From: Thomas Schwinge <thomas@codesourcery.com>
To: Richard Biener <richard.guenther@gmail.com>,
Bernd Schmidt <bernds@codesourcery.com>
Cc: Jakub Jelinek <jakub@redhat.com>, Jeff Law <law@redhat.com>,
GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: The nvptx port [1/11+] indirect jumps
Date: Wed, 22 Oct 2014 08:37:00 -0000 [thread overview]
Message-ID: <87d29kjyem.fsf@kepler.schwinge.homeip.net> (raw)
In-Reply-To: <CAFiYyc0eE1RvYfANE-z+CDcznxtHOs1X-1TzbWZHGVQKSYPSBg@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 1584 bytes --]
Hi!
On Wed, 22 Oct 2014 10:18:49 +0200, Richard Biener <richard.guenther@gmail.com> wrote:
> On Tue, Oct 21, 2014 at 11:32 PM, Bernd Schmidt <bernds@codesourcery.com> wrote:
> > On 10/21/2014 11:30 PM, Jakub Jelinek wrote:
> >>
> >> At least for OpenMP, the best would be if the #pragma omp target regions
> >> and/or #pragma omp declare target functions contain anything a particular
> >> offloading accelerator can't handle, instead of failing the whole
> >> compilation perhaps just emit some at least by default non-fatal warning
> >> and not emit anything for the particular offloading target, which would
> >> mean
> >> either host fallback, or, if some other offloading target succeeded, just
> >> that target.
> >
> >
> > I guess a test could be added to mkoffload if gcc were to return a different
> > value for a sorry vs. any other compilation failure. The tool could then
> > choose not to produce offloading support for that target.
>
> But that would be for the whole file instead of for the specific region?
I'm not sure that's what you're suggesting, but at least on non-shared
memory offloading devices, you can't switch arbitrarily between
offloading device(s) and host-fallback, for you have to do data
management between the non-shared memories.
> So maybe we should produce one LTO offload object for each offload
> function and make the symbols they are supposed to provide weak
> so a fail doesn't end up failing to link the main program?
>
> Looks like this gets somewhat awkward with the LTO setup.
Grüße,
Thomas
[-- Attachment #2: Type: application/pgp-signature, Size: 472 bytes --]
next prev parent reply other threads:[~2014-10-22 8:34 UTC|newest]
Thread overview: 83+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-10-20 14:19 The nvptx port [0/11+] Bernd Schmidt
2014-10-20 14:21 ` The nvptx port [1/11+] indirect jumps Bernd Schmidt
2014-10-21 18:29 ` Jeff Law
2014-10-21 21:03 ` Bernd Schmidt
2014-10-21 21:30 ` Jakub Jelinek
2014-10-21 21:37 ` Bernd Schmidt
2014-10-22 8:21 ` Richard Biener
2014-10-22 8:34 ` Jakub Jelinek
2014-10-22 8:37 ` Thomas Schwinge [this message]
2014-10-22 10:03 ` Richard Biener
2014-10-22 10:32 ` Jakub Jelinek
2014-11-04 15:35 ` Bernd Schmidt
2014-11-04 15:43 ` Richard Henderson
2014-10-20 14:22 ` The nvptx port [2/11+] No register allocation Bernd Schmidt
2014-10-20 14:24 ` The nvptx port [3/11+] Struct returns Bernd Schmidt
2014-10-21 18:41 ` Jeff Law
2014-10-20 14:24 ` The nvptx port [2/11+] No register allocation Bernd Schmidt
2014-10-21 18:36 ` Jeff Law
2014-10-20 14:27 ` The nvptx port [4/11+] Post-RA pipeline Bernd Schmidt
2014-10-21 18:42 ` Jeff Law
2024-06-28 15:07 ` Document 'pass_postreload' vs. 'pass_late_compilation' (was: The nvptx port [4/11+] Post-RA pipeline) Thomas Schwinge
2014-10-20 14:27 ` The nvptx port [5/11+] Variable declarations Bernd Schmidt
2014-10-21 18:44 ` Jeff Law
2014-10-20 14:31 ` The nvptx port [6/11+] Pseudo call args Bernd Schmidt
2014-10-21 18:56 ` Jeff Law
2014-10-20 14:32 ` The nvptx port [8/11+] Write undefined decls Bernd Schmidt
2014-10-21 22:07 ` Jeff Law
2014-10-21 22:30 ` Bernd Schmidt
2014-10-22 18:23 ` Jeff Law
2014-11-05 12:05 ` Bernd Schmidt
2014-11-05 20:05 ` Jeff Law
2014-10-20 14:32 ` The nvptx port [7/11+] Inform the port about call arguments Bernd Schmidt
2014-10-21 21:25 ` Jeff Law
2014-10-21 21:33 ` Bernd Schmidt
2014-10-21 21:55 ` Jeff Law
2014-10-21 22:16 ` Bernd Schmidt
2014-10-22 18:23 ` Jeff Law
2014-10-28 14:57 ` Bernd Schmidt
2014-10-29 23:42 ` Jeff Law
2014-10-20 14:35 ` The nvptx port [9/11+] Epilogues Bernd Schmidt
2014-10-21 22:08 ` Jeff Law
2014-10-20 14:50 ` The nvptx port [10/11+] Target files Bernd Schmidt
2014-10-22 18:12 ` Jeff Law
2014-10-28 15:10 ` Bernd Schmidt
2014-10-29 23:51 ` Jeff Law
2014-10-30 2:53 ` Bernd Schmidt
2014-10-30 3:09 ` Jeff Law
2014-11-10 16:33 ` Bernd Schmidt
2014-11-10 20:06 ` Jakub Jelinek
2014-11-10 20:37 ` H.J. Lu
2014-11-10 20:40 ` H.J. Lu
2014-11-10 20:42 ` Mike Stump
2014-12-12 20:18 ` Thomas Schwinge
2014-12-23 18:51 ` nvptx-tools and nvptx-newlib (was: The nvptx port [10/11+] Target files) Thomas Schwinge
2015-02-02 15:33 ` Thomas Schwinge
2015-02-04 9:43 ` Jakub Jelinek
2015-02-18 8:50 ` Thomas Schwinge
2015-02-18 9:03 ` Jakub Jelinek
2015-07-08 15:03 ` [nvptx offloading] Only 64-bit configurations are currently supported (was: nvptx-tools and nvptx-newlib) Thomas Schwinge
2015-07-14 20:10 ` [nvptx offloading] Only 64-bit configurations are currently supported Thomas Schwinge
2015-07-14 20:25 ` Richard Biener
2021-01-14 18:18 ` [nvptx libgomp plugin] Build only in supported configurations (was: [nvptx offloading] Only 64-bit configurations are currently supported) Thomas Schwinge
2021-03-04 8:52 ` [committed] libgomp: Use sizeof(void*) based checks instead of looking through $CC $CFLAGS for -m32/-mx32 Jakub Jelinek
2021-03-22 11:24 ` Thomas Schwinge
2014-11-04 16:48 ` The nvptx port [10/11+] Target files Richard Henderson
2014-11-04 16:55 ` Bernd Schmidt
2014-11-05 13:07 ` Bernd Schmidt
2014-10-20 14:58 ` The nvptx port [11/11] More tools Bernd Schmidt
2014-10-21 0:16 ` Joseph S. Myers
2014-10-22 20:40 ` Jeff Law
2014-10-22 21:16 ` Bernd Schmidt
2014-10-24 19:52 ` Jeff Law
2014-10-31 21:04 ` Jeff Law
[not found] ` <54542050.6010908@codesourcery.com>
2014-11-03 21:49 ` Jeff Law
2014-10-21 8:23 ` The nvptx port [0/11+] Richard Biener
2014-10-21 10:57 ` Bernd Schmidt
2014-10-21 11:27 ` Richard Biener
2014-10-21 9:17 ` Jakub Jelinek
2014-10-21 11:19 ` Bernd Schmidt
2014-11-12 12:36 ` Richard Biener
2014-11-12 21:39 ` Jeff Law
2015-02-18 7:48 ` nvptx-none: Define empty GOMP_SELF_SPECS (was: The nvptx port [0/11+]) Thomas Schwinge
2015-02-18 8:01 ` The nvptx port [0/11+] Thomas Schwinge
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=87d29kjyem.fsf@kepler.schwinge.homeip.net \
--to=thomas@codesourcery.com \
--cc=bernds@codesourcery.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=law@redhat.com \
--cc=richard.guenther@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).