public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Xi Ruoyao <xry111@xry111.site>
To: YunQiang Su <yunqiang.su@cipunited.com>, gcc-patches@gcc.gnu.org
Cc: macro@orcam.me.uk
Subject: Re: [PATCH] MIPS: improve -march=native arch detection
Date: Wed, 24 Aug 2022 13:56:50 +0800	[thread overview]
Message-ID: <c40416d5600d6b0c411a2bfb4e8f9a891df51832.camel@xry111.site> (raw)
In-Reply-To: <20220802111009.35536-1-yunqiang.su@cipunited.com>

On Tue, 2022-08-02 at 11:10 +0000, YunQiang Su wrote:

> If we cannot get info from options and cpuinfo, we try to get from:
>   1. getauxval(AT_BASE_PLATFORM), introduced since Linux 5.7
>   2. _MIPS_ARCH from host compiler.
> 
> This can fix the wrong loader usage on r5/r6 platform with
>  -march=native.

/* snip */

>    if (argc < 1)
> -    return NULL;
> +    goto fallback_cpu;

I don't think this should be changed, if argc < 1 it means the spec
(disambiguation: the thing printed by "gcc -dumpspecs", hard coded in
gnu-user.h) is wrong.  It cannot happen with the built-in spec, but if a
user specifies a bad custom spec with "-spec", we shouldn't be tricked.

>    arch = strcmp (argv[0], "arch") == 0;
>    if (!arch && strcmp (argv[0], "tune"))
> -    return NULL;
> +    goto fallback_cpu;

Likewise.

>    f = fopen ("/proc/cpuinfo", "r");
>    if (f == NULL)
> -    return NULL;
> +    goto fallback_cpu;

OK.

> +fallback_cpu:
> +/*FIXME: how about other OSes, like FreeBSD? */

https://reviews.freebsd.org/D12743 added elf_aux_info as a counterpart
of getauxinfo, but it looks like FreeBSD does not have AT_BASE_PLATFORM.

> +#ifdef __linux__
> +  /*Note: getauxval may return NULL as:
> +   * AT_BASE_PLATFORM is supported since Linux 5.7
> +   * Or from older version of qemu-user
> +   * */
> +  if (cpu == NULL)
> +    cpu = (const char *) getauxval (AT_BASE_PLATFORM);

getauxval is added in Glibc-2.16 so it will fail to build on hosts with
old glibc or other libc implementation.  Check if getauxval and
AT_BASE_PLATFORM are available (in gcc/configure.ac) instead of an
inaccurate "#ifdef __linux__".

> +#endif
> +
>    if (cpu == NULL)
> +#if defined (_MIPS_ARCH)
> +    cpu = _MIPS_ARCH;
> +#else

Ok.

>      return NULL;
> +#endif
>  
>    return concat ("-m", argv[0], "=", cpu, NULL);
>  }

  parent reply	other threads:[~2022-08-24  5:57 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-02 11:10 YunQiang Su
2022-08-14 10:36 ` YunQiang Su
2022-08-24  4:14 ` YunQiang Su
2022-08-24  5:56 ` Xi Ruoyao [this message]
2022-09-16 12:23 ` [PATCH v2] " YunQiang Su

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=c40416d5600d6b0c411a2bfb4e8f9a891df51832.camel@xry111.site \
    --to=xry111@xry111.site \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=macro@orcam.me.uk \
    --cc=yunqiang.su@cipunited.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).