{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:ES6FD7Y75ZGSP2R5RY2CIQORGF","short_pith_number":"pith:ES6FD7Y7","schema_version":"1.0","canonical_sha256":"24bc51ff1fee4d27ea3d8e342441d13148f30bfb51e843e50b9864065d044b78","source":{"kind":"arxiv","id":"2008.08708","version":2},"attestation_state":"computed","paper":{"title":"Synthesizing Optimal Collective Algorithms","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Jacob Nelson, Madan Musuvathi, Olli Saarikivi, Saeed Maleki, Todd Mytkowicz, Zhengyang Liu, Zixian Cai","submitted_at":"2020-08-19T23:51:23Z","abstract_excerpt":"Collective communication algorithms are an important component of distributed computation. Indeed, in the case of deep-learning, collective communication is the Amdahl's bottleneck of data-parallel training.\n  This paper introduces SCCL (for Synthesized Collective Communication Library), a systematic approach to synthesize collective communication algorithms that are explicitly tailored to a particular hardware topology. SCCL synthesizes algorithms along the Pareto-frontier spanning from latency-optimal to bandwidth-optimal implementations of a collective. The paper demonstrates how to encode "},"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":"2008.08708","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2020-08-19T23:51:23Z","cross_cats_sorted":[],"title_canon_sha256":"e30f3edc43cbb27e8188ac775bc097424751fef54d221e616031a69aec8db971","abstract_canon_sha256":"02b62dd593f49ac5be995a4738519717c0e7cb3f4f4ced6b4e5f0ac6e7d01883"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:33:54.570164Z","signature_b64":"tKLaN73Tr/47sXg4li5x4YEyvt2+ALi07ZdEY1jq9GBIu19HrmF4t7q9TsGF6ywaD8rRe0fL6lzipJvEix/WBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24bc51ff1fee4d27ea3d8e342441d13148f30bfb51e843e50b9864065d044b78","last_reissued_at":"2026-07-05T02:33:54.569617Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:33:54.569617Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Synthesizing Optimal Collective Algorithms","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Jacob Nelson, Madan Musuvathi, Olli Saarikivi, Saeed Maleki, Todd Mytkowicz, Zhengyang Liu, Zixian Cai","submitted_at":"2020-08-19T23:51:23Z","abstract_excerpt":"Collective communication algorithms are an important component of distributed computation. Indeed, in the case of deep-learning, collective communication is the Amdahl's bottleneck of data-parallel training.\n  This paper introduces SCCL (for Synthesized Collective Communication Library), a systematic approach to synthesize collective communication algorithms that are explicitly tailored to a particular hardware topology. SCCL synthesizes algorithms along the Pareto-frontier spanning from latency-optimal to bandwidth-optimal implementations of a collective. The paper demonstrates how to encode "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2008.08708","kind":"arxiv","version":2},"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/2008.08708/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":"2008.08708","created_at":"2026-07-05T02:33:54.569689+00:00"},{"alias_kind":"arxiv_version","alias_value":"2008.08708v2","created_at":"2026-07-05T02:33:54.569689+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2008.08708","created_at":"2026-07-05T02:33:54.569689+00:00"},{"alias_kind":"pith_short_12","alias_value":"ES6FD7Y75ZGS","created_at":"2026-07-05T02:33:54.569689+00:00"},{"alias_kind":"pith_short_16","alias_value":"ES6FD7Y75ZGSP2R5","created_at":"2026-07-05T02:33:54.569689+00:00"},{"alias_kind":"pith_short_8","alias_value":"ES6FD7Y7","created_at":"2026-07-05T02:33:54.569689+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2309.14509","citing_title":"DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models","ref_index":90,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF","json":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF.json","graph_json":"https://pith.science/api/pith-number/ES6FD7Y75ZGSP2R5RY2CIQORGF/graph.json","events_json":"https://pith.science/api/pith-number/ES6FD7Y75ZGSP2R5RY2CIQORGF/events.json","paper":"https://pith.science/paper/ES6FD7Y7"},"agent_actions":{"view_html":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF","download_json":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF.json","view_paper":"https://pith.science/paper/ES6FD7Y7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2008.08708&json=true","fetch_graph":"https://pith.science/api/pith-number/ES6FD7Y75ZGSP2R5RY2CIQORGF/graph.json","fetch_events":"https://pith.science/api/pith-number/ES6FD7Y75ZGSP2R5RY2CIQORGF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF/action/storage_attestation","attest_author":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF/action/author_attestation","sign_citation":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF/action/citation_signature","submit_replication":"https://pith.science/pith/ES6FD7Y75ZGSP2R5RY2CIQORGF/action/replication_record"}},"created_at":"2026-07-05T02:33:54.569689+00:00","updated_at":"2026-07-05T02:33:54.569689+00:00"}