{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QTJH5BAWVL2TMTKJOSQSO2TULT","short_pith_number":"pith:QTJH5BAW","schema_version":"1.0","canonical_sha256":"84d27e8416aaf5364d4974a1276a745cf20d03aa1dc22332e3b939cd2399fa92","source":{"kind":"arxiv","id":"2410.23668","version":1},"attestation_state":"computed","paper":{"title":"Kernel Looping: Eliminating Synchronization Boundaries for Peak Inference Performance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.CL","authors_text":"Angela Wang, Darshan Gandhi, David Koeplinger, Han Wang, Leon Zhang, Matheen Musaddiq, Matthew Shaffer, Mingran Wang, Nathan Sheeley, Pushkar Nandkar, Raghu Prabhakar, Reid Goodbar","submitted_at":"2024-10-31T06:32:47Z","abstract_excerpt":"Token generation speed is critical to power the next wave of AI inference applications. GPUs significantly underperform during token generation due to synchronization overheads at kernel boundaries, utilizing only 21% of their peak memory bandwidth. While recent dataflow architectures mitigate these overheads by enabling aggressive fusion of decoder layers into a single kernel, they too leave performance on the table due to synchronization penalties at layer boundaries.\n  This paper presents kernel looping, a specialized global optimization technique which exploits an optimization opportunity "},"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":"2410.23668","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-31T06:32:47Z","cross_cats_sorted":["cs.AI","cs.AR"],"title_canon_sha256":"b7487ea4f7d832ab4653ed1c5560cad0ed291f57a03234df7b6e6768b993d432","abstract_canon_sha256":"7b7416dc2dbeb13add6bc12b55b00fa68e8a7e68494170307dadb33023f1cde9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:29:03.635177Z","signature_b64":"KxszhKTL/4DtvhrGfdXfAPA8R+5f1vz/F2/h5niFL8UW+kd36ks4zF0Tair+Dd669rD8UlkJ7qiFv/IV+kX7Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"84d27e8416aaf5364d4974a1276a745cf20d03aa1dc22332e3b939cd2399fa92","last_reissued_at":"2026-07-05T09:29:03.634671Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:29:03.634671Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Kernel Looping: Eliminating Synchronization Boundaries for Peak Inference Performance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.CL","authors_text":"Angela Wang, Darshan Gandhi, David Koeplinger, Han Wang, Leon Zhang, Matheen Musaddiq, Matthew Shaffer, Mingran Wang, Nathan Sheeley, Pushkar Nandkar, Raghu Prabhakar, Reid Goodbar","submitted_at":"2024-10-31T06:32:47Z","abstract_excerpt":"Token generation speed is critical to power the next wave of AI inference applications. GPUs significantly underperform during token generation due to synchronization overheads at kernel boundaries, utilizing only 21% of their peak memory bandwidth. While recent dataflow architectures mitigate these overheads by enabling aggressive fusion of decoder layers into a single kernel, they too leave performance on the table due to synchronization penalties at layer boundaries.\n  This paper presents kernel looping, a specialized global optimization technique which exploits an optimization opportunity "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.23668","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/2410.23668/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":"2410.23668","created_at":"2026-07-05T09:29:03.634735+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.23668v1","created_at":"2026-07-05T09:29:03.634735+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.23668","created_at":"2026-07-05T09:29:03.634735+00:00"},{"alias_kind":"pith_short_12","alias_value":"QTJH5BAWVL2T","created_at":"2026-07-05T09:29:03.634735+00:00"},{"alias_kind":"pith_short_16","alias_value":"QTJH5BAWVL2TMTKJ","created_at":"2026-07-05T09:29:03.634735+00:00"},{"alias_kind":"pith_short_8","alias_value":"QTJH5BAW","created_at":"2026-07-05T09:29:03.634735+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/QTJH5BAWVL2TMTKJOSQSO2TULT","json":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT.json","graph_json":"https://pith.science/api/pith-number/QTJH5BAWVL2TMTKJOSQSO2TULT/graph.json","events_json":"https://pith.science/api/pith-number/QTJH5BAWVL2TMTKJOSQSO2TULT/events.json","paper":"https://pith.science/paper/QTJH5BAW"},"agent_actions":{"view_html":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT","download_json":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT.json","view_paper":"https://pith.science/paper/QTJH5BAW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.23668&json=true","fetch_graph":"https://pith.science/api/pith-number/QTJH5BAWVL2TMTKJOSQSO2TULT/graph.json","fetch_events":"https://pith.science/api/pith-number/QTJH5BAWVL2TMTKJOSQSO2TULT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT/action/storage_attestation","attest_author":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT/action/author_attestation","sign_citation":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT/action/citation_signature","submit_replication":"https://pith.science/pith/QTJH5BAWVL2TMTKJOSQSO2TULT/action/replication_record"}},"created_at":"2026-07-05T09:29:03.634735+00:00","updated_at":"2026-07-05T09:29:03.634735+00:00"}