This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Unaligned floating point access generation for powerpc


On Wednesday, October 22, 2003, at 08:09 AM, David Edelsohn wrote:
The programmer is overriding the ABI definitions for alignment.
The compiler assumes the programmer knows what s/he is doing. Violating
the ABI may produce inefficient code. Why do you expect the compiler to
do anything else?

?


The compiler is allowed to produce inefficient code, the compiler isn't allowed to produce code that doesn't work. To do so, would be a bug. Now, the bug may me that gcc always assumes the existence of a trap handler to do the move and hasn't been ported to a ppc without such a handler, or the bug may be that the backend can't assume that and must generate a move by pieces style code.

I'd call it a compiler bug in any event. Worst case, the target platform doesn't allow for trap handlers and gcc just has to be smart enough to generate code that will never fault.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]