The 3rd section contains information about every subroutine. For each
subroutine, the programmer needs to specify on which memory this
routine depends, and in what way (read, write or both). This helps
HyCuda to determine which data has to be available on which device at
any point in the algorithm. The syntax of the routine-section looks
like this:
function-identifier: memory-identifier (permissions), ...
For example, a function fun1 that will read from
array1 and will both read and write to array2 will
be declared as
fun1: array1 (r), array2 (rw)
The memory-identifiers should match those declared in the
memory-section, and all functions that are part of the algorithm
should be declared here.
Joren Heit
2013-12-17