From: Jonathan Cameron <jic23@kernel.org>
To: David Lechner <dlechner@baylibre.com>
Cc: "Michael Hennerich" <Michael.Hennerich@analog.com>,
"Nuno Sá" <nuno.sa@analog.com>,
linux-iio@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] iio: adc: ad7944: use spi_optimize_message()
Date: Sat, 30 Mar 2024 16:07:13 +0000 [thread overview]
Message-ID: <20240330160713.7b31c201@jic23-huawei> (raw)
In-Reply-To: <20240328-ad7944-spi-optimize-message-v2-1-a142b2576379@baylibre.com>
On Thu, 28 Mar 2024 16:16:59 -0500
David Lechner <dlechner@baylibre.com> wrote:
> This modifies the ad7944 driver to use spi_optimize_message() to reduce
> CPU usage and increase the max sample rate by avoiding repeating
> validation of the spi message on each transfer.
>
> Signed-off-by: David Lechner <dlechner@baylibre.com>
Applied to the togreg branch of iio.git and pushed out as testing.
Thanks,
Jonathan
> ---
> Changes in v2:
> - Fix wrong array index in ad7944_3wire_cs_mode_init_msg()
> - Fixed wrong number of xfers in ad7944_4wire_mode_init_msg()
> - Link to v1: https://lore.kernel.org/r/20240325-ad7944-spi-optimize-message-v1-1-cded69b9e27f@baylibre.com
> ---
> drivers/iio/adc/ad7944.c | 177 +++++++++++++++++++++++++++--------------------
> 1 file changed, 103 insertions(+), 74 deletions(-)
>
> diff --git a/drivers/iio/adc/ad7944.c b/drivers/iio/adc/ad7944.c
> index 261a3f645fd8..e09ab0c842d3 100644
> --- a/drivers/iio/adc/ad7944.c
> +++ b/drivers/iio/adc/ad7944.c
> @@ -51,6 +51,8 @@ static const char * const ad7944_spi_modes[] = {
> struct ad7944_adc {
> struct spi_device *spi;
> enum ad7944_spi_mode spi_mode;
> + struct spi_transfer xfers[3];
> + struct spi_message msg;
> /* Chip-specific timing specifications. */
> const struct ad7944_timing_spec *timing_spec;
> /* GPIO connected to CNV pin. */
> @@ -130,6 +132,88 @@ AD7944_DEFINE_CHIP_INFO(ad7985, ad7944, 16, 0);
> /* fully differential */
> AD7944_DEFINE_CHIP_INFO(ad7986, ad7986, 18, 1);
>
> +static void ad7944_unoptimize_msg(void *msg)
> +{
> + spi_unoptimize_message(msg);
> +}
> +
> +static int ad7944_3wire_cs_mode_init_msg(struct device *dev, struct ad7944_adc *adc,
> + const struct iio_chan_spec *chan)
> +{
> + unsigned int t_conv_ns = adc->always_turbo ? adc->timing_spec->turbo_conv_ns
> + : adc->timing_spec->conv_ns;
> + struct spi_transfer *xfers = adc->xfers;
> + int ret;
> +
> + /*
> + * NB: can get better performance from some SPI controllers if we use
> + * the same bits_per_word in every transfer.
> + */
> + xfers[0].bits_per_word = chan->scan_type.realbits;
> + /*
> + * CS is tied to CNV and we need a low to high transition to start the
> + * conversion, so place CNV low for t_QUIET to prepare for this.
> + */
> + xfers[0].delay.value = T_QUIET_NS;
> + xfers[0].delay.unit = SPI_DELAY_UNIT_NSECS;
> +
> + /*
> + * CS has to be high for full conversion time to avoid triggering the
> + * busy indication.
