[hatari-devel] CPU breakpoints and single-stepping

Eero Tamminen eerot at users.berlios.de
Mon Jun 8 21:17:49 CEST 2009


On Monday 08 June 2009, npomarede at corp.free.fr wrote:
> > Btw.  I was thinking that the conditional breakpoint checking will
> > be quite a bit slower than normal address based breakpoints, so
> > maybe we should have both?
> Well, I think that as soon as you use breakpoint, you don't care about
> speed anymore ; I'm only exagerating a little, but I really think that if
> the user choose to use advanced debugging functions, he must have a
> machine powerful enough to handle it, so splitting the code in 2 kinds of
> breakpoints seems to be extra work not necessarily needed to me. But if
> someone feels like doing it, why not :)

I'm mostly concerned about the DSP side.  Currently e.g. even the trivial
DSP_Run() is quite visible in profile because it's run 7x for each m68k
instruction and this adds a lot of code to execute at that place when
there're conditional breakpoints.  (And half of the time DSP emu would like
to take much more CPU than my machine has available.... :-))

Extra performance may also be nice for speeding up the emulation
while waiting the breakpoint to trigger.  But let's see whether this will be
a problem in practice.

	- Eero

PS. I added the current code as breakcond.c to repo.  I will shortly add
some more parsing code to it + do some other fine-tuning.

More information about the hatari-devel mailing list