From: Duy Nguyen <pclouds@gmail.com>
To: Jeff King <peff@peff.net>
Cc: Git Mailing List <git@vger.kernel.org>
Subject: Re: [PATCH 03/12] shallow.c: use commit-slab for commit depth instead of commit->util
Date: Sat, 12 May 2018 13:59:00 +0200 [thread overview]
Message-ID: <CACsJy8BKxKvu-M3xdp5RPt8YaXe_Q7dhdO22BQT-q0AiOaE7vQ@mail.gmail.com> (raw)
In-Reply-To: <20180512090748.GB28279@sigill.intra.peff.net>
On Sat, May 12, 2018 at 11:07 AM, Jeff King <peff@peff.net> wrote:
> On Sat, May 12, 2018 at 10:00:19AM +0200, Nguyễn Thái Ngọc Duy wrote:
>
>> @@ -82,25 +84,29 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
>> struct object_array stack = OBJECT_ARRAY_INIT;
>> struct commit *commit = NULL;
>> struct commit_graft *graft;
>> + struct commit_depth depths;
>>
>> + init_commit_depth(&depths);
>> while (commit || i < heads->nr || stack.nr) {
>> struct commit_list *p;
>> if (!commit) {
>> if (i < heads->nr) {
>> + int **depth_slot;
>> commit = (struct commit *)
>> deref_tag(heads->objects[i++].item, NULL, 0);
>> if (!commit || commit->object.type != OBJ_COMMIT) {
>> commit = NULL;
>> continue;
>> }
>> - if (!commit->util)
>> - commit->util = xmalloc(sizeof(int));
>> - *(int *)commit->util = 0;
>> + depth_slot = commit_depth_at(&depths, commit);
>> + if (!*depth_slot)
>> + *depth_slot = xmalloc(sizeof(int));
>> + **depth_slot = 0;
>
> It looks like we could save ourselves an extra layer of indirection (and
> tiny heap allocations) by just storing an "int" directly in the slab.
> Do we ever use the NULL as a sentinel value?
>
> Here we just allocate it if not set. Let's see if we can find some
> others...
>
>> @@ -116,25 +122,32 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
>> }
>> commit->object.flags |= not_shallow_flag;
>> for (p = commit->parents, commit = NULL; p; p = p->next) {
>> - if (!p->item->util) {
>> - int *pointer = xmalloc(sizeof(int));
>> - p->item->util = pointer;
>> - *pointer = cur_depth;
>> + int **depth_slot = commit_depth_at(&depths, p->item);
>> + if (!*depth_slot) {
>> + *depth_slot = xmalloc(sizeof(int));
>> + **depth_slot = cur_depth;
>> } else {
>> - int *pointer = p->item->util;
>> - if (cur_depth >= *pointer)
>> + if (cur_depth >= **depth_slot)
>> continue;
>> - *pointer = cur_depth;
>> + **depth_slot = cur_depth;
>> }
>
> Here we malloc again if it's not set. But we do behave slightly
> differently when we see NULL, in that we do not bother to even compare
> against cur_depth. So if we were to directly store ints, we'd see "0" as
> the sentinel depth here, which would not match our "cur_depth >=
> depth_slot" check.
>
> So no, it wouldn't work to directly store depths with the code as
> written. I'm not sure if the depth can ever be 0. If not, then it would
> be a suitable sentinel as:
>
> int *slot = commit_depth_at(&depths, p->item);
> if (!*slot || cur_depth < *slot)
> *slot = cur_depth;
>
> But somebody would have to dig into the possible values of cur_depth
> there (which would make sense to do as a separate patch anyway, since
> the point of this is to be a direct conversion).
I actually tried that first, going with storing int directly in the
slab instead of int*. And some shallow tests failed so I didn't bother
(the goal was to get rid of 'util' pointer, not to optimize more)
--
Duy
next prev parent reply other threads:[~2018-05-12 11:59 UTC|newest]
Thread overview: 82+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-05-12 8:00 [PATCH 00/12] Die commit->util, die! Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 01/12] blame: use commit-slab for blame suspects instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-12 9:22 ` Jeff King
2018-05-12 12:13 ` Duy Nguyen
2018-05-12 8:00 ` [PATCH 02/12] describe: use commit-slab for commit names " Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 03/12] shallow.c: use commit-slab for commit depth " Nguyễn Thái Ngọc Duy
2018-05-12 9:07 ` Jeff King
2018-05-12 9:18 ` Jeff King
2018-05-12 12:09 ` Duy Nguyen
2018-05-12 19:12 ` Jeff King
2018-05-12 11:59 ` Duy Nguyen [this message]
2018-05-12 8:00 ` [PATCH 04/12] sequencer.c: use commit-slab to mark seen commits Nguyễn Thái Ngọc Duy
2018-05-12 9:25 ` Jeff King
2018-05-12 13:43 ` Junio C Hamano
2018-05-12 14:00 ` Duy Nguyen
2018-05-12 8:00 ` [PATCH 05/12] sequencer.c: use commit-slab to associate todo items to commits Nguyễn Thái Ngọc Duy
2018-05-12 9:28 ` Jeff King
2018-05-12 8:00 ` [PATCH 06/12] revision.c: use commit-slab for show_source Nguyễn Thái Ngọc Duy
2018-05-12 9:33 ` Jeff King
2018-05-12 13:58 ` Junio C Hamano
2018-05-12 14:13 ` Duy Nguyen
2018-05-12 19:06 ` Jeff King
2018-05-12 8:00 ` [PATCH 07/12] bisect.c: use commit-slab for commit weight instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-12 13:59 ` Junio C Hamano
