Guilherme Cox
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Guilherme Cox has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Abhishek Bhattacharjee
4 shared publications- Abhishek Bhattacharjee
Yale University, New Haven, Connecticut, USA
3 shared publications Cristiana Bentes
State University of Rio de Janeiro
3 shared publicationsRicardo Farias
Federal University of Rio de Janeiro
3 shared publicationsQingyuan Deng
Facebook Inc., Menlo Park, CA
2 shared publicationsRicardo Bianchini
Microsoft Research, Redmond, WA
2 shared publications
Latest Publications
- Open Access
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 137-152
Forest: Access-aware GPU UVM Management
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 137-152
Open AccessForest: Access-aware GPU UVM Management
- Open Access
- 22 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 53-60
Prefetching Using Principles of Hippocampal-Neocortical Interaction
Memory prefetching improves performance across many systems layers. However, achieving high prefetch accuracy with low overhead is challenging, as memory hierarchies and application memory access patterns become more complicated. Furthermore, a prefetcher's ability to adapt to new access patterns as they emerge is becoming more crucial than ever. Recent work has demonstrated the use of deep learning techniques to improve prefetching accuracy, albeit with impractical compute and storage overheads. This paper suggests taking inspiration from the learning mechanisms and memory architecture of the human brain---specifically, the hippocampus and neocortex---to build resource-efficient, accurate, and adaptable prefetchers.
- 22 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 53-60
Open AccessPrefetching Using Principles of Hippocampal-Neocortical Interaction
Memory prefetching improves performance across many systems layers. However, achieving high prefetch accuracy with low overhead is challenging, as memory hierarchies and application memory access patterns become more complicated. Furthermore, a prefetcher's ability to adapt to new access patterns as they emerge is becoming more crucial than ever. Recent work has demonstrated the use of deep learning techniques to improve prefetching accuracy, albeit with impractical compute and storage overheads. This paper suggests taking inspiration from the learning mechanisms and memory architecture of the human brain---specifically, the hippocampus and neocortex---to build resource-efficient, accurate, and adaptable prefetchers.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 180-192
Scheduling Page Table Walks for Irregular GPU Applications
Recent studies on commercial hardware demonstrated that irregular GPU applications can bottleneck on virtual-to-physical address translations. In this work, we explore ways to reduce address translation overheads for such applications. We discover that the order of servicing GPU's address translation requests (specifically, page table walks) plays a key role in determining the amount of translation overhead experienced by an application. We find that different SIMD instructions executed by an application require vastly different amounts of work to service their address translation needs, primarily depending upon the number of distinct pages they access. We show that better forward progress is achieved by prioritizing translation requests from the instructions that require less work to service their address translation needs. Further, in the GPU's Single-Instruction-Multiple-Thread (SIMT) execution paradigm, all threads that execute in lockstep (wavefront) need to finish operating on their respective data elements (and thus, finish their address translations) before the execution moves ahead. Thus, batching walk requests originating from the same SIMD instruction could reduce unnecessary stalls. We demonstrate that the reordering of translation requests based on the above principles improves the performance of several irregular GPU applications by 30% on average.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 180-192
Scheduling Page Table Walks for Irregular GPU Applications
Recent studies on commercial hardware demonstrated that irregular GPU applications can bottleneck on virtual-to-physical address translations. In this work, we explore ways to reduce address translation overheads for such applications. We discover that the order of servicing GPU's address translation requests (specifically, page table walks) plays a key role in determining the amount of translation overhead experienced by an application. We find that different SIMD instructions executed by an application require vastly different amounts of work to service their address translation needs, primarily depending upon the number of distinct pages they access. We show that better forward progress is achieved by prioritizing translation requests from the instructions that require less work to service their address translation needs. Further, in the GPU's Single-Instruction-Multiple-Thread (SIMT) execution paradigm, all threads that execute in lockstep (wavefront) need to finish operating on their respective data elements (and thus, finish their address translations) before the execution moves ahead. Thus, batching walk requests originating from the same SIMD instruction could reduce unnecessary stalls. We demonstrate that the reordering of translation requests based on the above principles improves the performance of several irregular GPU applications by 30% on average.
