{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JY6E7UABYEP6V3FVABAHJMDE3U","short_pith_number":"pith:JY6E7UAB","schema_version":"1.0","canonical_sha256":"4e3c4fd001c11feaecb5004074b064dd2db06d0e3e273febc58b055c3c4f9bb9","source":{"kind":"arxiv","id":"2607.28824","version":1},"attestation_state":"computed","paper":{"title":"Characterizing LLM Kernel Access and Memory Interaction in Multi-Partition NUMA GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","cs.PF"],"primary_cat":"cs.AR","authors_text":"Bahar Asgari, Donghyeon Joo, Nuwan Jayasena, Sooraj Puthoor","submitted_at":"2026-07-30T20:26:58Z","abstract_excerpt":"Large language model (LLM) workloads motivate multi-partition GPUs as a path to scaling compute and memory capacity, but their non-uniform memory access characteristics and inter-partition communication can amplify contention and degrade locality, leading to suboptimal kernel latency. To address this, we analyze performance-critical LLM kernel implementations spanning weight projection, mixture-of-experts, and attention variants of state-of-the-art serving engines to present a characterization of data access patterns in multi-partition GPUs. First, we introduce memory trace analysis methodolog"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.28824","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AR","submitted_at":"2026-07-30T20:26:58Z","cross_cats_sorted":["cs.DC","cs.PF"],"title_canon_sha256":"4dfbae532a772e63f7f2c212c12602ed9437c28240169a1ec6c8aaa66fab43ea","abstract_canon_sha256":"911e275dc8baf3bfb7bd3800e8b4cbb3ae1b85933da813bfc69de551a7da9b1a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-03T01:16:53.450602Z","signature_b64":"Ug+RQe5wwxFFBI95VESxxPwQJcP2htx8dh/99STz7NxQ+tsBTOHRXVjdeHpAdpoFuHQLkqFGJIxwK2h8bPJWBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e3c4fd001c11feaecb5004074b064dd2db06d0e3e273febc58b055c3c4f9bb9","last_reissued_at":"2026-08-03T01:16:53.448794Z","signature_status":"signed_v1","first_computed_at":"2026-08-03T01:16:53.448794Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Characterizing LLM Kernel Access and Memory Interaction in Multi-Partition NUMA GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","cs.PF"],"primary_cat":"cs.AR","authors_text":"Bahar Asgari, Donghyeon Joo, Nuwan Jayasena, Sooraj Puthoor","submitted_at":"2026-07-30T20:26:58Z","abstract_excerpt":"Large language model (LLM) workloads motivate multi-partition GPUs as a path to scaling compute and memory capacity, but their non-uniform memory access characteristics and inter-partition communication can amplify contention and degrade locality, leading to suboptimal kernel latency. To address this, we analyze performance-critical LLM kernel implementations spanning weight projection, mixture-of-experts, and attention variants of state-of-the-art serving engines to present a characterization of data access patterns in multi-partition GPUs. First, we introduce memory trace analysis methodolog"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.28824","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.28824/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.28824","created_at":"2026-08-03T01:16:53.449799+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.28824v1","created_at":"2026-08-03T01:16:53.449799+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.28824","created_at":"2026-08-03T01:16:53.449799+00:00"},{"alias_kind":"pith_short_12","alias_value":"JY6E7UABYEP6","created_at":"2026-08-03T01:16:53.449799+00:00"},{"alias_kind":"pith_short_16","alias_value":"JY6E7UABYEP6V3FV","created_at":"2026-08-03T01:16:53.449799+00:00"},{"alias_kind":"pith_short_8","alias_value":"JY6E7UAB","created_at":"2026-08-03T01:16:53.449799+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U","json":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U.json","graph_json":"https://pith.science/api/pith-number/JY6E7UABYEP6V3FVABAHJMDE3U/graph.json","events_json":"https://pith.science/api/pith-number/JY6E7UABYEP6V3FVABAHJMDE3U/events.json","paper":"https://pith.science/paper/JY6E7UAB"},"agent_actions":{"view_html":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U","download_json":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U.json","view_paper":"https://pith.science/paper/JY6E7UAB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.28824&json=true","fetch_graph":"https://pith.science/api/pith-number/JY6E7UABYEP6V3FVABAHJMDE3U/graph.json","fetch_events":"https://pith.science/api/pith-number/JY6E7UABYEP6V3FVABAHJMDE3U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U/action/storage_attestation","attest_author":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U/action/author_attestation","sign_citation":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U/action/citation_signature","submit_replication":"https://pith.science/pith/JY6E7UABYEP6V3FVABAHJMDE3U/action/replication_record"}},"created_at":"2026-08-03T01:16:53.449799+00:00","updated_at":"2026-08-03T01:16:53.449799+00:00"}