linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
To: Peng Zhang <zhangpeng.00@bytedance.com>
Cc: akpm@linux-foundation.org, linux-mm@kvack.org,
	linux-kernel@vger.kernel.org, maple-tree@lists.infradead.org
Subject: Re: [PATCH v2 08/10] maple_tree: Rework mas_wr_slot_store() to be cleaner and more efficient.
Date: Wed, 17 May 2023 18:48:12 -0400	[thread overview]
Message-ID: <20230517224812.knkhqmrweggsbnuk@revolver> (raw)
In-Reply-To: <20230517085809.86522-9-zhangpeng.00@bytedance.com>

* Peng Zhang <zhangpeng.00@bytedance.com> [230517 04:59]:
> The code of mas_wr_slot_store() is messy, make it clearer and concise,
> and add comments. In addition, get whether the two gaps are empty to
> avoid calling mas_update_gap() all the time.
> 
> Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
> ---
>  lib/maple_tree.c | 59 ++++++++++++++++++++----------------------------
>  1 file changed, 24 insertions(+), 35 deletions(-)
> 
> diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> index bbe4c6f2858c..25a8b7d5d598 100644
> --- a/lib/maple_tree.c
> +++ b/lib/maple_tree.c
> @@ -4200,49 +4200,38 @@ static inline bool mas_wr_node_store(struct ma_wr_state *wr_mas)
>  static inline bool mas_wr_slot_store(struct ma_wr_state *wr_mas)
>  {
>  	struct ma_state *mas = wr_mas->mas;
> -	unsigned long lmax; /* Logical max. */
>  	unsigned char offset = mas->offset;
> +	unsigned char offset_end = wr_mas->offset_end;
> +	unsigned long lmax = wr_mas->end_piv; /* Logical max. */
> +	bool gap = false;
>  
> -	if ((wr_mas->r_max > mas->last) && ((wr_mas->r_min != mas->index) ||
> -				  (offset != wr_mas->node_end)))
> +	if (offset_end - offset != 1)
>  		return false;
>  
> -	if (offset == wr_mas->node_end - 1)
> -		lmax = mas->max;
> -	else
> -		lmax = wr_mas->pivots[offset + 1];
> -
> -	/* going to overwrite too many slots. */
> -	if (lmax < mas->last)
> -		return false;
> -
> -	if (wr_mas->r_min == mas->index) {
> -		/* overwriting two or more ranges with one. */
> -		if (lmax == mas->last)
> -			return false;
> -
> -		/* Overwriting all of offset and a portion of offset + 1. */
> +	if (mas->index == wr_mas->r_min && mas->last < lmax) {
> +		/* Overwriting the range and over a part of the next range. */
> +		gap |= !mt_slot_locked(mas->tree, wr_mas->slots, offset);
> +		gap |= !mt_slot_locked(mas->tree, wr_mas->slots, offset + 1);
>  		rcu_assign_pointer(wr_mas->slots[offset], wr_mas->entry);
>  		wr_mas->pivots[offset] = mas->last;
> -		goto done;
> -	}
> -
> -	/* Doesn't end on the next range end. */
> -	if (lmax != mas->last)
> +	} else if (mas->index > wr_mas->r_min && mas->last == lmax) {
> +		/* Overwriting a part of the range and over the next range */
> +		gap |= !mt_slot_locked(mas->tree, wr_mas->slots, offset);
> +		gap |= !mt_slot_locked(mas->tree, wr_mas->slots, offset + 1);
> +		rcu_assign_pointer(wr_mas->slots[offset + 1], wr_mas->entry);
> +		wr_mas->pivots[offset] = mas->index - 1;
> +		mas->offset++; /* Keep mas accurate. */
> +	} else {
>  		return false;

Again here, I think you have already verified this will be a slot store
by offset_end - offset == 1 and new_end == wr_mas->node_end.

The checking of the slots for the gap could be moved outside the
statements.  WDYT?

