From: Richard Biener <richard.guenther@gmail.com>
To: Robin Dapp <rdapp.gcc@gmail.com>
Cc: gcc-patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] genopinit: Split init_all_optabs [PR113575]
Date: Fri, 26 Jan 2024 13:21:55 +0100 [thread overview]
Message-ID: <CAFiYyc0rgfMt8gpxf3anCd+R8PFDS4s53nRNoLYWnZ735ZOwnw@mail.gmail.com> (raw)
In-Reply-To: <77b98581-e3ac-4288-8ee0-f31416b95323@gmail.com>
On Fri, Jan 26, 2024 at 9:17 AM Robin Dapp <rdapp.gcc@gmail.com> wrote:
>
> Hi,
>
> init_all_optabs initializes > 10000 patterns for riscv targets. This
> leads to pathological situations in dataflow analysis (which can occur
> with many adjacent stores).
> To alleviate this this patch makes genopinit split the init_all_optabs
> function into several init_optabs_xx functions that each initialize 1000
> patterns.
>
> With this change insn-opinit.cc's compilation time is reduced from 4+
> minutes to 1:30 and memory consumption decreases from 1.2G to 630M.
>
> Bootstrapped and regtested on x86 and aarch64 (where we do split) and
> on power10 (where we don't). Regtested on riscv.
OK.
Thanks,
Richard.
> Regards
> Robin
>
> gcc/ChangeLog:
>
> PR other/113575
>
> * genopinit.cc (main): Split init_all_optabs into functions
> of 1000 patterns each.
> ---
> gcc/genopinit.cc | 43 ++++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 38 insertions(+), 5 deletions(-)
>
> diff --git a/gcc/genopinit.cc b/gcc/genopinit.cc
> index 88ccafa5b2c..d8682b2a9ad 100644
> --- a/gcc/genopinit.cc
> +++ b/gcc/genopinit.cc
> @@ -367,11 +367,44 @@ main (int argc, const char **argv)
> fprintf (s_file, " { %#08x, CODE_FOR_%s },\n", p->sort_num, p->name);
> fprintf (s_file, "};\n\n");
>
> - fprintf (s_file, "void\ninit_all_optabs (struct target_optabs *optabs)\n{\n");
> - fprintf (s_file, " bool *ena = optabs->pat_enable;\n");
> - for (i = 0; patterns.iterate (i, &p); ++i)
> - fprintf (s_file, " ena[%u] = HAVE_%s;\n", i, p->name);
> - fprintf (s_file, "}\n\n");
> + /* Some targets like riscv have a large number of patterns. In order to
> + prevent pathological situations in dataflow analysis split the init
> + function into separate ones that initialize 1000 patterns each. */
> +
> + const int patterns_per_function = 1000;
> +
> + if (patterns.length () > patterns_per_function)
> + {
> + unsigned num_init_functions
> + = patterns.length () / patterns_per_function + 1;
> + for (i = 0; i < num_init_functions; i++)
> + {
> + fprintf (s_file, "static void\ninit_optabs_%02d "
> + "(struct target_optabs *optabs)\n{\n", i);
> + fprintf (s_file, " bool *ena = optabs->pat_enable;\n");
> + unsigned start = i * patterns_per_function;
> + unsigned end = MIN (patterns.length (),
> + (i + 1) * patterns_per_function);
> + for (j = start; j < end; ++j)
> + fprintf (s_file, " ena[%u] = HAVE_%s;\n", j, patterns[j].name);
> + fprintf (s_file, "}\n\n");
> + }
> +
> + fprintf (s_file, "void\ninit_all_optabs "
> + "(struct target_optabs *optabs)\n{\n");
> + for (i = 0; i < num_init_functions; ++i)
> + fprintf (s_file, " init_optabs_%02d (optabs);\n", i);
> + fprintf (s_file, "}\n\n");
> + }
> + else
> + {
> + fprintf (s_file, "void\ninit_all_optabs "
> + "(struct target_optabs *optabs)\n{\n");
> + fprintf (s_file, " bool *ena = optabs->pat_enable;\n");
> + for (i = 0; patterns.iterate (i, &p); ++i)
> + fprintf (s_file, " ena[%u] = HAVE_%s;\n", i, p->name);
> + fprintf (s_file, "}\n\n");
> + }
>
> fprintf (s_file,
> "/* Returns TRUE if the target supports any of the partial vector\n"
> --
> 2.43.0
prev parent reply other threads:[~2024-01-26 12:22 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-01-26 8:16 Robin Dapp
2024-01-26 12:21 ` Richard Biener [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=CAFiYyc0rgfMt8gpxf3anCd+R8PFDS4s53nRNoLYWnZ735ZOwnw@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=rdapp.gcc@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 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).