2018-05-12 8:00 ` [PATCH 08/12] name-rev: use commit-slab for rev-name " Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 09/12] show-branch: use commit-slab for commit-name " Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 10/12] log: use commit-slab in prepare_bases() " Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 11/12] merge: use commit-slab in merge remote desc " Nguyễn Thái Ngọc Duy
2018-05-12 8:00 ` [PATCH 12/12] commit.h: delete 'util' field in struct commit Nguyễn Thái Ngọc Duy
2018-05-12 9:41 ` [PATCH 00/12] Die commit->util, die! Jeff King
2018-05-12 18:50 ` Jakub Narebski
2018-05-13 5:39 ` Duy Nguyen
2018-05-13 5:51 ` [PATCH v2 00/14] " Nguyễn Thái Ngọc Duy
2018-05-13 5:51 ` [PATCH v2 01/14] commit-slab.h: code split Nguyễn Thái Ngọc Duy
2018-05-13 23:33 ` Junio C Hamano
2018-05-13 5:51 ` [PATCH v2 02/14] commit-slab: support shared commit-slab Nguyễn Thái Ngọc Duy
2018-05-13 23:36 ` Junio C Hamano
2018-05-13 5:51 ` [PATCH v2 03/14] blame: use commit-slab for blame suspects instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-13 23:42 ` Junio C Hamano
2018-05-13 5:51 ` [PATCH v2 04/14] describe: use commit-slab for commit names " Nguyễn Thái Ngọc Duy
2018-05-13 23:45 ` Junio C Hamano
2018-05-13 5:51 ` [PATCH v2 05/14] shallow.c: use commit-slab for commit depth " Nguyễn Thái Ngọc Duy
2018-05-13 5:52 ` [PATCH v2 06/14] sequencer.c: use commit-slab to mark seen commits Nguyễn Thái Ngọc Duy
2018-05-14 4:17 ` Junio C Hamano
2018-05-13 5:52 ` [PATCH v2 07/14] sequencer.c: use commit-slab to associate todo items to commits Nguyễn Thái Ngọc Duy
2018-05-13 5:52 ` [PATCH v2 08/14] revision.c: use commit-slab for show_source Nguyễn Thái Ngọc Duy
2018-05-14 5:10 ` Junio C Hamano
2018-05-14 5:37 ` Junio C Hamano
2018-05-13 5:52 ` [PATCH v2 09/14] bisect.c: use commit-slab for commit weight instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-14 6:16 ` Junio C Hamano
2018-05-13 5:52 ` [PATCH v2 10/14] name-rev: use commit-slab for rev-name " Nguyễn Thái Ngọc Duy
2018-05-13 5:52 ` [PATCH v2 11/14] show-branch: use commit-slab for commit-name " Nguyễn Thái Ngọc Duy
2018-05-14 6:45 ` Junio C Hamano
2018-05-19 4:51 ` Duy Nguyen
2018-05-13 5:52 ` [PATCH v2 12/14] log: use commit-slab in prepare_bases() " Nguyễn Thái Ngọc Duy
2018-05-13 5:52 ` [PATCH v2 13/14] merge: use commit-slab in merge remote desc " Nguyễn Thái Ngọc Duy
2018-05-18 2:16 ` Junio C Hamano
2018-05-18 17:54 ` Ramsay Jones
2018-05-13 5:52 ` [PATCH v2 14/14] commit.h: delete 'util' field in struct commit Nguyễn Thái Ngọc Duy
2018-05-14 7:52 ` Junio C Hamano
2018-05-14 16:07 ` Duy Nguyen
2018-05-14 17:30 ` Duy Nguyen
2018-05-14 18:14 ` Derrick Stolee
2018-05-19 5:41 ` Duy Nguyen
2018-05-19 5:28 ` [PATCH v3 00/15] Die commit->util, die! Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 01/15] commit-slab.h: code split Nguyễn Thái Ngọc Duy
2018-05-21 5:11 ` Junio C Hamano
2018-05-19 5:28 ` [PATCH v3 02/15] commit-slab: support shared commit-slab Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 03/15] blame: use commit-slab for blame suspects instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 04/15] describe: use commit-slab for commit names " Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 05/15] shallow.c: use commit-slab for commit depth " Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 06/15] sequencer.c: use commit-slab to mark seen commits Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 07/15] sequencer.c: use commit-slab to associate todo items to commits Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 08/15] revision.c: use commit-slab for show_source Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 09/15] bisect.c: use commit-slab for commit weight instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 10/15] name-rev: use commit-slab for rev-name " Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 11/15] show-branch: use commit-slab for commit-name " Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 12/15] show-branch: note about its object flags usage Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 13/15] log: use commit-slab in prepare_bases() instead of commit->util Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 14/15] merge: use commit-slab in merge remote desc " Nguyễn Thái Ngọc Duy
2018-05-19 5:28 ` [PATCH v3 15/15] commit.h: delete 'util' field in struct commit Nguyễn Thái Ngọc Duy
2018-05-22 22:40 ` [PATCH 00/12] Die commit->util, die! Stefan Beller
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CACsJy8BKxKvu-M3xdp5RPt8YaXe_Q7dhdO22BQT-q0AiOaE7vQ@mail.gmail.com \
--to=pclouds@gmail.com \
--cc=git@vger.kernel.org \
--cc=peff@peff.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).