Re: GC [was Re: Lua 5.1 (work4) now available]

[Mike Pall <mikelu-0501@...>]

Hi,

Roberto Ierusalimschy wrote:
> During the sweep phase, the estimate is corrected to subtract dead
> objects (see lgc.c:593 and lgc.c:604). So, that shouldn't be a problem.

Yes, I realized that 5 minutes after I sent the mail. But then I was
already in bed. ;-)

> + g->gcgenerational = 0;
> 
> Mike, is it easy to regenerate that graph with the above change?

See attachment. Looks much better now. Note that the baseline of the
estimate is oscillating slowly between 794 and 805. You see only the
first downward slope in the image. The estimate is 731 w/ full gc.

This needs to be verified with other test cases, though. I'm thinking
about a better way to interactively watch the GC behaviour. I'll keep
you posted when I come up with something.

Bye,
     Mike

Attachment: gcstat_inc_no_gen.png
Description: PNG image