Introduce omap_has_feature
Subsystems (5)
Tony Lindgren
maintainercover Introduce omap_has_feature
Hi,
Here's a quick experimental patches on testing features with omap_has_feature as discussed based on earlier patches by Nishant Menon.
These still need to be thought through a bit on how we want to implement it.
Regards,
Tony
---
Nishanth Menon (1):
omap2/3: id: fix sparse warning
Tony Lindgren (4):
omap: Implement common omap_has_feature
omap: Replace omap3_has_ macros with omap_has_feature
omap: Remove old omap3_has_ macros
omap: Allow testing for omap type with omap_has_feature
arch/arm/mach-omap2/clock3xxx_data.c | 2 - arch/arm/mach-omap2/id.c | 70 +++++++++++++++++++++++---------- arch/arm/mach-omap2/pm34xx.c | 2 - arch/arm/plat-omap/common.c | 15 +++++++ arch/arm/plat-omap/include/plat/cpu.h | 69 ++++++++++++++++++--------------- 5 files changed, 104 insertions(+), 54 deletions(-)
Signature · 1 line
Signature
Tony Lindgren
maintainerPATCH 1/5 omap2/3: id: fix sparse warning
From: Nishanth Menon <[email protected]>
omap24xx_check_revision, omap3_check_features, omap3_check_revision, omap4_check_revision, omap3_cpuinfo are not used elsewhere, it should be static
Also fixes the following sparse warnings:arch/arm/mach-omap2/id.c:105:13: warning: symbol 'omap24xx_check_revision' was not declared. Should it be static? arch/arm/mach-omap2/id.c:167:13: warning: symbol 'omap3_check_features' was not declared. Should it be static? arch/arm/mach-omap2/id.c:189:13: warning: symbol 'omap3_check_revision' was not declared. Should it be static? arch/arm/mach-omap2/id.c:270:13: warning: symbol 'omap4_check_revision' was not declared. Should it be static? arch/arm/mach-omap2/id.c:300:13: warning: symbol 'omap3_cpuinfo' was not declared. Should it be static?
--- arch/arm/mach-omap2/id.c | 10 +++++----- 1 files changed, 5 insertions(+), 5 deletions(-)
arch/arm/mach-omap2/id.c +5 -5
@@ -112,7 +112,7 @@ void omap_get_die_id(struct omap_die_id *odi) odi->id_3 =read_tap_reg (OMAP_TAP_DIE_ID_3 ); } -void __init omap24xx_check_revision (void ) +static void __init omap24xx_check_revision (void ) {int i, j;u32 idcode, prod_id;
@@ -172,7 +172,7 @@ void __init omap24xx_check_revision(void) omap3_features |= OMAP3_HAS_ ##feat; \ } -void __init omap3_check_features (void ) +static void __init omap3_check_features (void ) {u32 status;
@@ -196,7 +196,7 @@ void __init omap3_check_features(void)*/ } -void __init omap3_check_revision (void ) +static void __init omap3_check_revision (void ) {u32 cpuid, idcode;u16 hawkeye;
@@ -277,7 +277,7 @@ void __init omap3_check_revision(void) } } -void __init omap4_check_revision (void ) +static void __init omap4_check_revision (void ) {u32 idcode;u16 hawkeye;
@@ -307,7 +307,7 @@ void __init omap4_check_revision(void) if (omap3_has_ ##feat()) \ printk(#feat" "); -void __init omap3_cpuinfo (void ) +static void __init omap3_cpuinfo (void ) {u8 rev =GET_OMAP_REVISION ();char cpu_name[16 ], cpu_rev[16 ];
Tony Lindgren
maintainerPATCH 2/5 omap: Implement common omap_has_feature
In reply to Tony LindgrenImplement common omap_has_feature.
Intended to replace omap3_has_ functions
--- arch/arm/mach-omap2/id.c | 32 ++++++++++++++++++++++++++++++++ arch/arm/plat-omap/common.c | 15 +++++++++++++++ arch/arm/plat-omap/include/plat/cpu.h | 3 +++ 3 files changed, 50 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/id.c +32 -0
@@ -29,7 +29,9 @@static struct omap_chip_id omap_chip;static unsigned int omap_revision; +static u32 omap_features; +/* REVISIT: Get rid of omap3_features */ u32 omap3_features;unsigned int omap_rev (void )
@@ -112,6 +114,12 @@ void omap_get_die_id(struct omap_die_id *odi) odi->id_3 =read_tap_reg (OMAP_TAP_DIE_ID_3 ); } +u32 omap2_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap2 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +static void __init omap24xx_check_revision (void ) {int i, j;
@@ -164,6 +172,15 @@ static void __init omap24xx_check_revision(void)if ((omap_rev () >>8 ) &0x0f )pr_info ("ES%x" , (omap_rev () >>12 ) &0xf );pr_info ("\n" ); + + omap_features =0 ; +omap_init_features (omap2_has_feature); +} + +u32 omap3_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap3 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); }#define OMAP3_CHECK_FEATURE (status,feat) \
@@ -194,6 +211,11 @@ static void __init omap3_check_features(void)* TODO: Get additional info (where applicable) * e.g. Size of L2 cache. */ + +/* REVISIT: Get rid of omap3_features */ + omap_features = omap3_features; + +omap_init_features (omap3_has_feature); }static void __init omap3_check_revision (void )
@@ -277,6 +299,12 @@ static void __init omap3_check_revision(void) } } +u32 omap4_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap4 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +static void __init omap4_check_revision (void ) {u32 idcode;
@@ -297,6 +325,10 @@ static void __init omap4_check_revision(void) omap_revision =OMAP4430_REV_ES1_0 ; omap_chip.oc |=CHIP_IS_OMAP4430ES1 ;pr_info ("OMAP%04x %s\n" ,omap_rev () >>16 , rev_name); + + omap_features =0 ; +omap_init_features (omap4_has_feature); +return ; }
arch/arm/plat-omap/common.c +15 -0
@@ -89,6 +89,21 @@ void __init omap_reserve(void)omap_vram_reserve_sdram_lmb (); } +static int (*_omap_check_feature)(u32 feat_mask); + +u32 omap_has_feature (u32 feat_mask) +{ +if (!_omap_check_feature) +return 0 ; + +return _omap_check_feature (feat_mask); +} + +void __init omap_init_features (u32 (*check_feature)(u32 feat)) +{ + _omap_check_feature = check_feature; +} +/* * 32KHz clocksource ... always available, on pretty most chips except * OMAP 730 and 1510. Other timers could be used as clocksources, with
arch/arm/plat-omap/include/plat/cpu.h +3 -0
@@ -49,6 +49,9 @@ struct omap_chip_id {u8 type; }; +u32 omap_has_feature (u32 feat_mask); +void omap_init_features (u32 (*check_feature)(u32 feat)); +#define OMAP_CHIP_INIT (x) { .oc = x }/*
Re: [PATCH 2/5] omap: Implement common omap_has_feature
Tony Lindgren had written, on 07/08/2010 04:37 AM, the following:
Quoted 26 lines
Implement common
omap_has_feature.Intended to replace
omap3_has_functionsSigned-off-by Tony Lindgren <[email protected]>--- arch/arm/mach-omap2/id.c | 32 ++++++++++++++++++++++++++++++++ arch/arm/plat-omap/common.c | 15 +++++++++++++++ arch/arm/plat-omap/include/plat/cpu.h | 3 +++ 3 files changed, 50 insertions(+), 0 deletions(-)arch/arm/mach-omap2/id.c +2
