drm/i915: Correcting the reg definitions for PORT_DFT
Subsystems (2)
PATCH 1/1 drm/i915: Correcting the reg definitions for PORT_DFT
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for PORT_DFT_I9XX and PORT_DFT2_G4X were wrong.
MMIO offset needs to be added to these register offsets to fix them.
Issue: GMINL-6869
--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_reg.h +1 -1
@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31)
Signature · 17 lines
1.9.1 --------------------------------------------------------------------- Intel Corporation (UK) Limited Registered No. 1134945 (England) Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/i915_reg.h +0
-0
drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XX isn't used on VLV/CHV, so this doesn't change anything.
Quoted 21 lines
drivers/gpu/drm/i915/i915_reg.h +0
-0#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31)Signature · 17 lines
1.9.1 --------------------------------------------------------------------- Intel Corporation (UK) Limited Registered No. 1134945 (England) Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
This change is needed for some hardware composer tests in chv.
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Friday, June 5, 2015 2:08 PM
To: Mathai, Minu
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions for PORT_DFT
Quoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/gpu/drm/i915/i915_reg.hindex 7213224..c327c7c 100644
drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XX isn't used on VLV/CHV, so this doesn't change anything.
Quoted 16 lines
drivers/gpu/drm/i915/i915_reg.h +0
-1#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31) --1.9.1
---------------------------------------------------------------------
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Jani Nikula
maintainerQuoted 1 line · Minu
This change is needed for some hardware composer tests in chv.
As Ville said, this #define is not used by the upstream kernel on byt/chv. Maybe you have some out-of-tree patches using that?
BR,
Jani.
Quoted 63 lines
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Friday, June 5, 2015 2:08 PM
To: Mathai, Minu
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/gpu/drm/i915/i915_reg.hindex 7213224..c327c7c 100644drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 16 lines
drivers/gpu/drm/i915/i915_reg.h +0
-1#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31) --1.9.1
---------------------------------------------------------------------
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
The hardware composer test which found the problem with this register definition is used in Android testing and isn't open source. However fixing this register definition will stop potential problems for future use cases and would reduce the rebase effort for our Android tree.
-----Original Message-----
From: Jani Nikula [mailto:[email protected]]
Sent: Monday, June 8, 2015 2:48 PM
To: Mathai, Minu; Ville Syrjälä
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions for PORT_DFT
Quoted 1 line · Minu
This change is needed for some hardware composer tests in chv.
As Ville said, this #define is not used by the upstream kernel on byt/chv. Maybe you have some out-of-tree patches using that?
BR,
Jani.
Quoted 57 lines
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Friday, June 5, 2015 2:08 PM
To: Mathai, Minu
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/gpu/drm/i915/i915_reg.hindex 7213224..c327c7c 100644drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 16 lines
drivers/gpu/drm/i915/i915_reg.h +0
-1#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31) --1.9.1
---------------------------------------------------------------------
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 7 lines
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 2 lines · Minu
The hardware composer test which found the problem with this register definition is used in Android testing and isn't open source. However fixing this register definition will stop potential problems for future use cases and would reduce the rebase effort for our Android tree.
What are you testing with this register? The register isn't even listed in any VLV/CHV documentation so I have no idea what you would even do with it.
Quoted 84 lines
-----Original Message-----
From: Jani Nikula [mailto:[email protected]]
Sent: Monday, June 8, 2015 2:48 PM
To: Mathai, Minu; Ville Syrjälä
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 1 line · Minu
This change is needed for some hardware composer tests in chv.
As Ville said, this #define is not used by the upstream kernel on byt/chv. Maybe you have some out-of-tree patches using that?
BR,
Jani.Quoted 57 lines
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Friday, June 5, 2015 2:08 PM
To: Mathai, Minu
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/gpu/drm/i915/i915_reg.hindex 7213224..c327c7c 100644drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 16 lines
drivers/gpu/drm/i915/i915_reg.h +0
-1#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31) --1.9.1
---------------------------------------------------------------------
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfxQuoted 7 lines
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx--
Jani Nikula, Intel Open Source Technology Center
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Agreed .We don't need this
Thanks
Minu
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Tuesday, June 9, 2015 1:17 PM
To: Mathai, Minu
Cc: Jani Nikula; [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions for PORT_DFT
Quoted 2 lines · Minu
The hardware composer test which found the problem with this register definition is used in Android testing and isn't open source. However fixing this register definition will stop potential problems for future use cases and would reduce the rebase effort for our Android tree.
What are you testing with this register? The register isn't even listed in any VLV/CHV documentation so I have no idea what you would even do with it.
