Section: New Results

Optimizing Remote Accesses for HLS

Participants : Christophe Alias, Alain Darte, Alexandru Plesco [Compsys/Zettice] .

Some data- and compute-intensive applications can be accelerated by offloading portions of codes to platforms such as GPGPUs or FPGAs. However, to get high performance for these kernels, it is mandatory to restructure the application, to generate adequate communication mechanisms for the transfer of remote data, and to make good usage of the memory bandwidth. In the context of the high-level synthesis (HLS), from a C program, of hardware accelerators on FPGA, we showed how to automatically generate optimized remote accesses for an accelerator communicating to an external DDR memory. Loop tiling is used to enable block communications, suitable for DDR memories. Pipelined communication processes are generated to overlap communications and computations, thereby hiding some latencies, in a way similar to double buffering. Finally, not only intra-tile but also inter-tile data reuse is exploited to avoid remote accesses when data are already available in the local memory.

We showed how to generate the sets of data to be read from (resp. written to) the external memory just before (resp. after) each tile so as to reduce communications and reuse data as much as possible in the accelerator. The main difficulty arises when some data may be (re)defined in the accelerator and should be kept locally. We proposed an automatic optimized code generation scheme, entirely at source-level, i.e., in C, that allows us to compile all the necessary glue (the communication processes) with the same HLS tool as for the computation kernel. Our method, implemented in the tool Chuba (see Section  5.7 ) uses advanced polyhedral techniques for program analysis and transformation. Experiments with Altera HLS tools demonstrate how to use our techniques to efficiently map C kernels to FPGA.

This work, astride two different fields (compilation for high-performance computing and high-level synthesis) turned out to be very difficult to publish. It was finally accepted at PPoPP'12 [6] , but only as a short paper (2 pages). We requested to retain the copyright of this work to be able to publish a longer version. It was accepted at the IMPACT'12 workshop [7] , which makes paper available on the web, but with no copyright. It was finally accepted as a full publication at the DATE'13 conference [8] .