Patchwork [U-Boot,PATCHv4] Optimized nand_read_buf for kirkwood

login
register
mail settings
Submitter Phil Sutter
Date Aug. 26, 2013, 12:10 p.m.
Message ID <1377519056-25364-1-git-send-email-phil.sutter@viprinet.com>
Download mbox | patch
Permalink /patch/269873/
State Accepted
Delegated to: Scott Wood
Headers show

Comments

Phil Sutter - Aug. 26, 2013, 12:10 p.m.
From: Nico Erfurth <ne@erfurth.eu>

The basic idea is taken from the linux-kernel, but further optimized.

First align the buffer to 8 bytes, then use ldrd/strd to read and store
in 8 byte quantities, then do the final bytes.

Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
patch in place, reading the same amount of data was done in 27s
(~4.89MB/s). So read performance is increased by ~80%!

Signed-off-by: Nico Erfurth <ne@erfurth.eu>
Tested-by: Phil Sutter <phil.sutter@viprinet.com>
Cc: Prafulla Wadaskar <prafulla@marvell.com>
---
Changed since V3:
- fixed author
---
 drivers/mtd/nand/kirkwood_nand.c | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)
Scott Wood - Aug. 26, 2013, 3:43 p.m.
On Mon, 2013-08-26 at 14:10 +0200, Phil Sutter wrote:
> From: Nico Erfurth <ne@erfurth.eu>
> 
> The basic idea is taken from the linux-kernel, but further optimized.
> 
> First align the buffer to 8 bytes, then use ldrd/strd to read and store
> in 8 byte quantities, then do the final bytes.
> 
> Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> patch in place, reading the same amount of data was done in 27s
> (~4.89MB/s). So read performance is increased by ~80%!
> 
> Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> Cc: Prafulla Wadaskar <prafulla@marvell.com>
> ---
> Changed since V3:
> - fixed author

It needs your Signed-off-by: as well -- can I add that when applying?

-Scott
Phil Sutter - Aug. 26, 2013, 4 p.m.
On Mon, Aug 26, 2013 at 10:43:38AM -0500, Scott Wood wrote:
> On Mon, 2013-08-26 at 14:10 +0200, Phil Sutter wrote:
> > From: Nico Erfurth <ne@erfurth.eu>
> > 
> > The basic idea is taken from the linux-kernel, but further optimized.
> > 
> > First align the buffer to 8 bytes, then use ldrd/strd to read and store
> > in 8 byte quantities, then do the final bytes.
> > 
> > Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> > Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> > patch in place, reading the same amount of data was done in 27s
> > (~4.89MB/s). So read performance is increased by ~80%!
> > 
> > Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> > Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> > Cc: Prafulla Wadaskar <prafulla@marvell.com>
> > ---
> > Changed since V3:
> > - fixed author
> 
> It needs your Signed-off-by: as well -- can I add that when applying?

Yes, that's fine by me. So if I submit others' patches, I need to sign
them off as well?

Best wishes,

Phil Sutter
Software Engineer
Phil Sutter - Aug. 26, 2013, 4:06 p.m.
On Mon, Aug 26, 2013 at 06:00:39PM +0200, Phil Sutter wrote:
> On Mon, Aug 26, 2013 at 10:43:38AM -0500, Scott Wood wrote:
> > On Mon, 2013-08-26 at 14:10 +0200, Phil Sutter wrote:
> > > From: Nico Erfurth <ne@erfurth.eu>
> > > 
> > > The basic idea is taken from the linux-kernel, but further optimized.
> > > 
> > > First align the buffer to 8 bytes, then use ldrd/strd to read and store
> > > in 8 byte quantities, then do the final bytes.
> > > 
> > > Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> > > Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> > > patch in place, reading the same amount of data was done in 27s
> > > (~4.89MB/s). So read performance is increased by ~80%!
> > > 
> > > Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> > > Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> > > Cc: Prafulla Wadaskar <prafulla@marvell.com>
> > > ---
> > > Changed since V3:
> > > - fixed author
> > 
> > It needs your Signed-off-by: as well -- can I add that when applying?
> 
> Yes, that's fine by me. So if I submit others' patches, I need to sign
> them off as well?

Ah, nevermind. Reading 1.12 of Documentation/SubmittingPatches helps.

Best wishes,

Phil Sutter
Software Engineer
Scott Wood - Nov. 14, 2013, 1:31 a.m.
On Mon, Aug 26, 2013 at 02:10:56PM +0200, Phil Sutter wrote:
> From: Nico Erfurth <ne@erfurth.eu>
> 
> The basic idea is taken from the linux-kernel, but further optimized.
> 
> First align the buffer to 8 bytes, then use ldrd/strd to read and store
> in 8 byte quantities, then do the final bytes.
> 
> Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> patch in place, reading the same amount of data was done in 27s
> (~4.89MB/s). So read performance is increased by ~80%!
> 
> Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> Cc: Prafulla Wadaskar <prafulla@marvell.com>
> 
> ---
> Changed since V3:
> - fixed author
> ---
>  drivers/mtd/nand/kirkwood_nand.c | 32 ++++++++++++++++++++++++++++++++
>  1 file changed, 32 insertions(+)

I tried to build-test this, and I couldn't find any board that defines
CONFIG_NAND_KIRKWOOD.

The patch that removed it was commit
b5befd8211b54ae2d2fca3fbed061c879951ceaa ("arm/km: fix u-boot.kwb build
breakage"), over two years ago.  It's not clear whether the removal was
intentional.

What target did you use to test this?

-Scott
Holger Brunck - Nov. 14, 2013, 8:18 a.m.
Hi Scott,

On 11/14/2013 02:31 AM, Scott Wood wrote:
> On Mon, Aug 26, 2013 at 02:10:56PM +0200, Phil Sutter wrote:
>> From: Nico Erfurth <ne@erfurth.eu>
>>
>> The basic idea is taken from the linux-kernel, but further optimized.
>>
>> First align the buffer to 8 bytes, then use ldrd/strd to read and store
>> in 8 byte quantities, then do the final bytes.
>>
>> Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
>> Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
>> patch in place, reading the same amount of data was done in 27s
>> (~4.89MB/s). So read performance is increased by ~80%!
>>
>> Signed-off-by: Nico Erfurth <ne@erfurth.eu>
>> Tested-by: Phil Sutter <phil.sutter@viprinet.com>
>> Cc: Prafulla Wadaskar <prafulla@marvell.com>
>>
>> ---
>> Changed since V3:
>> - fixed author
>> ---
>>  drivers/mtd/nand/kirkwood_nand.c | 32 ++++++++++++++++++++++++++++++++
>>  1 file changed, 32 insertions(+)
> 
> I tried to build-test this, and I couldn't find any board that defines
> CONFIG_NAND_KIRKWOOD.
> 

it's not in board specific code defined it's defined in a common kirkwood header:

arch/arm/include/asm/arch-kirkwood/config.h:58:#define CONFIG_NAND_KIRKWOOD

> The patch that removed it was commit
> b5befd8211b54ae2d2fca3fbed061c879951ceaa ("arm/km: fix u-boot.kwb build
> breakage"), over two years ago.  It's not clear whether the removal was
> intentional.
>

yes it was. We include this common header and therefore we don't need to
redefine it in our board setup.

Regards
Holger
Phil Sutter - Nov. 14, 2013, 12:12 p.m.
Scott,

On Wed, Nov 13, 2013 at 07:31:02PM -0600, Scott Wood wrote:
> On Mon, Aug 26, 2013 at 02:10:56PM +0200, Phil Sutter wrote:
> > From: Nico Erfurth <ne@erfurth.eu>
> > 
> > The basic idea is taken from the linux-kernel, but further optimized.
> > 
> > First align the buffer to 8 bytes, then use ldrd/strd to read and store
> > in 8 byte quantities, then do the final bytes.
> > 
> > Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> > Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> > patch in place, reading the same amount of data was done in 27s
> > (~4.89MB/s). So read performance is increased by ~80%!
> > 
> > Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> > Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> > Cc: Prafulla Wadaskar <prafulla@marvell.com>
> > 
> > ---
> > Changed since V3:
> > - fixed author
> > ---
> >  drivers/mtd/nand/kirkwood_nand.c | 32 ++++++++++++++++++++++++++++++++
> >  1 file changed, 32 insertions(+)
> 
> I tried to build-test this, and I couldn't find any board that defines
> CONFIG_NAND_KIRKWOOD.
> 
> The patch that removed it was commit
> b5befd8211b54ae2d2fca3fbed061c879951ceaa ("arm/km: fix u-boot.kwb build
> breakage"), over two years ago.  It's not clear whether the removal was
> intentional.
> 
> What target did you use to test this?