-0@@ -29,7 +29,9 @@static struct omap_chip_id omap_chip;static unsigned int omap_revision; +static u32 omap_features; +/* REVISIT: Get rid of omap3_features */ u32 omap3_features;unsigned int omap_rev (void )
arch/arm/mach-omap2/id.c void omap_get_die_id(struct omap_die_id *odi) · line 122
Quoted 9 lines
arch/arm/mach-omap2/id.c +6
-0@@ -112,6 +114,12 @@ void omap_get_die_id(struct omap_die_id *odi) odi->id_3 =read_tap_reg (OMAP_TAP_DIE_ID_3 ); } +u32 omap2_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap2 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +
I did consider this path initially,
a) Additional functional call overhead here. some of the calls to has_feature() will get called through pretty active paths, we would like it to be minimized to compile time optimized inline function as much as possible.(no reason why this cant me a inline macro in cpu.h?) - the original series received a similar comment:
http://marc.info/?l=linux-omap&m=125018002127428&w=2
Quoted 86 lines
arch/arm/mach-omap2/id.c +24
-0static void __init omap24xx_check_revision (void ) {int i, j;@@ -164,6 +172,15 @@ static void __init omap24xx_check_revision(void)if ((omap_rev () >>8 ) &0x0f )pr_info ("ES%x" , (omap_rev () >>12 ) &0xf );pr_info ("\n" ); + + omap_features =0 ; +omap_init_features (omap2_has_feature); +} + +u32 omap3_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap3 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); }#define OMAP3_CHECK_FEATURE (status,feat) \@@ -194,6 +211,11 @@ static void __init omap3_check_features(void)* TODO: Get additional info (where applicable) * e.g. Size of L2 cache. */ + +/* REVISIT: Get rid of omap3_features */ + omap_features = omap3_features; + +omap_init_features (omap3_has_feature); }static void __init omap3_check_revision (void )@@ -277,6 +299,12 @@ static void __init omap3_check_revision(void) } } +u32 omap4_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap4 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +static void __init omap4_check_revision (void ) {u32 idcode;@@ -297,6 +325,10 @@ static void __init omap4_check_revision(void) omap_revision =OMAP4430_REV_ES1_0 ; omap_chip.oc |=CHIP_IS_OMAP4430ES1 ;pr_info ("OMAP%04x %s\n" ,omap_rev () >>16 , rev_name); + + omap_features =0 ; +omap_init_features (omap4_has_feature); +return ; }arch/arm/plat-omap/common.c +15
-0@@ -89,6 +89,21 @@ void __init omap_reserve(void)omap_vram_reserve_sdram_lmb (); } +static int (*_omap_check_feature)(u32 feat_mask); + +u32 omap_has_feature (u32 feat_mask) +{ +if (!_omap_check_feature) +return 0 ; + +return _omap_check_feature (feat_mask); +} + +void __init omap_init_features (u32 (*check_feature)(u32 feat)) +{ + _omap_check_feature = check_feature; +} +/* * 32KHz clocksource ... always available, on pretty most chips except * OMAP 730 and 1510. Other timers could be used as clocksources, with arch/arm/plat-omap/include/plat/cpu.h +0
-0
arch/arm/plat-omap/include/plat/cpu.h struct omap_chip_id { · line 52
Quoted 4 lines
arch/arm/plat-omap/include/plat/cpu.h +1
-0@@ -49,6 +49,9 @@ struct omap_chip_id {u8 type; }; +u32 omap_has_feature (u32 feat_mask);
the above crib -> it is better as an static inline function instead of explicit function call.
Quoted 6 lines
arch/arm/plat-omap/include/plat/cpu.h +2
-0+void omap_init_features (u32 (*check_feature)(u32 feat)); +#define OMAP_CHIP_INIT (x) { .oc = x }/*
Signature · 2 lines
Regards, Nishanth Menon
Tony Lindgren
maintainerRe: [PATCH 2/5] omap: Implement common omap_has_feature
Quoted 18 lines · Nishanth Menon
Tony Lindgren had written, on 07/08/2010 04:37 AM, the following:
arch/arm/mach-omap2/id.c void omap_get_die_id(struct omap_die_id *odi) · line 121
Quoted 10 lines
arch/arm/mach-omap2/id.c +6
-0@@ -112,6 +114,12 @@ void omap_get_die_id(struct omap_die_id *odi) odi->id_3 =read_tap_reg (OMAP_TAP_DIE_ID_3 ); } +u32 omap2_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap2 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +I did consider this path initially,
a) Additional functional call overhead here. some of the calls tohas_feature()will get called through pretty active paths, we would like it to be minimized to compile time optimized inline function as much as possible.(no reason why this cant me a inline macro incpu.h?) -
Well it should not matter how slow this code is, it should only be used to initialize things during __init.
Any code doing detection after initialization is wrong.
Regards,
Tony
Re: [PATCH 2/5] omap: Implement common omap_has_feature
Tony Lindgren had written, on 07/09/2010 02:08 AM, the following:
Quoted 18 lines · Nishanth Menon
Quoted 16 lines
Tony Lindgren had written, on 07/08/2010 04:37 AM, the following:
arch/arm/mach-omap2/id.c void omap_get_die_id(struct omap_die_id *odi) · line 121
Quoted 9 lines
arch/arm/mach-omap2/id.c +6
-0@@ -112,6 +114,12 @@ void omap_get_die_id(struct omap_die_id *odi) odi->id_3 =read_tap_reg (OMAP_TAP_DIE_ID_3 ); } +u32 omap2_has_feature (u32 feat_mask) +{ +/* REVISIT: Add necessary omap2 feature tests here */ +return ((feat_mask & omap_features) == feat_mask); +} +I did consider this path initially,
a) Additional functional call overhead here. some of the calls tohas_feature()will get called through pretty active paths, we would like it to be minimized to compile time optimized inline function as much as possible.(no reason why this cant me a inline macro incpu.h?) -
Quoted 4 lines
Well it should not matter how slow this code is, it should only be used to initialize things during
__init.Any code doing detection after initialization is wrong.
couple of reasons why this may not make sense:
a) drivers can be modules -e.g. lets say dss driver would like to know if vrfb OR tiler is used -> __init wont make sense
b) has_feature could be used as part of data flow to make decision what is to be done.
e.g.
lets say dma chaining as a feature OR dma list as a feature, we should allow for usage such as the following:
int chain_dma(...)
{
if (!omap_has_feature(OMAP_HAS_DMA_CHAINING)) {
return -EINVAL;
}
}this means that we would prefer these kind of path to be high performance paths (data paths) - aka inline functions make more sense than incurring a function call overhead.
