64-BitBrainstorm_2026/AMSS-NCKU
8
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
136 Commits 49 Branches 0 Tags
cjy-leonhardt
Commit Graph

136 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
CGH0S7
6fd7ef2b55 Cache GPU RHS symbols and zero vacuum sources once 2026-04-12 22:42:58 +08:00
CGH0S7
7064ebd5b4 Batch GPU stage downloads 2026-04-12 21:06:41 +08:00
CGH0S7
87c581ea7c Checkpoint stable GPU optimization baseline 2026-04-12 20:26:27 +08:00
CGH0S7
d702aa06b9 Trim GPU restrict sync overhead 2026-04-12 19:45:34 +08:00
CGH0S7
ce88c18265 Tune GPU RHS launch geometry 2026-04-12 18:59:59 +08:00
CGH0S7
db2d6978b2 Reduce final GPU host downloads 2026-04-12 18:46:42 +08:00
CGH0S7
c8977d8356 Optimize GPU RK4 stage sync path 2026-04-12 18:36:05 +08:00
CGH0S7
d9287ea530 Fix GPU RK4 boundary and sync correctness 2026-04-12 12:13:47 +08:00
CGH0S7
b78874ef21 Refine stable GPU AMR staging path 2026-04-10 23:37:36 +08:00
CGH0S7
a089041c3b Stabilize GPU AMR prolong/restrict paths 2026-04-10 21:57:58 +08:00
CGH0S7
c578a15ecd Fix GPU interpolation cache lifetime leaks 2026-04-10 10:29:04 +08:00
CGH0S7
e1a0bff43c Reduce redundant GPU host buffer preparation 2026-04-09 21:20:45 +08:00
CGH0S7
cf3c6d6218 Stabilize GPU buffer lifecycle around regrid 2026-04-09 20:48:06 +08:00
CGH0S7
46e94d1248 Trim constraint-only GPU downloads 2026-04-09 19:36:19 +08:00
CGH0S7
7cd2414faa Move constraint recomputation onto GPU path 2026-04-09 19:17:39 +08:00
CGH0S7
4463f1d23e Unpack intermediate sync stages directly to GPU 2026-04-09 19:01:12 +08:00
CGH0S7
4484635f0d Pack sync send buffers directly from GPU state 2026-04-09 18:49:11 +08:00
CGH0S7
b0dd069a2b Register GPU transfer buffers as pinned host memory 2026-04-09 18:36:10 +08:00
CGH0S7
5bc67ded06 Download staged GPU sync regions incrementally 2026-04-09 18:23:05 +08:00
CGH0S7
3b16795e78 Refresh synced GPU regions incrementally 2026-04-09 17:07:31 +08:00
CGH0S7
5b00d49070 Reduce staged GPU host-device copies 2026-04-09 16:44:08 +08:00
CGH0S7
42e851d19a Cache repeated interpolation plans 2026-04-09 15:21:01 +08:00
CGH0S7
06fa643365 Refine batched CUDA interpolation kernel 2026-04-09 15:06:11 +08:00
CGH0S7
c47349b7a9 Add batched CUDA patch interpolation path 2026-04-09 14:56:01 +08:00
CGH0S7
ad999e4c5a Add guarded GPU prolong3 path scaffold 2026-04-09 14:28:36 +08:00
CGH0S7
e1e3b4a448 Reduce GPU RK4 transfer overhead 2026-04-09 12:11:40 +08:00
CGH0S7
49409645c0 Stabilize GPU output path and MPI sync 2026-04-09 10:57:49 +08:00
CGH0S7
4e3946a4f0 Persist GPU RK4 stage caches 2026-04-08 20:59:15 +08:00
CGH0S7
a0af9b8804 Trim GPU main-path transfer overhead 2026-04-08 20:16:25 +08:00
CGH0S7
01ac1f9250 Cache GPU main-path device buffers 2026-04-08 19:43:17 +08:00
CGH0S7
ea470737db Add runnable GPU main-path prototype 2026-04-08 19:14:37 +08:00
CGH0S7
8c1f4d8108 迁移C算子的循环融合和临时量消除 2026-03-03 16:20:15 +08:00
CGH0S7
d310ef918b bssn_rhs(fortran): migrate C kernel loop-fusion optimizations 2026-03-03 16:20:15 +08:00
CGH0S7
b35e1b289f 设置开关关闭内存打印统计 2026-03-03 16:17:47 +08:00
CGH0S7
05851b2c59 关闭静态负载 2026-03-03 16:17:47 +08:00
ianchb
3b39583d67 fix(bssn_rhs) 2026-03-03 16:06:33 +08:00
gh0s7
688bdb6708 Merge pull request 'cjy-dystopia' (#3) from cjy-dystopia into main 
Reviewed-on: https://seele.tail3b303.ts.net:3000/64-BitBrainstorm_2026/AMSS-NCKU/pulls/3
 2026-03-02 21:36:26 +08:00
CGH0S7
5070134857 perf(transfer_cached): 将 per-call new/delete 的 req_node/req_is_recv/completed 数组移入 SyncCache 复用 
避免 transfer_cached 每次调用分配释放 3 个临时数组，减少堆操作开销。

