TII);
-  const ARMBaseRegisterInfo &RI = AII->getRegisterInfo();
   const uint16_t *SavedRegs = RI.getCalleeSavedRegs(MF);
   SmallVector<MachineBasicBlock*, 64> MBBLPads;
   for (SmallPtrSet<MachineBasicBlock*, 64>::iterator

Modified: llvm/branches/R600/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMInstrInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMInstrInfo.td Tue Nov 13 09:21:47 2012
@@ -283,15 +283,13 @@
 //  ARM specific transformation functions and pattern fragments.
 //
 
-// imm_neg_XFORM - Return a imm value packed into the format described for
-// imm_neg defs below.
+// imm_neg_XFORM - Return the negation of an i32 immediate value.
 def imm_neg_XFORM : SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant(-(int)N->getZExtValue(), MVT::i32);
 }]>;
 
-// so_imm_not_XFORM - Return a so_imm value packed into the format described for
-// so_imm_not def below.
-def so_imm_not_XFORM : SDNodeXForm<imm, [{
+// imm_not_XFORM - Return the complement of a i32 immediate value.
+def imm_not_XFORM : SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant(~(int)N->getZExtValue(), MVT::i32);
 }]>;
 
@@ -314,7 +312,7 @@
 def so_imm_not_asmoperand : AsmOperandClass { let Name = "ARMSOImmNot"; }
 def so_imm_not : Operand<i32>, PatLeaf<(imm), [{
     return ARM_AM::getSOImmVal(~(uint32_t)N->getZExtValue()) != -1;
-  }], so_imm_not_XFORM> {
+  }], imm_not_XFORM> {
   let ParserMatchClass = so_imm_not_asmoperand;
 }
 
@@ -3120,6 +3118,8 @@
 // for part of the negation.
 def : ARMPat<(ARMadde GPR:$src, so_imm_not:$imm, CPSR),
              (SBCri   GPR:$src, so_imm_not:$imm)>;
+def : ARMPat<(ARMadde GPR:$src, imm0_65535_neg:$imm, CPSR),
+             (SBCrr   GPR:$src, (MOVi16 (imm_not_XFORM imm:$imm)))>;
 
 // Note: These are implemented in C++ code, because they have to generate
 // ADD/SUBrs instructions, which use a complex pattern that a xform function
@@ -4743,21 +4743,13 @@
                                 Requires<[IsARM, IsIOS]>;
 }
 
-// eh.sjlj.dispatchsetup pseudo-instructions.
-// These pseudos are used for both ARM and Thumb2. Any differences are
-// handled when the pseudo is expanded (which happens before any passes
-// that need the instruction size).
-let Defs =
-  [ R0,  R1,  R2,  R3,  R4,  R5,  R6,  R7,  R8,  R9,  R10, R11, R12, LR, CPSR,
-    Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, Q10, Q11, Q12, Q13, Q14, Q15 ],
-  isBarrier = 1 in
+// eh.sjlj.dispatchsetup pseudo-instruction.
+// This pseudo is used for both ARM and Thumb. Any differences are handled when
+// the pseudo is expanded (which happens before any passes that need the
+// instruction size).
+let isBarrier = 1 in
 def Int_eh_sjlj_dispatchsetup : PseudoInst<(outs), (ins), NoItinerary, []>;
 
-let Defs =
-  [ R0,  R1,  R2,  R3,  R4,  R5,  R6,  R7,  R8,  R9,  R10, R11, R12, LR, CPSR ],
-  isBarrier = 1 in
-def Int_eh_sjlj_dispatchsetup_nofp : PseudoInst<(outs), (ins), NoItinerary, []>;
-
 
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns

Modified: llvm/branches/R600/lib/Target/ARM/ARMInstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMInstrNEON.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMInstrNEON.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMInstrNEON.td Tue Nov 13 09:21:47 2012
@@ -5140,23 +5140,25 @@
                                            GPR:$R, imm:$lane))]> {
   let Inst{21} = lane{0};
 }
-
-def VSETLNi8Q : PseudoNeonI<(outs QPR:$V),
-                             (ins QPR:$src1, GPR:$R, VectorIndex8:$lane),
-                             IIC_VMOVISL, "",
-                             [(set QPR:$V, (vector_insert (v16i8 QPR:$src1),
-                                           GPR:$R, imm:$lane))]>;
-def VSETLNi16Q : PseudoNeonI<(outs QPR:$V),
-                             (ins QPR:$src1, GPR:$R, VectorIndex16:$lane),
-                             IIC_VMOVISL, "",
-                             [(set QPR:$V, (vector_insert (v8i16 QPR:$src1),
-                                           GPR:$R, imm:$lane))]>;
 }
-
+def : Pat<(vector_insert (v16i8 QPR:$src1), GPR:$src2, imm:$lane),
+          (v16i8 (INSERT_SUBREG QPR:$src1,
+                  (v8i8 (VSETLNi8 (v8i8 (EXTRACT_SUBREG QPR:$src1,
+                                   (DSubReg_i8_reg imm:$lane))),
+                            GPR:$src2, (SubReg_i8_lane imm:$lane))),
+                  (DSubReg_i8_reg imm:$lane)))>;
+def : Pat<(vector_insert (v8i16 QPR:$src1), GPR:$src2, imm:$lane),
+          (v8i16 (INSERT_SUBREG QPR:$src1,
+                  (v4i16 (VSETLNi16 (v4i16 (EXTRACT_SUBREG QPR:$src1,
+                                     (DSubReg_i16_reg imm:$lane))),
+                             GPR:$src2, (SubReg_i16_lane imm:$lane))),
+                  (DSubReg_i16_reg imm:$lane)))>;
 def : Pat<(insertelt (v4i32 QPR:$src1), GPR:$src2, imm:$lane),
-         (v4i32 (INSERT_SUBREG QPR:$src1,
-                 GPR:$src2,
-                 (SSubReg_f32_reg imm:$lane)))>;
+          (v4i32 (INSERT_SUBREG QPR:$src1,
+                  (v2i32 (VSETLNi32 (v2i32 (EXTRACT_SUBREG QPR:$src1,
+                                     (DSubReg_i32_reg imm:$lane))),
+                             GPR:$src2, (SubReg_i32_lane imm:$lane))),
+                  (DSubReg_i32_reg imm:$lane)))>;
 
 def : Pat<(v2f32 (insertelt DPR:$src1, SPR:$src2, imm:$src3)),
           (INSERT_SUBREG (v2f32 (COPY_TO_REGCLASS DPR:$src1, DPR_VFP2)),

Modified: llvm/branches/R600/lib/Target/ARM/ARMInstrThumb.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMInstrThumb.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMInstrThumb.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMInstrThumb.td Tue Nov 13 09:21:47 2012
@@ -223,6 +223,7 @@
 def t_addrmode_pc : Operand<i32> {
   let EncoderMethod = "getAddrModePCOpValue";
   let DecoderMethod = "DecodeThumbAddrModePC";
+  let PrintMethod = "printThumbLdrLabelOperand";
 }
 
 //===----------------------------------------------------------------------===//
@@ -1246,10 +1247,6 @@
                               [(ARMeh_sjlj_longjmp GPR:$src, GPR:$scratch)]>,
                              Requires<[IsThumb, IsIOS]>;
 
-let Defs = [ R0,  R1,  R2,  R3,  R4,  R5,  R6,  R7, R12, CPSR ],
-    isBarrier = 1 in
-def tInt_eh_sjlj_dispatchsetup : PseudoInst<(outs), (ins), NoItinerary, []>;
-
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns
 //

Modified: llvm/branches/R600/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMInstrThumb2.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMInstrThumb2.td Tue Nov 13 09:21:47 2012
@@ -159,7 +159,7 @@
 // t2ldrlabel  := imm12
 def t2ldrlabel : Operand<i32> {
   let EncoderMethod = "getAddrModeImm12OpValue";
-  let PrintMethod = "printT2LdrLabelOperand";
+  let PrintMethod = "printThumbLdrLabelOperand";
 }
 
 def t2ldr_pcrel_imm12_asmoperand : AsmOperandClass {let Name = "MemPCRelImm12";}
@@ -1953,7 +1953,7 @@
 def : T2Pat<(ARMadde    rGPR:$src, t2_so_imm_not:$imm, CPSR),
             (t2SBCri    rGPR:$src, t2_so_imm_not:$imm)>;
 def : T2Pat<(ARMadde    rGPR:$src, imm0_65535_neg:$imm, CPSR),
-            (t2SBCrr    rGPR:$src, (t2MOVi16 (imm_neg_XFORM imm:$imm)))>;
+            (t2SBCrr    rGPR:$src, (t2MOVi16 (imm_not_XFORM imm:$imm)))>;
 
 // Select Bytes -- for disassembly only
 
@@ -3245,11 +3245,11 @@
   let Inst{15-14} = 0b10;
   let Inst{12} = 1;
 
-  bits<20> target;
+  bits<24> target;
   let Inst{26} = target{19};
   let Inst{11} = target{18};
   let Inst{13} = target{17};
-  let Inst{21-16} = target{16-11};
+  let Inst{25-16} = target{20-11};
   let Inst{10-0} = target{10-0};
   let DecoderMethod = "DecodeT2BInstruction";
 }

Modified: llvm/branches/R600/lib/Target/ARM/ARMRegisterInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMRegisterInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMRegisterInfo.td (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMRegisterInfo.td Tue Nov 13 09:21:47 2012
@@ -49,6 +49,9 @@
 def ssub_1  : SubRegIndex;
 def ssub_2  : SubRegIndex<[dsub_1, ssub_0]>;
 def ssub_3  : SubRegIndex<[dsub_1, ssub_1]>;
+
+def gsub_0  : SubRegIndex;
+def gsub_1  : SubRegIndex;
 // Let TableGen synthesize the remaining 12 ssub_* indices.
 // We don't need to name them.
 }
@@ -313,6 +316,17 @@
   let AltOrderSelect = [{ return 1; }];
 }
 
+// Pseudo-registers representing even-odd pairs of GPRs from R1 to R13/SP.
+// These are needed by instructions (e.g. ldrexd/strexd) requiring even-odd GPRs.
+def Tuples2R : RegisterTuples<[gsub_0, gsub_1],
+                              [(add R0, R2, R4, R6, R8, R10, R12),
+                               (add R1, R3, R5, R7, R9, R11, SP)]>;
+
+// Register class representing a pair of even-odd GPRs.
+def GPRPair : RegisterClass<"ARM", [untyped], 64, (add Tuples2R)> {
+  let Size = 64; // 2 x 32 bits, we have no predefined type of that size.
+}
+
 // Pseudo-registers representing 3 consecutive D registers.
 def Tuples3D : RegisterTuples<[dsub_0, dsub_1, dsub_2],
                               [(shl DPR, 0),

Modified: llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -68,11 +68,10 @@
                            "v128:64:128-v64:64:64-n32-S64") :
                std::string("e-p:32:32-f64:64:64-i64:64:64-"
                            "v128:64:128-v64:64:64-n32-S32")),
-    ELFWriterInfo(*this),
     TLInfo(*this),
     TSInfo(*this),
     FrameLowering(Subtarget),
-    STTI(&TLInfo) {
+    STTI(&TLInfo), VTTI(&TLInfo) {
   if (!Subtarget.hasARMOps())
     report_fatal_error("CPU: '" + Subtarget.getCPUString() + "' does not "
                        "support ARM mode execution!");
@@ -100,13 +99,12 @@
                std::string("e-p:32:32-f64:64:64-i64:64:64-"
                            "i16:16:32-i8:8:32-i1:8:32-"
                            "v128:64:128-v64:64:64-a:0:32-n32-S32")),
-    ELFWriterInfo(*this),
     TLInfo(*this),
     TSInfo(*this),
     FrameLowering(Subtarget.hasThumb2()
               ? new ARMFrameLowering(Subtarget)
               : (ARMFrameLowering*)new Thumb1FrameLowering(Subtarget)),
-    STTI(&TLInfo){
+    STTI(&TLInfo), VTTI(&TLInfo) {
 }
 
 namespace {

Modified: llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.h (original)
+++ llvm/branches/R600/lib/Target/ARM/ARMTargetMachine.h Tue Nov 13 09:21:47 2012
@@ -15,7 +15,6 @@
 #define ARMTARGETMACHINE_H
 
 #include "ARMInstrInfo.h"
-#include "ARMELFWriterInfo.h"
 #include "ARMFrameLowering.h"
 #include "ARMJITInfo.h"
 #include "ARMSubtarget.h"
@@ -64,7 +63,6 @@
   virtual void anchor();
   ARMInstrInfo        InstrInfo;
   const DataLayout    DL;       // Calculates type size & alignment
-  ARMELFWriterInfo    ELFWriterInfo;
   ARMTargetLowering   TLInfo;
   ARMSelectionDAGInfo TSInfo;
   ARMFrameLowering    FrameLowering;
@@ -99,9 +97,6 @@
   }
   virtual const ARMInstrInfo     *getInstrInfo() const { return &InstrInfo; }
   virtual const DataLayout       *getDataLayout() const { return &DL; }
-  virtual const ARMELFWriterInfo *getELFWriterInfo() const {
-    return Subtarget.isTargetELF() ? &ELFWriterInfo : 0;
-  }
 };
 
 /// ThumbTargetMachine - Thumb target machine.
@@ -113,7 +108,6 @@
   // Either Thumb1InstrInfo or Thumb2InstrInfo.
   OwningPtr<ARMBaseInstrInfo> InstrInfo;
   const DataLayout    DL;   // Calculates type size & alignment
-  ARMELFWriterInfo    ELFWriterInfo;
   ARMTargetLowering   TLInfo;
   ARMSelectionDAGInfo TSInfo;
   // Either Thumb1FrameLowering or ARMFrameLowering.
@@ -155,9 +149,6 @@
     return &VTTI;
   }
   virtual const DataLayout       *getDataLayout() const { return &DL; }
-  virtual const ARMELFWriterInfo *getELFWriterInfo() const {
-    return Subtarget.isTargetELF() ? &ELFWriterInfo : 0;
-  }
 };
 
 } // end namespace llvm

Modified: llvm/branches/R600/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/AsmParser/ARMAsmParser.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/AsmParser/ARMAsmParser.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/AsmParser/ARMAsmParser.cpp Tue Nov 13 09:21:47 2012
@@ -253,7 +253,8 @@
 
   // Implementation of the MCTargetAsmParser interface:
   bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc);
-  bool ParseInstruction(StringRef Name, SMLoc NameLoc,
+  bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
+                        SMLoc NameLoc,
                         SmallVectorImpl<MCParsedAsmOperand*> &Operands);
   bool ParseDirective(AsmToken DirectiveID);
 
@@ -4954,7 +4955,8 @@
 
 static void applyMnemonicAliases(StringRef &Mnemonic, unsigned Features);
 /// Parse an arm instruction mnemonic followed by its operands.
-bool ARMAsmParser::ParseInstruction(StringRef Name, SMLoc NameLoc,
+bool ARMAsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
+                                    SMLoc NameLoc,
                                SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
   // Apply mnemonic aliases before doing anything else, as the destination
   // mnemnonic may include suffices and we want to handle them normally.

Modified: llvm/branches/R600/lib/Target/ARM/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Target/ARM/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -22,7 +22,6 @@
   ARMCodeEmitter.cpp
   ARMConstantIslandPass.cpp
   ARMConstantPoolValue.cpp
-  ARMELFWriterInfo.cpp
   ARMExpandPseudoInsts.cpp
   ARMFastISel.cpp
   ARMFrameLowering.cpp

Modified: llvm/branches/R600/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/Disassembler/ARMDisassembler.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/Disassembler/ARMDisassembler.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/Disassembler/ARMDisassembler.cpp Tue Nov 13 09:21:47 2012
@@ -2095,16 +2095,28 @@
 static DecodeStatus
 DecodeT2BInstruction(MCInst &Inst, unsigned Insn,
                      uint64_t Address, const void *Decoder) {
-  DecodeStatus S = MCDisassembler::Success;
-  unsigned imm = (fieldFromInstruction(Insn, 0, 11) << 0) |
-                 (fieldFromInstruction(Insn, 11, 1) << 18) |
-                 (fieldFromInstruction(Insn, 13, 1) << 17) |
-                 (fieldFromInstruction(Insn, 16, 6) << 11) |
-                 (fieldFromInstruction(Insn, 26, 1) << 19);
-  if (!tryAddingSymbolicOperand(Address, Address + SignExtend32<20>(imm<<1) + 4,
+  DecodeStatus Status = MCDisassembler::Success;
+
+  // Note the J1 and J2 values are from the encoded instruction.  So here
+  // change them to I1 and I2 values via as documented:
+  // I1 = NOT(J1 EOR S);
+  // I2 = NOT(J2 EOR S);
+  // and build the imm32 with one trailing zero as documented:
+  // imm32 = SignExtend(S:I1:I2:imm10:imm11:'0', 32);
+  unsigned S = fieldFromInstruction(Insn, 26, 1);
+  unsigned J1 = fieldFromInstruction(Insn, 13, 1);
+  unsigned J2 = fieldFromInstruction(Insn, 11, 1);
+  unsigned I1 = !(J1 ^ S);
+  unsigned I2 = !(J2 ^ S);
+  unsigned imm10 = fieldFromInstruction(Insn, 16, 10);
+  unsigned imm11 = fieldFromInstruction(Insn, 0, 11);
+  unsigned tmp = (S << 23) | (I1 << 22) | (I2 << 21) | (imm10 << 11) | imm11;
+  int imm32 = SignExtend32<24>(tmp << 1);
+  if (!tryAddingSymbolicOperand(Address, Address + imm32 + 4,
                                 true, 4, Inst, Decoder))
-    Inst.addOperand(MCOperand::CreateImm(SignExtend32<20>(imm << 1)));
-  return S;
+    Inst.addOperand(MCOperand::CreateImm(imm32));
+
+  return Status;
 }
 
 static DecodeStatus

Modified: llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp (original)
+++ llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -39,7 +39,7 @@
 
 /// Prints the shift value with an immediate value.
 static void printRegImmShift(raw_ostream &O, ARM_AM::ShiftOpc ShOpc,
-                          unsigned ShImm) {
+                          unsigned ShImm, bool UseMarkup) {
   if (ShOpc == ARM_AM::no_shift || (ShOpc == ARM_AM::lsl && !ShImm))
     return;
   O << ", ";
@@ -47,8 +47,14 @@
   assert (!(ShOpc == ARM_AM::ror && !ShImm) && "Cannot have ror #0");
   O << getShiftOpcStr(ShOpc);
 
-  if (ShOpc != ARM_AM::rrx)
-    O << " #" << translateShiftImm(ShImm);
+  if (ShOpc != ARM_AM::rrx) {
+    O << " ";
+    if (UseMarkup)
+      O << "<imm:";
+    O << "#" << translateShiftImm(ShImm);
+    if (UseMarkup)
+      O << ">";
+  }
 }
 
 ARMInstPrinter::ARMInstPrinter(const MCAsmInfo &MAI,
@@ -61,7 +67,9 @@
 }
 
 void ARMInstPrinter::printRegName(raw_ostream &OS, unsigned RegNo) const {
-  OS << getRegisterName(RegNo);
+  OS << markup("<reg:")
+     << getRegisterName(RegNo)
+     << markup(">");
 }
 
 void ARMInstPrinter::printInst(const MCInst *MI, raw_ostream &O,
@@ -101,10 +109,13 @@
     printSBitModifierOperand(MI, 6, O);
     printPredicateOperand(MI, 4, O);
 
-    O << '\t' << getRegisterName(Dst.getReg())
-      << ", " << getRegisterName(MO1.getReg());
+    O << '\t';
+    printRegName(O, Dst.getReg());
+    O << ", ";
+    printRegName(O, MO1.getReg());
 
-    O << ", " << getRegisterName(MO2.getReg());
+    O << ", ";
+    printRegName(O, MO2.getReg());
     assert(ARM_AM::getSORegOffset(MO3.getImm()) == 0);
     printAnnotation(O, Annot);
     return;
@@ -120,15 +131,20 @@
     printSBitModifierOperand(MI, 5, O);
     printPredicateOperand(MI, 3, O);
 
-    O << '\t' << getRegisterName(Dst.getReg())
-      << ", " << getRegisterName(MO1.getReg());
+    O << '\t';
+    printRegName(O, Dst.getReg());
+    O << ", ";
+    printRegName(O, MO1.getReg());
 
     if (ARM_AM::getSORegShOp(MO2.getImm()) == ARM_AM::rrx) {
       printAnnotation(O, Annot);
       return;
     }
 
-    O << ", #" << translateShiftImm(ARM_AM::getSORegOffset(MO2.getImm()));
+    O << ", "
+      << markup("<imm:")
+      << "#" << translateShiftImm(ARM_AM::getSORegOffset(MO2.getImm()))
+      << markup(">");
     printAnnotation(O, Annot);
     return;
   }
@@ -152,7 +168,9 @@
       MI->getOperand(3).getImm() == -4) {
     O << '\t' << "push";
     printPredicateOperand(MI, 4, O);
-    O << "\t{" << getRegisterName(MI->getOperand(1).getReg()) << "}";
+    O << "\t{";
+    printRegName(O, MI->getOperand(1).getReg());
+    O << "}";
     printAnnotation(O, Annot);
     return;
   }
@@ -175,7 +193,9 @@
       MI->getOperand(4).getImm() == 4) {
     O << '\t' << "pop";
     printPredicateOperand(MI, 5, O);
-    O << "\t{" << getRegisterName(MI->getOperand(0).getReg()) << "}";
+    O << "\t{";
+    printRegName(O, MI->getOperand(0).getReg());
+    O << "}";
     printAnnotation(O, Annot);
     return;
   }
@@ -214,7 +234,8 @@
     O << "\tldm";
 
     printPredicateOperand(MI, 1, O);
-    O << '\t' << getRegisterName(BaseReg);
+    O << '\t';
+    printRegName(O, BaseReg);
     if (Writeback) O << "!";
     O << ", ";
     printRegisterList(MI, 3, O);
@@ -240,9 +261,11 @@
   const MCOperand &Op = MI->getOperand(OpNo);
   if (Op.isReg()) {
     unsigned Reg = Op.getReg();
-    O << getRegisterName(Reg);
+    printRegName(O, Reg);
   } else if (Op.isImm()) {
-    O << '#' << Op.getImm();
+    O << markup("<imm:")
+      << '#' << Op.getImm()
+      << markup(">");
   } else {
     assert(Op.isExpr() && "unknown operand kind in printOperand");
     // If a symbolic branch target was added as a constant expression then print
@@ -260,13 +283,16 @@
   }
 }
 
-void ARMInstPrinter::printT2LdrLabelOperand(const MCInst *MI, unsigned OpNum,
-                                       raw_ostream &O) {
+void ARMInstPrinter::printThumbLdrLabelOperand(const MCInst *MI, unsigned OpNum,
+                                               raw_ostream &O) {
   const MCOperand &MO1 = MI->getOperand(OpNum);
   if (MO1.isExpr())
     O << *MO1.getExpr();
-  else if (MO1.isImm())
-    O << "[pc, #" << MO1.getImm() << "]";
+  else if (MO1.isImm()) {
+    O << markup("<mem:") << "[pc, "
+      << markup("<imm:") << "#" << MO1.getImm()
+      << markup(">]>", "]");
+  }
   else
     llvm_unreachable("Unknown LDR label operand?");
 }
@@ -282,7 +308,7 @@
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
   const MCOperand &MO3 = MI->getOperand(OpNum+2);
 
-  O << getRegisterName(MO1.getReg());
+  printRegName(O, MO1.getReg());
 
   // Print the shift opc.
   ARM_AM::ShiftOpc ShOpc = ARM_AM::getSORegShOp(MO3.getImm());
@@ -290,7 +316,8 @@
   if (ShOpc == ARM_AM::rrx)
     return;
 
-  O << ' ' << getRegisterName(MO2.getReg());
+  O << ' ';
+  printRegName(O, MO2.getReg());
   assert(ARM_AM::getSORegOffset(MO3.getImm()) == 0);
 }
 
@@ -299,11 +326,11 @@
   const MCOperand &MO1 = MI->getOperand(OpNum);
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
-  O << getRegisterName(MO1.getReg());
+  printRegName(O, MO1.getReg());
 
   // Print the shift opc.
   printRegImmShift(O, ARM_AM::getSORegShOp(MO2.getImm()),
-                   ARM_AM::getSORegOffset(MO2.getImm()));
+                   ARM_AM::getSORegOffset(MO2.getImm()), UseMarkup);
 }
 
 
@@ -317,40 +344,51 @@
   const MCOperand &MO2 = MI->getOperand(Op+1);
   const MCOperand &MO3 = MI->getOperand(Op+2);
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   if (!MO2.getReg()) {
-    if (ARM_AM::getAM2Offset(MO3.getImm())) // Don't print +0.
-      O << ", #"
+    if (ARM_AM::getAM2Offset(MO3.getImm())) { // Don't print +0.
+      O << ", "
+        << markup("<imm:")
+        << "#"
         << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO3.getImm()))
-        << ARM_AM::getAM2Offset(MO3.getImm());
-    O << "]";
+        << ARM_AM::getAM2Offset(MO3.getImm())
+        << markup(">");
+    }
+    O << "]" << markup(">");
     return;
   }
 
-  O << ", "
-    << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO3.getImm()))
-    << getRegisterName(MO2.getReg());
+  O << ", ";
+  O << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO3.getImm()));
+  printRegName(O, MO2.getReg());
 
   printRegImmShift(O, ARM_AM::getAM2ShiftOpc(MO3.getImm()),
-                   ARM_AM::getAM2Offset(MO3.getImm()));
-  O << "]";
+                   ARM_AM::getAM2Offset(MO3.getImm()), UseMarkup);
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrModeTBB(const MCInst *MI, unsigned Op,
                                            raw_ostream &O) {
   const MCOperand &MO1 = MI->getOperand(Op);
   const MCOperand &MO2 = MI->getOperand(Op+1);
-  O << "[" << getRegisterName(MO1.getReg()) << ", "
-    << getRegisterName(MO2.getReg()) << "]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  O << ", ";
+  printRegName(O, MO2.getReg());
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrModeTBH(const MCInst *MI, unsigned Op,
                                            raw_ostream &O) {
   const MCOperand &MO1 = MI->getOperand(Op);
   const MCOperand &MO2 = MI->getOperand(Op+1);
-  O << "[" << getRegisterName(MO1.getReg()) << ", "
-    << getRegisterName(MO2.getReg()) << ", lsl #1]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  O << ", ";
+  printRegName(O, MO2.getReg());
+  O << ", lsl " << markup("<imm:") << "#1" << markup(">") << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrMode2Operand(const MCInst *MI, unsigned Op,
@@ -380,17 +418,18 @@
 
   if (!MO1.getReg()) {
     unsigned ImmOffs = ARM_AM::getAM2Offset(MO2.getImm());
-    O << '#'
-      << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO2.getImm()))
-      << ImmOffs;
+    O << markup("<imm:")
+      << '#' << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO2.getImm()))
+      << ImmOffs
+      << markup(">");
     return;
   }
 
-  O << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO2.getImm()))
-    << getRegisterName(MO1.getReg());
+  O << ARM_AM::getAddrOpcStr(ARM_AM::getAM2Op(MO2.getImm()));
+  printRegName(O, MO1.getReg());
 
   printRegImmShift(O, ARM_AM::getAM2ShiftOpc(MO2.getImm()),
-                   ARM_AM::getAM2Offset(MO2.getImm()));
+                   ARM_AM::getAM2Offset(MO2.getImm()), UseMarkup);
 }
 
 //===--------------------------------------------------------------------===//
@@ -403,18 +442,22 @@
   const MCOperand &MO2 = MI->getOperand(Op+1);
   const MCOperand &MO3 = MI->getOperand(Op+2);
 
-  O << "[" << getRegisterName(MO1.getReg()) << "], ";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  O << "], " << markup(">");
 
   if (MO2.getReg()) {
-    O << (char)ARM_AM::getAM3Op(MO3.getImm())
-    << getRegisterName(MO2.getReg());
+    O << (char)ARM_AM::getAM3Op(MO3.getImm());
+    printRegName(O, MO2.getReg());
     return;
   }
 
   unsigned ImmOffs = ARM_AM::getAM3Offset(MO3.getImm());
-  O << '#'
+  O << markup("<imm:")
+    << '#'
     << ARM_AM::getAddrOpcStr(ARM_AM::getAM3Op(MO3.getImm()))
-    << ImmOffs;
+    << ImmOffs
+    << markup(">");
 }
 
 void ARMInstPrinter::printAM3PreOrOffsetIndexOp(const MCInst *MI, unsigned Op,
@@ -423,11 +466,13 @@
   const MCOperand &MO2 = MI->getOperand(Op+1);
   const MCOperand &MO3 = MI->getOperand(Op+2);
 
-  O << '[' << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << '[';
+  printRegName(O, MO1.getReg());
 
   if (MO2.getReg()) {
-    O << ", " << getAddrOpcStr(ARM_AM::getAM3Op(MO3.getImm()))
-      << getRegisterName(MO2.getReg()) << ']';
+    O << ", " << getAddrOpcStr(ARM_AM::getAM3Op(MO3.getImm()));
+    printRegName(O, MO2.getReg());
+    O << ']' << markup(">");
     return;
   }
 
@@ -435,11 +480,15 @@
   unsigned ImmOffs = ARM_AM::getAM3Offset(MO3.getImm());
   ARM_AM::AddrOpc op = ARM_AM::getAM3Op(MO3.getImm());
 
-  if (ImmOffs || (op == ARM_AM::sub))
-    O << ", #"
+  if (ImmOffs || (op == ARM_AM::sub)) {
+    O << ", "
+      << markup("<imm:")
+      << "#"
       << ARM_AM::getAddrOpcStr(op)
-      << ImmOffs;
-  O << ']';
+      << ImmOffs
+      << markup(">");
+  }
+  O << ']' << markup(">");
 }
 
 void ARMInstPrinter::printAddrMode3Operand(const MCInst *MI, unsigned Op,
@@ -467,15 +516,15 @@
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
   if (MO1.getReg()) {
-    O << getAddrOpcStr(ARM_AM::getAM3Op(MO2.getImm()))
-      << getRegisterName(MO1.getReg());
+    O << getAddrOpcStr(ARM_AM::getAM3Op(MO2.getImm()));
+    printRegName(O, MO1.getReg());
     return;
   }
 
   unsigned ImmOffs = ARM_AM::getAM3Offset(MO2.getImm());
-  O << '#'
-    << ARM_AM::getAddrOpcStr(ARM_AM::getAM3Op(MO2.getImm()))
-    << ImmOffs;
+  O << markup("<imm:")
+    << '#' << ARM_AM::getAddrOpcStr(ARM_AM::getAM3Op(MO2.getImm())) << ImmOffs
+    << markup(">");
 }
 
 void ARMInstPrinter::printPostIdxImm8Operand(const MCInst *MI,
@@ -483,7 +532,9 @@
                                              raw_ostream &O) {
   const MCOperand &MO = MI->getOperand(OpNum);
   unsigned Imm = MO.getImm();
-  O << '#' << ((Imm & 256) ? "" : "-") << (Imm & 0xff);
+  O << markup("<imm:")
+    << '#' << ((Imm & 256) ? "" : "-") << (Imm & 0xff)
+    << markup(">");
 }
 
 void ARMInstPrinter::printPostIdxRegOperand(const MCInst *MI, unsigned OpNum,
@@ -491,7 +542,8 @@
   const MCOperand &MO1 = MI->getOperand(OpNum);
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
-  O << (MO2.getImm() ? "" : "-") << getRegisterName(MO1.getReg());
+  O << (MO2.getImm() ? "" : "-");
+  printRegName(O, MO1.getReg());
 }
 
 void ARMInstPrinter::printPostIdxImm8s4Operand(const MCInst *MI,
@@ -499,7 +551,9 @@
                                              raw_ostream &O) {
   const MCOperand &MO = MI->getOperand(OpNum);
   unsigned Imm = MO.getImm();
-  O << '#' << ((Imm & 256) ? "" : "-") << ((Imm & 0xff) << 2);
+  O << markup("<imm:")
+    << '#' << ((Imm & 256) ? "" : "-") << ((Imm & 0xff) << 2)
+    << markup(">");
 }
 
 
@@ -520,16 +574,20 @@
     return;
   }
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   unsigned ImmOffs = ARM_AM::getAM5Offset(MO2.getImm());
   unsigned Op = ARM_AM::getAM5Op(MO2.getImm());
   if (ImmOffs || Op == ARM_AM::sub) {
-    O << ", #"
+    O << ", "
+      << markup("<imm:")
+      << "#"
       << ARM_AM::getAddrOpcStr(ARM_AM::getAM5Op(MO2.getImm()))
-      << ImmOffs * 4;
+      << ImmOffs * 4
+      << markup(">");
   }
-  O << "]";
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrMode6Operand(const MCInst *MI, unsigned OpNum,
@@ -537,18 +595,21 @@
   const MCOperand &MO1 = MI->getOperand(OpNum);
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
   if (MO2.getImm()) {
     // FIXME: Both darwin as and GNU as violate ARM docs here.
     O << ", :" << (MO2.getImm() << 3);
   }
-  O << "]";
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrMode7Operand(const MCInst *MI, unsigned OpNum,
                                            raw_ostream &O) {
   const MCOperand &MO1 = MI->getOperand(OpNum);
-  O << "[" << getRegisterName(MO1.getReg()) << "]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printAddrMode6OffsetOperand(const MCInst *MI,
@@ -557,8 +618,10 @@
   const MCOperand &MO = MI->getOperand(OpNum);
   if (MO.getReg() == 0)
     O << "!";
-  else
-    O << ", " << getRegisterName(MO.getReg());
+  else {
+    O << ", ";
+    printRegName(O, MO.getReg());
+  }
 }
 
 void ARMInstPrinter::printBitfieldInvMaskImmOperand(const MCInst *MI,
@@ -569,7 +632,9 @@
   int32_t lsb = CountTrailingZeros_32(v);
   int32_t width = (32 - CountLeadingZeros_32 (v)) - lsb;
   assert(MO.isImm() && "Not a valid bf_inv_mask_imm value!");
-  O << '#' << lsb << ", #" << width;
+  O << markup("<imm:") << '#' << lsb << markup(">")
+    << ", "
+    << markup("<imm:") << '#' << width << markup(">");
 }
 
 void ARMInstPrinter::printMemBOption(const MCInst *MI, unsigned OpNum,
@@ -583,10 +648,18 @@
   unsigned ShiftOp = MI->getOperand(OpNum).getImm();
   bool isASR = (ShiftOp & (1 << 5)) != 0;
   unsigned Amt = ShiftOp & 0x1f;
-  if (isASR)
-    O << ", asr #" << (Amt == 0 ? 32 : Amt);
-  else if (Amt)
-    O << ", lsl #" << Amt;
+  if (isASR) {
+    O << ", asr "
+      << markup("<imm:")
+      << "#" << (Amt == 0 ? 32 : Amt)
+      << markup(">");
+  }
+  else if (Amt) {
+    O << ", lsl "
+      << markup("<imm:")
+      << "#" << Amt
+      << markup(">");
+  }
 }
 
 void ARMInstPrinter::printPKHLSLShiftImm(const MCInst *MI, unsigned OpNum,
@@ -595,7 +668,7 @@
   if (Imm == 0)
     return;
   assert(Imm > 0 && Imm < 32 && "Invalid PKH shift immediate value!");
-  O << ", lsl #" << Imm;
+  O << ", lsl " << markup("<imm:") << "#" << Imm << markup(">");
 }
 
 void ARMInstPrinter::printPKHASRShiftImm(const MCInst *MI, unsigned OpNum,
@@ -605,7 +678,7 @@
   if (Imm == 0)
     Imm = 32;
   assert(Imm > 0 && Imm <= 32 && "Invalid PKH shift immediate value!");
-  O << ", asr #" << Imm;
+  O << ", asr " << markup("<imm:") << "#" << Imm << markup(">");
 }
 
 void ARMInstPrinter::printRegisterList(const MCInst *MI, unsigned OpNum,
@@ -613,7 +686,7 @@
   O << "{";
   for (unsigned i = OpNum, e = MI->getNumOperands(); i != e; ++i) {
     if (i != OpNum) O << ", ";
-    O << getRegisterName(MI->getOperand(i).getReg());
+    printRegName(O, MI->getOperand(i).getReg());
   }
   O << "}";
 }
@@ -787,23 +860,29 @@
 
   int32_t OffImm = (int32_t)MO.getImm();
 
+  O << markup("<imm:");
   if (OffImm == INT32_MIN)
     O << "#-0";
   else if (OffImm < 0)
     O << "#-" << -OffImm;
   else
     O << "#" << OffImm;
+  O << markup(">");
 }
 
 void ARMInstPrinter::printThumbS4ImmOperand(const MCInst *MI, unsigned OpNum,
                                             raw_ostream &O) {
-  O << "#" << MI->getOperand(OpNum).getImm() * 4;
+  O << markup("<imm:")
+    << "#" << MI->getOperand(OpNum).getImm() * 4
+    << markup(">");
 }
 
 void ARMInstPrinter::printThumbSRImm(const MCInst *MI, unsigned OpNum,
                                      raw_ostream &O) {
   unsigned Imm = MI->getOperand(OpNum).getImm();
-  O << "#" << (Imm == 0 ? 32 : Imm);
+  O << markup("<imm:")
+    << "#" << (Imm == 0 ? 32 : Imm)
+    << markup(">");
 }
 
 void ARMInstPrinter::printThumbITMask(const MCInst *MI, unsigned OpNum,
@@ -833,10 +912,13 @@
     return;
   }
 
-  O << "[" << getRegisterName(MO1.getReg());
-  if (unsigned RegNum = MO2.getReg())
-    O << ", " << getRegisterName(RegNum);
-  O << "]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  if (unsigned RegNum = MO2.getReg()) {
+    O << ", ";
+    printRegName(O, RegNum);
+  }
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printThumbAddrModeImm5SOperand(const MCInst *MI,
@@ -851,10 +933,15 @@
     return;
   }
 
-  O << "[" << getRegisterName(MO1.getReg());
-  if (unsigned ImmOffs = MO2.getImm())
-    O << ", #" << ImmOffs * Scale;
-  O << "]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  if (unsigned ImmOffs = MO2.getImm()) {
+    O << ", "
+      << markup("<imm:")
+      << "#" << ImmOffs * Scale
+      << markup(">");
+  }
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printThumbAddrModeImm5S1Operand(const MCInst *MI,
@@ -890,12 +977,12 @@
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
   unsigned Reg = MO1.getReg();
-  O << getRegisterName(Reg);
+  printRegName(O, Reg);
 
   // Print the shift opc.
   assert(MO2.isImm() && "Not a valid t2_so_reg value!");
   printRegImmShift(O, ARM_AM::getSORegShOp(MO2.getImm()),
-                   ARM_AM::getSORegOffset(MO2.getImm()));
+                   ARM_AM::getSORegOffset(MO2.getImm()), UseMarkup);
 }
 
 void ARMInstPrinter::printAddrModeImm12Operand(const MCInst *MI, unsigned OpNum,
@@ -908,18 +995,27 @@
     return;
   }
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   int32_t OffImm = (int32_t)MO2.getImm();
   bool isSub = OffImm < 0;
   // Special value for #-0. All others are normal.
   if (OffImm == INT32_MIN)
     OffImm = 0;
-  if (isSub)
-    O << ", #-" << -OffImm;
-  else if (OffImm > 0)
-    O << ", #" << OffImm;
-  O << "]";
+  if (isSub) {
+    O << ", "
+      << markup("<imm:") 
+      << "#-" << -OffImm
+      << markup(">");
+  }
+  else if (OffImm > 0) {
+    O << ", "
+      << markup("<imm:") 
+      << "#" << OffImm
+      << markup(">");
+  }
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printT2AddrModeImm8Operand(const MCInst *MI,
@@ -928,17 +1024,24 @@
   const MCOperand &MO1 = MI->getOperand(OpNum);
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   int32_t OffImm = (int32_t)MO2.getImm();
   // Don't print +0.
+  if (OffImm != 0)
+    O << ", ";
+  if (OffImm != 0 && UseMarkup)
+    O << "<imm:";
   if (OffImm == INT32_MIN)
-    O << ", #-0";
+    O << "#-0";
   else if (OffImm < 0)
-    O << ", #-" << -OffImm;
+    O << "#-" << -OffImm;
   else if (OffImm > 0)
-    O << ", #" << OffImm;
-  O << "]";
+    O << "#" << OffImm;
+  if (OffImm != 0 && UseMarkup)
+    O << ">";
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printT2AddrModeImm8s4Operand(const MCInst *MI,
@@ -952,20 +1055,27 @@
     return;
   }
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   int32_t OffImm = (int32_t)MO2.getImm();
 
   assert(((OffImm & 0x3) == 0) && "Not a valid immediate!");
 
