From: Derrick Stolee <stolee@gmail.com>
To: "brian m. carlson" <sandals@crustytoothpaste.net>,
Jeff King <peff@peff.net>,
git@vger.kernel.org, Jacob Keller <jacob.keller@gmail.com>
Subject: Re: [PATCH 0/9] introducing oideq()
Date: Mon, 27 Aug 2018 08:41:32 -0400 [thread overview]
Message-ID: <dd26d177-7e37-83ac-7275-03db962a282c@gmail.com> (raw)
In-Reply-To: <20180826205620.GC873448@genre.crustytoothpaste.net>
On 8/26/2018 4:56 PM, brian m. carlson wrote:
> On Sat, Aug 25, 2018 at 04:00:31AM -0400, Jeff King wrote:
>> This is a follow-up to the discussion in:
>>
>> https://public-inbox.org/git/20180822030344.GA14684@sigill.intra.peff.net/
>>
>> The general idea is that the majority of callers don't care about actual
>> plus/minus ordering from oidcmp() and hashcmp(); they just want to know
>> if two oids are equal. The stricter equality check can be optimized
>> better by the compiler.
>>
>> Due to the simplicity of the current code and our inlining, the compiler
>> can usually figure this out for now. So I wouldn't expect this patch to
>> actually improve performance right away. But as that discussion shows,
>> we are likely to take a performance hit as we move to more runtime
>> determination of the_hash_algo parameters. Having these callers use the
>> more strict form will potentially help us recover that.
>>
>> So in that sense we _could_ simply punt on this series until then (and
>> it's certainly post-v2.19 material). But I think it's worth doing now,
>> simply from a readability/annotation standpoint. IMHO the resulting code
>> is more clear (though I've long since taught myself to read !foocmp() as
>> equality).
> I would quite like to see this series picked up for v2.20. If we want
> to minimize performance regressions with the SHA-256 work, I think it's
> important.
Seconded.
> Applying the following patch on top of this series causes gcc to inline
> both branches, which is pretty much the best we can expect. I haven't
> benchmarked it, though, so I can't say what the actual performance
> consequence is.
We should definitely measure the cost here, but when we have the option
to move to newhash, that cost should be acceptable.
Thanks,
-Stolee
next prev parent reply other threads:[~2018-08-27 12:41 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-25 8:00 [PATCH 0/9] introducing oideq() Jeff King
2018-08-25 8:03 ` [PATCH 1/9] coccinelle: use <...> for function exclusion Jeff King
2018-08-25 8:05 ` [PATCH 2/9] introduce hasheq() and oideq() Jeff King
2018-08-25 10:58 ` Andrei Rybak
2018-08-25 8:07 ` [PATCH 3/9] convert "oidcmp() == 0" to oideq() Jeff King
2018-08-25 8:36 ` Jeff King
2018-08-27 12:31 ` Derrick Stolee
2018-08-27 12:33 ` Derrick Stolee
2018-08-25 8:08 ` [PATCH 4/9] convert "hashcmp() == 0" to hasheq() Jeff King
2018-08-25 8:09 ` [PATCH 5/9] convert "oidcmp() != 0" to "!oideq()" Jeff King
2018-08-25 8:10 ` [PATCH 6/9] convert "hashcmp() != 0" to "!hasheq()" Jeff King
2018-08-25 8:14 ` [PATCH 7/9] convert hashmap comparison functions to oideq() Jeff King
2018-08-25 8:15 ` [PATCH 8/9] read-cache: use oideq() in ce_compare functions Jeff King
2018-08-25 8:17 ` [PATCH 9/9] show_dirstat: simplify same-content check Jeff King
2018-08-25 8:20 ` Eric Sunshine
2018-08-25 8:23 ` Jeff King
2018-08-26 20:56 ` [PATCH 0/9] introducing oideq() brian m. carlson
2018-08-27 12:41 ` Derrick Stolee [this message]
2018-08-28 21:21 ` Jeff King
2018-08-28 21:22 ` [PATCH v2 1/9] coccinelle: use <...> for function exclusion Jeff King
2018-08-28 21:22 ` [PATCH v2 2/9] introduce hasheq() and oideq() Jeff King
2018-08-28 21:22 ` [PATCH v2 3/9] convert "oidcmp() == 0" to oideq() Jeff King
2018-08-28 21:22 ` [PATCH v2 4/9] convert "hashcmp() == 0" to hasheq() Jeff King
2018-08-28 21:22 ` [PATCH v2 5/9] convert "oidcmp() != 0" to "!oideq()" Jeff King
2018-08-28 21:22 ` [PATCH v2 6/9] convert "hashcmp() != 0" to "!hasheq()" Jeff King
2018-08-28 21:22 ` [PATCH v2 7/9] convert hashmap comparison functions to oideq() Jeff King
2018-08-28 21:22 ` [PATCH v2 8/9] read-cache: use oideq() in ce_compare functions Jeff King
2018-08-28 21:23 ` [PATCH v2 9/9] show_dirstat: simplify same-content check Jeff King
2018-08-28 21:36 ` [PATCH 0/9] introducing oideq() Derrick Stolee
2018-08-29 0:08 ` brian m. carlson
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=dd26d177-7e37-83ac-7275-03db962a282c@gmail.com \
--to=stolee@gmail.com \
--cc=git@vger.kernel.org \
--cc=jacob.keller@gmail.com \
--cc=peff@peff.net \
--cc=sandals@crustytoothpaste.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).