Support independent CPU and GPU workloads on the same node #331
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Worklaods can now run CPU and GPU tasks on each host.
The caveat is that the fragments for these tasks need to be independent.
Implementation Notes ⚒️
GPU is called accel in most places because CPU and GPU only differ by one letter, which can lead to more programming errors.
Using accel instead of GPU guards against that.
The fragments file can now have GPU fragments. These fragments have the is_gpu column set to true. When the flag is set the cpu_usage column is instead used as GPU usage.
The topology now has accel and accelPowerModel fields. The options for these are the same as those for the cpu and cpuPowerModel fields.
Check
gpu/single_1_2000.json
topology file in experiments-base test resources for an example.The GPU (accel) implementation uses many classes with CPU in the name, as the required parameters are the same. For example, CpuPowerModel, HostCpuStats, etc. We can rename them to generalize better in the future.
External Dependencies 🍀
Breaking API Changes⚠️
Simply specify none (N/A) if not applicable.