   // Don't print +0.
+  if (OffImm != 0)
+    O << ", ";
+  if (OffImm != 0 && UseMarkup)
+    O << "<imm:";
   if (OffImm == INT32_MIN)
-    O << ", #-0";
+    O << "#-0";
   else if (OffImm < 0)
-    O << ", #-" << -OffImm;
+    O << "#-" << -OffImm;
   else if (OffImm > 0)
-    O << ", #" << OffImm;
-  O << "]";
+    O << "#" << OffImm;
+  if (OffImm != 0 && UseMarkup)
+    O << ">";
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printT2AddrModeImm0_1020s4Operand(const MCInst *MI,
@@ -974,10 +1084,15 @@
   const MCOperand &MO1 = MI->getOperand(OpNum);
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
 
-  O << "[" << getRegisterName(MO1.getReg());
-  if (MO2.getImm())
-    O << ", #" << MO2.getImm() * 4;
-  O << "]";
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
+  if (MO2.getImm()) {
+    O << ", "
+      << markup("<imm:")
+      << "#" << MO2.getImm() * 4
+      << markup(">");
+  }
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printT2AddrModeImm8OffsetOperand(const MCInst *MI,
@@ -985,11 +1100,12 @@
                                                       raw_ostream &O) {
   const MCOperand &MO1 = MI->getOperand(OpNum);
   int32_t OffImm = (int32_t)MO1.getImm();
-  // Don't print +0.
+  O << ", " << markup("<imm:");
   if (OffImm < 0)
-    O << ", #-" << -OffImm;
+    O << "#-" << -OffImm;
   else
-    O << ", #" << OffImm;
+    O << "#" << OffImm;
+  O << markup(">");
 }
 
 void ARMInstPrinter::printT2AddrModeImm8s4OffsetOperand(const MCInst *MI,
@@ -1001,12 +1117,18 @@
   assert(((OffImm & 0x3) == 0) && "Not a valid immediate!");
 
   // Don't print +0.
+  if (OffImm != 0)
+    O << ", ";
+  if (OffImm != 0 && UseMarkup)
+    O << "<imm:";
   if (OffImm == INT32_MIN)
-    O << ", #-0";
+    O << "#-0";
   else if (OffImm < 0)
-    O << ", #-" << -OffImm;
+    O << "#-" << -OffImm;
   else if (OffImm > 0)
-    O << ", #" << OffImm;
+    O << "#" << OffImm;
+  if (OffImm != 0 && UseMarkup)
+    O << ">";
 }
 
 void ARMInstPrinter::printT2AddrModeSoRegOperand(const MCInst *MI,
@@ -1016,23 +1138,30 @@
   const MCOperand &MO2 = MI->getOperand(OpNum+1);
   const MCOperand &MO3 = MI->getOperand(OpNum+2);
 
-  O << "[" << getRegisterName(MO1.getReg());
+  O << markup("<mem:") << "[";
+  printRegName(O, MO1.getReg());
 
   assert(MO2.getReg() && "Invalid so_reg load / store address!");
-  O << ", " << getRegisterName(MO2.getReg());
+  O << ", ";
+  printRegName(O, MO2.getReg());
 
   unsigned ShAmt = MO3.getImm();
   if (ShAmt) {
     assert(ShAmt <= 3 && "Not a valid Thumb2 addressing mode!");
-    O << ", lsl #" << ShAmt;
+    O << ", lsl "
+      << markup("<imm:")
+      << "#" << ShAmt
+      << markup(">");
   }
-  O << "]";
+  O << "]" << markup(">");
 }
 
 void ARMInstPrinter::printFPImmOperand(const MCInst *MI, unsigned OpNum,
                                        raw_ostream &O) {
   const MCOperand &MO = MI->getOperand(OpNum);
-  O << '#' << ARM_AM::getFPImmFloat(MO.getImm());
+  O << markup("<imm:")
+    << '#' << ARM_AM::getFPImmFloat(MO.getImm())
+    << markup(">");
 }
 
 void ARMInstPrinter::printNEONModImmOperand(const MCInst *MI, unsigned OpNum,
@@ -1040,14 +1169,18 @@
   unsigned EncodedImm = MI->getOperand(OpNum).getImm();
   unsigned EltBits;
   uint64_t Val = ARM_AM::decodeNEONModImm(EncodedImm, EltBits);
-  O << "#0x";
+  O << markup("<imm:")
+    << "#0x";
   O.write_hex(Val);
+  O << markup(">");
 }
 
 void ARMInstPrinter::printImmPlusOneOperand(const MCInst *MI, unsigned OpNum,
                                             raw_ostream &O) {
   unsigned Imm = MI->getOperand(OpNum).getImm();
-  O << "#" << Imm + 1;
+  O << markup("<imm:")
+    << "#" << Imm + 1
+    << markup(">");
 }
 
 void ARMInstPrinter::printRotImmOperand(const MCInst *MI, unsigned OpNum,
@@ -1055,23 +1188,30 @@
   unsigned Imm = MI->getOperand(OpNum).getImm();
   if (Imm == 0)
     return;
-  O << ", ror #";
+  O << ", ror "
+    << markup("<imm:")
+    << "#";
   switch (Imm) {
   default: assert (0 && "illegal ror immediate!");
   case 1: O << "8"; break;
   case 2: O << "16"; break;
   case 3: O << "24"; break;
   }
+  O << markup(">");
 }
 
 void ARMInstPrinter::printFBits16(const MCInst *MI, unsigned OpNum,
                                   raw_ostream &O) {
-  O << "#" << 16 - MI->getOperand(OpNum).getImm();
+  O << markup("<imm:")
+    << "#" << 16 - MI->getOperand(OpNum).getImm()
+    << markup(">");
 }
 
 void ARMInstPrinter::printFBits32(const MCInst *MI, unsigned OpNum,
                                   raw_ostream &O) {
-  O << "#" << 32 - MI->getOperand(OpNum).getImm();
+  O << markup("<imm:")
+    << "#" << 32 - MI->getOperand(OpNum).getImm()
+    << markup(">");
 }
 
 void ARMInstPrinter::printVectorIndex(const MCInst *MI, unsigned OpNum,
@@ -1081,7 +1221,9 @@
 
 void ARMInstPrinter::printVectorListOne(const MCInst *MI, unsigned OpNum,
                                         raw_ostream &O) {
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListTwo(const MCInst *MI, unsigned OpNum,
@@ -1089,7 +1231,11 @@
   unsigned Reg = MI->getOperand(OpNum).getReg();
   unsigned Reg0 = MRI.getSubReg(Reg, ARM::dsub_0);
   unsigned Reg1 = MRI.getSubReg(Reg, ARM::dsub_1);
-  O << "{" << getRegisterName(Reg0) << ", " << getRegisterName(Reg1) << "}";
+  O << "{";
+  printRegName(O, Reg0);
+  O << ", ";
+  printRegName(O, Reg1);
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListTwoSpaced(const MCInst *MI,
@@ -1098,7 +1244,11 @@
   unsigned Reg = MI->getOperand(OpNum).getReg();
   unsigned Reg0 = MRI.getSubReg(Reg, ARM::dsub_0);
   unsigned Reg1 = MRI.getSubReg(Reg, ARM::dsub_2);
-  O << "{" << getRegisterName(Reg0) << ", " << getRegisterName(Reg1) << "}";
+  O << "{";
+  printRegName(O, Reg0);
+  O << ", ";
+  printRegName(O, Reg1);
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListThree(const MCInst *MI, unsigned OpNum,
@@ -1106,9 +1256,13 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 1) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << "}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 1);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListFour(const MCInst *MI, unsigned OpNum,
@@ -1116,16 +1270,23 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 1) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 3) << "}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 1);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 3);
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListOneAllLanes(const MCInst *MI,
                                                 unsigned OpNum,
                                                 raw_ostream &O) {
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "[]}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListTwoAllLanes(const MCInst *MI,
@@ -1134,7 +1295,11 @@
   unsigned Reg = MI->getOperand(OpNum).getReg();
   unsigned Reg0 = MRI.getSubReg(Reg, ARM::dsub_0);
   unsigned Reg1 = MRI.getSubReg(Reg, ARM::dsub_1);
-  O << "{" << getRegisterName(Reg0) << "[], " << getRegisterName(Reg1) << "[]}";
+  O << "{";
+  printRegName(O, Reg0);
+  O << "[], ";
+  printRegName(O, Reg1);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListThreeAllLanes(const MCInst *MI,
@@ -1143,9 +1308,13 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 1) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << "[]}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 1);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListFourAllLanes(const MCInst *MI,
@@ -1154,10 +1323,15 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 1) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 3) << "[]}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 1);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 3);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListTwoSpacedAllLanes(const MCInst *MI,
@@ -1166,7 +1340,11 @@
   unsigned Reg = MI->getOperand(OpNum).getReg();
   unsigned Reg0 = MRI.getSubReg(Reg, ARM::dsub_0);
   unsigned Reg1 = MRI.getSubReg(Reg, ARM::dsub_2);
-  O << "{" << getRegisterName(Reg0) << "[], " << getRegisterName(Reg1) << "[]}";
+  O << "{";
+  printRegName(O, Reg0);
+  O << "[], ";
+  printRegName(O, Reg1);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListThreeSpacedAllLanes(const MCInst *MI,
@@ -1175,9 +1353,13 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << "[]}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O  << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 4);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListFourSpacedAllLanes(const MCInst *MI,
@@ -1186,10 +1368,15 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << "[], "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 6) << "[]}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 4);
+  O << "[], ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 6);
+  O << "[]}";
 }
 
 void ARMInstPrinter::printVectorListThreeSpaced(const MCInst *MI,
@@ -1198,9 +1385,13 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << "}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 4);
+  O << "}";
 }
 
 void ARMInstPrinter::printVectorListFourSpaced(const MCInst *MI,
@@ -1209,8 +1400,13 @@
   // Normally, it's not safe to use register enum values directly with
   // addition to get the next register, but for VFP registers, the
   // sort order is guaranteed because they're all of the form D<n>.
-  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << ", "
-    << getRegisterName(MI->getOperand(OpNum).getReg() + 6) << "}";
+  O << "{";
+  printRegName(O, MI->getOperand(OpNum).getReg());
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 2);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 4);
+  O << ", ";
+  printRegName(O, MI->getOperand(OpNum).getReg() + 6);
+  O << "}";
 }

Modified: llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.h (original)
+++ llvm/branches/R600/lib/Target/ARM/InstPrinter/ARMInstPrinter.h Tue Nov 13 09:21:47 2012
@@ -126,7 +126,8 @@
   void printRotImmOperand(const MCInst *MI, unsigned OpNum, raw_ostream &O);
 
   void printPCLabel(const MCInst *MI, unsigned OpNum, raw_ostream &O);
-  void printT2LdrLabelOperand(const MCInst *MI, unsigned OpNum, raw_ostream &O);
+  void printThumbLdrLabelOperand(const MCInst *MI, unsigned OpNum,
+                                 raw_ostream &O);
   void printFBits16(const MCInst *MI, unsigned OpNum, raw_ostream &O);
   void printFBits32(const MCInst *MI, unsigned OpNum, raw_ostream &O);
   void printVectorIndex(const MCInst *MI, unsigned OpNum, raw_ostream &O);

Modified: llvm/branches/R600/lib/Target/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Target/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -1,7 +1,6 @@
 add_llvm_library(LLVMTarget
   Mangler.cpp
   Target.cpp
-  TargetELFWriterInfo.cpp
   TargetInstrInfo.cpp
   TargetIntrinsicInfo.cpp
   TargetJITInfo.cpp

Modified: llvm/branches/R600/lib/Target/CellSPU/SPUTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/CellSPU/SPUTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/CellSPU/SPUTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/CellSPU/SPUTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -44,7 +44,7 @@
     TLInfo(*this),
     TSInfo(*this),
     InstrItins(Subtarget.getInstrItineraryData()),
-    STTI(&TLInfo){
+    STTI(&TLInfo), VTTI(&TLInfo) {
 }
 
 //===----------------------------------------------------------------------===//

Modified: llvm/branches/R600/lib/Target/CppBackend/CPPBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/CppBackend/CPPBackend.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/CppBackend/CPPBackend.cpp (original)
+++ llvm/branches/R600/lib/Target/CppBackend/CPPBackend.cpp Tue Nov 13 09:21:47 2012
@@ -476,11 +476,11 @@
       unsigned index = PAL.getSlot(i).Index;
       AttrBuilder attrs(PAL.getSlot(i).Attrs);
       Out << "PAWI.Index = " << index << "U;\n";
-      Out << "   AttrBuilder B;\n";
+      Out << " {\n    AttrBuilder B;\n";
 
 #define HANDLE_ATTR(X)                                     \
       if (attrs.hasAttribute(Attributes::X))               \
-        Out << "   B.addAttribute(Attributes::" #X ");\n"; \
+        Out << "    B.addAttribute(Attributes::" #X ");\n"; \
       attrs.removeAttribute(Attributes::X);
 
       HANDLE_ATTR(SExt);
@@ -507,13 +507,13 @@
       HANDLE_ATTR(ReturnsTwice);
       HANDLE_ATTR(UWTable);
       HANDLE_ATTR(NonLazyBind);
+      HANDLE_ATTR(MinSize);
 #undef HANDLE_ATTR
       if (attrs.hasAttribute(Attributes::StackAlignment))
-        Out << "B.addStackAlignmentAttr(" << attrs.getStackAlignment() << ")";
-      nl(Out);
+        Out << "    B.addStackAlignmentAttr(" << attrs.getStackAlignment() << ")\n";
       attrs.removeAttribute(Attributes::StackAlignment);
       assert(!attrs.hasAttributes() && "Unhandled attribute!");
-      Out << "PAWI.Attrs = Attributes::get(mod->getContext(), B);";
+      Out << "    PAWI.Attrs = Attributes::get(mod->getContext(), B);\n }";
       nl(Out);
       Out << "Attrs.push_back(PAWI);";
       nl(Out);

Modified: llvm/branches/R600/lib/Target/Hexagon/HexagonInstrFormats.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Hexagon/HexagonInstrFormats.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Hexagon/HexagonInstrFormats.td (original)
+++ llvm/branches/R600/lib/Target/Hexagon/HexagonInstrFormats.td Tue Nov 13 09:21:47 2012
@@ -56,6 +56,16 @@
   bits<1> isPredicated = 0;
   let TSFlags{6} = isPredicated;
 
+  // Dot new value store instructions.
+  bits<1> isNVStore = 0;
+  let TSFlags{8} = isNVStore;
+
+  // Fields used for relation models.
+  string BaseOpcode = "";
+  string CextOpcode = "";
+  string PredSense = "";
+  string PNewValue = "";
+  string InputType = "";    // Input is "imm" or "reg" type.
   // *** The code above must match HexagonBaseInfo.h ***
 }
 

Modified: llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.cpp Tue Nov 13 09:21:47 2012
@@ -25,6 +25,7 @@
 #include "llvm/CodeGen/PseudoSourceValue.h"
 #include "llvm/Support/MathExtras.h"
 #define GET_INSTRINFO_CTOR
+#define GET_INSTRMAP_INFO
 #include "HexagonGenInstrInfo.inc"
 #include "HexagonGenDFAPacketizer.inc"
 
@@ -1915,6 +1916,15 @@
 
 int HexagonInstrInfo::
 getMatchingCondBranchOpcode(int Opc, bool invertPredicate) const {
+  enum Hexagon::PredSense inPredSense;
+  inPredSense = invertPredicate ? Hexagon::PredSense_false :
+                                  Hexagon::PredSense_true;
+  int CondOpcode = Hexagon::getPredOpcode(Opc, inPredSense);
+  if (CondOpcode >= 0) // Valid Conditional opcode/instruction
+    return CondOpcode;
+
+  // This switch case will be removed once all the instructions have been
+  // modified to use relation maps.
   switch(Opc) {
   case Hexagon::TFR:
     return !invertPredicate ? Hexagon::TFR_cPt :
@@ -1934,24 +1944,6 @@
   case Hexagon::JMP_EQriPt_nv_V4:
     return !invertPredicate ? Hexagon::JMP_EQriPt_nv_V4 :
                               Hexagon::JMP_EQriNotPt_nv_V4;
-  case Hexagon::ADD_ri:
-    return !invertPredicate ? Hexagon::ADD_ri_cPt :
-                              Hexagon::ADD_ri_cNotPt;
-  case Hexagon::ADD_rr:
-    return !invertPredicate ? Hexagon::ADD_rr_cPt :
-                              Hexagon::ADD_rr_cNotPt;
-  case Hexagon::XOR_rr:
-    return !invertPredicate ? Hexagon::XOR_rr_cPt :
-                              Hexagon::XOR_rr_cNotPt;
-  case Hexagon::AND_rr:
-    return !invertPredicate ? Hexagon::AND_rr_cPt :
-                              Hexagon::AND_rr_cNotPt;
-  case Hexagon::OR_rr:
-    return !invertPredicate ? Hexagon::OR_rr_cPt :
-                              Hexagon::OR_rr_cNotPt;
-  case Hexagon::SUB_rr:
-    return !invertPredicate ? Hexagon::SUB_rr_cPt :
-                              Hexagon::SUB_rr_cNotPt;
   case Hexagon::COMBINE_rr:
     return !invertPredicate ? Hexagon::COMBINE_rr_cPt :
                               Hexagon::COMBINE_rr_cNotPt;

Modified: llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.td (original)
+++ llvm/branches/R600/lib/Target/Hexagon/HexagonInstrInfo.td Tue Nov 13 09:21:47 2012
@@ -15,6 +15,18 @@
 include "HexagonImmediates.td"
 
 //===----------------------------------------------------------------------===//
+// Classes used for relation maps.
+//===----------------------------------------------------------------------===//
+// PredRel - Filter class used to relate non-predicated instructions with their
+// predicated forms.
+class PredRel;
+// PredNewRel - Filter class used to relate predicated instructions with their
+// predicate-new forms.
+class PredNewRel: PredRel;
+// ImmRegRel - Filter class used to relate instructions having reg-reg form
+// with their reg-imm counterparts.
+class ImmRegRel;
+//===----------------------------------------------------------------------===//
 // Hexagon Instruction Predicate Definitions.
 //===----------------------------------------------------------------------===//
 def HasV2T                      : Predicate<"Subtarget.hasV2TOps()">;
@@ -148,37 +160,91 @@
 }
 
 //===----------------------------------------------------------------------===//
-// ALU32/ALU +
+// ALU32/ALU (Instructions with register-register form)
 //===----------------------------------------------------------------------===//
-// Add.
-let isCommutable = 1, isPredicable = 1 in
-def ADD_rr : ALU32_rr<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, IntRegs:$src2),
-            "$dst = add($src1, $src2)",
-            [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$src1),
-                                           (i32 IntRegs:$src2)))]>;
+multiclass ALU32_Pbase<string mnemonic, bit isNot,
+                       bit isPredNew> {
 
-let isPredicable = 1 in
-def ADD_ri : ALU32_ri<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, s16Imm:$src2),
-            "$dst = add($src1, #$src2)",
-            [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$src1),
-                                           s16ImmPred:$src2))]>;
+  let PNewValue = #!if(isPredNew, "new", "") in
+  def #NAME# : ALU32_rr<(outs IntRegs:$dst),
+            (ins PredRegs:$src1, IntRegs:$src2, IntRegs: $src3),
+            !if(isNot, "if (!$src1", "if ($src1")#!if(isPredNew,".new) $dst = ",
+            ") $dst = ")#mnemonic#"($src2, $src3)",
+            []>;
+}
 
-// Logical operations.
-let isPredicable = 1 in
-def XOR_rr : ALU32_rr<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, IntRegs:$src2),
-            "$dst = xor($src1, $src2)",
-            [(set (i32 IntRegs:$dst), (xor (i32 IntRegs:$src1),
-                                           (i32 IntRegs:$src2)))]>;
+multiclass ALU32_Pred<string mnemonic, bit PredNot> {
+  let PredSense = #!if(PredNot, "false", "true") in {
+    defm _c#NAME# : ALU32_Pbase<mnemonic, PredNot, 0>;
+    // Predicate new
+    defm _cdn#NAME# : ALU32_Pbase<mnemonic, PredNot, 1>;
+  }
+}
 
-let isCommutable = 1, isPredicable = 1 in
-def AND_rr : ALU32_rr<(outs IntRegs:$dst),
+let InputType = "reg" in
+multiclass ALU32_base<string mnemonic, string CextOp, SDNode OpNode> {
+  let CextOpcode = CextOp, BaseOpcode = CextOp#_rr in {
+    let isPredicable = 1 in
+    def #NAME# : ALU32_rr<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
-            "$dst = and($src1, $src2)",
-            [(set (i32 IntRegs:$dst), (and (i32 IntRegs:$src1),
-                                           (i32 IntRegs:$src2)))]>;
+            "$dst = "#mnemonic#"($src1, $src2)",
+            [(set (i32 IntRegs:$dst), (OpNode (i32 IntRegs:$src1),
+                                              (i32 IntRegs:$src2)))]>;
+
+    let neverHasSideEffects = 1, isPredicated = 1 in {
+      defm Pt : ALU32_Pred<mnemonic, 0>;
+      defm NotPt : ALU32_Pred<mnemonic, 1>;
+    }
+  }
+}
+
+let isCommutable = 1 in {
+  defm ADD_rr : ALU32_base<"add", "ADD", add>, ImmRegRel, PredNewRel;
+  defm AND_rr : ALU32_base<"and", "AND", and>, ImmRegRel, PredNewRel;
+  defm XOR_rr : ALU32_base<"xor", "XOR", xor>, ImmRegRel, PredNewRel;
+  defm OR_rr  : ALU32_base<"or", "OR", or>, ImmRegRel, PredNewRel;
+}
+
+defm SUB_rr : ALU32_base<"sub", "SUB", sub>, ImmRegRel, PredNewRel;
+
+//===----------------------------------------------------------------------===//
+// ALU32/ALU (ADD with register-immediate form)
+//===----------------------------------------------------------------------===//
+multiclass ALU32ri_Pbase<string mnemonic, bit isNot, bit isPredNew> {
+  let PNewValue = #!if(isPredNew, "new", "") in
+  def #NAME# : ALU32_ri<(outs IntRegs:$dst),
+            (ins PredRegs:$src1, IntRegs:$src2, s8Imm: $src3),
+            !if(isNot, "if (!$src1", "if ($src1")#!if(isPredNew,".new) $dst = ",
+            ") $dst = ")#mnemonic#"($src2, #$src3)",
+            []>;
+}
+
+multiclass ALU32ri_Pred<string mnemonic, bit PredNot> {
+  let PredSense = #!if(PredNot, "false", "true") in {
+    defm _c#NAME# : ALU32ri_Pbase<mnemonic, PredNot, 0>;
+    // Predicate new
+    defm _cdn#NAME# : ALU32ri_Pbase<mnemonic, PredNot, 1>;
+  }
+}
+
+let InputType = "imm" in
+multiclass ALU32ri_base<string mnemonic, string CextOp, SDNode OpNode> {
+  let CextOpcode = CextOp, BaseOpcode = CextOp#_ri in {
+    let isPredicable = 1 in
+    def #NAME# : ALU32_ri<(outs IntRegs:$dst),
+            (ins IntRegs:$src1, s16Imm:$src2),
+            "$dst = "#mnemonic#"($src1, #$src2)",
+            [(set (i32 IntRegs:$dst), (OpNode (i32 IntRegs:$src1),
+                                              (s16ImmPred:$src2)))]>;
+
+    let neverHasSideEffects = 1, isPredicated = 1 in {
+      defm Pt : ALU32ri_Pred<mnemonic, 0>;
+      defm NotPt : ALU32ri_Pred<mnemonic, 1>;
+    }
+  }
+}
+
+defm ADD_ri : ALU32ri_base<"add", "ADD", add>, ImmRegRel, PredNewRel;
 
 def OR_ri : ALU32_ri<(outs IntRegs:$dst),
             (ins IntRegs:$src1, s10Imm:$src2),
@@ -197,13 +263,6 @@
             [(set (i32 IntRegs:$dst), (and (i32 IntRegs:$src1),
                                            s10ImmPred:$src2))]>;
 
-let isCommutable = 1, isPredicable = 1 in
-def OR_rr : ALU32_rr<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, IntRegs:$src2),
-            "$dst = or($src1, $src2)",
-            [(set (i32 IntRegs:$dst), (or (i32 IntRegs:$src1),
-                                          (i32 IntRegs:$src2)))]>;
-
 // Negate.
 def NEG : ALU32_rr<(outs IntRegs:$dst), (ins IntRegs:$src1),
           "$dst = neg($src1)",
@@ -214,14 +273,6 @@
           "nop",
           []>;
 
-// Subtract.
-let isPredicable = 1 in
-def SUB_rr : ALU32_rr<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, IntRegs:$src2),
-            "$dst = sub($src1, $src2)",
-            [(set (i32 IntRegs:$dst), (sub (i32 IntRegs:$src1),
-                                           (i32 IntRegs:$src2)))]>;
-
 // Rd32=sub(#s10,Rs32)
 def SUB_ri : ALU32_ri<(outs IntRegs:$dst),
             (ins s10Imm:$src1, IntRegs:$src2),
@@ -348,56 +399,6 @@
 // ALU32/PRED +
 //===----------------------------------------------------------------------===//
 
-// Conditional add.
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_ri_cPt : ALU32_ri<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, s8Imm:$src3),
-            "if ($src1) $dst = add($src2, #$src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_ri_cNotPt : ALU32_ri<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, s8Imm:$src3),
-            "if (!$src1) $dst = add($src2, #$src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_ri_cdnPt : ALU32_ri<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, s8Imm:$src3),
-            "if ($src1.new) $dst = add($src2, #$src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_ri_cdnNotPt : ALU32_ri<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, s8Imm:$src3),
-            "if (!$src1.new) $dst = add($src2, #$src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_rr_cPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1) $dst = add($src2, $src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_rr_cNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1) $dst = add($src2, $src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_rr_cdnPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1.new) $dst = add($src2, $src3)",
-            []>;
-
-let neverHasSideEffects = 1, isPredicated = 1 in
-def ADD_rr_cdnNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1.new) $dst = add($src2, $src3)",
-            []>;
-
-
 // Conditional combine.
 
 let neverHasSideEffects = 1, isPredicated = 1 in
@@ -424,108 +425,6 @@
             "if (!$src1.new) $dst = combine($src2, $src3)",
             []>;
 
-// Conditional logical operations.
-
-let isPredicated = 1 in
-def XOR_rr_cPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1) $dst = xor($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def XOR_rr_cNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1) $dst = xor($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def XOR_rr_cdnPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1.new) $dst = xor($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def XOR_rr_cdnNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1.new) $dst = xor($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def AND_rr_cPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1) $dst = and($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def AND_rr_cNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1) $dst = and($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def AND_rr_cdnPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1.new) $dst = and($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def AND_rr_cdnNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1.new) $dst = and($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def OR_rr_cPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1) $dst = or($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def OR_rr_cNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1) $dst = or($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def OR_rr_cdnPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1.new) $dst = or($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def OR_rr_cdnNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1.new) $dst = or($src2, $src3)",
-            []>;
-
-
-// Conditional subtract.
-
-let isPredicated = 1 in
-def SUB_rr_cPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1) $dst = sub($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def SUB_rr_cNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1) $dst = sub($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def SUB_rr_cdnPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if ($src1.new) $dst = sub($src2, $src3)",
-            []>;
-
-let isPredicated = 1 in
-def SUB_rr_cdnNotPt : ALU32_rr<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3),
-            "if (!$src1.new) $dst = sub($src2, $src3)",
-            []>;
-
-
 // Conditional transfer.
 let neverHasSideEffects = 1, isPredicated = 1 in
 def TFR_cPt : ALU32_rr<(outs IntRegs:$dst), (ins PredRegs:$src1, IntRegs:$src2),
@@ -3546,4 +3445,31 @@
 // V5 Instructions -
 //===----------------------------------------------------------------------===//
 
+//===----------------------------------------------------------------------===//
+// Generate mapping table to relate non-predicate instructions with their
+// predicated formats - true and false.
+//
+
+def getPredOpcode : InstrMapping {
+  let FilterClass = "PredRel";
+  // Instructions with the same BaseOpcode and isNVStore values form a row.
+  let RowFields = ["BaseOpcode", "isNVStore", "PNewValue"];
+  // Instructions with the same predicate sense form a column.
+  let ColFields = ["PredSense"];
+  // The key column is the unpredicated instructions.
+  let KeyCol = [""];
+  // Value columns are PredSense=true and PredSense=false
+  let ValueCols = [["true"], ["false"]];
+}
 
+//===----------------------------------------------------------------------===//
+// Generate mapping table to relate predicated instructions with their .new
+// format.
+//
+def getPredNewOpcode : InstrMapping {
+  let FilterClass = "PredNewRel";
+  let RowFields = ["BaseOpcode", "PredSense", "isNVStore"];
+  let ColFields = ["PNewValue"];
+  let KeyCol = [""];
+  let ValueCols = [["new"]];
+}

Modified: llvm/branches/R600/lib/Target/Hexagon/HexagonMachineScheduler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Hexagon/HexagonMachineScheduler.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Hexagon/HexagonMachineScheduler.cpp (original)
+++ llvm/branches/R600/lib/Target/Hexagon/HexagonMachineScheduler.cpp Tue Nov 13 09:21:47 2012
@@ -31,8 +31,7 @@
       LastSequentialCall = &(SUnits[su]);
     // Look for a compare that defines a predicate.
     else if (SUnits[su].getInstr()->isCompare() && LastSequentialCall)
-      SUnits[su].addPred(SDep(LastSequentialCall, SDep::Order, 0, /*Reg=*/0,
-                              false));
+      SUnits[su].addPred(SDep(LastSequentialCall, SDep::Barrier));
   }
 }
 

Modified: llvm/branches/R600/lib/Target/Hexagon/HexagonTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Hexagon/HexagonTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Hexagon/HexagonTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/Hexagon/HexagonTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -75,7 +75,7 @@
     TSInfo(*this),
     FrameLowering(Subtarget),
     InstrItins(&Subtarget.getInstrItineraryData()),
-    STTI(&TLInfo) {
+    STTI(&TLInfo), VTTI(&TLInfo) {
   setMCUseCFI(false);
 }
 

Modified: llvm/branches/R600/lib/Target/MBlaze/AsmParser/MBlazeAsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/AsmParser/MBlazeAsmParser.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/AsmParser/MBlazeAsmParser.cpp (original)
+++ llvm/branches/R600/lib/Target/MBlaze/AsmParser/MBlazeAsmParser.cpp Tue Nov 13 09:21:47 2012
@@ -61,7 +61,8 @@
   MBlazeAsmParser(MCSubtargetInfo &_STI, MCAsmParser &_Parser)
     : MCTargetAsmParser(), Parser(_Parser) {}
 
-  virtual bool ParseInstruction(StringRef Name, SMLoc NameLoc,
+  virtual bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
+                                SMLoc NameLoc,
                                 SmallVectorImpl<MCParsedAsmOperand*> &Operands);
 
   virtual bool ParseDirective(AsmToken DirectiveID);
@@ -477,7 +478,7 @@
 
 /// Parse an mblaze instruction mnemonic followed by its operands.
 bool MBlazeAsmParser::
-ParseInstruction(StringRef Name, SMLoc NameLoc,
+ParseInstruction(ParseInstructionInfo &Info, StringRef Name, SMLoc NameLoc,
                  SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
   // The first operands is the token for the instruction name
   size_t dotLoc = Name.find('.');

Modified: llvm/branches/R600/lib/Target/MBlaze/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Target/MBlaze/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -27,7 +27,6 @@
   MBlazeSelectionDAGInfo.cpp
   MBlazeAsmPrinter.cpp
   MBlazeMCInstLower.cpp
-  MBlazeELFWriterInfo.cpp
   )
 
 add_dependencies(LLVMMBlazeCodeGen intrinsics_gen)

Removed: llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.cpp (removed)
@@ -1,107 +0,0 @@
-//===-- MBlazeELFWriterInfo.cpp - ELF Writer Info for the MBlaze backend --===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the MBlaze backend.
-//
-//===----------------------------------------------------------------------===//
-
-#include "MBlazeELFWriterInfo.h"
-#include "MBlazeRelocations.h"
-#include "llvm/Function.h"
-#include "llvm/Support/ELF.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/DataLayout.h"
-#include "llvm/Target/TargetMachine.h"
-
-using namespace llvm;
-
-//===----------------------------------------------------------------------===//
-//  Implementation of the MBlazeELFWriterInfo class
-//===----------------------------------------------------------------------===//
-
-MBlazeELFWriterInfo::MBlazeELFWriterInfo(TargetMachine &TM)
-  : TargetELFWriterInfo(TM.getDataLayout()->getPointerSizeInBits(0) == 64,
-                        TM.getDataLayout()->isLittleEndian()) {
-}
-
-MBlazeELFWriterInfo::~MBlazeELFWriterInfo() {}
-
-unsigned MBlazeELFWriterInfo::getRelocationType(unsigned MachineRelTy) const {
-  switch (MachineRelTy) {
-  case MBlaze::reloc_pcrel_word:
-    return ELF::R_MICROBLAZE_64_PCREL;
-  case MBlaze::reloc_absolute_word:
-    return ELF::R_MICROBLAZE_NONE;
-  default:
-    llvm_unreachable("unknown mblaze machine relocation type");
-  }
-}
-
-long int MBlazeELFWriterInfo::getDefaultAddendForRelTy(unsigned RelTy,
-                                                    long int Modifier) const {
-  switch (RelTy) {
-  case ELF::R_MICROBLAZE_32_PCREL:
-    return Modifier - 4;
-  case ELF::R_MICROBLAZE_32:
-    return Modifier;
-  default:
-    llvm_unreachable("unknown mblaze relocation type");
-  }
-}
-
-unsigned MBlazeELFWriterInfo::getRelocationTySize(unsigned RelTy) const {
-  // FIXME: Most of these sizes are guesses based on the name
-  switch (RelTy) {
-  case ELF::R_MICROBLAZE_32:
-  case ELF::R_MICROBLAZE_32_PCREL:
-  case ELF::R_MICROBLAZE_32_PCREL_LO:
-  case ELF::R_MICROBLAZE_32_LO:
-  case ELF::R_MICROBLAZE_SRO32:
-  case ELF::R_MICROBLAZE_SRW32:
-  case ELF::R_MICROBLAZE_32_SYM_OP_SYM:
-  case ELF::R_MICROBLAZE_GOTOFF_32:
-    return 32;
-
-  case ELF::R_MICROBLAZE_64_PCREL:
-  case ELF::R_MICROBLAZE_64:
-  case ELF::R_MICROBLAZE_GOTPC_64:
-  case ELF::R_MICROBLAZE_GOT_64:
-  case ELF::R_MICROBLAZE_PLT_64:
-  case ELF::R_MICROBLAZE_GOTOFF_64:
-    return 64;
-  }
-
-  return 0;
-}
-
-bool MBlazeELFWriterInfo::isPCRelativeRel(unsigned RelTy) const {
-  // FIXME: Most of these are guesses based on the name
-  switch (RelTy) {
-  case ELF::R_MICROBLAZE_32_PCREL:
-  case ELF::R_MICROBLAZE_64_PCREL:
-  case ELF::R_MICROBLAZE_32_PCREL_LO:
-  case ELF::R_MICROBLAZE_GOTPC_64:
-    return true;
-  }
-
-  return false;
-}
-
-unsigned MBlazeELFWriterInfo::getAbsoluteLabelMachineRelTy() const {
-  return MBlaze::reloc_absolute_word;
-}
-
-long int MBlazeELFWriterInfo::computeRelocation(unsigned SymOffset,
-                                                unsigned RelOffset,
-                                                unsigned RelTy) const {
-  assert((RelTy == ELF::R_MICROBLAZE_32_PCREL ||
-          RelTy == ELF::R_MICROBLAZE_64_PCREL) &&
-         "computeRelocation unknown for this relocation type");
-  return SymOffset - (RelOffset + 4);
-}

Removed: llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.h (original)
+++ llvm/branches/R600/lib/Target/MBlaze/MBlazeELFWriterInfo.h (removed)
@@ -1,59 +0,0 @@
-//===-- MBlazeELFWriterInfo.h - ELF Writer Info for MBlaze ------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the MBlaze backend.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef MBLAZE_ELF_WRITER_INFO_H
-#define MBLAZE_ELF_WRITER_INFO_H
-
-#include "llvm/Target/TargetELFWriterInfo.h"
-
-namespace llvm {
-  class TargetMachine;
-
-  class MBlazeELFWriterInfo : public TargetELFWriterInfo {
-  public:
-    MBlazeELFWriterInfo(TargetMachine &TM);
-    virtual ~MBlazeELFWriterInfo();
-
-    /// getRelocationType - Returns the target specific ELF Relocation type.
-    /// 'MachineRelTy' contains the object code independent relocation type
-    virtual unsigned getRelocationType(unsigned MachineRelTy) const;
-
-    /// hasRelocationAddend - True if the target uses an addend in the
-    /// ELF relocation entry.
-    virtual bool hasRelocationAddend() const { return false; }
-
-    /// getDefaultAddendForRelTy - Gets the default addend value for a
-    /// relocation entry based on the target ELF relocation type.
-    virtual long int getDefaultAddendForRelTy(unsigned RelTy,
-                                              long int Modifier = 0) const;
-
-    /// getRelTySize - Returns the size of relocatable field in bits
-    virtual unsigned getRelocationTySize(unsigned RelTy) const;
-
-    /// isPCRelativeRel - True if the relocation type is pc relative
-    virtual bool isPCRelativeRel(unsigned RelTy) const;
-
-    /// getJumpTableRelocationTy - Returns the machine relocation type used
-    /// to reference a jumptable.
-    virtual unsigned getAbsoluteLabelMachineRelTy() const;
-
-    /// computeRelocation - Some relocatable fields could be relocated
-    /// directly, avoiding the relocation symbol emission, compute the
-    /// final relocation value for this symbol.
-    virtual long int computeRelocation(unsigned SymOffset, unsigned RelOffset,
-                                       unsigned RelTy) const;
-  };
-
-} // end llvm namespace
-
-#endif // MBLAZE_ELF_WRITER_INFO_H

Modified: llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -41,8 +41,9 @@
     DL("E-p:32:32:32-i8:8:8-i16:16:16"),
     InstrInfo(*this),
     FrameLowering(Subtarget),
-    TLInfo(*this), TSInfo(*this), ELFWriterInfo(*this),
-    InstrItins(Subtarget.getInstrItineraryData()), STTI(&TLInfo) {
+    TLInfo(*this), TSInfo(*this),
+    InstrItins(Subtarget.getInstrItineraryData()),
+    STTI(&TLInfo), VTTI(&TLInfo) {
 }
 
 namespace {

Modified: llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.h (original)
+++ llvm/branches/R600/lib/Target/MBlaze/MBlazeTargetMachine.h Tue Nov 13 09:21:47 2012
@@ -20,7 +20,6 @@
 #include "MBlazeSelectionDAGInfo.h"
 #include "MBlazeIntrinsicInfo.h"
 #include "MBlazeFrameLowering.h"
-#include "MBlazeELFWriterInfo.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/DataLayout.h"
@@ -38,7 +37,6 @@
     MBlazeTargetLowering   TLInfo;
     MBlazeSelectionDAGInfo TSInfo;
     MBlazeIntrinsicInfo    IntrinsicInfo;
-    MBlazeELFWriterInfo    ELFWriterInfo;
     InstrItineraryData     InstrItins;
     ScalarTargetTransformImpl STTI;
     VectorTargetTransformImpl VTTI;
@@ -77,9 +75,6 @@
     const TargetIntrinsicInfo *getIntrinsicInfo() const
     { return &IntrinsicInfo; }
 
-    virtual const MBlazeELFWriterInfo *getELFWriterInfo() const {
-      return &ELFWriterInfo;
-    }
     virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const
     { return &STTI; }
     virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const

Modified: llvm/branches/R600/lib/Target/MSP430/MSP430ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MSP430/MSP430ISelLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MSP430/MSP430ISelLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/MSP430/MSP430ISelLowering.cpp Tue Nov 13 09:21:47 2012
@@ -881,7 +881,7 @@
 
   if (ReturnAddrIndex == 0) {
     // Set up a frame object for the return address.
-    uint64_t SlotSize = TD->getPointerSize(0);
+    uint64_t SlotSize = TD->getPointerSize();
     ReturnAddrIndex = MF.getFrameInfo()->CreateFixedObject(SlotSize, -SlotSize,
                                                            true);
     FuncInfo->setRAIndex(ReturnAddrIndex);
@@ -901,7 +901,7 @@
   if (Depth > 0) {
     SDValue FrameAddr = LowerFRAMEADDR(Op, DAG);
     SDValue Offset =
-      DAG.getConstant(TD->getPointerSize(0), MVT::i16);
+      DAG.getConstant(TD->getPointerSize(), MVT::i16);
     return DAG.getLoad(getPointerTy(), dl, DAG.getEntryNode(),
                        DAG.getNode(ISD::ADD, dl, getPointerTy(),
                                    FrameAddr, Offset),

Modified: llvm/branches/R600/lib/Target/MSP430/MSP430TargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/MSP430/MSP430TargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/MSP430/MSP430TargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/MSP430/MSP430TargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -36,7 +36,7 @@
     // FIXME: Check DataLayout string.
     DL("e-p:16:16:16-i8:8:8-i16:16:16-i32:16:32-n8:16"),
     InstrInfo(*this), TLInfo(*this), TSInfo(*this),
-    FrameLowering(Subtarget), STTI(&TLInfo) { }
+    FrameLowering(Subtarget), STTI(&TLInfo), VTTI(&TLInfo) { }
 
 namespace {
 /// MSP430 Code Generator Pass Configuration Options.