Quoted 84 lines
-----Original Message-----
From: Jani Nikula [mailto:[email protected]]
Sent: Monday, June 8, 2015 2:48 PM
To: Mathai, Minu; Ville Syrjälä
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 1 line · Minu
This change is needed for some hardware composer tests in chv.
As Ville said, this #define is not used by the upstream kernel on byt/chv. Maybe you have some out-of-tree patches using that?
BR,
Jani.Quoted 58 lines
-----Original Message-----
From: Ville Syrjälä [mailto:[email protected]]
Sent: Friday, June 5, 2015 2:08 PM
To: Mathai, Minu
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Correcting the reg definitions forPORT_DFTQuoted 17 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/gpu/drm/i915/i915_reg.hindex 7213224..c327c7c 100644drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160 -#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 17 lines
drivers/gpu/drm/i915/i915_reg.h +0
-1#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31) --1.9.1
-------------------------------------------------------------------
--
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfxQuoted 7 lines
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 2 lines
--
Jani Nikula, Intel Open Source Technology Center
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 21 lines · Ville Syrjälä
Quoted 20 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/i915_reg.h +0
-0@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160
Quoted 8 lines
drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 2 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1-#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 3 lines
drivers/gpu/drm/i915/i915_reg.h +0
-0#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31)
Regardless of whether it's used, we have an inconsistency between the definitions of PORT_DFT_I9XX and PORT_DFT2_G4X -- one includes the mmio_offset and the other doesn't. Personally I think the #define with an implicit dependency on an object called "dev_priv" is really ugly and we should move away from that style rather than adding mode of them, but that's a lot of work.
As Minu says PORT_DFT_I9XX isn't really needed after all, can we just delete it to remove the inconsistency?
.Dave.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 38 lines · Dave Gordon
Quoted 21 lines · Ville Syrjälä
Quoted 20 lines · Minu Mathai
From: Minu <[email protected]>
Display CRCs were not readable because the register defintions for
PORT_DFT_I9XXandPORT_DFT2_G4Xwere wrong.
MMIO offset needs to be added to these register offsets to fix them.Issue: GMINL-6869
Signed-off-by Minu Mathai <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/i915_reg.h +0
-0@@ -3193,7 +3193,7 @@ enum skl_disp_power_wells {#define PCH_HDMIC 0xe1150#define PCH_HDMID 0xe1160Quoted 8 lines
drivers/gpu/drm/i915/i915_reg.h enum skl_disp_power_wells { · line 3196
Quoted 2 lines
drivers/gpu/drm/i915/i915_reg.h +1
-1-#define PORT_DFT_I9XX 0x61150 +#define PORT_DFT_I9XX (dev_priv->info.display_mmio_offset + 0x61150)
PORT_DFT_I9XXisn't used on VLV/CHV, so this doesn't change anything.Quoted 3 lines
drivers/gpu/drm/i915/i915_reg.h +0
-0#define DC_BALANCE_RESET (1 << 25)#define PORT_DFT2_G4X (dev_priv->info.display_mmio_offset + 0x61154)#define DC_BALANCE_RESET_VLV (1 << 31)Regardless of whether it's used, we have an inconsistency between the definitions of
PORT_DFT_I9XXandPORT_DFT2_G4X-- one includes themmio_offsetand the other doesn't. Personally I think the #define with an implicit dependency on an object called "dev_priv" is really ugly and we should move away from that style rather than adding mode of them, but that's a lot of work.
Quoted 2 lines
As Minu says
PORT_DFT_I9XXisn't really needed after all, can we just delete it to remove the inconsistency?
No, it's used on g4x.
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Jani Nikula
maintainerQuoted 3 lines · Dave Gordon
Regardless of whether it's used, we have an inconsistency between the definitions of
PORT_DFT_I9XXandPORT_DFT2_G4X-- one includes themmio_offsetand the other doesn't.
It's not inconsistent, it's consistent on another level:
We've settled on including the mmio_offset only for macros that need it. If a macro is relevant only on platforms that all have the same mmio offset, the offset is included statically (currently 0 or VLV_DISPLAY_BASE). dev_priv->info.display_mmio_offset is only used when the macro is relevant on platforms with different mmio offsets. This we try to follow consistently.
Quoted 3 lines
Personally I think the #define with an implicit dependency on an object called "
dev_priv" is really ugly and we should move away from that style rather than adding mode of them, but that's a lot of work.
