The GP timer is fixed for 1msec and CONFIG_SYS_HZ to 1000.
PATCH 1/1 The GP timer is fixed for 1msec and CONFIG_SYS_HZ to 1000.
The header files for other OAMP3 boards have also been changed for CONFIG_SYS_HZ to be 1000.
--- cpu/arm_cortexa8/omap3/interrupts.c | 12 +++--------- include/configs/omap3_beagle.h | 2 +- include/configs/omap3_evm.h | 2 +- include/configs/omap3_overo.h | 2 +- include/configs/omap3_pandora.h | 2 +- 5 files changed, 7 insertions(+), 13 deletions(-)
cpu/arm_cortexa8/omap3/interrupts.c +3 -9
@@ -234,19 +234,13 @@ void reset_timer_masked(void)/* reset time, capture current incrementer value time */ lastinc =readl (&timer_base->tcrr); timestamp =0 ;/* start "advancing" time stamp from 0 */ +/* reset the timer count */ +writel (0x1 , &timer_base->ttgr); }ulong get_timer_masked (void ) { -ulong now =readl (&timer_base->tcrr);/* current tick value */ - -if (now >= lastinc)/* normal mode (non roll) */ -/* move stamp fordward with absoulte diff ticks */ - timestamp += (now - lastinc); -else /* we have rollover of incrementer */ - timestamp += (0xFFFFFFFF - lastinc) + now; - lastinc = now; -return timestamp; +return readl (&timer_base->tcrr) >>5 ;/* current tick value */ }/* waits specified delay value and resets timestamp */
include/configs/omap3_beagle.h +1 -1
@@ -227,7 +227,7 @@#define CONFIG_SYS_TIMERBASE (OMAP34XX_GPT2)#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000/*----------------------------------------------------------------------- * Stack sizes
include/configs/omap3_evm.h +1 -1
@@ -226,7 +226,7 @@#define CONFIG_SYS_TIMERBASE OMAP34XX_GPT2#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000/*----------------------------------------------------------------------- * Stack sizes
include/configs/omap3_overo.h +1 -1
@@ -220,7 +220,7 @@#define CONFIG_SYS_TIMERBASE (OMAP34XX_GPT2)#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000/*----------------------------------------------------------------------- * Stack sizes
include/configs/omap3_pandora.h +1 -1
@@ -222,7 +222,7 @@#define CONFIG_SYS_TIMERBASE (OMAP34XX_GPT2)#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000/*----------------------------------------------------------------------- * Stack sizes
Signature · 1 line
1.5.6
PATCH 1/1 The GP timer is fixed for 1msec and CONFIG_SYS_HZ to 1000.
Dear Mani,
some comments below.
I will have a look to this soon. Maybe I will come with a patch then, too.
Quoted 10 lines · Manikandan Pillai
The header files for other OAMP3 boards have also been changed for
CONFIG_SYS_HZto be 1000.Signed-off-by Manikandan Pillai <[email protected]>--- cpu/arm_cortexa8/omap3/interrupts.c | 12 +++--------- include/configs/omap3_beagle.h | 2 +- include/configs/omap3_evm.h | 2 +- include/configs/omap3_overo.h | 2 +- include/configs/omap3_pandora.h | 2 +-
zoom1 missing?
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 238
Quoted 12 lines
5 files changed, 7 insertions(+), 13 deletions(-)cpu/arm_cortexa8/omap3/interrupts.c +2
-0@@ -234,19 +234,13 @@ void reset_timer_masked(void)/* reset time, capture current incrementer value time */ lastinc =readl (&timer_base->tcrr); timestamp =0 ;/* start "advancing" time stamp from 0 */ +/* reset the timer count */ +writel (0x1 , &timer_base->ttgr);
Not sure if this is necessary (correct?) here.
