Logo: to the web site of Uppsala University

uu.sePublikasjoner fra Uppsala universitet
Endre søk
Link to record
Permanent link

Direct link
Aligholipour, Rashid, PhD StudentORCID iD iconorcid.org/0000-0001-9376-2925
Publikasjoner (5 av 5) Visa alla publikasjoner
Aligholipour, R., Kaxiras, S. & Yao, Y. (2026). DICE: Detailed Inter-Chiplet End-to-End PHY Modeling for Accurate Chiplet Simulation. In: 2026 ACM/IEEE 53RD ANNUAL INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE, ISCA: . Paper presented at 53rd International Symposium on Computer Architecture-ISCA-Annual, JUN 27-JUL 01, 2026, Raleigh, NC (pp. 1611-1627). IEEE
Åpne denne publikasjonen i ny fane eller vindu >>DICE: Detailed Inter-Chiplet End-to-End PHY Modeling for Accurate Chiplet Simulation
2026 (engelsk)Inngår i: 2026 ACM/IEEE 53RD ANNUAL INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE, ISCA, IEEE, 2026, s. 1611-1627Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Scaling monolithic multicores is increasingly constrained by power/thermal limits, yield, and rising manufacturing and testing costs. Chiplet designs address these challenges by partitioning large dies into smaller parts (typically multiple core-complex dies and an I/O die) linked via high-bandwidth physical fabrics (PHY). As bandwidth and wiring density scale, however, these short-reach links are pushed closer to their signal-integrity limits, increasing susceptibility to noise, crosstalk, and channel loss, motivating stronger link-level reliability mechanisms such as forward error correction (FEC). Despite this trend, state-of-the-art simulation infrastructures often approximate inter-chiplet links using oversimplified, fixed-latency models. Such abstractions overlook the inherently dynamic, runtime-dependent behavior of the PHY-including channel conditions (e.g., signal-to-noise ratio shifts, signal crosstalk, clock jitter), iterative decoder convergence and packet retransmissions, and application dynamics (e.g., LLC-misses that travel across chiplet boundaries)-all of which are hard to determine offline. We show that neglecting these effects distorts inter-chiplet packet-level timing and high-level performance metrics such as IPC, leading to off-trend simulation results. We present DICE, an in-simulation, runtime PHY modeling in gem5 that captures the end-to-end inter-chiplet datapath, including QC-LDPC encoding/decoding, PAM4 modulation, lossy-channel transmission, LLR-based demodulation, adaptive packet re-sending, and PHY-level flow control between chiplets. For fidelity, we calibrate component latencies via hardware synthesis (e.g., QC-LDPC decode paths and iteration budgets) and compare the integrated system against production chiplet processors. Compared with state-of-the-art fixed-latency inter-chiplet links such as HeteroGarnet, DICE reshapes packet-latency composition and shifts system IPC by an average of 6.8% and up to 27.6%, revealing variability driven by realistic PHY link behaviors, enabling more accurate co-evaluation of performance, reliability, and design trade-offs in modern chiplet systems.

sted, utgiver, år, opplag, sider
IEEE, 2026
HSV kategori
Identifikatorer
urn:nbn:se:uu:diva-602961 (URN)10.1109/ISCA66397.2026.00118 (DOI)001886295500102 ()2-s2.0-105047154818 (Scopus ID)979-8-3315-5066-0 (ISBN)979-8-3315-5065-3 (ISBN)
Konferanse
53rd International Symposium on Computer Architecture-ISCA-Annual, JUN 27-JUL 01, 2026, Raleigh, NC
Forskningsfinansiär
Swedish Research Council, 2025-04436Swedish Research Council, 2022-04959Swedish Foundation for Strategic Research, FUS21-0067Swedish Research Council
Tilgjengelig fra: 2026-11-09 Laget: 2026-11-09 Sist oppdatert: 2026-11-09bibliografisk kontrollert
Yao, Y., Aligholipour, R., Li, S. & Kaxiras, S. (2026). NoCWalk: In-Network PageWalks for Efficient Pointer-Chasing Workloads on Multicores. In: PROCEEDINGS OF THE 23RD ACM INTERNATIONAL CONFERENCE ON COMPUTING FRONTIERS 2026, CF 2026: . Paper presented at 23rd International Conference on Computing Frontiers-CF, MAY 19-21, 2026, Catania, ITALY (pp. 122-132). Association for Computing Machinery (ACM)
Åpne denne publikasjonen i ny fane eller vindu >>NoCWalk: In-Network PageWalks for Efficient Pointer-Chasing Workloads on Multicores
2026 (engelsk)Inngår i: PROCEEDINGS OF THE 23RD ACM INTERNATIONAL CONFERENCE ON COMPUTING FRONTIERS 2026, CF 2026, Association for Computing Machinery (ACM), 2026, s. 122-132Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Large-scale concurrent data structure (CDS) applications frequently exhibit highly irregular memory access patterns, which result in frequent page walks. Traditionally, these page walks are serviced by a centralized page walker within each core's memory management unit (MMU). However, due to the irregular and dispersed memory footprints of CDS workloads, we find that centralized page walks over distributed, radix-based page tables often incur substantial latency, significantly limiting the overall performance of CDS workloads. To overcome this limit, we propose NoCWalk, a light-weight mechanism that leverages cache coherence controllers (CCs) to perform distributed page walks inside the network-on-chip (NoC). With NoCWalk, CCs can directly handle page walk requests within the NoC without relying on centralized MMUs. Specifically, each page walk message carries the page table offsets extracted from the virtual address under translation. As the message traverses the NoC, the CC holding the current-level page table entry (PTE) checks the access right to the next level, computes the address for the new PTE, and forwards the request directly to the new PTE location-all without returning to the requesting core's MMU. This redirection continues iteratively until the final-level PTE is reached and the translation is completed. If a PTE is missing, or a memory access violation is detected at any point, a page fault message is generated on the spot and routed back to the originating core, which will be handled by the MMU as in the normal case. We evaluate NoCWalk using gem5 full-system simulations on workloads representing intensive CDS applications. Our results show that NoCWalk reduces page walk latency by 43.8% on average (up to 62.5%), and improves overall application performance by 9.1% on average (up to 21.7%).

