hello cudahttps://hellocuda.com/zh-CNhttps://hellocuda.com/posts/jetson-gstreamer-nvdec-seek/https://hellocuda.com/posts/jetson-gstreamer-nvdec-seek/Jetson 上用 nvv4l2decoder 硬解视频、拖动进度条 seek 必卡死进程——从把 SIGKILL 误判成 OOM,到加日志定位 seek_simple 永久阻塞,最后靠给硬解码器降 rank、让 decodebin 自动切软解(avdec)彻底解决。附硬解/软解、NVMM、decodebin rank 的背景知识。Sat, 04 Jul 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-10-idempotency-and-fixed-point/https://hellocuda.com/posts/compiler-10-idempotency-and-fixed-point/从一个 shell 脚本的'幂等'说起——同一个函数反复跑结果不变,这个朴素概念在编译器里有一个精确对应:不动点(fixed point)。本文讲清楚为什么 DCE、常量折叠这种优化 pass 总是写成'loop until no change',以及不动点之外的汇合性、单调性、终止性这几个相关概念。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cheatsheet/gpu-spec/https://hellocuda.com/posts/cheatsheet/gpu-spec/常用 GPU 的硬件数字:峰值算力、带宽、SM 数、内存层级延迟、临界算术强度。Roofline 分析和性能预估的基础参考。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/memory/bank-conflict/https://hellocuda.com/posts/memory/bank-conflict/shared memory 物理上是 32 个并行 bank,warp 内多个 lane 同时访问同一个 bank → 串行化。理解了 bank 划分规则,所有冲突场景都能口算。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/memory/coalescing/https://hellocuda.com/posts/memory/coalescing/warp 内 32 个 thread 同时访问 global memory,什么情况下能合并成 1 次事务、什么情况下退化成 32 次。CUDA 性能的两大支柱之一(另一个是 bank conflict)。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/threads/index-and-bounds/https://hellocuda.com/posts/threads/index-and-bounds/每个 kernel 必写的两行——算出全局 tid + tail guard。1D / 2D / 3D 索引模板,以及当数据量不是 blockDim 整数倍时的标准处理。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/threads/thread-block-grid/https://hellocuda.com/posts/threads/thread-block-grid/CUDA 的三层并行抽象。一行 kernel<<<g, b>>>() 启动多少线程、它们怎么分组、各组之间能不能通信。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/threads/warp-and-scheduling/https://hellocuda.com/posts/threads/warp-and-scheduling/block 是逻辑分组,warp 才是硬件执行单位——32 个 lane 严格锁步。理解 warp 是搞清 CUDA 性能的钥匙:延迟隐藏、分支分歧、shuffle 通通围绕它。Tue, 16 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-09-optimization-pass-glossary/https://hellocuda.com/posts/compiler-09-optimization-pass-glossary/编译器优化 pass 的名字一大堆——DCE、GVN、LICM、SCCP、SROA……每个都是一个独立的算法 + 一段术语。本文按'本地 / 函数级 / 循环 / 跨函数 / 后端 / 特殊'六类组织,每个 pass 给出英文全称、中文意思、做什么、为什么有用——读 LLVM 源码或自己写 pass 时当词典查。Mon, 15 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-18-custom-intrinsic-walkthrough/https://hellocuda.com/posts/cuda-oxide-18-custom-intrinsic-walkthrough/走一遍真实流程:在 cuda-oxide 里加一个返回常量 42 的自定义 intrinsic xxx(),最终在 PTX 中通过内联汇编 mov.u32 %r1, 42; 实现。五个 crate 要改、两个错误版本一定会踩、convergent 属性是正确性而不是性能开关。把整套实战拆给你看。Mon, 15 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-08-mem2reg-why-it-works/https://hellocuda.com/posts/compiler-08-mem2reg-why-it-works/mem2reg 把'栈上变量 + load/store'提升成 SSA 寄存器值——这是打开后续一切优化的钥匙。但它能这么做,靠的是一个非常脆弱的前提:'编译器看得见这块内存的所有读写'。一旦 alloca 的地址逃逸到 load/store 之外的地方,这个前提就崩了,promote 就会产生错误的程序。本文用三个具体场景把'地址逃逸'讲透。Sat, 13 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-17-ir-comparison-across-stages/https://hellocuda.com/posts/cuda-oxide-17-ir-comparison-across-stages/拆完 11 步流程之后,把同一段代码在每个阶段的真实输出并排放一起。整个 hello_kernel(3 行 Rust → 6 行 PTX)横跨 7 层 IR 的对照,加上两个单语句的逐层追踪——*out = 42 和 thread::xxx()。看信息怎么逐层丢失,形态怎么逐层接近硬件。Sat, 13 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-03-mlir-dialect-levels/https://hellocuda.com/posts/compiler-03-mlir-dialect-levels/MLIR 方言的'高低'不是官方分级,而是按抽象层次来理解的。判断标准就一条:它离'人怎么想'近,还是离'机器怎么执行'近?用同一个矩阵乘从一行变成几十行的例子串起来看。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-04-rust-monomorphization/https://hellocuda.com/posts/compiler-04-rust-monomorphization/Rust 的单态化跟很多人想的不一样——它不在编译期预先生成每个泛型实例的独立 MIR,而是用 Instance(def + args) 这种'定义 + 参数'的轻量结构,在 codegen 时按需替换。本文讲清楚 Instance 类型、惰性单态化、is_fully_monomorphized 的两个检查,以及 cuda-oxide 怎么用它过滤泛型代码。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-06-ptx-introduction/https://hellocuda.com/posts/compiler-06-ptx-introduction/PTX 是 NVIDIA GPU 的虚拟 ISA,介于 LLVM IR 和真正的 GPU 机器码 SASS 之间。它跟 x86 完全不同——指令格式统一、寄存器无限、地址空间显式。本文讲清楚 PTX 指令格式、state space(地址空间)、常用指令族,最后用一段 5 行的 hello_kernel PTX 逐行拆解。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-07-inline-ptx-templates/https://hellocuda.com/posts/compiler-07-inline-ptx-templates/会读 PTX 之后,下一步是从 Rust 里写自己的 PTX——这是实现自定义 GPU intrinsic 的必备技能。本文讲清楚 inline asm 的'模板 + 约束串'两段对应关系、临时寄存器作用域、谓词输出 selp 套路、format! 动态生成的陷阱,以及一个'读 shared memory 比对'的完整合成例子。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-05-mem2reg/https://hellocuda.com/posts/compiler-05-mem2reg/几乎所有编译器前端都用 alloca-load-store 模型生成中间 IR——简单但冗余。mem2reg(memory-to-register,把'内存槽'升级回'寄存器值')是把这些冗余消掉、恢复 SSA 形式的经典优化。本文讲清楚 promotable alloca 的判定、phi 节点(φ,SSA 里的汇合节点)的插入、它在 cuda-oxide pipeline 里的位置,以及为什么它从'可选'变成了'必需'。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-09-codegen-crate-entry/https://hellocuda.com/posts/cuda-oxide-09-codegen-crate-entry/拆 hello-constant 系列第四站。终于进入我们自己的代码——librustc_codegen_cuda.so 被 rustc dlopen 之后,__rustc_codegen_backend 返回一个包装了 LlvmCodegenBackend 的对象,后续 trait 调用大多转发,只在 codegen_crate 这一处拦截 device 路径。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-10-kernel-naming-convention/https://hellocuda.com/posts/cuda-oxide-10-kernel-naming-convention/拆 hello-constant 系列第五站。codegen_crate 检测到 device code 之后,先靠命名约定(KERNEL_PREFIX 加魔数 246e25db)识别 kernel,worklist BFS 遍历调用图收集所有可达的 device 函数,最后通过 rustc_public 把 internal Instance 转成 stable Instance 喂给 mir-importer。本文讲清楚整条 collector 流程,以及背后那串 246e25db 是怎么来的。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-11-mir-to-dialect-mir-statements/https://hellocuda.com/posts/cuda-oxide-11-mir-to-dialect-mir-statements/拆 hello-constant 系列第六站。stable MIR 喂进 mir-importer 后,逐函数翻译成 dialect-mir 形式。每个 local 都翻译成栈槽(mir.alloca),每条 statement 翻译成 1-3 条 dialect-mir op。用 *out = 42 这一行对照真实 dump 看完整过程。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-12-terminator-and-intrinsic-dispatch/https://hellocuda.com/posts/cuda-oxide-12-terminator-and-intrinsic-dispatch/拆 hello-constant 系列第七站。每个 basic block(基本块)末尾的控制流(rustc 内部叫 Terminator,终结符)是 translator 最复杂的一段。本文按 docs/run.log 真实 dump,讲清楚函数调用为什么必须落在块尾、translate_call 的三步流程、try_dispatch_intrinsic 怎么靠 FQDN(完全限定路径名)字符串识别 GPU intrinsic(编译器内建函数)、以及桩函数体为什么根本没翻译。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-13-mir-lower-dialect-conversion/https://hellocuda.com/posts/cuda-oxide-13-mir-lower-dialect-conversion/拆 hello-constant 系列第八站。dialect-mir + dialect-nvvm 的混合 IR 整体降级到 dialect-llvm,靠的是 pliron 的 DialectConversion 框架——你声明'我能转换什么 op、什么类型、怎么转换',driver 自动遍历 module。本文讲清楚 trait 分发机制、跟 Step 7 字符串 match 的对比、replace_operation 的关键作用、full conversion 模式,以及降级完之后 IR 真实长什么样。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-14-llvm-export-and-llc/https://hellocuda.com/posts/cuda-oxide-14-llvm-export-and-llc/拆 hello-constant 系列第九站。dialect-llvm module 序列化成文本 LLVM IR(.ll),然后调外部 llc 工具用 NVPTX backend 编译成文本 PTX(.ptx)。本文讲清楚 llvm-export 怎么序列化、`llvm_nvvm_*` → `llvm.nvvm.*` 的反向转换、!nvvm.annotations 元数据的作用、llc 调用参数、以及最终 PTX 里每条指令跟 LLVM IR 一一对应的关系。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-15-host-runtime-and-cuda-driver/https://hellocuda.com/posts/cuda-oxide-15-host-runtime-and-cuda-driver/拆 hello-constant 系列第十站。device 编译完成,host 端 binary 启动——通过 CUDA driver API 一步步连 GPU、分配显存、加载 PTX(driver 现场 JIT 编译成 SASS)、launch kernel。本文讲清楚 cuda-core 怎么包装 driver API,以及 cuModuleLoad 触发的 JIT 是整条链路里最'魔法'的一步。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-16-gpu-execution-and-result-copyback/https://hellocuda.com/posts/cuda-oxide-16-gpu-execution-and-result-copyback/拆 hello-constant 系列最终站。cuLaunchKernel 返回后 GPU 接管,256 个线程并行执行 SASS。本文讲清楚 SM 调度 + warp 锁步、printf buffer 的延迟 flush 机制、benign race(良性竞争)、cuStreamSynchronize + cuMemcpyDtoH 的回传链路,以及为什么 thread xxx: 42 在 Output: 42 之前出现。Fri, 12 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-01-nm-symbols/https://hellocuda.com/posts/compiler-01-nm-symbols/nm 输出每个符号前带一个字母,标记它所在段(Section)和可见性。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/compiler-02-codegen-backend-flag/https://hellocuda.com/posts/compiler-02-codegen-backend-flag/rustc 的 -Z codegen-backend 是 nightly 阶段的插件机制,允许把默认 LLVM 后端换成自定义实现——Cranelift、GCC、甚至 CUDA。详细讲它的协议、加载流程、CodegenBackend trait,以及自己写一个要注意什么。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-02-verify-codegen-backend/https://hellocuda.com/posts/cuda-oxide-02-verify-codegen-backend/用 nm + grep 验证 rustc-codegen-cuda 把 __rustc_codegen_backend 这个入口符号正确导出来了。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-03-composition-over-inheritance/https://hellocuda.com/posts/cuda-oxide-03-composition-over-inheritance/面向对象书里讲烂了的 Composition over Inheritance,在 cuda-oxide 的 CudaCodegenBackend 上是怎么落地的——为什么我们不去重写一个 LLVM 后端,而是把它包起来。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-04-hello-constant-intro/https://hellocuda.com/posts/cuda-oxide-04-hello-constant-intro/用最简单的 hello-constant example,把 cuda-oxide 从一条 cargo 命令到 GPU 输出 42 之间发生的所有事,拆成 11 步逐站讲透。这是开篇——讲清楚我们要拆什么、为什么挑这个例子、整条路线长什么样。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-05-hello-constant-pipeline-overview/https://hellocuda.com/posts/cuda-oxide-05-hello-constant-pipeline-overview/拆 hello-constant 系列的第一站。一条 cargo 命令背后启动了两个独立进程、七层 IR 翻译。这一篇画出全景图,后续每篇都会停在图上某个具体方框深入。