Lei Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Lei Liu has a total of 42 co-authors. Here are the top co-authors by number of shared publications:
- Limin Xiao
Beihang University
6 shared publications Xinglei Dou
ABSTACT Beihang University China
4 shared publications- Xinglei Dou
Beihang University
4 shared publications Li Chen
SKLCA, ICT, CAS, Beijing, China
3 shared publications- Liang Wang
Beihang University
3 shared publications Limin Xiao
ABSTACT Beihang University China
3 shared publications
Latest Publications
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 121-122
EvoInfer: A Self-Evolution and ML-Driven Approach for Optimizing LLM Performance on Mobiles
Deploying LLMs on mobile devices has emerged as a compelling trend. Yet, mobile systems’ resources are limited, and the resources available on a specific device vary during runtime, as many concurrent applications exhibit diverse requirements. Existing approaches for optimizing LLM performance, such as sparse attention mechanisms, often rely on static LLM workloads that do not change during a single inference process. In this paper, we design LLM-Ev, which dynamically adjusts the LLM workload based on real-time system status to achieve high throughput on mobile devices. Moreover, we observe that memory bandwidth contention from concurrent applications can be a bottleneck for on-device LLMs, and the existing Android default memory frequency governor can use improvements for LLMs. Therefore, we design LLM-Dfm, an ML-driven memory frequency scaling approach that uses a customized RL model to execute frequency-scaling actions based on real-time system status, improving bandwidth utilization. Finally, we design a Vertical Orchestration Engine to coordinate LLM-Ev and LLM-Dfm. We denote the three components together as EvoInfer. We implement EvoInfer on Google Pixel 8-series mobile phones. Experimental results show that, EvoInfer improves prefill throughput by 18.3% (up to 25.8%), decode throughput by 29.6% (up to 32.4%), on average.
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 121-122
Open AccessEvoInfer: A Self-Evolution and ML-Driven Approach for Optimizing LLM Performance on Mobiles
Deploying LLMs on mobile devices has emerged as a compelling trend. Yet, mobile systems’ resources are limited, and the resources available on a specific device vary during runtime, as many concurrent applications exhibit diverse requirements. Existing approaches for optimizing LLM performance, such as sparse attention mechanisms, often rely on static LLM workloads that do not change during a single inference process. In this paper, we design LLM-Ev, which dynamically adjusts the LLM workload based on real-time system status to achieve high throughput on mobile devices. Moreover, we observe that memory bandwidth contention from concurrent applications can be a bottleneck for on-device LLMs, and the existing Android default memory frequency governor can use improvements for LLMs. Therefore, we design LLM-Dfm, an ML-driven memory frequency scaling approach that uses a customized RL model to execute frequency-scaling actions based on real-time system status, improving bandwidth utilization. Finally, we design a Vertical Orchestration Engine to coordinate LLM-Ev and LLM-Dfm. We denote the three components together as EvoInfer. We implement EvoInfer on Google Pixel 8-series mobile phones. Experimental results show that, EvoInfer improves prefill throughput by 18.3% (up to 25.8%), decode throughput by 29.6% (up to 32.4%), on average.
- Open Access
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 22 (1) , 1-29
Is Intelligence the Right Direction in New OS Scheduling for Multiple Resources in Cloud Environments?
Making it intelligent is a promising way in System/OS design. This article proposes OSML+, a new ML-based resource scheduling mechanism for co-located cloud services. OSML+ intelligently schedules the cache and main memory bandwidth resources at the memory hierarchy and the computing core resources simultaneously. OSML+ uses a multi-model collaborative learning approach during its scheduling and thus can handle complicated cases, e.g., avoiding resource cliffs, sharing resources among applications, enabling different scheduling policies for applications with different priorities, and so on. OSML+ can converge faster using ML models than previous studies. Moreover, OSML+ can automatically learn on the fly and handle dynamically changing workloads accordingly. Using transfer learning technologies, we show our design can work well across various cloud servers, including the latest off-the-shelf large-scale servers. Our experimental results show that OSML+ supports higher loads and meets QoS targets with lower overheads than previous studies.
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 22 (1) , 1-29
Open AccessIs Intelligence the Right Direction in New OS Scheduling for Multiple Resources in Cloud Environments?
