Unverified50% confidenceFactExact time
64 is the upper threshold for the number of subtransaction IDs each backend process caches in the PGPROC structure; subtransactions exceeding this count overflow into the shared slot mechanism.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverifieddatasette-ip-rate-limit使用内存存储追踪IP状态，是进程级别的限流，重启后状态会丢失61% similarUnverifiedWhen the number of subtransactions exceeds 64 and a subxid overflow occurs, other transactions must frequently access the pg_subtrans SLRU cache during MVCC visibility checks, leading to intensified lock contention, increased disk I/O, and reduced throughput.60% similarUnverifiedThe LLM decoding process is memory bound, with the GPU spending most of its time waiting for data transfers during token-by-token generation.59% similarUnverifiedCompressing child node indices from 64-bit pointers to 32-bit offsets both saves memory and allows a single 64-byte cache line to hold more index information.57% similarUnverifiedAfter all 31 layers of the 26B model are offloaded to the GPU, ollama ps still reports 14%/86% CPU/GPU, which originates from mmap-mapped CPU-side address space tensor buffers rather than actual CPU computation.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/914606API
curl https://kongchang.com/api/v1/knowledge/claims/914606MCP
get_claim(id=914606)