From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jonathan Cameron Subject: Re: [PATCH 3/3] iio: adc: ti-ads7950: use SPI_CS_WORD to reduce CPU usage Date: Sat, 22 Sep 2018 10:12:04 +0100 Message-ID: <20180922101204.07ce82d5@archlinux> References: <20180918170850.2749-1-david@lechnology.com> <20180918170850.2749-4-david@lechnology.com> <7a70711c-5d38-5ba6-ec90-b2506dc0727d@lechnology.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: linux-spi@vger.kernel.org, linux-iio@vger.kernel.org, Hartmut Knaack , Lars-Peter Clausen , Peter Meerwald-Stadler , Mark Brown , Geert Uytterhoeven , linux-kernel@vger.kernel.org To: David Lechner Return-path: In-Reply-To: <7a70711c-5d38-5ba6-ec90-b2506dc0727d@lechnology.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-spi.vger.kernel.org On Tue, 18 Sep 2018 12:14:50 -0500 David Lechner wrote: > On 09/18/2018 12:08 PM, David Lechner wrote: > > This changes how the SPI message for the triggered buffer is setup in > > the TI ADS7950 A/DC driver. By using the SPI_CS_WORD flag, we can read > > multiple samples in a single SPI transfer. If the SPI controller > > supports DMA transfers, we can see a significant reduction in CPU usage. > > > > For example, on an ARM9 system running at 456MHz reading just 4 channels > > at 100Hz: before this change, top shows the CPU usage of the IRQ thread > > of this driver to be ~7.7%. After this change, the CPU usage drops to > > ~3.8%. > > > > The use of big-endian for the raw data was cargo culted from another > > driver when this driver was originally written. It used an SPI word size > > of 8 bits and big-endian byte ordering to effectively emulate 16 bit > > words. > > > > Now, in order to inject a CS toggle between each word, we need to use > > the correct word size, otherwise we would get a CS toggle half way > > through each word 16-bit. The SPI subsystem uses CPU byte ordering for > > multi-byte words. So, the data we get back from the SPI is going to be > > CPU endian now no matter what. Converting that to big endian will just > > add overhead on little endian systems so we opt to change the raw data > > format from big endian to CPU endian. > > > > There is a small risk that this could break some lazy userspace programs > > that use the raw data without checking the data format. We can address > > this if/when it actually comes up. > > > > Signed-off-by: David Lechner > > --- > > And I just realized I forgot to pick up... > > > Reviewed-by: Jonathan Cameron > I've applied this on top of the original series on the basis I doubt anyone will be testing this particular driver in such a way as to hit the changes you made in the last two patches + I don't think Mark has tagged those for me to easily pick up. If it's a problem I can apply them to the IIO tree as well and git will sort it out come merge window time. However, that's inelegant so I won't do it unless needed. Applied to the togreg branch of iio.git and pushed out as testing for the autobuilders to play with it Thanks, Jonathan