I tested using a custom board with Marvell Kirkwood SoC, but e.g. the
Marvell OpenRD Ultimate should be fine.

Best wishes,

Phil Sutter
Software Engineer
Scott Wood - Nov. 14, 2013, 11:59 p.m.
On Thu, 2013-11-14 at 09:18 +0100, Holger Brunck wrote:
> Hi Scott,
> 
> On 11/14/2013 02:31 AM, Scott Wood wrote:
> > On Mon, Aug 26, 2013 at 02:10:56PM +0200, Phil Sutter wrote:
> >> From: Nico Erfurth <ne@erfurth.eu>
> >>
> >> The basic idea is taken from the linux-kernel, but further optimized.
> >>
> >> First align the buffer to 8 bytes, then use ldrd/strd to read and store
> >> in 8 byte quantities, then do the final bytes.
> >>
> >> Tested using: 'date ; nand read.raw 0xE00000 0x0 0x10000 ; date'.
> >> Without this patch, NAND read of 132MB took 49s (~2.69MB/s). With this
> >> patch in place, reading the same amount of data was done in 27s
> >> (~4.89MB/s). So read performance is increased by ~80%!
> >>
> >> Signed-off-by: Nico Erfurth <ne@erfurth.eu>
> >> Tested-by: Phil Sutter <phil.sutter@viprinet.com>
> >> Cc: Prafulla Wadaskar <prafulla@marvell.com>
> >>
> >> ---
> >> Changed since V3:
> >> - fixed author
> >> ---
> >>  drivers/mtd/nand/kirkwood_nand.c | 32 ++++++++++++++++++++++++++++++++
> >>  1 file changed, 32 insertions(+)
> > 
> > I tried to build-test this, and I couldn't find any board that defines
> > CONFIG_NAND_KIRKWOOD.
> > 
> 
> it's not in board specific code defined it's defined in a common kirkwood header:
> 
> arch/arm/include/asm/arch-kirkwood/config.h:58:#define CONFIG_NAND_KIRKWOOD

Oops.  I thought I grepped the whole tree rather than just include/, but
apparently not.

-Scott

Patch

diff --git a/drivers/mtd/nand/kirkwood_nand.c b/drivers/mtd/nand/kirkwood_nand.c
index 0a99a10..85ea5d2 100644
--- a/drivers/mtd/nand/kirkwood_nand.c
+++ b/drivers/mtd/nand/kirkwood_nand.c
@@ -38,6 +38,37 @@  struct kwnandf_registers {
 static struct kwnandf_registers *nf_reg =
 	(struct kwnandf_registers *)KW_NANDF_BASE;
 
+
+/*
+ * The basic idea is stolen from the linux kernel, but the inner loop is
+ * optimized a bit more.
+ */
+static void kw_nand_read_buf(struct mtd_info *mtd, uint8_t *buf, int len)
+{
+	struct nand_chip *chip = mtd->priv;
+
+	while (len && (unsigned long)buf & 7) {
+		*buf++ = readb(chip->IO_ADDR_R);
+		len--;
+	};
+
+	/* This loop reads and writes 64bit per round. */
+	asm volatile (
+		"1:\n"
+		"  subs   %0, #8\n"
+		"  ldrpld r2, [%2]\n"
+		"  strpld r2, [%1], #8\n"
+		"  bhi    1b\n"
+		"  addne  %0, #8\n"
+		: "+&r" (len), "+&r" (buf)
+		: "r" (chip->IO_ADDR_R)
+		: "r2", "r3", "memory", "cc"
+	);
+
+	while (len--)
+		*buf++ = readb(chip->IO_ADDR_R);
+}
+
 /*
  * hardware specific access to control-lines/bits
  */
@@ -80,6 +111,7 @@  int board_nand_init(struct nand_chip *nand)
 	nand->ecc.mode = NAND_ECC_SOFT;
 #endif
 	nand->cmd_ctrl = kw_nand_hwcontrol;
+	nand->read_buf = kw_nand_read_buf;
 	nand->chip_delay = 40;
 	nand->select_chip = kw_nand_select_chip;
 	return 0;