From: Jason Merrill <jason@redhat.com>
To: Patrick Palka <patrick@parcs.ath.cx>
Cc: gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] Fix PR c++/70452 (regression in C++ parsing performance)
Date: Mon, 04 Apr 2016 15:39:00 -0000 [thread overview]
Message-ID: <57028AA7.6090807@redhat.com> (raw)
In-Reply-To: <alpine.DEB.2.20.11.1604021715510.1308@idea>
On 04/02/2016 05:18 PM, Patrick Palka wrote:
> Here's a version that uses a separate deletable table to cache the
> function copies. For simplicity I used a hash_map instead of a
> hash_table. Does this look OK to commit after bootstrap + regtest?
Thanks. Minor nits:
> +struct fundef_copies_table_t
> +{
> + hash_map<tree, fun_copy *> *map;
> +};
Why wrap the pointer in a struct?
> + maybe_initialize_fundef_copies_table ();
> + fun_copy *copy = get_fun_copy (fun);
Let's move the initialization call inside get_fun_copy.
> On a related note, I noticed that the constexpr_call_table is not marked
> deletable. Marking it deletable speeds up the test case in the PR by
> about 10% and saves about 10MB. Do you think doing so is a good idea?
Please.
> On another related note, I noticed that marking something as both
> GTY((deletable, cache)) doesn't work as intended, because the
> gt_cleare_cache functions run _after_ all deletable roots get
> zeroed out. So during GC the gt_cleare_cache function of a root
> marked "deletable, cache" would always be a no-op. Concretely I think
> this means that our cv_cache and fold_cache leak memory during GC
> because their underlying hash_map (allocated by operator new) is zeroed
> before gc_cleare_cache could clear it.
Hmm, I thought I remembered hitting the breakpoint in gt_cleare_cache
and it being non-null. But I guess we can get rid of the cache_map
class and use the approach you have here, of a deletable gc-allocated
hash_map pointer; I'd still use ->empty() for dumping the cache outside
of GC, though.
Jason
next prev parent reply other threads:[~2016-04-04 15:39 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-01 19:13 Patrick Palka
2016-04-01 19:25 ` Patrick Palka
2016-04-02 1:28 ` Jason Merrill
2016-04-02 1:55 ` Patrick Palka
2016-04-02 2:10 ` Patrick Palka
2016-04-02 21:18 ` Patrick Palka
2016-04-03 3:52 ` Trevor Saunders
2016-04-03 22:06 ` Patrick Palka
2016-04-04 15:39 ` Jason Merrill [this message]
2016-04-04 16:55 ` Patrick Palka
2016-04-05 13:06 ` Jason Merrill
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=57028AA7.6090807@redhat.com \
--to=jason@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=patrick@parcs.ath.cx \
/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 read-only IMAP folder(s) and NNTP newsgroup(s).