Making it intelligent is a promising way in System/OS design. This article proposes OSML+, a new ML-based resource scheduling mechanism for co-located cloud services. OSML+ intelligently schedules the cache and main memory bandwidth resources at the memory hierarchy and the computing core resources simultaneously. OSML+ uses a multi-model collaborative learning approach during its scheduling and thus can handle complicated cases, e.g., avoiding resource cliffs, sharing resources among applications, enabling different scheduling policies for applications with different priorities, and so on. OSML+ can converge faster using ML models than previous studies. Moreover, OSML+ can automatically learn on the fly and handle dynamically changing workloads accordingly. Using transfer learning technologies, we show our design can work well across various cloud servers, including the latest off-the-shelf large-scale servers. Our experimental results show that OSML+ supports higher loads and meets QoS targets with lower overheads than previous studies.
- Open Access
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 854-868
Amove: Accelerating LLMs through Mitigating Outliers and Salient Points via Fine-Grained Grouped Vectorized Data Type
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 854-868
Open AccessAmove: Accelerating LLMs through Mitigating Outliers and Salient Points via Fine-Grained Grouped Vectorized Data Type
- Open Access
- 19 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (3) , 1-24
LarQucut: A New Cutting and Mapping Approach for Large-sized Quantum Circuits in Distributed Quantum Computing (DQC) Environments
Distributed quantum computing (DQC) is a promising way to achieve large-scale quantum computing. However, mapping large-sized quantum circuits in DQC is a challenging job; for example, it is difficult to find an ideal cutting and mapping solution when many qubits, complicated qubit operations, and diverse QPUs are involved. In this study, we propose LarQucut, a new quantum circuit cutting and mapping approach for large-sized circuits in DQC. LarQucut has several new designs. (1) LarQucut can have cutting solutions that use fewer cuts, and it does not cut a circuit into independent sub-circuits, therefore reducing the overall cutting and computing overheads. (2) LarQucut finds isomorphic sub-circuits and reuses their execution results. So, LarQucut can reduce the number of sub-circuits that need to be executed to reconstruct the large circuit's output, reducing the time spent on sampling the sub-circuits. (3) We design an adaptive quantum circuit mapping approach, which identifies qubit interaction patterns and accordingly enables the best-fit mapping policy in DQC. The experimental results show that, for large circuits with hundreds to thousands of qubits in DQC, LarQucut can provide a better cutting and mapping solution with lower overall overheads and achieves results closer to the ground truth.
- 19 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (3) , 1-24
Open AccessLarQucut: A New Cutting and Mapping Approach for Large-sized Quantum Circuits in Distributed Quantum Computing (DQC) Environments
Distributed quantum computing (DQC) is a promising way to achieve large-scale quantum computing. However, mapping large-sized quantum circuits in DQC is a challenging job; for example, it is difficult to find an ideal cutting and mapping solution when many qubits, complicated qubit operations, and diverse QPUs are involved. In this study, we propose LarQucut, a new quantum circuit cutting and mapping approach for large-sized circuits in DQC. LarQucut has several new designs. (1) LarQucut can have cutting solutions that use fewer cuts, and it does not cut a circuit into independent sub-circuits, therefore reducing the overall cutting and computing overheads. (2) LarQucut finds isomorphic sub-circuits and reuses their execution results. So, LarQucut can reduce the number of sub-circuits that need to be executed to reconstruct the large circuit's output, reducing the time spent on sampling the sub-circuits. (3) We design an adaptive quantum circuit mapping approach, which identifies qubit interaction patterns and accordingly enables the best-fit mapping policy in DQC. The experimental results show that, for large circuits with hundreds to thousands of qubits in DQC, LarQucut can provide a better cutting and mapping solution with lower overall overheads and achieves results closer to the ground truth.
- Open Access
Is Intelligence the Right Direction in New OS Scheduling for Multiple Resources in Cloud Environments?
Making it intelligent is a promising way in System/OS design. This paper proposes OSML+, a new ML-based resource scheduling mechanism for co-located cloud services. OSML+ intelligently schedules the cache and main memory bandwidth resources at the memory hierarchy and the computing core resources simultaneously. OSML+ uses a multi-model collaborative learning approach during its scheduling and thus can handle complicated cases, e.g., avoiding resource cliffs, sharing resources among applications, enabling different scheduling policies for applications with different priorities, etc. OSML+ can converge faster using ML models than previous studies. Moreover, OSML+ can automatically learn on the fly and handle dynamically changing workloads accordingly. Using transfer learning technologies, we show our design can work well across various cloud servers, including the latest off-the-shelf large-scale servers. Our experimental results show that OSML+ supports higher loads and meets QoS targets with lower overheads than previous studies.
