You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As KhronosGroup/SYCL-CTS#870 notes, the CTS seems to be testing that vec::load and vec::store can take an accessor as a parameter. See here.
This is not in the spec, but it seems generally useful. I wonder if it was just an oversight that we didn't add it to the spec. The fact that we included it in the CTS seems to indicate that it was our intent to allow vec::load / vec::store to an accessor.
If we decide to added these overloads, I think they would look like this:
I'm not sure if the CTS is testing with target::local or target::constant_buffer. If so, we could add deprecated overloads for these. (Those two targets are deprecated in SYCL 2020.)
Although the CTS does not test local_accessor, I presume we would want to allow that also. I think the only reason the CTS does not test local_accessor is because these tests were written in the SYCL 1.2.1 days, and we added local_accessor in SYCL 2020.
The text was updated successfully, but these errors were encountered:
Thinking some more about this, we probably want to allow vec::load and vec::store to be used from a host task. In that case, we also need to allow target::host_task. Therefore, it might be better to add this as a template parameter with another constraint:
As KhronosGroup/SYCL-CTS#870 notes, the CTS seems to be testing that
vec::load
andvec::store
can take anaccessor
as a parameter. See here.This is not in the spec, but it seems generally useful. I wonder if it was just an oversight that we didn't add it to the spec. The fact that we included it in the CTS seems to indicate that it was our intent to allow
vec::load
/vec::store
to anaccessor
.If we decide to added these overloads, I think they would look like this:
I'm not sure if the CTS is testing with
target::local
ortarget::constant_buffer
. If so, we could add deprecated overloads for these. (Those two targets are deprecated in SYCL 2020.)Although the CTS does not test
local_accessor
, I presume we would want to allow that also. I think the only reason the CTS does not testlocal_accessor
is because these tests were written in the SYCL 1.2.1 days, and we addedlocal_accessor
in SYCL 2020.The text was updated successfully, but these errors were encountered: