On Tue, Sep 20, 2011 at 10:26 AM, Jonathan Nieder <jrnieder@xxxxxxxxx> wrote: > Junio C Hamano wrote: > >> I gave the series a cursory look, and the patches all looked like a good >> and straight forward rewrites. Provided if it is indeed a good idea >> overall to stuff more objects in-core, that is. > > Right, that's exactly the question I had. When and why is it a good > idea to stuff more objects in-core (or when might it be a bad idea, > for that matter)? The next step would be to replace sha1 with struct object_entry* in fast-import. So it'll be in struct tree_entry (twice, for each of versions[2]), branch, tag, hash_list (used to store merge from lists), last_object. Then some fields will be deleted as they can be accessed from object_entry: last_object->depth last_object->offset tree_content->delta_depth branch,tag->pack_id And it all even slightly decreased memory consumption (checked some time ago, but think it's still true). Probably because of tree nodes having NULL instead of null_sha1 and then ptr instead of sha1; and maybe because for huge imports each object is from our pack so for a sha1 there always is object_entry anyway. In short, if there is nothing bad with this patchset, it'll be absolutely natural one after switch to oe instead of sha1, but it's put before to split the big series. And of course this part may have a small speedup of it's own. If it's not too good to be accepted on it's own, I'll just include it into future series depending on it. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html