Open AccessIs Intelligence the Right Direction in New OS Scheduling for Multiple Resources in Cloud Environments?
Making it intelligent is a promising way in System/OS design. This paper proposes OSML+, a new ML-based resource scheduling mechanism for co-located cloud services. OSML+ intelligently schedules the cache and main memory bandwidth resources at the memory hierarchy and the computing core resources simultaneously. OSML+ uses a multi-model collaborative learning approach during its scheduling and thus can handle complicated cases, e.g., avoiding resource cliffs, sharing resources among applications, enabling different scheduling policies for applications with different priorities, etc. OSML+ can converge faster using ML models than previous studies. Moreover, OSML+ can automatically learn on the fly and handle dynamically changing workloads accordingly. Using transfer learning technologies, we show our design can work well across various cloud servers, including the latest off-the-shelf large-scale servers. Our experimental results show that OSML+ supports higher loads and meets QoS targets with lower overheads than previous studies.
- Open Access
- 20 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (1) , 1-27
An Intelligent Scheduling Approach on Mobile OS for Optimizing UI Smoothness and Power
Mobile devices need to respond quickly to diverse user inputs. The existing approaches often heuristically raise the CPU/GPU frequency according to the empirical rules when facing burst inputs and various changes. Although doing so can be effective sometimes, the existing approaches still need improvements. For instance, raising processors’ frequency can lead to high power consumption when the frequency is over-provisioned or fail to meet user demands when the frequency is under-provisioned. To this end, we propose MobiRL, a reinforcement learning-based scheduler for intelligently adjusting the CPU/GPU frequency to satisfy user demands accurately on mobile systems. MobiRL monitors the mobile system status and autonomously learns to optimize UI smoothness and power consumption by conducting CPU/GPU frequency-adjusting actions. The experimental results on the latest delivered smartphones show that MobiRL outperforms the widely used commercial scheduler on real devices—reducing the frame drop rate by 4.1% and reducing power consumption by 42.8%, respectively. Moreover, compared with a study using Q-Learning for CPU frequency scheduling, MobiRL achieves up to a 2.5% lower frame drop rate and reduces power consumption by 32.6%, respectively. Our approach has been deployed in mobile phone products.
- 20 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (1) , 1-27
Open AccessAn Intelligent Scheduling Approach on Mobile OS for Optimizing UI Smoothness and Power
Mobile devices need to respond quickly to diverse user inputs. The existing approaches often heuristically raise the CPU/GPU frequency according to the empirical rules when facing burst inputs and various changes. Although doing so can be effective sometimes, the existing approaches still need improvements. For instance, raising processors’ frequency can lead to high power consumption when the frequency is over-provisioned or fail to meet user demands when the frequency is under-provisioned. To this end, we propose MobiRL, a reinforcement learning-based scheduler for intelligently adjusting the CPU/GPU frequency to satisfy user demands accurately on mobile systems. MobiRL monitors the mobile system status and autonomously learns to optimize UI smoothness and power consumption by conducting CPU/GPU frequency-adjusting actions. The experimental results on the latest delivered smartphones show that MobiRL outperforms the widely used commercial scheduler on real devices—reducing the frame drop rate by 4.1% and reducing power consumption by 42.8%, respectively. Moreover, compared with a study using Q-Learning for CPU frequency scheduling, MobiRL achieves up to a 2.5% lower frame drop rate and reduces power consumption by 32.6%, respectively. Our approach has been deployed in mobile phone products.
- Open Access
LarQucut: A New Cutting and Mapping Approach for Large-sized Quantum Circuits in Distributed Quantum Computing (DQC) Environments
Distributed quantum computing (DQC) is a promising way to achieve large-scale quantum computing. However, mapping large-sized quantum circuits in DQC is a challenging job; for example, it is difficult to find an ideal cutting and mapping solution when many qubits, complicated qubit operations, and diverse QPUs are involved. In this study, we propose LarQucut, a new quantum circuit cutting and mapping approach for large-sized circuits in DQC. LarQucut has several new designs. (1) LarQucut can have cutting solutions that use fewer cuts, and it does not cut a circuit into independent sub-circuits, therefore reducing the overall cutting and computing overheads. (2) LarQucut finds isomorphic sub-circuits and reuses their execution results. So, LarQucut can reduce the number of sub-circuits that need to be executed to reconstruct the large circuit's output, reducing the time spent on sampling the sub-circuits. (3) We design an adaptive quantum circuit mapping approach, which identifies qubit interaction patterns and accordingly enables the best-fit mapping policy in DQC. The experimental results show that, for large circuits with hundreds to thousands of qubits in DQC, LarQucut can provide a better cutting and mapping solution with lower overall overheads and achieves results closer to the ground truth.