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 249
Quoted 11 lines
cpu/arm_cortexa8/omap3/interrupts.c +0
-9ulong get_timer_masked (void ) { -ulong now =readl (&timer_base->tcrr);/* current tick value */ - -if (now >= lastinc)/* normal mode (non roll) */ -/* move stamp fordward with absoulte diff ticks */ - timestamp += (now - lastinc); -else /* we have rollover of incrementer */ - timestamp += (0xFFFFFFFF - lastinc) + now; - lastinc = now; -return timestamp;
As already mentioned in
http://lists.denx.de/pipermail/u-boot/2009-March/048842.html
I don't understand why you remove the overflow handling here?
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 243
Quoted 1 line
cpu/arm_cortexa8/omap3/interrupts.c +1
-0+return readl (&timer_base->tcrr) >>5 ;/* current tick value */
Hmm, here you shift by 5. In your first patch (linked above) you had 6.
Let's have a look to the clocking used here: If I'm right, the timer runs with 13MHz. Then we divide by 256 (2^(7+1)), this results in 50781.25Hz. Is this correct?
With '>> 6' this will result in 793.457Hz, with '>> 5' we will get 1586.914Hz. Correct?
Both doesn't fit
CONFIG_SYS_HZ 1000
?
At the moment I think the best way to go is to have a look how DaVinci is dealing with this. The only difference I see at the moment between OMAP3 and DaVinci concept is that DaVinci timer runs with 27MHz.
include/configs/omap3_beagle.h line 230
Quoted 11 lines
cpu/arm_cortexa8/omap3/interrupts.c +0
-0/* waits specified delay value and resets timestamp */ include/configs/omap3_beagle.h +1
-1@@ -227,7 +227,7 @@#define CONFIG_SYS_TIMERBASE (OMAP34XX_GPT2)#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000
Looking at this in the config files and the comments around it, it seems to me that we should do some more clean up here. E.g. update the comment above these macros to OMAP3 and replace V_PVT directly with 7.
Dirk
Pillai
PATCH 1/1 The GP timer is fixed for 1msec and CONFIG_SYS_HZ to 1000.
Pls find my comments inlined.
Quoted 27 lines
-----Original Message-----
From: Dirk Behme [mailto:dirk.behme at googlemail.com]
Sent: Tuesday, March 17, 2009 12:15 AM
To: Pillai, Manikandan
Cc: u-boot at lists.denx.de
Subject: Re: [PATCH 1/1] The GP timer is fixed for 1msec andCONFIG_SYS_HZto 1000.Dear Mani,
some comments below.
I will have a look to this soon. Maybe I will come with a patch then, too.
Quoted 10 lines · Manikandan Pillai
The header files for other OAMP3 boards have also been changed for
CONFIG_SYS_HZto be 1000.Signed-off-by Manikandan Pillai <[email protected]>--- cpu/arm_cortexa8/omap3/interrupts.c | 12 +++--------- include/configs/omap3_beagle.h | 2 +- include/configs/omap3_evm.h | 2 +- include/configs/omap3_overo.h | 2 +- include/configs/omap3_pandora.h | 2 +-
Quoted 1 line
zoom1 missing?
[Pillai, Manikandan] OK. I have tested only on OMAP3 EVM since I do not have any of the other boards.
Quoted 5 lines
Review
Quoted 4 lines
5 files changed, 7 insertions(+), 13 deletions(-)(quoted diff) +0
-0
b/cpu/arm_cortexa8/omap3/interrupts.c
Quoted 11 lines
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 238
Quoted 9 lines
index 9e9817d..d16f7bc 100644
cpu/arm_cortexa8/omap3/interrupts.c +2
-0@@ -234,19 +234,13 @@ void reset_timer_masked(void)/* reset time, capture current incrementer value time */ lastinc =readl (&timer_base->tcrr); timestamp =0 ;/* start "advancing" time stamp from 0 */ +/* reset the timer count */ +writel (0x1 , &timer_base->ttgr);Not sure if this is necessary (correct?) here.
[Pillai, Manikandan] Writing into this register clears the register count. The reset_timer() functions needs to clear the count in this register.
