8bac3ebf49
Add immutable MITC4 drilling selection and direct director springs
2026-10-20 08:52:43 +09:00
dd54b0ae1e
Connect HHT field convergence to production acceptance and v2 logs
2026-10-19 02:49:53 +09:00
ad06d7ad6c
Implement private HHT field convergence policy and oracle tests
2026-10-19 02:20:58 +09:00
adb809bc01
엄격 수렴 기준 실행 보존과 한정 HHT 개발 계측 기록
2026-10-19 02:09:22 +09:00
d7f6d0ce74
chore: preserve approved MITC4 tolerance experiment baseline
2026-10-18 23:42:54 +09:00
945d899b2f
modify repository
2026-10-18 14:32:55 +09:00
0635cec53d
docs: align feature guidance with adopted Refactor structure
2026-10-18 03:35:15 +09:00
12d51480de
Integrate Refactor solver sources and dual MSVC CMake builds
2026-10-18 02:10:51 +09:00
9936994e49
Merge reviewed geometric packet optimization and FP fallback
2026-10-14 23:34:29 +09:00
0c29a9f8d8
fix(hex8): bridge nonfinite vector sums before error arithmetic
2026-10-14 23:14:21 +09:00
96772651d8
perf(hex8): packetize ordered gradient and geometric contractions
2026-10-14 22:37:39 +09:00
b11e12a146
Merge reviewed explicit constant-zero derivative seeds
2026-10-14 22:35:15 +09:00
727c3566f9
명시된 shell 영 seed의 도함수 provenance 보존
2026-10-14 22:33:12 +09:00
b8d1dff670
Merge reviewed vector pair guards and exact tail screen
2026-10-14 22:02:20 +09:00
0b5adf43e0
perf(hex8): vectorize packet guards and screen exact tail bounds
2026-10-14 21:52:30 +09:00
c7e3f67f26
Merge reviewed MKL shell products and complete-trial reuse
2026-10-14 21:49:14 +09:00
0e5b3e05dd
완성된 HHT trial의 constraint 자료 중복 평가 제거
2026-10-14 21:46:11 +09:00
446d3fa059
fix(hex8): isolate AVX2 packet code from baseline COMDATs
2026-10-14 21:10:27 +09:00
c397a147b0
perf(shell): skip proven zero derivative Hessian work
2026-10-14 20:14:18 +09:00
9b8e53dcfd
perf(hex8): prepare guarded private four-column paired tangent packet
2026-10-14 20:12:20 +09:00
8bfe97dbf8
perf(math): retain GEMM after scoped GER comparison
2026-10-14 19:46:35 +09:00
11af92809a
perf(math): use checked MKL GER for rank-one products
2026-10-14 19:44:29 +09:00
8021deefd3
perf(math): inline measured high-low arithmetic call boundaries
2026-10-14 19:27:46 +09:00
3576f3f6de
perf(math): scan least significand bit with MSVC x64 intrinsic
2026-10-14 18:58:00 +09:00
c58d499727
fix(hht): preserve metric association and complete MKL actions
2026-10-14 18:38:00 +09:00
45aac19c58
perf(hex8): integrate reviewed common finite classifier and build ignores
2026-10-14 18:30:06 +09:00
2853051896
Merge commit '85ae2eb' into dryforge/performance-t4
2026-10-14 18:30:04 +09:00
f9e0df66dc
perf(shell): route local products through MKL and reduce Hessian storage work
2026-10-14 18:07:16 +09:00
6bc5cd3932
style(hex8): make new standard dependencies explicit
2026-10-14 17:40:12 +09:00
47d2a81741
perf(math): classify finite high-low components by binary64 bits
2026-10-14 17:39:25 +09:00
6a6c8c33c8
perf(hex8): defer optional paired tangent until a valid linear trial
2026-10-14 17:38:27 +09:00
774bfbcbdd
perf(hex8): reuse geometric first products within each trial point
2026-10-14 17:19:25 +09:00
6d709fd572
perf(hex8): reuse identical paired Hessian subexpressions per trial
2026-10-14 17:14:05 +09:00
1d77e6be83
style(hex8): use repository path in private include guard
2026-10-14 17:14:03 +09:00
71e08169ec
fix(hex8): retain original contraction order in MKL entry products
2026-10-14 17:01:35 +09:00
eabaf6c893
perf(hex8): use MKL nodal mass and preserve rejected diagnostic priority
2026-10-14 16:56:11 +09:00
918eb50300
fix(hex8): retain tensor failure priority before MKL dispatch
2026-10-14 16:50:43 +09:00
e18271051c
perf(hex8): use MKL Hessian products and exact zero pair addition
2026-10-14 16:49:15 +09:00
dec6eb3199
perf(hex8): elide exact zero paired cross products
2026-10-14 16:48:20 +09:00
8c7c41729e
perf(assembly): use MKL entry products with ordered load diagnostics
2026-10-14 16:39:17 +09:00
9c4e6854ae
fix(math): preflight finite dense operands at dispatch boundary
2026-10-14 16:02:17 +09:00
770fa2a0d1
perf(solvers): scope default one-equation calls to one thread
2026-10-14 15:56:03 +09:00
4c4d943941
fix(eigen): validate basis operands before dense dispatch
2026-10-14 15:49:27 +09:00
590a93793f
fix(math): preserve certified sums and complete MKL algebra
2026-10-14 15:40:52 +09:00
59c8e93b07
perf(math): route ordinary products through MKL and preserve arithmetic
2026-10-14 15:09:53 +09:00
bf512d137c
feat(performance): add single-run timing and optimized symbol builds
2026-10-14 12:31:08 +09:00
eaa5cbeb21
feat: adopt lumped mass and reduced-coordinate shell HHT
...
Integrate Hex8 mass verification and fixed-null-basis shell dynamics with numerical and physics handoff evidence. Preserve interrupted reference and full-validation status; full integration validation is deferred until after performance optimization.
2026-10-13 16:57:19 +09:00
18b36c1c22
feat(hex8): use reference-weighted mean-volume energy
2026-10-10 23:34:14 +09:00
e2c3c5bc55
feat(hex8): connect Solid CLI and atomic HDF5 output
2026-10-10 01:26:16 +09:00
338902974c
feat(hex8): integrate paired HHT dynamic state and bounded verification
2026-10-10 00:36:16 +09:00
KOKO\Mimi