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-02 21:14:35 +08:00
CGH0S7
4012e9d068 perf(RestrictProlong): 用 Restrict_cached/OutBdLow2Hi_cached 替换非缓存版本，Sync_finish 改为渐进式解包 
- RestrictProlong/RestrictProlong_aux 中的 Restrict() 和 OutBdLow2Hi() 替换为 _cached 版本，
  复用 gridseg 列表和 MPI 缓冲区，避免每次调用重新分配
- 新增 sync_cache_restrict/sync_cache_outbd 两组 per-level 缓存
- Sync_finish 从 MPI_Waitall 改为 MPI_Waitsome 渐进式解包，降低尾延迟
- AsyncSyncState 扩展 req_node/req_is_recv/pending_recv 字段支持渐进解包

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-02 20:48:38 +08:00
ianchb
b3c367f15b prolong3 改为先算实际 stencil 窗口；只有窗口触及对称边界时才走全域 symmetry_bd，否则只复制必需窗口。restrict3 同样改成窗口判定，无触边时仅填 ii/jj/kk 必需窗口。 2026-03-02 17:38:56 +08:00
ianchb
e73911f292 perf(restrict3): shrink X-pass ii sweep to required overlap window 
- compute fi_min/fi_max from output i-range and derive ii_lo/ii_hi
 - replace full ii sweep (-1:extf(1)) with windowed sweep in Z/Y precompute passes
 - keep stencil math unchanged; add bounds sanity check for ii window
 2026-03-02 17:37:13 +08:00
ianchb
7543d3e8c7 perf(MPatch): 用空间 bin 索引加速 Interp_Points 的 block 归属查找 
- 为 Patch::Interp_Points 三个重载引入 BlockBinIndex（候选筛选 + 全扫回退）
  - 保持原 point-in-block 判定与后续插值/通信流程不变
  - 将逐点线性扫块从 O(N_points*N_blocks) 降为近似 O(N_points*k)
  - 测试：bin 上限如果太大，会引入不必要的索引构建开销。将 bins 上限设为 16。

Co-authored-by: gpt-5.3-codex
 2026-03-02 17:37:13 +08:00
ianchb
42c69fab24 refactor(Parallel): streamline MPI communication by consolidating request handling and memory management 2026-03-02 17:37:13 +08:00
CGH0S7
95220a05c8 optimize fdderivs core-region branch elimination for ghost_width=3 2026-03-02 17:33:26 +08:00
CGH0S7
466b084a58 fix prolong/restrict index bounds after cherry-pick 12e1f63 2026-03-02 13:59:47 +08:00
jaunatisblue
61ccef9f97 prolong3: 减少Z-pass 冗余计算 2026-03-02 13:58:52 +08:00
CGH0S7
e11363e06e Optimize fdderivs: skip redundant 2nd-order work in 4th-order overlap 2026-03-02 03:21:21 +08:00
jaunatisblue
f70e90f694 prolong3：提升cache命中率 2026-03-02 03:05:35 +08:00
jaunatisblue
75dd5353b0 修改prolong 2026-03-02 02:25:25 +08:00
jaunatisblue
23a82d063b 对prolong3做访存优化 2026-03-02 02:25:25 +08:00