From mboxrd@z Thu Jan  1 00:00:00 1970
From: Marek Vasut <marex@denx.de>
Date: Thu, 5 Jul 2012 21:58:04 +0200
Subject: [U-Boot] [PATCH] ehci-hcd.c, musb_core,
	usb.h: Add USB_DMA_MINALIGN define for cache alignment
In-Reply-To: <CAA3CPjUsKCrGSctvzHQdxsGbQHtwnaGO48c0egOYnZ=kkT=3=A@mail.gmail.com>
References: <1341407039-6018-1-git-send-email-ilya.yanok@cogentembedded.com>
	<201207042224.58803.marex@denx.de>
	<CAA3CPjUsKCrGSctvzHQdxsGbQHtwnaGO48c0egOYnZ=kkT=3=A@mail.gmail.com>
Message-ID: <201207052158.04460.marex@denx.de>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

Dear Ilya Yanok,

> Hi Marek,
> 
> On Thu, Jul 5, 2012 at 12:24 AM, Marek Vasut <marex@denx.de> wrote:
> > > -static struct QH qh_list __attribute__((aligned(32)));
> > > +static char __qh_list[ALIGN(sizeof(struct QH), USB_DMA_MINALIGN)]
> > > +                     __attribute__((aligned(USB_DMA_MINALIGN)));
> > > +static struct QH *qh_list = (struct QH *)__qh_list;
> > 
> > Maybe we should create DEFINE_ALIGNED_VARIABLE as a common.h macro?
> 
> Yep. I even thought about this but decided not to do... can't recall why.
> Now I think it's really a good idea.

Like what's in common.h -- ALLOC_CACHE_ALIGN_BUFFER ?

[..]

> Hm.. Maybe. Ideas? ;) Actually I also thought about moving all this stuff
> to a single proper aligned buffer and do flush/invalidate for a whole
> buffer at once. It can save us some space... but it's BSS anyway... Don't
> know if it's worth it...

But if you copy stuff back and forth, it'll cause performance hit.

> Regards, Ilya.

Best regards,
Marek Vasut