I know the whole GPGPU thing has been around for a few years… but with the upcoming release of the new AMD chips later this year has brought it back to the fore… In a slightly different way… The HSA foundation has some interesting backers other than AMD (Samsung and some other ARM chip makers)…The other thing is it’s aim is to make integration painless and cost effective (i.e. simple and not take too much time)… Of course a big part of this is the memory space is truly shared (heterogeneous)… Anyway… I’m hoping we may see some benefits of this in the years to come… but sooner rather than later 
Unless they have done something to dramatically reduce latency, use of a GPU is going to be limited to “offline” processing. Which could still be interesting, but not much more so than simply having another processor handy.
The address space is shared now, so you are supposed to be able to pass pointers between CPU and GPU processes.
The hardware is only just now sampling, so I don’t think anyone outside of AMD and a few hardware OEM partners can actually test the first implementation to determine latency numbers on real hardware.
Currently only the proprietary drivers have OpenCL support, not the open source drivers, so not sure how interesting this will be in a real time environment. Does the proprietary Catalyst driver play nicely with a full -RT kernel, or even a carefully tuned -preempt kernel?
The kind of application stacks they are talking about are facial recognition algorithms and gesture analysis there appears to be big interest in the device space, my assumption then is latency is an issue they are trying to address by the shared space and being able to directly address the GPU. Here is something that may be of interest… from the HSA foundation website.
“Key members of the foundation are driving key technologies into open source comunity like IOMMU support in the linux kernel as well as bring memory management enhancement, driver stacks, exceptions handling, user mode interfaces, graphic driver interop and arbritration for the support for HSA enablement in operating system. HSA core compiler for generating HSAIL will be based on LLVM. They are also bring open source standards template library for Parallel Algorithms and Data Structures and other tools like simulators into the open source community. This is only the start of the activities.”