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-06-cargo-oxide-driver/https://hellocuda.com/posts/cuda-oxide-06-cargo-oxide-driver/拆 hello-constant 系列第二站。停在全景图最上面那个方框,看 cargo-oxide driver 进程在 spawn rustc 之前到底做了哪些事——找 backend .so、拼 RUSTFLAGS、然后挂起等子进程退出。它不做编译,只做配置。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-07-rustc-frontend-to-mir/https://hellocuda.com/posts/cuda-oxide-07-rustc-frontend-to-mir/拆 hello-constant 系列第三站。进入子进程,看 rustc 的前端流程——五段流水线把 src/main.rs 变成 MIR。我们 backend 拿到的输入到底长什么样?为什么 MIR 是恰到好处的过渡点?这一篇讲清楚 backend 的'原料'。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-bonus-backend-dylib/https://hellocuda.com/posts/cuda-oxide-bonus-backend-dylib/聚焦一个 hello-constant 拆解正篇略过的问题:librustc_codegen_cuda.so 到底是怎么编出来的?Cargo.toml 里 dylib、空 [workspace]、#![feature(rustc_private)] 三个不寻常的设计各自解决了什么问题。Thu, 11 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cuda-oxide-01-intro/https://hellocuda.com/posts/cuda-oxide-01-intro/用 Rust 直接写 CUDA kernel,编译成 PTX 跑在 GPU 上——cuda-oxide 是什么、为什么值得研究、这个系列要讲什么。Wed, 10 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/start/hello-cuda/https://hellocuda.com/posts/start/hello-cuda/第一个 kernel,几十行跑起来。Wed, 10 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/start/setup-environment/https://hellocuda.com/posts/start/setup-environment/装驱动 + 装 Toolkit,两条命令验通过。Wed, 10 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/start/what-is-cuda/https://hellocuda.com/posts/start/what-is-cuda/CUDA 是什么、为什么用 GPU、CUDA 的四件套。Wed, 10 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/hello-cuda/https://hellocuda.com/posts/hello-cuda/HelloCUDA 上线了——一个入门 CUDA 编程的地方。Tue, 09 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cheatsheet/nvcc/https://hellocuda.com/posts/cheatsheet/nvcc/CUDA 编译器——基础编译、架构号、多架构、PTX/SASS、链接。Tue, 09 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/cheatsheet/nvidia-smi/https://hellocuda.com/posts/cheatsheet/nvidia-smi/GPU 状态查看工具——基础、监控、显存、拓扑、功耗。Tue, 09 Jun 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/ssa/https://hellocuda.com/posts/2026/ssa/SSA 是编译器领域的一个核心术语Sat, 23 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026-05-canvas-blend-modes/https://hellocuda.com/posts/2026-05-canvas-blend-modes/PixMojo 的 30 多个 look 背后其实只有 11 个 canvas blend 原语。讲清楚每个 blend 做什么，一个照片审美的技术骨架就出来了。Fri, 22 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/unsafe-%E4%B8%8E%E6%9C%AA%E5%AE%9A%E4%B9%89%E8%A1%8C%E4%B8%BAub/https://hellocuda.com/posts/2026/unsafe-%E4%B8%8E%E6%9C%AA%E5%AE%9A%E4%B9%89%E8%A1%8C%E4%B8%BAub/unsafe 与未定义行为 UBTue, 19 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/jetson-%E5%B8%B8%E7%94%A8%E5%91%BD%E4%BB%A4%E9%80%9F%E6%9F%A5%E8%A1%A8/https://hellocuda.com/posts/2026/jetson-%E5%B8%B8%E7%94%A8%E5%91%BD%E4%BB%A4%E9%80%9F%E6%9F%A5%E8%A1%A8/汇总 Jetson 平台开发调试中常用的命令，包括系统监控、GPIO 操作、引脚配置和电源管理Mon, 18 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/aarch64/https://hellocuda.com/posts/2026/aarch64/简要介绍aarch64架构Sun, 17 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/pinmux/https://hellocuda.com/posts/2026/pinmux/通过 Rust 程序和 busybox devmem 两种方式交叉验证 Jetson Orin Nano 40-Pin 接口 Pin 31 的 Pinmux 寄存器地址Sat, 16 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/i2c/https://hellocuda.com/posts/2026/i2c/I2C 是什么Wed, 13 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/jetson-40-pin/https://hellocuda.com/posts/2026/jetson-40-pin/jetson 40 pin上的缩写Wed, 13 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/jetson-read-pins/https://hellocuda.com/posts/2026/jetson-read-pins/Jetson 平台读取GPIO输入Tue, 12 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/opencv/https://hellocuda.com/posts/2026/opencv/OpenCV 是 一 个 开 源 的 计 算 机 视 觉 库Tue, 12 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/jetson-40pin/https://hellocuda.com/posts/2026/jetson-40pin/Jetson 入门Mon, 11 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/nms/https://hellocuda.com/posts/2026/nms/NMS (Non-Maximum Suppression) 分析Sat, 09 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/bytetrack-%E5%8D%A1%E5%B0%94%E6%9B%BC%E6%BB%A4%E6%B3%A2%E8%BE%B9%E7%95%8C%E6%A1%86%E8%AE%A1%E7%AE%97/https://hellocuda.com/posts/2026/bytetrack-%E5%8D%A1%E5%B0%94%E6%9B%BC%E6%BB%A4%E6%B3%A2%E8%BE%B9%E7%95%8C%E6%A1%86%E8%AE%A1%E7%AE%97/从卡尔曼状态向量计算边界框的实现解析Wed, 06 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/iou/https://hellocuda.com/posts/2026/iou/Wed, 06 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/lprnet/https://hellocuda.com/posts/2026/lprnet/LPRNet是一种端到端的轻量级车牌识别网络，无需字符分割和RNN，在CPU/GPU上均能实时运行。