* [PATCH] spi: spi-ti-qspi: switch to polling mode for better r/w performance
@ 2015-10-13 10:21 Vignesh R
[not found] ` <1444731665-11359-1-git-send-email-vigneshr-l0cyMroinI0@public.gmane.org>
0 siblings, 1 reply; 2+ messages in thread
From: Vignesh R @ 2015-10-13 10:21 UTC (permalink / raw)
To: Mark Brown
Cc: linux-spi-u79uwXL29TY76Z2rM5mHXA,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
linux-omap-u79uwXL29TY76Z2rM5mHXA, Vignesh R
Currently word completion interrupt is fired for transfer of every
word(8bit to 128bit in size). This adds a lot of overhead, and decreases
r/w throughput. It hardly takes 3us(@48MHz) for 128bit r/w to complete,
hence its better to poll on word complete bit to be set in
QSPI_SPI_STATUS_REG instead of using interrupts.
This increases the throughput by 30% in both read and write case.
So, switch to polling mode instead of interrupts to determine completion
of word transfer.
Signed-off-by: Vignesh R <vigneshr-l0cyMroinI0@public.gmane.org>
---
Tested on DRA74 Rev G EVM.
drivers/spi/spi-ti-qspi.c | 74 +++++++++++++----------------------------------
1 file changed, 20 insertions(+), 54 deletions(-)
diff --git a/drivers/spi/spi-ti-qspi.c b/drivers/spi/spi-ti-qspi.c
index 81b84858cfee..69c1a95b0615 100644
--- a/drivers/spi/spi-ti-qspi.c
+++ b/drivers/spi/spi-ti-qspi.c
@@ -39,8 +39,6 @@ struct ti_qspi_regs {
};
struct ti_qspi {
- struct completion transfer_complete;
-
/* list synchronization */
struct mutex list_lock;
@@ -62,10 +60,6 @@ struct ti_qspi {
#define QSPI_PID (0x0)
#define QSPI_SYSCONFIG (0x10)
-#define QSPI_INTR_STATUS_RAW_SET (0x20)
-#define QSPI_INTR_STATUS_ENABLED_CLEAR (0x24)
-#define QSPI_INTR_ENABLE_SET_REG (0x28)
-#define QSPI_INTR_ENABLE_CLEAR_REG (0x2c)
#define QSPI_SPI_CLOCK_CNTRL_REG (0x40)
#define QSPI_SPI_DC_REG (0x44)
#define QSPI_SPI_CMD_REG (0x48)
@@ -97,7 +91,6 @@ struct ti_qspi {
#define QSPI_RD_DUAL (3 << 16)
#define QSPI_RD_QUAD (7 << 16)
#define QSPI_INVAL (4 << 16)
-#define QSPI_WC_CMD_INT_EN (1 << 14)
#define QSPI_FLEN(n) ((n - 1) << 0)
#define QSPI_WLEN_MAX_BITS 128
#define QSPI_WLEN_MAX_BYTES 16
@@ -106,10 +99,6 @@ struct ti_qspi {
#define BUSY 0x01
#define WC 0x02
-/* INTERRUPT REGISTER */
-#define QSPI_WC_INT_EN (1 << 1)
-#define QSPI_WC_INT_DISABLE (1 << 1)
-
/* Device Control */
#define QSPI_DD(m, n) (m << (3 + n * 8))
#define QSPI_CKPHA(n) (1 << (2 + n * 8))
@@ -217,6 +206,24 @@ static inline u32 qspi_is_busy(struct ti_qspi *qspi)
return stat & BUSY;
}
+static inline int ti_qspi_poll_wc(struct ti_qspi *qspi)
+{
+ u32 stat;
+ unsigned long timeout = jiffies + QSPI_COMPLETION_TIMEOUT;
+
+ do {
+ stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
+ if (stat & WC)
+ return 0;
+ cpu_relax();
+ } while (time_after(timeout, jiffies));
+
+ stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
+ if (stat & WC)
+ return 0;
+ return -ETIMEDOUT;
+}
+
static int qspi_write_msg(struct ti_qspi *qspi, struct spi_transfer *t)
{
int wlen, count, xfer_len;
@@ -275,8 +282,7 @@ static int qspi_write_msg(struct ti_qspi *qspi, struct spi_transfer *t)
}
ti_qspi_write(qspi, cmd, QSPI_SPI_CMD_REG);
- if (!wait_for_completion_timeout(&qspi->transfer_complete,
- QSPI_COMPLETION_TIMEOUT)) {
+ if (ti_qspi_poll_wc(qspi)) {
dev_err(qspi->dev, "write timed out\n");
return -ETIMEDOUT;
}
