[cudadev] Use Cooperative Groups in the Fishbone kernel - #212

Draft

fwyzard wants to merge 8 commits into

cms-patatrack:masterfrom

fwyzard:fishbone_use_cooperative_groups

Commits on Aug 28, 2021

Improve variable names and code readability
fwyzard
committed
4aa01cc

Commits on Aug 29, 2021

Use a 1D grid for the fishbone kernel, with the inner loop stride fixed at compile time
fwyzard
committed
dfd5a10

Commits on Aug 30, 2021

Move fishbone buffers to shared memory
fwyzard
committed
1438eaa
Parallelise the first inner loop of the Fishbone algorithm
fwyzard
committed
1d3bdd6
Minimal, incomplete serial CPU implementation of Cooperative Groups
fwyzard
committed
416a7e2
Use Cooperative Groups for a more fine-grained synchronisation
fwyzard
committed
e607935
Ignore __launch_bounds__ in CPU code
fwyzard
committed
f7a1998
Optimise the fishbone kernel for a fixed block size
fwyzard
committed
c30ff98