| Message ID | 20180806222918.12644-8-jmkrzyszt@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [RFC,v2,01/12] mtd: rawnand: ams-delta: Assign mtd->dev.parent, not mtd->owner | expand |
On Tue, 7 Aug 2018 00:29:13 +0200 Janusz Krzysztofik <jmkrzyszt@gmail.com> wrote: > In its current shape, the driver sets data port direction before each > byte read/write operation, even during multi-byte transfers. Since > performance of the driver is completely not acceptable on Amstrad Delta > after it has been converted to GPIO bitbang, try to improve things a > bit by setting the port direction only on first byte of each transfer. > > Resulting performance on Amstrad Delta is still far from acceptable. > > Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> > --- > drivers/mtd/nand/raw/ams-delta.c | 58 ++++++++++++++++++++++++++++++++-------- > 1 file changed, 47 insertions(+), 11 deletions(-) > > diff --git a/drivers/mtd/nand/raw/ams-delta.c b/drivers/mtd/nand/raw/ams-delta.c > index 78996ddf82e0..d02c48c013e8 100644 > --- a/drivers/mtd/nand/raw/ams-delta.c > +++ b/drivers/mtd/nand/raw/ams-delta.c > @@ -69,6 +69,30 @@ static const struct mtd_partition partition_info[] = { > .size = 3 * SZ_256K }, > }; > > +static void ams_delta_write_commit(struct ams_delta_nand *priv) > +{ > + gpiod_set_value(priv->gpiod_nwe, 0); > + ndelay(40); > + gpiod_set_value(priv->gpiod_nwe, 1); > +} > + > +static void ams_delta_write_next_byte(struct mtd_info *mtd, u_char byte) > +{ > + struct nand_chip *this = mtd_to_nand(mtd); > + struct ams_delta_nand *priv = nand_get_controller_data(this); > + struct gpio_descs *data_gpiods = priv->data_gpiods; > + unsigned long bits = byte; > + int i, value_array[data_gpiods->ndescs]; > + > + for (i = 0; i < data_gpiods->ndescs; i++) > + value_array[i] = test_bit(i, &bits); > + > + gpiod_set_raw_array_value(data_gpiods->ndescs, data_gpiods->desc, > + value_array); > + > + ams_delta_write_commit(priv); > +} > + > static void ams_delta_write_byte(struct mtd_info *mtd, u_char byte) > { > struct nand_chip *this = mtd_to_nand(mtd); > @@ -81,12 +105,10 @@ static void ams_delta_write_byte(struct mtd_info *mtd, u_char byte) > gpiod_direction_output_raw(data_gpiods->desc[i], > test_bit(i, &bits)); > > - gpiod_set_value(priv->gpiod_nwe, 0); > - ndelay(40); > - gpiod_set_value(priv->gpiod_nwe, 1); > + ams_delta_write_commit(priv); > } > > -static u_char ams_delta_read_byte(struct mtd_info *mtd) > +static u_char ams_delta_read_next_byte(struct mtd_info *mtd) > { > struct nand_chip *this = mtd_to_nand(mtd); > struct ams_delta_nand *priv = nand_get_controller_data(this); > @@ -94,9 +116,6 @@ static u_char ams_delta_read_byte(struct mtd_info *mtd) > unsigned long bits = 0; > int i, value_array[data_gpiods->ndescs]; > > - for (i = 0; i < data_gpiods->ndescs; i++) > - gpiod_direction_input(data_gpiods->desc[i]); > - > gpiod_set_value(priv->gpiod_nre, 0); > ndelay(40); > > @@ -112,21 +131,38 @@ static u_char ams_delta_read_byte(struct mtd_info *mtd) > return bits; > } > > +static u_char ams_delta_read_byte(struct mtd_info *mtd) > +{ > + struct nand_chip *this = mtd_to_nand(mtd); > + struct ams_delta_nand *priv = nand_get_controller_data(this); > + struct gpio_descs *data_gpiods = priv->data_gpiods; > + int i; > + > + for (i = 0; i < data_gpiods->ndescs; i++) > + gpiod_direction_input(data_gpiods->desc[i]); > + > + return ams_delta_read_next_byte(mtd); > +} > + > static void ams_delta_write_buf(struct mtd_info *mtd, const u_char *buf, > int len) > { > int i; > > - for (i=0; i<len; i++) > - ams_delta_write_byte(mtd, buf[i]); > + if (len > 0) > + ams_delta_write_byte(mtd, buf[0]); > + for (i = 1; i < len; i++) > + ams_delta_write_next_byte(mtd, buf[i]); > } > > static void ams_delta_read_buf(struct mtd_info *mtd, u_char *buf, int len) > { > int i; > > - for (i=0; i<len; i++) > - buf[i] = ams_delta_read_byte(mtd); > + if (len > 0) > + buf[0] = ams_delta_read_byte(mtd); > + for (i = 1; i < len; i++) > + buf[i] = ams_delta_read_next_byte(mtd); > } I'd suggest a slightly different approach where the data pins direction state is stored in the the priv struct and only changed when required. This way you just have to add a test in ams_delta_read/write_byte(). -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Boris, On Tuesday, August 7, 2018 8:57:52 PM CEST Boris Brezillon wrote: > On Tue, 7 Aug 2018 00:29:13 +0200 > Janusz Krzysztofik <jmkrzyszt@gmail.com> wrote: > > > In its current shape, the driver sets data port direction before each > > byte read/write operation, even during multi-byte transfers. Since > > performance of the driver is completely not acceptable on Amstrad Delta > > after it has been converted to GPIO bitbang, try to improve things a > > bit by setting the port direction only on first byte of each transfer. > ... > I'd suggest a slightly different approach where the data pins > direction state is stored in the the priv struct and only changed when > required. This way you just have to add a test in > ams_delta_read/write_byte(). Good idea, I'm going to use it, thanks. Once done, may I also move that one earlier in the series so that it can be applied while our discussion on GPIO bitmap changes still continues? Thanks, Janusz -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Janusz, Janusz Krzysztofik <jmkrzyszt@gmail.com> wrote on Wed, 08 Aug 2018 18:55:35 +0200: > Hi Boris, > > On Tuesday, August 7, 2018 8:57:52 PM CEST Boris Brezillon wrote: > > On Tue, 7 Aug 2018 00:29:13 +0200 > > Janusz Krzysztofik <jmkrzyszt@gmail.com> wrote: > > > > > In its current shape, the driver sets data port direction before each > > > byte read/write operation, even during multi-byte transfers. Since > > > performance of the driver is completely not acceptable on Amstrad Delta > > > after it has been converted to GPIO bitbang, try to improve things a > > > bit by setting the port direction only on first byte of each transfer. > > ... > > I'd suggest a slightly different approach where the data pins > > direction state is stored in the the priv struct and only changed when > > required. This way you just have to add a test in > > ams_delta_read/write_byte(). > > Good idea, I'm going to use it, thanks. > > Once done, may I also move that one earlier in the series so that it can be > applied while our discussion on GPIO bitmap changes still continues? I think I may answer on his behalf: yes! You can move the GPIO bitmap changes at the end of the series, checking that you never break the bisectability. Then I could apply the major changes and let us iterate on the GPIO bitmap stuff only. Thanks, Miquèl -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/mtd/nand/raw/ams-delta.c b/drivers/mtd/nand/raw/ams-delta.c index 78996ddf82e0..d02c48c013e8 100644 --- a/drivers/mtd/nand/raw/ams-delta.c +++ b/drivers/mtd/nand/raw/ams-delta.c @@ -69,6 +69,30 @@ static const struct mtd_partition partition_info[] = { .size = 3 * SZ_256K }, }; +static void ams_delta_write_commit(struct ams_delta_nand *priv) +{ + gpiod_set_value(priv->gpiod_nwe, 0); + ndelay(40); + gpiod_set_value(priv->gpiod_nwe, 1); +} + +static void ams_delta_write_next_byte(struct mtd_info *mtd, u_char byte) +{ + struct nand_chip *this = mtd_to_nand(mtd); + struct ams_delta_nand *priv = nand_get_controller_data(this); + struct gpio_descs *data_gpiods = priv->data_gpiods; + unsigned long bits = byte; + int i, value_array[data_gpiods->ndescs]; + + for (i = 0; i < data_gpiods->ndescs; i++) + value_array[i] = test_bit(i, &bits); + + gpiod_set_raw_array_value(data_gpiods->ndescs, data_gpiods->desc, + value_array); + + ams_delta_write_commit(priv); +} + static void ams_delta_write_byte(struct mtd_info *mtd, u_char byte) { struct nand_chip *this = mtd_to_nand(mtd); @@ -81,12 +105,10 @@ static void ams_delta_write_byte(struct mtd_info *mtd, u_char byte) gpiod_direction_output_raw(data_gpiods->desc[i], test_bit(i, &bits)); - gpiod_set_value(priv->gpiod_nwe, 0); - ndelay(40); - gpiod_set_value(priv->gpiod_nwe, 1); + ams_delta_write_commit(priv); } -static u_char ams_delta_read_byte(struct mtd_info *mtd) +static u_char ams_delta_read_next_byte(struct mtd_info *mtd) { struct nand_chip *this = mtd_to_nand(mtd); struct ams_delta_nand *priv = nand_get_controller_data(this); @@ -94,9 +116,6 @@ static u_char ams_delta_read_byte(struct mtd_info *mtd) unsigned long bits = 0; int i, value_array[data_gpiods->ndescs]; - for (i = 0; i < data_gpiods->ndescs; i++) - gpiod_direction_input(data_gpiods->desc[i]); - gpiod_set_value(priv->gpiod_nre, 0); ndelay(40); @@ -112,21 +131,38 @@ static u_char ams_delta_read_byte(struct mtd_info *mtd) return bits; } +static u_char ams_delta_read_byte(struct mtd_info *mtd) +{ + struct nand_chip *this = mtd_to_nand(mtd); + struct ams_delta_nand *priv = nand_get_controller_data(this); + struct gpio_descs *data_gpiods = priv->data_gpiods; + int i; + + for (i = 0; i < data_gpiods->ndescs; i++) + gpiod_direction_input(data_gpiods->desc[i]); + + return ams_delta_read_next_byte(mtd); +} + static void ams_delta_write_buf(struct mtd_info *mtd, const u_char *buf, int len) { int i; - for (i=0; i<len; i++) - ams_delta_write_byte(mtd, buf[i]); + if (len > 0) + ams_delta_write_byte(mtd, buf[0]); + for (i = 1; i < len; i++) + ams_delta_write_next_byte(mtd, buf[i]); } static void ams_delta_read_buf(struct mtd_info *mtd, u_char *buf, int len) { int i; - for (i=0; i<len; i++) - buf[i] = ams_delta_read_byte(mtd); + if (len > 0) + buf[0] = ams_delta_read_byte(mtd); + for (i = 1; i < len; i++) + buf[i] = ams_delta_read_next_byte(mtd); } /*
In its current shape, the driver sets data port direction before each byte read/write operation, even during multi-byte transfers. Since performance of the driver is completely not acceptable on Amstrad Delta after it has been converted to GPIO bitbang, try to improve things a bit by setting the port direction only on first byte of each transfer. Resulting performance on Amstrad Delta is still far from acceptable. Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> --- drivers/mtd/nand/raw/ams-delta.c | 58 ++++++++++++++++++++++++++++++++-------- 1 file changed, 47 insertions(+), 11 deletions(-)