Re: [PATCH v2 1/6] ASoC: Intel: Add helper to poll register for DSP status

From: Mark Brown <broonie@kernel.org>
To: Vinod Koul <vinod.koul@intel.com>
Cc: alsa-devel@alsa-project.org, tiwai@suse.de,
	liam.r.girdwood@linux.intel.com, patches.audio@intel.com, "Kp,
	Jeeja" <jeeja.kp@intel.com>,
	"Subhransu S. Prusty" <subhransu.s.prusty@intel.com>
Subject: Re: [PATCH v2 1/6] ASoC: Intel: Add helper to poll register for DSP status
Date: Thu, 9 Jul 2015 11:48:55 +0100	[thread overview]
Message-ID: <20150709104855.GJ11162@sirena.org.uk> (raw)
In-Reply-To: <20150709042741.GF836@localhost>

[-- Attachment #1.1: Type: text/plain, Size: 749 bytes --]

On Thu, Jul 09, 2015 at 09:57:41AM +0530, Vinod Koul wrote:
> On Wed, Jul 08, 2015 at 07:36:21PM +0100, Mark Brown wrote:

> > > +	for (time = 0; time < timeout; time++) {
> > > +		if ((sst_dsp_shim_read_unlocked(ctx, offset) & mask) == expected_value)
> > > +			break;
> > > +
> > > +		mdelay(1);
> > > +	}

> > mdelay() not msleep()?  If we're waiting for multiple miliseconds that
> > could be lots of busy waiting.

> Usually this should get reflected in 1st iteration as the register update
> would get updated farrily quickly. msleep will add up lots of latency to
> this.

A common approach for that is to do a busy wait for say the first
milisecond (perhaps polling more often too) and then fall back to
something sleepy if things are slow.

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]