Method for SOC performance and power optimization

1. A semiconductor chip comprising:

a first processing unit and a second processing unit; and

a power manager; and

a memory;

wherein in response to determining a first activity level for the first processing unit is above a first threshold and a second activity level for the second processing unit is below a second threshold, the power manager is configured to place a limit on a highest power-performance state (P-state) used by the second processing unit;

wherein the power manager is further configured to reduce at least one of a memory width and a memory request rate of the second processing unit to the memory.

2. The semiconductor chip as recited in claim 1 , wherein the second processing unit comprises a cache, wherein the power manager is further configured to flush the cache within the second processing unit.

3. The semiconductor chip as recited in claim 1 , wherein each of the first processing unit and the second processing unit is at least one of the following: a general-purpose central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a digital signal processor (DSP), and a field programmable gate array (FPGA).

4. The semiconductor chip as recited in claim 3 , wherein at least one of the first activity level and the second activity level is a status of an instruction retirement pointer measured over a time interval.

5. The semiconductor chip as recited in claim 3 , wherein at least one of the first activity level and the second activity level is a number of utilized computation units in one or more pipeline stages measured over a time interval.

6. The semiconductor chip as recited in claim 1 , wherein the power manager is further configured to change a P-state used by the first processing unit to a higher P-state, wherein a sum of power consumed by the first processing unit at the higher P-state and the second processing unit at a P-state not above said limit is below a thermal design power (TDP) of the semiconductor chip.

7. The semiconductor chip as recited in claim 6 , wherein the power manager is further configured to:

compute an accumulated energy margin based on said sum of power and said TDP, at each of a first time interval; and

reduce the accumulated energy margin by a given amount, wherein the given amount corresponds to energy dissipating over time in the semiconductor chip, at each of a second time interval larger then the first time interval.

8. The semiconductor chip as recited in claim 7 , wherein the semiconductor chip is coupled to each one of a display device and a battery, wherein the power manager is further configured to be toggled on and off, wherein in response to the power manager is toggled on, the power manager causes:

the battery has less power drawn than when the power manager is toggled off; and

the display device is driven by the semiconductor chip for a longer period of time than when the power manager is toggled off.

9. The semiconductor chip as recited in claim 7 , wherein in response to determining the second activity level is above the second threshold, the power manager is further configured to:

change a P-state used by the second processing unit to a higher P-state, wherein said sum of power is below said TDP; and

allow the second processing unit to store data in the cache.

10. A method comprising:

determining a first activity level for a first processing unit of an integrated circuit (IC) is above a first threshold, and a second activity level for a second processing unit in the IC is below a second threshold; and

placing a limit on a highest power-performance state (P-state) used by the second processing unit;

wherein the IC further comprises a memory coupled to each of the first processing unit and the second processing unit and wherein the method further comprises reducing at least one of a memory width and a memory request rate of the second processing unit to the memory.

11. The method as recited in claim 10 , wherein the second processing unit comprises a cache, further comprising flushing the cache within the second processing unit.

12. The method as recited in claim 10 , wherein each of the first processing unit and the second processing unit is at least one of the following: a general-purpose central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a digital signal processor (DSP), and a field programmable gate array (FPGA).

13. The method as recited in claim 12 , wherein at least one of the first activity level and the second activity level is a status of an instruction retirement pointer measured over a time interval.

14. The method as recited in claim 10 , further comprising changing a P-state used by the first processing unit to a higher performance P-state, wherein a sum of power consumed by the first processing unit at the higher P-state and the second processing unit at a P-state not above said limit is below a thermal design power (TDP) of the IC.

15. The method as recited in claim 14 , further comprising:

computing an accumulated energy margin based on said sum of power and said TDP, at each of a first time interval; and

reducing the accumulated energy margin by a given amount, wherein the given amount corresponds to energy dissipating over time in the semiconductor chip, at each of a second time interval larger than the first time interval.

16. A power management unit (PMU) for use in an integrated circuit (IC) comprising:

an interface to each of a first processing unit and a second processing unit; and

control circuitry configured to in response to determining a first activity level for the first processing unit is above a first threshold and a second activity level for the second processing unit is below a second threshold, place a limit on a highest power-performance state (P-state) used by the second processing unit;

wherein the circuitry is further configured to reduce at least one of a memory width and a memory request rate of the second processing unit to a memory.

17. The PMU as recited in claim 16 , wherein the second processing unit comprises a cache, wherein the control circuitry is further configured to flush the cache within the second processing unit.