* [RFC] spi: Using Trigger number to transmit/receive data @ 2014-03-13 8:29 Cao Minh Hiep 2014-03-13 8:29 ` Cao Minh Hiep 2014-03-13 14:17 ` Mark Brown 0 siblings, 2 replies; 14+ messages in thread From: Cao Minh Hiep @ 2014-03-13 8:29 UTC (permalink / raw) To: broonie; +Cc: devicetree, linux-sh, kuninori.morimoto.gx, geert+renesas From: Hiep Cao Minh <cm-hiep@jinso.co.jp> Hi, This is the patch of QSPI that support for r8a7790 SOC. The pupose of this patch is to transmit and receive the 32 bytes of data when the data already has prepared on Transmit/Receive Buffer. Using Transmit/Receive Buffer Data Trigger Number to do that instead transmits/receives a byte of data. With this patch will improve the speed of transfer data. These patches was developed base on the upstream-v3.14-rc5 and have tested on the Lager board. Please consider the following patches for the r8a7790 Soc. Thank you! Cao Minh Hiep Hiep Cao Minh (1): spi: Using Trigger number to transmit/receive data drivers/spi/spi-rspi.c | 144 +++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 124 insertions(+), 20 deletions(-) -- 1.7.9.5 ^ permalink raw reply [flat|nested] 14+ messages in thread
* [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 8:29 [RFC] spi: Using Trigger number to transmit/receive data Cao Minh Hiep @ 2014-03-13 8:29 ` Cao Minh Hiep 2014-03-13 9:23 ` Geert Uytterhoeven 2014-03-13 9:27 ` Kuninori Morimoto 2014-03-13 14:17 ` Mark Brown 1 sibling, 2 replies; 14+ messages in thread From: Cao Minh Hiep @ 2014-03-13 8:29 UTC (permalink / raw) To: broonie; +Cc: devicetree, linux-sh, kuninori.morimoto.gx, geert+renesas From: Hiep Cao Minh <cm-hiep@jinso.co.jp> In order to transmit and receive data when have 32 bytes of data that ready has prepared on Transmit/Receive Buffer to transmit or receive. Instead transmits/receives a byte data using Transmit/Receive Buffer Data Triggering Number will improve the speed of transfer data. Signed-off-by: Hiep Cao Minh <cm-hiep@jinso.co.jp> --- drivers/spi/spi-rspi.c | 144 +++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 124 insertions(+), 20 deletions(-) diff --git a/drivers/spi/spi-rspi.c b/drivers/spi/spi-rspi.c index 28987d9..772a942 100644 --- a/drivers/spi/spi-rspi.c +++ b/drivers/spi/spi-rspi.c @@ -174,6 +174,12 @@ #define SPBFCR_RXRST 0x40 /* Receive Buffer Data Reset (qspi only) */ #define SPBFCR_TXTRG_MASK 0x30 /* Transmit Buffer Data Triggering Number */ #define SPBFCR_RXTRG_MASK 0x07 /* Receive Buffer Data Triggering Number */ +#define SPBFCR_TXTRG_31B 0x00 /* QSPI only */ +#define SPBFCR_TXTRG_0B 0x30 /* QSPI only */ +#define SPBFCR_RXTRG_1B 0x00 /* QSPI only */ +#define SPBFCR_RXTRG_32B 0x07 /* QSPI only */ + +#define QSPI_BUFFER_SIZE 32 #define DUMMY_DATA 0x00 @@ -224,6 +230,16 @@ static u16 rspi_read16(const struct rspi_data *rspi, u16 offset) return ioread16(rspi->addr + offset); } +static void qspi_update(const struct rspi_data *rspi, u8 mask, u8 val, u8 reg) +{ + u8 data; + + data = rspi_read8(rspi, reg); + data &= ~mask; + data |= (val & mask); + rspi_write8(rspi, data, reg); +} + /* optional functions */ struct spi_ops { int (*set_config_register)(const struct rspi_data *rspi, @@ -358,6 +374,44 @@ static void rspi_negate_ssl(const struct rspi_data *rspi) rspi_write8(rspi, rspi_read8(rspi, RSPI_SPCR) & ~SPCR_SPE, RSPI_SPCR); } +static int qspi_set_send_trigger(struct rspi_data *rspi, int remain) +{ + int n; + + n = min(remain, QSPI_BUFFER_SIZE); + + if (remain >= QSPI_BUFFER_SIZE) { + /* sets triggering number to 32 bytes */ + qspi_update(rspi, SPBFCR_TXTRG_MASK, + SPBFCR_TXTRG_0B, QSPI_SPBFCR); + } else { + /* sets triggering number to 1 byte */ + qspi_update(rspi, SPBFCR_TXTRG_MASK, + SPBFCR_TXTRG_31B, QSPI_SPBFCR); + } + + return n; +} + +static int qspi_set_receive_trigger(struct rspi_data *rspi, int remain) +{ + int n; + + n = min(remain, QSPI_BUFFER_SIZE); + + if (remain >= QSPI_BUFFER_SIZE) { + /* sets triggering number to 32 bytes */ + qspi_update(rspi, SPBFCR_RXTRG_MASK, + SPBFCR_RXTRG_32B, QSPI_SPBFCR); + } else { + /* sets triggering number to 1 byte */ + qspi_update(rspi, SPBFCR_RXTRG_MASK, + SPBFCR_RXTRG_1B, QSPI_SPBFCR); + } + + return n; +} + static int rspi_send_pio(struct rspi_data *rspi, struct spi_message *mesg, struct spi_transfer *t) { @@ -389,9 +443,10 @@ static int qspi_send_pio(struct rspi_data *rspi, struct spi_message *mesg, { int remain = t->len; const u8 *data = t->tx_buf; + int i, n; rspi_write8(rspi, SPBFCR_TXRST, QSPI_SPBFCR); - rspi_write8(rspi, 0x00, QSPI_SPBFCR); + rspi_write8(rspi, DUMMY_DATA, QSPI_SPBFCR); while (remain > 0) { @@ -400,16 +455,38 @@ static int qspi_send_pio(struct rspi_data *rspi, struct spi_message *mesg, "%s: tx empty timeout\n", __func__); return -ETIMEDOUT; } - rspi_write8(rspi, *data++, RSPI_SPDR); - - if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, SPCR_SPRIE) < 0) { - dev_err(&rspi->master->dev, - "%s: receive timeout\n", __func__); - return -ETIMEDOUT; + n = qspi_set_send_trigger(rspi, remain); + if (n >= QSPI_BUFFER_SIZE) { + for (i = 0; i < n; i++) + rspi_write8(rspi, *data++, RSPI_SPDR); + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, + SPCR_SPRIE) < 0) { + dev_err(&rspi->master->dev, + "%s: receive timeout\n", __func__); + return -ETIMEDOUT; + } + for (i = 0; i < n; i++) + rspi_read8(rspi, RSPI_SPDR); + } else { + for (i = 0; i < n; i++) { + if (rspi_wait_for_interrupt(rspi, SPSR_SPTEF, + SPCR_SPTIE) < 0) { + dev_err(&rspi->master->dev, + "%s: tx empty timeout\n", __func__); + return -ETIMEDOUT; + } + rspi_write8(rspi, *data++, RSPI_SPDR); + + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, + SPCR_SPRIE) < 0) { + dev_err(&rspi->master->dev, + "%s: receive timeout\n", __func__); + return -ETIMEDOUT; + } + rspi_read8(rspi, RSPI_SPDR); + } } - rspi_read8(rspi, RSPI_SPDR); - - remain--; + remain -= n; } /* Waiting for the last transmission */ @@ -603,6 +680,7 @@ static int qspi_receive_pio(struct rspi_data *rspi, struct spi_message *mesg, { int remain = t->len; u8 *data; + int i, n; qspi_receive_init(rspi); @@ -614,17 +692,43 @@ static int qspi_receive_pio(struct rspi_data *rspi, struct spi_message *mesg, "%s: tx empty timeout\n", __func__); return -ETIMEDOUT; } - /* dummy write for generate clock */ - rspi_write8(rspi, DUMMY_DATA, RSPI_SPDR); - - if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, SPCR_SPRIE) < 0) { - dev_err(&rspi->master->dev, - "%s: receive timeout\n", __func__); - return -ETIMEDOUT; + n = qspi_set_receive_trigger(rspi, remain); + if (n >= QSPI_BUFFER_SIZE) { + /* dummy write for generate clock */ + for (i = 0; i < n; i++) + rspi_write8(rspi, DUMMY_DATA, RSPI_SPDR); + + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, + SPCR_SPRIE) < 0) { + dev_err(&rspi->master->dev, + "%s: receive timeout\n", __func__); + return -ETIMEDOUT; + } + /* SPDR allows 8, 16 or 32-bit access */ + for (i = 0; i < n; i++) + *data++ = rspi_read8(rspi, RSPI_SPDR); + } else { + for (i = 0; i < n; i++) { + if (rspi_wait_for_interrupt(rspi, SPSR_SPTEF, + SPCR_SPTIE) < 0) { + dev_err(&rspi->master->dev, + "%s: tx empty timeout\n", __func__); + return -ETIMEDOUT; + } + /* dummy write for generate clock */ + rspi_write8(rspi, DUMMY_DATA, RSPI_SPDR); + + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, + SPCR_SPRIE) < 0) { + dev_err(&rspi->master->dev, + "%s: receive timeout\n", __func__); + return -ETIMEDOUT; + } + /* SPDR allows 8, 16 or 32-bit access */ + *data++ = rspi_read8(rspi, RSPI_SPDR); + } } - /* SPDR allows 8, 16 or 32-bit access */ - *data++ = rspi_read8(rspi, RSPI_SPDR); - remain--; + remain -= n; } return 0; -- 1.7.9.5 ^ permalink raw reply related [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 8:29 ` Cao Minh Hiep @ 2014-03-13 9:23 ` Geert Uytterhoeven 2014-03-17 1:11 ` カオ ミン ヒェップ 2014-03-19 1:17 ` Kuninori Morimoto 2014-03-13 9:27 ` Kuninori Morimoto 1 sibling, 2 replies; 14+ messages in thread From: Geert Uytterhoeven @ 2014-03-13 9:23 UTC (permalink / raw) To: Cao Minh Hiep Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> wrote: > In order to transmit and receive data when have 32 bytes of data that > ready has prepared on Transmit/Receive Buffer to transmit or receive. > Instead transmits/receives a byte data using Transmit/Receive Buffer > Data Triggering Number will improve the speed of transfer data. Thanks for your patch, this looks like a valuable improvement! Unfortunately your patch doesn't apply against spi/for-next anymore, as the driver has changed a lot lately, due to the addition of RZ/A1H and Quad/Dual support. > +static void qspi_update(const struct rspi_data *rspi, u8 mask, u8 val, u8 reg) > +{ > + u8 data; > + > + data = rspi_read8(rspi, reg); > + data &= ~mask; > + data |= (val & mask); I think this "& mask" is superfluous. > + rspi_write8(rspi, data, reg); > +} > +static int qspi_set_send_trigger(struct rspi_data *rspi, int remain) > +{ > + int n; > + > + n = min(remain, QSPI_BUFFER_SIZE); > + > + if (remain >= QSPI_BUFFER_SIZE) { > + /* sets triggering number to 32 bytes */ "0 bytes"? (comment copied from qspi_set_receive_trigger()) > + qspi_update(rspi, SPBFCR_TXTRG_MASK, > + SPBFCR_TXTRG_0B, QSPI_SPBFCR); > + } else { > + /* sets triggering number to 1 byte */ "31 bytes"? > + qspi_update(rspi, SPBFCR_TXTRG_MASK, > + SPBFCR_TXTRG_31B, QSPI_SPBFCR); > + } > @@ -389,9 +443,10 @@ static int qspi_send_pio(struct rspi_data *rspi, struct spi_message *mesg, > { > int remain = t->len; > const u8 *data = t->tx_buf; > + int i, n; > > rspi_write8(rspi, SPBFCR_TXRST, QSPI_SPBFCR); > - rspi_write8(rspi, 0x00, QSPI_SPBFCR); > + rspi_write8(rspi, DUMMY_DATA, QSPI_SPBFCR); I think this should stay "0x00", as it's not dummy data written to the transmit data register. Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 9:23 ` Geert Uytterhoeven @ 2014-03-17 1:11 ` カオ ミン ヒェップ 2014-03-17 7:54 ` Geert Uytterhoeven 2014-03-19 1:17 ` Kuninori Morimoto 1 sibling, 1 reply; 14+ messages in thread From: カオ ミン ヒェップ @ 2014-03-17 1:11 UTC (permalink / raw) To: Geert Uytterhoeven Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven On 03/13/2014 06:23 PM, Geert Uytterhoeven wrote: > On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> wrote: >> In order to transmit and receive data when have 32 bytes of data that >> ready has prepared on Transmit/Receive Buffer to transmit or receive. >> Instead transmits/receives a byte data using Transmit/Receive Buffer >> Data Triggering Number will improve the speed of transfer data. > Thanks for your patch, this looks like a valuable improvement! > > Unfortunately your patch doesn't apply