gpu_group.c: can_tile: separate stride detection from bound computation
commit28d210f2c93077c99245ce2f868c97241fa45bcd
authorSven Verdoolaege <sven.verdoolaege@gmail.com>
Tue, 15 May 2018 08:27:54 +0000 (15 10:27 +0200)
committerSven Verdoolaege <sven.verdoolaege@gmail.com>
Tue, 29 May 2018 14:56:33 +0000 (29 16:56 +0200)
treefc8136935b1e2754e81a07ea1f7a212b28232b35
parent60433ace28be2ecda9c1d30f4de3f832b2af0efa
gpu_group.c: can_tile: separate stride detection from bound computation

In particular, instead of looking at each individual dimension
in the access and determining a stride and a bound for each
separately, first determine (and remove) all strides and
only then compute bounds.
This will make it easier to promote the entire bound computation
functionality to isl.

The code also gets simplified a bit and it could be further simplified
by changing the way the shifts and the strides are stored.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>
gpu_group.c