[PATCH 1/1] mtd: fsl-quadspi: improve read performance by increase AHB transfer size

[PATCH 1/1] mtd: fsl-quadspi: improve read performance by increase AHB transfer size

[Frank.Li]

From: Allen Xu <b45815@freescale.com>

Set AHB transfer size to 1K which improved the read performance.

Add ahb_buf_size field in fsl_qspi_devtype_data to denote the size for
different SoC.

Before:

root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
16+0 records in
16+0 records out
16777216 bytes (17 MB) copied, 0.472183 s, 25.1 MB/s

After:

root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
16+0 records in
16+0 records out
16777216 bytes (17 MB) copied, 0.369439 s, 29.5 MB/s

Signed-off-by: Allen Xu <b45815@freescale.com>
Signed-off-by: Huang Shijie <shijie8@gmail.com>
Signed-off-by: Frank Li <Frank.Li@freescale.com>
---
 drivers/mtd/spi-nor/fsl-quadspi.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/mtd/spi-nor/fsl-quadspi.c b/drivers/mtd/spi-nor/fsl-quadspi.c
index 9d608007..75d6898 100644
--- a/drivers/mtd/spi-nor/fsl-quadspi.c
+++ b/drivers/mtd/spi-nor/fsl-quadspi.c
@@ -57,7 +57,9 @@
 
 #define QUADSPI_BUF3CR			0x1c
 #define QUADSPI_BUF3CR_ALLMST_SHIFT	31
-#define QUADSPI_BUF3CR_ALLMST		(1 << QUADSPI_BUF3CR_ALLMST_SHIFT)
+#define QUADSPI_BUF3CR_ALLMST_MASK	(1 << QUADSPI_BUF3CR_ALLMST_SHIFT)
+#define QUADSPI_BUF3CR_ADATSZ_SHIFT		8
+#define QUADSPI_BUF3CR_ADATSZ_MASK	(0xFF << QUADSPI_BUF3CR_ADATSZ_SHIFT)
 
 #define QUADSPI_BFGENCR			0x20
 #define QUADSPI_BFGENCR_PAR_EN_SHIFT	16
@@ -198,18 +200,21 @@ struct fsl_qspi_devtype_data {
 	enum fsl_qspi_devtype devtype;
 	int rxfifo;
 	int txfifo;
+	int ahb_buf_size;
 };
 
 static struct fsl_qspi_devtype_data vybrid_data = {
 	.devtype = FSL_QUADSPI_VYBRID,
 	.rxfifo = 128,
-	.txfifo = 64
+	.txfifo = 64,
+	.ahb_buf_size = 1024
 };
 
 static struct fsl_qspi_devtype_data imx6sx_data = {
 	.devtype = FSL_QUADSPI_IMX6SX,
 	.rxfifo = 128,
-	.txfifo = 512
+	.txfifo = 512,
+	.ahb_buf_size = 1024
 };
 
 #define FSL_QSPI_MAX_CHIP	4
@@ -583,7 +588,12 @@ static void fsl_qspi_init_abh_read(struct fsl_qspi *q)
 	writel(QUADSPI_BUFXCR_INVALID_MSTRID, base + QUADSPI_BUF0CR);
 	writel(QUADSPI_BUFXCR_INVALID_MSTRID, base + QUADSPI_BUF1CR);
 	writel(QUADSPI_BUFXCR_INVALID_MSTRID, base + QUADSPI_BUF2CR);
-	writel(QUADSPI_BUF3CR_ALLMST, base + QUADSPI_BUF3CR);
+	/*
+	 * Set ADATSZ with the maximum AHB buffer size to improve the
+	 * read performance.
+	 */
+	writel(QUADSPI_BUF3CR_ALLMST_MASK | ((q->devtype_data->ahb_buf_size / 8)
+			<< QUADSPI_BUF3CR_ADATSZ_SHIFT), base + QUADSPI_BUF3CR);
 
 	/* We only use the buffer3 */
 	writel(0, base + QUADSPI_BUF0IND);
-- 
1.9.1

Re: [PATCH 1/1] mtd: fsl-quadspi: improve read performance by increase AHB transfer size

[Huang Shijie]

On Wed, Jan 14, 2015 at 12:28:56AM +0800, Frank.Li@freescale.com wrote:
> From: Allen Xu <b45815@freescale.com>
> 
> Set AHB transfer size to 1K which improved the read performance.
> 
> Add ahb_buf_size field in fsl_qspi_devtype_data to denote the size for
> different SoC.
> 
> Before:
> 
> root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
> 16+0 records in
> 16+0 records out
> 16777216 bytes (17 MB) copied, 0.472183 s, 25.1 MB/s
> 
> After:
> 
> root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
> 16+0 records in
> 16+0 records out
> 16777216 bytes (17 MB) copied, 0.369439 s, 29.5 MB/s
> 
> Signed-off-by: Allen Xu <b45815@freescale.com>
> Signed-off-by: Huang Shijie <shijie8@gmail.com>
> Signed-off-by: Frank Li <Frank.Li@freescale.com>
This is a very good patch.

This patch has my signed-off, but if it is needed, please add my ack:
Acked-by: Huang Shijie <shijie.huang@intel.com>

Re: [PATCH 1/1] mtd: fsl-quadspi: improve read performance by increase AHB transfer size

[Brian Norris]

On Wed, Jan 14, 2015 at 12:28:56AM +0800, Frank.Li@freescale.com wrote:
> From: Allen Xu <b45815@freescale.com>
> 
> Set AHB transfer size to 1K which improved the read performance.
> 
> Add ahb_buf_size field in fsl_qspi_devtype_data to denote the size for
> different SoC.
> 
> Before:
> 
> root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
> 16+0 records in
> 16+0 records out
> 16777216 bytes (17 MB) copied, 0.472183 s, 25.1 MB/s
> 
> After:
> 
> root@imx6qdlsolo:~# dd if=/dev/mtd1 of=/dev/null bs=1M count=16
> 16+0 records in
> 16+0 records out
> 16777216 bytes (17 MB) copied, 0.369439 s, 29.5 MB/s
> 
> Signed-off-by: Allen Xu <b45815@freescale.com>
> Signed-off-by: Huang Shijie <shijie8@gmail.com>
> Signed-off-by: Frank Li <Frank.Li@freescale.com>

Pushed to l2-mtd.git. Thanks.

Brian