against spi/for-next anymore, as > the driver has changed a lot lately, due to the addition of RZ/A1H and > Quad/Dual support. > Thanks, Do you mean this patch did not need for next and I don't mind about it? >> +static void qspi_update(const struct rspi_data *rspi, u8 mask, u8 val, u8 reg) >> +{ >> + u8 data; >> + >> + data = rspi_read8(rspi, reg); >> + data &= ~mask; >> + data |= (val & mask); > I think this "& mask" is superfluous. thanks, I modified it. >> + rspi_write8(rspi, data, reg); >> +} >> +static int qspi_set_send_trigger(struct rspi_data *rspi, int remain) >> +{ >> + int n; >> + >> + n = min(remain, QSPI_BUFFER_SIZE); >> + >> + if (remain >= QSPI_BUFFER_SIZE) { >> + /* sets triggering number to 32 bytes */ > "0 bytes"? (comment copied from qspi_set_receive_trigger()) > >> + qspi_update(rspi, SPBFCR_TXTRG_MASK, >> + SPBFCR_TXTRG_0B, QSPI_SPBFCR); >> + } else { >> + /* sets triggering number to 1 byte */ > "31 bytes"? thanks, >> + qspi_update(rspi, SPBFCR_TXTRG_MASK, >> + SPBFCR_TXTRG_31B, QSPI_SPBFCR); >> + } >> @@ -389,9 +443,10 @@ static int qspi_send_pio(struct rspi_data *rspi, struct spi_message *mesg, >> { >> int remain = t->len; >> const u8 *data = t->tx_buf; >> + int i, n; >> >> rspi_write8(rspi, SPBFCR_TXRST, QSPI_SPBFCR); >> - rspi_write8(rspi, 0x00, QSPI_SPBFCR); >> + rspi_write8(rspi, DUMMY_DATA, QSPI_SPBFCR); > I think this should stay "0x00", as it's not dummy data written to the > transmit data register. Thanks, > Gr{oetje,eeting}s, > > Geert > > -- > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org > > In personal conversations with technical people, I call myself a hacker. But > when I'm talking to journalists I just say "programmer" or something like that. > -- Linus Torvalds > > ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-17 1:11 ` カオ ミン ヒェップ @ 2014-03-17 7:54 ` Geert Uytterhoeven 2014-03-17 12:01 ` カオ ミン ヒェップ 0 siblings, 1 reply; 14+ messages in thread From: Geert Uytterhoeven @ 2014-03-17 7:54 UTC (permalink / raw) To: Cao Minh Hiep Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven Hi Hiep, On Mon, Mar 17, 2014 at 2:11 AM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: > On 03/13/2014 06:23 PM, Geert Uytterhoeven wrote: >> On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> >> wrote: >>> In order to transmit and receive data when have 32 bytes of data that >>> ready has prepared on Transmit/Receive Buffer to transmit or receive. >>> Instead transmits/receives a byte data using Transmit/Receive Buffer >>> Data Triggering Number will improve the speed of transfer data. >> >> Thanks for your patch, this looks like a valuable improvement! >> >> Unfortunately your patch doesn't apply against spi/for-next anymore, as >> the driver has changed a lot lately, due to the addition of RZ/A1H and >> Quad/Dual support. >> > Thanks, Do you mean this patch did not need for next and I don't mind about > it? It means you will have to rework your patch to apply on top of https://git.kernel.org/cgit/linux/kernel/git/broonie/spi.git/tree/drivers/spi?h=for-next before the SPI maintainer can accept it. Thanks! Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-17 7:54 ` Geert Uytterhoeven @ 2014-03-17 12:01 ` カオ ミン ヒェップ 2014-03-17 19:48 ` Geert Uytterhoeven 0 siblings, 1 reply; 14+ messages in thread From: カオ ミン ヒェップ @ 2014-03-17 12:01 UTC (permalink / raw) To: Geert Uytterhoeven Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven Hi Geert On 03/17/2014 04:54 PM, Geert Uytterhoeven wrote: > Hi Hiep, > > On Mon, Mar 17, 2014 at 2:11 AM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: >> On 03/13/2014 06:23 PM, Geert Uytterhoeven wrote: >>> On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> >>> wrote: >>>> In order to transmit and receive data when have 32 bytes of data that >>>> ready has prepared on Transmit/Receive Buffer to transmit or receive. >>>> Instead transmits/receives a byte data using Transmit/Receive Buffer >>>> Data Triggering Number will improve the speed of transfer data. >>> Thanks for your patch, this looks like a valuable improvement! >>> >>> Unfortunately your patch doesn't apply against spi/for-next anymore, as >>> the driver has changed a lot lately, due to the addition of RZ/A1H and >>> Quad/Dual support. >>> >> Thanks, Do you mean this patch did not need for next and I don't mind about >> it? > It means you will have to rework your patch to apply on top of > https://git.kernel.org/cgit/linux/kernel/git/broonie/spi.git/tree/drivers/spi?h=for-next > before the SPI maintainer can accept it. Thanks, I understood, I will rework my patch and report to your later. Cao Minh Hiep. ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-17 12:01 ` カオ ミン ヒェップ @ 2014-03-17 19:48 ` Geert Uytterhoeven 2014-03-17 20:32 ` Mark Brown 2014-03-19 0:50 ` カオ ミン ヒェップ 0 siblings, 2 replies; 14+ messages in thread From: Geert Uytterhoeven @ 2014-03-17 19:48 UTC (permalink / raw) To: Cao Minh Hiep Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven Hi Hiep, On Mon, Mar 17, 2014 at 1:01 PM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: > On 03/17/2014 04:54 PM, Geert Uytterhoeven wrote: >> On Mon, Mar 17, 2014 at 2:11 AM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: >>> On 03/13/2014 06:23 PM, Geert Uytterhoeven wrote: >>>> On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> >>>> wrote: >>>>> In order to transmit and receive data when have 32 bytes of data that >>>>> ready has prepared on Transmit/Receive Buffer to transmit or receive. >>>>> Instead transmits/receives a byte data using Transmit/Receive Buffer >>>>> Data Triggering Number will improve the speed of transfer data. >>>> >>>> Thanks for your patch, this looks like a valuable improvement! >>>> >>>> Unfortunately your patch doesn't apply against spi/for-next anymore, as >>>> the driver has changed a lot lately, due to the addition of RZ/A1H and >>>> Quad/Dual support. >>>> >>> Thanks, Do you mean this patch did not need for next and I don't mind >>> about >>> it? >> >> It means you will have to rework your patch to apply on top of >> >> https://git.kernel.org/cgit/linux/kernel/git/broonie/spi.git/tree/drivers/spi?h=for-next >> before the SPI maintainer can accept it. > > Thanks, I understood, I will rework my patch and report to your later. I gave your patch a try on a v3.14-rc5 kernel on r8a7791/Koelsch, and it increased Single SPI read transfer speed from ca. 1 Mbps to 11 Mbps. So I'm looking forward to an updated version. Thanks! Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-17 19:48 ` Geert Uytterhoeven @ 2014-03-17 20:32 ` Mark Brown 2014-03-19 0:50 ` カオ ミン ヒェップ 1 sibling, 0 replies; 14+ messages in thread From: Mark Brown @ 2014-03-17 20:32 UTC (permalink / raw) To: Geert Uytterhoeven Cc: Cao Minh Hiep, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven [-- Attachment #1: Type: text/plain, Size: 270 bytes --] On Mon, Mar 17, 2014 at 08:48:43PM +0100, Geert Uytterhoeven wrote: > I gave your patch a try on a v3.14-rc5 kernel on r8a7791/Koelsch, > and it increased Single SPI read transfer speed from ca. 1 Mbps to 11 Mbps. > So I'm looking forward to an updated version. Nice! [-- Attachment #2: Digital signature --] [-- Type: application/pgp-signature, Size: 836 bytes --] ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-17 19:48 ` Geert Uytterhoeven 2014-03-17 20:32 ` Mark Brown @ 2014-03-19 0:50 ` カオ ミン ヒェップ 1 sibling, 0 replies; 14+ messages in thread From: カオ ミン ヒェップ @ 2014-03-19 0:50 UTC (permalink / raw) To: Geert Uytterhoeven Cc: Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven Hi Geert On 03/18/2014 04:48 AM, Geert Uytterhoeven wrote: > Hi Hiep, > > On Mon, Mar 17, 2014 at 1:01 PM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: >> On 03/17/2014 04:54 PM, Geert Uytterhoeven wrote: >>> On Mon, Mar 17, 2014 at 2:11 AM, カオ ミン ヒェップ <cm-hiep@jinso.co.jp> wrote: >>>> On 03/13/2014 06:23 PM, Geert Uytterhoeven wrote: >>>>> On Thu, Mar 13, 2014 at 9:29 AM, Cao Minh Hiep <cm-hiep@jinso.co.jp> >>>>> wrote: >>>>>> In order to transmit and receive data when have 32 bytes of data that >>>>>> ready has prepared on Transmit/Receive Buffer to transmit or receive. >>>>>> Instead transmits/receives a byte data using Transmit/Receive Buffer >>>>>> Data Triggering Number will improve the speed of transfer data. >>>>> Thanks for your patch, this looks like a valuable improvement! >>>>> >>>>> Unfortunately your patch doesn't apply against spi/for-next anymore, as >>>>> the driver has changed a lot lately, due to the addition of RZ/A1H and >>>>> Quad/Dual support. >>>>> >>>> Thanks, Do you mean this patch did not need for next and I don't mind >>>> about >>>> it? >>> It means you will have to rework your patch to apply on top of >>> >>> https://git.kernel.org/cgit/linux/kernel/git/broonie/spi.git/tree/drivers/spi?h=for-next >>> before the SPI maintainer can accept it. >> Thanks, I understood, I will rework my patch and report to your later. > I gave your patch a try on a v3.14-rc5 kernel on r8a7791/Koelsch, > and it increased Single SPI read transfer speed from ca. 1 Mbps to 11 Mbps. > So I'm looking forward to an updated version. Thanks, I am modifying my patch and will send it to you as soon as possible. Hiep ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 9:23 ` Geert Uytterhoeven 2014-03-17 1:11 ` カオ ミン ヒェップ @ 2014-03-19 1:17 ` Kuninori Morimoto 1 sibling, 0 replies; 14+ messages in thread From: Kuninori Morimoto @ 2014-03-19 1:17 UTC (permalink / raw) To: Geert Uytterhoeven Cc: Cao Minh Hiep, Mark Brown, devicetree@vger.kernel.org, Linux-sh list, Kuninori Morimoto, Geert Uytterhoeven Hi Geert, Hiep > > +static void qspi_update(const struct rspi_data *rspi, u8 mask, u8 val, u8 reg) > > +{ > > + u8 data; > > + > > + data = rspi_read8(rspi, reg); > > + data &= ~mask; > > + data |= (val & mask); > > I think this "& mask" is superfluous. This "& mask" protects data from un-anticipative overwrite IMO. Other driver is using same style. ex) snd_soc_update_bits() ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 8:29 ` Cao Minh Hiep 2014-03-13 9:23 ` Geert Uytterhoeven @ 2014-03-13 9:27 ` Kuninori Morimoto 2014-03-17 1:15 ` カオ ミン ヒェップ 1 sibling, 1 reply; 14+ messages in thread From: Kuninori Morimoto @ 2014-03-13 9:27 UTC (permalink / raw) To: Cao Minh Hiep, broonie; +Cc: devicetree, linux-sh, geert+renesas Hi Hiep > In order to transmit and receive data when have 32 bytes of data that > ready has prepared on Transmit/Receive Buffer to transmit or receive. > Instead transmits/receives a byte data using Transmit/Receive Buffer > Data