> + */
> + xfers[1].cs_off = 1;
> + xfers[1].delay.value = t_conv_ns;
> + xfers[1].delay.unit = SPI_DELAY_UNIT_NSECS;
> + xfers[1].bits_per_word = chan->scan_type.realbits;
> +
> + /* Then we can read the data during the acquisition phase */
> + xfers[2].rx_buf = &adc->sample.raw;
> + xfers[2].len = BITS_TO_BYTES(chan->scan_type.storagebits);
> + xfers[2].bits_per_word = chan->scan_type.realbits;
> +
> + spi_message_init_with_transfers(&adc->msg, xfers, 3);
> +
> + ret = spi_optimize_message(adc->spi, &adc->msg);
> + if (ret)
> + return ret;
> +
> + return devm_add_action_or_reset(dev, ad7944_unoptimize_msg, &adc->msg);
> +}
> +
> +static int ad7944_4wire_mode_init_msg(struct device *dev, struct ad7944_adc *adc,
> + const struct iio_chan_spec *chan)
> +{
> + unsigned int t_conv_ns = adc->always_turbo ? adc->timing_spec->turbo_conv_ns
> + : adc->timing_spec->conv_ns;
> + struct spi_transfer *xfers = adc->xfers;
> + int ret;
> +
> + /*
> + * NB: can get better performance from some SPI controllers if we use
> + * the same bits_per_word in every transfer.
> + */
> + xfers[0].bits_per_word = chan->scan_type.realbits;
> + /*
> + * CS has to be high for full conversion time to avoid triggering the
> + * busy indication.
> + */
> + xfers[0].cs_off = 1;
> + xfers[0].delay.value = t_conv_ns;
> + xfers[0].delay.unit = SPI_DELAY_UNIT_NSECS;
> +
> + xfers[1].rx_buf = &adc->sample.raw;
> + xfers[1].len = BITS_TO_BYTES(chan->scan_type.storagebits);
> + xfers[1].bits_per_word = chan->scan_type.realbits;
> +
> + spi_message_init_with_transfers(&adc->msg, xfers, 2);
> +
> + ret = spi_optimize_message(adc->spi, &adc->msg);
> + if (ret)
> + return ret;
> +
> + return devm_add_action_or_reset(dev, ad7944_unoptimize_msg, &adc->msg);
> +}
> +
> /*
> * ad7944_3wire_cs_mode_conversion - Perform a 3-wire CS mode conversion and
> * acquisition
> @@ -145,48 +229,7 @@ AD7944_DEFINE_CHIP_INFO(ad7986, ad7986, 18, 1);
> static int ad7944_3wire_cs_mode_conversion(struct ad7944_adc *adc,
> const struct iio_chan_spec *chan)
> {
> - unsigned int t_conv_ns = adc->always_turbo ? adc->timing_spec->turbo_conv_ns
> - : adc->timing_spec->conv_ns;
> - struct spi_transfer xfers[] = {
> - {
> - /*
> - * NB: can get better performance from some SPI
> - * controllers if we use the same bits_per_word
> - * in every transfer.
> - */
> - .bits_per_word = chan->scan_type.realbits,
> - /*
> - * CS is tied to CNV and we need a low to high
> - * transition to start the conversion, so place CNV
> - * low for t_QUIET to prepare for this.
> - */
> - .delay = {
> - .value = T_QUIET_NS,
> - .unit = SPI_DELAY_UNIT_NSECS,
> - },
> -
> - },
> - {
> - .bits_per_word = chan->scan_type.realbits,
> - /*
> - * CS has to be high for full conversion time to avoid
> - * triggering the busy indication.