Modified: llvm/branches/R600/lib/Target/Mips/AsmParser/MipsAsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/AsmParser/MipsAsmParser.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/AsmParser/MipsAsmParser.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/AsmParser/MipsAsmParser.cpp Tue Nov 13 09:21:47 2012
@@ -74,7 +74,8 @@
 
   bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc);
 
-  bool ParseInstruction(StringRef Name, SMLoc NameLoc,
+  bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
+                        SMLoc NameLoc,
                         SmallVectorImpl<MCParsedAsmOperand*> &Operands);
 
   bool parseMathOperation(StringRef Name, SMLoc NameLoc,
@@ -1056,7 +1057,7 @@
 }
 
 bool MipsAsmParser::
-ParseInstruction(StringRef Name, SMLoc NameLoc,
+ParseInstruction(ParseInstructionInfo &Info, StringRef Name, SMLoc NameLoc,
                  SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
   // floating point instructions: should register be treated as double?
   if (requestsDoubleOperand(Name)) {

Modified: llvm/branches/R600/lib/Target/Mips/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/CMakeLists.txt?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/CMakeLists.txt (original)
+++ llvm/branches/R600/lib/Target/Mips/CMakeLists.txt Tue Nov 13 09:21:47 2012
@@ -22,7 +22,6 @@
   MipsAsmPrinter.cpp
   MipsCodeEmitter.cpp
   MipsDelaySlotFiller.cpp
-  MipsELFWriterInfo.cpp
   MipsJITInfo.cpp
   MipsInstrInfo.cpp
   MipsISelDAGToDAG.cpp

Modified: llvm/branches/R600/lib/Target/Mips/Mips16FrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16FrameLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16FrameLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16FrameLowering.cpp Tue Nov 13 09:21:47 2012
@@ -41,6 +41,11 @@
   // Adjust stack.
   if (isInt<16>(-StackSize))
     BuildMI(MBB, MBBI, dl, TII.get(Mips::SaveRaF16)).addImm(StackSize);
+
+  if (hasFP(MF))
+    BuildMI(MBB, MBBI, dl, TII.get(Mips::MoveR3216), Mips::S0)
+      .addReg(Mips::SP);
+
 }
 
 void Mips16FrameLowering::emitEpilogue(MachineFunction &MF,
@@ -55,6 +60,10 @@
   if (!StackSize)
     return;
 
+  if (hasFP(MF))
+    BuildMI(MBB, MBBI, dl, TII.get(Mips::Move32R16), Mips::SP)
+      .addReg(Mips::S0);
+
   // Adjust stack.
   if (isInt<16>(StackSize))
     // assumes stacksize multiple of 8
@@ -106,8 +115,10 @@
 
 bool
 Mips16FrameLowering::hasReservedCallFrame(const MachineFunction &MF) const {
-  // FIXME: implement.
-  return true;
+  const MachineFrameInfo *MFI = MF.getFrameInfo();
+  // Reserve call frame if the size of the maximum call frame fits into 15-bit
+  // immediate field and there are no variable sized objects on the stack.
+  return isInt<15>(MFI->getMaxCallFrameSize()) && !MFI->hasVarSizedObjects();
 }
 
 void Mips16FrameLowering::

Modified: llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.cpp Tue Nov 13 09:21:47 2012
@@ -26,7 +26,7 @@
 
 Mips16InstrInfo::Mips16InstrInfo(MipsTargetMachine &tm)
   : MipsInstrInfo(tm, Mips::BimmX16),
-    RI(*tm.getSubtargetImpl()) {}
+    RI(*tm.getSubtargetImpl(), *this) {}
 
 const MipsRegisterInfo &Mips16InstrInfo::getRegisterInfo() const {
   return RI;
@@ -126,7 +126,7 @@
   default:
     return false;
   case Mips::RetRA16:
-    ExpandRetRA16(MBB, MI, Mips::JrRa16);
+    ExpandRetRA16(MBB, MI, Mips::JrcRa16);
     break;
   }
 
@@ -160,6 +160,22 @@
   return 0;
 }
 
+/// Adjust SP by Amount bytes.
+void Mips16InstrInfo::adjustStackPtr(unsigned SP, int64_t Amount,
+                                     MachineBasicBlock &MBB,
+                                     MachineBasicBlock::iterator I) const {
+  DebugLoc DL = I != MBB.end() ? I->getDebugLoc() : DebugLoc();
+  if (isInt<16>(Amount)) {
+    if (Amount < 0)
+      BuildMI(MBB, I, DL, get(Mips::SaveDecSpF16)). addImm(-Amount);
+    else if (Amount > 0)
+      BuildMI(MBB, I, DL, get(Mips::RestoreIncSpF16)).addImm(Amount);
+  }
+  else
+    // not implemented for large values yet
+    assert(false && "adjust stack pointer amount exceeded");
+}
+
 unsigned Mips16InstrInfo::GetAnalyzableBrOpc(unsigned Opc) const {
   return (Opc == Mips::BeqzRxImmX16   || Opc == Mips::BimmX16  ||
           Opc == Mips::BnezRxImmX16   || Opc == Mips::BteqzX16 ||

Modified: llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.h (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.h Tue Nov 13 09:21:47 2012
@@ -64,6 +64,10 @@
 
   virtual unsigned GetOppositeBranchOpc(unsigned Opc) const;
 
+  /// Adjust SP by Amount bytes.
+  void adjustStackPtr(unsigned SP, int64_t Amount, MachineBasicBlock &MBB,
+                      MachineBasicBlock::iterator I) const;
+
 private:
   virtual unsigned GetAnalyzableBrOpc(unsigned Opc) const;
 

Modified: llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.td (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16InstrInfo.td Tue Nov 13 09:21:47 2012
@@ -11,16 +11,40 @@
 //
 //===----------------------------------------------------------------------===//
 //
+//
+// Mips Address
+//
+def addr16 :
+  ComplexPattern<iPTR, 3, "SelectAddr16", [frameindex], [SDNPWantParent]>;
 
 //
 // Address operand
 def mem16 : Operand<i32> {
   let PrintMethod = "printMemOperand";
+  let MIOperandInfo = (ops CPU16Regs, simm16, CPU16Regs);
+  let EncoderMethod = "getMemEncoding";
+}
+
+def mem16_ea : Operand<i32> {
+  let PrintMethod = "printMemOperandEA";
   let MIOperandInfo = (ops CPU16Regs, simm16);
   let EncoderMethod = "getMemEncoding";
 }
 
 //
+// Compare a register and immediate and place result in CC
+// Implicit use of T8
+//
+// EXT-CCRR Instruction format
+//
+class FEXT_CCRXI16_ins<bits<5> _op, string asmstr,
+                       InstrItinClass itin>:
+  FEXT_RI16<_op, (outs CPU16Regs:$cc), (ins CPU16Regs:$rx, simm16:$imm),
+            !strconcat(asmstr, "\t$rx, $imm\n\tmove\t$cc, $$t8"), [], itin> {
+  let isCodeGenOnly=1;
+}
+
+//
 // EXT-I instruction format
 //
 class FEXT_I16_ins<bits<5> eop, string asmstr, InstrItinClass itin> :
@@ -44,6 +68,17 @@
 // Assembler formats in alphabetical order.
 // Natural and pseudos are mixed together.
 //
+// Compare two registers and place result in CC
+// Implicit use of T8
+//
+// CC-RR Instruction format
+//
+class FCCRR16_ins<bits<5> f, string asmstr, InstrItinClass itin> :
+  FRR16<f, (outs CPU16Regs:$cc), (ins CPU16Regs:$rx, CPU16Regs:$ry),
+        !strconcat(asmstr, "\t$rx, $ry\n\tmove\t$cc, $$t8"), [], itin> {
+  let isCodeGenOnly=1;
+}
+
 //
 // EXT-RI instruction format
 //
@@ -95,6 +130,16 @@
              !strconcat(asmstr, "\t$ry, $addr"), [], itin>;
 
 //
+//
+// EXT-RRI-A instruction format
+//
+
+class FEXT_RRI_A16_mem_ins<bits<1> op, string asmstr, Operand MemOpnd,
+                           InstrItinClass itin>:
+  FEXT_RRI_A16<op, (outs CPU16Regs:$ry), (ins  MemOpnd:$addr),
+               !strconcat(asmstr, "\t$ry, $addr"), [], itin>;
+
+//
 // EXT-SHIFT instruction format
 //
 class FEXT_SHIFT16_ins<bits<2> _f, string asmstr, InstrItinClass itin>:
@@ -170,6 +215,10 @@
         !strconcat(asmstr, "\t$rx, $ry"), [], itin> {
 }
 
+class FRRTR16_ins<bits<5> f, string asmstr, InstrItinClass itin> :
+  FRR16<f, (outs CPU16Regs:$rz), (ins CPU16Regs:$rx, CPU16Regs:$ry),
+        !strconcat(asmstr, "\t$rx, $ry\n\tmove\t$rz, $$t8"), [], itin> ;
+
 //
 // maybe refactor but need a $zero as a dummy first parameter
 //
@@ -177,6 +226,11 @@
   FRR16<f, (outs ), (ins CPU16Regs:$rx, CPU16Regs:$ry),
         !strconcat(asmstr, "\t$$zero, $rx, $ry"), [], itin> ;
 
+class FUnaryRR16_ins<bits<5> f, string asmstr, InstrItinClass itin> :
+  FRR16<f, (outs CPU16Regs:$rx), (ins CPU16Regs:$ry),
+        !strconcat(asmstr, "\t$rx, $ry"), [], itin> ;
+
+
 class FRR16_M_ins<bits<5> f, string asmstr,
                   InstrItinClass itin> :
   FRR16<f, (outs CPU16Regs:$rx), (ins),
@@ -196,6 +250,12 @@
   FRR16_JALRC<nd_, l_, 1, (outs), (ins), !strconcat(asmstr, "\t $$ra"),
               [], itin> ;
 
+
+class FRR16_JALRC_ins<bits<1> nd, bits<1> l, bits<1> ra,
+                      string asmstr, InstrItinClass itin>:
+  FRR16_JALRC<nd, l, ra, (outs), (ins CPU16Regs:$rx), 
+              !strconcat(asmstr, "\t $rx"), [], itin> ;
+
 //
 // RRR-type instruction format
 //
@@ -205,6 +265,95 @@
          !strconcat(asmstr, "\t$rz, $rx, $ry"), [], itin>;
 
 //
+// These Sel patterns support the generation of conditional move
+// pseudo instructions.
+//
+// The nomenclature uses the components making up the pseudo and may
+// be a bit counter intuitive when compared with the end result we seek.
+// For example using a bqez in the example directly below results in the
+// conditional move being done if the tested register is not zero.
+// I considered in easier to check by keeping the pseudo consistent with
+// it's components but it could have been done differently.
+//
+// The simplest case is when can test and operand directly and do the
+// conditional move based on a simple mips16 conditional
+//  branch instruction.
+// for example:
+// if $op == beqz or bnez:
+//
+// $op1 $rt, .+4
+// move $rd, $rs
+//
+// if $op == beqz, then if $rt != 0, then the conditional assignment
+// $rd = $rs is done.
+
+// if $op == bnez, then if $rt == 0, then the conditional assignment
+// $rd = $rs is done.
+//
+// So this pseudo class only has one operand, i.e. op
+//
+class Sel<bits<5> f1, string op, InstrItinClass itin>:
+  MipsInst16_32<(outs CPU16Regs:$rd_), (ins CPU16Regs:$rd, CPU16Regs:$rs,
+                CPU16Regs:$rt),
+                !strconcat(op, "\t$rt, .+4\n\t\n\tmove $rd, $rs"), [], itin,
+                Pseudo16> {
+  let isCodeGenOnly=1;
+  let Constraints = "$rd = $rd_";
+}
+
+//
+// The next two instruction classes allow for an operand which tests
+// two operands and returns a value in register T8 and
+//then does a conditional branch based on the value of T8
+//
+
+// op2 can be cmpi or slti/sltiu
+// op1 can bteqz or btnez
+// the operands for op2 are a register and a signed constant
+//
+// $op2 $t, $imm  ;test register t and branch conditionally
+// $op1 .+4       ;op1 is a conditional branch
+// move $rd, $rs
+//
+//
+class SeliT<bits<5> f1, string op1, bits<5> f2, string op2,
+                 InstrItinClass itin>:
+  MipsInst16_32<(outs CPU16Regs:$rd_), (ins CPU16Regs:$rd, CPU16Regs:$rs,
+                                        CPU16Regs:$rl, simm16:$imm),
+                 !strconcat(op2,
+                 !strconcat("\t$rl, $imm\n\t",
+                 !strconcat(op1, "\t.+4\n\tmove $rd, $rs"))), [], itin,
+                 Pseudo16> {
+  let isCodeGenOnly=1;
+  let Constraints = "$rd = $rd_";
+}
+
+//
+// op2 can be cmp or slt/sltu
+// op1 can be bteqz or btnez
+// the operands for op2 are two registers
+// op1 is a conditional branch
+//
+//
+// $op2 $rl, $rr  ;test registers rl,rr
+// $op1 .+4       ;op2 is a conditional branch
+// move $rd, $rs
+//
+//
+class SelT<bits<5> f1, string op1, bits<5> f2, string op2,
+           InstrItinClass itin>:
+  MipsInst16_32<(outs CPU16Regs:$rd_), (ins CPU16Regs:$rd, CPU16Regs:$rs,
+                CPU16Regs:$rl, CPU16Regs:$rr),
+                !strconcat(op2,
+                !strconcat("\t$rl, $rr\n\t",
+                !strconcat(op1, "\t.+4\n\tmove $rd, $rs"))), [], itin,
+                Pseudo16> {
+  let isCodeGenOnly=1;
+  let Constraints = "$rd = $rd_";
+}
+
+
+//
 // Some general instruction class info
 //
 //
@@ -245,6 +394,9 @@
 def AddiuRxRxImmX16: FEXT_2RI16_ins<0b01001, "addiu", IIAlu>,
   ArithLogic16Defs<0>;
 
+def AddiuRxRyOffMemX16:
+  FEXT_RRI_A16_mem_ins<0, "addiu", mem16_ea, IIAlu>;
+
 //
 
 // Format: ADDIU rx, pc, immediate MIPS16e
@@ -355,8 +507,27 @@
 // address register.
 //
 
-def JrRa16: FRR16_JALRC_RA_only_ins<0, 0, "jr", IIAlu>;
-
+def JrRa16: FRR16_JALRC_RA_only_ins<0, 0, "jr", IIAlu> {
+  let isBranch = 1;
+  let isIndirectBranch = 1;
+  let hasDelaySlot = 1;
+  let isTerminator=1;
+  let isBarrier=1;
+}
+
+def JrcRa16: FRR16_JALRC_RA_only_ins<0, 0, "jrc", IIAlu> {
+  let isBranch = 1;
+  let isIndirectBranch = 1;
+  let isTerminator=1;
+  let isBarrier=1;
+}
+
+def JrcRx16: FRR16_JALRC_ins<1, 1, 0, "jrc", IIAlu> {
+  let isBranch = 1;
+  let isIndirectBranch = 1;
+  let isTerminator=1;
+  let isBarrier=1;
+}
 //
 // Format: LB ry, offset(rx) MIPS16e
 // Purpose: Load Byte (Extended)
@@ -483,14 +654,14 @@
 // Purpose: Negate
 // To negate an integer value.
 //
-def NegRxRy16: FRR16_ins<0b11101, "neg", IIAlu>;
+def NegRxRy16: FUnaryRR16_ins<0b11101, "neg", IIAlu>;
 
 //
 // Format: NOT rx, ry MIPS16e
 // Purpose: Not
 // To complement an integer value
 //
-def NotRxRy16: FRR16_ins<0b01111, "not", IIAlu>;
+def NotRxRy16: FUnaryRR16_ins<0b01111, "not", IIAlu>;
 
 //
 // Format: OR rx, ry MIPS16e
@@ -515,7 +686,17 @@
 let ra=1, s=0,s0=1,s1=1 in
 def RestoreRaF16:
   FI8_SVRS16<0b1, (outs), (ins uimm16:$frame_size),
-             "restore \t$$ra,  $$s0, $$s1, $frame_size", [], IILoad >, MayLoad {
+             "restore\t$$ra,  $$s0, $$s1, $frame_size", [], IILoad >, MayLoad {
+  let isCodeGenOnly = 1;
+}
+
+// Use Restore to increment SP since SP is not a Mip 16 register, this
+// is an easy way to do that which does not require a register.
+//
+let ra=0, s=0,s0=0,s1=0 in
+def RestoreIncSpF16:
+  FI8_SVRS16<0b1, (outs), (ins uimm16:$frame_size),
+             "restore\t$frame_size", [], IILoad >, MayLoad {
   let isCodeGenOnly = 1;
 }
 
@@ -529,7 +710,18 @@
 let ra=1, s=1,s0=1,s1=1 in
 def SaveRaF16:
   FI8_SVRS16<0b1, (outs), (ins uimm16:$frame_size),
-             "save \t$$ra, $$s0, $$s1, $frame_size", [], IIStore >, MayStore {
+             "save\t$$ra, $$s0, $$s1, $frame_size", [], IIStore >, MayStore {
+  let isCodeGenOnly = 1;
+}
+
+//
+// Use Save to decrement the SP by a constant since SP is not
+// a Mips16 register.
+//
+let ra=0, s=0,s0=0,s1=0 in
+def SaveDecSpF16:
+  FI8_SVRS16<0b1, (outs), (ins uimm16:$frame_size),
+             "save\t$frame_size", [], IIStore >, MayStore {
   let isCodeGenOnly = 1;
 }
 //
@@ -541,6 +733,120 @@
   FEXT_RRI16_mem2_ins<0b11000, "sb", mem16, IIStore>, MayStore;
 
 //
+// The Sel(T) instructions are pseudos
+// T means that they use T8 implicitly.
+//
+//
+// Format: SelBeqZ rd, rs, rt
+// Purpose: if rt==0, do nothing
+//          else rs = rt
+//
+def SelBeqZ: Sel<0b00100, "beqz", IIAlu>;
+
+//
+// Format:  SelTBteqZCmp rd, rs, rl, rr
+// Purpose: b = Cmp rl, rr.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZCmp: SelT<0b000, "bteqz", 0b01010, "cmp", IIAlu>;
+
+//
+// Format:  SelTBteqZCmpi rd, rs, rl, rr
+// Purpose: b = Cmpi rl, imm.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZCmpi: SeliT<0b000, "bteqz", 0b01110, "cmpi", IIAlu>;
+
+//
+// Format:  SelTBteqZSlt rd, rs, rl, rr
+// Purpose: b = Slt rl, rr.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZSlt: SelT<0b000, "bteqz", 0b00010, "slt", IIAlu>;
+
+//
+// Format:  SelTBteqZSlti rd, rs, rl, rr
+// Purpose: b = Slti rl, imm.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZSlti: SeliT<0b000, "bteqz", 0b01010, "slti", IIAlu>;
+
+//
+// Format:  SelTBteqZSltu rd, rs, rl, rr
+// Purpose: b = Sltu rl, rr.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZSltu: SelT<0b000, "bteqz", 0b00011, "sltu", IIAlu>;
+
+//
+// Format:  SelTBteqZSltiu rd, rs, rl, rr
+// Purpose: b = Sltiu rl, imm.
+//          If b==0 then do nothing.
+//          if b!=0 then rd = rs
+//
+def SelTBteqZSltiu: SeliT<0b000, "bteqz", 0b01011, "sltiu", IIAlu>;
+
+//
+// Format: SelBnez rd, rs, rt
+// Purpose: if rt!=0, do nothing
+//          else rs = rt
+//
+def SelBneZ: Sel<0b00101, "bnez", IIAlu>;
+
+//
+// Format:  SelTBtneZCmp rd, rs, rl, rr
+// Purpose: b = Cmp rl, rr.
+//          If b!=0 then do nothing.
+//          if b0=0 then rd = rs
+//
+def SelTBtneZCmp: SelT<0b001, "btnez", 0b01010, "cmp", IIAlu>;
+
+//
+// Format:  SelTBtnezCmpi rd, rs, rl, rr
+// Purpose: b = Cmpi rl, imm.
+//          If b!=0 then do nothing.
+//          if b==0 then rd = rs
+//
+def SelTBtneZCmpi: SeliT<0b000, "btnez", 0b01110, "cmpi", IIAlu>;
+
+//
+// Format:  SelTBtneZSlt rd, rs, rl, rr
+// Purpose: b = Slt rl, rr.
+//          If b!=0 then do nothing.
+//          if b==0 then rd = rs
+//
+def SelTBtneZSlt: SelT<0b001, "btnez", 0b00010, "slt", IIAlu>;
+
+//
+// Format:  SelTBtneZSlti rd, rs, rl, rr
+// Purpose: b = Slti rl, imm.
+//          If b!=0 then do nothing.
+//          if b==0 then rd = rs
+//
+def SelTBtneZSlti: SeliT<0b001, "btnez", 0b01010, "slti", IIAlu>;
+
+//
+// Format:  SelTBtneZSltu rd, rs, rl, rr
+// Purpose: b = Sltu rl, rr.
+//          If b!=0 then do nothing.
+//          if b==0 then rd = rs
+//
+def SelTBtneZSltu: SelT<0b001, "btnez", 0b00011, "sltu", IIAlu>;
+
+//
+// Format:  SelTBtneZSltiu rd, rs, rl, rr
+// Purpose: b = Slti rl, imm.
+//          If b!=0 then do nothing.
+//          if b==0 then rd = rs
+//
+def SelTBtneZSltiu: SeliT<0b001, "btnez", 0b01011, "sltiu", IIAlu>;
+//
+//
 // Format: SH ry, offset(rx) MIPS16e
 // Purpose: Store Halfword (Extended)
 // To store a halfword to memory.
@@ -562,7 +868,39 @@
 //
 def SllvRxRy16 : FRxRxRy16_ins<0b00100, "sllv", IIAlu>;
 
+//
+// Format: SLTI rx, immediate MIPS16e
+// Purpose: Set on Less Than Immediate (Extended)
+// To record the result of a less-than comparison with a constant.
+//
+def SltiCCRxImmX16: FEXT_CCRXI16_ins<0b01010, "slti", IIAlu>;
+
+//
+// Format: SLTIU rx, immediate MIPS16e
+// Purpose: Set on Less Than Immediate Unsigned (Extended)
+// To record the result of a less-than comparison with a constant.
+//
+def SltiuCCRxImmX16: FEXT_CCRXI16_ins<0b01011, "sltiu", IIAlu>;
+
+//
+// Format: SLT rx, ry MIPS16e
+// Purpose: Set on Less Than
+// To record the result of a less-than comparison.
+//
+def SltRxRy16: FRR16_ins<0b00010, "slt", IIAlu>;
+
+def SltCCRxRy16: FCCRR16_ins<0b00010, "slt", IIAlu>;
+
+// Format: SLTU rx, ry MIPS16e
+// Purpose: Set on Less Than Unsigned
+// To record the result of an unsigned less-than comparison.
+//
+def SltuRxRyRz16: FRRTR16_ins<0b00011, "sltu", IIAlu> {
+  let isCodeGenOnly=1;
+}
+
 
+def SltuCCRxRy16: FCCRR16_ins<0b00011, "sltu", IIAlu>;
 //
 // Format: SRAV ry, rx MIPS16e
 // Purpose: Shift Word Right Arithmetic Variable
@@ -672,7 +1010,7 @@
 def: shift_rotate_reg16_pat<srl, SrlvRxRy16>;
 
 class LoadM16_pat<PatFrag OpNode, Instruction I> :
-  Mips16Pat<(OpNode addr:$addr), (I addr:$addr)>;
+  Mips16Pat<(OpNode addr16:$addr), (I addr16:$addr)>;
 
 def: LoadM16_pat<sextloadi8, LbRxRyOffMemX16>;
 def: LoadM16_pat<zextloadi8, LbuRxRyOffMemX16>;
@@ -681,7 +1019,8 @@
 def: LoadM16_pat<load, LwRxRyOffMemX16>;
 
 class StoreM16_pat<PatFrag OpNode, Instruction I> :
-  Mips16Pat<(OpNode CPU16Regs:$r, addr:$addr), (I CPU16Regs:$r, addr:$addr)>;
+  Mips16Pat<(OpNode CPU16Regs:$r, addr16:$addr),
+            (I CPU16Regs:$r, addr16:$addr)>;
 
 def: StoreM16_pat<truncstorei8, SbRxRyOffMemX16>;
 def: StoreM16_pat<truncstorei16, ShRxRyOffMemX16>;
@@ -693,11 +1032,17 @@
     let Predicates = [RelocPIC, InMips16Mode];
   }
 
+// Indirect branch
+def: Mips16Pat<
+  (brind CPU16Regs:$rs), 
+  (JrcRx16 CPU16Regs:$rs)>;  
+
+
 // Jump and Link (Call)
-let isCall=1, hasDelaySlot=1 in
+let isCall=1, hasDelaySlot=0 in
 def JumpLinkReg16:
   FRR16_JALRC<0, 0, 0, (outs), (ins CPU16Regs:$rs),
-              "jalr \t$rs", [(MipsJmpLink CPU16Regs:$rs)], IIBranch>;
+              "jalrc \t$rs", [(MipsJmpLink CPU16Regs:$rs)], IIBranch>;
 
 // Mips16 pseudos
 let isReturn=1, isTerminator=1, hasDelaySlot=1, isBarrier=1, hasCtrlDep=1,
@@ -705,6 +1050,34 @@
 def RetRA16 : MipsPseudo16<(outs), (ins), "", [(MipsRet)]>;
 
 
+// setcc patterns
+
+class SetCC_R16<PatFrag cond_op, Instruction I>:
+  Mips16Pat<(cond_op CPU16Regs:$rx, CPU16Regs:$ry),
+            (I CPU16Regs:$rx, CPU16Regs:$ry)>;
+
+class SetCC_I16<PatFrag cond_op, PatLeaf imm_type, Instruction I>:
+  Mips16Pat<(cond_op CPU16Regs:$rx, imm_type:$imm16),
+            (I CPU16Regs:$rx, imm_type:$imm16)>;
+
+
+def: Mips16Pat<(i32  addr16:$addr),
+               (AddiuRxRyOffMemX16  addr16:$addr)>;
+
+
+// Large (>16 bit) immediate loads
+def : Mips16Pat<(i32 imm:$imm),
+                (OrRxRxRy16 (SllX16 (LiRxImmX16 (HI16 imm:$imm)), 16),
+                (LiRxImmX16 (LO16 imm:$imm)))>;
+
+// Carry MipsPatterns
+def : Mips16Pat<(subc CPU16Regs:$lhs, CPU16Regs:$rhs),
+                (SubuRxRyRz16 CPU16Regs:$lhs, CPU16Regs:$rhs)>;
+def : Mips16Pat<(addc CPU16Regs:$lhs, CPU16Regs:$rhs),
+                (AdduRxRyRz16 CPU16Regs:$lhs, CPU16Regs:$rhs)>;
+def : Mips16Pat<(addc  CPU16Regs:$src, immSExt16:$imm),
+                (AddiuRxRxImmX16 CPU16Regs:$src, imm:$imm)>;
+
 //
 // Some branch conditional patterns are not generated by llvm at this time.
 // Some are for seemingly arbitrary reasons not used: i.e. with signed number
@@ -757,10 +1130,10 @@
 
 //
 // never called because compiler transforms a >= k to a > (k-1)
-//def: Mips16Pat
-//  <(brcond (i32 (setge CPU16Regs:$rx, immSExt16:$imm)), bb:$imm16),
-//   (BteqzT8SltiX16 CPU16Regs:$rx, immSExt16:$imm,  bb:$imm16)
-//  >;
+def: Mips16Pat
+  <(brcond (i32 (setge CPU16Regs:$rx, immSExt16:$imm)), bb:$imm16),
+   (BteqzT8SltiX16 CPU16Regs:$rx, immSExt16:$imm,  bb:$imm16)
+  >;
 
 //
 // bcond-setlt
@@ -858,6 +1231,312 @@
   <(MipsDivRemU CPU16Regs:$rx, CPU16Regs:$ry),
    (DivuRxRy16 CPU16Regs:$rx, CPU16Regs:$ry)>;
 
+//  signed a,b
+//  x = (a>=b)?x:y
+//
+//  if !(a < b) x = y
+//
+def : Mips16Pat<(select (i32 (setge CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBteqZSlt CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$a, CPU16Regs:$b)>;
+
+//  signed a,b
+//  x = (a>b)?x:y
+//
+//  if  (b < a) x = y
+//
+def : Mips16Pat<(select (i32 (setgt CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBtneZSlt CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+// unsigned a,b
+// x = (a>=b)?x:y
+//
+// if !(a < b) x = y;
+//
+def : Mips16Pat<
+  (select (i32 (setuge CPU16Regs:$a, CPU16Regs:$b)),
+   CPU16Regs:$x, CPU16Regs:$y),
+  (SelTBteqZSltu CPU16Regs:$x, CPU16Regs:$y,
+   CPU16Regs:$a, CPU16Regs:$b)>;
+
+//  unsigned a,b
+//  x = (a>b)?x:y
+//
+//  if (b < a) x = y
+//
+def : Mips16Pat<(select (i32 (setugt CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBtneZSltu CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+// signed
+// x = (a >= k)?x:y
+// due to an llvm optimization, i don't think that this will ever
+// be used. This is transformed into x = (a > k-1)?x:y
+//
+//
+
+//def : Mips16Pat<
+//  (select (i32 (setge CPU16Regs:$lhs, immSExt16:$rhs)),
+//   CPU16Regs:$T, CPU16Regs:$F),
+//  (SelTBteqZSlti CPU16Regs:$T, CPU16Regs:$F,
+//   CPU16Regs:$lhs, immSExt16:$rhs)>;
+
+//def : Mips16Pat<
+//  (select (i32 (setuge CPU16Regs:$lhs, immSExt16:$rhs)),
+//   CPU16Regs:$T, CPU16Regs:$F),
+//  (SelTBteqZSltiu CPU16Regs:$T, CPU16Regs:$F,
+//   CPU16Regs:$lhs, immSExt16:$rhs)>;
+
+// signed
+// x = (a < k)?x:y
+//
+// if !(a < k) x = y;
+//
+def : Mips16Pat<
+  (select (i32 (setlt CPU16Regs:$a, immSExt16:$b)),
+   CPU16Regs:$x, CPU16Regs:$y),
+  (SelTBtneZSlti CPU16Regs:$x, CPU16Regs:$y,
+   CPU16Regs:$a, immSExt16:$b)>;
+
+
+//
+//
+// signed
+// x = (a <= b)? x : y
+//
+// if  (b < a) x = y
+//
+def : Mips16Pat<(select (i32 (setle CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBteqZSlt CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+//
+// unnsigned
+// x = (a <= b)? x : y
+//
+// if  (b < a) x = y
+//
+def : Mips16Pat<(select (i32 (setule CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBteqZSltu CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+//
+// signed/unsigned
+// x = (a == b)? x : y
+//
+// if (a != b) x = y
+//
+def : Mips16Pat<(select (i32 (seteq CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBteqZCmp CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+//
+// signed/unsigned
+// x = (a == 0)? x : y
+//
+// if (a != 0) x = y
+//
+def : Mips16Pat<(select (i32 (seteq CPU16Regs:$a, 0)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelBeqZ CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$a)>;
+
+
+//
+// signed/unsigned
+// x = (a == k)? x : y
+//
+// if (a != k) x = y
+//
+def : Mips16Pat<(select (i32 (seteq CPU16Regs:$a, immZExt16:$k)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBteqZCmpi CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$a, immZExt16:$k)>;
+
+
+//
+// signed/unsigned
+// x = (a != b)? x : y
+//
+// if (a == b) x = y
+//
+//
+def : Mips16Pat<(select (i32 (setne CPU16Regs:$a, CPU16Regs:$b)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBtneZCmp CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$b, CPU16Regs:$a)>;
+
+//
+// signed/unsigned
+// x = (a != 0)? x : y
+//
+// if (a == 0) x = y
+//
+def : Mips16Pat<(select (i32 (setne CPU16Regs:$a, 0)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelBneZ CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$a)>;
+
+// signed/unsigned
+// x = (a)? x : y
+//
+// if (!a) x = y
+//
+def : Mips16Pat<(select  CPU16Regs:$a,
+                 CPU16Regs:$x, CPU16Regs:$y),
+      (SelBneZ CPU16Regs:$x, CPU16Regs:$y,
+       CPU16Regs:$a)>;
+
+
+//
+// signed/unsigned
+// x = (a != k)? x : y
+//
+// if (a == k) x = y
+//
+def : Mips16Pat<(select (i32 (setne CPU16Regs:$a, immZExt16:$k)),
+                 CPU16Regs:$x, CPU16Regs:$y),
+                (SelTBtneZCmpi CPU16Regs:$x, CPU16Regs:$y,
+                 CPU16Regs:$a, immZExt16:$k)>;
+
+//
+// When writing C code to test setxx these patterns,
+// some will be transformed into
+// other things. So we test using C code but using -O3 and -O0
+//
+// seteq
+//
+def : Mips16Pat
+  <(seteq CPU16Regs:$lhs,CPU16Regs:$rhs),
+   (SltiuCCRxImmX16 (XorRxRxRy16 CPU16Regs:$lhs, CPU16Regs:$rhs), 1)>;
+
+def : Mips16Pat
+  <(seteq CPU16Regs:$lhs, 0),
+   (SltiuCCRxImmX16 CPU16Regs:$lhs, 1)>;
+
+
+//
+// setge
+//
+
+def: Mips16Pat
+  <(setge CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (XorRxRxRy16 (SltCCRxRy16 CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (LiRxImmX16 1))>;
+
+//
+// For constants, llvm transforms this to:
+// x > (k -1) and then reverses the operands to use setlt. So this pattern
+// is not used now by the compiler. (Presumably checking that k-1 does not
+// overflow). The compiler never uses this at a the current time, due to
+// other optimizations.
+//
+//def: Mips16Pat
+//  <(setge CPU16Regs:$lhs, immSExt16:$rhs),
+//   (XorRxRxRy16 (SltiCCRxImmX16 CPU16Regs:$lhs, immSExt16:$rhs),
+//   (LiRxImmX16 1))>;
+
+// This catches the x >= -32768 case by transforming it to  x > -32769
+//
+def: Mips16Pat
+  <(setgt CPU16Regs:$lhs, -32769),
+   (XorRxRxRy16 (SltiCCRxImmX16 CPU16Regs:$lhs, -32768),
+   (LiRxImmX16 1))>;
+
+//
+// setgt
+//
+//
+
+def: Mips16Pat
+  <(setgt CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (SltCCRxRy16 CPU16Regs:$rhs, CPU16Regs:$lhs)>;
+
+//
+// setle
+//
+def: Mips16Pat
+  <(setle CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (XorRxRxRy16 (SltCCRxRy16 CPU16Regs:$rhs, CPU16Regs:$lhs), (LiRxImmX16 1))>;
+
+//
+// setlt
+//
+def: SetCC_R16<setlt, SltCCRxRy16>;
+
+def: SetCC_I16<setlt, immSExt16, SltiCCRxImmX16>;
+
+//
+// setne
+//
+def : Mips16Pat
+  <(setne CPU16Regs:$lhs,CPU16Regs:$rhs),
+   (SltuCCRxRy16 (LiRxImmX16 0),
+   (XorRxRxRy16 CPU16Regs:$lhs, CPU16Regs:$rhs))>;
+
+
+//
+// setuge
+//
+def: Mips16Pat
+  <(setuge CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (XorRxRxRy16 (SltuCCRxRy16 CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (LiRxImmX16 1))>;
+
+// this pattern will never be used because the compiler will transform
+// x >= k to x > (k - 1) and then use SLT
+//
+//def: Mips16Pat
+//  <(setuge CPU16Regs:$lhs, immZExt16:$rhs),
+//   (XorRxRxRy16 (SltiuCCRxImmX16 CPU16Regs:$lhs, immZExt16:$rhs),
+//   (LiRxImmX16 1))>;
+
+//
+// setugt
+//
+def: Mips16Pat
+  <(setugt CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (SltuCCRxRy16 CPU16Regs:$rhs, CPU16Regs:$lhs)>;
+
+//
+// setule
+//
+def: Mips16Pat
+  <(setule CPU16Regs:$lhs, CPU16Regs:$rhs),
+   (XorRxRxRy16 (SltuCCRxRy16 CPU16Regs:$rhs, CPU16Regs:$lhs), (LiRxImmX16 1))>;
+
+//
+// setult
+//
+def: SetCC_R16<setult, SltuCCRxRy16>;
+
+def: SetCC_I16<setult, immSExt16, SltiuCCRxImmX16>;
 
 def: Mips16Pat<(add CPU16Regs:$hi, (MipsLo tglobaladdr:$lo)),
                (AddiuRxRxImmX16 CPU16Regs:$hi, tglobaladdr:$lo)>;
+
+// hi/lo relocs
+
+def : Mips16Pat<(MipsHi tglobaltlsaddr:$in), 
+                (SllX16 (LiRxImmX16 tglobaltlsaddr:$in), 16)>;
+
+// wrapper_pic
+class Wrapper16Pat<SDNode node, Instruction ADDiuOp, RegisterClass RC>:
+  Mips16Pat<(MipsWrapper RC:$gp, node:$in),
+            (ADDiuOp RC:$gp, node:$in)>;
+
+
+def : Wrapper16Pat<tglobaladdr, AddiuRxRxImmX16, CPU16Regs>;
+def : Wrapper16Pat<tglobaltlsaddr, AddiuRxRxImmX16, CPU16Regs>;
+
+def : Mips16Pat<(i32 (extloadi8   addr16:$src)),
+                (LbuRxRyOffMemX16  addr16:$src)>;
+def : Mips16Pat<(i32 (extloadi16  addr16:$src)),
+                (LhuRxRyOffMemX16  addr16:$src)>;
\ No newline at end of file

Modified: llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.cpp Tue Nov 13 09:21:47 2012
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "Mips16RegisterInfo.h"
+#include "Mips16InstrInfo.h"
 #include "Mips.h"
 #include "MipsAnalyzeImmediate.h"
 #include "MipsInstrInfo.h"
@@ -38,15 +39,28 @@
 
 using namespace llvm;
 
-Mips16RegisterInfo::Mips16RegisterInfo(const MipsSubtarget &ST)
-  : MipsRegisterInfo(ST) {}
+Mips16RegisterInfo::Mips16RegisterInfo(const MipsSubtarget &ST,
+    const Mips16InstrInfo &I)
+  : MipsRegisterInfo(ST), TII(I) {}
 
 // This function eliminate ADJCALLSTACKDOWN,
 // ADJCALLSTACKUP pseudo instructions
 void Mips16RegisterInfo::
 eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
                               MachineBasicBlock::iterator I) const {
-  // Simply discard ADJCALLSTACKDOWN, ADJCALLSTACKUP instructions.
+  const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
+
+  if (!TFI->hasReservedCallFrame(MF)) {
+    int64_t Amount = I->getOperand(0).getImm();
+
+    if (I->getOpcode() == Mips::ADJCALLSTACKDOWN)
+      Amount = -Amount;
+
+    const Mips16InstrInfo *II = static_cast<const Mips16InstrInfo*>(&TII);
+
+    II->adjustStackPtr(Mips::SP, Amount, MBB, I);
+  }
+
   MBB.erase(I);
 }
 
@@ -54,51 +68,60 @@
                                      unsigned OpNo, int FrameIndex,
                                      uint64_t StackSize,
                                      int64_t SPOffset) const {
-      MachineInstr &MI = *II;
-      MachineFunction &MF = *MI.getParent()->getParent();
-      MachineFrameInfo *MFI = MF.getFrameInfo();
-
-      const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
-      int MinCSFI = 0;
-      int MaxCSFI = -1;
-
-      if (CSI.size()) {
-        MinCSFI = CSI[0].getFrameIdx();
-        MaxCSFI = CSI[CSI.size() - 1].getFrameIdx();
-      }
-
-      // The following stack frame objects are always
-      // referenced relative to $sp:
-      //  1. Outgoing arguments.
-      //  2. Pointer to dynamically allocated stack space.
-      //  3. Locations for callee-saved registers.
-      // Everything else is referenced relative to whatever register
-      // getFrameRegister() returns.
-      unsigned FrameReg;
-
-      if (FrameIndex >= MinCSFI && FrameIndex <= MaxCSFI)
-        FrameReg = Subtarget.isABI_N64() ? Mips::SP_64 : Mips::SP;
+  MachineInstr &MI = *II;
+  MachineFunction &MF = *MI.getParent()->getParent();
+  MachineFrameInfo *MFI = MF.getFrameInfo();
+
+  const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
+  int MinCSFI = 0;
+  int MaxCSFI = -1;
+
+  if (CSI.size()) {
+    MinCSFI = CSI[0].getFrameIdx();
+    MaxCSFI = CSI[CSI.size() - 1].getFrameIdx();
+  }
+
+  // The following stack frame objects are always
+  // referenced relative to $sp:
+  //  1. Outgoing arguments.
+  //  2. Pointer to dynamically allocated stack space.
+  //  3. Locations for callee-saved registers.
+  // Everything else is referenced relative to whatever register
+  // getFrameRegister() returns.
+  unsigned FrameReg;
+
+  if (FrameIndex >= MinCSFI && FrameIndex <= MaxCSFI)
+    FrameReg = Mips::SP;
+  else {
+    const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
+    if (TFI->hasFP(MF)) {
+      FrameReg = Mips::S0;
+    }
+    else {
+      if ((MI.getNumOperands()> OpNo+2) && MI.getOperand(OpNo+2).isReg())
+        FrameReg = MI.getOperand(OpNo+2).getReg();
       else
-        FrameReg = getFrameRegister(MF);
-
-      // Calculate final offset.
-      // - There is no need to change the offset if the frame object
-      //   is one of the
-      //   following: an outgoing argument, pointer to a dynamically allocated
-      //   stack space or a $gp restore location,
-      // - If the frame object is any of the following,
-      //   its offset must be adjusted
-      //   by adding the size of the stack:
-      //   incoming argument, callee-saved register location or local variable.
-      int64_t Offset;
+        FrameReg = Mips::SP;
+    }
+  }
+  // Calculate final offset.
+  // - There is no need to change the offset if the frame object
+  //   is one of the
+  //   following: an outgoing argument, pointer to a dynamically allocated
+  //   stack space or a $gp restore location,
+  // - If the frame object is any of the following,
+  //   its offset must be adjusted
+  //   by adding the size of the stack:
+  //   incoming argument, callee-saved register location or local variable.
+  int64_t Offset;
+  Offset = SPOffset + (int64_t)StackSize;
+  Offset += MI.getOperand(OpNo + 1).getImm();
 
-      Offset = SPOffset + (int64_t)StackSize;
-      Offset += MI.getOperand(OpNo + 1).getImm();
 
-      DEBUG(errs() << "Offset     : " << Offset << "\n" << "<--------->\n");
+  DEBUG(errs() << "Offset     : " << Offset << "\n" << "<--------->\n");
 
-      MI.getOperand(OpNo).ChangeToRegister(FrameReg, false);
-      MI.getOperand(OpNo + 1).ChangeToImmediate(Offset);
+  MI.getOperand(OpNo).ChangeToRegister(FrameReg, false);
+  MI.getOperand(OpNo + 1).ChangeToImmediate(Offset);
 
 
 }

Modified: llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.h (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips16RegisterInfo.h Tue Nov 13 09:21:47 2012
@@ -20,8 +20,9 @@
 class Mips16InstrInfo;
 
 class Mips16RegisterInfo : public MipsRegisterInfo {
+  const Mips16InstrInfo &TII;
 public:
-  Mips16RegisterInfo(const MipsSubtarget &Subtarget);
+  Mips16RegisterInfo(const MipsSubtarget &Subtarget, const Mips16InstrInfo &TII);
 
   void eliminateCallFramePseudoInstr(MachineFunction &MF,
                                      MachineBasicBlock &MBB,

Modified: llvm/branches/R600/lib/Target/Mips/Mips64InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/Mips64InstrInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/Mips64InstrInfo.td (original)
+++ llvm/branches/R600/lib/Target/Mips/Mips64InstrInfo.td Tue Nov 13 09:21:47 2012
@@ -86,7 +86,7 @@
 def DADDi    : ArithOverflowI<0x18, "daddi", add, simm16_64, immSExt16,
                            CPU64Regs>;
 def DADDiu   : ArithLogicI<0x19, "daddiu", add, simm16_64, immSExt16,
-                           CPU64Regs>;
+                           CPU64Regs>, IsAsCheapAsAMove;
 def DANDi    : ArithLogicI<0x0c, "andi", and, uimm16_64, immZExt16, CPU64Regs>;
 def SLTi64   : SetCC_I<0x0a, "slti", setlt, simm16_64, immSExt16, CPU64Regs>;
 def SLTiu64  : SetCC_I<0x0b, "sltiu", setult, simm16_64, immSExt16, CPU64Regs>;

Modified: llvm/branches/R600/lib/Target/Mips/MipsAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsAsmPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsAsmPrinter.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsAsmPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -239,8 +239,7 @@
 
     OutStreamer.EmitRawText(StringRef("\t.set\tnoreorder"));
     OutStreamer.EmitRawText(StringRef("\t.set\tnomacro"));
-    if (MipsFI->getEmitNOAT())
-      OutStreamer.EmitRawText(StringRef("\t.set\tnoat"));
+    OutStreamer.EmitRawText(StringRef("\t.set\tnoat"));
   }
 }
 
@@ -251,9 +250,7 @@
   // always be at the function end, and we can't emit and
   // break with BB logic.
   if (OutStreamer.hasRawTextSupport()) {
-    if (MipsFI->getEmitNOAT())
-      OutStreamer.EmitRawText(StringRef("\t.set\tat"));
-
+    OutStreamer.EmitRawText(StringRef("\t.set\tat"));
     OutStreamer.EmitRawText(StringRef("\t.set\tmacro"));
     OutStreamer.EmitRawText(StringRef("\t.set\treorder"));
     OutStreamer.EmitRawText("\t.end\t" + Twine(CurrentFnSym->getName()));

Modified: llvm/branches/R600/lib/Target/Mips/MipsCallingConv.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsCallingConv.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsCallingConv.td (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsCallingConv.td Tue Nov 13 09:21:47 2012
@@ -35,9 +35,6 @@
 //===----------------------------------------------------------------------===//
 
 def CC_MipsN : CallingConv<[
-   // Handles byval parameters.
-  CCIfByVal<CCCustom<"CC_Mips64Byval">>,
-
   // Promote i8/i16 arguments to i32.
   CCIfType<[i8, i16], CCPromoteToType<i32>>,
 
@@ -72,9 +69,6 @@
 // N32/64 variable arguments.
 // All arguments are passed in integer registers.
 def CC_MipsN_VarArg : CallingConv<[
-   // Handles byval parameters.
-  CCIfByVal<CCCustom<"CC_Mips64Byval">>,
-
   // Promote i8/i16 arguments to i32.
   CCIfType<[i8, i16], CCPromoteToType<i32>>,
 
@@ -211,12 +205,6 @@
 // Mips Calling Convention Dispatch
 //===----------------------------------------------------------------------===//
 
-def CC_Mips : CallingConv<[
-  CCIfSubtarget<"isABI_EABI()", CCDelegateTo<CC_MipsEABI>>,
-  CCIfSubtarget<"isABI_N32()", CCDelegateTo<CC_MipsN>>,
-  CCIfSubtarget<"isABI_N64()", CCDelegateTo<CC_MipsN>>
-]>;
-
 def RetCC_Mips : CallingConv<[
   CCIfSubtarget<"isABI_EABI()", CCDelegateTo<RetCC_MipsEABI>>,
   CCIfSubtarget<"isABI_N32()", CCDelegateTo<RetCC_MipsN>>,

Removed: llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.cpp (removed)
@@ -1,92 +0,0 @@
-//===-- MipsELFWriterInfo.cpp - ELF Writer Info for the Mips backend ------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the Mips backend.
-//
-//===----------------------------------------------------------------------===//
-
-#include "MipsELFWriterInfo.h"
-#include "MipsRelocations.h"
-#include "llvm/Function.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/DataLayout.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Support/ELF.h"
-
-using namespace llvm;
-
-//===----------------------------------------------------------------------===//
-//  Implementation of the MipsELFWriterInfo class
-//===----------------------------------------------------------------------===//
-
-MipsELFWriterInfo::MipsELFWriterInfo(bool is64Bit_, bool isLittleEndian_)
-  : TargetELFWriterInfo(is64Bit_, isLittleEndian_) {
-  EMachine = EM_MIPS;
-}
-
-MipsELFWriterInfo::~MipsELFWriterInfo() {}
-
-unsigned MipsELFWriterInfo::getRelocationType(unsigned MachineRelTy) const {
-  switch(MachineRelTy) {
-  case Mips::reloc_mips_pc16:
-    return ELF::R_MIPS_GOT16;
-  case Mips::reloc_mips_hi:
-    return ELF::R_MIPS_HI16;
-  case Mips::reloc_mips_lo:
-    return ELF::R_MIPS_LO16;
-  case Mips::reloc_mips_26:
-    return ELF::R_MIPS_26;
-  default:
-    llvm_unreachable("unknown Mips machine relocation type");
-  }
-}
-
-long int MipsELFWriterInfo::getDefaultAddendForRelTy(unsigned RelTy,
-                                                     long int Modifier) const {
-  switch(RelTy) {
-  case ELF::R_MIPS_26: return Modifier;
-  default:
-    llvm_unreachable("unknown Mips relocation type");
-  }
-}
-
-unsigned MipsELFWriterInfo::getRelocationTySize(unsigned RelTy) const {
-  switch(RelTy) {
-  case ELF::R_MIPS_GOT16:
-  case ELF::R_MIPS_26:
-      return 32;
-  default:
-    llvm_unreachable("unknown Mips relocation type");
-  }
-}
-
-bool MipsELFWriterInfo::isPCRelativeRel(unsigned RelTy) const {
-  switch(RelTy) {
-  case ELF::R_MIPS_GOT16:
-      return true;
-  case ELF::R_MIPS_26:
-      return false;
-  default:
-    llvm_unreachable("unknown Mips relocation type");
-  }
-}
-
-unsigned MipsELFWriterInfo::getAbsoluteLabelMachineRelTy() const {
-  return Mips::reloc_mips_26;
-}
-
-long int MipsELFWriterInfo::computeRelocation(unsigned SymOffset,
-                                              unsigned RelOffset,
-                                              unsigned RelTy) const {
-
-  if (RelTy == ELF::R_MIPS_GOT16)
-    return SymOffset - (RelOffset + 4);
-
-  llvm_unreachable("computeRelocation unknown for this relocation type");
-}

Removed: llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.h?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsELFWriterInfo.h (removed)
@@ -1,59 +0,0 @@
-//===-- MipsELFWriterInfo.h - ELF Writer Info for Mips ------------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements ELF writer information for the Mips backend.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef MIPS_ELF_WRITER_INFO_H
-#define MIPS_ELF_WRITER_INFO_H
-
-#include "llvm/Target/TargetELFWriterInfo.h"
-
-namespace llvm {
-
-  class MipsELFWriterInfo : public TargetELFWriterInfo {
-
-  public:
-    MipsELFWriterInfo(bool is64Bit_, bool isLittleEndian_);
-    virtual ~MipsELFWriterInfo();
-
-    /// getRelocationType - Returns the target specific ELF Relocation type.
-    /// 'MachineRelTy' contains the object code independent relocation type
-    virtual unsigned getRelocationType(unsigned MachineRelTy) const;
-
-    /// hasRelocationAddend - True if the target uses an addend in the
-    /// ELF relocation entry.
-    virtual bool hasRelocationAddend() const { return is64Bit ? true : false; }
-
-    /// getDefaultAddendForRelTy - Gets the default addend value for a
-    /// relocation entry based on the target ELF relocation type.
-    virtual long int getDefaultAddendForRelTy(unsigned RelTy,
-                                              long int Modifier = 0) const;
-
-    /// getRelTySize - Returns the size of relocatable field in bits
-    virtual unsigned getRelocationTySize(unsigned RelTy) const;
-
-    /// isPCRelativeRel - True if the relocation type is pc relative
-    virtual bool isPCRelativeRel(unsigned RelTy) const;
-
-    /// getJumpTableRelocationTy - Returns the machine relocation type used
-    /// to reference a jumptable.
-    virtual unsigned getAbsoluteLabelMachineRelTy() const;
-
-    /// computeRelocation - Some relocatable fields could be relocated
-    /// directly, avoiding the relocation symbol emission, compute the
-    /// final relocation value for this symbol.
-    virtual long int computeRelocation(unsigned SymOffset, unsigned RelOffset,
-                                       unsigned RelTy) const;
-  };
-
-} // end llvm namespace
-
-#endif // MIPS_ELF_WRITER_INFO_H

Modified: llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.cpp Tue Nov 13 09:21:47 2012
@@ -98,3 +98,37 @@
   return MF.getTarget().Options.DisableFramePointerElim(MF) ||
       MFI->hasVarSizedObjects() || MFI->isFrameAddressTaken();
 }
+
+uint64_t MipsFrameLowering::estimateStackSize(const MachineFunction &MF) const {
+  const MachineFrameInfo *MFI = MF.getFrameInfo();
+  const TargetRegisterInfo &TRI = *MF.getTarget().getRegisterInfo();
+
+  int64_t Offset = 0;
+
+  // Iterate over fixed sized objects.
+  for (int I = MFI->getObjectIndexBegin(); I != 0; ++I)
+    Offset = std::max(Offset, -MFI->getObjectOffset(I));
+
+  // Conservatively assume all callee-saved registers will be saved.
+  for (const uint16_t *R = TRI.getCalleeSavedRegs(&MF); *R; ++R) {
+    unsigned Size = TRI.getMinimalPhysRegClass(*R)->getSize();
+    Offset = RoundUpToAlignment(Offset + Size, Size);
+  }
+
+  unsigned MaxAlign = MFI->getMaxAlignment();
+
+  // Check that MaxAlign is not zero if there is a stack object that is not a
+  // callee-saved spill.
+  assert(!MFI->getObjectIndexEnd() || MaxAlign);
+
+  // Iterate over other objects.
+  for (unsigned I = 0, E = MFI->getObjectIndexEnd(); I != E; ++I)
+    Offset = RoundUpToAlignment(Offset + MFI->getObjectSize(I), MaxAlign);
+
+  // Call frame.
+  if (MFI->adjustsStack() && hasReservedCallFrame(MF))
+    Offset = RoundUpToAlignment(Offset + MFI->getMaxCallFrameSize(),
+                                std::max(MaxAlign, getStackAlignment()));
+
+  return RoundUpToAlignment(Offset, getStackAlignment());
+}

Modified: llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsFrameLowering.h Tue Nov 13 09:21:47 2012
@@ -34,6 +34,9 @@
                                          const MipsSubtarget &ST);
 
   bool hasFP(const MachineFunction &MF) const;
+
+protected:
+  uint64_t estimateStackSize(const MachineFunction &MF) const;
 };
 
 /// Create MipsInstrInfo objects.

Modified: llvm/branches/R600/lib/Target/Mips/MipsISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsISelDAGToDAG.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsISelDAGToDAG.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsISelDAGToDAG.cpp Tue Nov 13 09:21:47 2012
@@ -86,6 +86,10 @@
 
   SDNode *getGlobalBaseReg();
 
+  SDValue getMips16SPAliasReg();
+
+  void getMips16SPRefReg(SDNode *parent, SDValue &AliasReg);
+
   std::pair<SDNode*, SDNode*> SelectMULT(SDNode *N, unsigned Opc, DebugLoc dl,
                                          EVT Ty, bool HasLo, bool HasHi);
 
@@ -94,6 +98,9 @@
   // Complex Pattern.
   bool SelectAddr(SDNode *Parent, SDValue N, SDValue &Base, SDValue &Offset);
 
+  bool SelectAddr16(SDNode *Parent, SDValue N, SDValue &Base, SDValue &Offset,
+       SDValue &Alias);
+
   // getImm - Return a target constant with the specified value.
   inline SDValue getImm(const SDNode *Node, unsigned Imm) {
     return CurDAG->getTargetConstant(Imm, Node->getValueType(0));
@@ -102,6 +109,7 @@
   void ProcessFunctionAfterISel(MachineFunction &MF);
   bool ReplaceUsesWithZeroReg(MachineRegisterInfo *MRI, const MachineInstr&);
   void InitGlobalBaseReg(MachineFunction &MF);
+  void InitMips16SPAliasReg(MachineFunction &MF);
 
   virtual bool SelectInlineAsmMemoryOperand(const SDValue &Op,
                                             char ConstraintCode,
@@ -220,6 +228,26 @@
     .addReg(Mips::V0).addReg(Mips::T9);
 }
 
+// Insert instructions to initialize the Mips16 SP Alias register in the
+// first MBB of the function.
+//
+void MipsDAGToDAGISel::InitMips16SPAliasReg(MachineFunction &MF) {
+  MipsFunctionInfo *MipsFI = MF.getInfo<MipsFunctionInfo>();
+
+  if (!MipsFI->mips16SPAliasRegSet())
+    return;
+
+  MachineBasicBlock &MBB = MF.front();
+  MachineBasicBlock::iterator I = MBB.begin();
+  const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+  DebugLoc DL = I != MBB.end() ? I->getDebugLoc() : DebugLoc();
+  unsigned Mips16SPAliasReg = MipsFI->getMips16SPAliasReg();
+
+  BuildMI(MBB, I, DL, TII.get(Mips::MoveR3216), Mips16SPAliasReg)
+    .addReg(Mips::SP);
+}
+
+
 bool MipsDAGToDAGISel::ReplaceUsesWithZeroReg(MachineRegisterInfo *MRI,
                                               const MachineInstr& MI) {
   unsigned DstReg = 0, ZeroReg = 0;
@@ -260,6 +288,7 @@
 
 void MipsDAGToDAGISel::ProcessFunctionAfterISel(MachineFunction &MF) {
   InitGlobalBaseReg(MF);
+  InitMips16SPAliasReg(MF);
 
   MachineRegisterInfo *MRI = &MF.getRegInfo();
 
@@ -284,6 +313,14 @@
   return CurDAG->getRegister(GlobalBaseReg, TLI.getPointerTy()).getNode();
 }
 
+/// getMips16SPAliasReg - Output the instructions required to put the
+/// SP into a Mips16 accessible aliased register.
+SDValue MipsDAGToDAGISel::getMips16SPAliasReg() {
+  unsigned Mips16SPAliasReg =
+    MF->getInfo<MipsFunctionInfo>()->getMips16SPAliasReg();
+  return CurDAG->getRegister(Mips16SPAliasReg, TLI.getPointerTy());
+}
+
 /// ComplexPattern used on MipsInstrInfo
 /// Used on Mips Load/Store instructions
 bool MipsDAGToDAGISel::
@@ -362,6 +399,115 @@
   return true;
 }
 
+void MipsDAGToDAGISel::getMips16SPRefReg(SDNode *Parent, SDValue &AliasReg) {
+  SDValue AliasFPReg = CurDAG->getRegister(Mips::S0, TLI.getPointerTy());
+  if (Parent) {
+    switch (Parent->getOpcode()) {
+      case ISD::LOAD: {
+        LoadSDNode *SD = dyn_cast<LoadSDNode>(Parent);
+        switch (SD->getMemoryVT().getSizeInBits()) {
+        case 8:
+        case 16:
+          AliasReg = TM.getFrameLowering()->hasFP(*MF)?
+            AliasFPReg: getMips16SPAliasReg();
+          return;
+        }
+        break;
+      }
+      case ISD::STORE: {
+        StoreSDNode *SD = dyn_cast<StoreSDNode>(Parent);
+        switch (SD->getMemoryVT().getSizeInBits()) {
+        case 8:
+        case 16:
+          AliasReg = TM.getFrameLowering()->hasFP(*MF)?
+            AliasFPReg: getMips16SPAliasReg();
+          return;
+        }
+        break;
+      }
+    }
+  }
+  AliasReg = CurDAG->getRegister(Mips::SP, TLI.getPointerTy());
+  return;
+
+}
+bool MipsDAGToDAGISel::SelectAddr16(
+  SDNode *Parent, SDValue Addr, SDValue &Base, SDValue &Offset,
+  SDValue &Alias) {
+  EVT ValTy = Addr.getValueType();
+
+  Alias = CurDAG->getTargetConstant(0, ValTy);
+
+  // if Address is FI, get the TargetFrameIndex.
+  if (FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(Addr)) {
+    Base   = CurDAG->getTargetFrameIndex(FIN->getIndex(), ValTy);
+    Offset = CurDAG->getTargetConstant(0, ValTy);
+    getMips16SPRefReg(Parent, Alias);
+    return true;
+  }
+  // on PIC code Load GA
+  if (Addr.getOpcode() == MipsISD::Wrapper) {
+    Base   = Addr.getOperand(0);
+    Offset = Addr.getOperand(1);
+    return true;
+  }
+  if (TM.getRelocationModel() != Reloc::PIC_) {
+    if ((Addr.getOpcode() == ISD::TargetExternalSymbol ||
+        Addr.getOpcode() == ISD::TargetGlobalAddress))
+      return false;
+  }
+  // Addresses of the form FI+const or FI|const
+  if (CurDAG->isBaseWithConstantOffset(Addr)) {
+    ConstantSDNode *CN = dyn_cast<ConstantSDNode>(Addr.getOperand(1));
+    if (isInt<16>(CN->getSExtValue())) {
+
+      // If the first operand is a FI, get the TargetFI Node
+      if (FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>
+                                  (Addr.getOperand(0))) {
+        Base = CurDAG->getTargetFrameIndex(FIN->getIndex(), ValTy);
+        getMips16SPRefReg(Parent, Alias);
+      }
+      else
+        Base = Addr.getOperand(0);
+
+      Offset = CurDAG->getTargetConstant(CN->getZExtValue(), ValTy);
+      return true;
+    }
+  }
+  // Operand is a result from an ADD.
+  if (Addr.getOpcode() == ISD::ADD) {
+    // When loading from constant pools, load the lower address part in
+    // the instruction itself. Example, instead of:
+    //  lui $2, %hi($CPI1_0)
+    //  addiu $2, $2, %lo($CPI1_0)
+    //  lwc1 $f0, 0($2)
+    // Generate:
+    //  lui $2, %hi($CPI1_0)
+    //  lwc1 $f0, %lo($CPI1_0)($2)
+    if (Addr.getOperand(1).getOpcode() == MipsISD::Lo ||
+        Addr.getOperand(1).getOpcode() == MipsISD::GPRel) {
+      SDValue Opnd0 = Addr.getOperand(1).getOperand(0);
+      if (isa<ConstantPoolSDNode>(Opnd0) || isa<GlobalAddressSDNode>(Opnd0) ||
+          isa<JumpTableSDNode>(Opnd0)) {
+        Base = Addr.getOperand(0);
+        Offset = Opnd0;
+        return true;
+      }
+    }
+
+    // If an indexed floating point load/store can be emitted, return false.
+    const LSBaseSDNode *LS = dyn_cast<LSBaseSDNode>(Parent);
+
+    if (LS &&
+        (LS->getMemoryVT() == MVT::f32 || LS->getMemoryVT() == MVT::f64) &&
+        Subtarget.hasMips32r2Or64())
+      return false;
+  }
+  Base   = Addr;
+  Offset = CurDAG->getTargetConstant(0, ValTy);
+  return true;
+}
+
 /// Select multiply instructions.
 std::pair<SDNode*, SDNode*>
 MipsDAGToDAGISel::SelectMULT(SDNode *N, unsigned Opc, DebugLoc dl, EVT Ty,
@@ -413,6 +559,7 @@
 
   case ISD::SUBE:
   case ISD::ADDE: {
+    bool inMips16Mode = Subtarget.inMips16Mode();
     SDValue InFlag = Node->getOperand(2), CmpLHS;
     unsigned Opc = InFlag.getOpcode(); (void)Opc;
     assert(((Opc == ISD::ADDC || Opc == ISD::ADDE) ||
@@ -422,10 +569,16 @@
     unsigned MOp;
     if (Opcode == ISD::ADDE) {
       CmpLHS = InFlag.getValue(0);
-      MOp = Mips::ADDu;
+      if (inMips16Mode)
+        MOp = Mips::AdduRxRyRz16;
+      else
+        MOp = Mips::ADDu;
     } else {
       CmpLHS = InFlag.getOperand(0);
-      MOp = Mips::SUBu;
+      if (inMips16Mode)
+        MOp = Mips::SubuRxRyRz16;
+      else
+        MOp = Mips::SUBu;
     }
 
     SDValue Ops[] = { CmpLHS, InFlag.getOperand(1) };
@@ -434,8 +587,11 @@
     SDValue RHS = Node->getOperand(1);
 
     EVT VT = LHS.getValueType();
-    SDNode *Carry = CurDAG->getMachineNode(Mips::SLTu, dl, VT, Ops, 2);
-    SDNode *AddCarry = CurDAG->getMachineNode(Mips::ADDu, dl, VT,
+
+    unsigned Sltu_op = inMips16Mode? Mips::SltuRxRyRz16: Mips::SLTu;
+    SDNode *Carry = CurDAG->getMachineNode(Sltu_op, dl, VT, Ops, 2);
+    unsigned Addu_op = inMips16Mode? Mips::AdduRxRyRz16 : Mips::ADDu;
+    SDNode *AddCarry = CurDAG->getMachineNode(Addu_op, dl, VT,
                                               SDValue(Carry,0), RHS);
 
     return CurDAG->SelectNodeTo(Node, MOp, VT, MVT::Glue,

Modified: llvm/branches/R600/lib/Target/Mips/MipsISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsISelLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsISelLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsISelLowering.cpp Tue Nov 13 09:21:47 2012
@@ -46,6 +46,20 @@
 EnableMipsTailCalls("enable-mips-tail-calls", cl::Hidden,
                     cl::desc("MIPS: Enable tail calls."), cl::init(false));
 
+static const uint16_t O32IntRegs[4] = {
+  Mips::A0, Mips::A1, Mips::A2, Mips::A3
+};
+
+static const uint16_t Mips64IntRegs[8] = {
+  Mips::A0_64, Mips::A1_64, Mips::A2_64, Mips::A3_64,
+  Mips::T0_64, Mips::T1_64, Mips::T2_64, Mips::T3_64
+};
+
+static const uint16_t Mips64DPRegs[8] = {
+  Mips::D12_64, Mips::D13_64, Mips::D14_64, Mips::D15_64,
+  Mips::D16_64, Mips::D17_64, Mips::D18_64, Mips::D19_64
+};
+
 // If I is a shifted mask, set the size (Size) and the first bit of the
 // mask (Pos), and return true.
 // For example, if I is 0x003ff800, (Pos, Size) = (11, 11).
@@ -198,8 +212,14 @@
   setOperationAction(ISD::VASTART,            MVT::Other, Custom);
   setOperationAction(ISD::FCOPYSIGN,          MVT::f32,   Custom);
   setOperationAction(ISD::FCOPYSIGN,          MVT::f64,   Custom);
-  setOperationAction(ISD::MEMBARRIER,         MVT::Other, Custom);
-  setOperationAction(ISD::ATOMIC_FENCE,       MVT::Other, Custom);
+  if (Subtarget->inMips16Mode()) {
+    setOperationAction(ISD::MEMBARRIER,         MVT::Other, Expand);
+    setOperationAction(ISD::ATOMIC_FENCE,       MVT::Other, Expand);
+  }
+  else {
+    setOperationAction(ISD::MEMBARRIER,         MVT::Other, Custom);
+    setOperationAction(ISD::ATOMIC_FENCE,       MVT::Other, Custom);
+  }
   if (!Subtarget->inMips16Mode()) {
     setOperationAction(ISD::LOAD,               MVT::i32, Custom);
     setOperationAction(ISD::STORE,              MVT::i32, Custom);
@@ -227,6 +247,10 @@
     setOperationAction(ISD::SRL_PARTS,          MVT::i32,   Custom);
   }
 
+  setOperationAction(ISD::ADD,                MVT::i32,   Custom);
+  if (HasMips64)
+    setOperationAction(ISD::ADD,                MVT::i64,   Custom);
+
   setOperationAction(ISD::SDIV, MVT::i32, Expand);
   setOperationAction(ISD::SREM, MVT::i32, Expand);
   setOperationAction(ISD::UDIV, MVT::i32, Expand);
@@ -306,6 +330,21 @@
   setOperationAction(ISD::ATOMIC_STORE,      MVT::i32,    Expand);
   setOperationAction(ISD::ATOMIC_STORE,      MVT::i64,    Expand);
 
+  if (Subtarget->inMips16Mode()) {
+    setOperationAction(ISD::ATOMIC_CMP_SWAP,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_SWAP,           MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_ADD,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_SUB,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_AND,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_OR,        MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_XOR,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_NAND,      MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_MIN,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_MAX,       MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_UMIN,      MVT::i32,    Expand);
+    setOperationAction(ISD::ATOMIC_LOAD_UMAX,      MVT::i32,    Expand);
+  }
+
   setInsertFencesForAtomic(true);
 
   if (!Subtarget->hasSEInReg()) {
@@ -879,6 +918,7 @@
     case ISD::STORE:              return LowerSTORE(Op, DAG);
     case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG);
     case ISD::INTRINSIC_W_CHAIN:  return LowerINTRINSIC_W_CHAIN(Op, DAG);
+    case ISD::ADD:                return LowerADD(Op, DAG);
   }
   return SDValue();
 }
@@ -2506,6 +2546,27 @@
   }
 }
 
+SDValue MipsTargetLowering::LowerADD(SDValue Op, SelectionDAG &DAG) const {
+  if (Op->getOperand(0).getOpcode() != ISD::FRAMEADDR
+      || cast<ConstantSDNode>
+        (Op->getOperand(0).getOperand(0))->getZExtValue() != 0
+      || Op->getOperand(1).getOpcode() != ISD::FRAME_TO_ARGS_OFFSET)
+    return SDValue();
+
+  // The pattern
+  //   (add (frameaddr 0), (frame_to_args_offset))
+  // results from lowering llvm.eh.dwarf.cfa intrinsic. Transform it to
+  //   (add FrameObject, 0)
+  // where FrameObject is a fixed StackObject with offset 0 which points to
+  // the old stack pointer.
+  MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
+  EVT ValTy = Op->getValueType(0);
+  int FI = MFI->CreateFixedObject(Op.getValueSizeInBits() / 8, 0, false);
+  SDValue InArgsAddr = DAG.getFrameIndex(FI, ValTy);
+  return DAG.getNode(ISD::ADD, Op->getDebugLoc(), ValTy, InArgsAddr,
+                     DAG.getConstant(0, ValTy));
+}
+
 //===----------------------------------------------------------------------===//
 //                      Calling Convention Implementation
 //===----------------------------------------------------------------------===//
@@ -2541,16 +2602,9 @@
       Mips::D6, Mips::D7
   };
 
-  // ByVal Args
-  if (ArgFlags.isByVal()) {
-    State.HandleByVal(ValNo, ValVT, LocVT, LocInfo,
-                      1 /*MinSize*/, 4 /*MinAlign*/, ArgFlags);
-    unsigned NextReg = (State.getNextStackOffset() + 3) / 4;
-    for (unsigned r = State.getFirstUnallocated(IntRegs, IntRegsSize);
-         r < std::min(IntRegsSize, NextReg); ++r)
-      State.AllocateReg(IntRegs[r]);
-    return false;
-  }
+  // Do not process byval args here.
+  if (ArgFlags.isByVal())
+    return true;
 
   // Promote i8 and i16
   if (LocVT == MVT::i8 || LocVT == MVT::i16) {
@@ -2605,296 +2659,68 @@
   } else
     llvm_unreachable("Cannot handle this ValVT.");
 
-  unsigned SizeInBytes = ValVT.getSizeInBits() >> 3;
-  unsigned Offset;
-  if (!ArgFlags.isSRet())
-    Offset = State.AllocateStack(SizeInBytes, OrigAlign);
-  else
-    Offset = State.AllocateStack(SizeInBytes, SizeInBytes);
-
-  if (!Reg)
+  if (!Reg) {
+    unsigned Offset = State.AllocateStack(ValVT.getSizeInBits() >> 3,
+                                          OrigAlign);
     State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
-  else
+  } else
     State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
 
-  return false; // CC must always match
-}
-
-static const uint16_t Mips64IntRegs[8] =
-  {Mips::A0_64, Mips::A1_64, Mips::A2_64, Mips::A3_64,
-   Mips::T0_64, Mips::T1_64, Mips::T2_64, Mips::T3_64};
-static const uint16_t Mips64DPRegs[8] =
-  {Mips::D12_64, Mips::D13_64, Mips::D14_64, Mips::D15_64,
-   Mips::D16_64, Mips::D17_64, Mips::D18_64, Mips::D19_64};
-
-static bool CC_Mips64Byval(unsigned ValNo, MVT ValVT, MVT LocVT,
-                           CCValAssign::LocInfo LocInfo,
-                           ISD::ArgFlagsTy ArgFlags, CCState &State) {
-  unsigned Align = std::max(ArgFlags.getByValAlign(), (unsigned)8);
-  unsigned Size  = (ArgFlags.getByValSize() + 7) / 8 * 8;
-  unsigned FirstIdx = State.getFirstUnallocated(Mips64IntRegs, 8);
-
-  assert(Align <= 16 && "Cannot handle alignments larger than 16.");
-
-  // If byval is 16-byte aligned, the first arg register must be even.
-  if ((Align == 16) && (FirstIdx % 2)) {
-    State.AllocateReg(Mips64IntRegs[FirstIdx], Mips64DPRegs[FirstIdx]);
-    ++FirstIdx;
-  }
-
-  // Mark the registers allocated.
-  for (unsigned I = FirstIdx; Size && (I < 8); Size -= 8, ++I)
-    State.AllocateReg(Mips64IntRegs[I], Mips64DPRegs[I]);
-
-  // Allocate space on caller's stack.
-  unsigned Offset = State.AllocateStack(Size, Align);
-
-  if (FirstIdx < 8)
-    State.addLoc(CCValAssign::getReg(ValNo, ValVT, Mips64IntRegs[FirstIdx],
-                                     LocVT, LocInfo));
-  else
-    State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
-
-  return true;
+  return false;
 }
 
 #include "MipsGenCallingConv.inc"
 
-static void
-AnalyzeMips64CallOperands(CCState &CCInfo,
-                          const SmallVectorImpl<ISD::OutputArg> &Outs) {
-  unsigned NumOps = Outs.size();
-  for (unsigned i = 0; i != NumOps; ++i) {
-    MVT ArgVT = Outs[i].VT;
-    ISD::ArgFlagsTy ArgFlags = Outs[i].Flags;
-    bool R;
-
-    if (Outs[i].IsFixed)
-      R = CC_MipsN(i, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo);
-    else
-      R = CC_MipsN_VarArg(i, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo);
-
-    if (R) {
-#ifndef NDEBUG
-      dbgs() << "Call operand #" << i << " has unhandled type "
-             << EVT(ArgVT).getEVTString();
-#endif
-      llvm_unreachable(0);
-    }
-  }
-}
-
 //===----------------------------------------------------------------------===//
 //                  Call Calling Convention Implementation
 //===----------------------------------------------------------------------===//
 
 static const unsigned O32IntRegsSize = 4;
 
-static const uint16_t O32IntRegs[] = {
-  Mips::A0, Mips::A1, Mips::A2, Mips::A3
-};
-
 // Return next O32 integer argument register.
 static unsigned getNextIntArgReg(unsigned Reg) {
   assert((Reg == Mips::A0) || (Reg == Mips::A2));
   return (Reg == Mips::A0) ? Mips::A1 : Mips::A3;
 }
 
