DNA mapping using Processor-in-Memory architecture

Abstract

This paper presents the implementation of a mapping algorithm on a new Processing-in-Memory (PIM) architecture developed by UPMEM Company. UPMEM's solution consists in adding processing units into the DRAM, to minimize data access time and maximize bandwidth, in order to drastically accelerate data-consuming algorithms. The technology developed by UPMEM makes it possible to combine 256 cores with 16 GBytes of DRAM, on a standard DIMM module. An experimentation of DNA Mapping on Human genome dataset shows that a speed-up of 25 can be obtained with UPMEM technology compared to fast mapping software such as BWA, Bowtie2 or NextGenMap running on 16 Intel threads. Experimentation also highlight that data transfer from storage device limits the performances of the implementation. The use of SSD drives can boost the speed-up to 80.

Massively parallel skyline computation for processing-in-memory architectures

“…Once the required data is in-place, the host may initiate any number of transformations to be performed on the data using the embedded co-processors. This data-centric model favors the execution of fine grained data-parallel tasks [25]. Figure 3 illustrates the UPMEM's PIM architecture.…”

“…Processing-In-Memory (PIM) architectures [2,15,19,20,25,28,34,37,38,43] present a viable alternative for addressing this bottleneck leveraging on many processing cores that are embedded into DRAM. Moving processing closer to where data reside offers many advantages including but not limited to higher processing throughput, lower power consumption and increased scalability for well designed parallel algorithms (Table 1).…”

“…Moving processing closer to where data reside offers many advantages including but not limited to higher processing throughput, lower power consumption and increased scalability for well designed parallel algorithms (Table 1). In this paper we rely on UPMEM's architecture [25] , a commercially available PIM implementation that incorporates several of the aforementioned characteristics. Our skyline implementation presents a practical use case, that captures the important challenges associated with designing complex data processing algorithms using the PIM programming model.…”

“…Our skyline implementation presents a practical use case, that captures the important challenges associated with designing complex data processing algorithms using the PIM programming model. UPMEM's architectural implementation follows closely the fundamental characteristics of previous PIM systems [2,15,19,20,25,28,34,37,38,43], offering in addition an FPGA-based testing environment [1].…”

Software platform dedicated for in-memory computing circuit evaluation

“…They can be grouped in the following techniques: Processing-In-Memory, In-Memory-Computing, Logic-In-Memory. Processing-In-Memory (PIM) is a process based on DRAM, which consists in putting the computation unit next to the memory while keeping the two dissociated [14]. The objective behind bringing the processor near to the memory aims to reduce the time required to access the data inside the memory and therefore the power consumption due to that.…”