[PATCH] D80639: AMDGPU: Support non-entry block static sized allocas
Stanislav Mekhanoshin via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Wed May 27 10:50:26 PDT 2020
rampitec added a comment.
In D80639#2057740 <https://reviews.llvm.org/D80639#2057740>, @arsenm wrote:
> In D80639#2057705 <https://reviews.llvm.org/D80639#2057705>, @rampitec wrote:
>
> > Can you add a tests showing a kernel along with resulting ScratchSize please?
>
>
> I wasn't sure what to report for the size, so this just misses it entirely. The old code object has is_dynamic_callstack = 1, which I'm not sure actually did anything. I guess we could just pick a big number here like is already done for the external call case? I guess I could pick a smaller, large number?
Probably yes. We need to allocate it somehow. A large number does not seem unreasonable unless until we have something better.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D80639/new/
https://reviews.llvm.org/D80639
More information about the llvm-commits
mailing list