SMT Processor Kernel Scheduling for ST Workloads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SMT processor systems face inefficiencies due to time-consuming mode switching between SMT and ST modes, leading to suboptimal performance for workloads that require single-threaded processing, and idle processors in SMT mode, where ST workloads wait until SMT workloads complete.
Innovation Solution
The system manages mixed ST and SMT workloads by creating an exclusive set of processors for ST workloads, where only one hardware thread per processor is active, and the remaining threads are placed in a snooze mode, allowing ST workloads to run efficiently without switching modes, while SMT workloads continue on other processors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the operating system switches between SMT mode and ST mode based on workload type, then the performance efficiency for specific workload types is improved, but the mode switching process is time-consuming and complex
Solution Approach 1:
The patent segments the processor resources by creating an exclusive set of processors for ST workloads, separating them from the pool of processors available for SMT workloads. This allows ST workloads to be isolated to specific processors without requiring system-wide mode switching, thereby maintaining performance efficiency while eliminating the time-consuming switching process.
Solution Approach 2:
The patent implements dynamic processor allocation where processors can be dynamically assigned to exclusive sets for ST workloads or made available for SMT workloads based on current workload demands. This dynamic allocation eliminates the need for static mode switching between SMT and ST modes, resolving the contradiction between performance efficiency and switching time.
2Productivity
If the operating system switches between SMT mode and ST mode, then ST workloads can run efficiently, but the reconfiguration process requires dynamic reconfiguration events that are time-consuming
Solution Approach 1:
The patent creates an exclusive set of processors for ST workloads, segmenting the processor pool into dedicated ST processors and shared SMT processors. This segmentation allows ST workloads to run efficiently on dedicated processors without triggering complex system-wide reconfiguration events, thereby reducing reconfiguration complexity while maintaining ST workload efficiency.
Solution Approach 2:
The patent extracts ST workload requirements from the general SMT processor pool by creating a separate exclusive set of processors for ST workloads. This extraction eliminates the need for complex dynamic reconfiguration events that would otherwise be required to switch between SMT and ST modes, reducing reconfiguration complexity while preserving ST workload performance.
3Productivity
If all processors run in SMT mode, then hardware resource utilization is maximized, but ST workloads must wait until SMT workloads complete before mode switching can occur
Solution Approach 1:
The patent segments processors into an exclusive set for ST workloads and a shared set for SMT workloads. This segmentation allows both ST and SMT workloads to execute concurrently on separate processor sets, eliminating the waiting time ST workloads would otherwise experience while maintaining high hardware resource utilization through parallel execution of both workload types.
Solution Approach 2:
The patent implements dynamic processor set allocation where the exclusive set of processors for ST workloads is dynamically managed alongside the shared SMT processor pool. This dynamic management enables concurrent execution of ST and SMT workloads without mode switching, maximizing hardware resource utilization while eliminating ST workload waiting time.
4Adaptability or versatility
If dynamic reconfiguration events are performed on each processor, then hardware threads can be activated or removed, but the process becomes increasingly limited as the number of processors increases
Solution Approach 1:
The patent segments the processor configuration approach by creating an exclusive set of processors for ST workloads that can be independently managed without triggering reconfiguration events on other processors. This segmentation maintains hardware thread configuration flexibility for ST workloads while eliminating the productivity penalty associated with performing reconfiguration events on all processors, especially as the number of processors increases.
Data Source
AI summary
A SMT enabled processor system, having multiple processors each activated to interleave execution of multiple hardware threads on each processor, for ST workload, includes a kernel. The kernel of the SMT enabled processor system facilitates construction of an exclusive set of processors to simulate an ST mode for handling the tasks of the ST workload, wherein the ST workload runs more efficiently on single threaded processors. The kernel schedules the ST workload on the exclusive set of processors by selecting one hardware thread per processor within said exclusive set of processors to handle a separate one of the tasks of the ST workload, while requiring the remaining hardware threads per processor within the exclusive set to idle. The ST workload is executed on the SMT enabled processor system as if the exclusive set of processors run in ST mode, but without actually removing the remaining idle hardware threads per processor within the exclusive set of processors.