From: Martin Jansa <martin.jansa@gmail.com>
To: Phil Blundell <pb@pbcl.net>
Cc: openembedded-core@lists.openembedded.org
Subject: Re: [PATCHv2 11/16] arch-armv7ve: add tune include for armv7ve and use it from cortexa7 and cortexa15
Date: Wed, 18 Nov 2015 18:20:53 +0100 [thread overview]
Message-ID: <20151118172053.GF2213@jama> (raw)
In-Reply-To: <1447866323.1472.58.camel@pbcl.net>
[-- Attachment #1: Type: text/plain, Size: 2401 bytes --]
On Wed, Nov 18, 2015 at 05:05:23PM +0000, Phil Blundell wrote:
> On Wed, 2015-11-18 at 17:51 +0100, Martin Jansa wrote:
> > On Wed, Nov 18, 2015 at 04:38:34PM +0000, Phil Blundell wrote:
> > > On Tue, 2015-11-17 at 11:58 +0100, Martin Jansa wrote:
> > > > * be aware that this -march value is available only in gcc-4.9
> > > > and
> > > > newer:
> > > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57907
> > > >
> > >
> > > Can you clarify why this is an improvement? The previous patch
> > > already
> > > changed cortex-a15.inc to use -mcpu=cortexa15, which effectively
> > > implies -march=armv7ve already (and works on all gcc versions) so
> > > I'm
> > > not entirely sure what this second patch really buys you.
> >
> > Because -mcpu=cortexa15 and -march=armv7a conflict each other, so we
> > either have to stop putting -march in default CCARGS or at least set
> > compatible one like this patch does.
>
> Ah right, got it. I think it would be better to just stop passing
> -march= in cases where we are using -mcpu=. It seems a bit silly to
> create another huge maze of tune variables which all boil down to
> essentially a no-op.
The problem with this is that currently all the tunes are at least a bit
hierarchical, in this case
tune-cortexa15 was including
openembedded-core/meta/conf/machine/include/arm/arch-armv7a.inc
Which adds -march for anything with armv7a in TUNE_FEATURES:
TUNE_CCARGS .= "${@bb.utils.contains("TUNE_FEATURES", "armv7a", "-march=armv7-a", "", d)}"
and as bonus adds MACHINEOVERRIDES:
MACHINEOVERRIDES =. "${@bb.utils.contains("TUNE_FEATURES", "armv7a",
"armv7a:", "" ,d)}"
Then cortexa15 just uses tune features from "upper" tune-armv7a + just cortexa15, e.g.:
TUNE_FEATURES_tune-cortexa15 = "${TUNE_FEATURES_tune-armv7a} cortexa15"
So to get rid of -march we'll have to either
1) check in arch-armv7a.inc if TUNE_FEATURES also contains some TUNE_FEATURE which will append -mcpu
2) don't include armv7a TUNE_FEATURE in TUNE_FEATURES_tune-cortexa15t,
but the MACHINEOVERRIDES is still useful in many cases
3) use TUNE_CCARGS_remove to remove -march when adding -mcpu
4) rework whole tune files even more (probably by introducing another
variable which will hold just one of -march,-mcpu,-mtune and gets
appended to TUNE_CCARGS).
