From: Jonathan Tan <jonathantanmy@google.com>
To: "René Scharfe" <l.s.r@web.de>
Cc: git@vger.kernel.org, stolee@gmail.com
Subject: Re: [PATCH 2/2] packfile: refactor hash search with fanout table
Date: Fri, 9 Feb 2018 11:50:51 -0800 [thread overview]
Message-ID: <20180209115051.b9356543f3f7d07f3bae213f@google.com> (raw)
In-Reply-To: <cfbde137-dbac-8796-f49f-2a543303d33a@web.de>
On Fri, 9 Feb 2018 19:03:48 +0100
René Scharfe <l.s.r@web.de> wrote:
> Going from unsigned to signed int means the patch breaks support for
> more than 2G pack entries, which was put with 326bf39677 (Use uint32_t
> for all packed object counts.) in 2007.
Ah, good catch. I'll wait to see if there are any more comments, then
send out a new version.
> > +int bsearch_hash(const unsigned char *sha1, const void *fanout_,
> > + const void *table_, size_t stride)
> > +{
> > + const uint32_t *fanout = fanout_;
>
> Why hide the type? It doesn't make the function more generic.
I thought that the fanout_ parameter could come from a variety of
sources (e.g. direct mmap - void *, or mmap with some pointer arithmetic
- char *) so I just picked the generic one. But now I realize that that
could lead to unaligned reads, which is probably not a good idea. I'll
update it.
For consistency, I'll also update table_ to be unsigned char *.
(Unsigned because it is primarily interpreted as hashes, which use
"unsigned char *" in the Git code.)
> Why not use sha1_pos()? I guess because it avoids the overhead of the
> accessor function, right? And I wonder how much of difference it makes.
Yes, overhead of the accessor function. We would also need to modify
sha1_pos to take in a function that we can pass userdata to (to contain
the stride).
> A binary search function for embedded hashes just needs the key, a
> pointer to the first hash in the array, the stride and the number of
> elements. It can then be used with or without a fanout table, making it
> more versatile. Just a thought.
I specifically want to include the fanout table in the calculation here,
because it will be used by subsequent patches that also incorporate the
fanout table.
next prev parent reply other threads:[~2018-02-09 19:50 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-02-02 22:36 [PATCH 0/2] Refactor hash search with fanout table Jonathan Tan
2018-02-02 22:36 ` [PATCH 1/2] packfile: remove GIT_DEBUG_LOOKUP log statements Jonathan Tan
2018-02-02 22:36 ` [PATCH 2/2] packfile: refactor hash search with fanout table Jonathan Tan
2018-02-09 18:03 ` René Scharfe
2018-02-09 19:50 ` Jonathan Tan [this message]
2018-02-02 23:30 ` [PATCH 0/2] Refactor " Junio C Hamano
2018-02-03 2:09 ` Derrick Stolee
2018-02-13 18:39 ` [PATCH v2 " Jonathan Tan
2018-02-13 18:39 ` [PATCH v2 1/2] packfile: remove GIT_DEBUG_LOOKUP log statements Jonathan Tan
2018-02-13 18:39 ` [PATCH v2 2/2] packfile: refactor hash search with fanout table Jonathan Tan
2018-02-13 18:52 ` [PATCH v2 0/2] Refactor " Derrick Stolee
2018-02-13 19:57 ` Junio C Hamano
2018-02-13 20:15 ` Jonathan Tan
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
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180209115051.b9356543f3f7d07f3bae213f@google.com \
--to=jonathantanmy@google.com \
--cc=git@vger.kernel.org \
--cc=l.s.r@web.de \
--cc=stolee@gmail.com \
/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.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
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 read-only IMAP folder(s) and NNTP newsgroup(s).