> - */
> - .cs_off = 1,
> - .delay = {
> - .value = t_conv_ns,
> - .unit = SPI_DELAY_UNIT_NSECS,
> - },
> - },
> - {
> - /* Then we can read the data during the acquisition phase */
> - .rx_buf = &adc->sample.raw,
> - .len = BITS_TO_BYTES(chan->scan_type.storagebits),
> - .bits_per_word = chan->scan_type.realbits,
> - },
> - };
> -
> - return spi_sync_transfer(adc->spi, xfers, ARRAY_SIZE(xfers));
> + return spi_sync(adc->spi, &adc->msg);
> }
>
> /*
> @@ -200,33 +243,6 @@ static int ad7944_3wire_cs_mode_conversion(struct ad7944_adc *adc,
> static int ad7944_4wire_mode_conversion(struct ad7944_adc *adc,
> const struct iio_chan_spec *chan)
> {
> - unsigned int t_conv_ns = adc->always_turbo ? adc->timing_spec->turbo_conv_ns
> - : adc->timing_spec->conv_ns;
> - struct spi_transfer xfers[] = {
> - {
> - /*
> - * NB: can get better performance from some SPI
> - * controllers if we use the same bits_per_word
> - * in every transfer.
> - */
> - .bits_per_word = chan->scan_type.realbits,
> - /*
> - * CS has to be high for full conversion time to avoid
> - * triggering the busy indication.
> - */
> - .cs_off = 1,
> - .delay = {
> - .value = t_conv_ns,
> - .unit = SPI_DELAY_UNIT_NSECS,
> - },
> -
> - },
> - {
> - .rx_buf = &adc->sample.raw,
> - .len = BITS_TO_BYTES(chan->scan_type.storagebits),
> - .bits_per_word = chan->scan_type.realbits,
> - },
> - };
> int ret;
>
> /*
> @@ -234,7 +250,7 @@ static int ad7944_4wire_mode_conversion(struct ad7944_adc *adc,
> * and acquisition process.
> */
> gpiod_set_value_cansleep(adc->cnv, 1);
> - ret = spi_sync_transfer(adc->spi, xfers, ARRAY_SIZE(xfers));
> + ret = spi_sync(adc->spi, &adc->msg);
> gpiod_set_value_cansleep(adc->cnv, 0);
>
> return ret;
> @@ -395,10 +411,6 @@ static int ad7944_probe(struct spi_device *spi)
> adc->spi_mode = ret;
> }
>
> - if (adc->spi_mode == AD7944_SPI_MODE_CHAIN)
> - return dev_err_probe(dev, -EINVAL,
> - "chain mode is not implemented\n");
> -
> /*
> * Some chips use unusual word sizes, so check now instead of waiting
> * for the first xfer.
> @@ -491,6 +503,23 @@ static int ad7944_probe(struct spi_device *spi)
> return dev_err_probe(dev, -EINVAL,
> "cannot have both chain mode and always turbo\n");
>
> + switch (adc->spi_mode) {
> + case AD7944_SPI_MODE_DEFAULT:
> + ret = ad7944_4wire_mode_init_msg(dev, adc, &chip_info->channels[0]);
> + if (ret)
> + return ret;
> +
> + break;
> + case AD7944_SPI_MODE_SINGLE:
> + ret = ad7944_3wire_cs_mode_init_msg(dev, adc, &chip_info->channels[0]);
> + if (ret)
> + return ret;
> +
> + break;
> + case AD7944_SPI_MODE_CHAIN:
> + return dev_err_probe(dev, -EINVAL, "chain mode is not implemented\n");
> + }
> +
> indio_dev->name = chip_info->name;
> indio_dev->modes = INDIO_DIRECT_MODE;
> indio_dev->info = &ad7944_iio_info;
>
> ---
> base-commit: 526f7f17b651c78ead26fea7cea20948c00e47a5
> change-id: 20240325-ad7944-spi-optimize-message-82debaa2a5a7
>
> Best regards,
prev parent reply other threads:[~2024-03-30 16:07 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-28 21:16 [PATCH v2] iio: adc: ad7944: use spi_optimize_message() David Lechner
2024-03-30 16:07 ` Jonathan Cameron [this message]
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=20240330160713.7b31c201@jic23-huawei \
--to=jic23@kernel.org \
--cc=Michael.Hennerich@analog.com \
--cc=dlechner@baylibre.com \
--cc=linux-iio@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=nuno.sa@analog.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).