Home

fast

Q.U.I.C.K. Campaign

Submitted by nsrr on Tue, 2024-05-07 01:23
Short Name: 
campaign-quick

Operation Q.U.I.C.K.: Quickly Untying "Inconvenient" Chronological Knots

Adds a fast play campaign with a Point Zero themed quest line that sends the player to a series of dungeon-portions from the standard AoA campaign. Players will adventure through ~8 shortened zones, starting with teir 1 dungeons and progressing all the way through to the finale at High Peak.

The campaign is broken into 8 segments, with a reward before each. These bonuses largely offer rewards that could be gained from optional quests in AoA. You will also have a chance to shop after each segment, and this version of Point Zero has greatly expanded and improved shopping opportunities. There is also an NPC that can create randarts with one ego of your choice, for a price.

See the forum thread for more details: https://forums.te4.org/viewtopic.php?f=50&t=55067

Faster ToME4

Submitted by yutio888 on Sun, 2021-06-20 06:43
Short Name: 
faster
  1. English Version (Condensed)
    1. Faster ToME4 (v0.2.8) — Release Notes

Comprehensive performance & stability overhaul for **ToME 1.7.6** (Upstream: 0.0.1).

> **TL;DR**: Synchronous save pause reduced by **~68%** (459 ms → 145 ms); pure Lua FBO guard eliminates mid-frame rendering stalls; native memory leaks fixed. **100% pure Lua, zero external binaries, fully compatible with existing saves & achievements.**

      1. 1. Save Stutter Reductions
    1. Fast Snapshot Clone**: Skips memo lookups for primitive values; cuts cloning CPU time by **~33%** while preserving the exact object graph.
    2. Wait-Screen Refresh**: Pumps native wait-screen redraws during snapshot cloning; reduces save-time freezes from **171.4 ms to 70.5 ms (-58.9%)**.
    3. Fast Lossless Screenshots**: Direct GL pixel reads with Level-1 PNG compression; cuts screenshot capture time from **50.0 ms to 24.3 ms (-51.5%)**.
    4. Redundant Work Eliminated**: Omitted unused secondary party creation and offline character dumps, saving 1,700+ transient UID allocations and ~136 RNG calls per save.
      1. Rendering & Combat Smoothness
    1. Pure Lua FBO GC Guard**: Fixes engine defect where FBO finalizers corrupt active drawing targets mid-frame; defers cleanup to the next `use()` call, eliminating driver/LLVM stalls.
    2. Hit-Warning Rate Limiting**: Throttles directional arrows (`hit_warning`) to **1 per 500 ms**; prevents arrow volleys from flooding emitters and lagging the game.
    3. UI & Mask Batching**: Caches hotkey text rasterization (**-23.6%** display CPU); batches effect-mask geometry into single vertex calls.
      1. Engine & Upstream Bug Fixes
    1. Native Memory Leak Fixed**: Switched character export compressor to embedded `lzlib` with proper cleanup, eliminating **~12.8 MiB leak per 50 exports**.
    2. Log & Cache Hardening**: Fixed `CacheList` modulo bug (now utilizes all 5,000 slots); restored `truncate_printlog`; bounded chat text & bytecode caches.
    3. Particle Leaks Fixed**: Added finite lifetimes to `notice_enemy` and `dreamhammer` emitters that previously leaked indefinitely.
      1. 4. Optional Settings (`config.settings.faster_tome.*`)

All enabled by default. Toggles include: `fbo_gc_guard`, `hit_warning_interval_ms` (0 to disable), `snapshot_refresh`, `screenshot_png`, `save_clone`, `offline_chardump`, `unused_party_cleanup`, `export_gzip`.

---

  1. 中文版（精简版）
    1. Faster ToME4 (v0.2.8) — 版本更新说明

适配 **ToME 1.7.6** 的全面性能与稳定性大版本（对比上游基准：0.0.1）

> **核心摘要**：存盘主线程卡顿减少 **~68%**（459 ms → 145 ms），告别存盘假死；纯 Lua FBO 保护消除渲染卡顿与绘制错乱；根治多处内存泄漏。**全插件纯 Lua 实现，零 C 扩展依赖，100% 兼容原版存档、成就与规则。**

      1. 1. 存盘卡顿深度优化
    1. 快照浅克隆提速**：基础类型跳过哈希查表，对象图 100% 一致的前提下纯克隆 CPU 耗时减少 **~33%**。
    2. 存盘画面定期刷新**：同步快照期间定期刷新原生等待条与界面，最大停顿时间从 **171.4 ms 降至 70.5 ms（-58.9%）**，不再白屏假死。
    3. 极速无损截图**：底层像素直读 + 1 级无损 PNG 编码，截图耗时从 **50.0 ms 降至 24.3 ms（耗时减半）**。
    4. 冗余流程彻底剔除**：省去未使用的临时导出队伍克隆与离线角色卡序列化，单次存盘省去 1,700+ 临时 UID 与约 136 次随机数消耗。
      1. 2. 渲染与战斗流畅优化
    1. 纯 Lua FBO 回收保护**：修复引擎原生 FBO 析构重置绘制目标导致后续掉帧的缺陷，延后至下次 `use()` 统一安全释放，消除驱动重编译与长帧。
    2. 受击提示限速**：远程受击箭头默认 **500 ms 冷却节流**，杜绝多重箭雨或高频连击瞬间刷爆粒子导致的严重顿卡与视效干扰。
    3. 界面与特效合批**：快捷栏文字光栅化缓存（绘制 CPU **减少 23.6%**），地图效果遮罩改为顶点数组批量绘制。
      1. 3. 上游缺陷与内存泄漏修复
    1. 根治导出原生内存泄漏**：改用引擎内置 `lzlib` 并严格释放流状态，彻底消除每 50 次角色导出约 **12.8 MiB** 的原生堆内存常驻泄漏。
    2. 日志与缓存加固**：修正 `CacheList` 模运算差一错误（用满全部 5,000 槽位），恢复日志截断功能，加固聊天框与字节码缓存上限。
    3. 终结粒子常驻泄漏**：修复 `notice_enemy` 与 `dreamhammer` 粒子发射后在后台永久空转的缺陷。
      1. 4. 独立开关配置（`config.settings.faster_tome.*`）

全部优化默认开启。需要时可在配置中独立调整：`fbo_gc_guard`（FBO保护）、`hit_warning_interval_ms`（受击间隔，0为不限速）、`snapshot_refresh`（快照刷新）、`screenshot_png`（极速截图）、`save_clone`（克隆加速）、`offline_chardump`（跳过离线导出）、`unused_party_cleanup`（跳过临时队伍）、`export_gzip`（无泄漏压缩）。

Syndicate content