Signature · 2 lines
Regards, Nishanth Menon
Tony Lindgren
maintainerPATCH 3/5 omap: Replace omap3_has_ macros with omap_has_feature
In reply to Tony LindgrenReplace omap3_has_ macros with omap_has_feature
--- arch/arm/mach-omap2/clock3xxx_data.c | 2 +- arch/arm/mach-omap2/id.c | 22 +++++++++++----------- arch/arm/mach-omap2/pm34xx.c | 2 +- arch/arm/plat-omap/include/plat/cpu.h | 12 ++++++------ 4 files changed, 19 insertions(+), 19 deletions(-)
arch/arm/mach-omap2/clock3xxx_data.c +1 -1
@@ -3435,7 +3435,7 @@ int __init omap3xxx_clk_init(void) cpu_clkflg |=CK_3505 ; } -if (omap3_has_192mhz_clk ()) +if (omap_has_feature (OMAP3_HAS_192MHZ_CLK )) omap_96m_alwon_fck = omap_96m_alwon_fck_3630;if (cpu_is_omap3630 ()) {
arch/arm/mach-omap2/id.c +11 -11
@@ -336,7 +336,7 @@ static void __init omap4_check_revision(void) }#define OMAP3_SHOW_FEATURE (feat) \ - if (omap3_has_ ##feat()) \ + if (omap_has_feature(feat)) \ printk(#feat" ");static void __init omap3_cpuinfo (void )
@@ -356,20 +356,20 @@ static void __init omap3_cpuinfo(void)/* * AM35xx devices */ -if (omap3_has_sgx ()) { +if (omap_has_feature (OMAP3_HAS_SGX )) { omap_revision =OMAP3517_REV (rev);strcpy (cpu_name,"AM3517" ); }else {/* Already set in omap3_check_revision() */ strcpy (cpu_name,"AM3505" ); } - }else if (omap3_has_iva () &&omap3_has_sgx ()) { + }else if (omap_has_feature (OMAP3_HAS_IVA |OMAP3_HAS_SGX )) {/* OMAP3430, OMAP3525, OMAP3515, OMAP3503 devices */ strcpy (cpu_name,"OMAP3430/3530" ); - }else if (omap3_has_iva ()) { + }else if (omap_has_feature (OMAP3_HAS_IVA )) { omap_revision =OMAP3525_REV (rev);strcpy (cpu_name,"OMAP3525" ); - }else if (omap3_has_sgx ()) { + }else if (omap_has_feature (OMAP3_HAS_SGX )) { omap_revision =OMAP3515_REV (rev);strcpy (cpu_name,"OMAP3515" ); }else {
@@ -400,12 +400,12 @@ static void __init omap3_cpuinfo(void)/* Print verbose information */ pr_info ("%s ES%s (" , cpu_name, cpu_rev); -OMAP3_SHOW_FEATURE (l2cache); -OMAP3_SHOW_FEATURE (iva); -OMAP3_SHOW_FEATURE (sgx); -OMAP3_SHOW_FEATURE (neon); -OMAP3_SHOW_FEATURE (isp); - OMAP3_SHOW_FEATURE(192mhz_clk); +OMAP3_SHOW_FEATURE (OMAP3_HAS_L2CACHE ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_IVA ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_SGX ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_NEON ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_ISP ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_192MHZ_CLK );printk (")\n" ); }
arch/arm/mach-omap2/pm34xx.c +1 -1
@@ -385,7 +385,7 @@ void omap_sram_idle(void)/* Enable IO-PAD and IO-CHAIN wakeups */ per_next_state =pwrdm_read_next_pwrst (per_pwrdm); core_next_state =pwrdm_read_next_pwrst (core_pwrdm); -if (omap3_has_io_wakeup () && \ +if (omap_has_feature (OMAP3_HAS_IO_WAKEUP ) && \ (per_next_state <PWRDM_POWER_ON || core_next_state <PWRDM_POWER_ON )) {prm_set_mod_reg_bits (OMAP3430_EN_IO_MASK ,WKUP_MOD ,PM_WKEN );
arch/arm/plat-omap/include/plat/cpu.h +6 -6
@@ -334,14 +334,14 @@ IS_OMAP_TYPE(3517, 0x3517)# undef cpu_is_omap3517# define cpu_is_omap3430 () is_omap3430()# define cpu_is_omap3503 () (cpu_is_omap3430() && \ - (!omap3_has_iva()) && \ - (!omap3_has_sgx())) + (!omap_has_feature(OMAP3_HAS_IVA) && \ + (!omap_has_feature(OMAP3_HAS_SGX)))# define cpu_is_omap3515 () (cpu_is_omap3430() && \ - (!omap3_has_iva()) && \ - (omap3_has_sgx())) + (!omap_has_feature(OMAP3_HAS_IVA)) && \ + (omap_has_feature(OMAP3_HAS_SGX)))# define cpu_is_omap3525 () (cpu_is_omap3430() && \ - (!omap3_has_sgx()) && \ - (omap3_has_iva())) + (!omap_has_feature(OMAP3_HAS_SGX)) && \ + (omap_has_feature(OMAP3_HAS_IVA)))# define cpu_is_omap3530 () (cpu_is_omap3430())# define cpu_is_omap3505 () is_omap3505()# define cpu_is_omap3517 () is_omap3517()
Re: [PATCH 3/5] omap: Replace omap3_has_ macros with omap_has_feature
Tony Lindgren had written, on 07/08/2010 04:37 AM, the following:
Quoted 109 lines
Replace
omap3_has_macros withomap_has_featureSigned-off-by Tony Lindgren <[email protected]>--- arch/arm/mach-omap2/clock3xxx_data.c | 2 +- arch/arm/mach-omap2/id.c | 22 +++++++++++----------- arch/arm/mach-omap2/pm34xx.c | 2 +- arch/arm/plat-omap/include/plat/cpu.h | 12 ++++++------ 4 files changed, 19 insertions(+), 19 deletions(-)arch/arm/mach-omap2/clock3xxx_data.c +1
-1@@ -3435,7 +3435,7 @@ int __init omap3xxx_clk_init(void) cpu_clkflg |=CK_3505 ; } -if (omap3_has_192mhz_clk ()) +if (omap_has_feature (OMAP3_HAS_192MHZ_CLK )) omap_96m_alwon_fck = omap_96m_alwon_fck_3630;if (cpu_is_omap3630 ()) {arch/arm/mach-omap2/id.c +11
-11@@ -336,7 +336,7 @@ static void __init omap4_check_revision(void) }#define OMAP3_SHOW_FEATURE (feat) \ - if (omap3_has_ ##feat()) \ + if (omap_has_feature(feat)) \ printk(#feat" ");static void __init omap3_cpuinfo (void )@@ -356,20 +356,20 @@ static void __init omap3_cpuinfo(void)/* * AM35xx devices */ -if (omap3_has_sgx ()) { +if (omap_has_feature (OMAP3_HAS_SGX )) { omap_revision =OMAP3517_REV (rev);strcpy (cpu_name,"AM3517" ); }else {/* Already set in omap3_check_revision() */ strcpy (cpu_name,"AM3505" ); } - }else if (omap3_has_iva () &&omap3_has_sgx ()) { + }else if (omap_has_feature (OMAP3_HAS_IVA |OMAP3_HAS_SGX )) {/* OMAP3430, OMAP3525, OMAP3515, OMAP3503 devices */ strcpy (cpu_name,"OMAP3430/3530" ); - }else if (omap3_has_iva ()) { + }else if (omap_has_feature (OMAP3_HAS_IVA )) { omap_revision =OMAP3525_REV (rev);strcpy (cpu_name,"OMAP3525" ); - }else if (omap3_has_sgx ()) { + }else if (omap_has_feature (OMAP3_HAS_SGX )) { omap_revision =OMAP3515_REV (rev);strcpy (cpu_name,"OMAP3515" ); }else {@@ -400,12 +400,12 @@ static void __init omap3_cpuinfo(void)/* Print verbose information */ pr_info ("%s ES%s (" , cpu_name, cpu_rev); -OMAP3_SHOW_FEATURE (l2cache); -OMAP3_SHOW_FEATURE (iva); -OMAP3_SHOW_FEATURE (sgx); -OMAP3_SHOW_FEATURE (neon); -OMAP3_SHOW_FEATURE (isp); - OMAP3_SHOW_FEATURE(192mhz_clk); +OMAP3_SHOW_FEATURE (OMAP3_HAS_L2CACHE ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_IVA ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_SGX ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_NEON ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_ISP ); +OMAP3_SHOW_FEATURE (OMAP3_HAS_192MHZ_CLK );printk (")\n" ); }arch/arm/mach-omap2/pm34xx.c +1
-1@@ -385,7 +385,7 @@ void omap_sram_idle(void)/* Enable IO-PAD and IO-CHAIN wakeups */ per_next_state =pwrdm_read_next_pwrst (per_pwrdm); core_next_state =pwrdm_read_next_pwrst (core_pwrdm); -if (omap3_has_io_wakeup () && \ +if (omap_has_feature (OMAP3_HAS_IO_WAKEUP ) && \ (per_next_state <PWRDM_POWER_ON || core_next_state <PWRDM_POWER_ON )) {prm_set_mod_reg_bits (OMAP3430_EN_IO_MASK ,WKUP_MOD ,PM_WKEN );arch/arm/plat-omap/include/plat/cpu.h +2
-4@@ -334,14 +334,14 @@ IS_OMAP_TYPE(3517, 0x3517)# undef cpu_is_omap3517# define cpu_is_omap3430 () is_omap3430()# define cpu_is_omap3503 () (cpu_is_omap3430() && \ - (!omap3_has_iva()) && \ - (!omap3_has_sgx())) + (!omap_has_feature(OMAP3_HAS_IVA) && \ + (!omap_has_feature(OMAP3_HAS_SGX)))# define cpu_is_omap3515 () (cpu_is_omap3430() && \ - (!omap3_has_iva()) && \ - (omap3_has_sgx()))
arch/arm/plat-omap/include/plat/cpu.h IS_OMAP_TYPE(3517, 0x3517) · line 347
Quoted 10 lines
arch/arm/plat-omap/include/plat/cpu.h +4
-2+ (!omap_has_feature(OMAP3_HAS_IVA)) && \ + (omap_has_feature(OMAP3_HAS_SGX)))# define cpu_is_omap3525 () (cpu_is_omap3430() && \ - (!omap3_has_sgx()) && \ - (omap3_has_iva())) + (!omap_has_feature(OMAP3_HAS_SGX)) && \ + (omap_has_feature(OMAP3_HAS_IVA)))# define cpu_is_omap3530 () (cpu_is_omap3430())# define cpu_is_omap3505 () is_omap3505()# define cpu_is_omap3517 () is_omap3517()
Signature · 2 lines
Regards, Nishanth Menon
Tony Lindgren
maintainerPATCH 4/5 omap: Remove old omap3_has_ macros
In reply to Tony LindgrenRemove old omap3_has_ macros. Please use omap_has_feature() instead.
--- arch/arm/mach-omap2/id.c | 14 ++++---------- arch/arm/plat-omap/include/plat/cpu.h | 16 ---------------- 2 files changed, 4 insertions(+), 26 deletions(-)
arch/arm/mach-omap2/id.c +4 -10
@@ -31,9 +31,6 @@ static struct omap_chip_id omap_chip;static unsigned int omap_revision;static u32 omap_features; -/* REVISIT: Get rid of omap3_features */ -u32 omap3_features; -unsigned int omap_rev (void ) {return omap_revision;
@@ -186,14 +183,14 @@ u32 omap3_has_feature(u32 feat_mask)#define OMAP3_CHECK_FEATURE (status,feat) \ if (((status & OMAP3_ ##feat## _MASK) \ >> OMAP3_ ##feat## _SHIFT) != FEAT_ ##feat## _NONE) { \ - omap3_features |= OMAP3_HAS_ ##feat; \ + omap_features |= OMAP3_HAS_ ##feat; \ }static void __init omap3_check_features (void ) {u32 status; - omap3_features =0 ; + omap_features =0 ; status =omap_ctrl_readl (OMAP3_CONTROL_OMAP_STATUS );
@@ -203,18 +200,15 @@ static void __init omap3_check_features(void)OMAP3_CHECK_FEATURE (status,NEON );OMAP3_CHECK_FEATURE (status,ISP );if (cpu_is_omap3630 ()) - omap3_features |=OMAP3_HAS_192MHZ_CLK ; + omap_features |=OMAP3_HAS_192MHZ_CLK ;if (!cpu_is_omap3505 () && !cpu_is_omap3517 ()) - omap3_features |=OMAP3_HAS_IO_WAKEUP ; + omap_features |=OMAP3_HAS_IO_WAKEUP ;/* * TODO: Get additional info (where applicable) * e.g. Size of L2 cache. */ -/* REVISIT: Get rid of omap3_features */ - omap_features = omap3_features; -omap_init_features (omap3_has_feature); }
arch/arm/plat-omap/include/plat/cpu.h +0 -16
@@ -439,8 +439,6 @@ void omap2_check_revision(void);/* * Runtime detection of OMAP3 features */ -extern u32 omap3_features; -#define OMAP3_HAS_L2CACHE BIT(0)#define OMAP3_HAS_IVA BIT(1)#define OMAP3_HAS_SGX BIT(2)
@@ -449,18 +447,4 @@ extern u32 omap3_features;#define OMAP3_HAS_192MHZ_CLK BIT(5)#define OMAP3_HAS_IO_WAKEUP BIT(6) -#define OMAP3_HAS_FEATURE (feat,flag) \ -static inline unsigned int omap3_has_ ##feat(void) \ -{ \ - return (omap3_features & OMAP3_HAS_ ##flag); \ -} \ - -OMAP3_HAS_FEATURE(l2cache, L2CACHE) -OMAP3_HAS_FEATURE(sgx, SGX) -OMAP3_HAS_FEATURE(iva, IVA) -OMAP3_HAS_FEATURE(neon, NEON) -OMAP3_HAS_FEATURE(isp, ISP) -OMAP3_HAS_FEATURE(192mhz_clk, 192MHZ_CLK) -OMAP3_HAS_FEATURE(io_wakeup, IO_WAKEUP) -#endif
Re: [PATCH 4/5] omap: Remove old omap3_has_ macros
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 84 lines
Remove old
omap3_has_macros. Please useomap_has_feature()instead.Signed-off-by Tony Lindgren <[email protected]>--- arch/arm/mach-omap2/id.c | 14 ++++---------- arch/arm/plat-omap/include/plat/cpu.h | 16 ---------------- 2 files changed, 4 insertions(+), 26 deletions(-)arch/arm/mach-omap2/id.c +4
-10@@ -31,9 +31,6 @@ static struct omap_chip_id omap_chip;static unsigned int omap_revision;static u32 omap_features; -/* REVISIT: Get rid of omap3_features */ -u32 omap3_features; -unsigned int omap_rev (void ) {return omap_revision;@@ -186,14 +183,14 @@ u32 omap3_has_feature(u32 feat_mask)#define OMAP3_CHECK_FEATURE (status,feat) \ if (((status & OMAP3_ ##feat## _MASK) \ >> OMAP3_ ##feat## _SHIFT) != FEAT_ ##feat## _NONE) { \ - omap3_features |= OMAP3_HAS_ ##feat; \ + omap_features |= OMAP3_HAS_ ##feat; \ }static void __init omap3_check_features (void ) {u32 status; - omap3_features =0 ; + omap_features =0 ; status =omap_ctrl_readl (OMAP3_CONTROL_OMAP_STATUS );@@ -203,18 +200,15 @@ static void __init omap3_check_features(void)OMAP3_CHECK_FEATURE (status,NEON );OMAP3_CHECK_FEATURE (status,ISP );if (cpu_is_omap3630 ()) - omap3_features |=OMAP3_HAS_192MHZ_CLK ; + omap_features |=OMAP3_HAS_192MHZ_CLK ;if (!cpu_is_omap3505 () && !cpu_is_omap3517 ()) - omap3_features |=OMAP3_HAS_IO_WAKEUP ; + omap_features |=OMAP3_HAS_IO_WAKEUP ;/* * TODO: Get additional info (where applicable) * e.g. Size of L2 cache. */ -/* REVISIT: Get rid of omap3_features */ - omap_features = omap3_features; -omap_init_features (omap3_has_feature); }arch/arm/plat-omap/include/plat/cpu.h +0