- Open Access
- 13 March 2018
- Published by Association for Computing Machinery (ACM)
- p. 236-247
Secure, Consistent, and High-Performance Memory Snapshotting
- 13 March 2018
- Published by Association for Computing Machinery (ACM)
- p. 236-247
Open AccessSecure, Consistent, and High-Performance Memory Snapshotting
- Open Access
- 5 September 2017
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Modeling and Performance Evaluation of Computing Systems
- Vol. 2 (3) , 1-31
Fast Power and Energy Management for Future Many-Core Systems
Future servers will incorporate many active low-power modes for each core and for the main memory subsystem. Though these modes provide flexibility for power and/or energy management via Dynamic Voltage and Frequency Scaling (DVFS), prior work has shown that they must be managed in a coordinated manner. This requirement creates a combinatorial space of possible power mode configurations. As a result, it becomes increasingly challenging to quickly select the configuration that optimizes for both performance and power/energy efficiency. In this article, we propose a novel queuing model for working with the abundant active low-power modes in many-core systems. Based on the queuing model, we derive two fast algorithms that optimize for performance and efficiency using both CPU and memory DVFS. Our first algorithm, called FastCap, maximizes the performance of applications under a full-system power cap, while promoting fairness across applications. Our second algorithm, called FastEnergy, maximizes the full-system energy savings under predefined application performance loss bounds. Both FastCap and FastEnergy operate online and efficiently, using a small set of performance counters as input. To evaluate them, we simulate both algorithms for a many-core server running different types of workloads. Our results show that FastCap achieves better application performance and fairness than prior power capping techniques for the same power budget, whereas FastEnergy conserves more energy than prior energy management techniques for the same performance constraint. FastCap and FastEnergy together demonstrate the applicability of the queuing model for managing the abundant active low-power modes in many-core systems.
- 5 September 2017
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Modeling and Performance Evaluation of Computing Systems
- Vol. 2 (3) , 1-31
Open AccessFast Power and Energy Management for Future Many-Core Systems
Future servers will incorporate many active low-power modes for each core and for the main memory subsystem. Though these modes provide flexibility for power and/or energy management via Dynamic Voltage and Frequency Scaling (DVFS), prior work has shown that they must be managed in a coordinated manner. This requirement creates a combinatorial space of possible power mode configurations. As a result, it becomes increasingly challenging to quickly select the configuration that optimizes for both performance and power/energy efficiency. In this article, we propose a novel queuing model for working with the abundant active low-power modes in many-core systems. Based on the queuing model, we derive two fast algorithms that optimize for performance and efficiency using both CPU and memory DVFS. Our first algorithm, called FastCap, maximizes the performance of applications under a full-system power cap, while promoting fairness across applications. Our second algorithm, called FastEnergy, maximizes the full-system energy savings under predefined application performance loss bounds. Both FastCap and FastEnergy operate online and efficiently, using a small set of performance counters as input. To evaluate them, we simulate both algorithms for a many-core server running different types of workloads. Our results show that FastCap achieves better application performance and fairness than prior power capping techniques for the same power budget, whereas FastEnergy conserves more energy than prior energy management techniques for the same performance constraint. FastCap and FastEnergy together demonstrate the applicability of the queuing model for managing the abundant active low-power modes in many-core systems.
- Open Access
- 24 June 2017
- Published by Association for Computing Machinery (ACM)
- No. 1,p. 430-443
Hardware Translation Coherence for Virtualized Systems
To improve system performance, operating systems (OSes) often undertake activities that require modification of virtual-to-physical address translations. For example, the OS may migrate data between physical pages to manage heterogeneous memory devices. We refer to such activities as page remappings. Unfortunately, page remappings are expensive. We show that a big part of this cost arises from address translation coherence, particularly on systems employing virtualization. In response, we propose hardware translation invalidation and coherence or HATRIC, a readily implementable hardware mechanism to piggyback translation coherence atop existing cache coherence protocols. We perform detailed studies using KVM-based virtualization, showing that HATRIC achieves up to 30% performance and 10% energy benefits, for per-CPU area overheads of 0.2%. We also quantify HATRIC's benefits on systems running Xen and find up to 33% performance improvements.
