<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Light Space</title><description>记录 Systems、Performance、Hardware、Research 方向的技术笔记，关注 HPC、Quantum Computation、FPGA、AI Infrastructure 与工程决策。</description><link>https://frantice.com/</link><language>zh-cn</language><item><title>我如何设计一个 AI + GraphQL + OCR 的可扩展系统</title><link>https://frantice.com/posts/system-architecture-notes/designing-ai-graphql-ocr-systems/</link><guid isPermaLink="true">https://frantice.com/posts/system-architecture-notes/designing-ai-graphql-ocr-systems/</guid><description>从服务边界、任务流转到权限与部署，拆解一个可维护的 AI 数据处理系统。</description><pubDate>Thu, 12 Mar 2026 00:00:00 GMT</pubDate><category>system-design</category><category>graphql</category><category>fastapi</category><category>infra</category></item><item><title>为什么很多程序慢，不是因为算力不够，而是因为内存</title><link>https://frantice.com/posts/compute-bottlenecks/memory-bottlenecks-before-compute/</link><guid isPermaLink="true">https://frantice.com/posts/compute-bottlenecks/memory-bottlenecks-before-compute/</guid><description>讨论带宽、缓存与数据移动成本，解释为什么很多性能问题首先是内存问题。</description><pubDate>Tue, 10 Mar 2026 00:00:00 GMT</pubDate><category>performance</category><category>hpc</category><category>memory</category><category>gpu</category></item><item><title>量子模拟的瓶颈到底在哪里</title><link>https://frantice.com/posts/quantum-simulation-notes/quantum-simulation-memory-wall/</link><guid isPermaLink="true">https://frantice.com/posts/quantum-simulation-notes/quantum-simulation-memory-wall/</guid><description>从状态向量规模、内存访问模式和设备带宽解释量子模拟为何常常撞上内存墙。</description><pubDate>Sat, 07 Mar 2026 00:00:00 GMT</pubDate><category>quantum-simulation</category><category>hpc</category><category>performance</category><category>memory</category></item><item><title>FP16、FP32、FP64：精度选择背后的工程现实</title><link>https://frantice.com/posts/compute-bottlenecks/precision-is-a-system-choice/</link><guid isPermaLink="true">https://frantice.com/posts/compute-bottlenecks/precision-is-a-system-choice/</guid><description>精度不是数学洁癖，而是吞吐、误差、带宽和可维护性之间的系统级权衡。</description><pubDate>Tue, 03 Mar 2026 00:00:00 GMT</pubDate><category>precision</category><category>performance</category><category>gpu</category><category>hpc</category></item><item><title>为什么 FPGA 仍然是异构计算里最被低估的一环</title><link>https://frantice.com/posts/hardware-boundaries/fpga-is-still-undervalued/</link><guid isPermaLink="true">https://frantice.com/posts/hardware-boundaries/fpga-is-still-undervalued/</guid><description>不是所有问题都该交给 GPU，FPGA 的价值在于可预测延迟、数据路径控制与专用流水线。</description><pubDate>Sat, 28 Feb 2026 00:00:00 GMT</pubDate><category>fpga</category><category>hardware</category><category>pcie</category><category>acceleration</category></item><item><title>从 Docker 到 pipeline：我如何组织多服务计算系统</title><link>https://frantice.com/posts/system-architecture-notes/docker-pipelines-for-compute-systems/</link><guid isPermaLink="true">https://frantice.com/posts/system-architecture-notes/docker-pipelines-for-compute-systems/</guid><description>不是为了“微服务而微服务”，而是为了让任务流、依赖和部署路径保持可控。</description><pubDate>Tue, 24 Feb 2026 00:00:00 GMT</pubDate><category>docker</category><category>infra</category><category>pipeline</category><category>backend</category></item><item><title>GPU、CPU、FPGA 各自适合解决什么问题</title><link>https://frantice.com/posts/hardware-boundaries/cpu-gpu-fpga-boundaries/</link><guid isPermaLink="true">https://frantice.com/posts/hardware-boundaries/cpu-gpu-fpga-boundaries/</guid><description>对比三类计算资源的长处与边界，避免把硬件选择变成流行趋势投票。</description><pubDate>Thu, 19 Feb 2026 00:00:00 GMT</pubDate><category>hardware</category><category>gpu</category><category>fpga</category><category>system-design</category></item><item><title>做系统的人，应该如何判断一项技术是否值得投入</title><link>https://frantice.com/posts/standalone/technology-decisions-for-system-builders/</link><guid isPermaLink="true">https://frantice.com/posts/standalone/technology-decisions-for-system-builders/</guid><description>面对新框架、新芯片和新工具链时，系统工程师更需要判断成本结构，而不是追逐热度。</description><pubDate>Sat, 14 Feb 2026 00:00:00 GMT</pubDate><category>writing</category><category>product-thinking</category><category>system-design</category></item><item><title>如何安全管理 secrets：敏感信息管理的演化</title><link>https://frantice.com/posts/standalone/local-secrets-and-production-boundaries/</link><guid isPermaLink="true">https://frantice.com/posts/standalone/local-secrets-and-production-boundaries/</guid><description>从 .env 到密钥托管，梳理一种更适合团队协作与部署演进的 secrets 管理方式。</description><pubDate>Mon, 09 Feb 2026 00:00:00 GMT</pubDate><category>infra</category><category>docker</category><category>backend</category><category>security</category></item><item><title>Xilinx FPGA Toolchain避险指南</title><link>https://frantice.com/posts/hardware-boundaries/Xilinx-FPGA-Toolchain-Troubleshooting/</link><guid isPermaLink="true">https://frantice.com/posts/hardware-boundaries/Xilinx-FPGA-Toolchain-Troubleshooting/</guid><description>分享一些在安装和使用Vivado，Vitis等工具的过程中出现的问题和解决方案</description><pubDate>Fri, 28 Feb 2025 00:00:00 GMT</pubDate><category>fpga</category></item><item><title>关于实验进度追不上截稿日被迫使用GPU做异构计算这件事</title><link>https://frantice.com/posts/compute-bottlenecks/using-gpu-in-exp/</link><guid isPermaLink="true">https://frantice.com/posts/compute-bottlenecks/using-gpu-in-exp/</guid><description>NULL</description><pubDate>Sat, 21 Sep 2024 00:00:00 GMT</pubDate><category>GPU</category></item><item><title>GAS与Clasp的初步上手</title><link>https://frantice.com/posts/standalone/the-very-first-touch-on-gas-and-clasp/</link><guid isPermaLink="true">https://frantice.com/posts/standalone/the-very-first-touch-on-gas-and-clasp/</guid><description>初步认识Google App Script以及管理工具Clasp</description><pubDate>Thu, 30 May 2024 00:00:00 GMT</pubDate><category>gas</category><category>script</category></item></channel></rss>