> +	}
>  
> -	/* Overwriting a portion of offset and all of offset + 1 */
> -	if ((offset + 1 < mt_pivots[wr_mas->type]) &&
> -	    (wr_mas->entry || wr_mas->pivots[offset + 1]))
> -		wr_mas->pivots[offset + 1] = mas->last;
> -
> -	rcu_assign_pointer(wr_mas->slots[offset + 1], wr_mas->entry);
> -	wr_mas->pivots[offset] = mas->index - 1;
> -	mas->offset++; /* Keep mas accurate. */
> -
> -done:
>  	trace_ma_write(__func__, mas, 0, wr_mas->entry);
> -	mas_update_gap(mas);
> +	/*
> +	 * Only update gap when the new entry is empty or there is an empty
> +	 * entry in the original two ranges.
> +	 */
> +	if (!wr_mas->entry || gap)
> +		mas_update_gap(mas);
>  	return true;
>  }
>  
> @@ -4396,7 +4385,7 @@ static inline void mas_wr_modify(struct ma_wr_state *wr_mas)
>  	if (new_end == wr_mas->node_end + 1 && mas_wr_append(wr_mas))
>  		return;
>  
> -	if ((wr_mas->offset_end - mas->offset <= 1) && mas_wr_slot_store(wr_mas))
> +	if (new_end == wr_mas->node_end && mas_wr_slot_store(wr_mas))
>  		return;
>  	else if (mas_wr_node_store(wr_mas))
>  		return;
> -- 
> 2.20.1
> 


  reply	other threads:[~2023-05-17 22:48 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-17  8:57 [PATCH v2 00/10] Clean ups for maple tree Peng Zhang
2023-05-17  8:58 ` [PATCH v2 01/10] maple_tree: Rework mtree_alloc_{range,rrange}() Peng Zhang
2023-05-17 18:17   ` Liam R. Howlett
2023-05-18  6:10     ` Peng Zhang
2023-05-18 13:47       ` Liam R. Howlett
2023-05-17  8:58 ` [PATCH v2 02/10] maple_tree: Drop mas_{rev_}alloc() and mas_fill_gap() Peng Zhang
2023-05-17  8:58 ` [PATCH v2 03/10] maple_tree: Fix the arguments to __must_hold() Peng Zhang
2023-05-17  8:58 ` [PATCH v2 04/10] maple_tree: Simplify mas_is_span_wr() Peng Zhang
2023-05-17 18:35   ` Liam R. Howlett
2023-05-17  8:58 ` [PATCH v2 05/10] maple_tree: Make the code symmetrical in mas_wr_extend_null() Peng Zhang
2023-05-17  8:58 ` [PATCH v2 06/10] maple_tree: Add mas_wr_new_end() to calculate new_end accurately Peng Zhang
2023-05-17 19:21   ` Liam R. Howlett
2023-05-17  8:58 ` [PATCH v2 07/10] maple_tree: Add comments and some minor cleanups to mas_wr_append() Peng Zhang
2023-05-17 19:33   ` Liam R. Howlett
2023-05-18  9:27     ` Peng Zhang
2023-05-17  8:58 ` [PATCH v2 08/10] maple_tree: Rework mas_wr_slot_store() to be cleaner and more efficient Peng Zhang
2023-05-17 22:48   ` Liam R. Howlett [this message]
2023-05-18  9:27     ` Peng Zhang
2023-05-17  8:58 ` [PATCH v2 09/10] maple_tree: Simplify and clean up mas_wr_node_store() Peng Zhang
2023-05-17  8:58 ` [PATCH v2 10/10] maple_tree: Relocate the declaration of mas_empty_area_rev() Peng Zhang

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=20230517224812.knkhqmrweggsbnuk@revolver \
    --to=liam.howlett@oracle.com \
    --cc=akpm@linux-foundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=maple-tree@lists.infradead.org \
    --cc=zhangpeng.00@bytedance.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox