WEBVTT

NOTE
Draft cues converted from captions.srt. Final wording and timing are recorded later.

1
00:00:00.300 --> 00:00:23.850 line:95% align:center size:80%
I build this tree with the seams-core pixi environment. The two lines on screen are the install: meson setup build, then meson compile -C build. That produces the seams command and libseams, which is the library the C call sits on. I keep that build and use it for every count in this recording.

2
00:00:24.250 --> 00:00:39.470 line:95% align:center size:80%
The input is input/traj/ice1h.lammpstrj. seams read prints nop 128 frame 1. Those are the 128 oxygens in one frame. This is the cell docs_engine_openmm scores, and it is the cell I hand to C.

3
00:00:40.020 --> 00:00:57.320 line:95% align:center size:80%
Neighbours are a cutoff of 3.5 angstroms on type 1. seams cn prints nI 128, so every oxygen is in the list. Ice Ih keeps a four-neighbour shell inside that cutoff. The C neighbour call walks this same list.

4
00:00:57.670 --> 00:01:12.890 line:95% align:center size:80%
Primitive rings use that list. The histogram is 6 256, and the summary is rings 256. Two six-rings for each of the 128 waters, and no other size. docs_engine_openmm names that count six-rings 256.

5
00:01:13.390 --> 00:01:29.020 line:95% align:center size:80%
Local q6 on the same cutoff prints q6 0.625855. Published CHILL+ then prints hexagonal 128. Every water is hexagonal. I want the C result to show the same two figures before I trust the wrapper.

6
00:01:29.620 --> 00:01:49.420 line:95% align:center size:80%
The C program is docs/videos/c/ice.c. It calls the neighbour builder, the q6 routine, chill plus with the published rule, and the ring finder. On this cell it prints nop 128 nI 128 q6 0.625855 hexagonal 128 rings 256. That is the line I read back.
