Commit 70d7c4d

committed

[NVPTX] add combiner rule for expanding LOAD, LoadV2, LoadParam, LoadParamV2

To reduce the number of unpacking movs when the element type is i64 but all uses are of unpacked f32s.

1 parent e3a4a39 commit 70d7c4dCopy full SHA for 70d7c4d

1 file changed

+296

-92

lines changed

+296

-92

lines changed

Comments

(0)