All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ross Zwisler <zwisler@google.com>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-trace-devel@vger.kernel.org,
	Stevie Alvarez <stevie.6strings@gmail.com>
Subject: Re: [PATCH v2 07/17] libtraceeval: Convert hist array into a hash table
Date: Tue, 15 Aug 2023 12:44:21 -0600	[thread overview]
Message-ID: <20230815184421.GC780024@google.com> (raw)
In-Reply-To: <20230811053940.1408424-8-rostedt@goodmis.org>

On Fri, Aug 11, 2023 at 01:39:30AM -0400, Steven Rostedt wrote:
> From: "Steven Rostedt (Google)" <rostedt@goodmis.org>
> 
> The lookups need to be extremely fast. Instead of doing a linear search
> across all entries (which could be thousands), do a hash lookup instead.
> 
> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
> ---
>  include/traceeval-hist.h |   7 ++
>  src/histograms.c         | 172 +++++++++++++++++++++++++++++++++------
>  2 files changed, 153 insertions(+), 26 deletions(-)
> 
> diff --git a/include/traceeval-hist.h b/include/traceeval-hist.h
> index 412efdbe8681..4baed4237787 100644
> --- a/include/traceeval-hist.h
> +++ b/include/traceeval-hist.h
<>
> @@ -12,6 +13,14 @@
>  
>  #include <traceeval-hist.h>
>  
> +#define offset_of(type, field) (&(((type *)(NULL))->field))

This currently returns a pointer type, but the kernel implementation(s) have
this cast back to a size_t, which makes sense I think.

https://elixir.bootlin.com/linux/latest/source/tools/include/nolibc/types.h#L220

> +#define container_of(ptr, type, field) \
> +	(type *)((void *)(ptr) - (void *)offset_of(type, field))
> +
> +#define HASH_BITS 10	/* Start with 1K of buckets */
> +#define HASH_SIZE(bits)	(1 << (bits))
> +#define HASH_MASK(bits)	(HASH_SIZE(bits) - 1)
> +
>  /*
>   * Compare two integers of variable length.
>   *
<>
> @@ -365,16 +406,19 @@ static void clean_entry(struct entry *entry, struct traceeval *teval)
>   */
>  static void hist_table_release(struct traceeval *teval)

This probably wants to be 'hash_table_release()' now since 'hist_table' is
gone.

>  {
> -	struct hist_table *hist = teval->hist;
> +	struct hash_table *hist = teval->hist;
>  
>  	if (!hist)
>  		return;
>  
> -	for (size_t i = 0; i < hist->nr_entries; i++) {
> -		clean_entry(hist->map + i, teval);
> +	for (size_t i = 0; i < HASH_SIZE(hist->bits); i++) {
> +		if (!hist->hash[i])
> +			continue;
> +
> +		free_entries(teval, hist->hash[i]);
>  	}
>  
> -	free(hist->map);
> +	free(hist->hash);
>  	free(hist);
>  	teval->hist = NULL;
>  }

  reply	other threads:[~2023-08-15 18:45 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-08-11  5:39 [PATCH v2 00/17] libtraceeval histogram: Updates Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 01/17] libtraceeval histograms: Fix traceeval_results_release() error message Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 02/17] libtraceeval: Add sample task-eval program Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 03/17] libtraceeval hist: Add pointer and const string types Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 04/17] libtraceeval histogram: Have cmp and release functions be generic Steven Rostedt
2023-08-15 16:50   ` Ross Zwisler
2023-08-15 18:52     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 05/17] libtraceeval histograms: Add traceeval struct to compare function Steven Rostedt
2023-08-15 16:55   ` Ross Zwisler
2023-08-15 18:53     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 06/17] libtraceeval histogram: Remove comparing of traceeval and types Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 07/17] libtraceeval: Convert hist array into a hash table Steven Rostedt
2023-08-15 18:44   ` Ross Zwisler [this message]
2023-08-15 19:05     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 08/17] libtraceeval histograms: Move hash functions into their own file Steven Rostedt
2023-08-15 19:31   ` Ross Zwisler
2023-08-15 20:23     ` Steven Rostedt
2023-08-15 22:56       ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 09/17] libtraceeval histogram: Label and check keys and values Steven Rostedt
2023-08-15 19:48   ` Ross Zwisler
2023-08-15 20:24     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 10/17] libtraceeval histogram: Add updating of stats Steven Rostedt
2023-08-15 20:25   ` Ross Zwisler
2023-08-15 20:55     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 11/17] libtraceeval histogram: Add iterator APIs Steven Rostedt
2023-08-16 21:34   ` Ross Zwisler
2023-08-16 21:49     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 12/17] libtraceeval histogram: Add data copy callback Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 13/17] libtraceeval histogram: Do the release on updates Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 14/17] libtraceeval histogram: Use stack for old copy in update Steven Rostedt
2023-08-16 22:37   ` Ross Zwisler
2023-08-16 23:12     ` Steven Rostedt
2023-08-17  1:03       ` Steven Rostedt
2023-08-17  1:13       ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 15/17] libtraceeval histogram: Add traceeval_iterator_sort_custom() Steven Rostedt
2023-08-16 22:57   ` Ross Zwisler
2023-08-16 23:22     ` Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 16/17] libtraceeval histogram: Have traceeval_query() just give the pointer to results Steven Rostedt
2023-08-11  5:39 ` [PATCH v2 17/17] libtraceeval samples: Update task-eval to use the histogram logic Steven Rostedt

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=20230815184421.GC780024@google.com \
    --to=zwisler@google.com \
    --cc=linux-trace-devel@vger.kernel.org \
    --cc=rostedt@goodmis.org \
    --cc=stevie.6strings@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.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.