@@ -315,8 +321,7 @@ static int qspi_read_msg(struct ti_qspi *qspi, struct spi_transfer *t)
return -EBUSY;
ti_qspi_write(qspi, cmd, QSPI_SPI_CMD_REG);
- if (!wait_for_completion_timeout(&qspi->transfer_complete,
- QSPI_COMPLETION_TIMEOUT)) {
+ if (ti_qspi_poll_wc(qspi)) {
dev_err(qspi->dev, "read timed out\n");
return -ETIMEDOUT;
}
@@ -388,9 +393,7 @@ static int ti_qspi_start_transfer_one(struct spi_master *master,
qspi->cmd = 0;
qspi->cmd |= QSPI_EN_CS(spi->chip_select);
qspi->cmd |= QSPI_FLEN(frame_length);
- qspi->cmd |= QSPI_WC_CMD_INT_EN;
- ti_qspi_write(qspi, QSPI_WC_INT_EN, QSPI_INTR_ENABLE_SET_REG);
ti_qspi_write(qspi, qspi->dc, QSPI_SPI_DC_REG);
mutex_lock(&qspi->list_lock);
@@ -417,31 +420,6 @@ static int ti_qspi_start_transfer_one(struct spi_master *master,
return status;
}
-static irqreturn_t ti_qspi_isr(int irq, void *dev_id)
-{
- struct ti_qspi *qspi = dev_id;
- u16 int_stat;
- u32 stat;
-
- irqreturn_t ret = IRQ_HANDLED;
-
- int_stat = ti_qspi_read(qspi, QSPI_INTR_STATUS_ENABLED_CLEAR);
- stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
-
- if (!int_stat) {
- dev_dbg(qspi->dev, "No IRQ triggered\n");
- ret = IRQ_NONE;
- goto out;
- }
-
- ti_qspi_write(qspi, QSPI_WC_INT_DISABLE,
- QSPI_INTR_STATUS_ENABLED_CLEAR);
- if (stat & WC)
- complete(&qspi->transfer_complete);
-out:
- return ret;
-}
-
static int ti_qspi_runtime_resume(struct device *dev)
{
struct ti_qspi *qspi;
@@ -550,22 +528,12 @@ static int ti_qspi_probe(struct platform_device *pdev)
}
}
- ret = devm_request_irq(&pdev->dev, irq, ti_qspi_isr, 0,
- dev_name(&pdev->dev), qspi);
- if (ret < 0) {
- dev_err(&pdev->dev, "Failed to register ISR for IRQ %d\n",
- irq);
- goto free_master;
- }
-
qspi->fclk = devm_clk_get(&pdev->dev, "fck");
if (IS_ERR(qspi->fclk)) {
ret = PTR_ERR(qspi->fclk);
dev_err(&pdev->dev, "could not get clk: %d\n", ret);
}
- init_completion(&qspi->transfer_complete);
-
pm_runtime_use_autosuspend(&pdev->dev);
pm_runtime_set_autosuspend_delay(&pdev->dev, QSPI_AUTOSUSPEND_TIMEOUT);
pm_runtime_enable(&pdev->dev);
@@ -595,8 +563,6 @@ static int ti_qspi_remove(struct platform_device *pdev)
return ret;
}
- ti_qspi_write(qspi, QSPI_WC_INT_DISABLE, QSPI_INTR_ENABLE_CLEAR_REG);
-
pm_runtime_put(qspi->dev);
pm_runtime_disable(&pdev->dev);
--
2.6.1
--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Applied "spi: spi-ti-qspi: switch to polling mode for better r/w performance" to the spi tree
[not found] ` <1444731665-11359-1-git-send-email-vigneshr-l0cyMroinI0@public.gmane.org>
@ 2015-10-16 18:11 ` Mark Brown
0 siblings, 0 replies; 2+ messages in thread
From: Mark Brown @ 2015-10-16 18:11 UTC (permalink / raw)
To: Vignesh R, Mark Brown; +Cc: linux-spi-u79uwXL29TY76Z2rM5mHXA
The patch
spi: spi-ti-qspi: switch to polling mode for better r/w performance
has been applied to the spi tree at
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next
tree (usually sometime in the next 24 hours) and sent to Linus during
the next merge window (or sooner if it is a bug fix), however if
problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing
and review of the tree, please engage with people reporting problems and
send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they
should be sent as incremental updates against current git, existing
patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying
to this mail.
