[PATCH] [ARM] orion5x: accelerate NAND on the TS-78xx

[PATCH] [ARM] orion5x: accelerate NAND on the TS-78xx

[Alexander Clouter]

The NAND supports 32bit reads and writes so lets stop shunting 8bit
chunks across the bus.

Doing a dumb 'dd' benchmark, this increases performance roughly like so:
 * read: 1.3MB/s to 3.4MB/s
 * write: 614kB/s to 882kB/s

Signed-off-by: Alexander Clouter <alex@digriz.org.uk>
---
 arch/arm/mach-orion5x/ts78xx-setup.c |   42 ++++++++++++++++++++++++++++++++++
 1 files changed, 42 insertions(+), 0 deletions(-)

diff --git a/arch/arm/mach-orion5x/ts78xx-setup.c b/arch/arm/mach-orion5x/ts78xx-setup.c
index 9a5d1ef..bcc21fe 100644
--- a/arch/arm/mach-orion5x/ts78xx-setup.c
+++ b/arch/arm/mach-orion5x/ts78xx-setup.c
@@ -191,6 +191,46 @@ static int ts78xx_ts_nand_dev_ready(struct mtd_info *mtd)
 	return readb(TS_NAND_CTRL) & 0x20;
 }
 
+static void ts78xx_ts_nand_write_buf(struct mtd_info *mtd,
+			const uint8_t *buf, int len)
+{
+	struct nand_chip *chip = mtd->priv;
+	void __iomem *io_base = chip->IO_ADDR_W;
+	uint32_t *buf32;
+	int i = 0;
+
+	while (len && (unsigned int)buf & 3) {
+		writeb(*buf++, io_base);
+		len--;
+	}
+	buf32 = (uint32_t *)buf;
+	while (i < len/4)
+		writel(buf32[i++], io_base);
+	i *= 4;
+	while (i < len)
+		writeb(buf[i++], io_base);
+}
+
+static void ts78xx_ts_nand_read_buf(struct mtd_info *mtd,
+			uint8_t *buf, int len)
+{
+	struct nand_chip *chip = mtd->priv;
+	void __iomem *io_base = chip->IO_ADDR_R;
+	uint32_t *buf32;
+	int i = 0;
+
+	while (len && (unsigned int)buf & 3) {
+		*buf++ = readb(io_base);
+		len--;
+	}
+	buf32 = (uint32_t *)buf;
+	while (i < len/4)
+		buf32[i++] = readl(io_base);
+	i *= 4;
+	while (i < len)
+		buf[i++] = readb(io_base);
+}
+
 const char *ts_nand_part_probes[] = { "cmdlinepart", NULL };
 
 static struct mtd_partition ts78xx_ts_nand_parts[] = {
@@ -233,6 +273,8 @@ static struct platform_nand_data ts78xx_ts_nand_data = {
 		 */
 		.cmd_ctrl		= ts78xx_ts_nand_cmd_ctrl,
 		.dev_ready		= ts78xx_ts_nand_dev_ready,
+		.write_buf		= ts78xx_ts_nand_write_buf,
+		.read_buf		= ts78xx_ts_nand_read_buf,
 	},
 };
 
-- 
1.7.2.3

[PATCH] [ARM] orion5x: accelerate NAND on the TS-78xx

[Russell King - ARM Linux]

On Tue, Jan 04, 2011 at 11:51:59PM +0000, Alexander Clouter wrote:
> The NAND supports 32bit reads and writes so lets stop shunting 8bit
> chunks across the bus.
> 
> Doing a dumb 'dd' benchmark, this increases performance roughly like so:
>  * read: 1.3MB/s to 3.4MB/s
>  * write: 614kB/s to 882kB/s

Try something like the below.  It's slightly more typing, but the
underlying string IO functions should improve your transfer speed.
Note that they won't do endian conversions.

> 
> Signed-off-by: Alexander Clouter <alex@digriz.org.uk>
> ---
>  arch/arm/mach-orion5x/ts78xx-setup.c |   42 ++++++++++++++++++++++++++++++++++
>  1 files changed, 42 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/arm/mach-orion5x/ts78xx-setup.c b/arch/arm/mach-orion5x/ts78xx-setup.c
> index 9a5d1ef..bcc21fe 100644
> --- a/arch/arm/mach-orion5x/ts78xx-setup.c
> +++ b/arch/arm/mach-orion5x/ts78xx-setup.c
> @@ -191,6 +191,46 @@ static int ts78xx_ts_nand_dev_ready(struct mtd_info *mtd)
>  	return readb(TS_NAND_CTRL) & 0x20;
>  }
>  
> +static void ts78xx_ts_nand_write_buf(struct mtd_info *mtd,
> +			const uint8_t *buf, int len)
> +{
> +	struct nand_chip *chip = mtd->priv;
> +	void __iomem *io_base = chip->IO_ADDR_W;
> +	uint32_t *buf32;
> +	int i = 0;
> +
> +	while (len && (unsigned int)buf & 3) {
> +		writeb(*buf++, io_base);
> +		len--;
> +	}
> +	buf32 = (uint32_t *)buf;
> +	while (i < len/4)
> +		writel(buf32[i++], io_base);
> +	i *= 4;
> +	while (i < len)
> +		writeb(buf[i++], io_base);

