Hey! Andy Wingo <[email protected]> writes:
> On Tue 13 Jul 2010 00:57, [email protected] (Ludovic Courtès) writes: > >> The attached patch is a first stab at re-implementing SRFI-1 in >> Scheme. >> >> IOW, with the debug engine (currently the default) and for large lists >> ‘fold’ in Scheme is ~9% slower than in C; for small lists it’s ~17% >> slower. >> >> With the regular engine, Scheme is ~2% faster for large lists and still >> ~5% slower for small lists. >> >> I’m tempted to put this in > > Cool, looks good to me. Just to make sure, did you read the figures right? With the debug engine, the slowdown is noticeable. >> and then make the regular engine the default unless ‘--debug’ is >> specified. > > I would prefer to keep the debug VM as the default. AFAICS the only difference between the two engine is VM_USE_HOOKS. Hooks are only used in (system vm coverage) at this point, so we don’t lose much by disabling them. Am I overlooking something? > At the very least, one should be able to switch VMs. You mean switch VMs for the code being executed? (For code to be executed than can already be done using ‘vm-apply’.) Thanks, Ludo’.
