Details
-
Type:
Improvement
-
Status:
Open
-
Priority:
Minor
-
Resolution: Unresolved
-
Affects Version/s: None
-
Fix Version/s: 1000
-
Component/s: Compiler: Optimizing, Instruction Architecture: PowerPC
-
Labels:None
-
Number of attachments :
Description
Currently in the PPC assembler the type of a register is an int and the type of a displacement is an int. In the Intel compiler the registers are bytes, scales are shorts and the displacements ints. It is all too easy with the PPC assembler to pass a register as a displacement or vice versa. For example here are the definitions of STW and STWX:
public final void emitSTW (int RS, int D, int RA)
public final void emitSTWX (int RS, int RA, int RB)
creating the statement:
asm.emitSTW(T0, T1, T2)
would not create any warnings (the 2nd argument probably means this should be an STWX) whereas had the definition of the registers been bytes (or even enums) the developer could have been spared some head scratching.
This replaces SF RFE tracker 1658972.
Issue Links
- relates to
-
RVM-322
Intel assembler register naming improvement required
-
r13833 added stronger typing to the Intel assembler's emit methods. By using a common interface of Register, operations for loading the TIB from the object model could avoid conversion to an integer and then back into an enumeration. That is:
baselineEmitLoadTIB(VM_Assembler asm, int dest, int object)
where dest and object are registers encoded as an int, could become:
baselineEmitLoadTIB(VM_Assembler asm, Register dest, Register object)
where Register is an interface common to all enumerations implementing registers.