rustytorch

rtx-cfd legacy GPU modules: the launches that already sat in unsafe blocks are not wrapped twice

Documentation #738:Commit f429bdd415 pushed by osobh

main

2026-09-16 12:21:10 +00:00

6s

rtx-cfd legacy GPU modules: the launches that already sat in unsafe blocks are not wrapped twice

Performance Benchmarks #736:Commit f429bdd415 pushed by osobh

main

2026-09-16 12:21:03 +00:00

6s

rtx-cfd: rustfmt over the repaired GPU modules and tests

Documentation #735:Commit 7f4d437199 pushed by osobh

main

2026-09-16 12:20:10 +00:00

6s

rtx-cfd legacy GPU tests: one base struct per literal

Documentation #732:Commit 3dadf1ab5c pushed by osobh

main

2026-09-16 12:19:47 +00:00

8s

rtx-cfd legacy GPU tests: one base struct per literal

Performance Benchmarks #730:Commit 3dadf1ab5c pushed by osobh

main

2026-09-16 12:19:38 +00:00

7s

rtx-cfd legacy GPU tests: struct-update commas and the Poisson in-out buffer

Documentation #729:Commit 147e0c0422 pushed by osobh

main

2026-09-16 12:19:41 +00:00

17s

rtx-cfd legacy GPU tests: the kernels' output buffers are &mut

Documentation #726:Commit 534801e16a pushed by osobh

main

2026-09-16 12:19:18 +00:00

12s

rtx-cfd legacy GPU tests: the current CudaKernelManager API (Arc-held manager, allocate-on-copy, copy_from_device returning the host vector), MatrixOpsKernel::tridiagonal_matvec added over the existing kernel, PoissonKernel::solve_jacobi_2d kept as the…

Documentation #723:Commit 1381b5b633 pushed by osobh

main

2026-09-16 12:18:45 +00:00

7s

rtx-cfd k-ε GPU: memcpy_dtoh argument order

Documentation #720:Commit af19860c04 pushed by osobh

main

2026-09-16 12:17:16 +00:00

8s

rtx-cfd k-ε GPU: memcpy_dtoh argument order

Performance Benchmarks #718:Commit af19860c04 pushed by osobh

main

2026-09-16 12:17:04 +00:00

6s

rtx-cfd legacy GPU modules: cudarc 0.18 API — unsafe launches, the stream accessor, memcpy_htod argument order; the renamed k-ε constants and ::standard(); get_module clones; CfdConfig test literals take ..Default

Documentation #717:Commit 5e402918bb pushed by osobh

main

2026-09-16 12:17:03 +00:00

17s

rtx-cfd: the legacy GPU modules back in the cuda build (repair in progress)

Documentation #714:Commit de435eb91c pushed by osobh

main

2026-09-16 12:15:58 +00:00

23s

rtx-fsi: the interface noise probe reads RTX_FSI2_MG_DEVICE (the device V-cycle's regime gate)

Documentation #711:Commit 3fd4fb7b8c pushed by osobh

main

2026-09-16 06:25:00 +00:00

39s

PERF-2 P3-ii: the device V-cycle as the CG's preconditioner — poisson/device.rs (one CUDA runtime per process, persistent per-operator buffers, the mg_vcycle.cu kernels at K = 1; upload r, run the V-cycle, download z; the f64 CG unchanged), MultigridP…

Documentation #708:Commit fc556f8a88 pushed by osobh

main

2026-09-16 06:17:12 +00:00

10s

PERF-2 P3 bench: the coarsest level uses the CPU's red-black symmetric ordering (a nearly singular coarsest system settles on a smoother-dependent constant, which the prolongation doubled up every level: the 19 % disagreement)

Documentation #705:Commit 3b7f8fb362 pushed by osobh

main

2026-09-16 06:10:51 +00:00

31s

PERF-2 P3 bench: the coarsest level uses the CPU's red-black symmetric ordering (a nearly singular coarsest system settles on a smoother-dependent constant, which the prolongation doubled up every level: the 19 % disagreement)

Performance Benchmarks #703:Commit 3b7f8fb362 pushed by osobh

main

2026-09-16 06:10:29 +00:00

13s

PERF-2 P3 bench: per-level comparison of the device V-cycle's work vectors against the CPU's (vcycle_f32_work) to locate the disagreement

Documentation #702:Commit 06a60b5fed pushed by osobh

main

2026-09-16 06:08:49 +00:00

12s

rtx-cfd: the legacy GPU solver modules (piso_gpu, simple_gpu, d2q9_gpu, d3q19_gpu, smagorinsky_gpu) kept out of the cuda build — broken since the initial commit; the PERF-2 benchmark uses cudarc directly

Documentation #699:Commit cb1647c0ef pushed by osobh

main

2026-09-16 06:05:51 +00:00

8s

rtx-cfd: the cuda feature builds again — the duplicate CudaKernelManager::stream removed, the legacy k-epsilon GPU model (broken since the initial commit) kept out of the build

Documentation #696:Commit 956e9d38e2 pushed by osobh

main

2026-09-16 06:03:57 +00:00

13s

rtx-cfd: the cuda feature builds again — the duplicate CudaKernelManager::stream removed, the legacy k-epsilon GPU model (broken since the initial commit) kept out of the build

Performance Benchmarks #694:Commit 956e9d38e2 pushed by osobh

main

2026-09-16 06:03:44 +00:00

7s

PERF-2 P3: the batched device V-cycle microbenchmark — mg_vcycle.cu (masked variable-coefficient five-point red-black half-sweeps, residual, CSR restriction in a fixed order, prolongation, coarsest sweeps; [K][n] layout, blockIdx.y = march), LevelExpo…

Documentation #693:Commit 0dc95a8de5 pushed by osobh

main

2026-09-16 06:03:16 +00:00

8s

PERF-2 P3: the batched device V-cycle microbenchmark — mg_vcycle.cu (masked variable-coefficient five-point red-black half-sweeps, residual, CSR restriction in a fixed order, prolongation, coarsest sweeps; [K][n] layout, blockIdx.y = march), LevelExpo…

Performance Benchmarks #691:Commit 0dc95a8de5 pushed by osobh

main

2026-09-16 06:03:14 +00:00

7s

rtx-cfd: get_features compiles under --features cuda (the vector must be mutable)

Documentation #690:Commit 53078c2aa7 pushed by osobh

main

2026-09-16 06:00:45 +00:00

6s

PERF-2 P2-c: RTX_FSI2O_INNER_STOP knob — the background Poisson's inner stop as a multiple of the Schwarz tolerance (default 0.1; the band-gated ladder on the CG iteration count)

Documentation #687:Commit 8c130c55c8 pushed by osobh

main

2026-09-16 05:51:16 +00:00

19s

PERF-2 P2-c: RTX_FSI2O_INNER_STOP knob — the background Poisson's inner stop as a multiple of the Schwarz tolerance (default 0.1; the band-gated ladder on the CG iteration count)

Performance Benchmarks #685:Commit 8c130c55c8 pushed by osobh

main

2026-09-16 05:51:03 +00:00

7s

PERF-2 P2: the parallel maps run only on levels with ≥ 8,192 active cells and in tasks of ≥ 2,048 cells (a fork-join over a small level costs more than the map); values unchanged

Documentation #684:Commit 822041fb3f pushed by osobh

main

2026-09-16 05:12:08 +00:00

1m18s

PERF-2 P2: the parallel maps run only on levels with ≥ 8,192 active cells and in tasks of ≥ 2,048 cells (a fork-join over a small level costs more than the map); values unchanged

Performance Benchmarks #682:Commit 822041fb3f pushed by osobh

main

2026-09-16 05:10:55 +00:00

7s

PERF-2 P2: the red-black colour maps, the residual and the V-cycle's per-level maps on rayon threads (MultigridParameters::threads; RTX_THREADS in the harness; rtx_cfd::configure_threads) — each colour's values computed from the unchanged other colour…

Documentation #681:Commit 7f144e0005 pushed by osobh

main

2026-09-16 05:00:59 +00:00

1m22s

PERF-2 P2: the red-black colour maps, the residual and the V-cycle's per-level maps on rayon threads (MultigridParameters::threads; RTX_THREADS in the harness; rtx_cfd::configure_threads) — each colour's values computed from the unchanged other colour…

Performance Benchmarks #679:Commit 7f144e0005 pushed by osobh

main

2026-09-16 04:59:43 +00:00

8s

rtx-fsi: the interface noise probe reads RTX_FSI2_MG_RB itself (its solver is built outside the march's knob site; the first red-black probe run was vacuous)

Documentation #678:Commit 364766a54a pushed by osobh

main

2026-09-16 04:55:40 +00:00

1m37s