Wed, 06 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/usls/https://hellocuda.com/posts/2026/usls/usls 库参考指南Wed, 06 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89%E4%BA%94%E5%A4%A7%E4%BB%BB/https://hellocuda.com/posts/2026/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89%E4%BA%94%E5%A4%A7%E4%BB%BB/目标检测、实例分割、姿态估计、图像分类、有向边界框——五大计算机视觉核心任务的概念、输出及应用场景Tue, 05 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/ros2%E7%AC%94%E8%AE%B0/https://hellocuda.com/posts/2026/ros2%E7%AC%94%E8%AE%B0/ROS2学习笔记Fri, 01 May 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/ros2/https://hellocuda.com/posts/2026/ros2/在Ubuntu系统上运行第一个ROS2程序，了解ROS2的基本使用方法。Wed, 29 Apr 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/claudecode%E6%98%AF%E6%80%8E%E4%B9%88%E6%83%B3%E7%9A%84/https://hellocuda.com/posts/2026/claudecode%E6%98%AF%E6%80%8E%E4%B9%88%E6%83%B3%E7%9A%84/Claudecode 是怎么想的Wed, 01 Apr 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/inference/https://hellocuda.com/posts/2026/inference/Inference Engineering LearningThu, 26 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/%E7%9F%A9%E9%98%B5%E4%B9%98%E6%B3%95/https://hellocuda.com/posts/2026/%E7%9F%A9%E9%98%B5%E4%B9%98%E6%B3%95/矩阵乘法的图解说明Wed, 25 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/high-rank-trait-bound/https://hellocuda.com/posts/2026/high-rank-trait-bound/High-rank trait bound (HRTB) in RustTue, 24 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/agent-%E6%A6%82%E5%BF%B5%E6%80%A7%E7%9A%84%E8%AE%B0%E5%BD%95/https://hellocuda.com/posts/2026/agent-%E6%A6%82%E5%BF%B5%E6%80%A7%E7%9A%84%E8%AE%B0%E5%BD%95/Agent 概念性的记录Tue, 24 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/running-up-ironclaw/https://hellocuda.com/posts/2026/running-up-ironclaw/本地环境搭建IronclawMon, 23 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/rules-in-rust/https://hellocuda.com/posts/2026/rules-in-rust/Rust的规则：所有权、引用和生命周期Wed, 18 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/rust-overview/https://hellocuda.com/posts/2026/rust-overview/Rust 语言概述Wed, 18 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/cublas%E5%9F%BA%E7%A1%80/https://hellocuda.com/posts/2026/cublas%E5%9F%BA%E7%A1%80/全面介绍 cuBLAS 库的基础知识，包括其在 CUDA 编程中的作用、如何使用它进行矩阵运算，以及列主序与行主序的区别。Fri, 13 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/row-column-order/https://hellocuda.com/posts/2026/row-column-order/深入理解列主序（Column-Major Order）和行主序（Row-Major Order）的区别，以及它们在内存布局和性能优化中的影响。Fri, 13 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/agenticpattern/https://hellocuda.com/posts/2026/agenticpattern/Understanding AI Agent patterns and architecturesMon, 09 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/obb/https://hellocuda.com/posts/2026/obb/有向边界框的概念、格式及应用场景Mon, 09 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026-03-non-ai-generation-stance/https://hellocuda.com/posts/2026-03-non-ai-generation-stance/2026 年做一个不用 AI 生成图片的图片工具，看起来反潮流。但我认为这是唯一还没被抢占的 differentiation。Sun, 08 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/cuda-kernel%E7%9A%84%E6%9C%80%E4%BD%B3%E5%AE%9E%E8%B7%B5/https://hellocuda.com/posts/2026/cuda-kernel%E7%9A%84%E6%9C%80%E4%BD%B3%E5%AE%9E%E8%B7%B5/CUDA Kernel 开发的最佳实践指南Thu, 05 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/nvidia-smi%E5%91%BD%E4%BB%A4/https://hellocuda.com/posts/2026/nvidia-smi%E5%91%BD%E4%BB%A4/NVIDIA System Management Interface 常用命令指南Thu, 05 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/nvidia-gpu-%E6%9E%B6%E6%9E%84%E6%BC%94%E8%BF%9B/https://hellocuda.com/posts/2026/nvidia-gpu-%E6%9E%B6%E6%9E%84%E6%BC%94%E8%BF%9B/NVIDIA GPU 架构演进Tue, 03 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026/temperature%E5%8F%82%E6%95%B0%E5%9C%A8%E7%94%9F%E6%88%90%E5%BC%8Fai%E4%B8%AD%E4%BD%9C%E7%94%A8/https://hellocuda.com/posts/2026/temperature%E5%8F%82%E6%95%B0%E5%9C%A8%E7%94%9F%E6%88%90%E5%BC%8Fai%E4%B8%AD%E4%BD%9C%E7%94%A8/Temperature参数在生成式AI中作用Tue, 03 Mar 2026 00:00:00 GMThttps://hellocuda.com/posts/2026-01-02-pytorch20%E5%AD%A6%E4%B9%A0/https://hellocuda.com/posts/2026-01-02-pytorch20%E5%AD%A6%E4%B9%A0/PyTorch 2.0 学习笔记：张量维度、dim 参数、激活函数、损失函数和优化器Fri, 02 Jan 2026 00:00:00 GMThttps://hellocuda.com/posts/rerank-by-model/https://hellocuda.com/posts/rerank-by-model/dive into rerank_by_modelMon, 08 Dec 2025 00:00:00 GMThttps://hellocuda.com/posts/reset-mysql-password-docker/https://hellocuda.com/posts/reset-mysql-password-docker/当遗忘Docker中MySQL实例的密码时，如何安全有效地重置密码Tue, 25 Nov 2025 00:00:00 GMThttps://hellocuda.com/posts/2025-11-from-rust-to-canvas/https://hellocuda.com/posts/2025-11-from-rust-to-canvas/一个写了几年 Rust 系统代码的人，为什么开始押注一个浏览器端的图片审美工具。不是技术升级，是审美需求。