Yunho Oh

Yunho Oh

Auto-generated profile, based on publicly available information

Yunho Oh

School of Electrical Engineering, Korea University, Seoul, South Korea

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 2023 in ACM Transactions on Architecture and Code Optimization
Verified
Journal Article
Open access
AYAhmet Caner Yüzügüler
Ahmet Caner Yüzügüler
CSCanberk Sönmez
Canberk Sönmez
MDMario Drumond
Mario Drumond
YOYunho Oh
Yunho Oh
BFBabak Falsafi
Babak Falsafi
PFPascal Frossard
Pascal Frossard
Multi-pod systolic arrays are emerging as the architecture of choice in DNN inference accelerators. Despite their potential, designing multi-pod systolic arrays to maximize effective throughput/Watt—i.e., throughput/Watt adjusted when accounting for array utilization—poses a unique set of challenges. In this work, we study three key pillars in multi-pod systolic array designs, namely array granularity, interconnect, and tiling. We identify optimal array granularity across workloads and show that state-of-the-art commercial accelerators use suboptimal array sizes for single-tenancy workloads. We, then evaluate the bandwidth/latency trade-offs in interconnects and show that Butterfly networks offer a scalable topology for accelerators with a large number of pods. Finally, we introduce a novel data tiling scheme with custom partition size to maximize utilization in optimally sized pods. We propose Scale-out Systolic Arrays , a multi-pod inference accelerator for both single- and multi-tenancy based on these three pillars. We show that SOSA exhibits scaling of up to 600 TeraOps/s in effective throughput for state-of-the-art DNN inference workloads, and outperforms state-of-the-art multi-pod accelerators by a factor of 1.5 ×. 1

17 June 2023
Verified
Conference Paper
Open access
DHDongho Ha
Dongho Ha
YOYunho Oh
Yunho Oh
WRWon Ro
Won Ro
A generally used GPU programming methodology is that adjacent threads access data in neighbor or specific-stride memory addresses and perform computations with the fetched data. This paper demonstrates that the memory addresses often exhibit a simple linear value pattern across GPU threads, as each thread uses built-in variables and constant values to compute the memory addresses. However, since the threads compute their context data individually, GPUs incur a heavy instruction overhead to calculate the memory addresses, even though they exhibit a simple pattern. We propose a GPU architecture called Removing ReDunDancy Utilizing Linearity of Address Generation (R2D2), reducing a large amount of the dynamic instruction count by detecting such linear patterns in the memory addresses and exploiting them for kernel computations. R2D2 detects linearities of the memory addresses with software support and pre-computes them before the threads execute the instructions. With the proposed scheme, each thread is able to compute its memory addresses with fewer dynamic instructions than conventional GPUs. In our evaluation, R2D2 achieves dynamic instruction reduction by 28%, 1.25x speedup, and energy consumption reduction by 17% over baseline GPU.

25 March 2024 in IEEE Access
Verified
Journal Article
Open access
SCSangun Choi
Sangun Choi
SPSeongjun Park
Seongjun Park
JPJaeyong Park
Jaeyong Park
JKJongmin Kim
Jongmin Kim
GKGunjae Koo
Gunjae Koo
YOYunho Oh
Yunho Oh
Conventional DNN inference accelerators are designed with a few (up to four) large systolic arrays. As such a scale-up architecture often suffers from low utilization, a scale-out architecture, in which a single accelerator has tens of pods and each pod has a small systolic array, has been proposed. While the scale-out architecture is promising, it still incurs increasing off-chip memory access as the pods are supposed to access the duplicate tiles of tensors. Prior work has proposed a shared buffer structure to address the problem, but those architectures suffer from performance degradation due to shared buffer access latency. We make an observation that all the pods access the same rows of input and weights within a short time window. With the observation, we propose a new inference accelerator architecture, called Vectored Systolic Arrays (SAVector). SAVector consists of a new two-level on-chip buffer architecture and a tensor tile scheduling technique. In the new buffer architecture, global buffers are shared by all the pods and they keep the rows shared by the pods. And each pod has a tiny dedicated buffer. SAVector monitors the memory access behavior and timely determines to prefetch the data and flush it. In our evaluation, SAVector exhibits a very similar off-chip memory access count to the scale-up architecture and achieves 52% energy-delay-product (EDP) reduction. Also, SAVector achieves 27% EDP reduction over prior work by mitigating performance degradation from global buffer access latency.

31 March 2022 in IEEE Embedded Systems Letters
Verified
Journal Article
Closed access
YOYunho Oh
Yunho Oh
IJIpoom Jeong
Ipoom Jeong
WRWon Ro
Won Ro
MYMyung Kuk Yoon
Myung Kuk Yoon
Spin-Transfer Torque Magnetic Random-Access Memory (STT-MRAM) is an emerging non-volatile memory technology that has been received significant attention due to its higher density and lower leakage current over SRAM. One compelling use case is to employ STT-MRAM as a GPU Register File (RF) to reduce its massive energy consumption. One critical challenge is that STT-MRAM has longer access latency and higher dynamic power consumption than SRAM, which motivates the hierarchical RF that places a small SRAM-based Register Cache (RC) between functional units and STT-MRAM RF. The RC acts as the write buffer, so all the writes on the RF are first performed on the RC. In the presence of a conflict miss, the RC writes back the corresponding cache line into the RF. In this work, we observe that a large amount of such write-back operations are unnecessary because they include register values that are never used again. Leveraging this observation, we propose a Compiler-ASsisted Hierarchical Register File in GPUs (CASH-RF) that optimizes STT-MRAM accesses by removing dead register values. In CASH-RF, unnecessary write-back operations are detected by the compiler via the last consumer analysis. At runtime, the corresponding RC lines are discarded after the last references without being updated to the RF. Compared to the baseline GPUs, CASH-RF removes 59.5% of write-back operations, which leads to 54.7% lower RF energy consumption with only 2.6% of performance degradation.

20 June 2025
Verified
Conference Paper
Open access
MGMinseong Gil
Minseong Gil
DHDongho Ha
Dongho Ha
SHSimla Burcu Harma
Simla Burcu Harma
MYMyung Kuk Yoon
Myung Kuk Yoon
BFBabak Falsafi
Babak Falsafi
YOYunho Oh
Yunho Oh

3 August 2022 in IEEE Access
Verified
Journal Article
Open access
IJIpoom Jeong
Ipoom Jeong
YOYunho Oh
Yunho Oh
WRWon Ro
Won Ro
MYMyung Kuk Yoon
Myung Kuk Yoon
Graphics processing units (GPUs) achieve high throughput by exploiting a high degree of thread-level parallelism (TLP). To support such high TLP, GPUs have a large-sized register file to store the context of all threads, consuming around 20% of total GPU energy. Several previous studies have attempted to minimize the energy consumption of the register file by implementing an emerging non-volatile memory (NVM), leveraging its higher density and lower leakage power over SRAMs. To amortize the cost of long access latency of NVM, prior work adopts a hierarchical register file consisting of an SRAM-based register cache and NVM-based registers where the register cache works as a write buffer. To get the register cache index, they use the partially selected bits of warp ID and register ID. This work observes that such an index calculation causes three types of contentions leading to the underutilization of the register cache: inter-warp , intra-warp , and false contentions. To minimize such contentions, this paper proposes a thread context-aware register cache (TEA-RC) in GPUs. In TEA-RC, the cache index is calculated considering the high correlation between the number of scheduled threads and the register usage of threads. The proposed design shows 28.5% higher performance and 9.1 percentage point lower energy consumption over the conventional register cache that concatenates three bits of warp ID and five bits of register ID to compute the cache index.