Define a compatible string for the Mali Bifrost GPU found in
Mediatek's MT8183 SoCs.
Signed-off-by: Nicolas Boichat <[email protected]>
Reviewed-by: Alyssa Rosenzweig <[email protected]>
---
v4:
- Add power-domain-names description
(kept Alyssa's reviewed-by as the change is minor)
v3:
- No change
.../bindings/gpu/arm,mali-bifrost.yaml | 25 +++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
index 4ea6a8789699709..0d93b3981445977 100644
--- a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
+++ b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
@@ -17,6 +17,7 @@ properties:
items:
- enum:
- amlogic,meson-g12a-mali
+ - mediatek,mt8183-mali
- realtek,rtd1619-mali
- rockchip,px30-mali
- const: arm,mali-bifrost # Mali Bifrost GPU model/revision is fully discoverable
@@ -62,6 +63,30 @@ allOf:
minItems: 2
required:
- resets
+ - if:
+ properties:
+ compatible:
+ contains:
+ const: mediatek,mt8183-mali
+ then:
+ properties:
+ sram-supply: true
+ power-domains:
+ description:
+ List of phandle and PM domain specifier as documented in
+ Documentation/devicetree/bindings/power/power_domain.txt
+ minItems: 3
+ maxItems: 3
+ power-domain-names:
+ items:
+ - const: core0
+ - const: core1
+ - const: 2d
+
+ required:
+ - sram-supply
+ - power-domains
+ - power-domains-names
examples:
- |
--
2.25.0.341.g760bfbb309-goog
On Fri, Feb 07, 2020 at 01:26:21PM +0800, Nicolas Boichat wrote:
> Define a compatible string for the Mali Bifrost GPU found in
> Mediatek's MT8183 SoCs.
>
> Signed-off-by: Nicolas Boichat <[email protected]>
> Reviewed-by: Alyssa Rosenzweig <[email protected]>
> ---
>
> v4:
> - Add power-domain-names description
> (kept Alyssa's reviewed-by as the change is minor)
> v3:
> - No change
>
> .../bindings/gpu/arm,mali-bifrost.yaml | 25 +++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> index 4ea6a8789699709..0d93b3981445977 100644
> --- a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> +++ b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> @@ -17,6 +17,7 @@ properties:
> items:
> - enum:
> - amlogic,meson-g12a-mali
> + - mediatek,mt8183-mali
> - realtek,rtd1619-mali
> - rockchip,px30-mali
> - const: arm,mali-bifrost # Mali Bifrost GPU model/revision is fully discoverable
> @@ -62,6 +63,30 @@ allOf:
> minItems: 2
> required:
> - resets
> + - if:
> + properties:
> + compatible:
> + contains:
> + const: mediatek,mt8183-mali
> + then:
> + properties:
> + sram-supply: true
> + power-domains:
> + description:
> + List of phandle and PM domain specifier as documented in
> + Documentation/devicetree/bindings/power/power_domain.txt
> + minItems: 3
> + maxItems: 3
> + power-domain-names:
> + items:
> + - const: core0
> + - const: core1
> + - const: 2d
AFAIK, there's no '2d' block in bifrost GPUs. A power domain for each
core group is correct though.
Rob
+Nick Fan +Sj Huang @ MTK
On Wed, Feb 26, 2020 at 1:16 AM Rob Herring <[email protected]> wrote:
>
> On Fri, Feb 07, 2020 at 01:26:21PM +0800, Nicolas Boichat wrote:
> > Define a compatible string for the Mali Bifrost GPU found in
> > Mediatek's MT8183 SoCs.