Sat, 15 Nov 2025 00:00:00 GMThttps://hellocuda.com/posts/blas/https://hellocuda.com/posts/blas/Basic Linear Algebra SubprogramsThu, 24 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/similarity-vector/https://hellocuda.com/posts/similarity-vector/common similarity searchesWed, 23 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/gemm/https://hellocuda.com/posts/gemm/通用矩阵乘法（GEMM）详解Tue, 22 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/f4m3-f5m2/https://hellocuda.com/posts/f4m3-f5m2/FP8的两种格式（E4M3和E5M2）Mon, 21 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/dot-product-vllm/https://hellocuda.com/posts/dot-product-vllm/摘录vLLM中 dot product计算过程Mon, 21 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/fma/https://hellocuda.com/posts/fma/现代 CPU 和 GPU 中的一种基础算术指令Mon, 21 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/unrolled/https://hellocuda.com/posts/unrolled/pragma unroll的详解Mon, 21 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/dense-and-sparse/https://hellocuda.com/posts/dense-and-sparse/Dense and Sparse overviewSun, 20 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/dive-into-forward/https://hellocuda.com/posts/dive-into-forward/走读一下 forward 方法(Embedding)Sun, 20 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/display-in-rust/https://hellocuda.com/posts/display-in-rust/Display and Debug Traits in RustSat, 19 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/multi-head-attention/https://hellocuda.com/posts/multi-head-attention/Sat, 19 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/hyperparameter-tuning/https://hellocuda.com/posts/hyperparameter-tuning/Hyperparameter overviewFri, 18 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/overfitting-underfitting/https://hellocuda.com/posts/overfitting-underfitting/Overfitting and Underfitting in Machine LearningThu, 17 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/training-neural-network/https://hellocuda.com/posts/training-neural-network/A comprehensive guide on training neural networks using PyTorch, covering data loading, preprocessing, training loops, evaluation metrics, and model saving.Wed, 16 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/hello-world-cuda/https://hellocuda.com/posts/hello-world-cuda/A simple CUDA program that prints 'Hello, World!' from a GPU thread.Tue, 15 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/nn-module/https://hellocuda.com/posts/nn-module/An overview of the nn module in PyTorch, providing a high-level interface for building neural networks.Tue, 15 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/optimization-algo/https://hellocuda.com/posts/optimization-algo/An overview of optimization algorithms used in training neural networks.Tue, 15 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/loss-functions/https://hellocuda.com/posts/loss-functions/An overview of loss functions in neural networks and their importance in training models.Mon, 14 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/activation-functions/https://hellocuda.com/posts/activation-functions/An overview of activation functions in neural networksSun, 13 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/artificial-neurons/https://hellocuda.com/posts/artificial-neurons/An introduction to artificial neurons, their structure, and how they function in neural networks.Sat, 12 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/neural-networks/https://hellocuda.com/posts/neural-networks/An overview of neural networks, their architecture, and applications in AI.Fri, 11 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/2025-07-love-poetry/https://hellocuda.com/posts/2025-07-love-poetry/Exploring the essence of love through poetry, capturing its fleeting nature and enduring impact.Thu, 10 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/custom-autograd-functions/https://hellocuda.com/posts/custom-autograd-functions/Learn how to create custom autograd functions in PyTorch for complex operations.Thu, 10 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/bias-vector/https://hellocuda.com/posts/bias-vector/Understanding the role of bias vectors in multi-head attention mechanisms.Wed, 09 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/mean/https://hellocuda.com/posts/mean/Understanding mean and variance in deep learning and statistics.Wed, 09 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/gemini-embedding/https://hellocuda.com/posts/gemini-embedding/Explore Gemini's text embedding models for advanced NLP tasks.Tue, 08 