-// Write ByVal Arg to arg registers and stack.
-static void
-WriteByValArg(SDValue Chain, DebugLoc dl,
-              SmallVector<std::pair<unsigned, SDValue>, 16> &RegsToPass,
-              SmallVector<SDValue, 8> &MemOpChains, SDValue StackPtr,
-              MachineFrameInfo *MFI, SelectionDAG &DAG, SDValue Arg,
-              const CCValAssign &VA, const ISD::ArgFlagsTy &Flags,
-              MVT PtrType, bool isLittle) {
-  unsigned LocMemOffset = VA.getLocMemOffset();
-  unsigned Offset = 0;
-  uint32_t RemainingSize = Flags.getByValSize();
-  unsigned ByValAlign = Flags.getByValAlign();
-
-  // Copy the first 4 words of byval arg to registers A0 - A3.
-  // FIXME: Use a stricter alignment if it enables better optimization in passes
-  //        run later.
-  for (; RemainingSize >= 4 && LocMemOffset < 4 * 4;
-       Offset += 4, RemainingSize -= 4, LocMemOffset += 4) {
-    SDValue LoadPtr = DAG.getNode(ISD::ADD, dl, MVT::i32, Arg,
-                                  DAG.getConstant(Offset, MVT::i32));
-    SDValue LoadVal = DAG.getLoad(MVT::i32, dl, Chain, LoadPtr,
-                                  MachinePointerInfo(), false, false, false,
-                                  std::min(ByValAlign, (unsigned )4));
-    MemOpChains.push_back(LoadVal.getValue(1));
-    unsigned DstReg = O32IntRegs[LocMemOffset / 4];
-    RegsToPass.push_back(std::make_pair(DstReg, LoadVal));
-  }
-
-  if (RemainingSize == 0)
-    return;
-
-  // If there still is a register available for argument passing, write the
-  // remaining part of the structure to it using subword loads and shifts.
-  if (LocMemOffset < 4 * 4) {
-    assert(RemainingSize <= 3 && RemainingSize >= 1 &&
-           "There must be one to three bytes remaining.");
-    unsigned LoadSize = (RemainingSize == 3 ? 2 : RemainingSize);
-    SDValue LoadPtr = DAG.getNode(ISD::ADD, dl, MVT::i32, Arg,
-                                  DAG.getConstant(Offset, MVT::i32));
-    unsigned Alignment = std::min(ByValAlign, (unsigned )4);
-    SDValue LoadVal = DAG.getExtLoad(ISD::ZEXTLOAD, dl, MVT::i32, Chain,
-                                     LoadPtr, MachinePointerInfo(),
-                                     MVT::getIntegerVT(LoadSize * 8), false,
-                                     false, Alignment);
-    MemOpChains.push_back(LoadVal.getValue(1));
-
-    // If target is big endian, shift it to the most significant half-word or
-    // byte.
-    if (!isLittle)
-      LoadVal = DAG.getNode(ISD::SHL, dl, MVT::i32, LoadVal,
-                            DAG.getConstant(32 - LoadSize * 8, MVT::i32));
-
-    Offset += LoadSize;
-    RemainingSize -= LoadSize;
-
-    // Read second subword if necessary.
-    if (RemainingSize != 0)  {
-      assert(RemainingSize == 1 && "There must be one byte remaining.");
-      LoadPtr = DAG.getNode(ISD::ADD, dl, MVT::i32, Arg,
-                            DAG.getConstant(Offset, MVT::i32));
-      unsigned Alignment = std::min(ByValAlign, (unsigned )2);
-      SDValue Subword = DAG.getExtLoad(ISD::ZEXTLOAD, dl, MVT::i32, Chain,
-                                       LoadPtr, MachinePointerInfo(),
-                                       MVT::i8, false, false, Alignment);
-      MemOpChains.push_back(Subword.getValue(1));
-      // Insert the loaded byte to LoadVal.
-      // FIXME: Use INS if supported by target.
-      unsigned ShiftAmt = isLittle ? 16 : 8;
-      SDValue Shift = DAG.getNode(ISD::SHL, dl, MVT::i32, Subword,
-                                  DAG.getConstant(ShiftAmt, MVT::i32));
-      LoadVal = DAG.getNode(ISD::OR, dl, MVT::i32, LoadVal, Shift);
-    }
-
-    unsigned DstReg = O32IntRegs[LocMemOffset / 4];
-    RegsToPass.push_back(std::make_pair(DstReg, LoadVal));
-    return;
-  }
-
-  // Copy remaining part of byval arg using memcpy.
-  SDValue Src = DAG.getNode(ISD::ADD, dl, MVT::i32, Arg,
-                            DAG.getConstant(Offset, MVT::i32));
-  SDValue Dst = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr,
-                            DAG.getIntPtrConstant(LocMemOffset));
-  Chain = DAG.getMemcpy(Chain, dl, Dst, Src,
-                        DAG.getConstant(RemainingSize, MVT::i32),
-                        std::min(ByValAlign, (unsigned)4),
-                        /*isVolatile=*/false, /*AlwaysInline=*/false,
-                        MachinePointerInfo(0), MachinePointerInfo(0));
-  MemOpChains.push_back(Chain);
-}
-
-// Copy Mips64 byVal arg to registers and stack.
-void static
-PassByValArg64(SDValue Chain, DebugLoc dl,
-               SmallVector<std::pair<unsigned, SDValue>, 16> &RegsToPass,
-               SmallVector<SDValue, 8> &MemOpChains, SDValue StackPtr,
-               MachineFrameInfo *MFI, SelectionDAG &DAG, SDValue Arg,
-               const CCValAssign &VA, const ISD::ArgFlagsTy &Flags,
-               EVT PtrTy, bool isLittle) {
-  unsigned ByValSize = Flags.getByValSize();
-  unsigned Alignment = std::min(Flags.getByValAlign(), (unsigned)8);
-  bool IsRegLoc = VA.isRegLoc();
-  unsigned Offset = 0; // Offset in # of bytes from the beginning of struct.
-  unsigned LocMemOffset = 0;
-  unsigned MemCpySize = ByValSize;
-
-  if (!IsRegLoc)
-    LocMemOffset = VA.getLocMemOffset();
-  else {
-    const uint16_t *Reg = std::find(Mips64IntRegs, Mips64IntRegs + 8,
-                                    VA.getLocReg());
-    const uint16_t *RegEnd = Mips64IntRegs + 8;
-
-    // Copy double words to registers.
-    for (; (Reg != RegEnd) && (ByValSize >= Offset + 8); ++Reg, Offset += 8) {
-      SDValue LoadPtr = DAG.getNode(ISD::ADD, dl, PtrTy, Arg,
-                                    DAG.getConstant(Offset, PtrTy));
-      SDValue LoadVal = DAG.getLoad(MVT::i64, dl, Chain, LoadPtr,
-                                    MachinePointerInfo(), false, false, false,
-                                    Alignment);
-      MemOpChains.push_back(LoadVal.getValue(1));
-      RegsToPass.push_back(std::make_pair(*Reg, LoadVal));
-    }
-
-    // Return if the struct has been fully copied.
-    if (!(MemCpySize = ByValSize - Offset))
-      return;
-
-    // If there is an argument register available, copy the remainder of the
-    // byval argument with sub-doubleword loads and shifts.
-    if (Reg != RegEnd) {
-      assert((ByValSize < Offset + 8) &&
-             "Size of the remainder should be smaller than 8-byte.");
-      SDValue Val;
-      for (unsigned LoadSize = 4; Offset < ByValSize; LoadSize /= 2) {
-        unsigned RemSize = ByValSize - Offset;
-
-        if (RemSize < LoadSize)
-          continue;
-
-        SDValue LoadPtr = DAG.getNode(ISD::ADD, dl, PtrTy, Arg,
-                                      DAG.getConstant(Offset, PtrTy));
-        SDValue LoadVal =
-          DAG.getExtLoad(ISD::ZEXTLOAD, dl, MVT::i64, Chain, LoadPtr,
-                         MachinePointerInfo(), MVT::getIntegerVT(LoadSize * 8),
-                         false, false, Alignment);
-        MemOpChains.push_back(LoadVal.getValue(1));
-
-        // Offset in number of bits from double word boundary.
-        unsigned OffsetDW = (Offset % 8) * 8;
-        unsigned Shamt = isLittle ? OffsetDW : 64 - (OffsetDW + LoadSize * 8);
-        SDValue Shift = DAG.getNode(ISD::SHL, dl, MVT::i64, LoadVal,
-                                    DAG.getConstant(Shamt, MVT::i32));
-
-        Val = Val.getNode() ? DAG.getNode(ISD::OR, dl, MVT::i64, Val, Shift) :
-                              Shift;
-        Offset += LoadSize;
-        Alignment = std::min(Alignment, LoadSize);
-      }
-
-      RegsToPass.push_back(std::make_pair(*Reg, Val));
-      return;
-    }
-  }
-
-  assert(MemCpySize && "MemCpySize must not be zero.");
-
-  // Copy remainder of byval arg to it with memcpy.
-  SDValue Src = DAG.getNode(ISD::ADD, dl, PtrTy, Arg,
-                            DAG.getConstant(Offset, PtrTy));
-  SDValue Dst = DAG.getNode(ISD::ADD, dl, MVT::i64, StackPtr,
-                            DAG.getIntPtrConstant(LocMemOffset));
-  Chain = DAG.getMemcpy(Chain, dl, Dst, Src,
-                        DAG.getConstant(MemCpySize, PtrTy), Alignment,
-                        /*isVolatile=*/false, /*AlwaysInline=*/false,
-                        MachinePointerInfo(0), MachinePointerInfo(0));
-  MemOpChains.push_back(Chain);
-}
-
 /// IsEligibleForTailCallOptimization - Check whether the call is eligible
 /// for tail call optimization.
 bool MipsTargetLowering::
-IsEligibleForTailCallOptimization(CallingConv::ID CalleeCC,
-                                  unsigned NextStackOffset) const {
+IsEligibleForTailCallOptimization(const MipsCC &MipsCCInfo,
+                                  unsigned NextStackOffset,
+                                  const MipsFunctionInfo& FI) const {
   if (!EnableMipsTailCalls)
     return false;
 
-  // Do not tail-call optimize if there is an argument passed on stack.
-  if (IsO32 && (CalleeCC != CallingConv::Fast)) {
-    if (NextStackOffset > 16)
-      return false;
-  } else if (NextStackOffset)
+  // No tail call optimization for mips16.
+  if (Subtarget->inMips16Mode())
     return false;
 
-  return true;
+  // Return false if either the callee or caller has a byval argument.
+  if (MipsCCInfo.hasByValArg() || FI.hasByvalArg())
+    return false;
+
+  // Return true if the callee's argument area is no larger than the
+  // caller's.
+  return NextStackOffset <= FI.getIncomingArgSize();
+}
+
+SDValue
+MipsTargetLowering::passArgOnStack(SDValue StackPtr, unsigned Offset,
+                                   SDValue Chain, SDValue Arg, DebugLoc DL,
+                                   bool IsTailCall, SelectionDAG &DAG) const {
+  if (!IsTailCall) {
+    SDValue PtrOff = DAG.getNode(ISD::ADD, DL, getPointerTy(), StackPtr,
+                                 DAG.getIntPtrConstant(Offset));
+    return DAG.getStore(Chain, DL, Arg, PtrOff, MachinePointerInfo(), false,
+                        false, 0);
+  }
+
+  MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
+  int FI = MFI->CreateFixedObject(Arg.getValueSizeInBits() / 8, Offset, false);
+  SDValue FIN = DAG.getFrameIndex(FI, getPointerTy());
+  return DAG.getStore(Chain, DL, Arg, FIN, MachinePointerInfo(),
+                      /*isVolatile=*/ true, false, 0);
 }
 
 /// LowerCall - functions arguments are copied from virtual regs to
@@ -2922,30 +2748,18 @@
   SmallVector<CCValAssign, 16> ArgLocs;
   CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
                  getTargetMachine(), ArgLocs, *DAG.getContext());
+  MipsCC MipsCCInfo(CallConv, isVarArg, IsO32, CCInfo);
 
-  if (CallConv == CallingConv::Fast)
-    CCInfo.AnalyzeCallOperands(Outs, CC_Mips_FastCC);
-  else if (IsO32)
-    CCInfo.AnalyzeCallOperands(Outs, CC_MipsO32);
-  else if (HasMips64)
-    AnalyzeMips64CallOperands(CCInfo, Outs);
-  else
-    CCInfo.AnalyzeCallOperands(Outs, CC_Mips);
+  MipsCCInfo.analyzeCallOperands(Outs);
 
   // Get a count of how many bytes are to be pushed on the stack.
   unsigned NextStackOffset = CCInfo.getNextStackOffset();
-  unsigned StackAlignment = TFL->getStackAlignment();
-  NextStackOffset = RoundUpToAlignment(NextStackOffset, StackAlignment);
-
-  // Update size of the maximum argument space.
-  // For O32, a minimum of four words (16 bytes) of argument space is
-  // allocated.
-  if (IsO32 && (CallConv != CallingConv::Fast))
-    NextStackOffset = std::max(NextStackOffset, (unsigned)16);
 
   // Check if it's really possible to do a tail call.
   if (isTailCall)
-    isTailCall = IsEligibleForTailCallOptimization(CallConv, NextStackOffset);
+    isTailCall =
+      IsEligibleForTailCallOptimization(MipsCCInfo, NextStackOffset,
+                                        *MF.getInfo<MipsFunctionInfo>());
 
   if (isTailCall)
     ++NumTailCalls;
@@ -2953,6 +2767,8 @@
   // Chain is the output chain of the last Load/Store or CopyToReg node.
   // ByValChain is the output chain of the last Memcpy node created for copying
   // byval arguments to the stack.
+  unsigned StackAlignment = TFL->getStackAlignment();
+  NextStackOffset = RoundUpToAlignment(NextStackOffset, StackAlignment);
   SDValue NextStackOffsetVal = DAG.getIntPtrConstant(NextStackOffset, true);
 
   if (!isTailCall)
@@ -2965,6 +2781,7 @@
   // With EABI is it possible to have 16 args on registers.
   SmallVector<std::pair<unsigned, SDValue>, 16> RegsToPass;
   SmallVector<SDValue, 8> MemOpChains;
+  MipsCC::byval_iterator ByValArg = MipsCCInfo.byval_begin();
 
   // Walk the register/memloc assignments, inserting copies/loads.
   for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
@@ -2977,14 +2794,12 @@
     if (Flags.isByVal()) {
       assert(Flags.getByValSize() &&
              "ByVal args of size 0 should have been ignored by front-end.");
-      if (IsO32)
-        WriteByValArg(Chain, dl, RegsToPass, MemOpChains, StackPtr,
-                      MFI, DAG, Arg, VA, Flags, getPointerTy(),
-                      Subtarget->isLittle());
-      else
-        PassByValArg64(Chain, dl, RegsToPass, MemOpChains, StackPtr,
-                       MFI, DAG, Arg, VA, Flags, getPointerTy(),
-                       Subtarget->isLittle());
+      assert(ByValArg != MipsCCInfo.byval_end());
+      assert(!isTailCall &&
+             "Do not tail-call optimize if there is a byval argument.");
+      passByValArg(Chain, dl, RegsToPass, MemOpChains, StackPtr, MFI, DAG, Arg,
+                   MipsCCInfo, *ByValArg, Flags, Subtarget->isLittle());
+      ++ByValArg;
       continue;
     }
 
@@ -3034,10 +2849,8 @@
 
     // emit ISD::STORE whichs stores the
     // parameter value to a stack Location
-    SDValue PtrOff = DAG.getNode(ISD::ADD, dl, getPointerTy(), StackPtr,
-                                 DAG.getIntPtrConstant(VA.getLocMemOffset()));
-    MemOpChains.push_back(DAG.getStore(Chain, dl, Arg, PtrOff,
-                                       MachinePointerInfo(), false, false, 0));
+    MemOpChains.push_back(passArgOnStack(StackPtr, VA.getLocMemOffset(),
+                                         Chain, Arg, dl, isTailCall, DAG));
   }
 
   // Transform all store nodes into one single node because all store
@@ -3212,70 +3025,6 @@
 //===----------------------------------------------------------------------===//
 //             Formal Arguments Calling Convention Implementation
 //===----------------------------------------------------------------------===//
-static void ReadByValArg(MachineFunction &MF, SDValue Chain, DebugLoc dl,
-                         std::vector<SDValue> &OutChains,
-                         SelectionDAG &DAG, unsigned NumWords, SDValue FIN,
-                         const CCValAssign &VA, const ISD::ArgFlagsTy &Flags,
-                         const Argument *FuncArg) {
-  unsigned LocMem = VA.getLocMemOffset();
-  unsigned FirstWord = LocMem / 4;
-
-  // copy register A0 - A3 to frame object
-  for (unsigned i = 0; i < NumWords; ++i) {
-    unsigned CurWord = FirstWord + i;
-    if (CurWord >= O32IntRegsSize)
-      break;
-
-    unsigned SrcReg = O32IntRegs[CurWord];
-    unsigned Reg = AddLiveIn(MF, SrcReg, &Mips::CPURegsRegClass);
-    SDValue StorePtr = DAG.getNode(ISD::ADD, dl, MVT::i32, FIN,
-                                   DAG.getConstant(i * 4, MVT::i32));
-    SDValue Store = DAG.getStore(Chain, dl, DAG.getRegister(Reg, MVT::i32),
-                                 StorePtr, MachinePointerInfo(FuncArg, i * 4),
-                                 false, false, 0);
-    OutChains.push_back(Store);
-  }
-}
-
-// Create frame object on stack and copy registers used for byval passing to it.
-static unsigned
-CopyMips64ByValRegs(MachineFunction &MF, SDValue Chain, DebugLoc dl,
-                    std::vector<SDValue> &OutChains, SelectionDAG &DAG,
-                    const CCValAssign &VA, const ISD::ArgFlagsTy &Flags,
-                    MachineFrameInfo *MFI, bool IsRegLoc,
-                    SmallVectorImpl<SDValue> &InVals, MipsFunctionInfo *MipsFI,
-                    EVT PtrTy, const Argument *FuncArg) {
-  const uint16_t *Reg = Mips64IntRegs + 8;
-  int FOOffset; // Frame object offset from virtual frame pointer.
-
-  if (IsRegLoc) {
-    Reg = std::find(Mips64IntRegs, Mips64IntRegs + 8, VA.getLocReg());
-    FOOffset = (Reg - Mips64IntRegs) * 8 - 8 * 8;
-  }
-  else
-    FOOffset = VA.getLocMemOffset();
-
-  // Create frame object.
-  unsigned NumRegs = (Flags.getByValSize() + 7) / 8;
-  unsigned LastFI = MFI->CreateFixedObject(NumRegs * 8, FOOffset, true);
-  SDValue FIN = DAG.getFrameIndex(LastFI, PtrTy);
-  InVals.push_back(FIN);
-
-  // Copy arg registers.
-  for (unsigned I = 0; (Reg != Mips64IntRegs + 8) && (I < NumRegs);
-       ++Reg, ++I) {
-    unsigned VReg = AddLiveIn(MF, *Reg, &Mips::CPU64RegsRegClass);
-    SDValue StorePtr = DAG.getNode(ISD::ADD, dl, PtrTy, FIN,
-                                   DAG.getConstant(I * 8, PtrTy));
-    SDValue Store = DAG.getStore(Chain, dl, DAG.getRegister(VReg, MVT::i64),
-                                 StorePtr, MachinePointerInfo(FuncArg, I * 8),
-                                 false, false, 0);
-    OutChains.push_back(Store);
-  }
-
-  return LastFI;
-}
-
 /// LowerFormalArguments - transform physical registers into virtual registers
 /// and generate load operations for arguments places on the stack.
 SDValue
@@ -3299,20 +3048,21 @@
   SmallVector<CCValAssign, 16> ArgLocs;
   CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
                  getTargetMachine(), ArgLocs, *DAG.getContext());
+  MipsCC MipsCCInfo(CallConv, isVarArg, IsO32, CCInfo);
 
-  if (CallConv == CallingConv::Fast)
-    CCInfo.AnalyzeFormalArguments(Ins, CC_Mips_FastCC);
-  else if (IsO32)
-    CCInfo.AnalyzeFormalArguments(Ins, CC_MipsO32);
-  else
-    CCInfo.AnalyzeFormalArguments(Ins, CC_Mips);
+  MipsCCInfo.analyzeFormalArguments(Ins);
+  MipsFI->setFormalArgInfo(CCInfo.getNextStackOffset(),
+                           MipsCCInfo.hasByValArg());
 
   Function::const_arg_iterator FuncArg =
     DAG.getMachineFunction().getFunction()->arg_begin();
-  int LastFI = 0;// MipsFI->LastInArgFI is 0 at the entry of this function.
+  unsigned CurArgIdx = 0;
+  MipsCC::byval_iterator ByValArg = MipsCCInfo.byval_begin();
 
-  for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i, ++FuncArg) {
+  for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
     CCValAssign &VA = ArgLocs[i];
+    std::advance(FuncArg, Ins[i].OrigArgIndex - CurArgIdx);
+    CurArgIdx = Ins[i].OrigArgIndex;
     EVT ValVT = VA.getValVT();
     ISD::ArgFlagsTy Flags = Ins[i].Flags;
     bool IsRegLoc = VA.isRegLoc();
@@ -3320,18 +3070,10 @@
     if (Flags.isByVal()) {
       assert(Flags.getByValSize() &&
              "ByVal args of size 0 should have been ignored by front-end.");
-      if (IsO32) {
-        unsigned NumWords = (Flags.getByValSize() + 3) / 4;
-        LastFI = MFI->CreateFixedObject(NumWords * 4, VA.getLocMemOffset(),
-                                        true);
-        SDValue FIN = DAG.getFrameIndex(LastFI, getPointerTy());
-        InVals.push_back(FIN);
-        ReadByValArg(MF, Chain, dl, OutChains, DAG, NumWords, FIN, VA, Flags,
-                     &*FuncArg);
-      } else // N32/64
-        LastFI = CopyMips64ByValRegs(MF, Chain, dl, OutChains, DAG, VA, Flags,
-                                     MFI, IsRegLoc, InVals, MipsFI,
-                                     getPointerTy(), &*FuncArg);
+      assert(ByValArg != MipsCCInfo.byval_end());
+      copyByValRegs(Chain, dl, OutChains, DAG, Flags, InVals, &*FuncArg,
+                    MipsCCInfo, *ByValArg);
+      ++ByValArg;
       continue;
     }
 
@@ -3393,13 +3135,13 @@
       assert(VA.isMemLoc());
 
       // The stack pointer offset is relative to the caller stack frame.
-      LastFI = MFI->CreateFixedObject(ValVT.getSizeInBits()/8,
+      int FI = MFI->CreateFixedObject(ValVT.getSizeInBits()/8,
                                       VA.getLocMemOffset(), true);
 
       // Create load nodes to retrieve arguments from the stack
-      SDValue FIN = DAG.getFrameIndex(LastFI, getPointerTy());
+      SDValue FIN = DAG.getFrameIndex(FI, getPointerTy());
       InVals.push_back(DAG.getLoad(ValVT, dl, Chain, FIN,
-                                   MachinePointerInfo::getFixedStack(LastFI),
+                                   MachinePointerInfo::getFixedStack(FI),
                                    false, false, false, 0));
     }
   }
@@ -3418,48 +3160,8 @@
     Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Copy, Chain);
   }
 
-  if (isVarArg) {
-    unsigned NumOfRegs = IsO32 ? 4 : 8;
-    const uint16_t *ArgRegs = IsO32 ? O32IntRegs : Mips64IntRegs;
-    unsigned Idx = CCInfo.getFirstUnallocated(ArgRegs, NumOfRegs);
-    int FirstRegSlotOffset = IsO32 ? 0 : -64 ; // offset of $a0's slot.
-    const TargetRegisterClass *RC = IsO32 ?
-      (const TargetRegisterClass*)&Mips::CPURegsRegClass :
-      (const TargetRegisterClass*)&Mips::CPU64RegsRegClass;
-    unsigned RegSize = RC->getSize();
-    int RegSlotOffset = FirstRegSlotOffset + Idx * RegSize;
-
-    // Offset of the first variable argument from stack pointer.
-    int FirstVaArgOffset;
-
-    if (IsO32 || (Idx == NumOfRegs)) {
-      FirstVaArgOffset =
-        (CCInfo.getNextStackOffset() + RegSize - 1) / RegSize * RegSize;
-    } else
-      FirstVaArgOffset = RegSlotOffset;
-
-    // Record the frame index of the first variable argument
-    // which is a value necessary to VASTART.
-    LastFI = MFI->CreateFixedObject(RegSize, FirstVaArgOffset, true);
-    MipsFI->setVarArgsFrameIndex(LastFI);
-
-    // Copy the integer registers that have not been used for argument passing
-    // to the argument register save area. For O32, the save area is allocated
-    // in the caller's stack frame, while for N32/64, it is allocated in the
-    // callee's stack frame.
-    for (int StackOffset = RegSlotOffset;
-         Idx < NumOfRegs; ++Idx, StackOffset += RegSize) {
-      unsigned Reg = AddLiveIn(DAG.getMachineFunction(), ArgRegs[Idx], RC);
-      SDValue ArgValue = DAG.getCopyFromReg(Chain, dl, Reg,
-                                            MVT::getIntegerVT(RegSize * 8));
-      LastFI = MFI->CreateFixedObject(RegSize, StackOffset, true);
-      SDValue PtrOff = DAG.getFrameIndex(LastFI, getPointerTy());
-      OutChains.push_back(DAG.getStore(Chain, dl, ArgValue, PtrOff,
-                                       MachinePointerInfo(), false, false, 0));
-    }
-  }
-
-  MipsFI->setLastInArgFI(LastFI);
+  if (isVarArg)
+    writeVarArgRegs(OutChains, MipsCCInfo, Chain, dl, DAG);
 
   // All stores are grouped in one node to allow the matching between
   // the size of Ins and InVals. This only happens when on varg functions
@@ -3539,10 +3241,11 @@
     if (!Reg)
       llvm_unreachable("sret virtual register not created in the entry block");
     SDValue Val = DAG.getCopyFromReg(Chain, dl, Reg, getPointerTy());
+    unsigned V0 = IsN64 ? Mips::V0_64 : Mips::V0;
 
-    Chain = DAG.getCopyToReg(Chain, dl, IsN64 ? Mips::V0_64 : Mips::V0, Val,
-                             Flag);
+    Chain = DAG.getCopyToReg(Chain, dl, V0, Val, Flag);
     Flag = Chain.getValue(1);
+    MF.getRegInfo().addLiveOut(V0);
   }
 
   // Return on Mips is always a "jr $ra"
@@ -3809,3 +3512,316 @@
 
   return TargetLowering::getJumpTableEncoding();
 }
+
+MipsTargetLowering::MipsCC::MipsCC(CallingConv::ID CallConv, bool IsVarArg,
+                                   bool IsO32, CCState &Info) : CCInfo(Info) {
+  UseRegsForByval = true;
+
+  if (IsO32) {
+    RegSize = 4;
+    NumIntArgRegs = array_lengthof(O32IntRegs);
+    ReservedArgArea = 16;
+    IntArgRegs = ShadowRegs = O32IntRegs;
+    FixedFn = VarFn = CC_MipsO32;
+  } else {
+    RegSize = 8;
+    NumIntArgRegs = array_lengthof(Mips64IntRegs);
+    ReservedArgArea = 0;
+    IntArgRegs = Mips64IntRegs;
+    ShadowRegs = Mips64DPRegs;
+    FixedFn = CC_MipsN;
+    VarFn = CC_MipsN_VarArg;
+  }
+
+  if (CallConv == CallingConv::Fast) {
+    assert(!IsVarArg);
+    UseRegsForByval = false;
+    ReservedArgArea = 0;
+    FixedFn = VarFn = CC_Mips_FastCC;
+  }
+
+  // Pre-allocate reserved argument area.
+  CCInfo.AllocateStack(ReservedArgArea, 1);
+}
+
+void MipsTargetLowering::MipsCC::
+analyzeCallOperands(const SmallVectorImpl<ISD::OutputArg> &Args) {
+  unsigned NumOpnds = Args.size();
+
+  for (unsigned I = 0; I != NumOpnds; ++I) {
+    MVT ArgVT = Args[I].VT;
+    ISD::ArgFlagsTy ArgFlags = Args[I].Flags;
+    bool R;
+
+    if (ArgFlags.isByVal()) {
+      handleByValArg(I, ArgVT, ArgVT, CCValAssign::Full, ArgFlags);
+      continue;
+    }
+
+    if (Args[I].IsFixed)
+      R = FixedFn(I, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo);
+    else
+      R = VarFn(I, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo);
+
+    if (R) {
+#ifndef NDEBUG
+      dbgs() << "Call operand #" << I << " has unhandled type "
+             << EVT(ArgVT).getEVTString();
+#endif
+      llvm_unreachable(0);
+    }
+  }
+}
+
+void MipsTargetLowering::MipsCC::
+analyzeFormalArguments(const SmallVectorImpl<ISD::InputArg> &Args) {
+  unsigned NumArgs = Args.size();
+
+  for (unsigned I = 0; I != NumArgs; ++I) {
+    MVT ArgVT = Args[I].VT;
+    ISD::ArgFlagsTy ArgFlags = Args[I].Flags;
+
+    if (ArgFlags.isByVal()) {
+      handleByValArg(I, ArgVT, ArgVT, CCValAssign::Full, ArgFlags);
+      continue;
+    }
+
+    if (!FixedFn(I, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo))
+      continue;
+
+#ifndef NDEBUG
+    dbgs() << "Formal Arg #" << I << " has unhandled type "
+           << EVT(ArgVT).getEVTString();
+#endif
+    llvm_unreachable(0);
+  }
+}
+
+void
+MipsTargetLowering::MipsCC::handleByValArg(unsigned ValNo, MVT ValVT,
+                                           MVT LocVT,
+                                           CCValAssign::LocInfo LocInfo,
+                                           ISD::ArgFlagsTy ArgFlags) {
+  assert(ArgFlags.getByValSize() && "Byval argument's size shouldn't be 0.");
+
+  struct ByValArgInfo ByVal;
+  unsigned ByValSize = RoundUpToAlignment(ArgFlags.getByValSize(), RegSize);
+  unsigned Align = std::min(std::max(ArgFlags.getByValAlign(), RegSize),
+                            RegSize * 2);
+
+  if (UseRegsForByval)
+    allocateRegs(ByVal, ByValSize, Align);
+
+  // Allocate space on caller's stack.
+  ByVal.Address = CCInfo.AllocateStack(ByValSize - RegSize * ByVal.NumRegs,
+                                       Align);
+  CCInfo.addLoc(CCValAssign::getMem(ValNo, ValVT, ByVal.Address, LocVT,
+                                    LocInfo));
+  ByValArgs.push_back(ByVal);
+}
+
+void MipsTargetLowering::MipsCC::allocateRegs(ByValArgInfo &ByVal,
+                                              unsigned ByValSize,
+                                              unsigned Align) {
+  assert(!(ByValSize % RegSize) && !(Align % RegSize) &&
+         "Byval argument's size and alignment should be a multiple of"
+         "RegSize.");
+
+  ByVal.FirstIdx = CCInfo.getFirstUnallocated(IntArgRegs, NumIntArgRegs);
+
+  // If Align > RegSize, the first arg register must be even.
+  if ((Align > RegSize) && (ByVal.FirstIdx % 2)) {
+    CCInfo.AllocateReg(IntArgRegs[ByVal.FirstIdx], ShadowRegs[ByVal.FirstIdx]);
+    ++ByVal.FirstIdx;
+  }
+
+  // Mark the registers allocated.
+  for (unsigned I = ByVal.FirstIdx; ByValSize && (I < NumIntArgRegs);
+       ByValSize -= RegSize, ++I, ++ByVal.NumRegs)
+    CCInfo.AllocateReg(IntArgRegs[I], ShadowRegs[I]);
+}
+
+void MipsTargetLowering::
+copyByValRegs(SDValue Chain, DebugLoc DL, std::vector<SDValue> &OutChains,
+              SelectionDAG &DAG, const ISD::ArgFlagsTy &Flags,
+              SmallVectorImpl<SDValue> &InVals, const Argument *FuncArg,
+              const MipsCC &CC, const ByValArgInfo &ByVal) const {
+  MachineFunction &MF = DAG.getMachineFunction();
+  MachineFrameInfo *MFI = MF.getFrameInfo();
+  unsigned RegAreaSize = ByVal.NumRegs * CC.regSize();
+  unsigned FrameObjSize = std::max(Flags.getByValSize(), RegAreaSize);
+  int FrameObjOffset;
+
+  if (RegAreaSize)
+    FrameObjOffset = (int)CC.reservedArgArea() -
+      (int)((CC.numIntArgRegs() - ByVal.FirstIdx) * CC.regSize());
+  else
+    FrameObjOffset = ByVal.Address;
+
+  // Create frame object.
+  EVT PtrTy = getPointerTy();
+  int FI = MFI->CreateFixedObject(FrameObjSize, FrameObjOffset, true);
+  SDValue FIN = DAG.getFrameIndex(FI, PtrTy);
+  InVals.push_back(FIN);
+
+  if (!ByVal.NumRegs)
+    return;
+
+  // Copy arg registers.
+  EVT RegTy = MVT::getIntegerVT(CC.regSize() * 8);
+  const TargetRegisterClass *RC = getRegClassFor(RegTy);
+
+  for (unsigned I = 0; I < ByVal.NumRegs; ++I) {
+    unsigned ArgReg = CC.intArgRegs()[ByVal.FirstIdx + I];
+    unsigned VReg = AddLiveIn(MF, ArgReg, RC);
+    unsigned Offset = I * CC.regSize();
+    SDValue StorePtr = DAG.getNode(ISD::ADD, DL, PtrTy, FIN,
+                                   DAG.getConstant(Offset, PtrTy));
+    SDValue Store = DAG.getStore(Chain, DL, DAG.getRegister(VReg, RegTy),
+                                 StorePtr, MachinePointerInfo(FuncArg, Offset),
+                                 false, false, 0);
+    OutChains.push_back(Store);
+  }
+}
+
+// Copy byVal arg to registers and stack.
+void MipsTargetLowering::
+passByValArg(SDValue Chain, DebugLoc DL,
+             SmallVector<std::pair<unsigned, SDValue>, 16> &RegsToPass,
+             SmallVector<SDValue, 8> &MemOpChains, SDValue StackPtr,
+             MachineFrameInfo *MFI, SelectionDAG &DAG, SDValue Arg,
+             const MipsCC &CC, const ByValArgInfo &ByVal,
+             const ISD::ArgFlagsTy &Flags, bool isLittle) const {
+  unsigned ByValSize = Flags.getByValSize();
+  unsigned Offset = 0; // Offset in # of bytes from the beginning of struct.
+  unsigned RegSize = CC.regSize();
+  unsigned Alignment = std::min(Flags.getByValAlign(), RegSize);
+  EVT PtrTy = getPointerTy(), RegTy = MVT::getIntegerVT(RegSize * 8);
+
+  if (ByVal.NumRegs) {
+    const uint16_t *ArgRegs = CC.intArgRegs();
+    bool LeftoverBytes = (ByVal.NumRegs * RegSize > ByValSize);
+    unsigned I = 0;
+
+    // Copy words to registers.
+    for (; I < ByVal.NumRegs - LeftoverBytes; ++I, Offset += RegSize) {
+      SDValue LoadPtr = DAG.getNode(ISD::ADD, DL, PtrTy, Arg,
+                                    DAG.getConstant(Offset, PtrTy));
+      SDValue LoadVal = DAG.getLoad(RegTy, DL, Chain, LoadPtr,
+                                    MachinePointerInfo(), false, false, false,
+                                    Alignment);
+      MemOpChains.push_back(LoadVal.getValue(1));
+      unsigned ArgReg = ArgRegs[ByVal.FirstIdx + I];
+      RegsToPass.push_back(std::make_pair(ArgReg, LoadVal));
+    }
+
+    // Return if the struct has been fully copied.
+    if (ByValSize == Offset)
+      return;
+
+    // Copy the remainder of the byval argument with sub-word loads and shifts.
+    if (LeftoverBytes) {
+      assert((ByValSize > Offset) && (ByValSize < Offset + RegSize) &&
+             "Size of the remainder should be smaller than RegSize.");
+      SDValue Val;
+
+      for (unsigned LoadSize = RegSize / 2, TotalSizeLoaded = 0;
+           Offset < ByValSize; LoadSize /= 2) {
+        unsigned RemSize = ByValSize - Offset;
+
+        if (RemSize < LoadSize)
+          continue;
+
+        // Load subword.
+        SDValue LoadPtr = DAG.getNode(ISD::ADD, DL, PtrTy, Arg,
+                                      DAG.getConstant(Offset, PtrTy));
+        SDValue LoadVal =
+          DAG.getExtLoad(ISD::ZEXTLOAD, DL, RegTy, Chain, LoadPtr,
+                         MachinePointerInfo(), MVT::getIntegerVT(LoadSize * 8),
+                         false, false, Alignment);
+        MemOpChains.push_back(LoadVal.getValue(1));
+
+        // Shift the loaded value.
+        unsigned Shamt;
+
+        if (isLittle)
+          Shamt = TotalSizeLoaded;
+        else
+          Shamt = (RegSize - (TotalSizeLoaded + LoadSize)) * 8;
+
+        SDValue Shift = DAG.getNode(ISD::SHL, DL, RegTy, LoadVal,
+                                    DAG.getConstant(Shamt, MVT::i32));
+
+        if (Val.getNode())
+          Val = DAG.getNode(ISD::OR, DL, RegTy, Val, Shift);
+        else
+          Val = Shift;
+
+        Offset += LoadSize;
+        TotalSizeLoaded += LoadSize;
+        Alignment = std::min(Alignment, LoadSize);
+      }
+
+      unsigned ArgReg = ArgRegs[ByVal.FirstIdx + I];
+      RegsToPass.push_back(std::make_pair(ArgReg, Val));
+      return;
+    }
+  }
+
+  // Copy remainder of byval arg to it with memcpy.
+  unsigned MemCpySize = ByValSize - Offset;
+  SDValue Src = DAG.getNode(ISD::ADD, DL, PtrTy, Arg,
+                            DAG.getConstant(Offset, PtrTy));
+  SDValue Dst = DAG.getNode(ISD::ADD, DL, PtrTy, StackPtr,
+                            DAG.getIntPtrConstant(ByVal.Address));
+  Chain = DAG.getMemcpy(Chain, DL, Dst, Src,
+                        DAG.getConstant(MemCpySize, PtrTy), Alignment,
+                        /*isVolatile=*/false, /*AlwaysInline=*/false,
+                        MachinePointerInfo(0), MachinePointerInfo(0));
+  MemOpChains.push_back(Chain);
+}
+
+void
+MipsTargetLowering::writeVarArgRegs(std::vector<SDValue> &OutChains,
+                                    const MipsCC &CC, SDValue Chain,
+                                    DebugLoc DL, SelectionDAG &DAG) const {
+  unsigned NumRegs = CC.numIntArgRegs();
+  const uint16_t *ArgRegs = CC.intArgRegs();
+  const CCState &CCInfo = CC.getCCInfo();
+  unsigned Idx = CCInfo.getFirstUnallocated(ArgRegs, NumRegs);
+  unsigned RegSize = CC.regSize();
+  EVT RegTy = MVT::getIntegerVT(RegSize * 8);
+  const TargetRegisterClass *RC = getRegClassFor(RegTy);
+  MachineFunction &MF = DAG.getMachineFunction();
+  MachineFrameInfo *MFI = MF.getFrameInfo();
+  MipsFunctionInfo *MipsFI = MF.getInfo<MipsFunctionInfo>();
+
+  // Offset of the first variable argument from stack pointer.
+  int VaArgOffset;
+
+  if (NumRegs == Idx)
+    VaArgOffset = RoundUpToAlignment(CCInfo.getNextStackOffset(), RegSize);
+  else
+    VaArgOffset =
+      (int)CC.reservedArgArea() - (int)(RegSize * (NumRegs - Idx));
+
+  // Record the frame index of the first variable argument
+  // which is a value necessary to VASTART.
+  int FI = MFI->CreateFixedObject(RegSize, VaArgOffset, true);
+  MipsFI->setVarArgsFrameIndex(FI);
+
+  // Copy the integer registers that have not been used for argument passing
+  // to the argument register save area. For O32, the save area is allocated
+  // in the caller's stack frame, while for N32/64, it is allocated in the
+  // callee's stack frame.
+  for (unsigned I = Idx; I < NumRegs; ++I, VaArgOffset += RegSize) {
+    unsigned Reg = AddLiveIn(MF, ArgRegs[I], RC);
+    SDValue ArgValue = DAG.getCopyFromReg(Chain, DL, Reg, RegTy);
+    FI = MFI->CreateFixedObject(RegSize, VaArgOffset, true);
+    SDValue PtrOff = DAG.getFrameIndex(FI, getPointerTy());
+    SDValue Store = DAG.getStore(Chain, DL, ArgValue, PtrOff,
+                                 MachinePointerInfo(), false, false, 0);
+    cast<StoreSDNode>(Store.getNode())->getMemOperand()->setValue(0);
+    OutChains.push_back(Store);
+  }
+}

Modified: llvm/branches/R600/lib/Target/Mips/MipsISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsISelLowering.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsISelLowering.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsISelLowering.h Tue Nov 13 09:21:47 2012
@@ -17,6 +17,7 @@
 
 #include "Mips.h"
 #include "MipsSubtarget.h"
+#include "llvm/CodeGen/CallingConvLower.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/Target/TargetLowering.h"
 
@@ -140,6 +141,7 @@
   //===--------------------------------------------------------------------===//
   // TargetLowering Implementation
   //===--------------------------------------------------------------------===//
+  class MipsFunctionInfo;
 
   class MipsTargetLowering : public TargetLowering  {
   public:
@@ -171,6 +173,69 @@
 
     virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
   private:
+
+    /// ByValArgInfo - Byval argument information.
+    struct ByValArgInfo {
+      unsigned FirstIdx; // Index of the first register used.
+      unsigned NumRegs;  // Number of registers used for this argument.
+      unsigned Address;  // Offset of the stack area used to pass this argument.
+
+      ByValArgInfo() : FirstIdx(0), NumRegs(0), Address(0) {}
+    };
+
+    /// MipsCC - This class provides methods used to analyze formal and call
+    /// arguments and inquire about calling convention information.
+    class MipsCC {
+    public:
+      MipsCC(CallingConv::ID CallConv, bool IsVarArg, bool IsO32,
+             CCState &Info);
+
+      void analyzeCallOperands(const SmallVectorImpl<ISD::OutputArg> &Outs);
+      void analyzeFormalArguments(const SmallVectorImpl<ISD::InputArg> &Ins);
+      void handleByValArg(unsigned ValNo, MVT ValVT, MVT LocVT,
+                          CCValAssign::LocInfo LocInfo,
+                          ISD::ArgFlagsTy ArgFlags);
+
+      const CCState &getCCInfo() const { return CCInfo; }
+
+      /// hasByValArg - Returns true if function has byval arguments.
+      bool hasByValArg() const { return !ByValArgs.empty(); }
+
+      /// useRegsForByval - Returns true if the calling convention allows the
+      /// use of registers to pass byval arguments.
+      bool useRegsForByval() const { return UseRegsForByval; }
+
+      /// regSize - Size (in number of bits) of integer registers.
+      unsigned regSize() const { return RegSize; }
+
+      /// numIntArgRegs - Number of integer registers available for calls.
+      unsigned numIntArgRegs() const { return NumIntArgRegs; }
+
+      /// reservedArgArea - The size of the area the caller reserves for
+      /// register arguments. This is 16-byte if ABI is O32.
+      unsigned reservedArgArea() const { return ReservedArgArea; }
+
+      /// intArgRegs - Pointer to array of integer registers.
+      const uint16_t *intArgRegs() const { return IntArgRegs; }
+
+      typedef SmallVector<ByValArgInfo, 2>::const_iterator byval_iterator;
+      byval_iterator byval_begin() const { return ByValArgs.begin(); }
+      byval_iterator byval_end() const { return ByValArgs.end(); }
+
+    private:
+      void allocateRegs(ByValArgInfo &ByVal, unsigned ByValSize,
+                        unsigned Align);
+
+      CCState &CCInfo;
+      bool UseRegsForByval;
+      unsigned RegSize;
+      unsigned NumIntArgRegs;
+      unsigned ReservedArgArea;
+      const uint16_t *IntArgRegs, *ShadowRegs;
+      SmallVector<ByValArgInfo, 2> ByValArgs;
+      llvm::CCAssignFn *FixedFn, *VarFn;
+    };
+
     // Subtarget Info
     const MipsSubtarget *Subtarget;
 
@@ -207,11 +272,37 @@
     SDValue LowerSTORE(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, SelectionDAG &DAG) const;
+    SDValue LowerADD(SDValue Op, SelectionDAG &DAG) const;
 
     /// IsEligibleForTailCallOptimization - Check whether the call is eligible
     /// for tail call optimization.
-    bool IsEligibleForTailCallOptimization(CallingConv::ID CalleeCC,
-                                           unsigned NextStackOffset) const;
+    bool IsEligibleForTailCallOptimization(const MipsCC &MipsCCInfo,
+                                           unsigned NextStackOffset,
+                                           const MipsFunctionInfo& FI) const;
+
+    /// copyByValArg - Copy argument registers which were used to pass a byval
+    /// argument to the stack. Create a stack frame object for the byval
+    /// argument.
+    void copyByValRegs(SDValue Chain, DebugLoc DL,
+                       std::vector<SDValue> &OutChains, SelectionDAG &DAG,
+                       const ISD::ArgFlagsTy &Flags,
+                       SmallVectorImpl<SDValue> &InVals,
+                       const Argument *FuncArg,
+                       const MipsCC &CC, const ByValArgInfo &ByVal) const;
+
+    /// passByValArg - Pass a byval argument in registers or on stack.
+    void passByValArg(SDValue Chain, DebugLoc DL,
+                      SmallVector<std::pair<unsigned, SDValue>, 16> &RegsToPass,
+                      SmallVector<SDValue, 8> &MemOpChains, SDValue StackPtr,
+                      MachineFrameInfo *MFI, SelectionDAG &DAG, SDValue Arg,
+                      const MipsCC &CC, const ByValArgInfo &ByVal,
+                      const ISD::ArgFlagsTy &Flags, bool isLittle) const;
+
+    /// writeVarArgRegs - Write variable function arguments passed in registers
+    /// to the stack. Also create a stack frame object for the first variable
+    /// argument.
+    void writeVarArgRegs(std::vector<SDValue> &OutChains, const MipsCC &CC,
+                         SDValue Chain, DebugLoc DL, SelectionDAG &DAG) const;
 
     virtual SDValue
       LowerFormalArguments(SDValue Chain,
@@ -220,6 +311,10 @@
                            DebugLoc dl, SelectionDAG &DAG,
                            SmallVectorImpl<SDValue> &InVals) const;
 
+    SDValue passArgOnStack(SDValue StackPtr, unsigned Offset, SDValue Chain,
+                           SDValue Arg, DebugLoc DL, bool IsTailCall,
+                           SelectionDAG &DAG) const;
+
     virtual SDValue
       LowerCall(TargetLowering::CallLoweringInfo &CLI,
                 SmallVectorImpl<SDValue> &InVals) const;

Modified: llvm/branches/R600/lib/Target/Mips/MipsInstrFPU.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsInstrFPU.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsInstrFPU.td (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsInstrFPU.td Tue Nov 13 09:21:47 2012
@@ -182,20 +182,21 @@
 defm CEIL_L  : FFR1_L_M<0xa, "ceil">;
 defm FLOOR_W : FFR1_W_M<0xf, "floor">;
 defm FLOOR_L : FFR1_L_M<0xb, "floor">;
-defm CVT_W   : FFR1_W_M<0x24, "cvt">;
+defm CVT_W   : FFR1_W_M<0x24, "cvt">, NeverHasSideEffects;
 //defm CVT_L   : FFR1_L_M<0x25, "cvt">;
 
-def CVT_S_W : FFR1<0x20, 20, "cvt", "s.w", FGR32, FGR32>;
-def CVT_L_S : FFR1<0x25, 16, "cvt", "l.s", FGR64, FGR32>;
-def CVT_L_D64: FFR1<0x25, 17, "cvt", "l.d", FGR64, FGR64>;
+def CVT_S_W : FFR1<0x20, 20, "cvt", "s.w", FGR32, FGR32>, NeverHasSideEffects;
+def CVT_L_S : FFR1<0x25, 16, "cvt", "l.s", FGR64, FGR32>, NeverHasSideEffects;
+def CVT_L_D64: FFR1<0x25, 17, "cvt", "l.d", FGR64, FGR64>, NeverHasSideEffects;
 
-let Predicates = [NotFP64bit, HasStandardEncoding] in {
+let Predicates = [NotFP64bit, HasStandardEncoding], neverHasSideEffects = 1 in {
   def CVT_S_D32 : FFR1<0x20, 17, "cvt", "s.d", FGR32, AFGR64>;
   def CVT_D32_W : FFR1<0x21, 20, "cvt", "d.w", AFGR64, FGR32>;
   def CVT_D32_S : FFR1<0x21, 16, "cvt", "d.s", AFGR64, FGR32>;
 }
 
-let Predicates = [IsFP64bit, HasStandardEncoding], DecoderNamespace = "Mips64" in {
+let Predicates = [IsFP64bit, HasStandardEncoding], DecoderNamespace = "Mips64",
+    neverHasSideEffects = 1 in {
  def CVT_S_D64 : FFR1<0x20, 17, "cvt", "s.d", FGR32, FGR64>;
  def CVT_S_L   : FFR1<0x20, 21, "cvt", "s.l", FGR32, FGR64>;
  def CVT_D64_W : FFR1<0x21, 20, "cvt", "d.w", FGR64, FGR32>;

Modified: llvm/branches/R600/lib/Target/Mips/MipsInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsInstrInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsInstrInfo.td (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsInstrInfo.td Tue Nov 13 09:21:47 2012
@@ -200,6 +200,14 @@
   bit isCodeGenOnly = 1;
 }
 
+class IsAsCheapAsAMove {
+  bit isAsCheapAsAMove = 1;
+}
+
+class NeverHasSideEffects {
+  bit neverHasSideEffects = 1;
+}
+
 //===----------------------------------------------------------------------===//
 // Instruction format superclass
 //===----------------------------------------------------------------------===//
@@ -417,7 +425,7 @@
 // Load Upper Imediate
 class LoadUpper<bits<6> op, string instr_asm, RegisterClass RC, Operand Imm>:
   FI<op, (outs RC:$rt), (ins Imm:$imm16),
-     !strconcat(instr_asm, "\t$rt, $imm16"), [], IIAlu> {
+     !strconcat(instr_asm, "\t$rt, $imm16"), [], IIAlu>, IsAsCheapAsAMove {
   let rs = 0;
   let neverHasSideEffects = 1;
   let isReMaterializable = 1;
@@ -925,7 +933,8 @@
 //===----------------------------------------------------------------------===//
 
 /// Arithmetic Instructions (ALU Immediate)
-def ADDiu   : ArithLogicI<0x09, "addiu", add, simm16, immSExt16, CPURegs>;
+def ADDiu   : ArithLogicI<0x09, "addiu", add, simm16, immSExt16, CPURegs>,
+              IsAsCheapAsAMove;
 def ADDi    : ArithOverflowI<0x08, "addi", add, simm16, immSExt16, CPURegs>;
 def SLTi    : SetCC_I<0x0a, "slti", setlt, simm16, immSExt16, CPURegs>;
 def SLTiu   : SetCC_I<0x0b, "sltiu", setult, simm16, immSExt16, CPURegs>;

Modified: llvm/branches/R600/lib/Target/Mips/MipsLongBranch.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsLongBranch.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsLongBranch.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsLongBranch.cpp Tue Nov 13 09:21:47 2012
@@ -424,8 +424,6 @@
 
   // Compute basic block addresses.
   if (TM.getRelocationModel() == Reloc::PIC_) {
-    MF->getInfo<MipsFunctionInfo>()->setEmitNOAT();
-
     uint64_t Address = 0;
 
     for (I = MBBInfos.begin(); I != E; Address += I->Size, ++I)

Modified: llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.cpp Tue Nov 13 09:21:47 2012
@@ -43,4 +43,17 @@
   return GlobalBaseReg = MF.getRegInfo().createVirtualRegister(RC);
 }
 
+bool MipsFunctionInfo::mips16SPAliasRegSet() const {
+  return Mips16SPAliasReg;
+}
+unsigned MipsFunctionInfo::getMips16SPAliasReg() {
+  // Return if it has already been initialized.
+  if (Mips16SPAliasReg)
+    return Mips16SPAliasReg;
+
+  const TargetRegisterClass *RC;
+  RC=(const TargetRegisterClass*)&Mips::CPU16RegsRegClass;
+  return Mips16SPAliasReg = MF.getRegInfo().createVirtualRegister(RC);
+}
+
 void MipsFunctionInfo::anchor() { }

Modified: llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsMachineFunction.h Tue Nov 13 09:21:47 2012
@@ -39,38 +39,45 @@
   /// relocation models.
   unsigned GlobalBaseReg;
 
+  /// Mips16SPAliasReg - keeps track of the virtual register initialized for
+  /// use as an alias for SP for use in load/store of halfword/byte from/to
+  /// the stack
+  unsigned Mips16SPAliasReg;
+
   /// VarArgsFrameIndex - FrameIndex for start of varargs area.
   int VarArgsFrameIndex;
 
-  // Range of frame object indices.
-  // InArgFIRange: Range of indices of all frame objects created during call to
-  //               LowerFormalArguments.
-  std::pair<int, int> InArgFIRange;
+  /// True if function has a byval argument.
+  bool HasByvalArg;
 
-  bool EmitNOAT;
+  /// Size of incoming argument area.
+  unsigned IncomingArgSize;
 
 public:
   MipsFunctionInfo(MachineFunction& MF)
-  : MF(MF), SRetReturnReg(0), GlobalBaseReg(0),
-    VarArgsFrameIndex(0), InArgFIRange(std::make_pair(-1, 0)), EmitNOAT(false)
+   : MF(MF), SRetReturnReg(0), GlobalBaseReg(0), Mips16SPAliasReg(0),
+     VarArgsFrameIndex(0)
   {}
 
-  bool isInArgFI(int FI) const {
-    return FI <= InArgFIRange.first && FI >= InArgFIRange.second;
-  }
-  void setLastInArgFI(int FI) { InArgFIRange.second = FI; }
-
   unsigned getSRetReturnReg() const { return SRetReturnReg; }
   void setSRetReturnReg(unsigned Reg) { SRetReturnReg = Reg; }
 
   bool globalBaseRegSet() const;
   unsigned getGlobalBaseReg();
 
+  bool mips16SPAliasRegSet() const;
+  unsigned getMips16SPAliasReg();
+
   int getVarArgsFrameIndex() const { return VarArgsFrameIndex; }
   void setVarArgsFrameIndex(int Index) { VarArgsFrameIndex = Index; }
 
-  bool getEmitNOAT() const { return EmitNOAT; }
-  void setEmitNOAT() { EmitNOAT = true; }
+  bool hasByvalArg() const { return HasByvalArg; }
+  void setFormalArgInfo(unsigned Size, bool HasByval) {
+    IncomingArgSize = Size;
+    HasByvalArg = HasByval;
+  }
+
+  unsigned getIncomingArgSize() const { return IncomingArgSize; }
 };
 
 } // end of namespace llvm

Modified: llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.cpp Tue Nov 13 09:21:47 2012
@@ -81,11 +81,11 @@
 BitVector MipsRegisterInfo::
 getReservedRegs(const MachineFunction &MF) const {
   static const uint16_t ReservedCPURegs[] = {
-    Mips::ZERO, Mips::AT, Mips::K0, Mips::K1, Mips::SP
+    Mips::ZERO, Mips::K0, Mips::K1, Mips::SP
   };
 
   static const uint16_t ReservedCPU64Regs[] = {
-    Mips::ZERO_64, Mips::AT_64, Mips::K0_64, Mips::K1_64, Mips::SP_64
+    Mips::ZERO_64, Mips::K0_64, Mips::K1_64, Mips::SP_64
   };
 
   BitVector Reserved(getNumRegs());
@@ -94,29 +94,28 @@
   for (unsigned I = 0; I < array_lengthof(ReservedCPURegs); ++I)
     Reserved.set(ReservedCPURegs[I]);
 
-  if (Subtarget.hasMips64()) {
-    for (unsigned I = 0; I < array_lengthof(ReservedCPU64Regs); ++I)
-      Reserved.set(ReservedCPU64Regs[I]);
+  for (unsigned I = 0; I < array_lengthof(ReservedCPU64Regs); ++I)
+    Reserved.set(ReservedCPU64Regs[I]);
 
+  if (Subtarget.hasMips64()) {
     // Reserve all registers in AFGR64.
     for (RegIter Reg = Mips::AFGR64RegClass.begin(),
          EReg = Mips::AFGR64RegClass.end(); Reg != EReg; ++Reg)
       Reserved.set(*Reg);
   } else {
-    // Reserve all registers in CPU64Regs & FGR64.
-    for (RegIter Reg = Mips::CPU64RegsRegClass.begin(),
-         EReg = Mips::CPU64RegsRegClass.end(); Reg != EReg; ++Reg)
-      Reserved.set(*Reg);
-
+    // Reserve all registers in FGR64.
     for (RegIter Reg = Mips::FGR64RegClass.begin(),
          EReg = Mips::FGR64RegClass.end(); Reg != EReg; ++Reg)
       Reserved.set(*Reg);
   }
-
   // Reserve FP if this function should have a dedicated frame pointer register.
   if (MF.getTarget().getFrameLowering()->hasFP(MF)) {
-    Reserved.set(Mips::FP);
-    Reserved.set(Mips::FP_64);
+    if (Subtarget.inMips16Mode())
+      Reserved.set(Mips::S0);
+    else {
+      Reserved.set(Mips::FP);
+      Reserved.set(Mips::FP_64);
+    }
   }
 
   // Reserve hardware registers.
@@ -186,8 +185,12 @@
   const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
   bool IsN64 = Subtarget.isABI_N64();
 
-  return TFI->hasFP(MF) ? (IsN64 ? Mips::FP_64 : Mips::FP) :
-                          (IsN64 ? Mips::SP_64 : Mips::SP);
+  if (Subtarget.inMips16Mode())
+    return TFI->hasFP(MF) ? Mips::S0 : Mips::SP;
+  else
+    return TFI->hasFP(MF) ? (IsN64 ? Mips::FP_64 : Mips::FP) :
+                            (IsN64 ? Mips::SP_64 : Mips::SP);
+
 }
 
 unsigned MipsRegisterInfo::

Modified: llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.td (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsRegisterInfo.td Tue Nov 13 09:21:47 2012
@@ -73,7 +73,7 @@
 let Namespace = "Mips" in {
   // General Purpose Registers
   def ZERO : MipsGPRReg< 0, "zero">, DwarfRegNum<[0]>;
-  def AT   : MipsGPRReg< 1, "at">,   DwarfRegNum<[1]>;
+  def AT   : MipsGPRReg< 1, "1">,    DwarfRegNum<[1]>;
   def V0   : MipsGPRReg< 2, "2">,    DwarfRegNum<[2]>;
   def V1   : MipsGPRReg< 3, "3">,    DwarfRegNum<[3]>;
   def A0   : MipsGPRReg< 4, "4">,    DwarfRegNum<[4]>;
@@ -107,7 +107,7 @@
 
   // General Purpose 64-bit Registers
   def ZERO_64 : Mips64GPRReg< 0, "zero", [ZERO]>, DwarfRegNum<[0]>;
-  def AT_64   : Mips64GPRReg< 1, "at",   [AT]>, DwarfRegNum<[1]>;
+  def AT_64   : Mips64GPRReg< 1, "1",    [AT]>, DwarfRegNum<[1]>;
   def V0_64   : Mips64GPRReg< 2, "2",    [V0]>, DwarfRegNum<[2]>;
   def V1_64   : Mips64GPRReg< 3, "3",    [V1]>, DwarfRegNum<[3]>;
   def A0_64   : Mips64GPRReg< 4, "4",    [A0]>, DwarfRegNum<[4]>;

Modified: llvm/branches/R600/lib/Target/Mips/MipsSEFrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsSEFrameLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsSEFrameLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsSEFrameLowering.cpp Tue Nov 13 09:21:47 2012
@@ -22,6 +22,7 @@
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/RegisterScavenging.h"
 #include "llvm/DataLayout.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/Support/CommandLine.h"
@@ -202,6 +203,19 @@
   // Mark $fp as used if function has dedicated frame pointer.
   if (hasFP(MF))
     MRI.setPhysRegUsed(FP);
+
+  // Set scavenging frame index if necessary.
+  uint64_t MaxSPOffset = MF.getInfo<MipsFunctionInfo>()->getIncomingArgSize() +
+    estimateStackSize(MF);
+
+  if (isInt<16>(MaxSPOffset))
+    return;
+
+  const TargetRegisterClass *RC = STI.isABI_N64() ?
+    &Mips::CPU64RegsRegClass : &Mips::CPURegsRegClass;
+  int FI = MF.getFrameInfo()->CreateStackObject(RC->getSize(),
+                                                RC->getAlignment(), false);
+  RS->setScavengingFrameIndex(FI);
 }
 
 const MipsFrameLowering *

Modified: llvm/branches/R600/lib/Target/Mips/MipsSEInstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsSEInstrInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsSEInstrInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsSEInstrInfo.cpp Tue Nov 13 09:21:47 2012
@@ -260,9 +260,8 @@
   if (isInt<16>(Amount))// addi sp, sp, amount
     BuildMI(MBB, I, DL, get(ADDiu), SP).addReg(SP).addImm(Amount);
   else { // Expand immediate that doesn't fit in 16-bit.
-    MBB.getParent()->getInfo<MipsFunctionInfo>()->setEmitNOAT();
     unsigned Reg = loadImmediate(Amount, MBB, I, DL, 0);
-    BuildMI(MBB, I, DL, get(ADDu), SP).addReg(SP).addReg(Reg);
+    BuildMI(MBB, I, DL, get(ADDu), SP).addReg(SP).addReg(Reg, RegState::Kill);
   }
 }
 
@@ -274,10 +273,12 @@
                                unsigned *NewImm) const {
   MipsAnalyzeImmediate AnalyzeImm;
   const MipsSubtarget &STI = TM.getSubtarget<MipsSubtarget>();
+  MachineRegisterInfo &RegInfo = MBB.getParent()->getRegInfo();
   unsigned Size = STI.isABI_N64() ? 64 : 32;
   unsigned LUi = STI.isABI_N64() ? Mips::LUi64 : Mips::LUi;
   unsigned ZEROReg = STI.isABI_N64() ? Mips::ZERO_64 : Mips::ZERO;
-  unsigned ATReg = STI.isABI_N64() ? Mips::AT_64 : Mips::AT;
+  const TargetRegisterClass *RC = STI.isABI_N64() ?
+    &Mips::CPU64RegsRegClass : &Mips::CPURegsRegClass;
   bool LastInstrIsADDiu = NewImm;
 
   const MipsAnalyzeImmediate::InstSeq &Seq =
@@ -289,22 +290,23 @@
   // The first instruction can be a LUi, which is different from other
   // instructions (ADDiu, ORI and SLL) in that it does not have a register
   // operand.
+  unsigned Reg = RegInfo.createVirtualRegister(RC);
+
   if (Inst->Opc == LUi)
-    BuildMI(MBB, II, DL, get(LUi), ATReg)
-      .addImm(SignExtend64<16>(Inst->ImmOpnd));
+    BuildMI(MBB, II, DL, get(LUi), Reg).addImm(SignExtend64<16>(Inst->ImmOpnd));
   else
-    BuildMI(MBB, II, DL, get(Inst->Opc), ATReg).addReg(ZEROReg)
+    BuildMI(MBB, II, DL, get(Inst->Opc), Reg).addReg(ZEROReg)
       .addImm(SignExtend64<16>(Inst->ImmOpnd));
 
   // Build the remaining instructions in Seq.
   for (++Inst; Inst != Seq.end() - LastInstrIsADDiu; ++Inst)
-    BuildMI(MBB, II, DL, get(Inst->Opc), ATReg).addReg(ATReg)
+    BuildMI(MBB, II, DL, get(Inst->Opc), Reg).addReg(Reg, RegState::Kill)
       .addImm(SignExtend64<16>(Inst->ImmOpnd));
 
   if (LastInstrIsADDiu)
     *NewImm = Inst->ImmOpnd;
 
-  return ATReg;
+  return Reg;
 }
 
 unsigned MipsSEInstrInfo::GetAnalyzableBrOpc(unsigned Opc) const {

Modified: llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.cpp Tue Nov 13 09:21:47 2012
@@ -26,6 +26,7 @@
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
+#include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/Target/TargetFrameLowering.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetOptions.h"
@@ -43,6 +44,16 @@
                                        const MipsSEInstrInfo &I)
   : MipsRegisterInfo(ST), TII(I) {}
 
+bool MipsSERegisterInfo::
+requiresRegisterScavenging(const MachineFunction &MF) const {
+  return true;
+}
+
+bool MipsSERegisterInfo::
+requiresFrameIndexScavenging(const MachineFunction &MF) const {
+  return true;
+}
+
 // This function eliminate ADJCALLSTACKDOWN,
 // ADJCALLSTACKUP pseudo instructions
 void MipsSERegisterInfo::
@@ -72,7 +83,6 @@
   MachineInstr &MI = *II;
   MachineFunction &MF = *MI.getParent()->getParent();
   MachineFrameInfo *MFI = MF.getFrameInfo();
-  MipsFunctionInfo *MipsFI = MF.getInfo<MipsFunctionInfo>();
 
   const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
   int MinCSFI = 0;
@@ -103,6 +113,7 @@
   // - If the frame object is any of the following, its offset must be adjusted
   //   by adding the size of the stack:
   //   incoming argument, callee-saved register location or local variable.
+  bool IsKill = false;
   int64_t Offset;
 
   Offset = SPOffset + (int64_t)StackSize;
@@ -116,17 +127,17 @@
     MachineBasicBlock &MBB = *MI.getParent();
     DebugLoc DL = II->getDebugLoc();
     unsigned ADDu = Subtarget.isABI_N64() ? Mips::DADDu : Mips::ADDu;
-    unsigned ATReg = Subtarget.isABI_N64() ? Mips::AT_64 : Mips::AT;
     unsigned NewImm;
 
-    MipsFI->setEmitNOAT();
     unsigned Reg = TII.loadImmediate(Offset, MBB, II, DL, &NewImm);
-    BuildMI(MBB, II, DL, TII.get(ADDu), ATReg).addReg(FrameReg).addReg(Reg);
+    BuildMI(MBB, II, DL, TII.get(ADDu), Reg).addReg(FrameReg)
+      .addReg(Reg, RegState::Kill);
 
-    FrameReg = ATReg;
+    FrameReg = Reg;
     Offset = SignExtend64<16>(NewImm);
+    IsKill = true;
   }
 
-  MI.getOperand(OpNo).ChangeToRegister(FrameReg, false);
+  MI.getOperand(OpNo).ChangeToRegister(FrameReg, false, false, IsKill);
   MI.getOperand(OpNo + 1).ChangeToImmediate(Offset);
 }

Modified: llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsSERegisterInfo.h Tue Nov 13 09:21:47 2012
@@ -27,6 +27,10 @@
   MipsSERegisterInfo(const MipsSubtarget &Subtarget,
                      const MipsSEInstrInfo &TII);
 
+  bool requiresRegisterScavenging(const MachineFunction &MF) const;
+
+  bool requiresFrameIndexScavenging(const MachineFunction &MF) const;
+
   void eliminateCallFramePseudoInstr(MachineFunction &MF,
                                      MachineBasicBlock &MBB,
                                      MachineBasicBlock::iterator I) const;

Modified: llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -53,7 +53,7 @@
     InstrInfo(MipsInstrInfo::create(*this)),
     FrameLowering(MipsFrameLowering::create(*this, Subtarget)),
     TLInfo(*this), TSInfo(*this), JITInfo(),
-    ELFWriterInfo(false, isLittle), STTI(&TLInfo) {
+    STTI(&TLInfo), VTTI(&TLInfo) {
 }
 
 void MipsebTargetMachine::anchor() { }

Modified: llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.h (original)
+++ llvm/branches/R600/lib/Target/Mips/MipsTargetMachine.h Tue Nov 13 09:21:47 2012
@@ -20,7 +20,6 @@
 #include "MipsJITInfo.h"
 #include "MipsSelectionDAGInfo.h"
 #include "MipsSubtarget.h"
-#include "MipsELFWriterInfo.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/DataLayout.h"
 #include "llvm/Target/TargetFrameLowering.h"
@@ -38,9 +37,8 @@
   MipsTargetLowering  TLInfo;
   MipsSelectionDAGInfo TSInfo;
   MipsJITInfo JITInfo;
-  MipsELFWriterInfo   ELFWriterInfo;
   ScalarTargetTransformImpl STTI;
-  VectorTargetTransformInfo VTTI; 
+  VectorTargetTransformImpl VTTI;
 
 public:
   MipsTargetMachine(const Target &T, StringRef TT,
@@ -74,9 +72,6 @@
     return &TSInfo;
   }
 
-  virtual const MipsELFWriterInfo *getELFWriterInfo() const {
-    return &ELFWriterInfo;
-  }
   virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const {
     return &STTI;
   }

Modified: llvm/branches/R600/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/NVPTX/NVPTXAsmPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/NVPTX/NVPTXAsmPrinter.cpp (original)
+++ llvm/branches/R600/lib/Target/NVPTX/NVPTXAsmPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -126,10 +126,8 @@
       return Base;
 
     // Truncate/sext the offset to the pointer size.
-    unsigned AS = PtrVal->getType()->isPointerTy() ?
-      cast<PointerType>(PtrVal->getType())->getAddressSpace() : 0;
-    if (TD.getPointerSizeInBits(AS) != 64) {
-      int SExtAmount = 64-TD.getPointerSizeInBits(AS);
+    if (TD.getPointerSizeInBits() != 64) {
+      int SExtAmount = 64-TD.getPointerSizeInBits();
       Offset = (Offset << SExtAmount) >> SExtAmount;
     }
 
@@ -1380,7 +1378,7 @@
 
   const FunctionType *FTy = dyn_cast<FunctionType>(Ty);
   if (FTy)
-    return TD->getPointerPrefAlignment(0);
+    return TD->getPointerPrefAlignment();
   return TD->getPrefTypeAlignment(Ty);
 }
 

Modified: llvm/branches/R600/lib/Target/NVPTX/NVPTXTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/NVPTX/NVPTXTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/NVPTX/NVPTXTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/NVPTX/NVPTXTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -73,7 +73,7 @@
   Subtarget(TT, CPU, FS, is64bit),
   DL(Subtarget.getDataLayout()),
   InstrInfo(*this), TLInfo(*this), TSInfo(*this), FrameLowering(*this,is64bit),
-  STTI(&TLInfo)
+  STTI(&TLInfo), VTTI(&TLInfo)
 /*FrameInfo(TargetFrameInfo::StackGrowsUp, 8, 0)*/ {
 }
 

Modified: llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp Tue Nov 13 09:21:47 2012
@@ -29,9 +29,14 @@
   case FK_Data_1:
   case FK_Data_2:
   case FK_Data_4:
+  case FK_Data_8:
+  case PPC::fixup_ppc_toc:
     return Value;
+  case PPC::fixup_ppc_lo14:
+  case PPC::fixup_ppc_toc16_ds:
+    return (Value & 0xffff) << 2;
   case PPC::fixup_ppc_brcond14:
-    return Value & 0x3ffc;
+    return Value & 0xfffc;
   case PPC::fixup_ppc_br24:
     return Value & 0x3fffffc;
 #if 0
@@ -41,6 +46,7 @@
   case PPC::fixup_ppc_ha16:
     return ((Value >> 16) + ((Value & 0x8000) ? 1 : 0)) & 0xffff;
   case PPC::fixup_ppc_lo16:
+  case PPC::fixup_ppc_toc16:
     return Value & 0xffff;
   }
 }
@@ -72,7 +78,10 @@
       { "fixup_ppc_brcond14",    16,     14,   MCFixupKindInfo::FKF_IsPCRel },
       { "fixup_ppc_lo16",        16,     16,   0 },
       { "fixup_ppc_ha16",        16,     16,   0 },
-      { "fixup_ppc_lo14",        16,     14,   0 }
+      { "fixup_ppc_lo14",        16,     14,   0 },
+      { "fixup_ppc_toc",          0,     64,   0 },
+      { "fixup_ppc_toc16",       16,     16,   0 },
+      { "fixup_ppc_toc16_ds",    16,     14,   0 }
     };
 
     if (Kind < FirstTargetFixupKind)

Modified: llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp Tue Nov 13 09:21:47 2012
@@ -11,6 +11,8 @@
 #include "MCTargetDesc/PPCMCTargetDesc.h"
 #include "llvm/MC/MCELFObjectWriter.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCValue.h"
 
 using namespace llvm;
 
@@ -21,9 +23,15 @@
 
     virtual ~PPCELFObjectWriter();
   protected:
+    virtual unsigned getRelocTypeInner(const MCValue &Target,
+                                       const MCFixup &Fixup,
+                                       bool IsPCRel) const;
     virtual unsigned GetRelocType(const MCValue &Target, const MCFixup &Fixup,
                                   bool IsPCRel, bool IsRelocWithSymbol,
                                   int64_t Addend) const;
+    virtual const MCSymbol *undefinedExplicitRelSym(const MCValue &Target,
+                                                    const MCFixup &Fixup,
+                                                    bool IsPCRel) const;
     virtual void adjustFixupOffset(const MCFixup &Fixup, uint64_t &RelocOffset);
   };
 }
@@ -36,11 +44,13 @@
 PPCELFObjectWriter::~PPCELFObjectWriter() {
 }
 
-unsigned PPCELFObjectWriter::GetRelocType(const MCValue &Target,
-                                             const MCFixup &Fixup,
-                                             bool IsPCRel,
-                                             bool IsRelocWithSymbol,
-                                             int64_t Addend) const {
+unsigned PPCELFObjectWriter::getRelocTypeInner(const MCValue &Target,
+                                               const MCFixup &Fixup,
+                                               bool IsPCRel) const
+{
+  MCSymbolRefExpr::VariantKind Modifier = Target.isAbsolute() ?
+    MCSymbolRefExpr::VK_None : Target.getSymA()->getKind();
+
   // determine the type of the relocation
   unsigned Type;
   if (IsPCRel) {
@@ -61,7 +71,7 @@
       Type = ELF::R_PPC_ADDR24;
       break;
     case PPC::fixup_ppc_brcond14:
-      Type = ELF::R_PPC_ADDR14_BRTAKEN; // XXX: or BRNTAKEN?_
+      Type = ELF::R_PPC_ADDR14; // XXX: or BRNTAKEN?_
       break;
     case PPC::fixup_ppc_ha16:
       Type = ELF::R_PPC_ADDR16_HA;
@@ -72,6 +82,26 @@
     case PPC::fixup_ppc_lo14:
       Type = ELF::R_PPC_ADDR14;
       break;
+    case PPC::fixup_ppc_toc:
+      Type = ELF::R_PPC64_TOC;
+      break;
+    case PPC::fixup_ppc_toc16:
+      Type = ELF::R_PPC64_TOC16;
+      break;
+    case PPC::fixup_ppc_toc16_ds:
+      Type = ELF::R_PPC64_TOC16_DS;
+      break;
+    case FK_Data_8:
+      switch (Modifier) {
+      default: llvm_unreachable("Unsupported Modifier");
+      case MCSymbolRefExpr::VK_PPC_TOC:
+        Type = ELF::R_PPC64_TOC;
+        break;
+      case MCSymbolRefExpr::VK_None:
+        Type = ELF::R_PPC64_ADDR64;
+	break;
+      }
+      break;
     case FK_Data_4:
       Type = ELF::R_PPC_ADDR32;
       break;
@@ -83,11 +113,41 @@
   return Type;
 }
 
+unsigned PPCELFObjectWriter::GetRelocType(const MCValue &Target,
+                                          const MCFixup &Fixup,
+                                          bool IsPCRel,
+                                          bool IsRelocWithSymbol,
+                                          int64_t Addend) const {
+  return getRelocTypeInner(Target, Fixup, IsPCRel);
+}
+
+const MCSymbol *PPCELFObjectWriter::undefinedExplicitRelSym(const MCValue &Target,
+                                                            const MCFixup &Fixup,
+                                                            bool IsPCRel) const {
+  assert(Target.getSymA() && "SymA cannot be 0");
+  const MCSymbol &Symbol = Target.getSymA()->getSymbol().AliasedSymbol();
+
+  unsigned RelocType = getRelocTypeInner(Target, Fixup, IsPCRel);
+
+  // The .odp creation emits a relocation against the symbol ".TOC." which
+  // create a R_PPC64_TOC relocation. However the relocation symbol name
+  // in final object creation should be NULL, since the symbol does not
+  // really exist, it is just the reference to TOC base for the current
+  // object file.
+  bool EmitThisSym = RelocType != ELF::R_PPC64_TOC;
+
+  if (EmitThisSym && !Symbol.isTemporary())
+    return &Symbol;
+  return NULL;
+}
+
 void PPCELFObjectWriter::
 adjustFixupOffset(const MCFixup &Fixup, uint64_t &RelocOffset) {
   switch ((unsigned)Fixup.getKind()) {
     case PPC::fixup_ppc_ha16:
     case PPC::fixup_ppc_lo16:
+    case PPC::fixup_ppc_toc16:
+    case PPC::fixup_ppc_toc16_ds:
       RelocOffset += 2;
       break;
     default:

Modified: llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h (original)
+++ llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h Tue Nov 13 09:21:47 2012
@@ -34,6 +34,16 @@
   /// fixup_ppc_lo14 - A 14-bit fixup corresponding to lo16(_foo) for instrs
   /// like 'std'.
   fixup_ppc_lo14,
+
+  /// fixup_ppc_toc - Insert value of TOC base (.TOC.).
+  fixup_ppc_toc,
+
+  /// fixup_ppc_toc16 - A 16-bit signed fixup relative to the TOC base.
+  fixup_ppc_toc16,
+
+  /// fixup_ppc_toc16_ds - A 14-bit signed fixup relative to the TOC base with
+  /// implied 2 zero bits
+  fixup_ppc_toc16_ds,
   
   // Marker
   LastTargetFixupKind,

Modified: llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp Tue Nov 13 09:21:47 2012
@@ -15,7 +15,9 @@
 #include "MCTargetDesc/PPCBaseInfo.h"
 #include "MCTargetDesc/PPCFixupKinds.h"
 #include "llvm/MC/MCCodeEmitter.h"
+#include "llvm/MC/MCSubtargetInfo.h"
 #include "llvm/MC/MCInst.h"
+#include "llvm/MC/MCInstrInfo.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -28,13 +30,25 @@
   PPCMCCodeEmitter(const PPCMCCodeEmitter &) LLVM_DELETED_FUNCTION;
   void operator=(const PPCMCCodeEmitter &) LLVM_DELETED_FUNCTION;
 
+  const MCSubtargetInfo &STI;
+  Triple TT;
+
 public:
   PPCMCCodeEmitter(const MCInstrInfo &mcii, const MCSubtargetInfo &sti,
-                   MCContext &ctx) {
+                   MCContext &ctx)
+    : STI(sti), TT(STI.getTargetTriple()) {
   }
   
   ~PPCMCCodeEmitter() {}
 
+  bool is64BitMode() const {
+    return (STI.getFeatureBits() & PPC::Feature64Bit) != 0;
+  }
+
+  bool isSVR4ABI() const {
+    return TT.isMacOSX() == 0;
+  }
+
   unsigned getDirectBrEncoding(const MCInst &MI, unsigned OpNo,
                                SmallVectorImpl<MCFixup> &Fixups) const;
   unsigned getCondBrEncoding(const MCInst &MI, unsigned OpNo,
@@ -61,11 +75,19 @@
                                  SmallVectorImpl<MCFixup> &Fixups) const;
   void EncodeInstruction(const MCInst &MI, raw_ostream &OS,
                          SmallVectorImpl<MCFixup> &Fixups) const {
-    unsigned Bits = getBinaryCodeForInstr(MI, Fixups);
+    uint64_t Bits = getBinaryCodeForInstr(MI, Fixups);
+
+    // BL8_NOPELF and BLA8_NOP_ELF is both size of 8 bacause of the
+    // following 'nop'.
+    unsigned Size = 4; // FIXME: Have Desc.getSize() return the correct value!
+    unsigned Opcode = MI.getOpcode();
+    if (Opcode == PPC::BL8_NOP_ELF || Opcode == PPC::BLA8_NOP_ELF)
+      Size = 8;
     
     // Output the constant in big endian byte order.
-    for (unsigned i = 0; i != 4; ++i) {
-      OS << (char)(Bits >> 24);
+    int ShiftValue = (Size * 8) - 8;
+    for (unsigned i = 0; i != Size; ++i) {
+      OS << (char)(Bits >> ShiftValue);
       Bits <<= 8;
     }
     
@@ -140,8 +162,12 @@
     return (getMachineOpValue(MI, MO, Fixups) & 0xFFFF) | RegBits;
   
   // Add a fixup for the displacement field.
-  Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
-                                   (MCFixupKind)PPC::fixup_ppc_lo16));
+  if (isSVR4ABI() && is64BitMode())
+    Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
+                                     (MCFixupKind)PPC::fixup_ppc_toc16));
+  else
+    Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
+                                     (MCFixupKind)PPC::fixup_ppc_lo16));
   return RegBits;
 }
 
@@ -158,8 +184,12 @@
     return (getMachineOpValue(MI, MO, Fixups) & 0x3FFF) | RegBits;
   
   // Add a fixup for the branch target.
-  Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
-                                   (MCFixupKind)PPC::fixup_ppc_lo14));
+  if (isSVR4ABI() && is64BitMode())
+    Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
+                                     (MCFixupKind)PPC::fixup_ppc_toc16_ds));
+  else
+    Fixups.push_back(MCFixup::Create(0, MO.getExpr(),
+                                     (MCFixupKind)PPC::fixup_ppc_lo14));
   return RegBits;
 }
 