Triggering Number will improve the speed of transfer data. > > Signed-off-by: Hiep Cao Minh <cm-hiep@jinso.co.jp> > --- (snip) > +static int qspi_set_send_trigger(struct rspi_data *rspi, int remain) > +{ > + int n; > + > + n = min(remain, QSPI_BUFFER_SIZE); > + > + if (remain >= QSPI_BUFFER_SIZE) { > + /* sets triggering number to 32 bytes */ > + qspi_update(rspi, SPBFCR_TXTRG_MASK, > + SPBFCR_TXTRG_0B, QSPI_SPBFCR); > + } else { > + /* sets triggering number to 1 byte */ > + qspi_update(rspi, SPBFCR_TXTRG_MASK, > + SPBFCR_TXTRG_31B, QSPI_SPBFCR); > + } > + > + return n; > +} (snip) > + n = qspi_set_send_trigger(rspi, remain); > + if (n >= QSPI_BUFFER_SIZE) { > + for (i = 0; i < n; i++) > + rspi_write8(rspi, *data++, RSPI_SPDR); > + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, > + SPCR_SPRIE) < 0) { > + dev_err(&rspi->master->dev, > + "%s: receive timeout\n", __func__); > + return -ETIMEDOUT; > + } > + for (i = 0; i < n; i++) > + rspi_read8(rspi, RSPI_SPDR); > + } else { > + for (i = 0; i < n; i++) { > + if (rspi_wait_for_interrupt(rspi, SPSR_SPTEF, > + SPCR_SPTIE) < 0) { > + dev_err(&rspi->master->dev, > + "%s: tx empty timeout\n", __func__); > + return -ETIMEDOUT; > + } > + rspi_write8(rspi, *data++, RSPI_SPDR); > + > + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, > + SPCR_SPRIE) < 0) { > + dev_err(&rspi->master->dev, > + "%s: receive timeout\n", __func__); > + return -ETIMEDOUT; > + } > + rspi_read8(rspi, RSPI_SPDR); > + } I guess, this "qspi_set_send_trigger()" is mode "settings", and, "transfer" is done here. I think "setting" and "tranfer" in 1 function is readable. ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 9:27 ` Kuninori Morimoto @ 2014-03-17 1:15 ` カオ ミン ヒェップ 0 siblings, 0 replies; 14+ messages in thread From: カオ ミン ヒェップ @ 2014-03-17 1:15 UTC (permalink / raw) To: Kuninori Morimoto, broonie; +Cc: devicetree, linux-sh, geert+renesas Hi Morimoto On 03/13/2014 06:27 PM, Kuninori Morimoto wrote: > Hi Hiep > >> In order to transmit and receive data when have 32 bytes of data that >> ready has prepared on Transmit/Receive Buffer to transmit or receive. >> Instead transmits/receives a byte data using Transmit/Receive Buffer >> Data Triggering Number will improve the speed of transfer data. >> >> Signed-off-by: Hiep Cao Minh <cm-hiep@jinso.co.jp> >> --- > (snip) >> +static int qspi_set_send_trigger(struct rspi_data *rspi, int remain) >> +{ >> + int n; >> + >> + n = min(remain, QSPI_BUFFER_SIZE); >> + >> + if (remain >= QSPI_BUFFER_SIZE) { >> + /* sets triggering number to 32 bytes */ >> + qspi_update(rspi, SPBFCR_TXTRG_MASK, >> + SPBFCR_TXTRG_0B, QSPI_SPBFCR); >> + } else { >> + /* sets triggering number to 1 byte */ >> + qspi_update(rspi, SPBFCR_TXTRG_MASK, >> + SPBFCR_TXTRG_31B, QSPI_SPBFCR); >> + } >> + >> + return n; >> +} > (snip) >> + n = qspi_set_send_trigger(rspi, remain); >> + if (n >= QSPI_BUFFER_SIZE) { >> + for (i = 0; i < n; i++) >> + rspi_write8(rspi, *data++, RSPI_SPDR); >> + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, >> + SPCR_SPRIE) < 0) { >> + dev_err(&rspi->master->dev, >> + "%s: receive timeout\n", __func__); >> + return -ETIMEDOUT; >> + } >> + for (i = 0; i < n; i++) >> + rspi_read8(rspi, RSPI_SPDR); >> + } else { >> + for (i = 0; i < n; i++) { >> + if (rspi_wait_for_interrupt(rspi, SPSR_SPTEF, >> + SPCR_SPTIE) < 0) { >> + dev_err(&rspi->master->dev, >> + "%s: tx empty timeout\n", __func__); >> + return -ETIMEDOUT; >> + } >> + rspi_write8(rspi, *data++, RSPI_SPDR); >> + >> + if (rspi_wait_for_interrupt(rspi, SPSR_SPRF, >> + SPCR_SPRIE) < 0) { >> + dev_err(&rspi->master->dev, >> + "%s: receive timeout\n", __func__); >> + return -ETIMEDOUT; >> + } >> + rspi_read8(rspi, RSPI_SPDR); >> + } > I guess, this "qspi_set_send_trigger()" is mode "settings", > and, "transfer" is done here. > > I think "setting" and "tranfer" in 1 function is readable. > Thanks for your comments, I will do so. ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 8:29 [RFC] spi: Using Trigger number to transmit/receive data Cao Minh Hiep 2014-03-13 8:29 ` Cao Minh Hiep @ 2014-03-13 14:17 ` Mark Brown 2014-03-17 1:16 ` カオ ミン ヒェップ 1 sibling, 1 reply; 14+ messages in thread From: Mark Brown @ 2014-03-13 14:17 UTC (permalink / raw) To: Cao Minh Hiep; +Cc: devicetree, linux-sh, kuninori.morimoto.gx, geert+renesas [-- Attachment #1: Type: text/plain, Size: 711 bytes --] On Thu, Mar 13, 2014 at 05:29:23PM +0900, Cao Minh Hiep wrote: > From: Hiep Cao Minh <cm-hiep@jinso.co.jp> > > Hi, > > This is the patch of QSPI that support for r8a7790 SOC. The pupose of > this patch is to transmit and receive the 32 bytes of data when the data > already has prepared on Transmit/Receive Buffer. Using Transmit/Receive > Buffer Data Trigger Number to do that instead transmits/receives a byte > of data. With this patch will improve the speed of transfer data. Please don't send cover letters for single patches, if there's anything that needs to be said put it in the changelog or after the --- in the patch. That way important information won't get lost from history. [-- Attachment #2: Digital signature --] [-- Type: application/pgp-signature, Size: 836 bytes --] ^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [RFC] spi: Using Trigger number to transmit/receive data 2014-03-13 14:17 ` Mark Brown @ 2014-03-17 1:16 ` カオ ミン ヒェップ 0 siblings, 0 replies; 14+ messages in thread From: カオ ミン ヒェップ @ 2014-03-17 1:16 UTC (permalink / raw) To: Mark Brown; +Cc: devicetree, linux-sh, kuninori.morimoto.gx, geert+renesas On 03/13/2014 11:17 PM, Mark Brown wrote: > On Thu, Mar 13, 2014 at 05:29:23PM +0900, Cao Minh Hiep wrote: >> From: Hiep Cao Minh <cm-hiep@jinso.co.jp> >> >> Hi, >> >> This is the patch of QSPI that support for r8a7790 SOC. The pupose of >> this patch is to transmit and receive the 32 bytes of data when the data >> already has prepared on Transmit/Receive Buffer. Using Transmit/Receive >> Buffer Data Trigger Number to do that instead transmits/receives a byte >> of data. With this patch will improve the speed of transfer data. > Please don't send cover letters for single patches, if there's anything > that needs to be said put it in the changelog or after the --- in the > patch. That way important information won't get lost from history. Thanks, I will do so in next time. ^ permalink raw reply [flat|nested] 14+ messages in thread
end of thread, other threads:[~2014-03-19 1:17 UTC | newest] Thread overview: 14+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2014-03-13 8:29 [RFC] spi: Using Trigger number to transmit/receive data Cao Minh Hiep 2014-03-13 8:29 ` Cao Minh Hiep 2014-03-13 9:23 ` Geert Uytterhoeven 2014-03-17 1:11 ` カオ ミン ヒェップ 2014-03-17 7:54 ` Geert Uytterhoeven 2014-03-17 12:01 ` カオ ミン ヒェップ 2014-03-17 19:48 ` Geert Uytterhoeven 2014-03-17 20:32 ` Mark Brown 2014-03-19 0:50 ` カオ ミン ヒェップ 2014-03-19 1:17 ` Kuninori Morimoto 2014-03-13 9:27 ` Kuninori Morimoto 2014-03-17 1:15 ` カオ ミン ヒェップ 2014-03-13 14:17 ` Mark Brown 2014-03-17 1:16 ` カオ ミン ヒェップ
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).