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/gpu-memory-calculate/https://hellocuda.com/posts/gpu-memory-calculate/GPU Memory Requirement Calculator for AI ModelsTue, 08 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/multihead-attention/https://hellocuda.com/posts/multihead-attention/因果自注意力和高效多头注意力模块的PyTorch实现Tue, 08 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/a-complete-self-attention/https://hellocuda.com/posts/a-complete-self-attention/Single Head Self-Attention的完整实现Mon, 07 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/a-simple-self-attention-mechanism/https://hellocuda.com/posts/a-simple-self-attention-mechanism/一个简单的自注意力机制实现Mon, 07 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/fish-pytorch/https://hellocuda.com/posts/fish-pytorch/一个 PyTorch FishData 例子Mon, 07 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/scaled-dot-product/https://hellocuda.com/posts/scaled-dot-product/simple self attention -> trainable weight self AttentionMon, 07 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/google-machine-learning/https://hellocuda.com/posts/google-machine-learning/什么是机器学习？Sun, 06 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/pytorch-cheetsheet/https://hellocuda.com/posts/pytorch-cheetsheet/PyTorch常用操作速查表Sun, 06 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/broadcasting-in-pytorch/https://hellocuda.com/posts/broadcasting-in-pytorch/详解PyTorch Broadcasting（广播机制）Sat, 05 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/rank-ndim-in-pytorch/https://hellocuda.com/posts/rank-ndim-in-pytorch/详解PyTorch张量中rank（秩）与ndim（维度数）的关键区别Sat, 05 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/training-with-multiple-gpus/https://hellocuda.com/posts/training-with-multiple-gpus/Training with multiple GPUs - template code with pytorchFri, 04 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/ai-abbr/https://hellocuda.com/posts/ai-abbr/AI 词汇表 - 常见缩写与术语Thu, 03 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/parameters-in-model/https://hellocuda.com/posts/parameters-in-model/代码片段计算model的参数大小Thu, 03 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/greedy-search-llm/https://hellocuda.com/posts/greedy-search-llm/一个Greedy search decoding的例子Wed, 02 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/python-dictionary-comprehension/https://hellocuda.com/posts/python-dictionary-comprehension/Python中字典推导式的基本用法Wed, 02 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/python-list-comprehension/https://hellocuda.com/posts/python-list-comprehension/Python中列表推导式的基本用法Wed, 02 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/2025-07/https://hellocuda.com/posts/2025-07/Book reading records for July 2025Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/chat-with-llamacpp/https://hellocuda.com/posts/chat-with-llamacpp/A hands-on guide to creating an interactive chat interface using llama.cpp, covering model loading, prompt engineering, and response streaming.Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/llama-vocab-type/https://hellocuda.com/posts/llama-vocab-type/llama.cpp 分词器类型对比.Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/llamacpp-simple/https://hellocuda.com/posts/llamacpp-simple/A llama.cpp Helloworld.Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/setup-and-run-llamacpp/https://hellocuda.com/posts/setup-and-run-llamacpp/详细指南：从零开始配置 llama.cpp 开发环境，编译并运行 simple 示例程序。Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/sft-lora/https://hellocuda.com/posts/sft-lora/深入解析监督微调(SFT)与低秩适应(LoRA)的核心原理、应用场景及性能对比，帮助开发者选择最佳微调方案。Tue, 01 Jul 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-checkp2p/https://hellocuda.com/posts/cuda-checkp2p/检查CUDA设备之间的P2P通信能力，了解如何优化GPU间的数据传输。Wed, 18 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-nsys/https://hellocuda.com/posts/cuda-nsys/使用 NVIDIA nsys Profiler 分析 CUDA 程序性能，了解如何优化 GPU 计算效率。Wed, 18 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-nsys-copy/https://hellocuda.com/posts/cuda-nsys-copy/使用 NVIDIA nsys Profiler 分析 CUDA 程序性能，了解如何优化 GPU 计算效率。Wed, 18 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-thrust-basic/https://hellocuda.com/posts/cuda-thrust-basic/介绍CUDA Thrust库的基本用法，了解如何在GPU上进行高效的数据处理和算法实现。Wed, 18 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-architecture/https://hellocuda.com/posts/cuda-architecture/深入了解CUDA架构，探索GPU的硬件结构、计算单元及其在并行计算中的应用。Tue, 17 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-dynamic-parralle/https://hellocuda.com/posts/cuda-dynamic-parralle/深入探讨CUDA动态并行编程模型，了解如何在GPU上实现更灵活的并行计算。Tue, 17 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/one-hot-encoding/https://hellocuda.com/posts/one-hot-encoding/了解独热编码的概念、应用场景以及在机器学习中的重要性。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-overview-copy/https://hellocuda.com/posts/cuda-overview-copy/深入了解CUDA编程模型，探索GPU的并行计算能力及其在数据处理中的应用。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-overview/https://hellocuda.com/posts/cuda-overview/深入了解CUDA编程模型，探索GPU的并行计算能力及其在数据处理中的应用。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-parallel/https://hellocuda.com/posts/cuda-parallel/深入探讨CUDA并行编程模型，了解如何高效利用GPU进行大规模数据处理和计算任务。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-profiling/https://hellocuda.com/posts/cuda-profiling/深入探讨CUDA性能分析工具，了解如何优化GPU应用程序的性能和资源利用率。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/with-torch-no-grad/https://hellocuda.com/posts/with-torch-no-grad/本文深入探讨 PyTorch 中的 `with torch.no_grad()` 上下文管理器，解释其作用、使用场景及等价实现方式。Mon, 16 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/ternary-in-python/https://hellocuda.com/posts/ternary-in-python/了解 Python 中的三元条件表达式。Sun, 15 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/define-fun-in-ocaml/https://hellocuda.com/posts/define-fun-in-ocaml/Learn how to define functions in OCaml, including syntax, examples, and best practices.Sat, 14 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/fnn-cnn-rnn/https://hellocuda.com/posts/fnn-cnn-rnn/介绍前馈神经网络（FNN）、卷积神经网络（CNN）和循环神经网络（RNN）的基本概念、结构和应用场景。Thu, 12 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/pub-model-torchserver/https://hellocuda.com/posts/pub-model-torchserver/The comprehensive guide should get you started with deploying your FNN model using TorchServe.Thu, 12 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/setup-pytorch-cuda-in-conda/https://hellocuda.com/posts/setup-pytorch-cuda-in-conda/搭建 PyTorch CUDA 环境的详细步骤，包括创建 Conda 环境、安装 PyTorch 及其 CUDA 支持，并验证安装。Thu, 12 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/torch-basic-operations/https://hellocuda.com/posts/torch-basic-operations/介绍 PyTorch 的基本操作，包括张量的创建、形状变换、切片、连接、转置以及矩阵运算等。Thu, 12 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/cast-in-cxx/https://hellocuda.com/posts/cast-in-cxx/C++ 中的类型转换操作符总结，包括 `static_cast`、`dynamic_cast`、`const_cast`、`reinterpret_cast`、`bit_cast`、`duration_cast` 和 C 风格的类型转换。每种转换的用途、特点和示例代码。Wed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/install-conda-on-ubuntu/https://hellocuda.com/posts/install-conda-on-ubuntu/Install conda on UbuntuWed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/mathematics-in-machine-learning/https://hellocuda.com/posts/mathematics-in-machine-learning/Understanding the role of mathematics in machine learningWed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/maybe_unused-in-cxx/https://hellocuda.com/posts/maybe_unused-in-cxx/介绍C++17引入的[[maybe_unused]]属性，如何优雅地处理未使用的变量和函数。Wed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/rust-cheatsheet/https://hellocuda.com/posts/rust-cheatsheet/Rust is a systems programming language focused on safety, speed, and concurrency. It uses a unique ownership model to manage memory without a garbage collector.Wed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/six-special-function-in-cxx/https://hellocuda.com/posts/six-special-function-in-cxx/Wed, 11 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/add-cxx-header-path/https://hellocuda.com/posts/add-cxx-header-path/在 VSCode 中配置 C/C++ 项目的头文件搜索路径。Tue, 10 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/oust-gut-boots/https://hellocuda.com/posts/oust-gut-boots/分析不同媒体在报道RFK Jr.解雇疫苗专家小组时使用的动词差异及其隐含立场。Tue, 10 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/2025-06/https://hellocuda.com/posts/2025-06/Book reading records for June 2025Sun, 01 Jun 2025 00:00:00 GMThttps://hellocuda.com/posts/fine-tune-llm/https://hellocuda.com/posts/fine-tune-llm/本文详细介绍如何通过微调OpenAI模型创建符合企业专业术语和写作风格的邮件自动回复系统，包含完整代码实现和最佳实践。Tue, 27 May 2025 00:00:00 GMThttps://hellocuda.com/posts/nvidaarch/https://hellocuda.com/posts/nvidaarch/A comprehensive guide to NVIDIA's GPU architectures, exploring key innovations from Tesla to Ampere for developers and tech enthusiasts.Sun, 18 May 2025 00:00:00 GMThttps://hellocuda.com/posts/modern-cxx-summary/https://hellocuda.com/posts/modern-cxx-summary/Concise reference for key C++ features across modern standards, helping developers write efficient and maintainable code.Sun, 18 May 2025 00:00:00 GMThttps://hellocuda.com/posts/blanket-and-bound-trait/https://hellocuda.com/posts/blanket-and-bound-trait/A comprehensive guide to Rust's trait system, covering blanket implementations and trait bounds with practical examples. Explores advanced patterns, performance implications, and standard library internals.Thu, 15 May 2025 00:00:00 GMThttps://hellocuda.com/posts/cuda-memory/https://hellocuda.com/posts/cuda-memory/深入解析CUDA内存体系结构，涵盖寄存器、共享内存、全局内存等所有内存类型的工作原理与优化策略。