- 24 June 2017
- Published by Association for Computing Machinery (ACM)
- No. 1,p. 430-443
Open AccessHardware Translation Coherence for Virtualized Systems
To improve system performance, operating systems (OSes) often undertake activities that require modification of virtual-to-physical address translations. For example, the OS may migrate data between physical pages to manage heterogeneous memory devices. We refer to such activities as page remappings. Unfortunately, page remappings are expensive. We show that a big part of this cost arises from address translation coherence, particularly on systems employing virtualization. In response, we propose hardware translation invalidation and coherence or HATRIC, a readily implementable hardware mechanism to piggyback translation coherence atop existing cache coherence protocols. We perform detailed studies using KVM-based virtualization, showing that HATRIC achieves up to 30% performance and 10% energy benefits, for per-CPU area overheads of 0.2%. We also quantify HATRIC's benefits on systems running Xen and find up to 33% performance improvements.
- Open Access
- 4 April 2017
- Published by Association for Computing Machinery (ACM) in ACM SIGARCH Computer Architecture News
- Vol. 45 (1) , 435-448
Efficient Address Translation for Architectures with Multiple Page Sizes
Processors and operating systems (OSes) support multiple memory page sizes. Superpages increase Translation Lookaside Buffer (TLB) hits, while small pages provide fine-grained memory protection. Ideally, TLBs should perform well for any distribution of page sizes. In reality, set-associative TLBs -- used frequently for their energy efficiency compared to fully-associative TLBs -- cannot (easily) support multiple page sizes concurrently. Instead, commercial systems typically implement separate set-associative TLBs for different page sizes. This means that when superpages are allocated aggressively, TLB misses may, counter intuitively, increase even if entries for small pages remain unused (and vice-versa). We invent MIX TLBs, energy-frugal set-associative structures that concurrently support all page sizes by exploiting superpage allocation patterns. MIX TLBs boost the performance (often by 10-30%) of big-memory applications on native CPUs, virtualized CPUs, and GPUs. MIX TLBs are simple and require no OS or program changes.
- 4 April 2017
- Published by Association for Computing Machinery (ACM) in ACM SIGARCH Computer Architecture News
- Vol. 45 (1) , 435-448
Open AccessEfficient Address Translation for Architectures with Multiple Page Sizes
Processors and operating systems (OSes) support multiple memory page sizes. Superpages increase Translation Lookaside Buffer (TLB) hits, while small pages provide fine-grained memory protection. Ideally, TLBs should perform well for any distribution of page sizes. In reality, set-associative TLBs -- used frequently for their energy efficiency compared to fully-associative TLBs -- cannot (easily) support multiple page sizes concurrently. Instead, commercial systems typically implement separate set-associative TLBs for different page sizes. This means that when superpages are allocated aggressively, TLB misses may, counter intuitively, increase even if entries for small pages remain unused (and vice-versa). We invent MIX TLBs, energy-frugal set-associative structures that concurrently support all page sizes by exploiting superpage allocation patterns. MIX TLBs boost the performance (often by 10-30%) of big-memory applications on native CPUs, virtualized CPUs, and GPUs. MIX TLBs are simple and require no OS or program changes.
- Open Access
- 4 April 2017
- Published by Association for Computing Machinery (ACM) in ACM SIGOPS Operating Systems Review
- Vol. 51 (2) , 435-448
Efficient Address Translation for Architectures with Multiple Page Sizes
- 4 April 2017
- Published by Association for Computing Machinery (ACM) in ACM SIGOPS Operating Systems Review
- Vol. 51 (2) , 435-448
Open AccessEfficient Address Translation for Architectures with Multiple Page Sizes
- Open Access
- 4 April 2017
- Published by Association for Computing Machinery (ACM)
- p. 435-448
Efficient Address Translation for Architectures with Multiple Page Sizes
Processors and operating systems (OSes) support multiple memory page sizes. Superpages increase Translation Lookaside Buffer (TLB) hits, while small pages provide fine-grained memory protection. Ideally, TLBs should perform well for any distribution of page sizes. In reality, set-associative TLBs -- used frequently for their energy efficiency compared to fully-associative TLBs -- cannot (easily) support multiple page sizes concurrently. Instead, commercial systems typically implement separate set-associative TLBs for different page sizes. This means that when superpages are allocated aggressively, TLB misses may, counter intuitively, increase even if entries for small pages remain unused (and vice-versa). We invent MIX TLBs, energy-frugal set-associative structures that concurrently support all page sizes by exploiting superpage allocation patterns. MIX TLBs boost the performance (often by 10-30%) of big-memory applications on native CPUs, virtualized CPUs, and GPUs. MIX TLBs are simple and require no OS or program changes.
