CU2CL: A CUDA-to-OpenCL Translator for Multi- and Many-Core Architectures

Abstract

The use of graphics processing units (GPUs) in high-performance parallel computing continues to steadily become more prevalent, often as part of a heterogeneous system. For years, CUDA has been the de facto programming environment for nearly all general-purpose GPU (GPGPU) applications. In spite of this, the framework is available only on NVIDIA GPUs, traditionally requiring reimplementation in other frameworks in order to utilize additional multi-or many-core devices. On the other hand, OpenCL provides an open and vendorneutral programming environment and run-time system. With implementations available for CPUs, GPUs, and other types of accelerators, OpenCL therefore holds the promise of a "write once, run anywhere" ecosystem for heterogeneous computing.Given the many similarities between CUDA and OpenCL, manually porting a CUDA application to OpenCL is almost straightforward, albeit tedious and error-prone. In response to this issue, we created CU2CL, an automated CUDA-to-OpenCL source-to-source translator that possesses a novel design and clever reuse of the Clang compiler framework. Currently, the CU2CL translator covers the primary constructs found in the CUDA Runtime API, and we have successfully translated several applications from the CUDA SDK and Rodinia benchmark suite. CU2CL's translation times are reasonable, allowing for many applications to be translated at once. The number of manual changes required after executing our translator on CUDA source is minimal, with some compiling and working with no changes at all. The performance of our automatically translated applications via CU2CL is on par with their manually ported counterparts.

MetaMorph: A Library Framework for Interoperable Kernels on Multi- and Many-Core Clusters

“…It provides a compilation infrastructure and helper APIs that handle the boilerplate initialization and compilation and simplify data exchange between the host and accelerators, such that MetaMorph developers can focus on developing the new kernels. In addition, we used our source-to-source translator [7], [8] to largely automate the generation of MetaMorph kernels, and in the future, we aim to release a family of such source-to-source translators and leverage the LLVM Just-InTime compiler to simplify the expansion of the MetaMorph library with new design patterns.…”

Lost in Translation: Challenges in Automating CUDA-to-OpenCL Translation

“…In addition, its source code has not been released, and as such, cannot be compared directly to our translator, CU2CL. We note that both CUDAtoOpenCL and CU2CL have recursion-based expression rewriting, propagation of the cl mem type, and support for the most commonly-used environment setup, memory management, and kernel launch functions [8] [9].…”

“…While a full discussion of CU2CL's implementation is outside the scope of this paper, we provide a brief introduction to its high-level construction as context. A more detailed exposition is available in [9].…”

Molecular Dynamics Simulation of Iron — A Review

“…[196][197][198] Though such a translation is trivial in view of the similar architecture of both CUDA and OpenCL, challenges exist for a robust porting between them. 199 For example, separate compilation of CUDA source¯les is possible, yet it is quite di±cult to link the code translated in OpenCL format due to the required reorganization of the initialization code throughout all the source¯les.…”