From: Martin Sebor <msebor@gmail.com>
To: Jan Hubicka <hubicka@kam.mff.cuni.cz>,
gcc-patches@gcc.gnu.org, msebor@gcc.gnu.org
Subject: Re: Enable more type attributes for signature changes
Date: Mon, 15 Nov 2021 08:38:25 -0700 [thread overview]
Message-ID: <969fcccd-e6cd-6e18-cfe2-57dab6103ff6@gmail.com> (raw)
In-Reply-To: <20211113144438.GA952@kam.mff.cuni.cz>
On 11/13/21 7:44 AM, Jan Hubicka wrote:
> Hi,
> this patch whitelists attributes that are safe for attribute changes and
> also makes access attribute dropped if function sigunature is changed.
> We could do better by updating the attribute, but doing so seems to be
> bit snowballing since with LTO the warnings produced seems bit confused.
> We would also like to output original name of function
> instead of mangledname.constprop or so. I looked into what attributes
> are dorpped in bootstrap and it does not look too bad.
>
> Bootstrapped/regtested x86_64-linux, will commit it shortly.
I've always intended the access attribute to eventually benefit
optimization so please feel free (and encouraged :) to use it
for that purpose.
Martin
>
> Honza
>
> gcc/ChangeLog:
>
> * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
> * ipa-param-manipulation.c (ipa_param_adjustments::type_attribute_allowed_p):
> New member function.
> (drop_type_attribute_if_params_changed_p): New function.
> (build_adjusted_function_type): Use it.
> * ipa-param-manipulation.h: Add type_attribute_allowed_p.
>
> diff --git a/gcc/ipa-fnsummary.c b/gcc/ipa-fnsummary.c
> index 94a80d3ec90..7e9201a554a 100644
> --- a/gcc/ipa-fnsummary.c
> +++ b/gcc/ipa-fnsummary.c
> @@ -3141,8 +3141,8 @@ compute_fn_summary (struct cgraph_node *node, bool early)
> modref summaries. */
> for (tree list = TYPE_ATTRIBUTES (TREE_TYPE (node->decl));
> list && !no_signature; list = TREE_CHAIN (list))
> - if (!flag_ipa_modref
> - || !is_attribute_p ("fn spec", get_attribute_name (list)))
> + if (!ipa_param_adjustments::type_attribute_allowed_p
> + (get_attribute_name (list)))
> {
> if (dump_file)
> {
> diff --git a/gcc/ipa-param-manipulation.c b/gcc/ipa-param-manipulation.c
> index 991db0d9b1b..29268fa5a58 100644
> --- a/gcc/ipa-param-manipulation.c
> +++ b/gcc/ipa-param-manipulation.c
> @@ -279,6 +279,32 @@ fill_vector_of_new_param_types (vec<tree> *new_types, vec<tree> *otypes,
> }
> }
>
> +/* Return false if given attribute should prevent type adjustments. */
> +
> +bool
> +ipa_param_adjustments::type_attribute_allowed_p (tree name)
> +{
> + if ((is_attribute_p ("fn spec", name) && flag_ipa_modref)
> + || is_attribute_p ("access", name)
> + || is_attribute_p ("returns_nonnull", name)
> + || is_attribute_p ("assume_aligned", name)
> + || is_attribute_p ("nocf_check", name)
> + || is_attribute_p ("warn_unused_result", name))
> + return true;
> + return false;
> +}
> +
> +/* Return true if attribute should be dropped if parameter changed. */
> +
> +static bool
> +drop_type_attribute_if_params_changed_p (tree name)
> +{
> + if (is_attribute_p ("fn spec", name)
> + || is_attribute_p ("access", name))
> + return true;
> + return false;
> +}
> +
> /* Build and return a function type just like ORIG_TYPE but with parameter
> types given in NEW_PARAM_TYPES - which can be NULL if, but only if,
> ORIG_TYPE itself has NULL TREE_ARG_TYPEs. If METHOD2FUNC is true, also make
> @@ -337,16 +363,19 @@ build_adjusted_function_type (tree orig_type, vec<tree> *new_param_types,
> if (skip_return)
> TREE_TYPE (new_type) = void_type_node;
> }
> - /* We only support one fn spec attribute on type. Be sure to remove it.
> - Once we support multiple attributes we will need to be able to unshare
> - the list. */
> if (args_modified && TYPE_ATTRIBUTES (new_type))
> {
> - gcc_checking_assert
> - (!TREE_CHAIN (TYPE_ATTRIBUTES (new_type))
> - && (is_attribute_p ("fn spec",
> - get_attribute_name (TYPE_ATTRIBUTES (new_type)))));
> + tree t = TYPE_ATTRIBUTES (new_type);
> + tree *last = &TYPE_ATTRIBUTES (new_type);
> TYPE_ATTRIBUTES (new_type) = NULL;
> + for (;t; t = TREE_CHAIN (t))
> + if (!drop_type_attribute_if_params_changed_p
> + (get_attribute_name (t)))
> + {
> + *last = copy_node (t);
> + TREE_CHAIN (*last) = NULL;
> + last = &TREE_CHAIN (*last);
> + }
> }
>
> return new_type;
> diff --git a/gcc/ipa-param-manipulation.h b/gcc/ipa-param-manipulation.h
> index 9440cbfc56c..5adf8a22356 100644
> --- a/gcc/ipa-param-manipulation.h
> +++ b/gcc/ipa-param-manipulation.h
> @@ -254,6 +254,7 @@ public:
> /* If true, make the function not return any value. */
> bool m_skip_return;
>
> + static bool type_attribute_allowed_p (tree);
> private:
> ipa_param_adjustments () {}
>
>
prev parent reply other threads:[~2021-11-15 15:38 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-11-13 14:44 Jan Hubicka
2021-11-15 15:38 ` Martin Sebor [this message]
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=969fcccd-e6cd-6e18-cfe2-57dab6103ff6@gmail.com \
--to=msebor@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=hubicka@kam.mff.cuni.cz \
--cc=msebor@gcc.gnu.org \
/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).