Quoted 15 lines
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 249
Quoted 12 lines
cpu/arm_cortexa8/omap3/interrupts.c +0
-9ulong get_timer_masked (void ) { -ulong now =readl (&timer_base->tcrr);/* current tick value */ - -if (now >= lastinc)/* normal mode (non roll) */ -/* move stamp fordward with absoulte diff ticks */ - timestamp += (now - lastinc); -else /* we have rollover of incrementer */ - timestamp += (0xFFFFFFFF - lastinc) + now; - lastinc = now; -return timestamp;As already mentioned in
Quoted 3 lines
http://lists.denx.de/pipermail/u-boot/2009-March/048842.html
I don't understand why you remove the overflow handling here?
[Pillai, Manikandan] Is overflow handling really required here ???
Quoted 5 lines
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 243
Quoted 2 lines
cpu/arm_cortexa8/omap3/interrupts.c +1
-0+return readl (&timer_base->tcrr) >>5 ;/* current tick value */ Hmm, here you shift by 5. In your first patch (linked above) you had 6.
Quoted 8 lines
Let's have a look to the clocking used here: If I'm right, the timer runs with 13MHz. Then we divide by 256 (2^(7+1)), this results in 50781.25Hz. Is this correct?
With '>> 6' this will result in 793.457Hz, with '>> 5' we will get 1586.914Hz. Correct?
Both doesn't fit
[Pillai, Manikandan] I knew it doesn't give the correct values. This was tested using sleep() function. Since 5 gives a > time than 1 second, I picked up 5 here.
Quoted 26 lines
CONFIG_SYS_HZ1000?
At the moment I think the best way to go is to have a look how DaVinci is dealing with this. The only difference I see at the moment between OMAP3 and DaVinci concept is that DaVinci timer runs with 27MHz.
include/configs/omap3_beagle.h line 230
Quoted 11 lines
cpu/arm_cortexa8/omap3/interrupts.c +0
-0/* waits specified delay value and resets timestamp */ include/configs/omap3_beagle.h +1
-1@@ -227,7 +227,7 @@#define CONFIG_SYS_TIMERBASE (OMAP34XX_GPT2)#define CONFIG_SYS_PVT V_PVT/* 2^(pvt+1) */ -#define CONFIG_SYS_HZ ((V_SCLK) / (2 << CONFIG_SYS_PVT)) +#define CONFIG_SYS_HZ 1000Looking at this in the config files and the comments around it, it seems to me that we should do some more clean up here. E.g. update the comment above these macros to OMAP3 and replace
V_PVTdirectly with 7.Dirk
PATCH 1/1 The GP timer is fixed for 1msec and CONFIG_SYS_HZ to 1000.
Dear "Pillai, Manikandan",
Quoted 13 lines · In message <[email protected]> you…
Quoted 13 lines
cpu/arm_cortexa8/omap3/interrupts.c void reset_timer_masked(void) · line 249
Quoted 10 lines
cpu/arm_cortexa8/omap3/interrupts.c +0
-9-ulong now =readl (&timer_base->tcrr);/* current tick value */ - -if (now >= lastinc)/* normal mode (non roll) */ -/* move stamp fordward with absoulte diff ticks */ - timestamp += (now - lastinc); -else /* we have rollover of incrementer */ - timestamp += (0xFFFFFFFF - lastinc) + now; - lastinc = now; -return timestamp;As already mentioned in
Quoted 4 lines
Quoted 3 lines
http://lists.denx.de/pipermail/u-boot/2009-March/048842.html
I don't understand why you remove the overflow handling here?
[Pillai, Manikandan] Is overflow handling really required here ???
What makes you sure it is NOT needed?
Quoted 6 lines
Quoted 4 lines
With '>> 6' this will result in 793.457Hz, with '>> 5' we will get 1586.914Hz. Correct?
Both doesn't fit
[Pillai, Manikandan] I knew it doesn't give the correct values. This was tested using
sleep()function. Since 5 gives a > time than 1 second, I picked up 5 here.
Argh... We don't need realy high precision, but I think we should attempt to not worse than +/- 2% or so. THis implementation is not acceptable.
Best regards,
Wolfgang Denk
Signature · 6 lines
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de "I can call spirits from the vasty deep." "Why so can I, or so can any man; but will they come when you do call for them?" - Shakespeare, 1 King Henry IV, Act III, Scene I.