All of lore.kernel.org
 help / color / mirror / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: "Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
Cc: git@vger.kernel.org
Subject: Re: [PATCH 2/2] index-pack: kill union delta_base to save memory
Date: Fri, 03 Jul 2015 09:51:57 -0700	[thread overview]
Message-ID: <xmqqoajt5glu.fsf@gitster.dls.corp.google.com> (raw)
In-Reply-To: <1429354025-24659-3-git-send-email-pclouds@gmail.com> ("Nguyễn	Thái Ngọc Duy"'s message of "Sat, 18 Apr 2015 17:47:05 +0700")

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Once we know the number of objects in the input pack, we allocate an
> array of nr_objects of struct delta_entry. On x86-64, this struct is
> 32 bytes long. The union delta_base, which is part of struct
> delta_entry, provides enough space to store either ofs-delta (8 bytes)
> or ref-delta (20 bytes).

Sorry for responding to a patch 7000+ messages ago, but some kind
folks at Google were puzzled by this code, and I think they found a
small bug.

>  static void fix_unresolved_deltas(struct sha1file *f, int nr_unresolved)
>  {
> -	struct delta_entry **sorted_by_pos;
> +	struct ref_delta_entry **sorted_by_pos;
>  	int i, n = 0;
>  
>  	/*
> @@ -1282,28 +1344,25 @@ static void fix_unresolved_deltas(struct sha1file *f, int nr_unresolved)
>  	 * resolving deltas in the same order as their position in the pack.
>  	 */
>  	sorted_by_pos = xmalloc(nr_unresolved * sizeof(*sorted_by_pos));
> -	for (i = 0; i < nr_deltas; i++) {
> -		if (objects[deltas[i].obj_no].real_type != OBJ_REF_DELTA)
> -			continue;
> -		sorted_by_pos[n++] = &deltas[i];
> -	}
> +	for (i = 0; i < nr_ref_deltas; i++)
> +		sorted_by_pos[n++] = &ref_deltas[i];
>  	qsort(sorted_by_pos, n, sizeof(*sorted_by_pos), delta_pos_compare);

You allocate an array of nr_unresolved (which is the sum of
nr_ref_deltas and nr_ofs_deltas in the new world order with patch)
entries, fill only the first nr_ref_deltas entries of it, and then
sort that first n (= nr_ref_deltas) entries.  The qsort and the
subsequent loop only looks at the first n entries, so nothing is
filling or reading these nr_ofs_deltas entres at the end.

I do not see any wrong behaviour other than temporary wastage of
nr_ofs_deltas pointers that would be caused by this, but this
allocation is misleading.

Also, the old code before this change had to use 'i' and 'n' because
some of the things we see in the (old) deltas[] array we scanned
with 'i' would not make it into the sorted_by_pos[] array in the old
world order, but now because you have only ref delta in a separate
ref_deltas[] array, they increment lock&step.  That also puzzled me
while re-reading this code.

Perhaps something like this is needed?


 builtin/index-pack.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/builtin/index-pack.c b/builtin/index-pack.c
index 48fa472..d6c48cd 100644
--- a/builtin/index-pack.c
+++ b/builtin/index-pack.c
@@ -1334,7 +1334,7 @@ static int delta_pos_compare(const void *_a, const void *_b)
 static void fix_unresolved_deltas(struct sha1file *f, int nr_unresolved)
 {
 	struct ref_delta_entry **sorted_by_pos;
-	int i, n = 0;
+	int i;
 
 	/*
 	 * Since many unresolved deltas may well be themselves base objects
@@ -1346,12 +1346,12 @@ static void fix_unresolved_deltas(struct sha1file *f, int nr_unresolved)
 	 * before deltas depending on them, a good heuristic is to start
 	 * resolving deltas in the same order as their position in the pack.
 	 */
-	sorted_by_pos = xmalloc(nr_unresolved * sizeof(*sorted_by_pos));
+	sorted_by_pos = xmalloc(nr_ref_deltas * sizeof(*sorted_by_pos));
 	for (i = 0; i < nr_ref_deltas; i++)
-		sorted_by_pos[n++] = &ref_deltas[i];
-	qsort(sorted_by_pos, n, sizeof(*sorted_by_pos), delta_pos_compare);
+		sorted_by_pos[i] = &ref_deltas[i];
+	qsort(sorted_by_pos, nr_ref_deltas, sizeof(*sorted_by_pos), delta_pos_compare);
 
-	for (i = 0; i < n; i++) {
+	for (i = 0; i < nr_ref_deltas; i++) {
 		struct ref_delta_entry *d = sorted_by_pos[i];
 		enum object_type type;
 		struct base_data *base_obj = alloc_base_data();

  reply	other threads:[~2015-07-03 16:52 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-04-18 10:47 [PATCH 0/2] nd/slim-index-pack-memory-usage update Nguyễn Thái Ngọc Duy
2015-04-18 10:47 ` [PATCH 1/2] index-pack: reduce object_entry size to save memory Nguyễn Thái Ngọc Duy
2015-04-18 10:47 ` [PATCH 2/2] index-pack: kill union delta_base " Nguyễn Thái Ngọc Duy
2015-07-03 16:51   ` Junio C Hamano [this message]
2015-07-03 18:29     ` Duy Nguyen
2015-07-04  1:21       ` [PATCH] index-pack: fix overallocation of sorted_by_pos array Junio C Hamano
2015-07-04 22:30         ` [PATCH] index-pack: fix allocation " Junio C Hamano
2015-07-06 23:23           ` Junio C Hamano
2015-07-07  0:36           ` Duy Nguyen
2015-07-07 15:49             ` Junio C Hamano
2015-07-07 16:06               ` Jeff King
2015-07-08 11:56                 ` [PATCH 1/2] index-pack: rename the field "type" to "in_pack_type" Nguyễn Thái Ngọc Duy
2015-07-08 11:56                   ` [PATCH 2/2] pack-objects: rename the field "type" to "real_type" Nguyễn Thái Ngọc Duy
2015-07-08 13:47                     ` Jeff King
2015-07-08 13:57                       ` Duy Nguyen
2015-07-08 14:11                         ` Jeff King
2015-07-04 22:24       ` [PATCH 2/2] index-pack: kill union delta_base to save memory Junio C Hamano
  -- strict thread matches above, loose matches on Subject: below --
2015-02-20  1:58 [PATCH 0/2] nd/slim-index-pack-memory-usage updates Nguyễn Thái Ngọc Duy
2015-02-20  1:58 ` [PATCH 2/2] index-pack: kill union delta_base to save memory Nguyễn Thái Ngọc Duy

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=xmqqoajt5glu.fsf@gitster.dls.corp.google.com \
    --to=gitster@pobox.com \
    --cc=git@vger.kernel.org \
    --cc=pclouds@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.