-7@@ -439,8 +439,6 @@ void omap2_check_revision(void);/* * Runtime detection of OMAP3 features */ -extern u32 omap3_features; -#define OMAP3_HAS_L2CACHE BIT(0)#define OMAP3_HAS_IVA BIT(1)#define OMAP3_HAS_SGX BIT(2)@@ -449,18 +447,4 @@ extern u32 omap3_features;#define OMAP3_HAS_192MHZ_CLK BIT(5)#define OMAP3_HAS_IO_WAKEUP BIT(6) -#define OMAP3_HAS_FEATURE (feat,flag) \ -static inline unsigned int omap3_has_ ##feat(void) \ -{ \ - return (omap3_features & OMAP3_HAS_ ##flag); \ -} \
arch/arm/plat-omap/include/plat/cpu.h extern u32 omap3_features; · line 450
Quoted 10 lines
arch/arm/plat-omap/include/plat/cpu.h +0
-9- -OMAP3_HAS_FEATURE(l2cache, L2CACHE) -OMAP3_HAS_FEATURE(sgx, SGX) -OMAP3_HAS_FEATURE(iva, IVA) -OMAP3_HAS_FEATURE(neon, NEON) -OMAP3_HAS_FEATURE(isp, ISP) -OMAP3_HAS_FEATURE(192mhz_clk, 192MHZ_CLK) -OMAP3_HAS_FEATURE(io_wakeup, IO_WAKEUP) -#endif
Signature · 2 lines
Regards, Nishanth Menon
Tony Lindgren
maintainerPATCH 5/5 omap: Allow testing for omap type with omap_has_feature
In reply to Tony LindgrenAllow testing for omap type with omap_has_feature. This can be used to leave out cpu_is_omapxxxx checks.
--- arch/arm/plat-omap/include/plat/cpu.h | 38 ++++++++++++++++++++++++++------- 1 files changed, 30 insertions(+), 8 deletions(-)
arch/arm/plat-omap/include/plat/cpu.h +30 -8
@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci);void omap2_check_revision (void );/* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 | OMAP_HAS_IVA +#define OMAP2_HAS_SGX OMAP_FEAT_CLASS_OMAP2 | OMAP_HAS_SGX + +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_L2CACHE +#define OMAP3_HAS_IVA OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_IVA +#define OMAP3_HAS_SGX OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_SGX +#define OMAP3_HAS_NEON OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_NEON +#define OMAP3_HAS_ISP OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_ISP +#define OMAP3_HAS_192MHZ_CLK OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_192MHZ_CLK +#define OMAP3_HAS_IO_WAKEUP OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_IOWAKEUP + +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_L2CACHE +#define OMAP4_HAS_IVA OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_IVA +#define OMAP4_HAS_SGX OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_SGX +#define OMAP4_HAS_NEON OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_NEON +#define OMAP4_HAS_ISP OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_ISP#endif
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 48 lines
Allow testing for omap type with
omap_has_feature. This can be used to leave outcpu_is_omapxxxxchecks.Signed-off-by Tony Lindgren <[email protected]>--- arch/arm/plat-omap/include/plat/cpu.h | 38 ++++++++++++++++++++++++++------- 1 files changed, 30 insertions(+), 8 deletions(-)arch/arm/plat-omap/include/plat/cpu.h +22
-8@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci);void omap2_check_revision (void );/* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 | OMAP_HAS_IVA +#define OMAP2_HAS_SGX OMAP_FEAT_CLASS_OMAP2 | OMAP_HAS_SGX + +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_L2CACHE +#define OMAP3_HAS_IVA OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_IVA +#define OMAP3_HAS_SGX OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_SGX +#define OMAP3_HAS_NEON OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_NEON +#define OMAP3_HAS_ISP OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_ISP
arch/arm/plat-omap/include/plat/cpu.h int omap_chip_is(struct omap_chip_id oci); · line 472
Quoted 10 lines
arch/arm/plat-omap/include/plat/cpu.h +8
-0+#define OMAP3_HAS_192MHZ_CLK OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_192MHZ_CLK +#define OMAP3_HAS_IO_WAKEUP OMAP_FEAT_CLASS_OMAP3 | OMAP_HAS_IOWAKEUP + +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_L2CACHE +#define OMAP4_HAS_IVA OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_IVA +#define OMAP4_HAS_SGX OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_SGX +#define OMAP4_HAS_NEON OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_NEON +#define OMAP4_HAS_ISP OMAP_FEAT_CLASS_OMAP4 | OMAP_HAS_ISP#endif
here is my contention:
there will be two ways to use this:omap_has_feature(OMAP_HAS_SGX) and omap_has_feature(OMAP3_HAS_SGX)
OMAP_HAS_SGX should return true or false no matter what omap silicon it is.
OMAP3_HAS_SGX usage is meant for what? it is a mixture of cpu_is_omap3() and omap_has_feature(OMAP_HAS_SGX) - tries to do two things in one shot. which defeats why we are trying to introduce a generic omap_has_feature in the first place.
a) confusing as there seems to be two standards
b) redundant information use cpu_is_omapxyz() if needed.
IMHO:
+#define OMAP_HAS_L2CACHE BIT(0)
+#define OMAP_HAS_IVA BIT(1)
+#define OMAP_HAS_SGX BIT(2)
+#define OMAP_HAS_NEON BIT(3)
+#define OMAP_HAS_ISP BIT(4)
+#define OMAP3_HAS_192MHZ_CLK BIT(5)
+#define OMAP_HAS_IO_WAKEUP BIT(6)
and later if needed
+#define OMAP4_SOME_NEW_OMAP4ONLY_FEATURE BIT(7)
where OMAP3_HAS is indicative that this is a OMAP3 *only* feature and should be used to differentiate between various omap3 silicon.
Benefits:
a) distinction b/w omap generic and omap family specific features
b) you get to define 32 features instead of reserving 24-32 for OMAP classes.
Signature · 2 lines
Regards, Nishanth Menon
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Quoted 98 lines · Nishanth Menon
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 64 lines
Allow testing for omap type with
omap_has_feature. This can be used to leave outcpu_is_omapxxxxchecks.Signed-off-by Tony Lindgren <[email protected]>---
arch/arm/plat-omap/include/plat/cpu.h | 38
++++++++++++++++++++++++++-------1 files changed, 30 insertions(+), 8 deletions(-)arch/arm/plat-omap/include/plat/cpu.h +15
-8@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci); void omap2_check_revision(void); /* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 |
OMAP_HAS_IVA
+#defineOMAP2_HAS_SGXOMAP_FEAT_CLASS_OMAP2 |OMAP_HAS_SGXarch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 |
OMAP_HAS_L2CACHE
+#defineOMAP3_HAS_IVAOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_IVA
+#defineOMAP3_HAS_SGXOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_SGX
+#defineOMAP3_HAS_NEONOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_NEON
+#defineOMAP3_HAS_ISPOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_ISP
+#defineOMAP3_HAS_192MHZ_CLKOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_192MHZ_CLK
+#defineOMAP3_HAS_IO_WAKEUPOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_IOWAKEUParch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 |
OMAP_HAS_L2CACHE
+#defineOMAP4_HAS_IVAOMAP_FEAT_CLASS_OMAP4 |Quoted 9 lines
OMAP_HAS_IVA
+#defineOMAP4_HAS_SGXOMAP_FEAT_CLASS_OMAP4 |OMAP_HAS_SGX
+#defineOMAP4_HAS_NEONOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_NEON
+#defineOMAP4_HAS_ISPOMAP_FEAT_CLASS_OMAP4 |OMAP_HAS_ISP
#endifhere is my contention:
there will be two ways to use this:omap_has_feature(OMAP_HAS_SGX)andomap_has_feature(OMAP3_HAS_SGX)
OMAP_HAS_SGXshould return true or false no matter what omap silicon it is.
OMAP3_HAS_SGXusage is meant for what? it is a mixture ofcpu_is_omap3()andomap_has_feature(OMAP_HAS_SGX)- tries to do two things in one shot. which defeats why we are trying to introduce a genericomap_has_featurein the first place.
a) confusing as there seems to be two standards
b) redundant information usecpu_is_omapxyz()if needed.IMHO:
arch/arm/plat-omap/include/plat/cpu.h +7
-0+#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP3_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6)and later if needed
+#defineOMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)
Quoted 8 lines
where
OMAP3_HASis indicative that this is a OMAP3 *only* feature and should be used to differentiate between various omap3 silicon.Benefits:
a) distinction b/w omap generic and omap family specific features
b) you get to define 32 features instead of reserving 24-32 for OMAP classes.
I still can't grok the need for the distinction in (a), and for "" +#define OMAP4_SOME_NEW_OMAP4ONLY_FEATURE BIT(7)"" etc.
If that OMAP4ONLY_FEATURE has to be checked, then the code to use it will also be OMAP4 specific.
IOW, as a user, there are 2 ways to use omap_has_xxxx()
void a_generic_funciton_for_all_omaps() {
if (cpu_has_xxxx_feature()
/* Do generic stuff */
}void a_omap_4_specific_function() {
if (omap_has_that_new_feature()
/* Do omap_4 specific stuff */
}In a_generic_function_for_all_omaps(), if there is a need for checking OMAP4_has_xxxx,
then the code will eventually be ugly. There is going to be a cpu_is_xxxx() overload, for things
not expressed through features framework.
I did read the other thread
http://marc.info/?l=linux-omap&m=127858108626850&w=2
and it's been discussed before as well. But I can't see a genuine use case yet, and what is the loss involved
if a omap3_has_192mhz_clk is expressed as omap_has_192mhz_clk, even if it is omap3 specific.
Thanks,
Venkat.
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
S, Venkatraman had written, on 07/08/2010 11:15 AM, the following:
Quoted 101 lines · Nishanth Menon
Quoted 100 lines
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 63 lines
Allow testing for omap type with
omap_has_feature. This can be used to leave outcpu_is_omapxxxxchecks.Signed-off-by Tony Lindgren <[email protected]>--- arch/arm/plat-omap/include/plat/cpu.h | 38 ++++++++++++++++++++++++++------- 1 files changed, 30 insertions(+), 8 deletions(-)arch/arm/plat-omap/include/plat/cpu.h +15
-8@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci);void omap2_check_revision (void );/* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 |
OMAP_HAS_IVA
+#defineOMAP2_HAS_SGXOMAP_FEAT_CLASS_OMAP2|OMAP_HAS_SGXarch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 |
OMAP_HAS_L2CACHE
+#defineOMAP3_HAS_IVAOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_IVA
+#defineOMAP3_HAS_SGXOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_SGX
+#defineOMAP3_HAS_NEONOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_NEON
+#defineOMAP3_HAS_ISPOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_ISP
+#defineOMAP3_HAS_192MHZ_CLKOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_192MHZ_CLK
+#defineOMAP3_HAS_IO_WAKEUPOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_IOWAKEUParch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 |
OMAP_HAS_L2CACHE
+#defineOMAP4_HAS_IVAOMAP_FEAT_CLASS_OMAP4|Quoted 9 lines
OMAP_HAS_IVA
+#defineOMAP4_HAS_SGXOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_SGX
+#defineOMAP4_HAS_NEONOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_NEON
+#defineOMAP4_HAS_ISPOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_ISP
#endifhere is my contention:
there will be two ways to use this:omap_has_feature(OMAP_HAS_SGX)andomap_has_feature(OMAP3_HAS_SGX)
OMAP_HAS_SGXshould return true or false no matter what omap silicon it is.
OMAP3_HAS_SGXusage is meant for what? it is a mixture ofcpu_is_omap3()andomap_has_feature(OMAP_HAS_SGX)- tries to do two things in one shot. which defeats why we are trying to introduce a genericomap_has_featurein the first place.
a) confusing as there seems to be two standards
b) redundant information usecpu_is_omapxyz()if needed.IMHO:
arch/arm/plat-omap/include/plat/cpu.h +7
-0+#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP3_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6)and later if needed
+#defineOMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)where
OMAP3_HASis indicative that this is a OMAP3 *only* feature and should be used to differentiate between various omap3 silicon.
Quoted 9 lines
Quoted 5 lines
Benefits:
a) distinction b/w omap generic and omap family specific features
b) you get to define 32 features instead of reserving 24-32 for OMAP classes.I still can't grok the need for the distinction in (a), and for "" +#define
OMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)"" etc.
OMAP_HAS_192MHZ_CLK -> does not indicate if this is omap3 ONLY feature (e.g. 3430 does not have it, 3630 has it) but we know that omap4, 2, 1 etc dont need it.
in terms of readability, when i see omap_has_feature(OMAP3_HAS_xyz), I can immediately review the code/read the code with the context of omap3 alone Vs if this code was used in omap4/2/1 context question why it is so and we can all improve.
e.g. if a generic clock code meant for all omaps used 192MHZ, I would question why is cpu specific feature being used there. which is easier with a OMAP3_ tag.
Quoted 10 lines
If that
OMAP4ONLY_FEATUREhas to be checked, then the code to use it will also be OMAP4 specific.IOW, as a user, there are 2 ways to use
omap_has_xxxx()void a_generic_funciton_for_all_omaps() { if (cpu_has_xxxx_feature() /* Do generic stuff */ }
Quoted 4 lines
void a_omap_4_specific_function() {
if (omap_has_that_new_feature()
/* Do omap_4 specific stuff */
}See above explanation. sitting in the shoes of a reviewer who looks through code not written by self, it helps to have some differentiation in definitions to aid review.
Quoted 6 lines
In
a_generic_function_for_all_omaps(), if there is a need for checkingOMAP4_has_xxxx,
then the code will eventually be ugly. There is going to be acpu_is_xxxx()overload, for things
not expressed through features framework.
No matter how elegant the framework is, someone definitely will find a crappy way to use it.. we've all been there, seen that and some of us (including yours truely) done that.
Quoted 10 lines
I did read the other thread
http://marc.info/?l=linux-omap&m=127858108626850&w=2
and it's been discussed before as well. But I can't see a genuine use case yet, and what is the loss involved
if aomap3_has_192mhz_clkis expressed asomap_has_192mhz_clk, even if it is omap3 specific.Thanks,
Venkat.
Signature · 2 lines
Regards, Nishanth Menon
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Quoted 123 lines · Nishanth Menon
S, Venkatraman had written, on 07/08/2010 11:15 AM, the following:
Quoted 108 lines · Nishanth Menon
Quoted 106 lines
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 64 lines
Allow testing for omap type with
omap_has_feature. This can be used to leave outcpu_is_omapxxxxchecks.Signed-off-by Tony Lindgren <[email protected]>---
arch/arm/plat-omap/include/plat/cpu.h | 38
++++++++++++++++++++++++++-------1 files changed, 30 insertions(+), 8 deletions(-)arch/arm/plat-omap/include/plat/cpu.h +15
-8@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci); void omap2_check_revision(void); /* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 |
OMAP_HAS_IVA
+#defineOMAP2_HAS_SGXOMAP_FEAT_CLASS_OMAP2 |OMAP_HAS_SGXarch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 |
OMAP_HAS_L2CACHE
+#defineOMAP3_HAS_IVAOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_IVA
+#defineOMAP3_HAS_SGXOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_SGX
+#defineOMAP3_HAS_NEONOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_NEON
+#defineOMAP3_HAS_ISPOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_ISP
+#defineOMAP3_HAS_192MHZ_CLKOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_192MHZ_CLK
+#defineOMAP3_HAS_IO_WAKEUPOMAP_FEAT_CLASS_OMAP3 |OMAP_HAS_IOWAKEUParch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 |
OMAP_HAS_L2CACHE
+#defineOMAP4_HAS_IVAOMAP_FEAT_CLASS_OMAP4 |Quoted 9 lines
OMAP_HAS_IVA
+#defineOMAP4_HAS_SGXOMAP_FEAT_CLASS_OMAP4 |OMAP_HAS_SGX
+#defineOMAP4_HAS_NEONOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_NEON
+#defineOMAP4_HAS_ISPOMAP_FEAT_CLASS_OMAP4 |OMAP_HAS_ISP
#endifhere is my contention:
there will be two ways to use this:omap_has_feature(OMAP_HAS_SGX)andomap_has_feature(OMAP3_HAS_SGX)
OMAP_HAS_SGXshould return true or false no matter what omap silicon it is.
OMAP3_HAS_SGXusage is meant for what? it is a mixture ofcpu_is_omap3()andomap_has_feature(OMAP_HAS_SGX)- tries to do two things in one shot. which
defeats why we are trying to introduce a genericomap_has_featurein the first place.
a) confusing as there seems to be two standards
b) redundant information usecpu_is_omapxyz()if needed.IMHO:
arch/arm/plat-omap/include/plat/cpu.h +7
-0+#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP3_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6)and later if needed
+#defineOMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)where
OMAP3_HASis indicative that this is a OMAP3 *only* feature and should
be used to differentiate between various omap3 silicon.Quoted 9 lines
Quoted 5 lines
Benefits:
a) distinction b/w omap generic and omap family specific features
b) you get to define 32 features instead of reserving 24-32 for OMAP classes.I still can't grok the need for the distinction in (a), and for "" +#define
OMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)"" etc.
OMAP_HAS_192MHZ_CLK-> does not indicate if this is omap3 ONLY feature (e.g. 3430 does not have it, 3630 has it) but we know that omap4, 2, 1 etc dont need it.
Quoted 8 lines
in terms of readability, when i see
omap_has_feature(OMAP3_HAS_xyz), I can immediately review the code/read the code with the context of omap3 alone Vs if this code was used in omap4/2/1 context question why it is so and we can all improve.e.g. if a generic clock code meant for all omaps used 192MHZ, I would question why is cpu specific feature being used there. which is easier with a
OMAP3_tag.
If we extend this analogy, I can write omap_dma_driver_init(OMAP3_NUM_CHANNELS) only to make it "more readable" and provide a omap3 specific context to the reader. The whole point was that the user code doesn't have such klux.
Quoted 12 lines
Quoted 12 lines
If that
OMAP4ONLY_FEATUREhas to be checked, then the code to use it will also be OMAP4 specific.IOW, as a user, there are 2 ways to use
omap_has_xxxx()void
a_generic_funciton_for_all_omaps(){
if (cpu_has_xxxx_feature()
/* Do generic stuff */
}
Quoted 9 lines
Quoted 4 lines
void
a_omap_4_specific_function(){
if (omap_has_that_new_feature()
/* Doomap_4specific stuff */
}See above explanation. sitting in the shoes of a reviewer who looks through code not written by self, it helps to have some differentiation in definitions to aid review.
I think this 'lazy reviewability' comes at the cost of very
abstraction the features framework is intended to provide, not to mention the question of correct selection (is this a OMAP4 specific feature or is OMAP5 expected to have it ?). and upgradation.
As mentioned before, the surrounding context of the use ofomap_has_feature() will provide enough clues about the cpu specific nature of a feature, if at all needed.
Thanks,
Venkat.
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Venkatraman S had written, on 07/08/2010 02:28 PM, the following:
Quoted 124 lines · Nishanth Menon
Quoted 123 lines
S, Venkatraman had written, on 07/08/2010 11:15 AM, the following:
Quoted 105 lines · Nishanth Menon
Quoted 104 lines
Tony Lindgren had written, on 07/08/2010 04:38 AM, the following:
Quoted 63 lines
Allow testing for omap type with
omap_has_feature. This can be used to leave outcpu_is_omapxxxxchecks.Signed-off-by Tony Lindgren <[email protected]>--- arch/arm/plat-omap/include/plat/cpu.h | 38 ++++++++++++++++++++++++++------- 1 files changed, 30 insertions(+), 8 deletions(-)arch/arm/plat-omap/include/plat/cpu.h +15
-8@@ -437,14 +437,36 @@ int omap_chip_is(struct omap_chip_id oci);void omap2_check_revision (void );/* -* Runtime detection of OMAP3 features +* Runtime detection of OMAP features */ -#define OMAP3_HAS_L2CACHE BIT(0) -#define OMAP3_HAS_IVA BIT(1) -#define OMAP3_HAS_SGX BIT(2) -#define OMAP3_HAS_NEON BIT(3) -#define OMAP3_HAS_ISP BIT(4) -#define OMAP3_HAS_192MHZ_CLK BIT(5) -#define OMAP3_HAS_IO_WAKEUP BIT(6) +#define OMAP_FEAT_CLASS_OMAP1 BIT(24) +#define OMAP_FEAT_CLASS_OMAP2 BIT(25) +#define OMAP_FEAT_CLASS_OMAP3 BIT(26) +#define OMAP_FEAT_CLASS_OMAP4 BIT(27) + +#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6) + +#define OMAP2_HAS_IVA OMAP_FEAT_CLASS_OMAP2 |
OMAP_HAS_IVA
+#defineOMAP2_HAS_SGXOMAP_FEAT_CLASS_OMAP2|OMAP_HAS_SGXarch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP3_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP3 |
OMAP_HAS_L2CACHE
+#defineOMAP3_HAS_IVAOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_IVA
+#defineOMAP3_HAS_SGXOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_SGX
+#defineOMAP3_HAS_NEONOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_NEON
+#defineOMAP3_HAS_ISPOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_ISP
+#defineOMAP3_HAS_192MHZ_CLKOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_192MHZ_CLK
+#defineOMAP3_HAS_IO_WAKEUPOMAP_FEAT_CLASS_OMAP3|OMAP_HAS_IOWAKEUParch/arm/plat-omap/include/plat/cpu.h +2
-0+ +#define OMAP4_HAS_L2CACHE OMAP_FEAT_CLASS_OMAP4 |
OMAP_HAS_L2CACHE
+#defineOMAP4_HAS_IVAOMAP_FEAT_CLASS_OMAP4|Quoted 9 lines
OMAP_HAS_IVA
+#defineOMAP4_HAS_SGXOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_SGX
+#defineOMAP4_HAS_NEONOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_NEON
+#defineOMAP4_HAS_ISPOMAP_FEAT_CLASS_OMAP4|OMAP_HAS_ISP
#endifhere is my contention:
there will be two ways to use this:omap_has_feature(OMAP_HAS_SGX)andomap_has_feature(OMAP3_HAS_SGX)
OMAP_HAS_SGXshould return true or false no matter what omap silicon it is.
