Mixed cells plan: AhmedML times measured with the perf branches; point gradients split out (#389's review)
Mixed cells plan: questions 1 and 2 answered, polyhedra= for faces=, which faces are refused (#389's review)
Mixed cells plan: polygons read beside triangles and quads (#389)
CI test management plan: done (#390, #391, #392, #394, #395, #397, #398)
CI test management plan: steps 5 and 6 in review (#395, #397, #398)
CI test management plan: steps 1 and 3 done (#390); 2 and 4 in review (#391, #392)
Array views (#335, in review as #393)
CI test management plan: step 2 runs 3.13 in full, with VTK, and the quick tier without it
Mixed cells plan: in review as #389, the performance work on two stacked branches
CI test management plan (#369): path filter, one Python version in full, timing tests, flakes, selection, merge queue
Device memory plan: host access uses the spilled copy; only the device restores; the host copy is not handed out
Device memory plan (#387): accounting, and spilling under a budget or on a failed allocation
Compile time plan: in review as #384; --Ofast-compile not pursued
Compile time plan: CUDA measured (Tack's share halved; NVRTC's cache, CUBIN, Ofast-compile)
Compile time plan: what the disk cache leaves, and first calls with it warm
Compile time plan: MSL generation (0.86 to 0.54 s); status 10.1 s CPU, 9.2 s Metal
Compile time plan: Metal measured; status at 10.3 s (CPU), 9.5 s (Metal)
Compile time plan: step 4 (branches constants decide) done
Compile time plan: copy propagation and part of lowering done; LLVM's fixed cost measured
Compile time plan: step 3 (verify less) done
Compile time plan (#322): measurements over AhmedML's sweep and the steps
Mixed cells plan: point gradients at bad corners
Mixed cells plan: phase 6 done
Mixed cells plan: phase 5 done
Mixed cells plan: phase 4 done
Mixed cells plan: phase 3 done
Mixed cells plan: phases 1 and 2 done
Mixed cells plan (#307)
Initial Home page