<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.sitemaps.org/schemas/sitemap/0.9 http://www.sitemaps.org/schemas/sitemap/0.9/sitemap.xsd" xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
<url>
<loc>https://blog.cptz.space/posts/build-your-own-image-bed/</loc>
<lastmod>2023-05-10T20:55:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/fundamentals-of-distributed-systems/</loc>
<lastmod>2023-05-12T20:23:15+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-learning-series/</loc>
<lastmod>2026-07-09T09:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-stack-topology/</loc>
<lastmod>2026-07-09T09:05:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-rank-process-group/</loc>
<lastmod>2026-07-09T09:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-stack-version-boundary/</loc>
<lastmod>2026-07-10T09:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-cuda-async-completion/</loc>
<lastmod>2026-07-10T09:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-collectives/</loc>
<lastmod>2026-07-10T09:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-group-ordering/</loc>
<lastmod>2026-07-10T09:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-collective-decomposition/</loc>
<lastmod>2026-07-10T09:40:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-performance-model-statistics/</loc>
<lastmod>2026-07-10T09:50:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-tests-benchmark/</loc>
<lastmod>2026-07-10T10:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-latency-bandwidth-model/</loc>
<lastmod>2026-07-10T10:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-parameter-sweeps/</loc>
<lastmod>2026-07-10T10:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-ring-allreduce-steps/</loc>
<lastmod>2026-07-10T10:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-tree-double-tree-hierarchy/</loc>
<lastmod>2026-07-10T16:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-simple-ll-ll128-protocols/</loc>
<lastmod>2026-07-10T17:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-channel-chunk-slice-step-buffer/</loc>
<lastmod>2026-07-10T20:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-tuning-model-auto-selection-plugin/</loc>
<lastmod>2026-07-10T20:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-bootstrap-communicator-init-state-machine/</loc>
<lastmod>2026-07-11T10:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-topology-xml-path-computation/</loc>
<lastmod>2026-07-11T12:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-graph-search-ring-tree-channel-construction/</loc>
<lastmod>2026-07-11T15:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-transport-selection-connector-lifecycle/</loc>
<lastmod>2026-07-11T20:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-p2p-cuda-ipc-shm-copy-engine/</loc>
<lastmod>2026-07-11T21:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-enqueue-task-plan-work-batch-cuda-graph/</loc>
<lastmod>2026-07-11T22:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-device-kernel-primitives-step-flag/</loc>
<lastmod>2026-07-11T22:50:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-cpu-proxy-net-progress-socket/</loc>
<lastmod>2026-07-11T23:50:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-socket-infiniband-roce-verbs/</loc>
<lastmod>2026-07-11T23:59:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-gpudirect-rdma-mr-registration-cache/</loc>
<lastmod>2026-07-12T00:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-multinic-rail-crossnic-pxn/</loc>
<lastmod>2026-07-12T00:25:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-collnet-sharp-nvls-plugin-fallback/</loc>
<lastmod>2026-07-12T00:35:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-processgroupnccl-stream-event-work/</loc>
<lastmod>2026-07-12T00:50:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-ddp-reducer-bucket-overlap/</loc>
<lastmod>2026-07-12T01:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-zero-fsdp-state-sharding/</loc>
<lastmod>2026-07-12T01:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-tp-pp-ep-multidim-process-groups/</loc>
<lastmod>2026-07-12T02:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-collective-mismatch-hang-rank-crash-root-cause/</loc>
<lastmod>2026-07-12T02:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-debug-nvtx-nsight-flight-recorder/</loc>
<lastmod>2026-07-12T02:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-production-baseline-regression-version-governance/</loc>
<lastmod>2026-07-12T02:35:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-allreduce-end-to-end-capstone/</loc>
<lastmod>2026-07-12T02:40:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-interview-guide/</loc>
<lastmod>2026-07-15T09:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nccl-interview-deep-dive-labs/</loc>
<lastmod>2026-07-17T16:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-series/</loc>
<lastmod>2026-07-20T00:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-1-foundations/</loc>
<lastmod>2026-07-20T00:11:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-2-lifecycle/</loc>
<lastmod>2026-07-20T00:12:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-3-scheduling/</loc>
<lastmod>2026-07-20T00:13:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-4-kv-runner/</loc>
<lastmod>2026-07-20T00:14:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-5-advanced-distributed/</loc>
<lastmod>2026-07-20T00:15:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-level-6-system-design/</loc>
<lastmod>2026-07-20T00:16:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-comprehensive-guide/</loc>
<lastmod>2026-07-20T00:17:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-interview-source-labs/</loc>
<lastmod>2026-07-20T00:18:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-request-lifecycle-source-guide/</loc>
<lastmod>2026-07-20T00:19:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/sglang-request-lifecycle-source-guide/</loc>
<lastmod>2026-07-20T00:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-series/</loc>
<lastmod>2026-07-20T16:00:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-1-gpu-performance/</loc>
<lastmod>2026-07-20T16:01:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-2-training-runtime/</loc>
<lastmod>2026-07-20T16:02:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-3-model-parallel/</loc>
<lastmod>2026-07-20T16:03:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-4-inference-serving/</loc>
<lastmod>2026-07-20T16:04:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-5-kernel-compiler/</loc>
<lastmod>2026-07-20T16:05:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-6-production-reliability/</loc>
<lastmod>2026-07-20T16:06:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-7-project-defense/</loc>
<lastmod>2026-07-20T16:07:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-8-system-design/</loc>
<lastmod>2026-07-20T16:08:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-9-engineering-deep-dive/</loc>
<lastmod>2026-07-20T16:09:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-level-10-runtime-contracts/</loc>
<lastmod>2026-07-20T16:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/aiinfra-interview-practical-guide/</loc>
<lastmod>2026-07-20T16:11:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-series/</loc>
<lastmod>2026-07-20T16:12:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-0-architecture/</loc>
<lastmod>2026-07-20T16:13:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-1-model-dataflow/</loc>
<lastmod>2026-07-20T16:14:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-2-tensor-sequence-parallel/</loc>
<lastmod>2026-07-20T16:15:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-3-pipeline-parallel/</loc>
<lastmod>2026-07-20T16:16:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-4-distributed-optimizer/</loc>
<lastmod>2026-07-20T16:17:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-5-context-long-sequence/</loc>
<lastmod>2026-07-20T16:18:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-6-moe-expert-parallel/</loc>
<lastmod>2026-07-20T16:19:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-7-performance-precision/</loc>
<lastmod>2026-07-20T16:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-8-checkpoint-reliability/</loc>
<lastmod>2026-07-20T16:21:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-level-9-expert-systems/</loc>
<lastmod>2026-07-20T16:22:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/megatron-interview-study-labs/</loc>
<lastmod>2026-07-20T16:23:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/vllm-sglang-scheduler-source-deep-dive/</loc>
