vof_block_packing Namespace Reference

Functions
	build (n, seeds, device, pool)
	seeds_64 (n)
	run (n, seeds, device, pool, steps, warm)
	main ()

Detailed Description

WO-V9 item 4 (second half) — the block gather/scatter packing, device vs host, on a QUIET GPU.

WO-W12 measured 2.731 ms (device) vs 4.626 ms (host) = 1.69x on a comparatively quiet GPU and
79.4 vs 63.4 ms = 0.80x with five other jobs on it, and asked for the number to be re-taken on an
idle machine.  This is that measurement: the 64-marker swarm of the W12 `swarm` gate, kinematic
block advection (`advect_vof_blocks`), device staging on and off, alternated so a drift in machine
state shows up as scatter rather than as a result.

    PYTHONPATH=$PWD/build_cuda python tests/study/vof_block_packing.py

Function Documentation

build()

vof_block_packing.build	(		n,
			seeds,
			device,
			pool
	)

Definition at line 20 of file vof_block_packing.py.

seeds_64()

vof_block_packing.seeds_64

(

n

)

Definition at line 35 of file vof_block_packing.py.

Referenced by main().

run()

vof_block_packing.run	(		n,
			seeds,
			device,
			pool,
			steps,
			warm
	)

Definition at line 46 of file vof_block_packing.py.

Referenced by main().

main()

vof_block_packing.main

(

)

Definition at line 60 of file vof_block_packing.py.

References main(), run(), and seeds_64().

Referenced by main().