- 4 April 2017
- Published by Association for Computing Machinery (ACM)
- p. 435-448
Open AccessEfficient Address Translation for Architectures with Multiple Page Sizes
Processors and operating systems (OSes) support multiple memory page sizes. Superpages increase Translation Lookaside Buffer (TLB) hits, while small pages provide fine-grained memory protection. Ideally, TLBs should perform well for any distribution of page sizes. In reality, set-associative TLBs -- used frequently for their energy efficiency compared to fully-associative TLBs -- cannot (easily) support multiple page sizes concurrently. Instead, commercial systems typically implement separate set-associative TLBs for different page sizes. This means that when superpages are allocated aggressively, TLB misses may, counter intuitively, increase even if entries for small pages remain unused (and vice-versa). We invent MIX TLBs, energy-frugal set-associative structures that concurrently support all page sizes by exploiting superpage allocation patterns. MIX TLBs boost the performance (often by 10-30%) of big-memory applications on native CPUs, virtualized CPUs, and GPUs. MIX TLBs are simple and require no OS or program changes.
- 1 April 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 57-68
FastCap: An efficient and fair algorithm for power capping in many-core systems
Future servers will incorporate many active low-power modes for different system components, such as cores and memory. Though these modes provide flexibility for power management via Dynamic Voltage and Frequency Scaling (DVFS), they must be operated in a coordinated manner. Such coordinated control creates a combinatorial space of possible power mode configurations. Given the rapid growth of the number of cores, it is becoming increasingly challenging to quickly select the configuration that maximizes the performance under a given power budget. Prior power capping techniques do not scale well to large numbers of cores, and none of those works has considered memory DVFS. In this paper, we present FastCap, our optimization approach for system-wide power capping, using both CPU and memory DVFS. Based on a queuing model, FastCap formulates power capping as a non-linear optimization problem where we seek to maximize the system performance under a power budget, while promoting fairness across applications. Our FastCap algorithm solves the optimization online and efficiently (low complexity on the number of cores), using a small set of performance counters as input. To evaluate FastCap, we simulate it for a many-core server running different types of workloads. Our results show that FastCap caps power draw accurately, while producing better application performance and fairness than many existing CPU power capping methods (even after they are extended to use of memory DVFS as well).
- 1 April 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 57-68
FastCap: An efficient and fair algorithm for power capping in many-core systems
Future servers will incorporate many active low-power modes for different system components, such as cores and memory. Though these modes provide flexibility for power management via Dynamic Voltage and Frequency Scaling (DVFS), they must be operated in a coordinated manner. Such coordinated control creates a combinatorial space of possible power mode configurations. Given the rapid growth of the number of cores, it is becoming increasingly challenging to quickly select the configuration that maximizes the performance under a given power budget. Prior power capping techniques do not scale well to large numbers of cores, and none of those works has considered memory DVFS. In this paper, we present FastCap, our optimization approach for system-wide power capping, using both CPU and memory DVFS. Based on a queuing model, FastCap formulates power capping as a non-linear optimization problem where we seek to maximize the system performance under a power budget, while promoting fairness across applications. Our FastCap algorithm solves the optimization online and efficiently (low complexity on the number of cores), using a small set of performance counters as input. To evaluate FastCap, we simulate it for a many-core server running different types of workloads. Our results show that FastCap caps power draw accurately, while producing better application performance and fairness than many existing CPU power capping methods (even after they are extended to use of memory DVFS as well).