Thanks,
Mark
>From 57c2ecd9bf971946ea0c6ae90a79c90a22159c73 Mon Sep 17 00:00:00 2001
From: Vignesh R <vigneshr-l0cyMroinI0@public.gmane.org>
Date: Tue, 13 Oct 2015 15:51:05 +0530
Subject: [PATCH] spi: spi-ti-qspi: switch to polling mode for better r/w
performance
Currently word completion interrupt is fired for transfer of every
word(8bit to 128bit in size). This adds a lot of overhead, and decreases
r/w throughput. It hardly takes 3us(@48MHz) for 128bit r/w to complete,
hence its better to poll on word complete bit to be set in
QSPI_SPI_STATUS_REG instead of using interrupts.
This increases the throughput by 30% in both read and write case.
So, switch to polling mode instead of interrupts to determine completion
of word transfer.
Signed-off-by: Vignesh R <vigneshr-l0cyMroinI0@public.gmane.org>
Signed-off-by: Mark Brown <broonie-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
---
drivers/spi/spi-ti-qspi.c | 74 +++++++++++++----------------------------------
1 file changed, 20 insertions(+), 54 deletions(-)
diff --git a/drivers/spi/spi-ti-qspi.c b/drivers/spi/spi-ti-qspi.c
index aa6d284131e0..89cf0c821524 100644
--- a/drivers/spi/spi-ti-qspi.c
+++ b/drivers/spi/spi-ti-qspi.c
@@ -39,8 +39,6 @@ struct ti_qspi_regs {
};
struct ti_qspi {
- struct completion transfer_complete;
-
/* list synchronization */
struct mutex list_lock;
@@ -62,10 +60,6 @@ struct ti_qspi {
#define QSPI_PID (0x0)
#define QSPI_SYSCONFIG (0x10)
-#define QSPI_INTR_STATUS_RAW_SET (0x20)
-#define QSPI_INTR_STATUS_ENABLED_CLEAR (0x24)
-#define QSPI_INTR_ENABLE_SET_REG (0x28)
-#define QSPI_INTR_ENABLE_CLEAR_REG (0x2c)
#define QSPI_SPI_CLOCK_CNTRL_REG (0x40)
#define QSPI_SPI_DC_REG (0x44)
#define QSPI_SPI_CMD_REG (0x48)
@@ -97,7 +91,6 @@ struct ti_qspi {
#define QSPI_RD_DUAL (3 << 16)
#define QSPI_RD_QUAD (7 << 16)
#define QSPI_INVAL (4 << 16)
-#define QSPI_WC_CMD_INT_EN (1 << 14)
#define QSPI_FLEN(n) ((n - 1) << 0)
#define QSPI_WLEN_MAX_BITS 128
#define QSPI_WLEN_MAX_BYTES 16
@@ -106,10 +99,6 @@ struct ti_qspi {
#define BUSY 0x01
#define WC 0x02
-/* INTERRUPT REGISTER */
-#define QSPI_WC_INT_EN (1 << 1)
-#define QSPI_WC_INT_DISABLE (1 << 1)
-
/* Device Control */
#define QSPI_DD(m, n) (m << (3 + n * 8))
#define QSPI_CKPHA(n) (1 << (2 + n * 8))
@@ -217,6 +206,24 @@ static inline u32 qspi_is_busy(struct ti_qspi *qspi)
return stat & BUSY;
}
+static inline int ti_qspi_poll_wc(struct ti_qspi *qspi)
+{
+ u32 stat;
+ unsigned long timeout = jiffies + QSPI_COMPLETION_TIMEOUT;
+
+ do {
+ stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
+ if (stat & WC)
+ return 0;
+ cpu_relax();
+ } while (time_after(timeout, jiffies));
+
+ stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
+ if (stat & WC)
+ return 0;
+ return -ETIMEDOUT;
+}
+
static int qspi_write_msg(struct ti_qspi *qspi, struct spi_transfer *t)
{
int wlen, count, xfer_len;
@@ -275,8 +282,7 @@ static int qspi_write_msg(struct ti_qspi *qspi, struct spi_transfer *t)
}
ti_qspi_write(qspi, cmd, QSPI_SPI_CMD_REG);