<lastmod>2026-07-21T00:05:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/nvidia-gpu-volta-to-blackwell/</loc>
<lastmod>2026-07-21T10:30:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-01-unified-memory/</loc>
<lastmod>2026-07-22T08:01:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-02-cuda-graphs/</loc>
<lastmod>2026-07-22T08:02:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-03-stream-ordered-memory-allocator/</loc>
<lastmod>2026-07-22T08:03:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-04-cooperative-groups/</loc>
<lastmod>2026-07-22T08:04:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-05-programmatic-dependent-launch/</loc>
<lastmod>2026-07-22T08:05:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-06-green-contexts/</loc>
<lastmod>2026-07-22T08:06:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-07-lazy-loading/</loc>
<lastmod>2026-07-22T08:07:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-08-error-log-management/</loc>
<lastmod>2026-07-22T08:08:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-09-asynchronous-barriers/</loc>
<lastmod>2026-07-22T08:09:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-10-pipelines/</loc>
<lastmod>2026-07-22T08:10:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-11-asynchronous-data-copies/</loc>
<lastmod>2026-07-22T08:11:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-12-cluster-launch-control/</loc>
<lastmod>2026-07-22T08:12:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-13-l2-cache-control/</loc>
<lastmod>2026-07-22T08:13:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-14-memory-synchronization-domains/</loc>
<lastmod>2026-07-22T08:14:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-15-interprocess-communication/</loc>
<lastmod>2026-07-22T08:15:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-16-virtual-memory-management/</loc>
<lastmod>2026-07-22T08:16:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-17-extended-gpu-memory/</loc>
<lastmod>2026-07-22T08:17:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-18-dynamic-parallelism/</loc>
<lastmod>2026-07-22T08:18:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-19-graphics-interop/</loc>
<lastmod>2026-07-22T08:19:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/posts/cuda-features-20-driver-entry-point-access/</loc>
<lastmod>2026-07-22T08:20:00+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/categories/</loc>
<lastmod>2026-07-22T11:25:11+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/tags/</loc>
<lastmod>2026-07-22T11:25:11+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/archives/</loc>
<lastmod>2026-07-22T11:25:11+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/about/</loc>
<lastmod>2026-07-22T11:25:11+08:00</lastmod>
</url>
<url>
<loc>https://blog.cptz.space/readme/2019-08-08-text-and-typography.html</loc>
</url>
<url>
<loc>https://blog.cptz.space/readme/2019-08-08-write-a-new-post.html</loc>
</url>
<url>
<loc>https://blog.cptz.space/readme/2019-08-09-getting-started.html</loc>
</url>
<url>
<loc>https://blog.cptz.space/readme/2019-08-11-customize-the-favicon.html</loc>
</url>
<url>
<loc>https://blog.cptz.space/readme/2021-01-03-enable-google-pv.html</loc>
</url>
<url>
<loc>https://blog.cptz.space/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/minio/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/image-bed/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/typora/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/distributed-system/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/notes/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nccl/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/gpu/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ddp/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/infra/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvlink/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/pcie/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/numa/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/topology/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/rank/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/communicator/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/process-group/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/torchrun/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-visible-devices/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/pytorch/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/elf/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/dynamic-linker/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/stream/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/event/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/asynchronous/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/collective/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/all-reduce/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/all-gather/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/reduce-scatter/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/all-to-all/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/group/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ordering/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/timeout/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/watchdog/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/processgroup/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/fsdp/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/zero/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nccl-tests/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/algbw/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/busbw/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/benchmark/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/statistics/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/performance/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/latency/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/bandwidth/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tuning/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/performance-model/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/regression/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/transport/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/channel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/troubleshooting/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ring/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tree/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/double-tree/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/allreduce/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/simple/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ll/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ll128/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/protocol/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/chunk/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/slice/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/buffer/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/algorithm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/plugin/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/bootstrap/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/debugging/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/xml/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/graph/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/p2p/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/shm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/network/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/connector/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-ipc/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cumem/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/enqueue/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/planner/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-graph/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvtx/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nsys/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-kernel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/primitives/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/synchronization/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/proxy/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/socket/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cpu-affinity/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nsight/