Co-Array Fortran

Co-Array Fortran [36], formerly called F$--$, is a simple and small set of extensions to Fortran 95 for Single Program Multiple Data, SPMD, parallel processing. It feels and look similar to Fortran and has a few new extensions associated with work distribution and data distribution. We review the work distribution first. A single program is replicated a fixed number of times. Each replication, called an image, has its own data objects. Each image is asynchronously executed with the normal rules of Fortran, so the execution path may differ from image to image. The programmer decides the actual path for the image through a unique image index, by using normal Fortran control constructs and by explicit synchronizations. For data distribution, Co-Array Fortran lets the programmer choose data distribution in a syntax very much like normal Fortran array syntax. Array indices in parentheses follow the normal Fortran rules within one memory image. Array indices in square brackets provide convenient notation for accessing objects across images and follow similar rules. The programmer uses co-array syntax only where it is needed. A reference to a co-array with no square brackets attached to it is a reference to the object in the local memory of the executing image. Since most references to data objects in a parallel code should be to the local part, co-array syntax should appear only in isolated parts of the code. If not, the syntax acts as a visual flag to the programmer that too much communication among images may be taking place. It also acts as a flag to the compiler to generate code that avoids latency whenever possible. Co-arrays provides a way of expressing remote memory operations. For example:

$$\begin{minipage}[t]{\linewidth}\small\begin{verbatim} X = Y[PE] !... ... ! S scalar, promoted to array of shape (1:M,1:N)\end{verbatim}\end{minipage}$$

The approach is different to HPJava. In Co-Array Fortran, array subscripting is local by default, or involves a combination of local subscripts and explicit process ids. There is no analogue of global subscripts, or HPF-like distribution formats. In Co-Array Fortran the logical model of communication is built into the language--remote memory access with intrinsics for synchronization. In HPJava, there are no communication primitives in the language itself. We follow the MPI philosophy of providing communication through separate libraries. While Co-Array Fortran and HPJava share an underlying programming model, we claim that our framework may offer greater opportunities for exploiting established software technologies.