	unsigned long off = ((unsigned long)buf & 3);
	int sz;

	if (off) {
		sz = min(4 - off, len);
		writesb(io_base, buf, sz);
		buf += sz;
		len -= sz;
	}

	sz = len >> 2;
	if (sz) {
		u32 *buf32 = (u32 *)buf;
		writesl(io_base, buf32, sz);
		buf += sz << 2;
		len -= sz << 2;
	}

	if (len)
		writesb(io_base, buf, len);

> +}
> +
> +static void ts78xx_ts_nand_read_buf(struct mtd_info *mtd,
> +			uint8_t *buf, int len)
> +{
> +	struct nand_chip *chip = mtd->priv;
> +	void __iomem *io_base = chip->IO_ADDR_R;
> +	uint32_t *buf32;
> +	int i = 0;
> +
> +	while (len && (unsigned int)buf & 3) {
> +		*buf++ = readb(io_base);
> +		len--;
> +	}
> +	buf32 = (uint32_t *)buf;
> +	while (i < len/4)
> +		buf32[i++] = readl(io_base);
> +	i *= 4;
> +	while (i < len)
> +		buf[i++] = readb(io_base);

	unsigned long off = ((unsigned long)buf & 3);
	int sz;

	if (off) {
		sz = min(4 - off, len);
		readsb(io_base, buf, sz);
		buf += sz;
		len -= sz;
	}

	sz = len >> 2;
	if (sz) {
		u32 *buf32 = (u32 *)buf;
		readsl(io_base, buf32, sz);
		buf += sz << 2;
		len -= sz << 2;
	}

	if (len)
		readsb(io_base, buf, len);

> +}
> +
>  const char *ts_nand_part_probes[] = { "cmdlinepart", NULL };
>  
>  static struct mtd_partition ts78xx_ts_nand_parts[] = {
> @@ -233,6 +273,8 @@ static struct platform_nand_data ts78xx_ts_nand_data = {
>  		 */
>  		.cmd_ctrl		= ts78xx_ts_nand_cmd_ctrl,
>  		.dev_ready		= ts78xx_ts_nand_dev_ready,
> +		.write_buf		= ts78xx_ts_nand_write_buf,
> +		.read_buf		= ts78xx_ts_nand_read_buf,
>  	},
>  };
>  
> -- 
> 1.7.2.3
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH] [ARM] orion5x: accelerate NAND on the TS-78xx

[Alexander Clouter]

Russell King - ARM Linux <linux@arm.linux.org.uk> wrote:
>
> On Tue, Jan 04, 2011 at 11:51:59PM +0000, Alexander Clouter wrote:
>> The NAND supports 32bit reads and writes so lets stop shunting 8bit
>> chunks across the bus.
>> 
>> Doing a dumb 'dd' benchmark, this increases performance roughly like so:
>>  * read: 1.3MB/s to 3.4MB/s
>>  * write: 614kB/s to 882kB/s
> 
> Try something like the below.  It's slightly more typing, but the
> underlying string IO functions should improve your transfer speed.
> Note that they won't do endian conversions.
>
I get roughly the same speeds with your approach...I am happy to swing 
either your way or my original implementation; well actually it was 
stolen from Nico's orion_nand implementation :)

What should I submit?  These functions will at somestage get DMA support 
added.
 
Cheers

-- 
Alexander Clouter
.sigmonster says: Live Free or Live in Massachusetts.

[PATCH] [ARM] orion5x: accelerate NAND on the TS-78xx

[Alexander Clouter]

Alexander Clouter <alex@digriz.org.uk> wrote:
>>
>> On Tue, Jan 04, 2011 at 11:51:59PM +0000, Alexander Clouter wrote:
>>> The NAND supports 32bit reads and writes so lets stop shunting 8bit
>>> chunks across the bus.
>>> 
>>> Doing a dumb 'dd' benchmark, this increases performance roughly like so:
>>>  * read: 1.3MB/s to 3.4MB/s
>>>  * write: 614kB/s to 882kB/s
>> 
>> Try something like the below.  It's slightly more typing, but the
>> underlying string IO functions should improve your transfer speed.
>> Note that they won't do endian conversions.
>>
> I get roughly the same speeds with your approach...I am happy to swing 
> either your way or my original implementation; well actually it was 
> stolen from Nico's orion_nand implementation :)
> 
> What should I submit?  These functions will at somestage get DMA support 
> added.
> 
I decided I prefer your version and it 'hides' the while loops and is 
easier to read.  

Cheers

-- 
Alexander Clouter
.sigmonster says: What!?  Me worry?
                  		-- Alfred E. Newman