Archives Discussions

[ryta1203]

Can we please get local arrays supported in Brook+?

I have seen too many embarassingly parallel apps not being portable (in a practical sense) to Brook+ simply because of the lack of local arrays.

Is this feature going to be added in the new release? If not, what new features can we look forward to? When can we expect the new release? Also, is AMD still committed to GPU computing in a serious way?

Thanks.

[ryta1203]

If I could make another request, it would be to either:

A. allow multiple scatter streams in a kernel AND allow the out scatter streams to be of different variable names and types

B. allow a struct scatter stream, so long as the actually assignment involved 128 bits (ie. can't assign whole structures, etc). This would allow us to get around the multiple scatter streams in a kernel problem simply by having a structure of different variables we wanted to output and controlling the domain size either by execDomain or within the kernel itself (ie. if statements).

Either of these would drastically increase the usability of Brook+, IMO along with the aforementioned request also.

Are any of these things going to be added to v1.2 BETA???