Agreed in principle, but I think we've lost the battle. It's way too much churn, and makes a lot of code really ugly. Compare these:
I915_WRITE(FOOBAR, I915_READ(FOOBAR) | FOOBAR_ENABLE); I915_WRITE(dev_priv, FOOBAR(dev_priv), I915_READ(dev_priv, FOOBAR(dev_priv)) | FOOBAR_ENABLE);
We'll try to focus on only requiring "dev_priv" implicitly and nothing else, and where a parameter does get passed, we'll try to make it "dev_priv" as that pretty much has to be around anyway.
BR,
Jani.
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 5 lines · Jani Nikula
Quoted 3 lines · Dave Gordon
Regardless of whether it's used, we have an inconsistency between the definitions of
PORT_DFT_I9XXandPORT_DFT2_G4X-- one includes themmio_offsetand the other doesn't.
Quoted 8 lines
It's not inconsistent, it's consistent on another level:
We've settled on including the
mmio_offsetonly for macros that need it. If a macro is relevant only on platforms that all have the same mmio offset, the offset is included statically (currently 0 orVLV_DISPLAY_BASE).dev_priv->info.display_mmio_offsetis only used when the macro is relevant on platforms with different mmio offsets. This we try to follow consistently.
So you only have a problem when a previously-statically located block turns into something that can be (and is) moved around on different platforms ... I suppose that shouldn't happen too often.
Quoted 11 lines
Quoted 3 lines
Personally I think the #define with an implicit dependency on an object called "
dev_priv" is really ugly and we should move away from that style rather than adding more of them, but that's a lot of work.Agreed in principle, but I think we've lost the battle. It's way too much churn, and makes a lot of code really ugly. Compare these:
I915_WRITE(FOOBAR, I915_READ(FOOBAR) | FOOBAR_ENABLE); I915_WRITE(dev_priv, FOOBAR(dev_priv), I915_READ(dev_priv, FOOBAR(dev_priv)) | FOOBAR_ENABLE);
I'd rather write
uint32_t foobar = I915_DISP_REG_READ(dev_priv, FOOBAR); I915_DISP_REG_WRITE(dev_priv, foobar | FOOBAR_ENABLE);
And if there were very many places where that read-modify-write was used, we could define that as
#define I915_DISP_REG_BITSET(dev_priv, reg, bits) ...
Of course it would require people to use the right macro according to whether the register was part of the display block, the GT block, or however many other areas there are that might be independently relocated ... but it might be more future proof :)
Quoted 6 lines
We'll try to focus on only requiring "
dev_priv" implicitly and nothing else, and where a parameter does get passed, we'll try to make it "dev_priv" as that pretty much has to be around anyway.BR,
Jani.
Well I guess we're stuck with 'dev_priv' for the foreseeable future :(
Hmmm ... we have quite a bit of code that passes 'dev' around rather than dev_priv, where the (usually static) called function immediately uses it to derive dev_priv, and where the only other uses in the called function are in calls to INTEL_INFO(dev) and its derivatives. For example:
static int get_context_size(struct drm_device *dev)
{
struct drm_i915_private *dev_priv = dev->dev_private;
int ret;
u32 reg;
switch (INTEL_INFO(dev)->gen) {
case 6:
reg = I915_READ(CXT_SIZE);
ret = GEN6_CXT_TOTAL_SIZE(reg) * 64;
break;
case 7:
reg = I915_READ(GEN7_CXT_SIZE);
if (IS_HASWELL(dev))
ret = HSW_CXT_TOTAL_SIZE;
else
ret = GEN7_CXT_TOTAL_SIZE(reg) * 64;
break;
case 8:
ret = GEN8_CXT_TOTAL_SIZE;
break;
default:
BUG();
}
return ret;
}But since INTEL_INFO() can take 'dev_priv' as a parameter (as an alternative to 'dev'), all the uses of 'dev' here could be replaced by 'dev_priv' and then the parameter changed to pass that directly, thus potentially eliminating quite a few extra memory references. Applied driver-wide, that micro-optimisation might add up to something useful :)
.Dave.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Jani Nikula
maintainerQuoted 5 lines · Dave Gordon
But since
INTEL_INFO()can take 'dev_priv' as a parameter (as an alternative to 'dev'), all the uses of 'dev' here could be replaced by 'dev_priv' and then the parameter changed to pass that directly, thus potentially eliminating quite a few extra memory references. Applied driver-wide, that micro-optimisation might add up to something useful :)
I think we haven't done that driver-wide because it would be more disruptive than beneficial, but we're slowly moving at that direction whenever we change stuff. And this is exactly the reason why Chris hacked INTEL_INFO et al. to accept either dev or dev_priv.
BR,
Jani.
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx