search

riscv-non-isa / rvv-intrinsic-doc

https://jira.riscv.org/browse/RVG-153
BSD 3-Clause "New" or "Revised" License
299 stars 89 forks source link

Passing vl as arguments of intrinsics #8

Closed Hsiangkai closed 4 years ago

Hsiangkai commented 4 years ago

We provide intrinsics with/without vl at the same time.

vop_vv_type(a, b) vop_vv_type_vl(a, b, gvl)

rdolbeau commented 4 years ago

Hello, It seems to me we are in somewhat of an agreement here - having both versions of the intrinsics, with the VL-explicit version (with an integer or integer-like parameter from #20) being essentially a shortcut for adding an explicit vsetvl() before the instruction (i.e., the intrinsics might compute on fewer elements if the requested VL is too big and/or somehow illegitimate, user beware). And as 0 is a legitimate value for VL, 0 is not a special case. So - maybe we can close this one as well to move things forward with the proposal? Cordially,

Hsiangkai commented 4 years ago

Agree. In current stage, we keep both versions of intrinsics as described in the RFC.