I actually just started collaborating with these guys http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6546013 who recently moved from MIT to Berkeley. They're using Chisel, Bluespec, and a custom compiler that takes a graph representation of an algorithm and determines a hardware layout and schedule for deployment on an FPGA or ASIC. I'm trying to convince them that Julia would be the best choice as a high-level algorithm description language to generate inputs to their tools, but haven't won them over yet. It's still early going in terms of making it usable with a high level BLAS-like interface.
On Sunday, June 1, 2014 9:02:18 AM UTC-7, David Ainish wrote: > > @Jameson > Apparently IBM's Liquid Metal project (LIME programming language) is > already taking this approach. > > It seems LIME can compile into heterogeneous hardware with both CPUs, GPUs > and FPGAs and customize the hardware according to the application that is > running, executing tasks in the most efficient hardware for each kind of > task. Things would be transparent for the programmer. > > From their website: "Our long-term goal is to 'JIT the hardware' to > dynamically select methods for compilation and synthesis to hardware, > potentially taking advantage of dynamic information in the same way that > multi-level JIT compilers do today for software." > > IBM's Liquid Metal project. > <http://domino.research.ibm.com/library/cyberdig.nsf/1e4115aea78b6e7c85256b360066f0d4/43e23ec91521c1e3852577bc0048ea37!OpenDocument> > > > For those not familiar with FPGAs/ASICs here is a very good article that > discusses the benefits and related technologies, like High-Level synthesis, > which turns high-level language code into integrated circuits. It discusses > Liquid Metal and shares thoughts about the future of programming languages > and the hardware architecture they'll run on: > http://queue.acm.org/detail.cfm?id=2443836 > > And an open source project for hardware construction from code (Chisel): > https://chisel.eecs.berkeley.edu/ > > > Julia seems perfect for this level of dynamism. I'd rather see Julia's > simplicity, clean and flexible style taking advantage of pioneering in this > field to help it become a standard rather than a not so optimal adaptation > of any legacy language just for the sake of compatibility. > > > On Friday, May 30, 2014 8:58:29 PM UTC-3, Jameson wrote: > > >> but is anyone doing stuff like that already? >> >