Open AccessLarQucut: A New Cutting and Mapping Approach for Large-sized Quantum Circuits in Distributed Quantum Computing (DQC) Environments
Distributed quantum computing (DQC) is a promising way to achieve large-scale quantum computing. However, mapping large-sized quantum circuits in DQC is a challenging job; for example, it is difficult to find an ideal cutting and mapping solution when many qubits, complicated qubit operations, and diverse QPUs are involved. In this study, we propose LarQucut, a new quantum circuit cutting and mapping approach for large-sized circuits in DQC. LarQucut has several new designs. (1) LarQucut can have cutting solutions that use fewer cuts, and it does not cut a circuit into independent sub-circuits, therefore reducing the overall cutting and computing overheads. (2) LarQucut finds isomorphic sub-circuits and reuses their execution results. So, LarQucut can reduce the number of sub-circuits that need to be executed to reconstruct the large circuit's output, reducing the time spent on sampling the sub-circuits. (3) We design an adaptive quantum circuit mapping approach, which identifies qubit interaction patterns and accordingly enables the best-fit mapping policy in DQC. The experimental results show that, for large circuits with hundreds to thousands of qubits in DQC, LarQucut can provide a better cutting and mapping solution with lower overall overheads and achieves results closer to the ground truth.
- Open Access
- 14 September 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (3) , 1-24
iSwap: A New Memory Page Swap Mechanism for Reducing Ineffective I/O Operations in Cloud Environments
This article proposes iSwap , a new memory page swap mechanism that reduces the ineffective I/O swap operations and improves the QoS for applications with a high priority in cloud environments. iSwap works in the OS kernel. iSwap accurately learns the reuse patterns for memory pages and makes the swap decisions accordingly to avoid ineffective operations. In the cases where memory pressure is high, iSwap compresses pages that belong to the latency-critical (LC) applications (or high-priority applications) and keeps them in main memory, avoiding I/O operations for these LC applications to ensure QoS, and iSwap evicts low-priority applications’ pages out of main memory. iSwap has a low overhead and works well for cloud applications with large memory footprints. We evaluate iSwap on Intel x86 and ARM platforms. The experimental results show that iSwap can significantly reduce ineffective swap operations (8.0%–19.2%) and improve the QoS for LC applications (36.8%–91.3%) in cases where memory pressure is high, compared with the latest LRU-based approach widely used in modern OSes.
- 14 September 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (3) , 1-24
Open AccessiSwap: A New Memory Page Swap Mechanism for Reducing Ineffective I/O Operations in Cloud Environments
This article proposes iSwap , a new memory page swap mechanism that reduces the ineffective I/O swap operations and improves the QoS for applications with a high priority in cloud environments. iSwap works in the OS kernel. iSwap accurately learns the reuse patterns for memory pages and makes the swap decisions accordingly to avoid ineffective operations. In the cases where memory pressure is high, iSwap compresses pages that belong to the latency-critical (LC) applications (or high-priority applications) and keeps them in main memory, avoiding I/O operations for these LC applications to ensure QoS, and iSwap evicts low-priority applications’ pages out of main memory. iSwap has a low overhead and works well for cloud applications with large memory footprints. We evaluate iSwap on Intel x86 and ARM platforms. The experimental results show that iSwap can significantly reduce ineffective swap operations (8.0%–19.2%) and improve the QoS for LC applications (36.8%–91.3%) in cases where memory pressure is high, compared with the latest LRU-based approach widely used in modern OSes.
