Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fast Optimization for SKU Variations #88

Open
4 tasks
marenz2569 opened this issue Nov 19, 2024 · 0 comments
Open
4 tasks

Fast Optimization for SKU Variations #88

marenz2569 opened this issue Nov 19, 2024 · 0 comments
Labels
backlog optimization improvements Backlog of things that improve the optimization e.g., runtime or effectiveness of the results

Comments

@marenz2569
Copy link
Collaborator

marenz2569 commented Nov 19, 2024

Currently FIRESTARTER needs to be optimized for the highest power consumption with an algorithm that will take a couple hours. This optimization will create highest power consumption on a specific SKU of a CPU generation. This ticket shows a way to optimize FS in two steps that should allow SKU variations and create very high (not highest) power consumption.

  1. Optimize FS on a specific SKU with a lower number of cores. This will allow us to have a bigger power budget per core making it more ideal for those SKUs. For SKUs with a lower budget per core FS will push the CPU into power limit anyway.
  2. This optimization should be done with only cache and register accesses. Prefetches to RAM should be added to causes a level of power consumption from the memory controllers.
  3. The second part of the optimization will be done on any SKU. The results of the previous steps will be taken and RAM instruction incrementally added. Once peak power consumption is reached, the optimization is finished.

DoD

  • Implement a way to use a specific number of thread equally spaced out in the CPU. (i.e. a specific number of cores per CCX on Zen or physically spaced out on the 2D grid on Intel CPUs)
  • Allow optimization of caches and registers with a specific number of RAM prefetches
  • Implement 'fast-optimization' that adjusts the number of RAM accesses.
  • Evaluate the effectiveness of this optimization on different SKUs vs. optimizing for each SKU.
@marenz2569 marenz2569 added the backlog optimization improvements Backlog of things that improve the optimization e.g., runtime or effectiveness of the results label Nov 19, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
backlog optimization improvements Backlog of things that improve the optimization e.g., runtime or effectiveness of the results
Projects
None yet
Development

No branches or pull requests

1 participant