On 12.07.2023 22:39, Krzysztof Kozlowski wrote: > On 11/07/2023 14:18, Konrad Dybcio wrote: >> It makes zero (or less) sense to consume BCM voters per interconnect >> provider. They are shared throughout the entire system and it's enough >> to keep a single reference to each of them. >> >> Storing them in a shared array at fixed indices will let us improve both >> the representation of the RPMh architecture (every RSC can hold a resource >> vote on any bus, they're not limited in that regard) and save as much as >> kilobytes worth of RAM. >> >> Signed-off-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> >> --- >> include/dt-bindings/interconnect/qcom,icc.h | 8 ++++++++ >> 1 file changed, 8 insertions(+) >> >> diff --git a/include/dt-bindings/interconnect/qcom,icc.h b/include/dt-bindings/interconnect/qcom,icc.h >> index cd34f36daaaa..9c13ef8a044e 100644 >> --- a/include/dt-bindings/interconnect/qcom,icc.h >> +++ b/include/dt-bindings/interconnect/qcom,icc.h >> @@ -23,4 +23,12 @@ >> #define QCOM_ICC_TAG_ALWAYS (QCOM_ICC_TAG_AMC | QCOM_ICC_TAG_WAKE |\ >> QCOM_ICC_TAG_SLEEP) >> >> +#define ICC_BCM_VOTER_APPS 0 >> +#define ICC_BCM_VOTER_DISP 1 >> +#define ICC_BCM_VOTER_CAM0 2 >> +#define ICC_BCM_VOTER_CAM1 3 >> +#define ICC_BCM_VOTER_CAM2 4 >> + >> +#define ICC_BCM_VOTER_MAX 64 > > I proposed to skip the max. If you actually use it, you won't be able to > change it ever. I guess I can just add the max in the driver. Konrad > > > Best regards, > Krzysztof >