--
Martin 'JaMa' Jansa jabber: Martin.Jansa@gmail.com
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]
next prev parent reply other threads:[~2015-11-18 17:20 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-11-17 10:57 [PATCHv2 00/16] Various fixes from LGE Martin Jansa
2015-11-17 10:57 ` [PATCHv2 01/16] icu: force arm mode Martin Jansa
2015-11-17 10:57 ` [PATCHv2 02/16] guile, mailx, gcc, opensp, gstreamer1.0-libav, libunwind: disable thumb where it fails for qemuarm Martin Jansa
2015-11-17 10:57 ` [PATCHv2 03/16] directfb: add fPIC to CFLAGS Martin Jansa
2015-11-17 10:57 ` [PATCHv2 04/16] libunwind: fix build for qemuarm Martin Jansa
2015-11-17 10:57 ` [PATCHv2 05/16] gstreamer1.0-plugins-base: add PACKAGECONFIG for libvisual Martin Jansa
2015-11-17 10:57 ` [PATCHv2 06/16] gstreamer1.0-plugins-bad: fix dependencies for uvch264 PACKAGECONFIG Martin Jansa
2015-11-17 16:47 ` Burton, Ross
2015-11-17 10:57 ` [PATCHv2 07/16] gstreamer1.0-plugins-good: fix PACKAGECONFIG for gudev and add one for v4l2 and libv4l2 Martin Jansa
2015-11-17 14:08 ` Burton, Ross
2015-12-09 8:44 ` Martin Jansa
2015-11-17 10:58 ` [PATCHv2 08/16] gst-plugins-bad: improve FILES variables Martin Jansa
2015-11-17 10:58 ` [PATCHv2 09/16] cortexa{7,15,17}: add VFPv4 tunes Martin Jansa
2015-11-18 16:36 ` Phil Blundell
2015-11-18 16:49 ` Martin Jansa
2015-11-17 10:58 ` [PATCHv2 10/16] tune-*: use mcpu instead of mtune for ARM tunes Martin Jansa
2015-11-17 10:58 ` [PATCHv2 11/16] arch-armv7ve: add tune include for armv7ve and use it from cortexa7 and cortexa15 Martin Jansa
2015-11-18 16:38 ` Phil Blundell
2015-11-18 16:51 ` Martin Jansa
2015-11-18 17:05 ` Phil Blundell
2015-11-18 17:20 ` Martin Jansa [this message]
2015-11-18 17:31 ` Phil Blundell
2015-11-17 10:58 ` [PATCHv2 12/16] feature-arm-vfp.inc: respect vfpv4 when setting -mfloat-abi and ARMPKGSFX_EABI Martin Jansa
2015-11-18 16:41 ` Phil Blundell
2015-11-18 18:37 ` [PATCHv3 1/6] meta/conf/machine: use ' inside quoted values Martin Jansa
2015-11-18 18:37 ` [PATCHv3 2/6] cortexa{7,15,17}: add VFPv4 tunes Martin Jansa
2015-11-18 22:31 ` Phil Blundell
2015-11-18 18:37 ` [PATCHv3 3/6] tune-*: use mcpu instead of mtune for ARM tunes Martin Jansa
2015-11-18 18:37 ` [PATCHv3 4/6] arch-armv7ve: add tune include for armv7ve and use it from cortexa7 and cortexa15 Martin Jansa
2015-11-18 18:37 ` [PATCHv3 5/6] feature-arm-vfp.inc: respect vfpv4 when setting -mfloat-abi and ARMPKGSFX_EABI Martin Jansa
2015-12-07 22:35 ` Martin Jansa
2015-12-08 10:34 ` Richard Purdie
2015-12-08 14:03 ` Martin Jansa
2015-11-18 18:37 ` [PATCHv3 6/6] feature-arm-vfp.inc: add -mfpu=vfpv4 when vfpv4 is in TUNE_FEATURES Martin Jansa
2015-11-17 10:58 ` [PATCHv2 13/16] " Martin Jansa
2015-11-17 10:58 ` [PATCHv2 14/16] rootfs_*.bbclass: don't add BUILDNAME to do_rootfs vardepsexclude Martin Jansa
2015-11-17 10:58 ` [PATCHv2 15/16] libsecret: add dependency on intltool-native Martin Jansa
2015-11-17 10:58 ` [PATCHv2 16/16] libsdl2: require GLES when building Wayland support Martin Jansa
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=20151118172053.GF2213@jama \
--to=martin.jansa@gmail.com \
--cc=openembedded-core@lists.openembedded.org \
--cc=pb@pbcl.net \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.