> >
> > Signed-off-by: Nicolas Boichat <[email protected]>
> > Reviewed-by: Alyssa Rosenzweig <[email protected]>
> > ---
> >
> > v4:
> > - Add power-domain-names description
> > (kept Alyssa's reviewed-by as the change is minor)
> > v3:
> > - No change
> >
> > .../bindings/gpu/arm,mali-bifrost.yaml | 25 +++++++++++++++++++
> > 1 file changed, 25 insertions(+)
> >
> > diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > index 4ea6a8789699709..0d93b3981445977 100644
> > --- a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > +++ b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > @@ -17,6 +17,7 @@ properties:
> > items:
> > - enum:
> > - amlogic,meson-g12a-mali
> > + - mediatek,mt8183-mali
> > - realtek,rtd1619-mali
> > - rockchip,px30-mali
> > - const: arm,mali-bifrost # Mali Bifrost GPU model/revision is fully discoverable
> > @@ -62,6 +63,30 @@ allOf:
> > minItems: 2
> > required:
> > - resets
> > + - if:
> > + properties:
> > + compatible:
> > + contains:
> > + const: mediatek,mt8183-mali
> > + then:
> > + properties:
> > + sram-supply: true
> > + power-domains:
> > + description:
> > + List of phandle and PM domain specifier as documented in
> > + Documentation/devicetree/bindings/power/power_domain.txt
> > + minItems: 3
> > + maxItems: 3
> > + power-domain-names:
> > + items:
> > + - const: core0
> > + - const: core1
> > + - const: 2d
>
> AFAIK, there's no '2d' block in bifrost GPUs. A power domain for each
> core group is correct though.
Good question... Hopefully Nick/SJ@MTK can comment, the non-upstream DTS has:
gpu: mali@13040000 {
compatible = "mediatek,mt8183-mali", "arm,mali-bifrost";
power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE0>;
...
}
gpu_core1: mali_gpu_core1 {
compatible = "mediatek,gpu_core1";
power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE1>;
};
gpu_core2: mali_gpu_core2 {
compatible = "mediatek,gpu_core2";
power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_2D>;
};
So I picked core0/core1/2d as names, but looking at this, it's likely
core2 is more appropriate (and MT8183_POWER_DOMAIN_MFG_2D might just
be a internal/legacy name, if there is no real 2d domain).
Thanks.
> Rob
Sorry for my late reply.
I have checked internally.
The MT8183_POWER_DOMAIN_MFG_2D is just a legacy name, not really 2D
domain.
If the naming too confusing, we can change this name to
MT8183_POWER_DOMAIN_MFG_CORE2 for consistency.
Thanks
Nick Fan
On Wed, 2020-02-26 at 08:55 +0800, Nicolas Boichat wrote:
> +Nick Fan +Sj Huang @ MTK
>
> On Wed, Feb 26, 2020 at 1:16 AM Rob Herring <[email protected]> wrote:
> >
> > On Fri, Feb 07, 2020 at 01:26:21PM +0800, Nicolas Boichat wrote:
> > > Define a compatible string for the Mali Bifrost GPU found in
> > > Mediatek's MT8183 SoCs.
> > >
> > > Signed-off-by: Nicolas Boichat <[email protected]>
> > > Reviewed-by: Alyssa Rosenzweig <[email protected]>
> > > ---
> > >
> > > v4:
> > > - Add power-domain-names description
> > > (kept Alyssa's reviewed-by as the change is minor)
> > > v3:
> > > - No change
> > >
> > > .../bindings/gpu/arm,mali-bifrost.yaml | 25 +++++++++++++++++++
> > > 1 file changed, 25 insertions(+)
> > >
> > > diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > index 4ea6a8789699709..0d93b3981445977 100644
> > > --- a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > +++ b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > @@ -17,6 +17,7 @@ properties:
> > > items:
> > > - enum:
> > > - amlogic,meson-g12a-mali
> > > + - mediatek,mt8183-mali
> > > - realtek,rtd1619-mali
> > > - rockchip,px30-mali
> > > - const: arm,mali-bifrost # Mali Bifrost GPU model/revision is fully discoverable
> > > @@ -62,6 +63,30 @@ allOf:
> > > minItems: 2
> > > required:
> > > - resets
> > > + - if:
> > > + properties:
> > > + compatible:
> > > + contains:
> > > + const: mediatek,mt8183-mali
> > > + then:
> > > + properties:
> > > + sram-supply: true
> > > + power-domains:
> > > + description:
> > > + List of phandle and PM domain specifier as documented in
> > > + Documentation/devicetree/bindings/power/power_domain.txt
> > > + minItems: 3
> > > + maxItems: 3
> > > + power-domain-names:
> > > + items:
> > > + - const: core0
> > > + - const: core1
> > > + - const: 2d
> >
> > AFAIK, there's no '2d' block in bifrost GPUs. A power domain for each
> > core group is correct though.
>
> Good question... Hopefully Nick/SJ@MTK can comment, the non-upstream DTS has:
> gpu: mali@13040000 {
> compatible = "mediatek,mt8183-mali", "arm,mali-bifrost";
> power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE0>;
> ...
