llvm-svn: 973

llvm-svn: 972

1. Forward propogate a loaded constant to make debugging slightly less annoying
2. Fix an extremely nasty hard to find bug that really irritated me for the better
   part of 6 hours and was causing linking to die and bizarre and mysterious ways.
   Things should be much more stable now that this one liner has been fixed. GRR

llvm-svn: 971

Changes to get some meaningful feedback from the bytecode reader.  At some point this stuff should all be exception driven, but for now it is not.

llvm-svn: 970

llvm-svn: 969

llvm-svn: 968

llvm-svn: 967

llvm-svn: 965

llvm-svn: 964

llvm-svn: 963

llvm-svn: 962

Fixed a LONG standing, SCARY problem with bytecode encoding.  It turns out to be an endian problem that only shows up with type 0 instructions in LARGE programs.

llvm-svn: 961

llvm-svn: 960

Allow unresolved/opaque types to be read and written to bytecode files

llvm-svn: 959

Fix another annoying bug that took forever to track down.  This one involves abstract type resolution at a bad time that broke symbol tables.

llvm-svn: 958

llvm-svn: 956

llvm-svn: 955

This will use the CallGraph only if one is provided.

llvm-svn: 950

llvm-svn: 949

llvm-svn: 948

llvm-svn: 947

llvm-svn: 946

useful instructions already inserted into delay slots.

llvm-svn: 945

llvm-svn: 944

llvm-svn: 943

llvm-svn: 942

Added a function to map between the caller's and callee's register windows.

llvm-svn: 941

be used to distinguish those uses of the `jmpl' instruction.

llvm-svn: 940

Use JMPLRET instead of RETURN for method returns so that

llvm-svn: 939

which have the same opcode and operands but different flags.

llvm-svn: 938

Added code to map between caller's and callee's register windows.
Added frame layout class.

llvm-svn: 937

llvm-svn: 936

%list = type {%list *}
%list = type {%list *}

not being accepted (broken testmisc.ll)

llvm-svn: 935

declare int "call_operand"      (%rtx_def*, int)        ;; Prototype for: call_operand
declare int "restore_operand"   (%rtx_def*, int)        ;; Prototype for: restore_operand

%rtx_def = type opaque
%rtx_def = type int
implementation

llvm-svn: 934

1. Delete type handle regardless of whether a collision occured
2. Remove a MAJOR pessimization of runtime performance (thought be be an optimization at the time).

This second one was causing a 105k llvm file (from gcc) to parse in 58 seconds... without the
'optimization' it now parses in 3.64 seconds.  I suck.

llvm-svn: 933

declare int "malloc"(...)

...

%reg112 = call int (...) * %malloc( uint %cast1007 )

llvm-svn: 932

%list = type opaque
%list = type %list

llvm-svn: 931

llvm-svn: 930

llvm-svn: 929

llvm-svn: 928