OMAP3_HAS_SGXusage is meant for what? it is a mixture ofcpu_is_omap3()andomap_has_feature(OMAP_HAS_SGX)- tries to do two things in one shot. which
defeats why we are trying to introduce a genericomap_has_featurein the first place.
a) confusing as there seems to be two standards
b) redundant information usecpu_is_omapxyz()if needed.IMHO:
arch/arm/plat-omap/include/plat/cpu.h +7
-0+#define OMAP_HAS_L2CACHE BIT(0) +#define OMAP_HAS_IVA BIT(1) +#define OMAP_HAS_SGX BIT(2) +#define OMAP_HAS_NEON BIT(3) +#define OMAP_HAS_ISP BIT(4) +#define OMAP3_HAS_192MHZ_CLK BIT(5) +#define OMAP_HAS_IO_WAKEUP BIT(6)and later if needed
+#defineOMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)where
OMAP3_HASis indicative that this is a OMAP3 *only* feature and should
be used to differentiate between various omap3 silicon.Quoted 8 lines
Quoted 5 lines
Benefits:
a) distinction b/w omap generic and omap family specific features
b) you get to define 32 features instead of reserving 24-32 for OMAP classes.I still can't grok the need for the distinction in (a), and for "" +#define
OMAP4_SOME_NEW_OMAP4ONLY_FEATUREBIT(7)"" etc.
OMAP_HAS_192MHZ_CLK-> does not indicate if this is omap3 ONLY feature (e.g. 3430 does not have it, 3630 has it) but we know that omap4, 2, 1 etc dont need it.in terms of readability, when i see
omap_has_feature(OMAP3_HAS_xyz), I can immediately review the code/read the code with the context of omap3 alone Vs if this code was used in omap4/2/1 context question why it is so and we can all improve.
Quoted 8 lines
Quoted 3 lines
e.g. if a generic clock code meant for all omaps used 192MHZ, I would question why is cpu specific feature being used there. which is easier with a
OMAP3_tag.If we extend this analogy, I can write
omap_dma_driver_init(OMAP3_NUM_CHANNELS)only to make it "more readable" and provide a omap3 specific context to the reader. The whole point was that the user code doesn't have such klux.
I must sadly disagree on this. different concept entirely.
in the case of a dma driver, num_channels is defintely something you'd like to abstract out. the initialization makes sense, but from a driver context, using omap specific - nak.
Quoted 21 lines
Quoted 12 lines
Quoted 12 lines
If that
OMAP4ONLY_FEATUREhas to be checked, then the code to use it will also be OMAP4 specific.IOW, as a user, there are 2 ways to use
omap_has_xxxx()void a_generic_funciton_for_all_omaps() { if (cpu_has_xxxx_feature() /* Do generic stuff */ }Quoted 8 lines
Quoted 4 lines
void a_omap_4_specific_function() { if (omap_has_that_new_feature() /* Do omap_4 specific stuff */ }See above explanation. sitting in the shoes of a reviewer who looks through code not written by self, it helps to have some differentiation in definitions to aid review.
Quoted 8 lines
I think this 'lazy reviewability' comes at the cost of very
abstraction the features framework is intended to provide, not to mention the question of correct selection (is this a OMAP4 specific feature or is OMAP5 expected to have it ?). and upgradation.As mentioned before, the surrounding context of the use of
omap_has_feature()will provide enough clues about the cpu specific nature of a feature, if at all needed.
Does it really? when a new feature is added, dont we want to know if it is generic feature or a omap specific feature? where is the flag?
1 year down the line, neither of us will remember this discussion in detail. lets take an example what might happen:
assume OMAP_HAS_192MHZ is defined - with our good intentions of being omap3.
developer Y comes along for OMAP5, sees that there is a flag for OMAP_HAS_192MHZ and says - cool i need this and since it is already present,
writes code for
if(omap_has_feature(OMAP_HAS_192MHZ)) {
/* enable some cruft errata */
}in omap5 code. guess what he posts the patch to us, a quick reviewer has forgotten that OMAP_HAS_192MHZ was omap3 only feature, ignores the fact that the feature was not enabled in omap5's check_revision.. darn.. we have a bug to fix.
instead, the developer is now able to create two patches:
a) convert OMAP3_HAS_192MHZ into OMAP_HAS_192MHZ after adding the omap5 check_revision
b) do his crufty code..
same applies for reviewers.
Signature · 2 lines
Regards, Nishanth Menon
Tony Lindgren
maintainerRe: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Quoted 6 lines · Nishanth Menon
Quoted 6 lines
I think this 'lazy reviewability' comes at the cost of very
abstraction the features framework is intended to provide, not to mention the question of correct selection (is this a OMAP4 specific feature or is OMAP5 expected to have it ?). and upgradation.
Quoted 6 lines
Quoted 3 lines
As mentioned before, the surrounding context of the use of
omap_has_feature()will provide enough clues about the cpu specific nature of a feature, if at all needed.Does it really? when a new feature is added, dont we want to know if it is generic feature or a omap specific feature? where is the flag?
Yeah I don't know what we should do with these defines.. Kind of just threw the patch out there.
If we already have omap specific omap_has_feature functions, we don't need cpu_is_omapxxxx in most cases.
I suggest we only use the generic defines now, then look at it again when we run out of the bits to define.
Regards,
Tony
Re: [PATCH 5/5] omap: Allow testing for omap type with omap_has_feature
Tony Lindgren had written, on 07/09/2010 02:04 AM, the following:
Quoted 12 lines · Nishanth Menon
Quoted 10 lines
Quoted 8 lines
I think this 'lazy reviewability' comes at the cost of very
abstraction the features framework is intended to provide, not to mention the question of correct selection (is this a OMAP4 specific feature or is OMAP5 expected to have it ?). and upgradation.As mentioned before, the surrounding context of the use of
omap_has_feature()will provide enough clues about the cpu specific nature of a feature, if at all needed.Does it really? when a new feature is added, dont we want to know if it is generic feature or a omap specific feature? where is the flag?
Quoted 8 lines
Yeah I don't know what we should do with these defines.. Kind of just threw the patch out there.
If we already have omap specific
omap_has_featurefunctions, we don't needcpu_is_omapxxxxin most cases.I suggest we only use the generic defines now, then look at it again when we run out of the bits to define.
ack.
Signature · 2 lines
Regards, Nishanth Menon