sted, utgiver, år, opplag, sider
Association for Computing Machinery (ACM), 2026
Emneord
Page walk, network-on-chip, coherence protocol
HSV kategori
Identifikatorer
urn:nbn:se:uu:diva-602965 (URN)10.1145/3801487.3801813 (DOI)001900982600017 ()2-s2.0-105044225194 (Scopus ID)979-8-4007-2568-5 (ISBN)
Konferanse
23rd International Conference on Computing Frontiers-CF, MAY 19-21, 2026, Catania, ITALY
Forskningsfinansiär
Swedish Research Council, 2025-04436Swedish Research Council, 2022-04959Swedish Foundation for Strategic Research, FUS21006
Tilgjengelig fra: 2026-11-10 Laget: 2026-11-10 Sist oppdatert: 2026-11-10bibliografisk kontrollert
Aligholipour, R., Aimoniotis, P., Kaxiras, S. & Yao, Y. (2025). RXT: RefleXive Address Translation for Pointer-Chasing Workloads. In: 2025 IEEE International Parallel and Distributed Processing Symposium (IPDPS): . Paper presented at International Symposium on Parallel and Distributed Processing (IPDPS), 3-7 June, 2025, Milano, Italy (pp. 1062-1073). Institute of Electrical and Electronics Engineers (IEEE)
Åpne denne publikasjonen i ny fane eller vindu >>RXT: RefleXive Address Translation for Pointer-Chasing Workloads
2025 (engelsk)Inngår i: 2025 IEEE International Parallel and Distributed Processing Symposium (IPDPS), Institute of Electrical and Electronics Engineers (IEEE), 2025, s. 1062-1073Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

With increasingly irregular memory access patterns in Indirect Memory Access (IMA) and Graph Processing (GP) applications, Virtual Address Translation (VAT) not only has become a major performance bottleneck, but also a key contributor to high power consumption in out-of-order (OoO) multi-core processor pipeline. We find that this problem can be attributed to a large extent to an inefficient utilization of TLBs in pointer-chasing instructions (i.e. load-to-load), which constitute a significant portion of workloads in IMA/GP. Conventionally, VATs for pointer-chasing are performed via a series of TLB look-ups: one for each load in the pointer chain. However, this approach is sub-optimal as it overlooks the correlations between pointers, missing opportunities to perform VATs through more costeffective calculations instead of relying on the more expensive TLB look-ups. This work draws on the insight that in pointer-chasing workloads, the physical page holding an upstream pointer is often at a fixed distance to the physical page where a downstream pointer resides. Building on this insight, we introduce RefleXive address Translation (RXT), which encodes the physical page distance (termed PageDist) between an upstream and downstream pointer into the unused upper 16 bits of the upstream pointer's virtual address. Consequently, RXT can compute the translation of a pointer by directly adding the PageDist to its residing address (the physical address where the pointer is stored). This process can be recursively applied throughout the pointer-chasing sequence, transforming address translation from a sequence of TLB accesses to computes. Through gem5 full-system simulation, RXT reduces TLB energy consumption for pointer-chasing applications by an average of 41.48 % (up to 62.50 %) and decreases overall core power density by 4.65 % (up to 13.09 %), without compromising application execution time. In fact, RXT even improves program runtime by an average of 0.71 % (up to 2.09 %).

