Related papers: Scalable and Efficient Virtual Memory Sharing in H…
Distributed inference of large language models (LLMs) using tensor parallelism can introduce communication overheads of $20$% even over GPUs connected via NVLink, a high-speed GPU interconnect. Several techniques have been proposed to…
Serving numerous users and requests concurrently requires good fairness in Large Language Models (LLMs) serving system. This ensures that, at the same cost, the system can meet the Service Level Objectives (SLOs) of more users , such as…
Common implementations of core memory allocation components, like the Linux buddy system, handle concurrent allocation/release requests by synchronizing threads via spin-locks. This approach is clearly not prone to scale with large thread…
Serverless computing is renowned for its computation elasticity, yet its full potential is often constrained by the requirement for functions to operate within local and dedicated background environments, resulting in limited memory…
As high-performance computing (HPC) moves into the exascale era, computer scientists and engineers must find innovative ways of transferring and processing unprecedented amounts of data. As the scale and complexity of the applications…
Offloading compute-intensive kernels to hardware accelerators relies on the large degree of parallelism offered by these platforms. However, the effective bandwidth of the memory interface often causes a bottleneck, hindering the…
Hybrid language models like Jamba mix attention layers with State Space Models (SSMs), creating two memory cache types with opposite profiles: Key-Value (KV) caches grow linearly with sequence length, while SSM states stay fixed per layer.…
Non-volatile memory (NVM) technologies such as spin-transfer torque magnetic random access memory (STT-MRAM) and spin-orbit torque magnetic random access memory (SOT-MRAM) have significant advantages compared to conventional SRAM due to…
The sizes of GPU applications are rapidly growing. They are exhausting the compute and memory resources of a single GPU, and are demanding the move to multiple GPUs. However, the performance of these applications scales sub-linearly with…
Top-K SpMV is a key component of similarity-search on sparse embeddings. This sparse workload does not perform well on general-purpose NUMA systems that employ traditional caching strategies. Instead, modern FPGA accelerator cards have a…
This paper addresses the problem of universal synchronization primitives that can support scalable thread synchronization for large-scale many-core architectures. The universal synchronization primitives that have been deployed widely in…
Large Language Models (LLMs) exhibit pronounced memory-bound characteristics during inference due to High Bandwidth Memory (HBM) bandwidth constraints. In this paper, we propose an L2 Cache-oriented asynchronous KV Cache prefetching method…
Bandwidth-starved multicore chips have become ubiquitous. It is well known that the performance of stencil codes can be improved by temporal blocking, lessening the pressure on the memory interface. We introduce a new pipelined approach…
In this paper, we present a SRAM-PCM hybrid cache design, along with a cache replacement policy, named dead fast block (DFB) to manage the hybrid cache. This design aims to leverage the best features of both SRAM and PCM devices. Compared…
In current microarchitectures, due to the complex memory hierarchies and different latencies on memory accesses, thread and data mapping are important issues to improve application performance. Software transactional memory (STM) is an…
Large persistent memories such as NVDIMM have been perceived as a disruptive memory technology, because they can maintain the state of a system even after a power failure and allow the system to recover quickly. However, overheads incurred…
Software Transactional Memory Systems (STM) are a promising alternative to lock based systems for concurrency control in shared memory systems. In multiversion STM systems, each write on a transaction object produces a new version of that…
Solid-state storage architectures based on NAND or emerging memory devices (SSD), are fundamentally architected and optimized for both reliability and performance. Achieving these simultaneous goals requires co-design of memory components…
Non-volatile memory (NVM) is a promising technology for low-energy and high-capacity main memory of computers. The characteristics of NVM devices, however, tend to be fundamentally different from those of DRAM (i.e., the memory device…
Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to Autoregressive Models (ARMs), utilizing parallel decoding to overcome sequential bottlenecks. However, existing research focuses primarily on kernel-level…