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/infiniband/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/roce/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/rdma/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/verbs/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/gpudirect-rdma/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/dmabuf/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/memory-registration/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/iommu/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/multi-nic/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/rail/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/pxn/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/collnet/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/sharp/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvls/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvswitch/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/processgroupnccl/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-stream/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/worknccl/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/reducer/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/bucket/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/overlap/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/optimizer/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/sharding/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tensor-parallel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/pipeline-parallel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/expert-parallel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/moe/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/failure/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/root-cause/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/debug/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nsight-systems/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/flight-recorder/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/observability/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/production/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/canary/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/baseline/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/governance/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/capstone/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/callgraph/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/diagnosis/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/interview/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/vllm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/sglang/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/experiment/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/llm-inference/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/kv-cache/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/scheduler/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/architecture/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ai-infra/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/distributed-training/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/megatron/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/megatron-core/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/model-parallel/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/continuous-batching/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/chunked-prefill/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvidia/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/gpu-architecture/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tensor-core/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/volta/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/turing/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ampere/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ada/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/hopper/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/blackwell/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tma/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/wgmma/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/tmem/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/unified-memory/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/hmm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/memory/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-graphs/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/launch-overhead/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/runtime/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/memory-pool/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cuda-malloc-async/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/allocator/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cooperative-groups/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/collectives/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/dependent-launch/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/green-context/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/qos/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/resource-partition/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/lazy-loading/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/module/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/startup/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/logging/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/diagnostics/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/driver-api/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/barrier/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/pipeline/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/double-buffering/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/async-copy/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cp-async/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/stas/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cluster-launch-control/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/work-stealing/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/scheduling/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/l2-cache/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/persisting-cache/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/memory-model/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/fence/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/ipc/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/vmm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/multi-gpu/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/virtual-memory/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/multicast/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/egm/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvlink-c2c/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/grace-hopper/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/dynamic-parallelism/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/cdp2/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/device-runtime/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/vulkan/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/opengl/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/direct3d/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/nvsci/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/interop/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/abi/</loc>
</url>
<url>
<loc>https://blog.cptz.space/tags/dynamic-loading/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/experience/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/tutorial/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/reading-notes/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/%E6%B7%B1%E5%85%A5%E7%90%86%E8%A7%A3%E5%88%86%E5%B8%83%E5%BC%8F%E7%B3%BB%E7%BB%9F/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/ai-infra/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/nccl/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/inference/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/interview/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/megatron/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/gpu/</loc>
</url>
<url>
<loc>https://blog.cptz.space/categories/cuda-features/</loc>
</url>
<url>
<loc>https://blog.cptz.space/page2/</loc>
</url>
<url>
<loc>https://blog.cptz.space/page3/</loc>
</url>
<url>
<loc>https://blog.cptz.space/page4/</loc>
</url>
<url>
<loc>https://blog.cptz.space/page5/</loc>
</url>
</urlset>