sted, utgiver, år, opplag, sider
Institute of Electrical and Electronics Engineers (IEEE), 2025
Serie
IEEE International Parallel and Distributed Processing Symposium, ISSN 1530-2075, E-ISSN 1530-2075
Emneord
Virtual address translation, indirect memory access, graph applications
HSV kategori
Identifikatorer
urn:nbn:se:uu:diva-575490 (URN)10.1109/IPDPS64566.2025.00098 (DOI)001552207700090 ()979-8-3315-3237-6 (ISBN)979-8-3315-3238-3 (ISBN)
Konferanse
International Symposium on Parallel and Distributed Processing (IPDPS), 3-7 June, 2025, Milano, Italy
Tilgjengelig fra: 2026-01-12 Laget: 2026-01-12 Sist oppdatert: 2026-02-10bibliografisk kontrollert
Aligholipour, R. & Yao, Y. (2025). The Fake-Busy and True-Idle Problems of Running Graph Applications on Chiplet-Based Multi-cores. In: 2025 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS): . Paper presented at IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), Ghent, Belgium, 11-13 May, 2025 (pp. 347-349). Institute of Electrical and Electronics Engineers (IEEE)
Åpne denne publikasjonen i ny fane eller vindu >>The Fake-Busy and True-Idle Problems of Running Graph Applications on Chiplet-Based Multi-cores
2025 (engelsk)Inngår i: 2025 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), Institute of Electrical and Electronics Engineers (IEEE), 2025, s. 347-349Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

We introduce the fake-busy and true-idle problems encountered when running large graph workloads on chipletbased Out-of-Order (OoO) multi-cores. Caused by high interchiplet communication latency and irregular memory access patterns, these issues lead to inefficient use of core pipeline resources such as the Reorder Buffer (RoB) and Load Queue (LQ). Our evaluation shows that reducing RoB and LQ sizes has minimal impact on application performance, revealing new performance optimization opportunities for large graph workloads.

sted, utgiver, år, opplag, sider
Institute of Electrical and Electronics Engineers (IEEE), 2025
Serie
IEEE International Symposium on Performance Analysis of Software and Systems, ISSN 2994-9513, E-ISSN 2766-0486
HSV kategori
Identifikatorer
urn:nbn:se:uu:diva-575493 (URN)10.1109/ISPASS64960.2025.00039 (DOI)001555755100029 ()979-8-3315-0294-2 (ISBN)979-8-3315-0295-9 (ISBN)
Konferanse
IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), Ghent, Belgium, 11-13 May, 2025
Tilgjengelig fra: 2026-01-12 Laget: 2026-01-12 Sist oppdatert: 2026-02-10bibliografisk kontrollert
Aligholipour, R. & Yao, Y. (2025). The Fake-Busy and True-Idle Problems of Running Graph Applications on Chiplet-Based Multi-Cores. In: 2025 IEEE International Symposium on Workload Characterization (IISWC): . Paper presented at IEEE International Symposium on Workload Characterization (IISWC), 12-14 October, 2025, Irvine, CA, USA (pp. 42-55). Institute of Electrical and Electronics Engineers (IEEE)
Åpne denne publikasjonen i ny fane eller vindu >>The Fake-Busy and True-Idle Problems of Running Graph Applications on Chiplet-Based Multi-Cores
2025 (engelsk)Inngår i: 2025 IEEE International Symposium on Workload Characterization (IISWC), Institute of Electrical and Electronics Engineers (IEEE), 2025, s. 42-55Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Graph applications, characterized by irregular memory footprints and low branch prediction accuracy, pose significant challenges for modern computing systems. This work examines the performance of state-of-the-art chiplet-based multi-core architectures featuring out-of-order (OoO) cores with deep pipeline components, such as reorder buffers (ROB) and load queues (LQ), when executing graph workloads.Through detailed micro-architectural profiling, we find that OoO cores are often inefficiently utilized. We identify two dominant causes: fake-busy and true-idle. In fake-busy, a core waiting for a branch-resolving load speculatively skips the current branch, enters the next loop, and repeats this process— iteratively issuing a cascade of speculative instructions. In true-idle, a core waiting for a branch-resolving load speculatively executes code of the current branch, where instructions often depend on the branch-resolving load itself, resulting in unproductive pipeline stalls and instructions accumulating in the ROB. Due to high load latency in chiplet architectures and low branch prediction accuracy typical of graph workloads, both scenarios suffer from deep speculation and frequent pipeline squashing, leading to inefficient utilization of OoO pipeline resources.Our results show that halving the ROB and LQ sizes leads to only minor performance degradation—average slowdowns of 4.3% and 1.0%, respectively—across most applications, with the exception of two cases that exhibit more substantial slowdowns of 37.3% (half-ROB) and 5.6% (half-LQ). Moreover, operating at a lower core frequency (e.g., 2.0 GHz) yields an energy-delay product (EDP) comparable to that of higher frequencies (e.g., 2.5 GHz and 3.0 GHz) across all applications. Together, these findings highlight new opportunities for performance and power co-optimization when running graph workloads on chiplet-based multi-core systems.

sted, utgiver, år, opplag, sider
Institute of Electrical and Electronics Engineers (IEEE), 2025
HSV kategori
Identifikatorer
urn:nbn:se:uu:diva-575492 (URN)10.1109/IISWC66894.2025.00014 (DOI)001828212200004 ()2-s2.0-105029039453 (Scopus ID)979-8-3315-4917-6 (ISBN)
Konferanse
IEEE International Symposium on Workload Characterization (IISWC), 12-14 October, 2025, Irvine, CA, USA
Tilgjengelig fra: 2026-01-12 Laget: 2026-01-12 Sist oppdatert: 2026-09-18bibliografisk kontrollert
Organisasjoner
Identifikatorer
ORCID-id: ORCID iD iconorcid.org/0000-0001-9376-2925