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCAsmPrinter.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCAsmPrinter.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCAsmPrinter.cpp Tue Nov 13 09:21:47 2012
@@ -284,8 +284,22 @@
                                           unsigned AsmVariant,
                                           const char *ExtraCode,
                                           raw_ostream &O) {
-  if (ExtraCode && ExtraCode[0])
-    return true; // Unknown modifier.
+  if (ExtraCode && ExtraCode[0]) {
+    if (ExtraCode[1] != 0) return true; // Unknown modifier.
+
+    switch (ExtraCode[0]) {
+    default: return true;  // Unknown modifier.
+    case 'y': // A memory reference for an X-form instruction
+      {
+        const char *RegName = "r0";
+        if (!Subtarget.isDarwin()) RegName = stripRegisterPrefix(RegName);
+        O << RegName << ", ";
+        printOperand(MI, OpNo, O);
+        return false;
+      }
+    }
+  }
+
   assert(MI->getOperand(OpNo).isReg());
   O << "0(";
   printOperand(MI, OpNo, O);
@@ -420,10 +434,14 @@
   OutStreamer.EmitValueToAlignment(8);
   MCSymbol *Symbol1 = 
     OutContext.GetOrCreateSymbol(".L." + Twine(CurrentFnSym->getName()));
-  MCSymbol *Symbol2 = OutContext.GetOrCreateSymbol(StringRef(".TOC. at tocbase"));
+  // Generates a R_PPC64_ADDR64 (from FK_DATA_8) relocation for the function
+  // entry point.
   OutStreamer.EmitValue(MCSymbolRefExpr::Create(Symbol1, OutContext),
                         8/*size*/, 0/*addrspace*/);
-  OutStreamer.EmitValue(MCSymbolRefExpr::Create(Symbol2, OutContext),
+  MCSymbol *Symbol2 = OutContext.GetOrCreateSymbol(StringRef(".TOC."));
+  // Generates a R_PPC64_TOC relocation for TOC base insertion.
+  OutStreamer.EmitValue(MCSymbolRefExpr::Create(Symbol2,
+                        MCSymbolRefExpr::VK_PPC_TOC, OutContext),
                         8/*size*/, 0/*addrspace*/);
   // Emit a null environment pointer.
   OutStreamer.EmitIntValue(0, 8 /* size */, 0 /* addrspace */);
@@ -439,7 +457,7 @@
 bool PPCLinuxAsmPrinter::doFinalization(Module &M) {
   const DataLayout *TD = TM.getDataLayout();
 
-  bool isPPC64 = TD->getPointerSizeInBits(0) == 64;
+  bool isPPC64 = TD->getPointerSizeInBits() == 64;
 
   if (isPPC64 && !TOC.empty()) {
     const MCSectionELF *Section = OutStreamer.getContext().getELFSection(".toc",
@@ -545,7 +563,7 @@
 
 void PPCDarwinAsmPrinter::
 EmitFunctionStubs(const MachineModuleInfoMachO::SymbolListTy &Stubs) {
-  bool isPPC64 = TM.getDataLayout()->getPointerSizeInBits(0) == 64;
+  bool isPPC64 = TM.getDataLayout()->getPointerSizeInBits() == 64;
   
   const TargetLoweringObjectFileMachO &TLOFMacho = 
     static_cast<const TargetLoweringObjectFileMachO &>(getObjFileLowering());
@@ -640,7 +658,7 @@
 
 
 bool PPCDarwinAsmPrinter::doFinalization(Module &M) {
-  bool isPPC64 = TM.getDataLayout()->getPointerSizeInBits(0) == 64;
+  bool isPPC64 = TM.getDataLayout()->getPointerSizeInBits() == 64;
 
   // Darwin/PPC always uses mach-o.
   const TargetLoweringObjectFileMachO &TLOFMacho = 

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCCallingConv.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCCallingConv.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCCallingConv.td (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCCallingConv.td Tue Nov 13 09:21:47 2012
@@ -12,12 +12,19 @@
 //
 //===----------------------------------------------------------------------===//
 
+/// CCIfSubtarget - Match if the current subtarget has a feature F.
+class CCIfSubtarget<string F, CCAction A>
+ : CCIf<!strconcat("State.getTarget().getSubtarget<PPCSubtarget>().", F), A>;
+
 //===----------------------------------------------------------------------===//
 // Return Value Calling Convention
 //===----------------------------------------------------------------------===//
 
 // Return-value convention for PowerPC
 def RetCC_PPC : CallingConv<[
+  // On PPC64, integer return values are always promoted to i64
+  CCIfType<[i32], CCIfSubtarget<"isPPC64()", CCPromoteToType<i64>>>,
+
   CCIfType<[i32], CCAssignToReg<[R3, R4, R5, R6, R7, R8, R9, R10]>>,
   CCIfType<[i64], CCAssignToReg<[X3, X4, X5, X6]>>,
   

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCISelDAGToDAG.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCISelDAGToDAG.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCISelDAGToDAG.cpp Tue Nov 13 09:21:47 2012
@@ -623,6 +623,88 @@
   }
 }
 
+// getVCmpInst: return the vector compare instruction for the specified
+// vector type and condition code. Since this is for altivec specific code,
+// only support the altivec types (v16i8, v8i16, v4i32, and v4f32).
+static unsigned int getVCmpInst(MVT::SimpleValueType VecVT, ISD::CondCode CC) {
+  switch (CC) {
+    case ISD::SETEQ:
+    case ISD::SETUEQ:
+    case ISD::SETNE:
+    case ISD::SETUNE:
+      if (VecVT == MVT::v16i8)
+        return PPC::VCMPEQUB;
+      else if (VecVT == MVT::v8i16)
+        return PPC::VCMPEQUH;
+      else if (VecVT == MVT::v4i32)
+        return PPC::VCMPEQUW;
+      // v4f32 != v4f32 could be translate to unordered not equal
+      else if (VecVT == MVT::v4f32)
+        return PPC::VCMPEQFP;
+      break;
+    case ISD::SETLT:
+    case ISD::SETGT:
+    case ISD::SETLE:
+    case ISD::SETGE:
+      if (VecVT == MVT::v16i8)
+        return PPC::VCMPGTSB;
+      else if (VecVT == MVT::v8i16)
+        return PPC::VCMPGTSH;
+      else if (VecVT == MVT::v4i32)
+        return PPC::VCMPGTSW;
+      else if (VecVT == MVT::v4f32)
+        return PPC::VCMPGTFP;
+      break;
+    case ISD::SETULT:
+    case ISD::SETUGT:
+    case ISD::SETUGE:
+    case ISD::SETULE:
+      if (VecVT == MVT::v16i8)
+        return PPC::VCMPGTUB;
+      else if (VecVT == MVT::v8i16)
+        return PPC::VCMPGTUH;
+      else if (VecVT == MVT::v4i32)
+        return PPC::VCMPGTUW;
+      break;
+    case ISD::SETOEQ:
+      if (VecVT == MVT::v4f32)
+        return PPC::VCMPEQFP;
+      break;
+    case ISD::SETOLT:
+    case ISD::SETOGT:
+    case ISD::SETOLE:
+      if (VecVT == MVT::v4f32)
+        return PPC::VCMPGTFP;
+      break;
+    case ISD::SETOGE:
+      if (VecVT == MVT::v4f32)
+        return PPC::VCMPGEFP;
+      break;
+    default:
+      break;
+  }
+  llvm_unreachable("Invalid integer vector compare condition");
+}
+
+// getVCmpEQInst: return the equal compare instruction for the specified vector
+// type. Since this is for altivec specific code, only support the altivec
+// types (v16i8, v8i16, v4i32, and v4f32).
+static unsigned int getVCmpEQInst(MVT::SimpleValueType VecVT) {
+  switch (VecVT) {
+    case MVT::v16i8:
+      return PPC::VCMPEQUB;
+    case MVT::v8i16:
+      return PPC::VCMPEQUH;
+    case MVT::v4i32:
+      return PPC::VCMPEQUW;
+    case MVT::v4f32:
+      return PPC::VCMPEQFP;
+    default:
+      llvm_unreachable("Invalid integer vector compare condition");
+  }
+}
+
+
 SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) {
   DebugLoc dl = N->getDebugLoc();
   unsigned Imm;
@@ -706,20 +788,58 @@
   SDValue LHS = N->getOperand(0);
   SDValue RHS = N->getOperand(1);
 
-  // Altivec Vector compare instructions do not set any CR register by default
+  // Altivec Vector compare instructions do not set any CR register by default and
+  // vector compare operations return the same type as the operands.
   if (LHS.getValueType().isVector()) {
-    unsigned int Opc;
-    if (LHS.getValueType() == MVT::v16i8)
-      Opc = PPC::VCMPEQUB;
-    else if (LHS.getValueType() == MVT::v4i32)
-      Opc = PPC::VCMPEQUW;
-    else if (LHS.getValueType() == MVT::v8i16)
-      Opc = PPC::VCMPEQUH;
-    else if (LHS.getValueType() == MVT::v4f32)
-      Opc = PPC::VCMPEQFP;
-    else
-      llvm_unreachable("Invalid vector compare type: should be expanded by legalize");
-    return CurDAG->SelectNodeTo(N, Opc, LHS.getValueType(), LHS, RHS);
+    EVT VecVT = LHS.getValueType();
+    MVT::SimpleValueType VT = VecVT.getSimpleVT().SimpleTy;
+    unsigned int VCmpInst = getVCmpInst(VT, CC);
+
+    switch (CC) {
+      case ISD::SETEQ:
+      case ISD::SETOEQ:
+      case ISD::SETUEQ:
+        return CurDAG->SelectNodeTo(N, VCmpInst, VecVT, LHS, RHS);
+      case ISD::SETNE:
+      case ISD::SETONE:
+      case ISD::SETUNE: {
+        SDValue VCmp(CurDAG->getMachineNode(VCmpInst, dl, VecVT, LHS, RHS), 0);
+        return CurDAG->SelectNodeTo(N, PPC::VNOR, VecVT, VCmp, VCmp);
+      } 
+      case ISD::SETLT:
+      case ISD::SETOLT:
+      case ISD::SETULT:
+        return CurDAG->SelectNodeTo(N, VCmpInst, VecVT, RHS, LHS);
+      case ISD::SETGT:
+      case ISD::SETOGT:
+      case ISD::SETUGT:
+        return CurDAG->SelectNodeTo(N, VCmpInst, VecVT, LHS, RHS);
+      case ISD::SETGE:
+      case ISD::SETOGE:
+      case ISD::SETUGE: {
+        // Small optimization: Altivec provides a 'Vector Compare Greater Than
+        // or Equal To' instruction (vcmpgefp), so in this case there is no
+        // need for extra logic for the equal compare.
+        if (VecVT.getSimpleVT().isFloatingPoint()) {
+          return CurDAG->SelectNodeTo(N, VCmpInst, VecVT, LHS, RHS);
+        } else {
+          SDValue VCmpGT(CurDAG->getMachineNode(VCmpInst, dl, VecVT, LHS, RHS), 0);
+          unsigned int VCmpEQInst = getVCmpEQInst(VT);
+          SDValue VCmpEQ(CurDAG->getMachineNode(VCmpEQInst, dl, VecVT, LHS, RHS), 0);
+          return CurDAG->SelectNodeTo(N, PPC::VOR, VecVT, VCmpGT, VCmpEQ);
+        }
+      }
+      case ISD::SETLE:
+      case ISD::SETOLE:
+      case ISD::SETULE: {
+        SDValue VCmpLE(CurDAG->getMachineNode(VCmpInst, dl, VecVT, RHS, LHS), 0);
+        unsigned int VCmpEQInst = getVCmpEQInst(VT);
+        SDValue VCmpEQ(CurDAG->getMachineNode(VCmpEQInst, dl, VecVT, LHS, RHS), 0);
+        return CurDAG->SelectNodeTo(N, PPC::VOR, VecVT, VCmpLE, VCmpEQ);
+      }
+      default:
+        llvm_unreachable("Invalid vector compare type: should be expanded by legalize");
+    }
   }
 
   bool Inv;

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.cpp Tue Nov 13 09:21:47 2012
@@ -361,6 +361,22 @@
       setOperationAction(ISD::CTLZ_ZERO_UNDEF, VT, Expand);
       setOperationAction(ISD::CTTZ, VT, Expand);
       setOperationAction(ISD::CTTZ_ZERO_UNDEF, VT, Expand);
+      setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Expand);
+
+      for (unsigned j = (unsigned)MVT::FIRST_VECTOR_VALUETYPE;
+           j <= (unsigned)MVT::LAST_VECTOR_VALUETYPE; ++j) {
+        MVT::SimpleValueType InnerVT = (MVT::SimpleValueType)j;
+        setTruncStoreAction(VT, InnerVT, Expand);
+      }
+      setLoadExtAction(ISD::SEXTLOAD, VT, Expand);
+      setLoadExtAction(ISD::ZEXTLOAD, VT, Expand);
+      setLoadExtAction(ISD::EXTLOAD, VT, Expand);
+    }
+
+    for (unsigned i = (unsigned)MVT::FIRST_FP_VECTOR_VALUETYPE;
+         i <= (unsigned)MVT::LAST_FP_VECTOR_VALUETYPE; ++i) {
+      MVT::SimpleValueType VT = (MVT::SimpleValueType)i;
+      setOperationAction(ISD::FSQRT, VT, Expand);
     }
 
     // We can custom expand all VECTOR_SHUFFLEs to VPERM, others we can handle
@@ -396,6 +412,14 @@
     setOperationAction(ISD::BUILD_VECTOR, MVT::v8i16, Custom);
     setOperationAction(ISD::BUILD_VECTOR, MVT::v4i32, Custom);
     setOperationAction(ISD::BUILD_VECTOR, MVT::v4f32, Custom);
+
+    // Altivec does not contain unordered floating-point compare instructions
+    setCondCodeAction(ISD::SETUO, MVT::v4f32, Expand);
+    setCondCodeAction(ISD::SETUEQ, MVT::v4f32, Expand);
+    setCondCodeAction(ISD::SETUGT, MVT::v4f32, Expand);
+    setCondCodeAction(ISD::SETUGE, MVT::v4f32, Expand);
+    setCondCodeAction(ISD::SETULT, MVT::v4f32, Expand);
+    setCondCodeAction(ISD::SETULE, MVT::v4f32, Expand);
   }
 
   if (Subtarget->has64BitSupport()) {
@@ -1953,6 +1977,48 @@
   return Chain;
 }
 
+// PPC64 passes i8, i16, and i32 values in i64 registers. Promote
+// value to MVT::i64 and then truncate to the correct register size.
+SDValue
+PPCTargetLowering::extendArgForPPC64(ISD::ArgFlagsTy Flags, EVT ObjectVT,
+                                     SelectionDAG &DAG, SDValue ArgVal,
+                                     DebugLoc dl) const {
+  if (Flags.isSExt())
+    ArgVal = DAG.getNode(ISD::AssertSext, dl, MVT::i64, ArgVal,
+                         DAG.getValueType(ObjectVT));
+  else if (Flags.isZExt())
+    ArgVal = DAG.getNode(ISD::AssertZext, dl, MVT::i64, ArgVal,
+                         DAG.getValueType(ObjectVT));
+  
+  return DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, ArgVal);
+}
+
+// Set the size that is at least reserved in caller of this function.  Tail
+// call optimized functions' reserved stack space needs to be aligned so that
+// taking the difference between two stack areas will result in an aligned
+// stack.
+void
+PPCTargetLowering::setMinReservedArea(MachineFunction &MF, SelectionDAG &DAG,
+                                      unsigned nAltivecParamsAtEnd,
+                                      unsigned MinReservedArea,
+                                      bool isPPC64) const {
+  PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
+  // Add the Altivec parameters at the end, if needed.
+  if (nAltivecParamsAtEnd) {
+    MinReservedArea = ((MinReservedArea+15)/16)*16;
+    MinReservedArea += 16*nAltivecParamsAtEnd;
+  }
+  MinReservedArea =
+    std::max(MinReservedArea,
+             PPCFrameLowering::getMinCallFrameSize(isPPC64, true));
+  unsigned TargetAlign
+    = DAG.getMachineFunction().getTarget().getFrameLowering()->
+        getStackAlignment();
+  unsigned AlignMask = TargetAlign-1;
+  MinReservedArea = (MinReservedArea + AlignMask) & ~AlignMask;
+  FI->setMinReservedArea(MinReservedArea);
+}
+
 SDValue
 PPCTargetLowering::LowerFormalArguments_64SVR4(
                                       SDValue Chain,
@@ -2034,6 +2100,19 @@
       // ObjSize is the true size, ArgSize rounded up to multiple of registers.
       ObjSize = Flags.getByValSize();
       ArgSize = ((ObjSize + PtrByteSize - 1)/PtrByteSize) * PtrByteSize;
+      // Empty aggregate parameters do not take up registers.  Examples:
+      //   struct { } a;
+      //   union  { } b;
+      //   int c[0];
+      // etc.  However, we have to provide a place-holder in InVals, so
+      // pretend we have an 8-byte item at the current address for that
+      // purpose.
+      if (!ObjSize) {
+        int FI = MFI->CreateFixedObject(PtrByteSize, ArgOffset, true);
+        SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
+        InVals.push_back(FIN);
+        continue;
+      }
       // All aggregates smaller than 8 bytes must be passed right-justified.
       if (ObjSize < PtrByteSize)
         CurArgOffset = CurArgOffset + (PtrByteSize - ObjSize);
@@ -2041,25 +2120,42 @@
       int FI = MFI->CreateFixedObject(ObjSize, CurArgOffset, true);
       SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
       InVals.push_back(FIN);
-      if (ObjSize==1 || ObjSize==2 || ObjSize==4) {
+
+      if (ObjSize < 8) {
         if (GPR_idx != Num_GPR_Regs) {
-          unsigned VReg;
-          VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass);
+          unsigned VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass);
           SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT);
-          EVT ObjType = (ObjSize == 1 ? MVT::i8 :
-                         (ObjSize == 2 ? MVT::i16 : MVT::i32));
-          SDValue Store = DAG.getTruncStore(Val.getValue(1), dl, Val, FIN,
-                                            MachinePointerInfo(FuncArg,
-                                              CurArgOffset),
-                                            ObjType, false, false, 0);
+          SDValue Store;
+
+          if (ObjSize==1 || ObjSize==2 || ObjSize==4) {
+            EVT ObjType = (ObjSize == 1 ? MVT::i8 :
+                           (ObjSize == 2 ? MVT::i16 : MVT::i32));
+            Store = DAG.getTruncStore(Val.getValue(1), dl, Val, FIN,
+                                      MachinePointerInfo(FuncArg, CurArgOffset),
+                                      ObjType, false, false, 0);
+          } else {
+            // For sizes that don't fit a truncating store (3, 5, 6, 7),
+            // store the whole register as-is to the parameter save area
+            // slot.  The address of the parameter was already calculated
+            // above (InVals.push_back(FIN)) to be the right-justified
+            // offset within the slot.  For this store, we need a new
+            // frame index that points at the beginning of the slot.
+            int FI = MFI->CreateFixedObject(PtrByteSize, ArgOffset, true);
+            SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
+            Store = DAG.getStore(Val.getValue(1), dl, Val, FIN,
+                                 MachinePointerInfo(FuncArg, ArgOffset),
+                                 false, false, 0);
+          }
+
           MemOps.push_back(Store);
           ++GPR_idx;
         }
-
+        // Whether we copied from a register or not, advance the offset
+        // into the parameter save area by a full doubleword.
         ArgOffset += PtrByteSize;
-
         continue;
       }
+
       for (unsigned j = 0; j < ArgSize; j += PtrByteSize) {
         // Store whatever pieces of the object are in registers
         // to memory.  ArgOffset will be the address of the beginning
@@ -2070,16 +2166,7 @@
           int FI = MFI->CreateFixedObject(PtrByteSize, ArgOffset, true);
           SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
           SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT);
-          SDValue Shifted = Val;
-
-          // For 64-bit SVR4, small structs come in right-adjusted.
-          // Shift them left so the following logic works as expected.
-          if (ObjSize < 8) {
-            SDValue ShiftAmt = DAG.getConstant(64 - 8 * ObjSize, PtrVT);
-            Shifted = DAG.getNode(ISD::SHL, dl, PtrVT, Val, ShiftAmt);
-          }
-
-          SDValue Store = DAG.getStore(Val.getValue(1), dl, Shifted, FIN,
+          SDValue Store = DAG.getStore(Val.getValue(1), dl, Val, FIN,
                                        MachinePointerInfo(FuncArg, ArgOffset),
                                        false, false, 0);
           MemOps.push_back(Store);
@@ -2101,18 +2188,10 @@
         unsigned VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass);
         ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i64);
 
-        if (ObjectVT == MVT::i32) {
+        if (ObjectVT == MVT::i32)
           // PPC64 passes i8, i16, and i32 values in i64 registers. Promote
           // value to MVT::i64 and then truncate to the correct register size.
-          if (Flags.isSExt())
-            ArgVal = DAG.getNode(ISD::AssertSext, dl, MVT::i64, ArgVal,
-                                 DAG.getValueType(ObjectVT));
-          else if (Flags.isZExt())
-            ArgVal = DAG.getNode(ISD::AssertZext, dl, MVT::i64, ArgVal,
-                                 DAG.getValueType(ObjectVT));
-
-          ArgVal = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, ArgVal);
-        }
+          ArgVal = extendArgForPPC64(Flags, ObjectVT, DAG, ArgVal, dl);
 
         ++GPR_idx;
       } else {
@@ -2190,24 +2269,10 @@
   }
 
   // Set the size that is at least reserved in caller of this function.  Tail
-  // call optimized function's reserved stack space needs to be aligned so that
+  // call optimized functions' reserved stack space needs to be aligned so that
   // taking the difference between two stack areas will result in an aligned
   // stack.
-  PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
-  // Add the Altivec parameters at the end, if needed.
-  if (nAltivecParamsAtEnd) {
-    MinReservedArea = ((MinReservedArea+15)/16)*16;
-    MinReservedArea += 16*nAltivecParamsAtEnd;
-  }
-  MinReservedArea =
-    std::max(MinReservedArea,
-             PPCFrameLowering::getMinCallFrameSize(true, true));
-  unsigned TargetAlign
-    = DAG.getMachineFunction().getTarget().getFrameLowering()->
-        getStackAlignment();
-  unsigned AlignMask = TargetAlign-1;
-  MinReservedArea = (MinReservedArea + AlignMask) & ~AlignMask;
-  FI->setMinReservedArea(MinReservedArea);
+  setMinReservedArea(MF, DAG, nAltivecParamsAtEnd, MinReservedArea, true);
 
   // If the function takes variable number of arguments, make a frame index for
   // the start of the first vararg value... for expansion of llvm.va_start.
@@ -2215,8 +2280,7 @@
     int Depth = ArgOffset;
 
     FuncInfo->setVarArgsFrameIndex(
-      MFI->CreateFixedObject(PtrVT.getSizeInBits()/8,
-                             Depth, true));
+      MFI->CreateFixedObject(PtrByteSize, Depth, true));
     SDValue FIN = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT);
 
     // If this function is vararg, store any remaining integer argument regs
@@ -2229,7 +2293,7 @@
                                    MachinePointerInfo(), false, false, 0);
       MemOps.push_back(Store);
       // Increment the address by four for the next argument to store
-      SDValue PtrOff = DAG.getConstant(PtrVT.getSizeInBits()/8, PtrVT);
+      SDValue PtrOff = DAG.getConstant(PtrByteSize, PtrVT);
       FIN = DAG.getNode(ISD::ADD, dl, PtrOff.getValueType(), FIN, PtrOff);
     }
   }
@@ -2394,8 +2458,7 @@
           else
             VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::GPRCRegClass);
           SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT);
-          EVT ObjType = (ObjSize == 1 ? MVT::i8 :
-                         (ObjSize == 2 ? MVT::i16 : MVT::i32));
+          EVT ObjType = ObjSize == 1 ? MVT::i8 : MVT::i16;
           SDValue Store = DAG.getTruncStore(Val.getValue(1), dl, Val, FIN,
                                             MachinePointerInfo(FuncArg,
                                               CurArgOffset),
@@ -2457,18 +2520,10 @@
         unsigned VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass);
         ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i64);
 
-        if (ObjectVT == MVT::i32) {
+        if (ObjectVT == MVT::i32)
           // PPC64 passes i8, i16, and i32 values in i64 registers. Promote
           // value to MVT::i64 and then truncate to the correct register size.
-          if (Flags.isSExt())
-            ArgVal = DAG.getNode(ISD::AssertSext, dl, MVT::i64, ArgVal,
-                                 DAG.getValueType(ObjectVT));
-          else if (Flags.isZExt())
-            ArgVal = DAG.getNode(ISD::AssertZext, dl, MVT::i64, ArgVal,
-                                 DAG.getValueType(ObjectVT));
-
-          ArgVal = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, ArgVal);
-        }
+          ArgVal = extendArgForPPC64(Flags, ObjectVT, DAG, ArgVal, dl);
 
         ++GPR_idx;
       } else {
@@ -2555,23 +2610,10 @@
   }
 
   // Set the size that is at least reserved in caller of this function.  Tail
-  // call optimized function's reserved stack space needs to be aligned so that
+  // call optimized functions' reserved stack space needs to be aligned so that
   // taking the difference between two stack areas will result in an aligned
   // stack.
-  PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
-  // Add the Altivec parameters at the end, if needed.
-  if (nAltivecParamsAtEnd) {
-    MinReservedArea = ((MinReservedArea+15)/16)*16;
-    MinReservedArea += 16*nAltivecParamsAtEnd;
-  }
-  MinReservedArea =
-    std::max(MinReservedArea,
-             PPCFrameLowering::getMinCallFrameSize(isPPC64, true));
-  unsigned TargetAlign = DAG.getMachineFunction().getTarget().getFrameLowering()->
-    getStackAlignment();
-  unsigned AlignMask = TargetAlign-1;
-  MinReservedArea = (MinReservedArea + AlignMask) & ~AlignMask;
-  FI->setMinReservedArea(MinReservedArea);
+  setMinReservedArea(MF, DAG, nAltivecParamsAtEnd, MinReservedArea, isPPC64);
 
   // If the function takes variable number of arguments, make a frame index for
   // the start of the first vararg value... for expansion of llvm.va_start.
@@ -3021,7 +3063,7 @@
       // Thus for a call through a function pointer, the following actions need
       // to be performed:
       //   1. Save the TOC of the caller in the TOC save area of its stack
-      //      frame (this is done in LowerCall_Darwin_Or_64SVR4()).
+      //      frame (this is done in LowerCall_Darwin() or LowerCall_64SVR4()).
       //   2. Load the address of the function entry point from the function
       //      descriptor.
       //   3. Load the TOC of the callee from the function descriptor into r2.
@@ -3135,12 +3177,32 @@
   // Copy all of the result registers out of their specified physreg.
   for (unsigned i = 0, e = RVLocs.size(); i != e; ++i) {
     CCValAssign &VA = RVLocs[i];
-    EVT VT = VA.getValVT();
     assert(VA.isRegLoc() && "Can only return in registers!");
-    Chain = DAG.getCopyFromReg(Chain, dl,
-                               VA.getLocReg(), VT, InFlag).getValue(1);
-    InVals.push_back(Chain.getValue(0));
-    InFlag = Chain.getValue(2);
+
+    SDValue Val = DAG.getCopyFromReg(Chain, dl,
+                                     VA.getLocReg(), VA.getLocVT(), InFlag);
+    Chain = Val.getValue(1);
+    InFlag = Val.getValue(2);
+
+    switch (VA.getLocInfo()) {
+    default: llvm_unreachable("Unknown loc info!");
+    case CCValAssign::Full: break;
+    case CCValAssign::AExt:
+      Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val);
+      break;
+    case CCValAssign::ZExt:
+      Val = DAG.getNode(ISD::AssertZext, dl, VA.getLocVT(), Val,
+                        DAG.getValueType(VA.getValVT()));
+      Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val);
+      break;
+    case CCValAssign::SExt:
+      Val = DAG.getNode(ISD::AssertSext, dl, VA.getLocVT(), Val,
+                        DAG.getValueType(VA.getValVT()));
+      Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val);
+      break;
+    }
+
+    InVals.push_back(Val);
   }
 
   return Chain;
@@ -3271,14 +3333,20 @@
     isTailCall = IsEligibleForTailCallOptimization(Callee, CallConv, isVarArg,
                                                    Ins, DAG);
 
-  if (PPCSubTarget.isSVR4ABI() && !PPCSubTarget.isPPC64())
-    return LowerCall_32SVR4(Chain, Callee, CallConv, isVarArg,
-                            isTailCall, Outs, OutVals, Ins,
-                            dl, DAG, InVals);
-
-  return LowerCall_Darwin_Or_64SVR4(Chain, Callee, CallConv, isVarArg,
-                                    isTailCall, Outs, OutVals, Ins,
-                                    dl, DAG, InVals);
+  if (PPCSubTarget.isSVR4ABI()) {
+    if (PPCSubTarget.isPPC64())
+      return LowerCall_64SVR4(Chain, Callee, CallConv, isVarArg,
+                              isTailCall, Outs, OutVals, Ins,
+                              dl, DAG, InVals);
+    else
+      return LowerCall_32SVR4(Chain, Callee, CallConv, isVarArg,
+                              isTailCall, Outs, OutVals, Ins,
+                              dl, DAG, InVals);
+  }
+
+  return LowerCall_Darwin(Chain, Callee, CallConv, isVarArg,
+                          isTailCall, Outs, OutVals, Ins,
+                          dl, DAG, InVals);
 }
 
 SDValue
@@ -3494,8 +3562,27 @@
                     Ins, InVals);
 }
 
+// Copy an argument into memory, being careful to do this outside the
+// call sequence for the call to which the argument belongs.
 SDValue
-PPCTargetLowering::LowerCall_Darwin_Or_64SVR4(SDValue Chain, SDValue Callee,
+PPCTargetLowering::createMemcpyOutsideCallSeq(SDValue Arg, SDValue PtrOff,
+                                              SDValue CallSeqStart,
+                                              ISD::ArgFlagsTy Flags,
+                                              SelectionDAG &DAG,
+                                              DebugLoc dl) const {
+  SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, PtrOff,
+                        CallSeqStart.getNode()->getOperand(0),
+                        Flags, DAG, dl);
+  // The MEMCPY must go outside the CALLSEQ_START..END.
+  SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall,
+                             CallSeqStart.getNode()->getOperand(1));
+  DAG.ReplaceAllUsesWith(CallSeqStart.getNode(),
+                         NewCallSeqStart.getNode());
+  return NewCallSeqStart;
+}
+
+SDValue
+PPCTargetLowering::LowerCall_64SVR4(SDValue Chain, SDValue Callee,
                                     CallingConv::ID CallConv, bool isVarArg,
                                     bool isTailCall,
                                     const SmallVectorImpl<ISD::OutputArg> &Outs,
@@ -3504,13 +3591,10 @@
                                     DebugLoc dl, SelectionDAG &DAG,
                                     SmallVectorImpl<SDValue> &InVals) const {
 
-  bool isSVR4ABI = PPCSubTarget.isSVR4ABI();
-
-  unsigned NumOps  = Outs.size();
+  unsigned NumOps = Outs.size();
 
   EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy();
-  bool isPPC64 = PtrVT == MVT::i64;
-  unsigned PtrByteSize = isPPC64 ? 8 : 4;
+  unsigned PtrByteSize = 8;
 
   MachineFunction &MF = DAG.getMachineFunction();
 
@@ -3526,12 +3610,13 @@
   unsigned nAltivecParamsAtEnd = 0;
 
   // Count how many bytes are to be pushed on the stack, including the linkage
-  // area, and parameter passing area.  We start with 24/48 bytes, which is
-  // prereserved space for [SP][CR][LR][3 x unused].
+  // area, and parameter passing area.  We start with at least 48 bytes, which
+  // is reserved space for [SP][CR][LR][3 x unused].
+  // NOTE: For PPC64, nAltivecParamsAtEnd always remains zero as a result
+  // of this call.
   unsigned NumBytes =
-    CalculateParameterAndLinkageAreaSize(DAG, isPPC64, isVarArg, CallConv,
-                                         Outs, OutVals,
-                                         nAltivecParamsAtEnd);
+    CalculateParameterAndLinkageAreaSize(DAG, true, isVarArg, CallConv,
+                                         Outs, OutVals, nAltivecParamsAtEnd);
 
   // Calculate by how many bytes the stack has to be adjusted in case of tail
   // call optimization.
@@ -3556,24 +3641,16 @@
   // Set up a copy of the stack pointer for use loading and storing any
   // arguments that may not fit in the registers available for argument
   // passing.
-  SDValue StackPtr;
-  if (isPPC64)
-    StackPtr = DAG.getRegister(PPC::X1, MVT::i64);
-  else
-    StackPtr = DAG.getRegister(PPC::R1, MVT::i32);
+  SDValue StackPtr = DAG.getRegister(PPC::X1, MVT::i64);
 
   // Figure out which arguments are going to go in registers, and which in
   // memory.  Also, if this is a vararg function, floating point operations
   // must be stored to our stack, and loaded into integer regs as well, if
   // any integer regs are available for argument passing.
-  unsigned ArgOffset = PPCFrameLowering::getLinkageSize(isPPC64, true);
+  unsigned ArgOffset = PPCFrameLowering::getLinkageSize(true, true);
   unsigned GPR_idx = 0, FPR_idx = 0, VR_idx = 0;
 
-  static const uint16_t GPR_32[] = {           // 32-bit registers.
-    PPC::R3, PPC::R4, PPC::R5, PPC::R6,
-    PPC::R7, PPC::R8, PPC::R9, PPC::R10,
-  };
-  static const uint16_t GPR_64[] = {           // 64-bit registers.
+  static const uint16_t GPR[] = {
     PPC::X3, PPC::X4, PPC::X5, PPC::X6,
     PPC::X7, PPC::X8, PPC::X9, PPC::X10,
   };
@@ -3583,12 +3660,10 @@
     PPC::V2, PPC::V3, PPC::V4, PPC::V5, PPC::V6, PPC::V7, PPC::V8,
     PPC::V9, PPC::V10, PPC::V11, PPC::V12, PPC::V13
   };
-  const unsigned NumGPRs = array_lengthof(GPR_32);
+  const unsigned NumGPRs = array_lengthof(GPR);
   const unsigned NumFPRs = 13;
   const unsigned NumVRs  = array_lengthof(VR);
 
-  const uint16_t *GPR = isPPC64 ? GPR_64 : GPR_32;
-
   SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass;
   SmallVector<TailCallArgumentInfo, 8> TailCallArguments;
 
@@ -3605,8 +3680,8 @@
 
     PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff);
 
-    // On PPC64, promote integers to 64-bit values.
-    if (isPPC64 && Arg.getValueType() == MVT::i32) {
+    // Promote integers to 64-bit values.
+    if (Arg.getValueType() == MVT::i32) {
       // FIXME: Should this use ANY_EXTEND if neither sext nor zext?
       unsigned ExtOp = Flags.isSExt() ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
       Arg = DAG.getNode(ExtOp, dl, MVT::i64, Arg);
@@ -3620,14 +3695,16 @@
       //   struct x { short a; char b; }
       // will have Size = 4.  With #pragma pack(1), it will have Size = 3.
       // These are the proper values we need for right-justifying the
-      // aggregate in a parameter register for 64-bit SVR4.
+      // aggregate in a parameter register.
       unsigned Size = Flags.getByValSize();
-      // FOR DARWIN ONLY:  Very small objects are passed right-justified.
-      // Everything else is passed left-justified.
-      // FOR 64-BIT SVR4:  All aggregates smaller than 8 bytes must
-      // be passed right-justified.
-      if (Size==1 || Size==2 ||
-          (Size==4 && isSVR4ABI)) {
+
+      // An empty aggregate parameter takes up no storage and no
+      // registers.
+      if (Size == 0)
+        continue;
+
+      // All aggregates smaller than 8 bytes must be passed right-justified.
+      if (Size==1 || Size==2 || Size==4) {
         EVT VT = (Size==1) ? MVT::i8 : ((Size==2) ? MVT::i16 : MVT::i32);
         if (GPR_idx != NumGPRs) {
           SDValue Load = DAG.getExtLoad(ISD::EXTLOAD, dl, PtrVT, Chain, Arg,
@@ -3637,39 +3714,17 @@
           RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load));
 
           ArgOffset += PtrByteSize;
-        } else {
-          SDValue Const = DAG.getConstant(PtrByteSize - Size,
-                                          PtrOff.getValueType());
-          SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const);
-          SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, AddPtr,
-                                CallSeqStart.getNode()->getOperand(0),
-                                Flags, DAG, dl);
-          // The MEMCPY must go outside the CALLSEQ_START..END.
-          SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall,
-                               CallSeqStart.getNode()->getOperand(1));
-          DAG.ReplaceAllUsesWith(CallSeqStart.getNode(),
-                                 NewCallSeqStart.getNode());
-          Chain = CallSeqStart = NewCallSeqStart;
-          ArgOffset += PtrByteSize;
+          continue;
         }
-        continue;
-      } else if (isSVR4ABI && GPR_idx == NumGPRs && Size < 8) {
-        // Case: Size is 3, 5, 6, or 7 for SVR4 and we're out of registers.
-        // This is the same case as 1, 2, and 4 for SVR4 with no registers.
-        // FIXME: Separate into 64-bit SVR4 and Darwin versions of this
-        // function, and combine the duplicated code chunks.
+      }
+
+      if (GPR_idx == NumGPRs && Size < 8) {
         SDValue Const = DAG.getConstant(PtrByteSize - Size,
                                         PtrOff.getValueType());
         SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const);
-        SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, AddPtr,
-                              CallSeqStart.getNode()->getOperand(0),
-                              Flags, DAG, dl);
-        // The MEMCPY must go outside the CALLSEQ_START..END.
-        SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall,
-                                    CallSeqStart.getNode()->getOperand(1));
-        DAG.ReplaceAllUsesWith(CallSeqStart.getNode(),
-                               NewCallSeqStart.getNode());
-        Chain = CallSeqStart = NewCallSeqStart;
+        Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, AddPtr,
+                                                          CallSeqStart,
+                                                          Flags, DAG, dl);
         ArgOffset += PtrByteSize;
         continue;
       }
@@ -3678,29 +3733,21 @@
       // registers.  (This is not what the doc says.)
 