包含Bank Conflict解决方案、内存访问模式优化技巧及实际性能对比分析。Thu, 15 May 2025 00:00:00 GMThttps://hellocuda.com/posts/vulkan-002-pipeline-basic/https://hellocuda.com/posts/vulkan-002-pipeline-basic/Detailed examination of each Vulkan graphics pipeline stage and their rendering responsibilitiesThu, 15 May 2025 00:00:00 GMThttps://hellocuda.com/posts/vulkan-001-overview/https://hellocuda.com/posts/vulkan-001-overview/From VkInstance to vkCmdDraw: The essential steps to render your first Vulkan triangleMon, 12 May 2025 00:00:00 GMThttps://hellocuda.com/posts/50-universal-technical-phrases-for-developers-set-2/https://hellocuda.com/posts/50-universal-technical-phrases-for-developers-set-2/Designed to help you articulate ideas clearly and professionally in technical discussions, interviews, or presentations.Sun, 04 May 2025 00:00:00 GMThttps://hellocuda.com/posts/partialeq-eq/https://hellocuda.com/posts/partialeq-eq/全面解析Rust中PartialEq与Eq特质的区别，从数学定义到实际应用场景，帮助开发者正确实现和使用这两种相等性关系。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/rust-must_use/https://hellocuda.com/posts/rust-must_use/Rust官方文档对#[must_use]属性的完整说明，包含类型、函数、trait等场景的应用规范。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/send--sync/https://hellocuda.com/posts/send--sync/全面解析Rust中Send与Sync特质的工作原理、实现机制及实际应用场景，涵盖自动实现规则、手动实现要点以及常见并发模式的最佳实践。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/send-sync-examples/https://hellocuda.com/posts/send-sync-examples/通过基础代码示例演示Send特质的所有权跨线程转移和Sync特质的共享不可变数据访问。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/agave-fee/https://hellocuda.com/posts/agave-fee/Solana 账户获得的奖励信息。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/trait-add-clone/https://hellocuda.com/posts/trait-add-clone/trait AppendVecScan: Send + Sync + CloneSat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/%E5%8D%8F%E5%8F%98-%E9%80%86%E5%8F%98-%E4%B8%8D%E5%8F%98/https://hellocuda.com/posts/%E5%8D%8F%E5%8F%98-%E9%80%86%E5%8F%98-%E4%B8%8D%E5%8F%98/剖析Rust类型系统中协变(&T)、逆变(fn(T))和不变(&mut T)的核心区别，掌握安全泛型编程的关键规则。Sat, 03 May 2025 00:00:00 GMThttps://hellocuda.com/posts/50-universal-technical-phrases-for-developers/https://hellocuda.com/posts/50-universal-technical-phrases-for-developers/Designed to help you articulate ideas clearly and professionally in technical discussions, interviews, or presentations.Fri, 02 May 2025 00:00:00 GMThttps://hellocuda.com/posts/rust%E7%BC%96%E8%AF%91%E6%97%B6%E4%B8%8E%E8%BF%90%E8%A1%8C%E6%97%B6%E4%BB%A3%E7%A0%81-%E6%A0%B8%E5%BF%83%E7%89%B9%E5%BE%81%E4%B8%8E%E5%88%A4%E6%96%AD/https://hellocuda.com/posts/rust%E7%BC%96%E8%AF%91%E6%97%B6%E4%B8%8E%E8%BF%90%E8%A1%8C%E6%97%B6%E4%BB%A3%E7%A0%81-%E6%A0%B8%E5%BF%83%E7%89%B9%E5%BE%81%E4%B8%8E%E5%88%A4%E6%96%AD/深度解析Rust语言中编译时(compile-time)与运行时(runtime)代码的执行机制与判断方法，包含实际案例和优化技巧。Fri, 02 May 2025 00:00:00 GMThttps://hellocuda.com/posts/aigc-hoopics/https://hellocuda.com/posts/aigc-hoopics/AIGC Hoopics is a comprehensive and general-purpose AI-generated content service. It is designed for scalability, flexibility, and efficiency, leveraging the power of Rust and Shell scripting to empower users in creating and managing AI-powered solutions.Thu, 01 May 2025 00:00:00 GMThttps://hellocuda.com/posts/hoopics-api-service/https://hellocuda.com/posts/hoopics-api-service/A performant admin backend for hoopics image sharing platform, rebuilt in Rust with Actix-Web and Diesel ORM. Features multi-database support and follows EggJS-inspired architecture.Thu, 01 May 2025 00:00:00 GMThttps://hellocuda.com/posts/sgx-attacker/https://hellocuda.com/posts/sgx-attacker/SGX Attacker is an experimental project designed to explore vulnerabilities in Intel's Software Guard Extensions (SGX). It uses a combination of C++, Shell scripting, and Makefile to simulate and analyze potential attack vectors.Thu, 01 May 2025 00:00:00 GMThttps://hellocuda.com/posts/the-time-machine/https://hellocuda.com/posts/the-time-machine/The Time Traveller (for so it will be convenient to speak of him) was expounding a recondite matter to us. His pale grey eyes shone and twinkled, and his usually pale face was flushed and animated...Wed, 15 Aug 2018 00:00:00 GMThttps://hellocuda.com/posts/2016-09-01/https://hellocuda.com/posts/2016-09-01/Daily English book reading project started September 1, 2016Wed, 01 Jun 2016 00:00:00 GMT