Could you elaborate on this statement? > In some cases the assembler, using 32-bit arithmetic, will generate > instructions that produce mathematically incorrect results in 64-bit > mode.
Expression evaluation is indeed done using 32-bit arithmetic, but I can't see how that would "generate instructions that produce mathematically incorrect results in 64-bit mode": Do you mean (a) 64-bit assembly-time expressions, (b) 64-bit execution-time evaluations, or (c) executing in 64-bit addressing mode? John Ehrman