- Open Access
- 18 January 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (1) , 1-27
QuCloud+: A Holistic Qubit Mapping Scheme for Single/Multi-programming on 2D/3D NISQ Quantum Computers
Qubit mapping for NISQ superconducting quantum computers is essential to fidelity and resource utilization. The existing qubit mapping schemes meet challenges, e.g., crosstalk, SWAP overheads, diverse device topologies, etc., leading to qubit resource underutilization and low fidelity in computing results. This article introduces QuCloud+, a new qubit mapping scheme that tackles these challenges. QuCloud+ has several new designs. (1) QuCloud+ supports single/multi-programming quantum computing on quantum chips with 2D/3D topology. (2) QuCloud+ partitions physical qubits for concurrent quantum programs with the crosstalk-aware community detection technique and further allocates qubits according to qubit degree, improving fidelity, and resource utilization. (3) QuCloud+ includes an X-SWAP mechanism that avoids SWAPs with high crosstalk errors and enables inter-program SWAPs to reduce the SWAP overheads. (4) QuCloud+ schedules concurrent quantum programs to be mapped and executed based on estimated fidelity for the best practice. Experimental results show that, compared with the existing typical multi-programming study [ 12 ], QuCloud+ achieves up to 9.03% higher fidelity and saves on the required SWAPs during mapping, reducing the number of CNOT gates inserted by 40.92%. Compared with a recent study [ 30 ] that enables post-mapping gate optimizations to further reduce gates, QuCloud+ reduces the post-mapping circuit depth by 21.91% while using a similar number of gates.
- 18 January 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (1) , 1-27
Open AccessQuCloud+: A Holistic Qubit Mapping Scheme for Single/Multi-programming on 2D/3D NISQ Quantum Computers
Qubit mapping for NISQ superconducting quantum computers is essential to fidelity and resource utilization. The existing qubit mapping schemes meet challenges, e.g., crosstalk, SWAP overheads, diverse device topologies, etc., leading to qubit resource underutilization and low fidelity in computing results. This article introduces QuCloud+, a new qubit mapping scheme that tackles these challenges. QuCloud+ has several new designs. (1) QuCloud+ supports single/multi-programming quantum computing on quantum chips with 2D/3D topology. (2) QuCloud+ partitions physical qubits for concurrent quantum programs with the crosstalk-aware community detection technique and further allocates qubits according to qubit degree, improving fidelity, and resource utilization. (3) QuCloud+ includes an X-SWAP mechanism that avoids SWAPs with high crosstalk errors and enables inter-program SWAPs to reduce the SWAP overheads. (4) QuCloud+ schedules concurrent quantum programs to be mapped and executed based on estimated fidelity for the best practice. Experimental results show that, compared with the existing typical multi-programming study [ 12 ], QuCloud+ achieves up to 9.03% higher fidelity and saves on the required SWAPs during mapping, reducing the number of CNOT gates inserted by 40.92%. Compared with a recent study [ 30 ] that enables post-mapping gate optimizations to further reduce gates, QuCloud+ reduces the post-mapping circuit depth by 21.91% while using a similar number of gates.
- 29 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43 (5) , 1429-1441
ATA-Cache: Contention Mitigation for GPU Shared L1 Cache With Aggregated Tag Array
To fully exploit the locality of GPU applications, the GPU shared L1 cache architecture, which shares L1 cache among multiple GPU cores, is a promising architecture while still suffering from high-resource contentions. We present a GPU shared L1 cache architecture with an aggregated tag array that minimizes the L1 cache contentions and takes full advantage of inter-core locality. The key idea is to decouple and aggregate the tag arrays of multiple L1 caches so that the cache requests can be compared with all tag arrays in parallel to probe the replicated data in other caches. The GPU caches are only accessed by other GPU cores when replicated data exists, filtering out unnecessary cache accesses that cause high-resource contentions. We also develop a two-level thread-block scheduling policy adapted for the shared L1 cache architecture to maximize the available locality. The experimental results show that GPU performance can be improved by 14.5% on average for applications with a high inter-core locality.
- 29 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43 (5) , 1429-1441
ATA-Cache: Contention Mitigation for GPU Shared L1 Cache With Aggregated Tag Array
To fully exploit the locality of GPU applications, the GPU shared L1 cache architecture, which shares L1 cache among multiple GPU cores, is a promising architecture while still suffering from high-resource contentions. We present a GPU shared L1 cache architecture with an aggregated tag array that minimizes the L1 cache contentions and takes full advantage of inter-core locality. The key idea is to decouple and aggregate the tag arrays of multiple L1 caches so that the cache requests can be compared with all tag arrays in parallel to probe the replicated data in other caches. The GPU caches are only accessed by other GPU cores when replicated data exists, filtering out unnecessary cache accesses that cause high-resource contentions. We also develop a two-level thread-block scheduling policy adapted for the shared L1 cache architecture to maximize the available locality. The experimental results show that GPU performance can be improved by 14.5% on average for applications with a high inter-core locality.