mv_ddr: ddr3: fix tRAS timimg parameter
PATCH 1/2 mv_ddr: ddr3: fix tRAS timimg parameter
From: Chris Packham <[email protected]>
Based on the JEDEC standard JESD79-3F. The tRAS timings should include the highest speed bins at a given frequency. This is similar to commit 683c67b ("mv_ddr: ddr3: fix tfaw timimg parameter") where the wrong comparison was used in the initial implementation.
---
drivers/ddr/marvell/a38x/ddr3_training_db.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/ddr/marvell/a38x/ddr3_training_db.c +4 -4
@@ -420,13 +420,13 @@ unsigned int mv_ddr_speed_bin_timing_get(enum mv_ddr_speed_bin index, enum mv_dd result = speed_bin_table_t_rcd_t_rp[index];break ;case SPEED_BIN_TRAS : -if (index <SPEED_BIN_DDR_1066G ) +if (index <=SPEED_BIN_DDR_1066G ) result =37500 ; -else if (index <SPEED_BIN_DDR_1333J ) +else if (index <=SPEED_BIN_DDR_1333J ) result =36000 ; -else if (index <SPEED_BIN_DDR_1600K ) +else if (index <=SPEED_BIN_DDR_1600K ) result =35000 ; -else if (index <SPEED_BIN_DDR_1866M ) +else if (index <=SPEED_BIN_DDR_1866M ) result =34000 ;else result =33000 ;
Signature · 1 line
2.21.0
PATCH 2/2 mv_ddr: ddr3: only use active chip-selects when tuning ODT
From: Chris Packham <[email protected]>
Inactive chip-selects will give invalid values for read_sample so don't consider them when trying to determine the overall min/max read sample.
---
Hi Baruch,
Does this help with your boards? It's not the custom g_odt_config you were after but it might mean that the code gets the right ODT value so you don't have to override it.
drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c +2 -1
@@ -50,6 +50,7 @@ int ddr3_tip_write_additional_odt_setting(u32 dev_num, u32 if_id)int max_phase =MIN_VALUE , current_phase;enum hws_access_type access_type =ACCESS_TYPE_UNICAST ;u32 octets_per_if_num =ddr3_tip_dev_attr_get (dev_num,MV_ATTR_OCTET_PER_INTERFACE ); +unsigned int max_cs =mv_ddr_cs_num_get (); CHECK_STATUS(ddr3_tip_if_write(dev_num, access_type, if_id, DUNIT_ODT_CTRL_REG,
@@ -59,7 +60,7 @@ int ddr3_tip_write_additional_odt_setting(u32 dev_num, u32 if_id) data_read, MASK_ALL_BITS)); val = data_read[if_id]; -for (cs_num =0 ; cs_num <MAX_CS_NUM ; cs_num++) { +for (cs_num =0 ; cs_num < max_cs; cs_num++) { read_sample[cs_num] =GET_RD_SAMPLE_DELAY (val, cs_num);/* find maximum of read_samples */
Signature · 1 line
2.21.0
PATCH 2/2 mv_ddr: ddr3: only use active chip-selects when tuning ODT
Hi Chris,
Quoted 7 lines · Chris Packham
From: Chris Packham <[email protected]>
Inactive chip-selects will give invalid values for
read_sampleso don't consider them when trying to determine the overall min/max read sample.Signed-off-by Chris Packham <[email protected]>
Quoted 8 lines
[https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell/pull/18]Signed-off-by Chris Packham <[email protected]>---
Hi Baruch,Does this help with your boards? It's not the custom
g_odt_configyou were after but it might mean that the code gets the right ODT value so you don't have to override it.
Maybe. The code looks pretty opaque to me.
Ideally we would like to have ODT_CONFIG (offset 0x1494) set to 0x30000 for all Clearfog platforms, since only ODT0 is connected. This value is good for both 1CS and 2CS systems, although the current 0x10000 is also fine for 1CS.
The current 2CS default hard-coded value of 0x120012 is good when both ODT0/ODT1 are connected and both chip-selects are used. 0x30000 is not optimal for these systems.
baruch
Quoted 16 lines
drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c +1
-0@@ -50,6 +50,7 @@ int ddr3_tip_write_additional_odt_setting(u32 dev_num, u32 if_id)int max_phase =MIN_VALUE , current_phase;enum hws_access_type access_type =ACCESS_TYPE_UNICAST ;u32 octets_per_if_num =ddr3_tip_dev_attr_get (dev_num,MV_ATTR_OCTET_PER_INTERFACE ); +unsigned int max_cs =mv_ddr_cs_num_get (); CHECK_STATUS(ddr3_tip_if_write(dev_num, access_type, if_id, DUNIT_ODT_CTRL_REG,
drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c int ddr3_tip_write_additional_odt_setting(u32 dev_num, u32 if_id) · line 66
Quoted 8 lines
drivers/ddr/marvell/a38x/ddr3_training_hw_algo.c +1
-1@@ -59,7 +60,7 @@ int ddr3_tip_write_additional_odt_setting(u32 dev_num, u32 if_id) data_read, MASK_ALL_BITS)); val = data_read[if_id]; -for (cs_num =0 ; cs_num <MAX_CS_NUM ; cs_num++) { +for (cs_num =0 ; cs_num < max_cs; cs_num++) { read_sample[cs_num] =GET_RD_SAMPLE_DELAY (val, cs_num);/* find maximum of read_samples */
--
http://baruch.siach.name/blog/ ~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- baruch at tkos.co.il - tel: +972.52.368.4656, http://www.tkos.co.il -
PATCH 2/2 mv_ddr: ddr3: only use active chip-selects when tuning ODT
In reply to Chris PackhamQuoted 9 lines · Chris Packham
From: Chris Packham <[email protected]>
Inactive chip-selects will give invalid values for
read_sampleso don't consider them when trying to determine the overall min/max read sample.Signed-off-by Chris Packham <[email protected]>[https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell/pull/18]Signed-off-by Chris Packham <[email protected]>
Applied to u-boot-marvell/master.
Thanks,
Stefan
PATCH 1/2 mv_ddr: ddr3: fix tRAS timimg parameter
In reply to Chris PackhamQuoted 11 lines · Chris Packham
From: Chris Packham <[email protected]>
Based on the JEDEC standard JESD79-3F. The tRAS timings should include the highest speed bins at a given frequency. This is similar to commit 683c67b ("
mv_ddr: ddr3: fix tfaw timimg parameter") where the wrong comparison was used in the initial implementation.Signed-off-by Chris Packham <[email protected]>[https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell/pull/15]Signed-off-by Chris Packham <[email protected]>
Applied to u-boot-marvell/master.
Thanks,
Stefan