64-BitBrainstorm_2026/AMSS-NCKU

cjy-dystopia #2

Merged

gh0s7 merged 26 commits from cjy-dystopia into main

2026-03-01 19:22:09 +08:00

gh0s7 commented

2026-03-01 19:21:59 +08:00

Owner

Make a fortune quietly :D

Make a fortune quietly :D

gh0s7 added 26 commits 2026-03-01 19:21:59 +08:00

bssn_rhs_c: 融合逆度规+Gamma约束+Christoffel三循环为一循环 (Modify 2) 2c60533501

逆度规计算结果保留在局部标量中直接复用，减少对gupxx..gupzz数组的重复读取，每步加速0.01秒

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

bssn_rhs_c: 融合A^{ij}升指标+Gamma_rhs_part1两循环为一循环 (Modify 3) c2b676abf2

A^{ij}六分量保留在局部标量中直接复用于Gamma_rhs计算，减少Rxx..Ryz数组的额外读取

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

bssn_rhs_c: 融合fxx/Gamxa+Gamma_rhs_part2两循环为一循环 (Modify 4) 808387aa11

fxx/fxy/fxz和Gamxa/ya/za保留在局部标量中直接复用于Gamma_rhs part2，减少数组读写

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

bssn_rhs_c: 融合Christoffel修正+trK_rhs两循环为一循环 (Modify 5) 38e691f013

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

kodiss_c: 收紧循环范围消除边界无用迭代和分支判断 8010ad27ed

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

fdderivs_c: 全量清零改为只清零边界面，减少无效内存写入 34d6922a66

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

bssn_rhs_c: 融合最终RHS两循环为一循环，用局部变量传递fij中间值 (Modify 6) 51dead090e

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

bssn_rhs_c: merge lopsided+kodis with shared symmetry buffer bc457d981e

derivs_c: split low/high stencil regions to reduce branch overhead 6024708a48

bssn_rhs_c: remove gxx/gyy/gzz temporaries in favor of dxx/dyy/dzz+1 1fbc213888

bssn_rhs_c: compute div_beta on-the-fly to remove temp array 169986cde1

bssn_rhs_c: recompute contracted Gamma terms to remove temp arrays 6493101ca0

build: switch allocator option to oneTBB tbbmalloc e29ca2dca9

Add switchable C RK4 kernel and build toggle b91cfff301

先关闭绑核心，发现速度对比：不绑定核心+SCX>绑核心+SCX 1eba73acbe

Optimize prolong3: precompute coarse index/parity maps 2c0a3055d4

Optimize prolong3: replace parity branches with coefficient lookup d4ec69c98a

Optimize prolong3: hoist bounds check out of inner loop 3046a0ccde

Optimize average2: replace array expression with explicit loops 1dc622e516

Optimize average2: use DO CONCURRENT loop form 1a518cd3f6

perf(polint): add ordn=6 specialized neville path a766e49ff0

perf(polin3): switch to lagrange-weight tensor contraction e5231849ee

perf(polint): add switchable barycentric ordn=6 path cca3c16c2b

perf(polint): add uniform-grid fast path for barycentric n=6 03416a7b28

Optimize symmetry_bd with stride-based fast paths 16013081e0

预赛最终版本v1.0.0: 确定PGO和原负载均衡方案在当前版本造成负优化已经回退 44efb2e08c

gh0s7 merged commit 524d1d1512 into main

2026-03-01 19:22:09 +08:00

gh0s7 referenced this issue from a commit

2026-03-01 19:22:10 +08:00

Merge pull request 'cjy-dystopia' (#2) from cjy-dystopia into main

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: 64-BitBrainstorm_2026/AMSS-NCKU#2