       // FIXME: The above statement is likely due to a misunderstanding of the
-      // documents.  At least for 64-bit SVR4, all arguments must be copied
-      // into the parameter area BY THE CALLEE in the event that the callee
-      // takes the address of any formal argument.  That has not yet been
-      // implemented.  However, it is reasonable to use the stack area as a
-      // staging area for the register load.
-
-      // Skip this for small aggregates under 64-bit SVR4, as we will use
-      // the same slot for a right-justified copy, below.
-      if (Size >= 8 || !isSVR4ABI) {
-        SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, PtrOff,
-                              CallSeqStart.getNode()->getOperand(0),
-                              Flags, DAG, dl);
-        // This must go outside the CALLSEQ_START..END.
-        SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall,
-                                   CallSeqStart.getNode()->getOperand(1));
-        DAG.ReplaceAllUsesWith(CallSeqStart.getNode(),
-                               NewCallSeqStart.getNode());
-        Chain = CallSeqStart = NewCallSeqStart;
-      }
-
-      // FOR 64-BIT SVR4:  When a register is available, pass the
-      // aggregate right-justified.
-      if (isSVR4ABI && Size < 8 && GPR_idx != NumGPRs) {
+      // documents.  All arguments must be copied into the parameter area BY
+      // THE CALLEE in the event that the callee takes the address of any
+      // formal argument.  That has not yet been implemented.  However, it is
+      // reasonable to use the stack area as a staging area for the register
+      // load.
+
+      // Skip this for small aggregates, as we will use the same slot for a
+      // right-justified copy, below.
+      if (Size >= 8)
+        Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, PtrOff,
+                                                          CallSeqStart,
+                                                          Flags, DAG, dl);
+
+      // When a register is available, pass a small aggregate right-justified.
+      if (Size < 8 && GPR_idx != NumGPRs) {
         // The easiest way to get this right-justified in a register
         // is to copy the structure into the rightmost portion of a
         // local variable slot, then load the whole slot into the
@@ -3711,16 +3758,9 @@
         // parameter save area instead of a new local variable.
         SDValue Const = DAG.getConstant(8 - Size, PtrOff.getValueType());
         SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const);
-        SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, AddPtr,
-                              CallSeqStart.getNode()->getOperand(0),
-                              Flags, DAG, dl);
-
-        // Place the memcpy outside the CALLSEQ_START..END.
-        SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall,
-                                   CallSeqStart.getNode()->getOperand(1));
-        DAG.ReplaceAllUsesWith(CallSeqStart.getNode(), 
-                               NewCallSeqStart.getNode());
-        Chain = CallSeqStart = NewCallSeqStart;
+        Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, AddPtr,
+                                                          CallSeqStart,
+                                                          Flags, DAG, dl);
 
         // Load the slot into the register.
         SDValue Load = DAG.getLoad(PtrVT, dl, Chain, PtrOff,
@@ -3734,6 +3774,335 @@
         continue;
       }
 
+      // For aggregates larger than PtrByteSize, copy the pieces of the
+      // object that fit into registers from the parameter save area.
+      for (unsigned j=0; j<Size; j+=PtrByteSize) {
+        SDValue Const = DAG.getConstant(j, PtrOff.getValueType());
+        SDValue AddArg = DAG.getNode(ISD::ADD, dl, PtrVT, Arg, Const);
+        if (GPR_idx != NumGPRs) {
+          SDValue Load = DAG.getLoad(PtrVT, dl, Chain, AddArg,
+                                     MachinePointerInfo(),
+                                     false, false, false, 0);
+          MemOpChains.push_back(Load.getValue(1));
+          RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load));
+          ArgOffset += PtrByteSize;
+        } else {
+          ArgOffset += ((Size - j + PtrByteSize-1)/PtrByteSize)*PtrByteSize;
+          break;
+        }
+      }
+      continue;
+    }
+
+    switch (Arg.getValueType().getSimpleVT().SimpleTy) {
+    default: llvm_unreachable("Unexpected ValueType for argument!");
+    case MVT::i32:
+    case MVT::i64:
+      if (GPR_idx != NumGPRs) {
+        RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Arg));
+      } else {
+        LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset,
+                         true, isTailCall, false, MemOpChains,
+                         TailCallArguments, dl);
+      }
+      ArgOffset += PtrByteSize;
+      break;
+    case MVT::f32:
+    case MVT::f64:
+      if (FPR_idx != NumFPRs) {
+        RegsToPass.push_back(std::make_pair(FPR[FPR_idx++], Arg));
+
+        if (isVarArg) {
+          // A single float or an aggregate containing only a single float
+          // must be passed right-justified in the stack doubleword, and
+          // in the GPR, if one is available.
+          SDValue StoreOff;
+          if (Arg.getValueType().getSimpleVT().SimpleTy == MVT::f32) {
+            SDValue ConstFour = DAG.getConstant(4, PtrOff.getValueType());
+            StoreOff = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, ConstFour);
+          } else
+            StoreOff = PtrOff;
+
+          SDValue Store = DAG.getStore(Chain, dl, Arg, StoreOff,
+                                       MachinePointerInfo(), false, false, 0);
+          MemOpChains.push_back(Store);
+
+          // Float varargs are always shadowed in available integer registers
+          if (GPR_idx != NumGPRs) {
+            SDValue Load = DAG.getLoad(PtrVT, dl, Store, PtrOff,
+                                       MachinePointerInfo(), false, false,
+                                       false, 0);
+            MemOpChains.push_back(Load.getValue(1));
+            RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load));
+          }
+        } else if (GPR_idx != NumGPRs)
+          // If we have any FPRs remaining, we may also have GPRs remaining.
+          ++GPR_idx;
+      } else {
+        // Single-precision floating-point values are mapped to the
+        // second (rightmost) word of the stack doubleword.
+        if (Arg.getValueType() == MVT::f32) {
+          SDValue ConstFour = DAG.getConstant(4, PtrOff.getValueType());
+          PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, ConstFour);
+        }
+
+        LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset,
+                         true, isTailCall, false, MemOpChains,
+                         TailCallArguments, dl);
+      }
+      ArgOffset += 8;
+      break;
+    case MVT::v4f32:
+    case MVT::v4i32:
+    case MVT::v8i16:
+    case MVT::v16i8:
+      if (isVarArg) {
+        // These go aligned on the stack, or in the corresponding R registers
+        // when within range.  The Darwin PPC ABI doc claims they also go in
+        // V registers; in fact gcc does this only for arguments that are
+        // prototyped, not for those that match the ...  We do it for all
+        // arguments, seems to work.
+        while (ArgOffset % 16 !=0) {
+          ArgOffset += PtrByteSize;
+          if (GPR_idx != NumGPRs)
+            GPR_idx++;
+        }
+        // We could elide this store in the case where the object fits
+        // entirely in R registers.  Maybe later.
+        PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr,
+                            DAG.getConstant(ArgOffset, PtrVT));
+        SDValue Store = DAG.getStore(Chain, dl, Arg, PtrOff,
+                                     MachinePointerInfo(), false, false, 0);
+        MemOpChains.push_back(Store);
+        if (VR_idx != NumVRs) {
+          SDValue Load = DAG.getLoad(MVT::v4f32, dl, Store, PtrOff,
+                                     MachinePointerInfo(),
+                                     false, false, false, 0);
+          MemOpChains.push_back(Load.getValue(1));
+          RegsToPass.push_back(std::make_pair(VR[VR_idx++], Load));
+        }
+        ArgOffset += 16;
+        for (unsigned i=0; i<16; i+=PtrByteSize) {
+          if (GPR_idx == NumGPRs)
+            break;
+          SDValue Ix = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff,
+                                  DAG.getConstant(i, PtrVT));
+          SDValue Load = DAG.getLoad(PtrVT, dl, Store, Ix, MachinePointerInfo(),
+                                     false, false, false, 0);
+          MemOpChains.push_back(Load.getValue(1));
+          RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load));
+        }
+        break;
+      }
+
+      // Non-varargs Altivec params generally go in registers, but have
+      // stack space allocated at the end.
+      if (VR_idx != NumVRs) {
+        // Doesn't have GPR space allocated.
+        RegsToPass.push_back(std::make_pair(VR[VR_idx++], Arg));
+      } else {
+        LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset,
+                         true, isTailCall, true, MemOpChains,
+                         TailCallArguments, dl);
+        ArgOffset += 16;
+      }
+      break;
+    }
+  }
+
+  if (!MemOpChains.empty())
+    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                        &MemOpChains[0], MemOpChains.size());
+
+  // Check if this is an indirect call (MTCTR/BCTRL).
+  // See PrepareCall() for more information about calls through function
+  // pointers in the 64-bit SVR4 ABI.
+  if (!isTailCall &&
+      !dyn_cast<GlobalAddressSDNode>(Callee) &&
+      !dyn_cast<ExternalSymbolSDNode>(Callee) &&
+      !isBLACompatibleAddress(Callee, DAG)) {
+    // Load r2 into a virtual register and store it to the TOC save area.
+    SDValue Val = DAG.getCopyFromReg(Chain, dl, PPC::X2, MVT::i64);
+    // TOC save area offset.
+    SDValue PtrOff = DAG.getIntPtrConstant(40);
+    SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff);
+    Chain = DAG.getStore(Val.getValue(1), dl, Val, AddPtr, MachinePointerInfo(),
+                         false, false, 0);
+    // R12 must contain the address of an indirect callee.  This does not
+    // mean the MTCTR instruction must use R12; it's easier to model this
+    // as an extra parameter, so do that.
+    RegsToPass.push_back(std::make_pair((unsigned)PPC::X12, Callee));
+  }
+
+  // Build a sequence of copy-to-reg nodes chained together with token chain
+  // and flag operands which copy the outgoing args into the appropriate regs.
+  SDValue InFlag;
+  for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
+    Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first,
+                             RegsToPass[i].second, InFlag);
+    InFlag = Chain.getValue(1);
+  }
+
+  if (isTailCall)
+    PrepareTailCall(DAG, InFlag, Chain, dl, true, SPDiff, NumBytes, LROp,
+                    FPOp, true, TailCallArguments);
+
+  return FinishCall(CallConv, dl, isTailCall, isVarArg, DAG,
+                    RegsToPass, InFlag, Chain, Callee, SPDiff, NumBytes,
+                    Ins, InVals);
+}
+
+SDValue
+PPCTargetLowering::LowerCall_Darwin(SDValue Chain, SDValue Callee,
+                                    CallingConv::ID CallConv, bool isVarArg,
+                                    bool isTailCall,
+                                    const SmallVectorImpl<ISD::OutputArg> &Outs,
+                                    const SmallVectorImpl<SDValue> &OutVals,
+                                    const SmallVectorImpl<ISD::InputArg> &Ins,
+                                    DebugLoc dl, SelectionDAG &DAG,
+                                    SmallVectorImpl<SDValue> &InVals) const {
+
+  unsigned NumOps = Outs.size();
+
+  EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy();
+  bool isPPC64 = PtrVT == MVT::i64;
+  unsigned PtrByteSize = isPPC64 ? 8 : 4;
+
+  MachineFunction &MF = DAG.getMachineFunction();
+
+  // Mark this function as potentially containing a function that contains a
+  // tail call. As a consequence the frame pointer will be used for dynamicalloc
+  // and restoring the callers stack pointer in this functions epilog. This is
+  // done because by tail calling the called function might overwrite the value
+  // in this function's (MF) stack pointer stack slot 0(SP).
+  if (getTargetMachine().Options.GuaranteedTailCallOpt &&
+      CallConv == CallingConv::Fast)
+    MF.getInfo<PPCFunctionInfo>()->setHasFastCall();
+
+  unsigned nAltivecParamsAtEnd = 0;
+
+  // Count how many bytes are to be pushed on the stack, including the linkage
+  // area, and parameter passing area.  We start with 24/48 bytes, which is
+  // prereserved space for [SP][CR][LR][3 x unused].
+  unsigned NumBytes =
+    CalculateParameterAndLinkageAreaSize(DAG, isPPC64, isVarArg, CallConv,
+                                         Outs, OutVals,
+                                         nAltivecParamsAtEnd);
+
+  // Calculate by how many bytes the stack has to be adjusted in case of tail
+  // call optimization.
+  int SPDiff = CalculateTailCallSPDiff(DAG, isTailCall, NumBytes);
+
+  // To protect arguments on the stack from being clobbered in a tail call,
+  // force all the loads to happen before doing any other lowering.
+  if (isTailCall)
+    Chain = DAG.getStackArgumentTokenFactor(Chain);
+
+  // Adjust the stack pointer for the new arguments...
+  // These operations are automatically eliminated by the prolog/epilog pass
+  Chain = DAG.getCALLSEQ_START(Chain, DAG.getIntPtrConstant(NumBytes, true));
+  SDValue CallSeqStart = Chain;
+
+  // Load the return address and frame pointer so it can be move somewhere else
+  // later.
+  SDValue LROp, FPOp;
+  Chain = EmitTailCallLoadFPAndRetAddr(DAG, SPDiff, Chain, LROp, FPOp, true,
+                                       dl);
+
+  // Set up a copy of the stack pointer for use loading and storing any
+  // arguments that may not fit in the registers available for argument
+  // passing.
+  SDValue StackPtr;
+  if (isPPC64)
+    StackPtr = DAG.getRegister(PPC::X1, MVT::i64);
+  else
+    StackPtr = DAG.getRegister(PPC::R1, MVT::i32);
+
+  // Figure out which arguments are going to go in registers, and which in
+  // memory.  Also, if this is a vararg function, floating point operations
+  // must be stored to our stack, and loaded into integer regs as well, if
+  // any integer regs are available for argument passing.
+  unsigned ArgOffset = PPCFrameLowering::getLinkageSize(isPPC64, true);
+  unsigned GPR_idx = 0, FPR_idx = 0, VR_idx = 0;
+
+  static const uint16_t GPR_32[] = {           // 32-bit registers.
+    PPC::R3, PPC::R4, PPC::R5, PPC::R6,
+    PPC::R7, PPC::R8, PPC::R9, PPC::R10,
+  };
+  static const uint16_t GPR_64[] = {           // 64-bit registers.
+    PPC::X3, PPC::X4, PPC::X5, PPC::X6,
+    PPC::X7, PPC::X8, PPC::X9, PPC::X10,
+  };
+  static const uint16_t *FPR = GetFPR();
+
+  static const uint16_t VR[] = {
+    PPC::V2, PPC::V3, PPC::V4, PPC::V5, PPC::V6, PPC::V7, PPC::V8,
+    PPC::V9, PPC::V10, PPC::V11, PPC::V12, PPC::V13
+  };
+  const unsigned NumGPRs = array_lengthof(GPR_32);
+  const unsigned NumFPRs = 13;
+  const unsigned NumVRs  = array_lengthof(VR);
+
+  const uint16_t *GPR = isPPC64 ? GPR_64 : GPR_32;
+
+  SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass;
+  SmallVector<TailCallArgumentInfo, 8> TailCallArguments;
+
+  SmallVector<SDValue, 8> MemOpChains;
+  for (unsigned i = 0; i != NumOps; ++i) {
+    SDValue Arg = OutVals[i];
+    ISD::ArgFlagsTy Flags = Outs[i].Flags;
+
+    // PtrOff will be used to store the current argument to the stack if a
+    // register cannot be found for it.
+    SDValue PtrOff;
+
+    PtrOff = DAG.getConstant(ArgOffset, StackPtr.getValueType());
+
+    PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff);
+
+    // On PPC64, promote integers to 64-bit values.
+    if (isPPC64 && Arg.getValueType() == MVT::i32) {
+      // FIXME: Should this use ANY_EXTEND if neither sext nor zext?
+      unsigned ExtOp = Flags.isSExt() ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
+      Arg = DAG.getNode(ExtOp, dl, MVT::i64, Arg);
+    }
+
+    // FIXME memcpy is used way more than necessary.  Correctness first.
+    // Note: "by value" is code for passing a structure by value, not
+    // basic types.
+    if (Flags.isByVal()) {
+      unsigned Size = Flags.getByValSize();
+      // Very small objects are passed right-justified.  Everything else is
+      // passed left-justified.
+      if (Size==1 || Size==2) {
+        EVT VT = (Size==1) ? MVT::i8 : MVT::i16;
+        if (GPR_idx != NumGPRs) {
+          SDValue Load = DAG.getExtLoad(ISD::EXTLOAD, dl, PtrVT, Chain, Arg,
+                                        MachinePointerInfo(), VT,
+                                        false, false, 0);
+          MemOpChains.push_back(Load.getValue(1));
+          RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load));
+
+          ArgOffset += PtrByteSize;
+        } else {
+          SDValue Const = DAG.getConstant(PtrByteSize - Size,
+                                          PtrOff.getValueType());
+          SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const);
+          Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, AddPtr,
+                                                            CallSeqStart,
+                                                            Flags, DAG, dl);
+          ArgOffset += PtrByteSize;
+        }
+        continue;
+      }
+      // Copy entire object into memory.  There are cases where gcc-generated
+      // code assumes it is there, even if it could be put entirely into
+      // registers.  (This is not what the doc says.)
+      Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, PtrOff,
+                                                        CallSeqStart,
+                                                        Flags, DAG, dl);
+
       // For small aggregates (Darwin only) and aggregates >= PtrByteSize,
       // copy the pieces of the object that fit into registers from the
       // parameter save area.
@@ -3805,18 +4174,10 @@
               !isPPC64)  // PPC64 has 64-bit GPR's obviously :)
             ++GPR_idx;
         }
-      } else {
-        // Single-precision floating-point values are mapped to the
-        // second (rightmost) word of the stack doubleword.
-        if (Arg.getValueType() == MVT::f32 && isPPC64 && isSVR4ABI) {
-          SDValue ConstFour = DAG.getConstant(4, PtrOff.getValueType());
-          PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, ConstFour);
-        }
-
+      } else
         LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset,
                          isPPC64, isTailCall, false, MemOpChains,
                          TailCallArguments, dl);
-      }
       if (isPPC64)
         ArgOffset += 8;
       else
@@ -3911,22 +4272,6 @@
     Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                         &MemOpChains[0], MemOpChains.size());
 
-  // Check if this is an indirect call (MTCTR/BCTRL).
-  // See PrepareCall() for more information about calls through function
-  // pointers in the 64-bit SVR4 ABI.
-  if (!isTailCall && isPPC64 && PPCSubTarget.isSVR4ABI() &&
-      !dyn_cast<GlobalAddressSDNode>(Callee) &&
-      !dyn_cast<ExternalSymbolSDNode>(Callee) &&
-      !isBLACompatibleAddress(Callee, DAG)) {
-    // Load r2 into a virtual register and store it to the TOC save area.
-    SDValue Val = DAG.getCopyFromReg(Chain, dl, PPC::X2, MVT::i64);
-    // TOC save area offset.
-    SDValue PtrOff = DAG.getIntPtrConstant(40);
-    SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff);
-    Chain = DAG.getStore(Val.getValue(1), dl, Val, AddPtr, MachinePointerInfo(),
-                         false, false, 0);
-  }
-
   // On Darwin, R12 must contain the address of an indirect callee.  This does
   // not mean the MTCTR instruction must use R12; it's easier to model this as
   // an extra parameter, so do that.
@@ -3991,8 +4336,24 @@
   for (unsigned i = 0; i != RVLocs.size(); ++i) {
     CCValAssign &VA = RVLocs[i];
     assert(VA.isRegLoc() && "Can only return in registers!");
-    Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
-                             OutVals[i], Flag);
+
+    SDValue Arg = OutVals[i];
+
+    switch (VA.getLocInfo()) {
+    default: llvm_unreachable("Unknown loc info!");
+    case CCValAssign::Full: break;
+    case CCValAssign::AExt:
+      Arg = DAG.getNode(ISD::ANY_EXTEND, dl, VA.getLocVT(), Arg);
+      break;
+    case CCValAssign::ZExt:
+      Arg = DAG.getNode(ISD::ZERO_EXTEND, dl, VA.getLocVT(), Arg);
+      break;
+    case CCValAssign::SExt:
+      Arg = DAG.getNode(ISD::SIGN_EXTEND, dl, VA.getLocVT(), Arg);
+      break;
+    }
+
+    Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), Arg, Flag);
     Flag = Chain.getValue(1);
   }
 
@@ -6118,6 +6479,14 @@
     case 'v':
     case 'y':
       return C_RegisterClass;
+    case 'Z':
+      // FIXME: While Z does indicate a memory constraint, it specifically
+      // indicates an r+r address (used in conjunction with the 'y' modifier
+      // in the replacement string). Currently, we're forcing the base
+      // register to be r0 in the asm printer (which is interpreted as zero)
+      // and forming the complete address in the second register. This is
+      // suboptimal.
+      return C_Memory;
     }
   }
   return TargetLowering::getConstraintType(Constraint);
@@ -6160,6 +6529,9 @@
   case 'y':
     weight = CW_Register;
     break;
+  case 'Z':
+    weight = CW_Memory;
+    break;
   }
   return weight;
 }
@@ -6176,9 +6548,9 @@
         return std::make_pair(0U, &PPC::G8RCRegClass);
       return std::make_pair(0U, &PPC::GPRCRegClass);
     case 'f':
-      if (VT == MVT::f32)
+      if (VT == MVT::f32 || VT == MVT::i32)
         return std::make_pair(0U, &PPC::F4RCRegClass);
-      if (VT == MVT::f64)
+      if (VT == MVT::f64 || VT == MVT::i64)
         return std::make_pair(0U, &PPC::F8RCRegClass);
       break;
     case 'v':

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.h (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCISelLowering.h Tue Nov 13 09:21:47 2012
@@ -467,6 +467,15 @@
                   DebugLoc dl, SelectionDAG &DAG) const;
 
     SDValue
+      extendArgForPPC64(ISD::ArgFlagsTy Flags, EVT ObjectVT, SelectionDAG &DAG,
+                        SDValue ArgVal, DebugLoc dl) const;
+
+    void
+      setMinReservedArea(MachineFunction &MF, SelectionDAG &DAG,
+                         unsigned nAltivecParamsAtEnd,
+                         unsigned MinReservedArea, bool isPPC64) const;
+
+    SDValue
       LowerFormalArguments_Darwin(SDValue Chain,
                                   CallingConv::ID CallConv, bool isVarArg,
                                   const SmallVectorImpl<ISD::InputArg> &Ins,
@@ -486,7 +495,21 @@
                                   SmallVectorImpl<SDValue> &InVals) const;
 
     SDValue
-      LowerCall_Darwin_Or_64SVR4(SDValue Chain, SDValue Callee,
+      createMemcpyOutsideCallSeq(SDValue Arg, SDValue PtrOff,
+                                 SDValue CallSeqStart, ISD::ArgFlagsTy Flags,
+                                 SelectionDAG &DAG, DebugLoc dl) const;
+
+    SDValue
+      LowerCall_Darwin(SDValue Chain, SDValue Callee,
+                       CallingConv::ID CallConv,
+                       bool isVarArg, bool isTailCall,
+                       const SmallVectorImpl<ISD::OutputArg> &Outs,
+                       const SmallVectorImpl<SDValue> &OutVals,
+                       const SmallVectorImpl<ISD::InputArg> &Ins,
+                       DebugLoc dl, SelectionDAG &DAG,
+                       SmallVectorImpl<SDValue> &InVals) const;
+    SDValue
+      LowerCall_64SVR4(SDValue Chain, SDValue Callee,
                        CallingConv::ID CallConv,
                        bool isVarArg, bool isTailCall,
                        const SmallVectorImpl<ISD::OutputArg> &Outs,

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCInstr64Bit.td
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCInstr64Bit.td?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCInstr64Bit.td (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCInstr64Bit.td Tue Nov 13 09:21:47 2012
@@ -494,16 +494,16 @@
 
 // Rotate instructions.
 def RLDCL  : MDForm_1<30, 0,
-                      (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB, u6imm:$MB),
-                      "rldcl $rA, $rS, $rB, $MB", IntRotateD,
+                      (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB, u6imm:$MBE),
+                      "rldcl $rA, $rS, $rB, $MBE", IntRotateD,
                       []>, isPPC64;
 def RLDICL : MDForm_1<30, 0,
-                      (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$MB),
-                      "rldicl $rA, $rS, $SH, $MB", IntRotateDI,
+                      (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$MBE),
+                      "rldicl $rA, $rS, $SH, $MBE", IntRotateDI,
                       []>, isPPC64;
 def RLDICR : MDForm_1<30, 1,
-                      (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$ME),
-                      "rldicr $rA, $rS, $SH, $ME", IntRotateDI,
+                      (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$MBE),
+                      "rldicr $rA, $rS, $SH, $MBE", IntRotateDI,
                       []>, isPPC64;
 
 def RLWINM8 : MForm_2<21,
@@ -639,13 +639,13 @@
                      (PPCtoc_entry tconstpool:$disp, G8RC:$reg))]>, isPPC64;
 
 let hasSideEffects = 1 in { 
-let RST = 2, DS_RA = 0 in // FIXME: Should be a pseudo.
-def LDinto_toc: DSForm_1<58, 0, (outs), (ins G8RC:$reg),
+let RST = 2, DS = 2 in
+def LDinto_toc: DSForm_1a<58, 0, (outs), (ins G8RC:$reg),
                     "ld 2, 8($reg)", LdStLD,
                     [(PPCload_toc G8RC:$reg)]>, isPPC64;
                     
-let RST = 2, DS_RA = 0 in // FIXME: Should be a pseudo.
-def LDtoc_restore : DSForm_1<58, 0, (outs), (ins),
+let RST = 2, DS = 10, RA = 1 in
+def LDtoc_restore : DSForm_1a<58, 0, (outs), (ins),
                     "ld 2, 40(1)", LdStLD,
                     [(PPCtoc_restore)]>, isPPC64;
 }

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCRegisterInfo.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCRegisterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCRegisterInfo.cpp Tue Nov 13 09:21:47 2012
@@ -498,7 +498,7 @@
     } else if (CRSpillFrameIdx) {
       FrameIdx = CRSpillFrameIdx;
     } else {
-      MachineFrameInfo *MFI = (const_cast<MachineFunction &>(MF)).getFrameInfo();
+      MachineFrameInfo *MFI = ((MachineFunction &)MF).getFrameInfo();
       FrameIdx = MFI->CreateFixedObject((uint64_t)4, (int64_t)-4, true);
       CRSpillFrameIdx = FrameIdx;
     }

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.cpp Tue Nov 13 09:21:47 2012
@@ -54,19 +54,26 @@
     CPUName = sys::getHostCPUName();
 #endif
 
-  // Parse features string.
-  ParseSubtargetFeatures(CPUName, FS);
-
   // Initialize scheduling itinerary for the specified CPU.
   InstrItins = getInstrItineraryForCPU(CPUName);
 
+  // Make sure 64-bit features are available when CPUname is generic
+  std::string FullFS = FS;
+
   // If we are generating code for ppc64, verify that options make sense.
   if (is64Bit) {
     Has64BitSupport = true;
     // Silently force 64-bit register use on ppc64.
     Use64BitRegs = true;
+    if (!FullFS.empty())
+      FullFS = "+64bit," + FullFS;
+    else
+      FullFS = "+64bit";
   }
 
+  // Parse features string.
+  ParseSubtargetFeatures(CPUName, FullFS);
+
   // If the user requested use of 64-bit regs, but the cpu selected doesn't
   // support it, ignore.
   if (use64BitRegs() && !has64BitSupport())

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.h
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.h?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.h (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCSubtarget.h Tue Nov 13 09:21:47 2012
@@ -33,34 +33,34 @@
   enum {
     DIR_NONE,
     DIR_32,
-    DIR_440, 
-    DIR_601, 
-    DIR_602, 
-    DIR_603, 
+    DIR_440,
+    DIR_601,
+    DIR_602,
+    DIR_603,
     DIR_7400,
-    DIR_750, 
-    DIR_970, 
+    DIR_750,
+    DIR_970,
     DIR_A2,
     DIR_E500mc,
     DIR_E5500,
     DIR_PWR6,
     DIR_PWR7,
-    DIR_64  
+    DIR_64
   };
 }
 
 class GlobalValue;
 class TargetMachine;
-  
+
 class PPCSubtarget : public PPCGenSubtargetInfo {
 protected:
   /// stackAlignment - The minimum alignment known to hold of the stack frame on
   /// entry to the function and which must be maintained by every function.
   unsigned StackAlignment;
-  
+
   /// Selected instruction itineraries (one entry per itinerary class.)
   InstrItineraryData InstrItins;
-  
+
   /// Which cpu directive was used.
   unsigned DarwinDirective;
 
@@ -76,7 +76,7 @@
   bool IsBookE;
   bool HasLazyResolverStubs;
   bool IsJITCodeModel;
-  
+
   /// TargetTriple - What processor and OS we're targeting.
   Triple TargetTriple;
 
@@ -86,11 +86,11 @@
   ///
   PPCSubtarget(const std::string &TT, const std::string &CPU,
                const std::string &FS, bool is64Bit);
-  
-  /// ParseSubtargetFeatures - Parses features string setting specified 
+
+  /// ParseSubtargetFeatures - Parses features string setting specified
   /// subtarget options.  Definition of function is auto generated by tblgen.
   void ParseSubtargetFeatures(StringRef CPU, StringRef FS);
-  
+
   /// SetJITMode - This is called to inform the subtarget info that we are
   /// producing code for the JIT.
   void SetJITMode();
@@ -99,12 +99,12 @@
   /// stack frame on entry to the function and which must be maintained by every
   /// function for this subtarget.
   unsigned getStackAlignment() const { return StackAlignment; }
-  
+
   /// getDarwinDirective - Returns the -m directive specified for the cpu.
   ///
   unsigned getDarwinDirective() const { return DarwinDirective; }
-  
-  /// getInstrItins - Return the instruction itineraies based on subtarget 
+
+  /// getInstrItins - Return the instruction itineraies based on subtarget
   /// selection.
   const InstrItineraryData &getInstrItineraryData() const { return InstrItins; }
 
@@ -113,6 +113,13 @@
   const char *getDataLayoutString() const {
     // Note, the alignment values for f64 and i64 on ppc64 in Darwin
     // documentation are wrong; these are correct (i.e. "what gcc does").
+    if (isPPC64() && isSVR4ABI()) {
+      if (TargetTriple.getOS() == llvm::Triple::FreeBSD)
+        return "E-p:64:64-f64:64:64-i64:64:64-f128:64:64-v128:128:128-n32:64";
+      else
+        return "E-p:64:64-f64:64:64-i64:64:64-f128:128:128-v128:128:128-n32:64";
+    }
+
     return isPPC64() ? "E-p:64:64-f64:64:64-i64:64:64-f128:64:128-n32:64"
                      : "E-p:32:32-f64:64:64-i64:64:64-f128:64:128-n32";
   }
@@ -120,22 +127,22 @@
   /// isPPC64 - Return true if we are generating code for 64-bit pointer mode.
   ///
   bool isPPC64() const { return IsPPC64; }
-  
+
   /// has64BitSupport - Return true if the selected CPU supports 64-bit
   /// instructions, regardless of whether we are in 32-bit or 64-bit mode.
   bool has64BitSupport() const { return Has64BitSupport; }
-  
+
   /// use64BitRegs - Return true if in 64-bit mode or if we should use 64-bit
   /// registers in 32-bit mode when possible.  This can only true if
   /// has64BitSupport() returns true.
   bool use64BitRegs() const { return Use64BitRegs; }
-  
+
   /// hasLazyResolverStub - Return true if accesses to the specified global have
   /// to go through a dyld lazy resolution stub.  This means that an extra load
   /// is required to get the address of the global.
-  bool hasLazyResolverStub(const GlobalValue *GV, 
+  bool hasLazyResolverStub(const GlobalValue *GV,
                            const TargetMachine &TM) const;
-  
+
   // isJITCodeModel - True if we're generating code for the JIT
   bool isJITCodeModel() const { return IsJITCodeModel; }
 

Modified: llvm/branches/R600/lib/Target/PowerPC/PPCTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/PowerPC/PPCTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/PowerPC/PPCTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/PowerPC/PPCTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -44,7 +44,7 @@
     FrameLowering(Subtarget), JITInfo(*this, is64Bit),
     TLInfo(*this), TSInfo(*this),
     InstrItins(Subtarget.getInstrItineraryData()),
-    STTI(&TLInfo){
+    STTI(&TLInfo), VTTI(&TLInfo) {
 
   // The binutils for the BG/P are too old for CFI.
   if (Subtarget.isBGP())

Modified: llvm/branches/R600/lib/Target/Sparc/SparcTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/Sparc/SparcTargetMachine.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/Sparc/SparcTargetMachine.cpp (original)
+++ llvm/branches/R600/lib/Target/Sparc/SparcTargetMachine.cpp Tue Nov 13 09:21:47 2012
@@ -36,7 +36,7 @@
     DL(Subtarget.getDataLayout()),
     InstrInfo(Subtarget),
     TLInfo(*this), TSInfo(*this),
-    FrameLowering(Subtarget),STTI(&TLInfo) {
+    FrameLowering(Subtarget), STTI(&TLInfo), VTTI(&TLInfo) {
 }
 
 namespace {

Removed: llvm/branches/R600/lib/Target/TargetELFWriterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/TargetELFWriterInfo.cpp?rev=167837&view=auto
==============================================================================
--- llvm/branches/R600/lib/Target/TargetELFWriterInfo.cpp (original)
+++ llvm/branches/R600/lib/Target/TargetELFWriterInfo.cpp (removed)
@@ -1,25 +0,0 @@
-//===-- lib/Target/TargetELFWriterInfo.cpp - ELF Writer Info --0-*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements the TargetELFWriterInfo class.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/Function.h"
-#include "llvm/Target/TargetELFWriterInfo.h"
-#include "llvm/DataLayout.h"
-#include "llvm/Target/TargetMachine.h"
-using namespace llvm;
-
-TargetELFWriterInfo::TargetELFWriterInfo(bool is64Bit_, bool isLittleEndian_) :
-  is64Bit(is64Bit_), isLittleEndian(isLittleEndian_) {
-}
-
-TargetELFWriterInfo::~TargetELFWriterInfo() {}
-

Modified: llvm/branches/R600/lib/Target/TargetTransformImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/R600/lib/Target/TargetTransformImpl.cpp?rev=167838&r1=167837&r2=167838&view=diff
==============================================================================
--- llvm/branches/R600/lib/Target/TargetTransformImpl.cpp (original)
+++ llvm/branches/R600/lib/Target/TargetTransformImpl.cpp Tue Nov 13 09:21:47 2012
@@ -9,9 +9,16 @@
 
 #include "llvm/Target/TargetTransformImpl.h"
 #include "llvm/Target/TargetLowering.h"
+#include <utility>
 
 using namespace llvm;
 
+//===----------------------------------------------------------------------===//
+//
+// Calls used by scalar transformations.
+//
+//===----------------------------------------------------------------------===//
+
 bool ScalarTargetTransformImpl::isLegalAddImmediate(int64_t imm) const {
   return TLI->isLegalAddImmediate(imm);
 }
@@ -21,7 +28,7 @@
 }
 
 bool ScalarTargetTransformImpl::isLegalAddressingMode(const AddrMode &AM,
-                                                    Type *Ty) const {
+                                                      Type *Ty) const {
   return TLI->isLegalAddressingMode(AM, Ty);
 }
 
@@ -41,3 +48,298 @@
 unsigned ScalarTargetTransformImpl::getJumpBufSize() const {
   return TLI->getJumpBufSize();
 }
+
+bool ScalarTargetTransformImpl::shouldBuildLookupTables() const {
+  return TLI->supportJumpTables() &&
+      (TLI->isOperationLegalOrCustom(ISD::BR_JT, MVT::Other) ||
+       TLI->isOperationLegalOrCustom(ISD::BRIND, MVT::Other));
+}
+
+//===----------------------------------------------------------------------===//
+//
+// Calls used by the vectorizers.
+//
+//===----------------------------------------------------------------------===//
+int VectorTargetTransformImpl::InstructionOpcodeToISD(unsigned Opcode) const {
+  enum InstructionOpcodes {
+#define HANDLE_INST(NUM, OPCODE, CLASS) OPCODE = NUM,
+#define LAST_OTHER_INST(NUM) InstructionOpcodesCount = NUM
+#include "llvm/Instruction.def"
+  };
+  switch (static_cast<InstructionOpcodes>(Opcode)) {
+  case Ret:            return 0;
+  case Br:             return 0;
+  case Switch:         return 0;
+  case IndirectBr:     return 0;
+  case Invoke:         return 0;
+  case Resume:         return 0;
+  case Unreachable:    return 0;
+  case Add:            return ISD::ADD;
+  case FAdd:           return ISD::FADD;
+  case Sub:            return ISD::SUB;
+  case FSub:           return ISD::FSUB;
+  case Mul:            return ISD::MUL;
+  case FMul:           return ISD::FMUL;
+  case UDiv:           return ISD::UDIV;
+  case SDiv:           return ISD::UDIV;
+  case FDiv:           return ISD::FDIV;
+  case URem:           return ISD::UREM;
+  case SRem:           return ISD::SREM;
+  case FRem:           return ISD::FREM;
+  case Shl:            return ISD::SHL;
+  case LShr:           return ISD::SRL;
+  case AShr:           return ISD::SRA;
+  case And:            return ISD::AND;
+  case Or:             return ISD::OR;
+  case Xor:            return ISD::XOR;
+  case Alloca:         return 0;
+  case Load:           return ISD::LOAD;
+  case Store:          return ISD::STORE;
+  case GetElementPtr:  return 0;
+  case Fence:          return 0;
+  case AtomicCmpXchg:  return 0;
+  case AtomicRMW:      return 0;
+  case Trunc:          return ISD::TRUNCATE;
+  case ZExt:           return ISD::ZERO_EXTEND;
+  case SExt:           return ISD::SIGN_EXTEND;
+  case FPToUI:         return ISD::FP_TO_UINT;
+  case FPToSI:         return ISD::FP_TO_SINT;
+  case UIToFP:         return ISD::UINT_TO_FP;
+  case SIToFP:         return ISD::SINT_TO_FP;
+  case FPTrunc:        return ISD::FP_ROUND;
+  case FPExt:          return ISD::FP_EXTEND;
+  case PtrToInt:       return ISD::BITCAST;
+  case IntToPtr:       return ISD::BITCAST;
+  case BitCast:        return ISD::BITCAST;
+  case ICmp:           return ISD::SETCC;
+  case FCmp:           return ISD::SETCC;
+  case PHI:            return 0;
+  case Call:           return 0;
+  case Select:         return ISD::SELECT;
+  case UserOp1:        return 0;
+  case UserOp2:        return 0;
+  case VAArg:          return 0;
+  case ExtractElement: return ISD::EXTRACT_VECTOR_ELT;
+  case InsertElement:  return ISD::INSERT_VECTOR_ELT;
+  case ShuffleVector:  return ISD::VECTOR_SHUFFLE;
+  case ExtractValue:   return ISD::MERGE_VALUES;
+  case InsertValue:    return ISD::MERGE_VALUES;
+  case LandingPad:     return 0;
+  }
+
+  llvm_unreachable("Unknown instruction type encountered!");
+}
+
+std::pair<unsigned, MVT>
+VectorTargetTransformImpl::getTypeLegalizationCost(Type *Ty) const {
+
+  LLVMContext &C = Ty->getContext();
+  EVT MTy = TLI->getValueType(Ty);
+
+  unsigned Cost = 1;
+  // We keep legalizing the type until we find a legal kind. We assume that
+  // the only operation that costs anything is the split. After splitting
+  // we need to handle two types.
+  while (true) {
+    TargetLowering::LegalizeKind LK = TLI->getTypeConversion(C, MTy);
+
+    if (LK.first == TargetLowering::TypeLegal)
+      return std::make_pair(Cost, MTy.getSimpleVT());
+
+    if (LK.first == TargetLowering::TypeSplitVector ||
+        LK.first == TargetLowering::TypeExpandInteger)
+      Cost *= 2;
+
+    // Keep legalizing the type.
+    MTy = LK.second;
+  }
+}
+
+unsigned
+VectorTargetTransformImpl::getScalarizationOverhead(Type *Ty,
+                                                    bool Insert,
+                                                    bool Extract) const {
+  assert (Ty->isVectorTy() && "Can only scalarize vectors");
+  unsigned Cost = 0;
+
+  for (int i = 0, e = Ty->getVectorNumElements(); i < e; ++i) {
+    if (Insert)
+      Cost += getVectorInstrCost(Instruction::InsertElement, Ty, i);
+    if (Extract)
+      Cost += getVectorInstrCost(Instruction::ExtractElement, Ty, i);
+  }
+
+  return Cost;
+}
+
+unsigned VectorTargetTransformImpl::getArithmeticInstrCost(unsigned Opcode,
+                                                           Type *Ty) const {
+  // Check if any of the operands are vector operands.
+  int ISD = InstructionOpcodeToISD(Opcode);
+  assert(ISD && "Invalid opcode");
+
+  std::pair<unsigned, MVT> LT = getTypeLegalizationCost(Ty);
+
+  if (!TLI->isOperationExpand(ISD, LT.second)) {
+    // The operation is legal. Assume it costs 1. Multiply
+    // by the type-legalization overhead.
+    return LT.first * 1;
+  }
+
+  // Else, assume that we need to scalarize this op.
+  if (Ty->isVectorTy()) {
+    unsigned Num = Ty->getVectorNumElements();
+    unsigned Cost = getArithmeticInstrCost(Opcode, Ty->getScalarType());
+    // return the cost of multiple scalar invocation plus the cost of inserting
+    // and extracting the values.
+    return getScalarizationOverhead(Ty, true, true) + Num * Cost;
+  }
+
+  // We don't know anything about this scalar instruction.
+  return 1;
+}
+
+unsigned VectorTargetTransformImpl::getBroadcastCost(Type *Tp) const {
+  return 1;
+}
+
+unsigned VectorTargetTransformImpl::getCastInstrCost(unsigned Opcode, Type *Dst,
+                                  Type *Src) const {
+  int ISD = InstructionOpcodeToISD(Opcode);
+  assert(ISD && "Invalid opcode");
+
+  std::pair<unsigned, MVT> SrcLT = getTypeLegalizationCost(Src);
+  std::pair<unsigned, MVT> DstLT = getTypeLegalizationCost(Dst);
+
+  // Handle scalar conversions.
+  if (!Src->isVectorTy() && !Dst->isVectorTy()) {
+
+    // Scalar bitcasts and truncs are usually free.
+    if (Opcode == Instruction::BitCast || Opcode == Instruction::Trunc)
+      return 0;
+
+    // Just check the op cost. If the operation is legal then assume it costs 1.
+    if (!TLI->isOperationExpand(ISD, DstLT.second))
+      return  1;
+
+    // Assume that illegal scalar instruction are expensive.
+    return 4;
+  }
+
+  // Check vector-to-vector casts.
+  if (Dst->isVectorTy() && Src->isVectorTy()) {
+
+    // If the cast is between same-sized registers, then the check is simple.
+    if (SrcLT.first == DstLT.first &&
+        SrcLT.second.getSizeInBits() == DstLT.second.getSizeInBits()) {
+
+      // Bitcast between types that are legalized to the same type are free.
+      if (Opcode == Instruction::BitCast || Opcode == Instruction::Trunc)
+        return 0;
+
+      // Assume that Zext is done using AND.
+      if (Opcode == Instruction::ZExt)
+        return 1;
+
+      // Assume that sext is