Isotonic additive models in workload characterization
A method for modeling a system configuration, including isotonizing an unconstrained additive model of a cache architecture to obtain an isotonic additive model for the cache architecture, wherein the isotonic additive model is of the same functional form as the unconstrained additive model, smoothing the isotonic additive model using a flat spot technique to obtain a characterization of the cache architecture, and modeling a system configuration using the characterization of the cache architecture.
1. A method for modeling a system configuration, comprising:
isotonizing an unconstrained additive model of a cache architecture to obtain an isonotic additive model for the cache architecture, wherein the isotonic addictive model is of the same functional form as the unconstrained additive model;
smoothing the isotonic additive model using a flat spot technique to obtain a characterization of the cache architecture; and
modeling a system configuration using the characterization of the cache architecture.
2. The method of claim 1 , further comprising:
adding a plurality of interpretability components to the characterization of the cache architecture, wherein the plurality of interpretability components and the characterization of the cache architecture are used in modeling the system configuration.
3. The method of claim 1 , wherein isotonizing comprises:
obtaining a first target additive component in the unconstrained additive model;
isotonizing the first target additive component to obtain a first isotonized additive component;
obtaining a second target additive component from the unconstrained additive model and the first isotonized additive component; and
isotonizing the second target additive component to obtain a second isotonized additive component.
4. The method of claim 3 , wherein isotonizing uses a pool-adjacent violators algorithm.
5. The method of claim 4 , wherein the pool-adjacent violators algorithm uses a least squares distance as an optimization criterion.
6. The method of claim 4 , wherein the pool-adjacent violators algorithm uses a weighted least squares as an optimization criterion.
7. The method of claim 3 , wherein isotonizing further comprises:
computing a first target coordinate effect for the first target additive component;
applying a pool-adjacent violators algorithm in a direction of the first target coordinate effect for the first target additive component; and
applying the pool-adjacent violators algorithm in a direction of a second target coordinate.
8. A computer-readable medium having recorded thereon instructions executable by a processor to perform a cache architecture simulation, the instructions for:
isotonizing an unconstrained additive model of a cache architecture to obtain an isotonic additive model for the cache architecture, wherein the isotonic additive model is of the same functional form as the unconstrained additive model;
smoothing the isotonic additive model using a flat spot technique to obtain a characterization of the cache architecture; and
modeling a system configuration using the characterization of the cache architecture.
9. The computer-readable medium of claim 8 , further comprising instructions for:
adding a plurality of interpretability components to the characterization of the cache architecture, wherein the plurality of interpretability components and the characterization of the cache architecture are used in modeling the system configuration.
10. The computer-readable medium of claim 8 , further comprising instructions for:
obtaining a first target additive component in the unconstrained additive model;
isotonizing the first target additive component to obtain a first isotonized additive component;
obtaining a second target additive component from the unconstrained additive model and the first isotonized additive component; and
isotonizing the second target additive component to obtain a second isotonized additive component.
11. The computer-readable medium of claim 10 , wherein isotonizing uses a pool-adjacent violators algorithm.
12. The computer-readable medium of claim 11 , wherein the pool-adjacent violators algorithm uses a least squares distance as an optimization criterion.
13. The computer-readable medium of claim 10 , further comprising instructions for:
computing a first target coordinate effect for the first target additive component;
applying a pool-adjacent violators algorithm in a direction of the first target coordinate effect for the first target additive component; and
applying the pool-adjacent violators algorithm in a direction of a second target coordinate.
14. A computer system for cache architecture simulation, comprising:
a processor;
a memory;
a display device; and
software instructions stored in the memory for enabling the computer system under control of the processor, to perform:
isotonizing an unconstrained additive model of a cache architecture to obtain an isotonic additive model for the cache architecture, wherein the isotonic additive model is of the same functional form as the unconstrained additive model;
smoothing the isotonic additive model using a flat spot technique to obtain a characterization of the cache architecture; and
modeling a system configuration using the characterization of the cache architecture.
15. The computer system of claim 14 , further comprising instructions for: adding a plurality of interpretability components to the characterization of the cache architecture, wherein the plurality of interpretability components and the characterization of the cache architecture are used in modeling the system configuration.
16. The computer system of claim 14 , further comprising instructions for:
obtaining a first target additive component in the unconstrained additive model;
isotonizing the first target additive component to obtain a first isotonized additive component;
obtaining a second target additive component from the unconstrained additive model and the first isotonized additive component; and
isotonizing the second target additive component to obtain a second isotonized additive component.
17. The computer system of claim 16 , wherein isotonizing uses a pool adjacent violators algorithm.
18. The computer system of claim 17 , wherein pool-adjacent violators algorithm uses a least squares distance as an optimization criterion.
19. The computer system of claim 17 , wherein pool-adjacent violators algorithm uses a weighted least squares as an optimization criterion.
20. The computer system of claim 16 , further comprising instructions for:
computing a first target coordinate effect for the first target additive component;
applying a pool-adjacent violators algorithm in a direction of the first target coordinate effect for the first target additive component; and
applying the pool-adjacent violators algorithm in a direction of a second target
coordinate.