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
During the office hour we discussed a case where all control flow ends with discard, and whether we want to support it.
It conflicts with this rule in the behavioral spec:
If the function has a return type, the behavior of its body must be one of {Return} or {Return, Discard}.
I raised the concern that this restriction feels artificial and unhelpful, and that we should also allow {Discard} behavior. So here comes this issue.
To support this notion, here is a real-world use case which we'd have to reject if we keep the rule unchanged: gfx-rs/wgpu#4458
The text was updated successfully, but these errors were encountered:
AB: Would like to leave as is. We know Naga has seen case from code generator. Given problems in metal backend feel like changing code gen is easier to not do unconditional discard. Not big change and makes a bunch of stuff easier to explain.
KG: As long as has satisfactory workaround, fine for v1?
AB: Yea
KG: Happy with that. Will copy notes and mark closed.
Highly related to gfx-rs/naga#2378
During the office hour we discussed a case where all control flow ends with
discard
, and whether we want to support it.It conflicts with this rule in the behavioral spec:
I raised the concern that this restriction feels artificial and unhelpful, and that we should also allow
{Discard}
behavior. So here comes this issue.To support this notion, here is a real-world use case which we'd have to reject if we keep the rule unchanged: gfx-rs/wgpu#4458
The text was updated successfully, but these errors were encountered: