暂时存档(描述内容为更新 2.0,待我有时间制作时适用)
Lumen Terrain
Lumen Terrain (LTE) 是一款面向 Fabric 的服务端性能模组,旨在通过更有效地利用可用硬件资源来加速 Minecraft 地形生成。LTE 不单纯依赖标准生成路径,而是智能地将地形处理任务分配到可用的多核 CPU 和 OpenCL 兼容 GPU 上。
⚡ 功能特性
- 智能硬件调度: 基于实时基准测试和工作负载大小,动态选择最佳的处理后端(CPU 或 GPU)。
- 异构计算: 在有益的情况下,同时在多个 CPU 和 GPU 上并行处理工作负载。
- 故障转移安全机制: 当硬件或原生库发生故障时,可安全回退至 CPU 或原版生成方式,防止服务器崩溃。
- 即插即用原生库: 捆绑的原生代码会根据您的操作系统和架构自动提取并加载。
? 智能硬件调度
LTE 不会盲目地将所有任务发送给 GPU。它会基准测试可用硬件,评估工作负载大小、队列压力和处理开销,从而智能地调度任务。如果 CPU 对特定地形工作负载更快,LTE 会将其选为最优路径。
? 多核 CPU 处理
LTE 跨多个 CPU 工作线程并发处理独立的地形工作负载,利用硬件感知调度技术最大化多核效率,同时避免系统过载。
? 多 GPU 处理
引擎设计用于枚举多个支持 OpenCL 的 GPU 并并发处理独立的地形工作负载。调度器考虑设备健康状况、测量出的性能以及当前队列压力,以在异构设备之间平衡负载。
⚡ CPU + GPU 协同处理
LTE 能够结合 CPU 和 GPU 处理来提高整体吞吐量。通过对可用处理路径进行基准测试,LTE 会将任务无缝委派给对当前工作负载最有效的硬件。
? 硬件基准测试
LTE 检测 CPU 拓扑结构和 GPU 能力,缓存硬件配置档案和基准测试结果。通过测量实际性能而非依赖假设指标,LTE 确保始终选择最高效的处理路径。
?️ 可靠性与回退机制
加速失败不会导致服务器崩溃。如果 GPU 故障或原生处理不可用,LTE 会将该设备标记为不健康,并回退到更安全的处理路径——首先尝试使用 CPU 工作线程,必要时最终回退到标准的 Minecraft 原版地形生成。
? 跨平台支持
LTE 采用专为现代服务器环境设计的平台感知架构。原生库已构建并支持以下平台:
- Windows x86-64
- Windows ARM64
- Linux x86-64
- Linux ARM64
- macOS ARM64
? 自包含原生库
兼容的 .dll、.so 和 .dylib 文件直接捆绑在模组 JAR 包内。LTE 会在运行时检测您的平台和架构,自动提取并加载正确的原生库,无需用户进行任何手动安装。
? 轻量级设计
LTE 通过低空闲 CPU 占用率、有界队列和缓存、高效的原生资源管理以及减少不必要的内存分配来最小化其体积。硬件配置档案会被缓存,不必要的日志记录被限制在最低限度,以保持低开销。
?️ Fabric 服务端集成
LTE 直接集成到 Fabric 服务端的地形生成管线中。它拦截标准的 Minecraft 地形数据,并通过专用管线进行处理:16×16 高度图 → 坡度图 → 地形掩码 → 高度修改 → 处理后地形。CPU 和 GPU 实现均产生等效且可预测的地形结果。
? 安装
将 LTE .jar 文件放入你的 Fabric 服务端的 mods 文件夹中。无需外部依赖项或手动安装原生库。
? 配置
通过以下配置选项自定义 LTE 以适应你的服务器硬件:
- 自动或显式后端选择
- 每周期区块数和处理速率限制
- 工作线程配置
- 用于诊断的详细日志记录
? 链接
GitHub:
联系方式:
astryon@proton.me
? 许可证
MIT License
ARCHIVED FOR NOW (DESCRIPTION IS FOR UPDATE 2.0 WHEN I GET TIME TO MAKE IT)
Lumen Terrain
Lumen Terrain (LTE) is a Fabric server-side performance mod designed to accelerate Minecraft terrain generation by making better use of available hardware resources. Instead of relying solely on standard generation paths, LTE intelligently distributes terrain processing across available multi-core CPUs and OpenCL-capable GPUs.
⚡ Features
- Intelligent Hardware Scheduling: Dynamically selects the best processing backend (CPU or GPU) based on live benchmarking and workload size.
- Heterogeneous Compute: Processes workloads simultaneously across multiple CPUs and GPUs when beneficial.
- Failover Safety: Hardware or native library failures safely fall back to CPU or vanilla generation to prevent server crashes.
- Plug-and-Play Native Libraries: Bundled native code automatically extracts and loads based on your operating system and architecture.
? Intelligent Hardware Scheduling
LTE does not blindly send every task to the GPU. It benchmarks available hardware and assesses workload size, queue pressure, and processing overhead to schedule tasks intelligently. If the CPU is faster for a specific terrain workload, LTE will select it as the optimal path.
? Multi-Core CPU Processing
LTE processes independent terrain workloads concurrently across multiple CPU workers, utilizing hardware-aware scheduling to maximize multi-core efficiency without overloading the system.
? Multi-GPU Processing
The engine is designed to enumerate multiple OpenCL-capable GPUs and process independent terrain workloads concurrently. The scheduler considers device health, measured performance, and current queue pressure to balance the load across heterogeneous devices.
⚡ CPU + GPU Processing
LTE is capable of combining CPU and GPU processing to improve overall throughput. By benchmarking available processing paths, LTE seamlessly delegates tasks to whichever hardware is most efficient for the current workload.
? Hardware Benchmarking
LTE detects CPU topology and GPU capabilities, caching hardware profiles and benchmark results. By measuring actual performance rather than relying on assumed metrics, LTE ensures the most efficient processing path is always chosen.
?️ Reliability & Fallbacks
Acceleration failures will not crash your server. If a GPU fails or native processing is unavailable, LTE marks the device as unhealthy and falls back to a safer processing path—first attempting CPU workers, and ultimately falling back to standard vanilla Minecraft terrain generation if necessary.
? Cross-Platform Support
LTE features a platform-aware architecture built for modern server environments. Native libraries are built and supported for:
- Windows x86-64
- Windows ARM64
- Linux x86-64
- Linux ARM64
- macOS ARM64
? Self-Contained Native Libraries
Compatible .dll, .so, and .dylib files are bundled directly within the mod JAR. LTE detects your platform and architecture at runtime, automatically extracting and loading the correct native libraries without requiring any manual installation by the user.
? Lightweight Design
LTE minimizes its footprint with low idle CPU usage, bounded queues and caches, efficient native resource management, and reduced unnecessary memory allocations. Hardware profiles are cached, and unnecessary logging is kept to a minimum to maintain low overhead.
?️ Fabric Server Integration
LTE integrates directly into the Fabric server terrain-generation pipeline. It intercepts standard Minecraft terrain data and processes it through a dedicated pipeline: 16×16 heightmap → slope map → terrain mask → height modification → processed terrain. Both CPU and GPU implementations produce equivalent, predictable terrain results.
? Installation
Drop the LTE .jar file into your Fabric server's mods folder. No external dependencies or manual native library installations are required.
? Configuration
Customize LTE to fit your server hardware through configuration options including:
- Automatic or explicit backend selection
- Chunks-per-cycle and processing rate limits
- Worker configuration
- Verbose logging for diagnostics
? Links
GitHub:
Contact:
astryon@proton.me
? License
MIT License
评论 0
登录后参与评分
还没有评论,来抢沙发吧