- if (!wait_for_completion_timeout(&qspi->transfer_complete,
- QSPI_COMPLETION_TIMEOUT)) {
+ if (ti_qspi_poll_wc(qspi)) {
dev_err(qspi->dev, "write timed out\n");
return -ETIMEDOUT;
}
@@ -315,8 +321,7 @@ static int qspi_read_msg(struct ti_qspi *qspi, struct spi_transfer *t)
return -EBUSY;
ti_qspi_write(qspi, cmd, QSPI_SPI_CMD_REG);
- if (!wait_for_completion_timeout(&qspi->transfer_complete,
- QSPI_COMPLETION_TIMEOUT)) {
+ if (ti_qspi_poll_wc(qspi)) {
dev_err(qspi->dev, "read timed out\n");
return -ETIMEDOUT;
}
@@ -388,9 +393,7 @@ static int ti_qspi_start_transfer_one(struct spi_master *master,
qspi->cmd = 0;
qspi->cmd |= QSPI_EN_CS(spi->chip_select);
qspi->cmd |= QSPI_FLEN(frame_length);
- qspi->cmd |= QSPI_WC_CMD_INT_EN;
- ti_qspi_write(qspi, QSPI_WC_INT_EN, QSPI_INTR_ENABLE_SET_REG);
ti_qspi_write(qspi, qspi->dc, QSPI_SPI_DC_REG);
mutex_lock(&qspi->list_lock);
@@ -418,31 +421,6 @@ static int ti_qspi_start_transfer_one(struct spi_master *master,
return status;
}
-static irqreturn_t ti_qspi_isr(int irq, void *dev_id)
-{
- struct ti_qspi *qspi = dev_id;
- u16 int_stat;
- u32 stat;
-
- irqreturn_t ret = IRQ_HANDLED;
-
- int_stat = ti_qspi_read(qspi, QSPI_INTR_STATUS_ENABLED_CLEAR);
- stat = ti_qspi_read(qspi, QSPI_SPI_STATUS_REG);
-
- if (!int_stat) {
- dev_dbg(qspi->dev, "No IRQ triggered\n");
- ret = IRQ_NONE;
- goto out;
- }
-
- ti_qspi_write(qspi, QSPI_WC_INT_DISABLE,
- QSPI_INTR_STATUS_ENABLED_CLEAR);
- if (stat & WC)
- complete(&qspi->transfer_complete);
-out:
- return ret;
-}
-
static int ti_qspi_runtime_resume(struct device *dev)
{
struct ti_qspi *qspi;
@@ -551,22 +529,12 @@ static int ti_qspi_probe(struct platform_device *pdev)
}
}
- ret = devm_request_irq(&pdev->dev, irq, ti_qspi_isr, 0,
- dev_name(&pdev->dev), qspi);
- if (ret < 0) {
- dev_err(&pdev->dev, "Failed to register ISR for IRQ %d\n",
- irq);
- goto free_master;
- }
-
qspi->fclk = devm_clk_get(&pdev->dev, "fck");
if (IS_ERR(qspi->fclk)) {
ret = PTR_ERR(qspi->fclk);
dev_err(&pdev->dev, "could not get clk: %d\n", ret);
}
- init_completion(&qspi->transfer_complete);
-
pm_runtime_use_autosuspend(&pdev->dev);
pm_runtime_set_autosuspend_delay(&pdev->dev, QSPI_AUTOSUSPEND_TIMEOUT);
pm_runtime_enable(&pdev->dev);
@@ -596,8 +564,6 @@ static int ti_qspi_remove(struct platform_device *pdev)
return ret;
}
- ti_qspi_write(qspi, QSPI_WC_INT_DISABLE, QSPI_INTR_ENABLE_CLEAR_REG);
-
pm_runtime_put(qspi->dev);
pm_runtime_disable(&pdev->dev);
--
2.6.1
--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 2+ messages in thread
end of thread, other threads:[~2015-10-16 18:11 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-10-13 10:21 [PATCH] spi: spi-ti-qspi: switch to polling mode for better r/w performance Vignesh R
[not found] ` <1444731665-11359-1-git-send-email-vigneshr-l0cyMroinI0@public.gmane.org>
2015-10-16 18:11 ` Applied "spi: spi-ti-qspi: switch to polling mode for better r/w performance" to the spi tree Mark Brown
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 NNTP newsgroup(s).