whether can I explicitly bind a certain thread block to a certain SM? block, sm, cuda, 295

[azaonline]

Is there any API support these functions

1. whether can I explicitly bind a certain thread block to a certain SM?

2)if 1 can, whether can I keep the result of block(kernel) on shared memory and let next block(kernel) access it?

my GPU is G295

Thanks,

[avidday]

1. whether can I explicitly bind a certain thread block to a certain SM?

No.

2)if 1 can, whether can I keep the result of block(kernel) on shared memory and let next block(kernel) access it?

Also no.

[tera]

BUT you can process several work packages in a loop within one block and reuse your shared memory setup between them. No need to invent fancy new constructs.

[azaonline]

Thanks