Re: Block reording.

On Mar 8, 5:52 pm, Edward.Ros...@gmail.com wrote:


> The code is simple to generate if I only use long distance jumps for
> je, since then every block is the same size and every jump is more
> or less equal in cost. However, a jump to an arbitrary point is 6
> bytes of machine code, where as a jump to a nearby point can be sone
> in only 2 bytes. Naturally, I'd rather use as many short jumps as
> possible, but this requires some strategy to reorder the blocks and
> switching from a 6 byte jump to a 2 byte has the potential to affect
> the distances of a large number of the other jumps. And the
> shrinking of the code size may allow other short jumps to be used,
> etc...


Just to clarify, I have a simple scheme scans through the blocks,
inserting short jumps if possible (noting that the switch to a short
jump may make it possible). This process is repeated until it can't
continue.


-Ed

Return to the comp.compilers page.
Search the comp.compilers archives again.