> }
>
> gpu_core1: mali_gpu_core1 {
> compatible = "mediatek,gpu_core1";
> power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE1>;
> };
>
> gpu_core2: mali_gpu_core2 {
> compatible = "mediatek,gpu_core2";
> power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_2D>;
> };
>
> So I picked core0/core1/2d as names, but looking at this, it's likely
> core2 is more appropriate (and MT8183_POWER_DOMAIN_MFG_2D might just
> be a internal/legacy name, if there is no real 2d domain).
>
> Thanks.
>
> > Rob
On Fri, Mar 6, 2020 at 10:34 AM Nick Fan <[email protected]> wrote:
>
> Sorry for my late reply.
> I have checked internally.
> The MT8183_POWER_DOMAIN_MFG_2D is just a legacy name, not really 2D
> domain.
>
> If the naming too confusing, we can change this name to
> MT8183_POWER_DOMAIN_MFG_CORE2 for consistency.
Thanks! I think I'll keep MT8183_POWER_DOMAIN_MFG_2D (that's fine if
that's the domain name you use internally in your HW design), but I'll
modify power-domain-names to core0/1/2 in the binding.
> Thanks
>
> Nick Fan
>
> On Wed, 2020-02-26 at 08:55 +0800, Nicolas Boichat wrote:
>
> > +Nick Fan +Sj Huang @ MTK
> >
> > On Wed, Feb 26, 2020 at 1:16 AM Rob Herring <[email protected]> wrote:
> > >
> > > On Fri, Feb 07, 2020 at 01:26:21PM +0800, Nicolas Boichat wrote:
> > > > Define a compatible string for the Mali Bifrost GPU found in
> > > > Mediatek's MT8183 SoCs.
> > > >
> > > > Signed-off-by: Nicolas Boichat <[email protected]>
> > > > Reviewed-by: Alyssa Rosenzweig <[email protected]>
> > > > ---
> > > >
> > > > v4:
> > > > - Add power-domain-names description
> > > > (kept Alyssa's reviewed-by as the change is minor)
> > > > v3:
> > > > - No change
> > > >
> > > > .../bindings/gpu/arm,mali-bifrost.yaml | 25 +++++++++++++++++++
> > > > 1 file changed, 25 insertions(+)
> > > >
> > > > diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > > index 4ea6a8789699709..0d93b3981445977 100644
> > > > --- a/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > > +++ b/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml
> > > > @@ -17,6 +17,7 @@ properties:
> > > > items:
> > > > - enum:
> > > > - amlogic,meson-g12a-mali
> > > > + - mediatek,mt8183-mali
> > > > - realtek,rtd1619-mali
> > > > - rockchip,px30-mali
> > > > - const: arm,mali-bifrost # Mali Bifrost GPU model/revision is fully discoverable
> > > > @@ -62,6 +63,30 @@ allOf:
> > > > minItems: 2
> > > > required:
> > > > - resets
> > > > + - if:
> > > > + properties:
> > > > + compatible:
> > > > + contains:
> > > > + const: mediatek,mt8183-mali
> > > > + then:
> > > > + properties:
> > > > + sram-supply: true
> > > > + power-domains:
> > > > + description:
> > > > + List of phandle and PM domain specifier as documented in
> > > > + Documentation/devicetree/bindings/power/power_domain.txt
> > > > + minItems: 3
> > > > + maxItems: 3
> > > > + power-domain-names:
> > > > + items:
> > > > + - const: core0
> > > > + - const: core1
> > > > + - const: 2d
> > >
> > > AFAIK, there's no '2d' block in bifrost GPUs. A power domain for each
> > > core group is correct though.
> >
> > Good question... Hopefully Nick/SJ@MTK can comment, the non-upstream DTS has:
> > gpu: mali@13040000 {
> > compatible = "mediatek,mt8183-mali", "arm,mali-bifrost";
> > power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE0>;
> > ...
> > }
> >
> > gpu_core1: mali_gpu_core1 {
> > compatible = "mediatek,gpu_core1";
> > power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_CORE1>;
> > };
> >
> > gpu_core2: mali_gpu_core2 {
> > compatible = "mediatek,gpu_core2";
> > power-domains = <&scpsys MT8183_POWER_DOMAIN_MFG_2D>;
> > };
> >
> > So I picked core0/core1/2d as names, but looking at this, it's likely
> > core2 is more appropriate (and MT8183_POWER_DOMAIN_MFG_2D might just
> > be a internal/legacy name, if there is no real 2d domain).
> >
> > Thanks.
> >
> > > Rob
>
On Thu, Mar 5, 2020 at 8:34 PM Nick Fan <[email protected]> wrote:
>
> Sorry for my late reply.
> I have checked internally.
> The MT8183_POWER_DOMAIN_MFG_2D is just a legacy name, not really 2D
> domain.
>
> If the naming too confusing, we can change this name to
> MT8183_POWER_DOMAIN_MFG_CORE2 for consistency.
Can you clarify what's in each domain? Are there actually 3 shader
cores (IIRC, that should be discoverable)?
Rob
On Fri, Mar 06, 2020 at 02:13:08PM +0000, Rob Herring wrote:
> On Thu, Mar 5, 2020 at 8:34 PM Nick Fan <[email protected]> wrote:
> >
> > Sorry for my late reply.
> > I have checked internally.
> > The MT8183_POWER_DOMAIN_MFG_2D is just a legacy name, not really 2D
> > domain.
> >
> > If the naming too confusing, we can change this name to
> > MT8183_POWER_DOMAIN_MFG_CORE2 for consistency.
>
> Can you clarify what's in each domain? Are there actually 3 shader
> cores (IIRC, that should be discoverable)?
The cover letter from Nicolas includes:
> [ 501.321752] panfrost 13040000.gpu: shader_present=0x7 l2_present=0x1
0x7 is three bits set, so it certainly looks like there are 3 shader
cores. Of course I wouldn't guarantee that it is as simple as each power
domain has a shader core in. The job manager and tiler also need to be
powered somehow, so they are either sharing with a shader core or
there's something more complex going on.
Steve
On Fri, 2020-03-06 at 14:43 +0000, Steven Price wrote:
> On Fri, Mar 06, 2020 at 02:13:08PM +0000, Rob Herring wrote:
> > On Thu, Mar 5, 2020 at 8:34 PM Nick Fan <[email protected]> wrote:
> > >
> > > Sorry for my late reply.
> > > I have checked internally.
> > > The MT8183_POWER_DOMAIN_MFG_2D is just a legacy name, not really 2D
> > > domain.
> > >
> > > If the naming too confusing, we can change this name to
> > > MT8183_POWER_DOMAIN_MFG_CORE2 for consistency.
> >
> > Can you clarify what's in each domain? Are there actually 3 shader
> > cores (IIRC, that should be discoverable)?
>
> The cover letter from Nicolas includes:
>
> > [ 501.321752] panfrost 13040000.gpu: shader_present=0x7 l2_present=0x1
>
> 0x7 is three bits set, so it certainly looks like there are 3 shader
> cores. Of course I wouldn't guarantee that it is as simple as each power
> domain has a shader core in. The job manager and tiler also need to be
> powered somehow, so they are either sharing with a shader core or
> there's something more complex going on.
>
> Steve
>
There are actually five power domains in total for MT8183 GPU.
There are 3 shader cores in MT8183.
They can be listed as following for each power domain:
1.MT8183_POWER_DOMAIN_MFG_ASYNC : SOC bus logic
2.MT8183_POWER_DOMAIN_MFG : GPU job manager & tiler
3.MT8183_POWER_DOMAIN_MFG_CORE0 : GPU shader core 0
4.MT8183_POWER_DOMAIN_MFG_CORE1 : GPU shader core 1
5.MT8183_POWER_DOMAIN_MFG_2D : GPU shader core 2
There are other power domain dependency can be reference in the
following link.
https://lkml.org/lkml/2019/2/1/166
You can check the power domain dependencies as following
==========================================================
+static const struct scp_subdomain scp_subdomain_mt8183[] = {
+ {MT8183_POWER_DOMAIN_MFG_ASYNC, MT8183_POWER_DOMAIN_MFG},
+ {MT8183_POWER_DOMAIN_MFG, MT8183_POWER_DOMAIN_MFG_2D},
+ {MT8183_POWER_DOMAIN_MFG, MT8183_POWER_DOMAIN_MFG_CORE0},
+ {MT8183_POWER_DOMAIN_MFG, MT8183_POWER_DOMAIN_MFG_CORE1},
==========================================================
Thanks
Nick Fan